diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,173240 @@
+{
+  "best_metric": 0.5751461982727051,
+  "best_model_checkpoint": "clean_output/checkpoint-28860",
+  "epoch": 0.9999913381666363,
+  "eval_steps": 5772,
+  "global_step": 28862,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.1039,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.059,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-06,
+      "loss": 1.0878,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.0608,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 1.0106,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-05,
+      "loss": 0.9002,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4e-05,
+      "loss": 0.8761,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.8031,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-05,
+      "loss": 0.7787,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 0.7707,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999999940718676e-05,
+      "loss": 0.7745,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999999762874704e-05,
+      "loss": 0.7176,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999946646808e-05,
+      "loss": 0.7355,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999999051498816e-05,
+      "loss": 0.7355,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999851796691e-05,
+      "loss": 0.7445,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999997865872376e-05,
+      "loss": 0.7311,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999709521521e-05,
+      "loss": 0.7301,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999996205995435e-05,
+      "loss": 0.7276,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999995198213053e-05,
+      "loss": 0.7548,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999994071868078e-05,
+      "loss": 0.695,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999992826960523e-05,
+      "loss": 0.6671,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999991463490405e-05,
+      "loss": 0.7124,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999989981457735e-05,
+      "loss": 0.7277,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999988380862533e-05,
+      "loss": 0.7145,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999986661704822e-05,
+      "loss": 0.6726,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999984823984616e-05,
+      "loss": 0.6888,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999982867701943e-05,
+      "loss": 0.7278,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999998079285682e-05,
+      "loss": 0.6938,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999978599449277e-05,
+      "loss": 0.6874,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999976287479336e-05,
+      "loss": 0.7285,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999973856947025e-05,
+      "loss": 0.6764,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999997130785238e-05,
+      "loss": 0.7189,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999968640195417e-05,
+      "loss": 0.6773,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999996585397618e-05,
+      "loss": 0.7218,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999962949194697e-05,
+      "loss": 0.6938,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999959925851006e-05,
+      "loss": 0.656,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999995678394514e-05,
+      "loss": 0.7196,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999953523477132e-05,
+      "loss": 0.7177,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999950144447027e-05,
+      "loss": 0.7175,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999994664685486e-05,
+      "loss": 0.6418,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999943030700678e-05,
+      "loss": 0.6706,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999939295984525e-05,
+      "loss": 0.7355,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999935442706436e-05,
+      "loss": 0.7043,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999993147086647e-05,
+      "loss": 0.7332,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999927380464657e-05,
+      "loss": 0.6892,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999923171501062e-05,
+      "loss": 0.7006,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999991884397572e-05,
+      "loss": 0.7,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99999143978887e-05,
+      "loss": 0.7009,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999909833240038e-05,
+      "loss": 0.7425,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999905150029798e-05,
+      "loss": 0.6946,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999990034825803e-05,
+      "loss": 0.7225,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999895427924797e-05,
+      "loss": 0.72,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999989038903015e-05,
+      "loss": 0.7409,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999885231574153e-05,
+      "loss": 0.6759,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999987995555687e-05,
+      "loss": 0.6848,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999874560978358e-05,
+      "loss": 0.6368,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999869047838683e-05,
+      "loss": 0.6673,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999986341613791e-05,
+      "loss": 0.6344,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999985766587611e-05,
+      "loss": 0.7001,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999851797053345e-05,
+      "loss": 0.6998,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999845809669685e-05,
+      "loss": 0.67,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999839703725206e-05,
+      "loss": 0.6732,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999833479219974e-05,
+      "loss": 0.6711,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999982713615407e-05,
+      "loss": 0.6905,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999820674527565e-05,
+      "loss": 0.7061,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999814094340534e-05,
+      "loss": 0.6912,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999807395593057e-05,
+      "loss": 0.7137,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999800578285217e-05,
+      "loss": 0.741,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999793642417087e-05,
+      "loss": 0.6492,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999786587988757e-05,
+      "loss": 0.6843,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999779415000303e-05,
+      "loss": 0.6669,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999772123451818e-05,
+      "loss": 0.6905,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999764713343382e-05,
+      "loss": 0.725,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999757184675087e-05,
+      "loss": 0.6838,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999974953744702e-05,
+      "loss": 0.6803,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999741771659276e-05,
+      "loss": 0.6766,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999973388731194e-05,
+      "loss": 0.6656,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999972588440511e-05,
+      "loss": 0.6682,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999971776293888e-05,
+      "loss": 0.6963,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999709522913346e-05,
+      "loss": 0.6526,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999701164328607e-05,
+      "loss": 0.6638,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999969268718476e-05,
+      "loss": 0.6587,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999684091481907e-05,
+      "loss": 0.6862,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999967537722015e-05,
+      "loss": 0.6593,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999966654439959e-05,
+      "loss": 0.6967,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999657593020333e-05,
+      "loss": 0.7101,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999648523082488e-05,
+      "loss": 0.6885,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999963933458616e-05,
+      "loss": 0.6838,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999630027531457e-05,
+      "loss": 0.7213,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999962060191849e-05,
+      "loss": 0.695,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999961105774737e-05,
+      "loss": 0.6578,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999601395018216e-05,
+      "loss": 0.6334,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999959161373113e-05,
+      "loss": 0.7443,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999958171388624e-05,
+      "loss": 0.6855,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999571695483664e-05,
+      "loss": 0.6794,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999561558523514e-05,
+      "loss": 0.6686,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999551303005908e-05,
+      "loss": 0.6625,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999540928930974e-05,
+      "loss": 0.6574,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999530436298834e-05,
+      "loss": 0.5893,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999951982510961e-05,
+      "loss": 0.6895,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999950909536343e-05,
+      "loss": 0.6567,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999498247060417e-05,
+      "loss": 0.659,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999487280200703e-05,
+      "loss": 0.6648,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999476194784423e-05,
+      "loss": 0.7062,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99994649908117e-05,
+      "loss": 0.6743,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999453668282668e-05,
+      "loss": 0.657,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999944222719747e-05,
+      "loss": 0.7071,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999943066755623e-05,
+      "loss": 0.6972,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999418989359093e-05,
+      "loss": 0.6402,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999940719260619e-05,
+      "loss": 0.6965,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999939527729767e-05,
+      "loss": 0.6946,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999383243433665e-05,
+      "loss": 0.6989,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999371091014325e-05,
+      "loss": 0.7167,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999935882003979e-05,
+      "loss": 0.715,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999934643051021e-05,
+      "loss": 0.6478,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999333922425727e-05,
+      "loss": 0.6596,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999932129578649e-05,
+      "loss": 0.7149,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999308550592655e-05,
+      "loss": 0.6807,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999295686844363e-05,
+      "loss": 0.6778,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999282704541772e-05,
+      "loss": 0.6956,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999269603685035e-05,
+      "loss": 0.6992,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999925638427431e-05,
+      "loss": 0.6458,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999924304630975e-05,
+      "loss": 0.6654,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999229589791517e-05,
+      "loss": 0.6751,
+      "step": 124
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999216014719766e-05,
+      "loss": 0.6912,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999202321094658e-05,
+      "loss": 0.6785,
+      "step": 126
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999918850891636e-05,
+      "loss": 0.6706,
+      "step": 127
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999917457818503e-05,
+      "loss": 0.6205,
+      "step": 128
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999916052890084e-05,
+      "loss": 0.7155,
+      "step": 129
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999146361063954e-05,
+      "loss": 0.6338,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999132074674534e-05,
+      "loss": 0.7315,
+      "step": 131
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999117669732756e-05,
+      "loss": 0.7305,
+      "step": 132
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999103146238793e-05,
+      "loss": 0.6543,
+      "step": 133
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999088504192808e-05,
+      "loss": 0.6456,
+      "step": 134
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999073743594984e-05,
+      "loss": 0.6311,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999905886444549e-05,
+      "loss": 0.6616,
+      "step": 136
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999043866744502e-05,
+      "loss": 0.684,
+      "step": 137
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999028750492204e-05,
+      "loss": 0.6658,
+      "step": 138
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999901351568877e-05,
+      "loss": 0.7399,
+      "step": 139
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9998998162334383e-05,
+      "loss": 0.6358,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9998982690429224e-05,
+      "loss": 0.7194,
+      "step": 141
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9998967099973475e-05,
+      "loss": 0.6876,
+      "step": 142
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999895139096732e-05,
+      "loss": 0.7008,
+      "step": 143
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999893556341095e-05,
+      "loss": 0.6821,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998919617304552e-05,
+      "loss": 0.6604,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999890355264831e-05,
+      "loss": 0.669,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999888736944242e-05,
+      "loss": 0.6264,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999887106768707e-05,
+      "loss": 0.6856,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998854647382454e-05,
+      "loss": 0.6742,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999883810852877e-05,
+      "loss": 0.7097,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999882145112621e-05,
+      "loss": 0.671,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998804675174973e-05,
+      "loss": 0.7166,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998787780675258e-05,
+      "loss": 0.6844,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998770767627268e-05,
+      "loss": 0.6494,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99987536360312e-05,
+      "loss": 0.6819,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999873638588726e-05,
+      "loss": 0.6262,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999871901719565e-05,
+      "loss": 0.6497,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999870152995658e-05,
+      "loss": 0.7246,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998683924170255e-05,
+      "loss": 0.6237,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999866619983688e-05,
+      "loss": 0.6849,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998648356956672e-05,
+      "loss": 0.669,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998630395529838e-05,
+      "loss": 0.6396,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998612315556593e-05,
+      "loss": 0.7026,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999859411703715e-05,
+      "loss": 0.649,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998575799971726e-05,
+      "loss": 0.6342,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998557364360537e-05,
+      "loss": 0.6696,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99985388102038e-05,
+      "loss": 0.758,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999852013750174e-05,
+      "loss": 0.6685,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999850134625458e-05,
+      "loss": 0.6515,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999848243646253e-05,
+      "loss": 0.6701,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999846340812583e-05,
+      "loss": 0.7115,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998444261244695e-05,
+      "loss": 0.6693,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998424995819354e-05,
+      "loss": 0.6831,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998405611850036e-05,
+      "loss": 0.6355,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998386109336974e-05,
+      "loss": 0.6397,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998366488280398e-05,
+      "loss": 0.666,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998346748680534e-05,
+      "loss": 0.7048,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998326890537622e-05,
+      "loss": 0.692,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99983069138519e-05,
+      "loss": 0.5899,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99982868186236e-05,
+      "loss": 0.6942,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998266604852965e-05,
+      "loss": 0.6529,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999824627254023e-05,
+      "loss": 0.6145,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999822582168563e-05,
+      "loss": 0.6123,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998205252289425e-05,
+      "loss": 0.6338,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998184564351846e-05,
+      "loss": 0.5915,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998163757873138e-05,
+      "loss": 0.6495,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998142832853554e-05,
+      "loss": 0.6947,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998121789293342e-05,
+      "loss": 0.6541,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998100627192743e-05,
+      "loss": 0.6577,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998079346552014e-05,
+      "loss": 0.6453,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999805794737141e-05,
+      "loss": 0.642,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999803642965118e-05,
+      "loss": 0.6259,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999801479339158e-05,
+      "loss": 0.6648,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997993038592866e-05,
+      "loss": 0.6516,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99979711652553e-05,
+      "loss": 0.6825,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997949173379138e-05,
+      "loss": 0.648,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999792706296464e-05,
+      "loss": 0.7087,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997904834012067e-05,
+      "loss": 0.6998,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997882486521686e-05,
+      "loss": 0.7002,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997860020493763e-05,
+      "loss": 0.6872,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999783743592856e-05,
+      "loss": 0.6346,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997814732826345e-05,
+      "loss": 0.6779,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997791911187394e-05,
+      "loss": 0.6303,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999776897101197e-05,
+      "loss": 0.7138,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999774591230035e-05,
+      "loss": 0.6549,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99977227350528e-05,
+      "loss": 0.6514,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997699439269603e-05,
+      "loss": 0.7101,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999767602495103e-05,
+      "loss": 0.7045,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997652492097362e-05,
+      "loss": 0.6947,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997628840708877e-05,
+      "loss": 0.6573,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997605070785856e-05,
+      "loss": 0.6927,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999758118232858e-05,
+      "loss": 0.7084,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997557175337327e-05,
+      "loss": 0.637,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999753304981239e-05,
+      "loss": 0.6459,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997508805754055e-05,
+      "loss": 0.6703,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99974844431626e-05,
+      "loss": 0.6845,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997459962038325e-05,
+      "loss": 0.7065,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997435362381513e-05,
+      "loss": 0.6916,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999741064419246e-05,
+      "loss": 0.6717,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997385807471455e-05,
+      "loss": 0.6564,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997360852218795e-05,
+      "loss": 0.6881,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997335778434776e-05,
+      "loss": 0.6871,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997310586119694e-05,
+      "loss": 0.7128,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997285275273847e-05,
+      "loss": 0.6689,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997259845897537e-05,
+      "loss": 0.6682,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997234297991065e-05,
+      "loss": 0.6399,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997208631554736e-05,
+      "loss": 0.6845,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999718284658885e-05,
+      "loss": 0.6598,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997156943093716e-05,
+      "loss": 0.6531,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997130921069638e-05,
+      "loss": 0.6963,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999710478051693e-05,
+      "loss": 0.6536,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997078521435897e-05,
+      "loss": 0.6364,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997052143826852e-05,
+      "loss": 0.6612,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9997025647690106e-05,
+      "loss": 0.6359,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996999033025973e-05,
+      "loss": 0.6627,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996972299834774e-05,
+      "loss": 0.6965,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999694544811682e-05,
+      "loss": 0.6281,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996918477872432e-05,
+      "loss": 0.6628,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999689138910193e-05,
+      "loss": 0.6152,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996864181805634e-05,
+      "loss": 0.7181,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996836855983866e-05,
+      "loss": 0.6723,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999680941163695e-05,
+      "loss": 0.6562,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996781848765217e-05,
+      "loss": 0.676,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999675416736899e-05,
+      "loss": 0.6982,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999672636744859e-05,
+      "loss": 0.6426,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996698449004355e-05,
+      "loss": 0.6921,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996670412036615e-05,
+      "loss": 0.6698,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996642256545703e-05,
+      "loss": 0.7375,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996613982531947e-05,
+      "loss": 0.7026,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996585589995693e-05,
+      "loss": 0.6962,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996557078937264e-05,
+      "loss": 0.6152,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999652844935701e-05,
+      "loss": 0.6328,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999649970125526e-05,
+      "loss": 0.6748,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996470834632367e-05,
+      "loss": 0.6736,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996441849488664e-05,
+      "loss": 0.6989,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996412745824497e-05,
+      "loss": 0.7381,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996383523640214e-05,
+      "loss": 0.6741,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996354182936157e-05,
+      "loss": 0.7372,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999632472371267e-05,
+      "loss": 0.6564,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996295145970116e-05,
+      "loss": 0.6507,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996265449708834e-05,
+      "loss": 0.6899,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999623563492918e-05,
+      "loss": 0.654,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999620570163151e-05,
+      "loss": 0.6122,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999617564981617e-05,
+      "loss": 0.68,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996145479483528e-05,
+      "loss": 0.6505,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996115190633937e-05,
+      "loss": 0.629,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999608478326775e-05,
+      "loss": 0.6506,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996054257385334e-05,
+      "loss": 0.6576,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9996023612987054e-05,
+      "loss": 0.6864,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995992850073265e-05,
+      "loss": 0.6596,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995961968644337e-05,
+      "loss": 0.7001,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995930968700633e-05,
+      "loss": 0.6973,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995899850242525e-05,
+      "loss": 0.7016,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999586861327038e-05,
+      "loss": 0.6668,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995837257784564e-05,
+      "loss": 0.6838,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995805783785457e-05,
+      "loss": 0.6747,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995774191273427e-05,
+      "loss": 0.6399,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995742480248847e-05,
+      "loss": 0.7069,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995710650712096e-05,
+      "loss": 0.6573,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999567870266355e-05,
+      "loss": 0.6539,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999564663610359e-05,
+      "loss": 0.6717,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995614451032593e-05,
+      "loss": 0.6501,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995582147450943e-05,
+      "loss": 0.681,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995549725359022e-05,
+      "loss": 0.6866,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995517184757217e-05,
+      "loss": 0.6517,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995484525645907e-05,
+      "loss": 0.669,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995451748025486e-05,
+      "loss": 0.6985,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999541885189634e-05,
+      "loss": 0.6871,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999538583725886e-05,
+      "loss": 0.6825,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995352704113434e-05,
+      "loss": 0.6675,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995319452460457e-05,
+      "loss": 0.6176,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999528608230033e-05,
+      "loss": 0.6468,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995252593633436e-05,
+      "loss": 0.6797,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999521898646018e-05,
+      "loss": 0.6408,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999518526078096e-05,
+      "loss": 0.7437,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995151416596174e-05,
+      "loss": 0.6256,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995117453906224e-05,
+      "loss": 0.6528,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995083372711513e-05,
+      "loss": 0.6495,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995049173012448e-05,
+      "loss": 0.6608,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9995014854809427e-05,
+      "loss": 0.6979,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994980418102862e-05,
+      "loss": 0.669,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994945862893162e-05,
+      "loss": 0.6566,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994911189180734e-05,
+      "loss": 0.6375,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994876396965987e-05,
+      "loss": 0.6745,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994841486249342e-05,
+      "loss": 0.6077,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994806457031206e-05,
+      "loss": 0.6828,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994771309311995e-05,
+      "loss": 0.6229,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994736043092126e-05,
+      "loss": 0.692,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999470065837202e-05,
+      "loss": 0.725,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994665155152093e-05,
+      "loss": 0.67,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994629533432767e-05,
+      "loss": 0.6253,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994593793214467e-05,
+      "loss": 0.5849,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999455793449761e-05,
+      "loss": 0.6863,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994521957282627e-05,
+      "loss": 0.6389,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994485861569946e-05,
+      "loss": 0.6762,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994449647359987e-05,
+      "loss": 0.6483,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999441331465319e-05,
+      "loss": 0.6637,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994376863449978e-05,
+      "loss": 0.6964,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994340293750784e-05,
+      "loss": 0.6615,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994303605556045e-05,
+      "loss": 0.6757,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994266798866195e-05,
+      "loss": 0.7322,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994229873681663e-05,
+      "loss": 0.6741,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994192830002896e-05,
+      "loss": 0.6331,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994155667830333e-05,
+      "loss": 0.6537,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9994118387164413e-05,
+      "loss": 0.6324,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999408098800557e-05,
+      "loss": 0.7125,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999404347035426e-05,
+      "loss": 0.6934,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999400583421092e-05,
+      "loss": 0.6976,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993968079576002e-05,
+      "loss": 0.7064,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993930206449943e-05,
+      "loss": 0.6366,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993892214833203e-05,
+      "loss": 0.6612,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993854104726228e-05,
+      "loss": 0.6447,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993815876129474e-05,
+      "loss": 0.6572,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993777529043384e-05,
+      "loss": 0.6659,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993739063468424e-05,
+      "loss": 0.6001,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993700479405044e-05,
+      "loss": 0.6715,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993661776853704e-05,
+      "loss": 0.6839,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999362295581486e-05,
+      "loss": 0.6628,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993584016288973e-05,
+      "loss": 0.6574,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993544958276506e-05,
+      "loss": 0.6423,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993505781777923e-05,
+      "loss": 0.6907,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993466486793685e-05,
+      "loss": 0.6485,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993427073324262e-05,
+      "loss": 0.6791,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993387541370116e-05,
+      "loss": 0.7051,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993347890931723e-05,
+      "loss": 0.6364,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999330812200955e-05,
+      "loss": 0.6741,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993268234604063e-05,
+      "loss": 0.6663,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999322822871574e-05,
+      "loss": 0.6749,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993188104345054e-05,
+      "loss": 0.6691,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993147861492486e-05,
+      "loss": 0.6295,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993107500158508e-05,
+      "loss": 0.6685,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993067020343597e-05,
+      "loss": 0.6714,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9993026422048238e-05,
+      "loss": 0.6694,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992985705272906e-05,
+      "loss": 0.6869,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999294487001809e-05,
+      "loss": 0.6464,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999290391628427e-05,
+      "loss": 0.6531,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992862844071933e-05,
+      "loss": 0.6748,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992821653381568e-05,
+      "loss": 0.678,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999278034421366e-05,
+      "loss": 0.662,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99927389165687e-05,
+      "loss": 0.6468,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992697370447177e-05,
+      "loss": 0.6266,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999265570584959e-05,
+      "loss": 0.6633,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992613922776424e-05,
+      "loss": 0.6867,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992572021228183e-05,
+      "loss": 0.6307,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992530001205356e-05,
+      "loss": 0.6664,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999248786270845e-05,
+      "loss": 0.6466,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999244560573796e-05,
+      "loss": 0.6394,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992403230294384e-05,
+      "loss": 0.6983,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992360736378226e-05,
+      "loss": 0.6417,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992318123989995e-05,
+      "loss": 0.6915,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999227539313019e-05,
+      "loss": 0.6626,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992232543799322e-05,
+      "loss": 0.6646,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992189575997893e-05,
+      "loss": 0.6502,
+      "step": 373
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992146489726417e-05,
+      "loss": 0.6562,
+      "step": 374
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992103284985407e-05,
+      "loss": 0.6618,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992059961775373e-05,
+      "loss": 0.6231,
+      "step": 376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9992016520096825e-05,
+      "loss": 0.7064,
+      "step": 377
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991972959950284e-05,
+      "loss": 0.6356,
+      "step": 378
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999192928133626e-05,
+      "loss": 0.6997,
+      "step": 379
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999188548425528e-05,
+      "loss": 0.715,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991841568707854e-05,
+      "loss": 0.6737,
+      "step": 381
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991797534694507e-05,
+      "loss": 0.6466,
+      "step": 382
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999175338221576e-05,
+      "loss": 0.6137,
+      "step": 383
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991709111272138e-05,
+      "loss": 0.7074,
+      "step": 384
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991664721864167e-05,
+      "loss": 0.6301,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991620213992366e-05,
+      "loss": 0.6615,
+      "step": 386
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999157558765727e-05,
+      "loss": 0.6592,
+      "step": 387
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991530842859408e-05,
+      "loss": 0.6536,
+      "step": 388
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991485979599305e-05,
+      "loss": 0.6525,
+      "step": 389
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99914409978775e-05,
+      "loss": 0.6904,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991395897694523e-05,
+      "loss": 0.6809,
+      "step": 391
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991350679050904e-05,
+      "loss": 0.7001,
+      "step": 392
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999130534194719e-05,
+      "loss": 0.6559,
+      "step": 393
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991259886383907e-05,
+      "loss": 0.7366,
+      "step": 394
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99912143123616e-05,
+      "loss": 0.6639,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999116861988081e-05,
+      "loss": 0.6234,
+      "step": 396
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9991122808942075e-05,
+      "loss": 0.6446,
+      "step": 397
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999107687954594e-05,
+      "loss": 0.6334,
+      "step": 398
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999103083169295e-05,
+      "loss": 0.6646,
+      "step": 399
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990984665383654e-05,
+      "loss": 0.6418,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990938380618594e-05,
+      "loss": 0.6635,
+      "step": 401
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999089197739832e-05,
+      "loss": 0.6835,
+      "step": 402
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990845455723385e-05,
+      "loss": 0.6757,
+      "step": 403
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990798815594336e-05,
+      "loss": 0.6286,
+      "step": 404
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999075205701173e-05,
+      "loss": 0.6338,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990705179976123e-05,
+      "loss": 0.7339,
+      "step": 406
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990658184488062e-05,
+      "loss": 0.6696,
+      "step": 407
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990611070548115e-05,
+      "loss": 0.6464,
+      "step": 408
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990563838156835e-05,
+      "loss": 0.6404,
+      "step": 409
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990516487314777e-05,
+      "loss": 0.6409,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990469018022513e-05,
+      "loss": 0.6854,
+      "step": 411
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99904214302806e-05,
+      "loss": 0.757,
+      "step": 412
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990373724089602e-05,
+      "loss": 0.6987,
+      "step": 413
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990325899450086e-05,
+      "loss": 0.6783,
+      "step": 414
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999027795636262e-05,
+      "loss": 0.6768,
+      "step": 415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999022989482777e-05,
+      "loss": 0.6386,
+      "step": 416
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999018171484611e-05,
+      "loss": 0.6743,
+      "step": 417
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990133416418206e-05,
+      "loss": 0.705,
+      "step": 418
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9990084999544628e-05,
+      "loss": 0.6619,
+      "step": 419
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999003646422596e-05,
+      "loss": 0.6282,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998998781046277e-05,
+      "loss": 0.6502,
+      "step": 421
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998993903825564e-05,
+      "loss": 0.6198,
+      "step": 422
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9989890147605143e-05,
+      "loss": 0.6751,
+      "step": 423
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998984113851186e-05,
+      "loss": 0.6718,
+      "step": 424
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9989792010976376e-05,
+      "loss": 0.677,
+      "step": 425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9989742764999268e-05,
+      "loss": 0.6689,
+      "step": 426
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9989693400581122e-05,
+      "loss": 0.6629,
+      "step": 427
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998964391772253e-05,
+      "loss": 0.648,
+      "step": 428
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9989594316424065e-05,
+      "loss": 0.652,
+      "step": 429
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9989544596686325e-05,
+      "loss": 0.6425,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99894947585099e-05,
+      "loss": 0.6606,
+      "step": 431
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9989444801895374e-05,
+      "loss": 0.6397,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9989394726843342e-05,
+      "loss": 0.6531,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9989344533354403e-05,
+      "loss": 0.6618,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9989294221429146e-05,
+      "loss": 0.6447,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998924379106817e-05,
+      "loss": 0.646,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998919324227207e-05,
+      "loss": 0.6244,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998914257504145e-05,
+      "loss": 0.6767,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9989091789376905e-05,
+      "loss": 0.6886,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9989040885279044e-05,
+      "loss": 0.6911,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998898986274847e-05,
+      "loss": 0.6695,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998893872178578e-05,
+      "loss": 0.6494,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988887462391585e-05,
+      "loss": 0.6884,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988836084566495e-05,
+      "loss": 0.6403,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988784588311118e-05,
+      "loss": 0.6263,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988732973626065e-05,
+      "loss": 0.7104,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988681240511944e-05,
+      "loss": 0.6558,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998862938896937e-05,
+      "loss": 0.7346,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988577418998963e-05,
+      "loss": 0.7029,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988525330601335e-05,
+      "loss": 0.6575,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99884731237771e-05,
+      "loss": 0.6277,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988420798526885e-05,
+      "loss": 0.6278,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988368354851303e-05,
+      "loss": 0.6512,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988315792750977e-05,
+      "loss": 0.6582,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988263112226534e-05,
+      "loss": 0.6563,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99882103132786e-05,
+      "loss": 0.6312,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988157395907796e-05,
+      "loss": 0.671,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998810436011475e-05,
+      "loss": 0.7301,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9988051205900088e-05,
+      "loss": 0.7009,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987997933264446e-05,
+      "loss": 0.684,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998794454220846e-05,
+      "loss": 0.6436,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998789103273275e-05,
+      "loss": 0.6916,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987837404837957e-05,
+      "loss": 0.6431,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987783658524717e-05,
+      "loss": 0.6918,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987729793793668e-05,
+      "loss": 0.6749,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987675810645452e-05,
+      "loss": 0.6272,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987621709080698e-05,
+      "loss": 0.6625,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998756748910006e-05,
+      "loss": 0.6728,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998751315070417e-05,
+      "loss": 0.6727,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998745869389368e-05,
+      "loss": 0.7283,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987404118669233e-05,
+      "loss": 0.6727,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987349425031475e-05,
+      "loss": 0.6526,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987294612981055e-05,
+      "loss": 0.718,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987239682518628e-05,
+      "loss": 0.6871,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987184633644836e-05,
+      "loss": 0.6442,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998712946636034e-05,
+      "loss": 0.6572,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987074180665784e-05,
+      "loss": 0.6738,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9987018776561835e-05,
+      "loss": 0.6871,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986963254049144e-05,
+      "loss": 0.6738,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998690761312837e-05,
+      "loss": 0.7051,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986851853800174e-05,
+      "loss": 0.657,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986795976065213e-05,
+      "loss": 0.6326,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986739979924153e-05,
+      "loss": 0.6632,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986683865377658e-05,
+      "loss": 0.6244,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986627632426393e-05,
+      "loss": 0.6894,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986571281071024e-05,
+      "loss": 0.669,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998651481131222e-05,
+      "loss": 0.6596,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986458223150648e-05,
+      "loss": 0.735,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986401516586982e-05,
+      "loss": 0.6984,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986344691621892e-05,
+      "loss": 0.6437,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998628774825605e-05,
+      "loss": 0.6491,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998623068649014e-05,
+      "loss": 0.7071,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998617350632483e-05,
+      "loss": 0.6113,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986116207760802e-05,
+      "loss": 0.5655,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986058790798735e-05,
+      "loss": 0.661,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9986001255439307e-05,
+      "loss": 0.7126,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99859436016832e-05,
+      "loss": 0.6351,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99858858295311e-05,
+      "loss": 0.724,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985827938983695e-05,
+      "loss": 0.6657,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998576993004166e-05,
+      "loss": 0.6827,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985711802705695e-05,
+      "loss": 0.6548,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985653556976485e-05,
+      "loss": 0.7073,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985595192854718e-05,
+      "loss": 0.6465,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998553671034109e-05,
+      "loss": 0.6026,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985478109436294e-05,
+      "loss": 0.7391,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998541939014102e-05,
+      "loss": 0.714,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985360552455973e-05,
+      "loss": 0.6015,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998530159638184e-05,
+      "loss": 0.7485,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985242521919325e-05,
+      "loss": 0.7235,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985183329069132e-05,
+      "loss": 0.643,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985124017831956e-05,
+      "loss": 0.6729,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985064588208502e-05,
+      "loss": 0.6805,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9985005040199482e-05,
+      "loss": 0.7176,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998494537380559e-05,
+      "loss": 0.7054,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984885589027546e-05,
+      "loss": 0.7019,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984825685866048e-05,
+      "loss": 0.683,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998476566432181e-05,
+      "loss": 0.7019,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984705524395545e-05,
+      "loss": 0.6943,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998464526608797e-05,
+      "loss": 0.6749,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998458488939979e-05,
+      "loss": 0.6843,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984524394331727e-05,
+      "loss": 0.7013,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984463780884492e-05,
+      "loss": 0.6499,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984403049058813e-05,
+      "loss": 0.6786,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984342198855407e-05,
+      "loss": 0.6882,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984281230274993e-05,
+      "loss": 0.7124,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984220143318292e-05,
+      "loss": 0.6805,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984158937986033e-05,
+      "loss": 0.5811,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984097614278938e-05,
+      "loss": 0.6887,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9984036172197738e-05,
+      "loss": 0.6854,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998397461174316e-05,
+      "loss": 0.6233,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998391293291593e-05,
+      "loss": 0.6501,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998385113571679e-05,
+      "loss": 0.6383,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9983789220146458e-05,
+      "loss": 0.6274,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9983727186205676e-05,
+      "loss": 0.6285,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998366503389518e-05,
+      "loss": 0.7192,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9983602763215705e-05,
+      "loss": 0.6715,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998354037416799e-05,
+      "loss": 0.6664,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9983477866752775e-05,
+      "loss": 0.6312,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99834152409708e-05,
+      "loss": 0.6394,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998335249682281e-05,
+      "loss": 0.6965,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9983289634309546e-05,
+      "loss": 0.6828,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9983226653431753e-05,
+      "loss": 0.6856,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9983163554190182e-05,
+      "loss": 0.6557,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9983100336585576e-05,
+      "loss": 0.6143,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998303700061869e-05,
+      "loss": 0.6438,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998297354629027e-05,
+      "loss": 0.6269,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998290997360107e-05,
+      "loss": 0.6387,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998284628255184e-05,
+      "loss": 0.6978,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982782473143343e-05,
+      "loss": 0.7045,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982718545376332e-05,
+      "loss": 0.6099,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982654499251565e-05,
+      "loss": 0.6118,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982590334769796e-05,
+      "loss": 0.687,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982526051931795e-05,
+      "loss": 0.6417,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982461650738317e-05,
+      "loss": 0.6631,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998239713119013e-05,
+      "loss": 0.6352,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982332493287998e-05,
+      "loss": 0.6242,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982267737032685e-05,
+      "loss": 0.6579,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998220286242496e-05,
+      "loss": 0.6848,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982137869465596e-05,
+      "loss": 0.6569,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982072758155352e-05,
+      "loss": 0.6796,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9982007528495015e-05,
+      "loss": 0.6234,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998194218048535e-05,
+      "loss": 0.6612,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9981876714127132e-05,
+      "loss": 0.6573,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998181112942114e-05,
+      "loss": 0.6689,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9981745426368147e-05,
+      "loss": 0.6617,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998167960496894e-05,
+      "loss": 0.6799,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9981613665224292e-05,
+      "loss": 0.6492,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9981547607134985e-05,
+      "loss": 0.6006,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9981481430701806e-05,
+      "loss": 0.7015,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998141513592554e-05,
+      "loss": 0.6743,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998134872280697e-05,
+      "loss": 0.6573,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9981282191346886e-05,
+      "loss": 0.6176,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9981215541546072e-05,
+      "loss": 0.6557,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9981148773405322e-05,
+      "loss": 0.7244,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998108188692543e-05,
+      "loss": 0.6493,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998101488210718e-05,
+      "loss": 0.6381,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980947758951376e-05,
+      "loss": 0.6693,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980880517458813e-05,
+      "loss": 0.6679,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980813157630283e-05,
+      "loss": 0.6581,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980745679466587e-05,
+      "loss": 0.7063,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980678082968527e-05,
+      "loss": 0.7262,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980610368136898e-05,
+      "loss": 0.7239,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980542534972512e-05,
+      "loss": 0.6892,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980474583476167e-05,
+      "loss": 0.6459,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980406513648673e-05,
+      "loss": 0.6314,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.998033832549083e-05,
+      "loss": 0.6919,
+      "step": 586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980270019003452e-05,
+      "loss": 0.6956,
+      "step": 587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980201594187353e-05,
+      "loss": 0.6729,
+      "step": 588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980133051043336e-05,
+      "loss": 0.6526,
+      "step": 589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9980064389572215e-05,
+      "loss": 0.6537,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997999560977481e-05,
+      "loss": 0.7245,
+      "step": 591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979926711651928e-05,
+      "loss": 0.6586,
+      "step": 592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997985769520439e-05,
+      "loss": 0.6778,
+      "step": 593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979788560433016e-05,
+      "loss": 0.6635,
+      "step": 594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979719307338624e-05,
+      "loss": 0.6056,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979649935922033e-05,
+      "loss": 0.6642,
+      "step": 596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997958044618407e-05,
+      "loss": 0.6176,
+      "step": 597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979510838125557e-05,
+      "loss": 0.6504,
+      "step": 598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997944111174732e-05,
+      "loss": 0.6681,
+      "step": 599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979371267050184e-05,
+      "loss": 0.668,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979301304034978e-05,
+      "loss": 0.6805,
+      "step": 601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979231222702527e-05,
+      "loss": 0.5749,
+      "step": 602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997916102305367e-05,
+      "loss": 0.6246,
+      "step": 603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979090705089232e-05,
+      "loss": 0.6548,
+      "step": 604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9979020268810054e-05,
+      "loss": 0.6684,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978949714216965e-05,
+      "loss": 0.6862,
+      "step": 606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978879041310805e-05,
+      "loss": 0.6373,
+      "step": 607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978808250092408e-05,
+      "loss": 0.6562,
+      "step": 608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997873734056262e-05,
+      "loss": 0.6862,
+      "step": 609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978666312722273e-05,
+      "loss": 0.6813,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978595166572216e-05,
+      "loss": 0.6866,
+      "step": 611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978523902113292e-05,
+      "loss": 0.6897,
+      "step": 612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978452519346343e-05,
+      "loss": 0.6791,
+      "step": 613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978381018272218e-05,
+      "loss": 0.6791,
+      "step": 614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978309398891758e-05,
+      "loss": 0.596,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997823766120582e-05,
+      "loss": 0.7414,
+      "step": 616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978165805215255e-05,
+      "loss": 0.6518,
+      "step": 617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9978093830920912e-05,
+      "loss": 0.7035,
+      "step": 618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997802173832364e-05,
+      "loss": 0.687,
+      "step": 619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99779495274243e-05,
+      "loss": 0.6969,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977877198223746e-05,
+      "loss": 0.6914,
+      "step": 621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997780475072284e-05,
+      "loss": 0.6692,
+      "step": 622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997773218492243e-05,
+      "loss": 0.6312,
+      "step": 623
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977659500823387e-05,
+      "loss": 0.6515,
+      "step": 624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977586698426568e-05,
+      "loss": 0.6731,
+      "step": 625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977513777732837e-05,
+      "loss": 0.6746,
+      "step": 626
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977440738743057e-05,
+      "loss": 0.666,
+      "step": 627
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977367581458095e-05,
+      "loss": 0.6846,
+      "step": 628
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977294305878823e-05,
+      "loss": 0.6311,
+      "step": 629
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977220912006104e-05,
+      "loss": 0.7214,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997714739984081e-05,
+      "loss": 0.6589,
+      "step": 631
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997707376938381e-05,
+      "loss": 0.6794,
+      "step": 632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9977000020635978e-05,
+      "loss": 0.6651,
+      "step": 633
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9976926153598196e-05,
+      "loss": 0.6242,
+      "step": 634
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997685216827133e-05,
+      "loss": 0.636,
+      "step": 635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997677806465626e-05,
+      "loss": 0.6702,
+      "step": 636
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9976703842753865e-05,
+      "loss": 0.7023,
+      "step": 637
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997662950256503e-05,
+      "loss": 0.7261,
+      "step": 638
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9976555044090628e-05,
+      "loss": 0.6894,
+      "step": 639
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9976480467331546e-05,
+      "loss": 0.6068,
+      "step": 640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9976405772288667e-05,
+      "loss": 0.6743,
+      "step": 641
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997633095896288e-05,
+      "loss": 0.6967,
+      "step": 642
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9976256027355068e-05,
+      "loss": 0.6662,
+      "step": 643
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997618097746612e-05,
+      "loss": 0.6928,
+      "step": 644
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9976105809296925e-05,
+      "loss": 0.7312,
+      "step": 645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9976030522848378e-05,
+      "loss": 0.6117,
+      "step": 646
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975955118121367e-05,
+      "loss": 0.661,
+      "step": 647
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975879595116793e-05,
+      "loss": 0.6827,
+      "step": 648
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975803953835543e-05,
+      "loss": 0.6607,
+      "step": 649
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997572819427852e-05,
+      "loss": 0.6394,
+      "step": 650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975652316446618e-05,
+      "loss": 0.6767,
+      "step": 651
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975576320340742e-05,
+      "loss": 0.6378,
+      "step": 652
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975500205961786e-05,
+      "loss": 0.6641,
+      "step": 653
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975423973310657e-05,
+      "loss": 0.634,
+      "step": 654
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975347622388256e-05,
+      "loss": 0.6461,
+      "step": 655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975271153195496e-05,
+      "loss": 0.6987,
+      "step": 656
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997519456573327e-05,
+      "loss": 0.629,
+      "step": 657
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975117860002497e-05,
+      "loss": 0.6563,
+      "step": 658
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9975041036004087e-05,
+      "loss": 0.6754,
+      "step": 659
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997496409373894e-05,
+      "loss": 0.583,
+      "step": 660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997488703320798e-05,
+      "loss": 0.6448,
+      "step": 661
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9974809854412114e-05,
+      "loss": 0.6472,
+      "step": 662
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9974732557352258e-05,
+      "loss": 0.6221,
+      "step": 663
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9974655142029333e-05,
+      "loss": 0.6354,
+      "step": 664
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997457760844425e-05,
+      "loss": 0.6744,
+      "step": 665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9974499956597933e-05,
+      "loss": 0.6218,
+      "step": 666
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99744221864913e-05,
+      "loss": 0.6858,
+      "step": 667
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9974344298125277e-05,
+      "loss": 0.7109,
+      "step": 668
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9974266291500785e-05,
+      "loss": 0.6245,
+      "step": 669
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9974188166618743e-05,
+      "loss": 0.7175,
+      "step": 670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997410992348009e-05,
+      "loss": 0.6277,
+      "step": 671
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997403156208574e-05,
+      "loss": 0.6664,
+      "step": 672
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9973953082436635e-05,
+      "loss": 0.5916,
+      "step": 673
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9973874484533696e-05,
+      "loss": 0.6305,
+      "step": 674
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997379576837786e-05,
+      "loss": 0.7272,
+      "step": 675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9973716933970053e-05,
+      "loss": 0.6461,
+      "step": 676
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997363798131122e-05,
+      "loss": 0.659,
+      "step": 677
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9973558910402287e-05,
+      "loss": 0.6459,
+      "step": 678
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99734797212442e-05,
+      "loss": 0.6481,
+      "step": 679
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9973400413837893e-05,
+      "loss": 0.6818,
+      "step": 680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9973320988184312e-05,
+      "loss": 0.6896,
+      "step": 681
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997324144428439e-05,
+      "loss": 0.705,
+      "step": 682
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9973161782139074e-05,
+      "loss": 0.6069,
+      "step": 683
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997308200174931e-05,
+      "loss": 0.649,
+      "step": 684
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9973002103116045e-05,
+      "loss": 0.6244,
+      "step": 685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997292208624022e-05,
+      "loss": 0.6134,
+      "step": 686
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997284195112279e-05,
+      "loss": 0.6361,
+      "step": 687
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9972761697764708e-05,
+      "loss": 0.6589,
+      "step": 688
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9972681326166913e-05,
+      "loss": 0.6722,
+      "step": 689
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997260083633037e-05,
+      "loss": 0.6137,
+      "step": 690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9972520228256023e-05,
+      "loss": 0.655,
+      "step": 691
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997243950194484e-05,
+      "loss": 0.6788,
+      "step": 692
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9972358657397768e-05,
+      "loss": 0.6674,
+      "step": 693
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997227769461577e-05,
+      "loss": 0.6437,
+      "step": 694
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9972196613599802e-05,
+      "loss": 0.6946,
+      "step": 695
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997211541435083e-05,
+      "loss": 0.6197,
+      "step": 696
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9972034096869816e-05,
+      "loss": 0.6597,
+      "step": 697
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971952661157727e-05,
+      "loss": 0.6862,
+      "step": 698
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971871107215518e-05,
+      "loss": 0.6711,
+      "step": 699
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971789435044166e-05,
+      "loss": 0.6499,
+      "step": 700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971707644644636e-05,
+      "loss": 0.667,
+      "step": 701
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971625736017894e-05,
+      "loss": 0.6668,
+      "step": 702
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997154370916492e-05,
+      "loss": 0.6213,
+      "step": 703
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971461564086677e-05,
+      "loss": 0.7301,
+      "step": 704
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971379300784143e-05,
+      "loss": 0.6449,
+      "step": 705
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971296919258297e-05,
+      "loss": 0.673,
+      "step": 706
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997121441951011e-05,
+      "loss": 0.6323,
+      "step": 707
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971131801540563e-05,
+      "loss": 0.7592,
+      "step": 708
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9971049065350635e-05,
+      "loss": 0.6485,
+      "step": 709
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9970966210941303e-05,
+      "loss": 0.664,
+      "step": 710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997088323831356e-05,
+      "loss": 0.6895,
+      "step": 711
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997080014746838e-05,
+      "loss": 0.6652,
+      "step": 712
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997071693840675e-05,
+      "loss": 0.666,
+      "step": 713
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997063361112966e-05,
+      "loss": 0.6135,
+      "step": 714
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9970550165638094e-05,
+      "loss": 0.6552,
+      "step": 715
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9970466601933042e-05,
+      "loss": 0.7183,
+      "step": 716
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99703829200155e-05,
+      "loss": 0.6539,
+      "step": 717
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9970299119886453e-05,
+      "loss": 0.6946,
+      "step": 718
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9970215201546897e-05,
+      "loss": 0.6823,
+      "step": 719
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.997013116499783e-05,
+      "loss": 0.6968,
+      "step": 720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9970047010240237e-05,
+      "loss": 0.6087,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9969962737275136e-05,
+      "loss": 0.6283,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996987834610351e-05,
+      "loss": 0.6209,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996979383672636e-05,
+      "loss": 0.6954,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9969709209144697e-05,
+      "loss": 0.6573,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996962446335952e-05,
+      "loss": 0.6562,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9969539599371832e-05,
+      "loss": 0.6646,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996945461718264e-05,
+      "loss": 0.7025,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9969369516792952e-05,
+      "loss": 0.652,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996928429820378e-05,
+      "loss": 0.6749,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9969198961416127e-05,
+      "loss": 0.6288,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996911350643101e-05,
+      "loss": 0.6815,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9969027933249445e-05,
+      "loss": 0.6958,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968942241872443e-05,
+      "loss": 0.6204,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968856432301016e-05,
+      "loss": 0.7274,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968770504536187e-05,
+      "loss": 0.6811,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968684458578973e-05,
+      "loss": 0.6852,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968598294430396e-05,
+      "loss": 0.6714,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968512012091473e-05,
+      "loss": 0.6398,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996842561156323e-05,
+      "loss": 0.675,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968339092846693e-05,
+      "loss": 0.6102,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968252455942886e-05,
+      "loss": 0.6682,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968165700852835e-05,
+      "loss": 0.675,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9968078827577572e-05,
+      "loss": 0.6789,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967991836118123e-05,
+      "loss": 0.6685,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967904726475524e-05,
+      "loss": 0.6302,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99678174986508e-05,
+      "loss": 0.6546,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967730152644994e-05,
+      "loss": 0.6236,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967642688459133e-05,
+      "loss": 0.7276,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967555106094263e-05,
+      "loss": 0.6164,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967467405551418e-05,
+      "loss": 0.6692,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996737958683164e-05,
+      "loss": 0.6237,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967291649935966e-05,
+      "loss": 0.7058,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967203594865437e-05,
+      "loss": 0.6501,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9967115421621107e-05,
+      "loss": 0.6314,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996702713020401e-05,
+      "loss": 0.6275,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9966938720615202e-05,
+      "loss": 0.6657,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9966850192855728e-05,
+      "loss": 0.6485,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9966761546926633e-05,
+      "loss": 0.6413,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9966672782828975e-05,
+      "loss": 0.6994,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99665839005638e-05,
+      "loss": 0.571,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9966494900132165e-05,
+      "loss": 0.6537,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996640578153513e-05,
+      "loss": 0.6617,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9966316544773742e-05,
+      "loss": 0.697,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9966227189849068e-05,
+      "loss": 0.6728,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996613771676216e-05,
+      "loss": 0.6724,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996604812551408e-05,
+      "loss": 0.6865,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99659584161059e-05,
+      "loss": 0.6182,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965868588538672e-05,
+      "loss": 0.664,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965778642813464e-05,
+      "loss": 0.6478,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965688578931344e-05,
+      "loss": 0.6195,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965598396893376e-05,
+      "loss": 0.62,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965508096700638e-05,
+      "loss": 0.6355,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965417678354193e-05,
+      "loss": 0.6178,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965327141855116e-05,
+      "loss": 0.6532,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965236487204476e-05,
+      "loss": 0.6551,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965145714403355e-05,
+      "loss": 0.5818,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9965054823452824e-05,
+      "loss": 0.7093,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9964963814353964e-05,
+      "loss": 0.5974,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996487268710785e-05,
+      "loss": 0.678,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9964781441715563e-05,
+      "loss": 0.6419,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996469007817819e-05,
+      "loss": 0.6929,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996459859649681e-05,
+      "loss": 0.6679,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9964506996672506e-05,
+      "loss": 0.6728,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996441527870637e-05,
+      "loss": 0.6373,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9964323442599486e-05,
+      "loss": 0.6557,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9964231488352942e-05,
+      "loss": 0.6519,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9964139415967825e-05,
+      "loss": 0.6264,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9964047225445235e-05,
+      "loss": 0.6571,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9963954916786258e-05,
+      "loss": 0.7012,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9963862489991992e-05,
+      "loss": 0.7036,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996376994506353e-05,
+      "loss": 0.6963,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9963677282001975e-05,
+      "loss": 0.6553,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996358450080842e-05,
+      "loss": 0.7194,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9963491601483966e-05,
+      "loss": 0.6633,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9963398584029712e-05,
+      "loss": 0.6564,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996330544844677e-05,
+      "loss": 0.7003,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9963212194736237e-05,
+      "loss": 0.6697,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9963118822899218e-05,
+      "loss": 0.6766,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9963025332936824e-05,
+      "loss": 0.7033,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996293172485016e-05,
+      "loss": 0.6664,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9962837998640335e-05,
+      "loss": 0.6186,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9962744154308466e-05,
+      "loss": 0.6845,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996265019185566e-05,
+      "loss": 0.6507,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9962556111283033e-05,
+      "loss": 0.6528,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9962461912591702e-05,
+      "loss": 0.6812,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9962367595782782e-05,
+      "loss": 0.6321,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9962273160857394e-05,
+      "loss": 0.5897,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996217860781665e-05,
+      "loss": 0.6878,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996208393666168e-05,
+      "loss": 0.6513,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961989147393602e-05,
+      "loss": 0.6111,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996189424001354e-05,
+      "loss": 0.6622,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961799214522624e-05,
+      "loss": 0.6152,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961704070921974e-05,
+      "loss": 0.6572,
+      "step": 814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961608809212717e-05,
+      "loss": 0.7035,
+      "step": 815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961513429395988e-05,
+      "loss": 0.6253,
+      "step": 816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961417931472918e-05,
+      "loss": 0.6329,
+      "step": 817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961322315444636e-05,
+      "loss": 0.6807,
+      "step": 818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961226581312277e-05,
+      "loss": 0.6351,
+      "step": 819
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961130729076975e-05,
+      "loss": 0.6808,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9961034758739866e-05,
+      "loss": 0.6376,
+      "step": 821
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9960938670302096e-05,
+      "loss": 0.6665,
+      "step": 822
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996084246376479e-05,
+      "loss": 0.7131,
+      "step": 823
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99607461391291e-05,
+      "loss": 0.6868,
+      "step": 824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9960649696396165e-05,
+      "loss": 0.6807,
+      "step": 825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9960553135567123e-05,
+      "loss": 0.7006,
+      "step": 826
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9960456456643125e-05,
+      "loss": 0.6536,
+      "step": 827
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996035965962532e-05,
+      "loss": 0.6801,
+      "step": 828
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9960262744514844e-05,
+      "loss": 0.6191,
+      "step": 829
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996016571131286e-05,
+      "loss": 0.6118,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9960068560020507e-05,
+      "loss": 0.6537,
+      "step": 831
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9959971290638945e-05,
+      "loss": 0.6816,
+      "step": 832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9959873903169322e-05,
+      "loss": 0.6334,
+      "step": 833
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9959776397612796e-05,
+      "loss": 0.6519,
+      "step": 834
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995967877397052e-05,
+      "loss": 0.6921,
+      "step": 835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9959581032243653e-05,
+      "loss": 0.7153,
+      "step": 836
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9959483172433354e-05,
+      "loss": 0.6587,
+      "step": 837
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9959385194540784e-05,
+      "loss": 0.6968,
+      "step": 838
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9959287098567105e-05,
+      "loss": 0.6617,
+      "step": 839
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995918888451348e-05,
+      "loss": 0.7113,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9959090552381067e-05,
+      "loss": 0.6243,
+      "step": 841
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995899210217104e-05,
+      "loss": 0.6312,
+      "step": 842
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9958893533884566e-05,
+      "loss": 0.6603,
+      "step": 843
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9958794847522806e-05,
+      "loss": 0.6749,
+      "step": 844
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9958696043086938e-05,
+      "loss": 0.6222,
+      "step": 845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9958597120578133e-05,
+      "loss": 0.686,
+      "step": 846
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995849807999756e-05,
+      "loss": 0.6389,
+      "step": 847
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995839892134639e-05,
+      "loss": 0.6612,
+      "step": 848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9958299644625807e-05,
+      "loss": 0.6377,
+      "step": 849
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9958200249836984e-05,
+      "loss": 0.6757,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9958100736981102e-05,
+      "loss": 0.6573,
+      "step": 851
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9958001106059337e-05,
+      "loss": 0.6419,
+      "step": 852
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995790135707287e-05,
+      "loss": 0.6547,
+      "step": 853
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995780149002289e-05,
+      "loss": 0.7177,
+      "step": 854
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9957701504910573e-05,
+      "loss": 0.6589,
+      "step": 855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9957601401737108e-05,
+      "loss": 0.6633,
+      "step": 856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9957501180503686e-05,
+      "loss": 0.6528,
+      "step": 857
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9957400841211486e-05,
+      "loss": 0.6204,
+      "step": 858
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9957300383861708e-05,
+      "loss": 0.6685,
+      "step": 859
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9957199808455533e-05,
+      "loss": 0.6273,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995709911499416e-05,
+      "loss": 0.6372,
+      "step": 861
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9956998303478783e-05,
+      "loss": 0.588,
+      "step": 862
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9956897373910593e-05,
+      "loss": 0.6503,
+      "step": 863
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995679632629079e-05,
+      "loss": 0.6776,
+      "step": 864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9956695160620572e-05,
+      "loss": 0.6681,
+      "step": 865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9956593876901137e-05,
+      "loss": 0.6227,
+      "step": 866
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9956492475133685e-05,
+      "loss": 0.6932,
+      "step": 867
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995639095531942e-05,
+      "loss": 0.678,
+      "step": 868
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9956289317459547e-05,
+      "loss": 0.6568,
+      "step": 869
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9956187561555273e-05,
+      "loss": 0.6827,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9956085687607793e-05,
+      "loss": 0.7149,
+      "step": 871
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9955983695618327e-05,
+      "loss": 0.6758,
+      "step": 872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9955881585588077e-05,
+      "loss": 0.6459,
+      "step": 873
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995577935751826e-05,
+      "loss": 0.724,
+      "step": 874
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9955677011410082e-05,
+      "loss": 0.6303,
+      "step": 875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9955574547264763e-05,
+      "loss": 0.6415,
+      "step": 876
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995547196508351e-05,
+      "loss": 0.6768,
+      "step": 877
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9955369264867545e-05,
+      "loss": 0.5706,
+      "step": 878
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9955266446618083e-05,
+      "loss": 0.6614,
+      "step": 879
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995516351033634e-05,
+      "loss": 0.6229,
+      "step": 880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9955060456023547e-05,
+      "loss": 0.6771,
+      "step": 881
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9954957283680916e-05,
+      "loss": 0.6815,
+      "step": 882
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995485399330967e-05,
+      "loss": 0.5953,
+      "step": 883
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9954750584911043e-05,
+      "loss": 0.6611,
+      "step": 884
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9954647058486252e-05,
+      "loss": 0.628,
+      "step": 885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995454341403653e-05,
+      "loss": 0.6283,
+      "step": 886
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99544396515631e-05,
+      "loss": 0.7106,
+      "step": 887
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9954335771067196e-05,
+      "loss": 0.6685,
+      "step": 888
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995423177255005e-05,
+      "loss": 0.6758,
+      "step": 889
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9954127656012892e-05,
+      "loss": 0.6501,
+      "step": 890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9954023421456962e-05,
+      "loss": 0.6767,
+      "step": 891
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995391906888349e-05,
+      "loss": 0.6599,
+      "step": 892
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9953814598293716e-05,
+      "loss": 0.62,
+      "step": 893
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9953710009688884e-05,
+      "loss": 0.6284,
+      "step": 894
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995360530307022e-05,
+      "loss": 0.5902,
+      "step": 895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9953500478438976e-05,
+      "loss": 0.6139,
+      "step": 896
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9953395535796397e-05,
+      "loss": 0.6801,
+      "step": 897
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9953290475143715e-05,
+      "loss": 0.6635,
+      "step": 898
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995318529648219e-05,
+      "loss": 0.6484,
+      "step": 899
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995307999981306e-05,
+      "loss": 0.6516,
+      "step": 900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9952974585137576e-05,
+      "loss": 0.6761,
+      "step": 901
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9952869052456986e-05,
+      "loss": 0.6572,
+      "step": 902
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9952763401772547e-05,
+      "loss": 0.649,
+      "step": 903
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9952657633085505e-05,
+      "loss": 0.6165,
+      "step": 904
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9952551746397117e-05,
+      "loss": 0.6816,
+      "step": 905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995244574170864e-05,
+      "loss": 0.6575,
+      "step": 906
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9952339619021326e-05,
+      "loss": 0.6408,
+      "step": 907
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9952233378336436e-05,
+      "loss": 0.6727,
+      "step": 908
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995212701965523e-05,
+      "loss": 0.6615,
+      "step": 909
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995202054297897e-05,
+      "loss": 0.6523,
+      "step": 910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9951913948308913e-05,
+      "loss": 0.6624,
+      "step": 911
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995180723564633e-05,
+      "loss": 0.6226,
+      "step": 912
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9951700404992484e-05,
+      "loss": 0.6941,
+      "step": 913
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995159345634864e-05,
+      "loss": 0.6587,
+      "step": 914
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995148638971607e-05,
+      "loss": 0.6709,
+      "step": 915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9951379205096034e-05,
+      "loss": 0.6155,
+      "step": 916
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9951271902489814e-05,
+      "loss": 0.6745,
+      "step": 917
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9951164481898675e-05,
+      "loss": 0.5887,
+      "step": 918
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9951056943323892e-05,
+      "loss": 0.6985,
+      "step": 919
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9950949286766745e-05,
+      "loss": 0.6759,
+      "step": 920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99508415122285e-05,
+      "loss": 0.6747,
+      "step": 921
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995073361971044e-05,
+      "loss": 0.6896,
+      "step": 922
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9950625609213855e-05,
+      "loss": 0.6642,
+      "step": 923
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995051748074001e-05,
+      "loss": 0.7297,
+      "step": 924
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995040923429019e-05,
+      "loss": 0.6708,
+      "step": 925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.995030086986568e-05,
+      "loss": 0.6658,
+      "step": 926
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9950192387467774e-05,
+      "loss": 0.5721,
+      "step": 927
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9950083787097744e-05,
+      "loss": 0.6448,
+      "step": 928
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994997506875689e-05,
+      "loss": 0.7138,
+      "step": 929
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994986623244649e-05,
+      "loss": 0.7068,
+      "step": 930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9949757278167835e-05,
+      "loss": 0.6277,
+      "step": 931
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9949648205922228e-05,
+      "loss": 0.6207,
+      "step": 932
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9949539015710953e-05,
+      "loss": 0.6545,
+      "step": 933
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9949429707535306e-05,
+      "loss": 0.6758,
+      "step": 934
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9949320281396583e-05,
+      "loss": 0.6425,
+      "step": 935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9949210737296084e-05,
+      "loss": 0.6833,
+      "step": 936
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9949101075235107e-05,
+      "loss": 0.6376,
+      "step": 937
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9948991295214947e-05,
+      "loss": 0.6198,
+      "step": 938
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9948881397236913e-05,
+      "loss": 0.6531,
+      "step": 939
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99487713813023e-05,
+      "loss": 0.5996,
+      "step": 940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9948661247412423e-05,
+      "loss": 0.616,
+      "step": 941
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994855099556858e-05,
+      "loss": 0.6328,
+      "step": 942
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994844062577208e-05,
+      "loss": 0.6399,
+      "step": 943
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994833013802423e-05,
+      "loss": 0.6709,
+      "step": 944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994821953232634e-05,
+      "loss": 0.6997,
+      "step": 945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9948108808679727e-05,
+      "loss": 0.7059,
+      "step": 946
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99479979670857e-05,
+      "loss": 0.6378,
+      "step": 947
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994788700754557e-05,
+      "loss": 0.6833,
+      "step": 948
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994777593006066e-05,
+      "loss": 0.6088,
+      "step": 949
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9947664734632275e-05,
+      "loss": 0.6755,
+      "step": 950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9947553421261743e-05,
+      "loss": 0.6743,
+      "step": 951
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9947441989950384e-05,
+      "loss": 0.6672,
+      "step": 952
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9947330440699516e-05,
+      "loss": 0.6452,
+      "step": 953
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994721877351046e-05,
+      "loss": 0.7122,
+      "step": 954
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9947106988384542e-05,
+      "loss": 0.6517,
+      "step": 955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994699508532309e-05,
+      "loss": 0.6411,
+      "step": 956
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994688306432743e-05,
+      "loss": 0.6134,
+      "step": 957
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9946770925398885e-05,
+      "loss": 0.6593,
+      "step": 958
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9946658668538787e-05,
+      "loss": 0.6626,
+      "step": 959
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9946546293748474e-05,
+      "loss": 0.6254,
+      "step": 960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9946433801029266e-05,
+      "loss": 0.5781,
+      "step": 961
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9946321190382503e-05,
+      "loss": 0.6458,
+      "step": 962
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9946208461809523e-05,
+      "loss": 0.6922,
+      "step": 963
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9946095615311657e-05,
+      "loss": 0.6487,
+      "step": 964
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9945982650890247e-05,
+      "loss": 0.6865,
+      "step": 965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994586956854663e-05,
+      "loss": 0.6574,
+      "step": 966
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994575636828215e-05,
+      "loss": 0.652,
+      "step": 967
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994564305009815e-05,
+      "loss": 0.6921,
+      "step": 968
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9945529613995962e-05,
+      "loss": 0.6655,
+      "step": 969
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9945416059976944e-05,
+      "loss": 0.6271,
+      "step": 970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9945302388042438e-05,
+      "loss": 0.6198,
+      "step": 971
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9945188598193787e-05,
+      "loss": 0.6322,
+      "step": 972
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994507469043235e-05,
+      "loss": 0.6348,
+      "step": 973
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994496066475947e-05,
+      "loss": 0.6567,
+      "step": 974
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99448465211765e-05,
+      "loss": 0.688,
+      "step": 975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9944732259684794e-05,
+      "loss": 0.6371,
+      "step": 976
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9944617880285708e-05,
+      "loss": 0.6764,
+      "step": 977
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9944503382980594e-05,
+      "loss": 0.6615,
+      "step": 978
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9944388767770818e-05,
+      "loss": 0.654,
+      "step": 979
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994427403465773e-05,
+      "loss": 0.6322,
+      "step": 980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994415918364269e-05,
+      "loss": 0.6361,
+      "step": 981
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994404421472707e-05,
+      "loss": 0.7075,
+      "step": 982
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9943929127912227e-05,
+      "loss": 0.6598,
+      "step": 983
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994381392319952e-05,
+      "loss": 0.6476,
+      "step": 984
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9943698600590325e-05,
+      "loss": 0.6707,
+      "step": 985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9943583160085998e-05,
+      "loss": 0.6571,
+      "step": 986
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994346760168792e-05,
+      "loss": 0.6246,
+      "step": 987
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9943351925397453e-05,
+      "loss": 0.6975,
+      "step": 988
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994323613121597e-05,
+      "loss": 0.6214,
+      "step": 989
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9943120219144845e-05,
+      "loss": 0.6481,
+      "step": 990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9943004189185452e-05,
+      "loss": 0.6499,
+      "step": 991
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9942888041339164e-05,
+      "loss": 0.6718,
+      "step": 992
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9942771775607364e-05,
+      "loss": 0.6454,
+      "step": 993
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9942655391991423e-05,
+      "loss": 0.647,
+      "step": 994
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9942538890492725e-05,
+      "loss": 0.6505,
+      "step": 995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9942422271112655e-05,
+      "loss": 0.6634,
+      "step": 996
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994230553385259e-05,
+      "loss": 0.6945,
+      "step": 997
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9942188678713915e-05,
+      "loss": 0.6659,
+      "step": 998
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9942071705698016e-05,
+      "loss": 0.6857,
+      "step": 999
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9941954614806282e-05,
+      "loss": 0.6576,
+      "step": 1000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9941837406040097e-05,
+      "loss": 0.6682,
+      "step": 1001
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9941720079400855e-05,
+      "loss": 0.7274,
+      "step": 1002
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9941602634889942e-05,
+      "loss": 0.667,
+      "step": 1003
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9941485072508756e-05,
+      "loss": 0.6634,
+      "step": 1004
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994136739225869e-05,
+      "loss": 0.6417,
+      "step": 1005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9941249594141136e-05,
+      "loss": 0.6436,
+      "step": 1006
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994113167815749e-05,
+      "loss": 0.6596,
+      "step": 1007
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9941013644309154e-05,
+      "loss": 0.6626,
+      "step": 1008
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9940895492597527e-05,
+      "loss": 0.6991,
+      "step": 1009
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994077722302401e-05,
+      "loss": 0.7163,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9940658835590002e-05,
+      "loss": 0.6366,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.994054033029691e-05,
+      "loss": 0.637,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9940421707146138e-05,
+      "loss": 0.6907,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9940302966139086e-05,
+      "loss": 0.5981,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9940184107277172e-05,
+      "loss": 0.6173,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99400651305618e-05,
+      "loss": 0.6585,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9939946035994385e-05,
+      "loss": 0.618,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993982682357633e-05,
+      "loss": 0.691,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993970749330906e-05,
+      "loss": 0.624,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993958804519398e-05,
+      "loss": 0.6919,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9939468479232512e-05,
+      "loss": 0.7062,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9939348795426068e-05,
+      "loss": 0.6964,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9939228993776076e-05,
+      "loss": 0.6498,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9939109074283947e-05,
+      "loss": 0.5902,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9938989036951108e-05,
+      "loss": 0.6298,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9938868881778983e-05,
+      "loss": 0.6774,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993874860876899e-05,
+      "loss": 0.6484,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9938628217922566e-05,
+      "loss": 0.6886,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993850770924113e-05,
+      "loss": 0.634,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993838708272611e-05,
+      "loss": 0.6976,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993826633837894e-05,
+      "loss": 0.6509,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9938145476201053e-05,
+      "loss": 0.6619,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9938024496193876e-05,
+      "loss": 0.6289,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9937903398358852e-05,
+      "loss": 0.7019,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9937782182697405e-05,
+      "loss": 0.607,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9937660849210982e-05,
+      "loss": 0.7335,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9937539397901022e-05,
+      "loss": 0.6915,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9937417828768958e-05,
+      "loss": 0.6806,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9937296141816235e-05,
+      "loss": 0.6355,
+      "step": 1039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9937174337044295e-05,
+      "loss": 0.6796,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9937052414454584e-05,
+      "loss": 0.6276,
+      "step": 1041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9936930374048544e-05,
+      "loss": 0.6477,
+      "step": 1042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9936808215827624e-05,
+      "loss": 0.6175,
+      "step": 1043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9936685939793274e-05,
+      "loss": 0.6403,
+      "step": 1044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9936563545946945e-05,
+      "loss": 0.6573,
+      "step": 1045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993644103429008e-05,
+      "loss": 0.6349,
+      "step": 1046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9936318404824137e-05,
+      "loss": 0.6539,
+      "step": 1047
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9936195657550573e-05,
+      "loss": 0.648,
+      "step": 1048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993607279247084e-05,
+      "loss": 0.6935,
+      "step": 1049
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9935949809586394e-05,
+      "loss": 0.6749,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9935826708898697e-05,
+      "loss": 0.6792,
+      "step": 1051
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99357034904092e-05,
+      "loss": 0.6409,
+      "step": 1052
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993558015411937e-05,
+      "loss": 0.6528,
+      "step": 1053
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993545670003067e-05,
+      "loss": 0.6629,
+      "step": 1054
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9935333128144564e-05,
+      "loss": 0.6447,
+      "step": 1055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993520943846252e-05,
+      "loss": 0.5856,
+      "step": 1056
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993508563098599e-05,
+      "loss": 0.6782,
+      "step": 1057
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993496170571646e-05,
+      "loss": 0.6574,
+      "step": 1058
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9934837662655385e-05,
+      "loss": 0.7085,
+      "step": 1059
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9934713501804242e-05,
+      "loss": 0.6474,
+      "step": 1060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9934589223164508e-05,
+      "loss": 0.6584,
+      "step": 1061
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9934464826737647e-05,
+      "loss": 0.6658,
+      "step": 1062
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993434031252514e-05,
+      "loss": 0.6565,
+      "step": 1063
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9934215680528463e-05,
+      "loss": 0.6951,
+      "step": 1064
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993409093074909e-05,
+      "loss": 0.6446,
+      "step": 1065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9933966063188505e-05,
+      "loss": 0.67,
+      "step": 1066
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9933841077848184e-05,
+      "loss": 0.6092,
+      "step": 1067
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993371597472961e-05,
+      "loss": 0.6321,
+      "step": 1068
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9933590753834265e-05,
+      "loss": 0.6517,
+      "step": 1069
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9933465415163644e-05,
+      "loss": 0.6269,
+      "step": 1070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9933339958719216e-05,
+      "loss": 0.6794,
+      "step": 1071
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993321438450248e-05,
+      "loss": 0.6821,
+      "step": 1072
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993308869251492e-05,
+      "loss": 0.6778,
+      "step": 1073
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9932962882758032e-05,
+      "loss": 0.6271,
+      "step": 1074
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99328369552333e-05,
+      "loss": 0.6348,
+      "step": 1075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9932710909942224e-05,
+      "loss": 0.6483,
+      "step": 1076
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9932584746886293e-05,
+      "loss": 0.6324,
+      "step": 1077
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9932458466067005e-05,
+      "loss": 0.6037,
+      "step": 1078
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993233206748586e-05,
+      "loss": 0.698,
+      "step": 1079
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993220555114435e-05,
+      "loss": 0.6798,
+      "step": 1080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9932078917043983e-05,
+      "loss": 0.7007,
+      "step": 1081
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993195216518625e-05,
+      "loss": 0.6736,
+      "step": 1082
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9931825295572665e-05,
+      "loss": 0.6602,
+      "step": 1083
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9931698308204727e-05,
+      "loss": 0.6736,
+      "step": 1084
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993157120308394e-05,
+      "loss": 0.6636,
+      "step": 1085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9931443980211817e-05,
+      "loss": 0.676,
+      "step": 1086
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993131663958986e-05,
+      "loss": 0.6531,
+      "step": 1087
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9931189181219582e-05,
+      "loss": 0.6498,
+      "step": 1088
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993106160510249e-05,
+      "loss": 0.6325,
+      "step": 1089
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99309339112401e-05,
+      "loss": 0.6227,
+      "step": 1090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993080609963393e-05,
+      "loss": 0.6344,
+      "step": 1091
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9930678170285486e-05,
+      "loss": 0.6421,
+      "step": 1092
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9930550123196295e-05,
+      "loss": 0.6824,
+      "step": 1093
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9930421958367865e-05,
+      "loss": 0.5911,
+      "step": 1094
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9930293675801726e-05,
+      "loss": 0.638,
+      "step": 1095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993016527549939e-05,
+      "loss": 0.6445,
+      "step": 1096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.993003675746238e-05,
+      "loss": 0.6421,
+      "step": 1097
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992990812169223e-05,
+      "loss": 0.7435,
+      "step": 1098
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9929779368190453e-05,
+      "loss": 0.6612,
+      "step": 1099
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992965049695858e-05,
+      "loss": 0.6776,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9929521507998144e-05,
+      "loss": 0.7012,
+      "step": 1101
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9929392401310665e-05,
+      "loss": 0.6209,
+      "step": 1102
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992926317689768e-05,
+      "loss": 0.6061,
+      "step": 1103
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992913383476072e-05,
+      "loss": 0.6861,
+      "step": 1104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9929004374901315e-05,
+      "loss": 0.6645,
+      "step": 1105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9928874797321004e-05,
+      "loss": 0.6859,
+      "step": 1106
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9928745102021325e-05,
+      "loss": 0.6822,
+      "step": 1107
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992861528900381e-05,
+      "loss": 0.6597,
+      "step": 1108
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9928485358270003e-05,
+      "loss": 0.6764,
+      "step": 1109
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992835530982144e-05,
+      "loss": 0.6531,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992822514365967e-05,
+      "loss": 0.6519,
+      "step": 1111
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9928094859786228e-05,
+      "loss": 0.6316,
+      "step": 1112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9927964458202664e-05,
+      "loss": 0.6524,
+      "step": 1113
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9927833938910522e-05,
+      "loss": 0.6522,
+      "step": 1114
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992770330191135e-05,
+      "loss": 0.6765,
+      "step": 1115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992757254720669e-05,
+      "loss": 0.7068,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992744167479811e-05,
+      "loss": 0.6484,
+      "step": 1117
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9927310684687145e-05,
+      "loss": 0.6971,
+      "step": 1118
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9927179576875353e-05,
+      "loss": 0.6314,
+      "step": 1119
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992704835136429e-05,
+      "loss": 0.641,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992691700815551e-05,
+      "loss": 0.67,
+      "step": 1121
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9926785547250575e-05,
+      "loss": 0.6263,
+      "step": 1122
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9926653968651037e-05,
+      "loss": 0.6828,
+      "step": 1123
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992652227235846e-05,
+      "loss": 0.6033,
+      "step": 1124
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9926390458374406e-05,
+      "loss": 0.6729,
+      "step": 1125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9926258526700433e-05,
+      "loss": 0.6233,
+      "step": 1126
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992612647733811e-05,
+      "loss": 0.6485,
+      "step": 1127
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9925994310289e-05,
+      "loss": 0.6452,
+      "step": 1128
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992586202555467e-05,
+      "loss": 0.6819,
+      "step": 1129
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9925729623136695e-05,
+      "loss": 0.6846,
+      "step": 1130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992559710303663e-05,
+      "loss": 0.6311,
+      "step": 1131
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9925464465256064e-05,
+      "loss": 0.6223,
+      "step": 1132
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9925331709796556e-05,
+      "loss": 0.6002,
+      "step": 1133
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992519883665969e-05,
+      "loss": 0.6376,
+      "step": 1134
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992506584584703e-05,
+      "loss": 0.6576,
+      "step": 1135
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9924932737360162e-05,
+      "loss": 0.6465,
+      "step": 1136
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992479951120066e-05,
+      "loss": 0.6861,
+      "step": 1137
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9924666167370108e-05,
+      "loss": 0.6449,
+      "step": 1138
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992453270587008e-05,
+      "loss": 0.6907,
+      "step": 1139
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9924399126702166e-05,
+      "loss": 0.6556,
+      "step": 1140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9924265429867943e-05,
+      "loss": 0.6818,
+      "step": 1141
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9924131615369e-05,
+      "loss": 0.646,
+      "step": 1142
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9923997683206925e-05,
+      "loss": 0.6933,
+      "step": 1143
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99238636333833e-05,
+      "loss": 0.6312,
+      "step": 1144
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992372946589972e-05,
+      "loss": 0.6395,
+      "step": 1145
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9923595180757773e-05,
+      "loss": 0.6488,
+      "step": 1146
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992346077795905e-05,
+      "loss": 0.6489,
+      "step": 1147
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9923326257505145e-05,
+      "loss": 0.6232,
+      "step": 1148
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992319161939766e-05,
+      "loss": 0.6601,
+      "step": 1149
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9923056863638183e-05,
+      "loss": 0.6524,
+      "step": 1150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992292199022831e-05,
+      "loss": 0.6275,
+      "step": 1151
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992278699916965e-05,
+      "loss": 0.5776,
+      "step": 1152
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9922651890463793e-05,
+      "loss": 0.6589,
+      "step": 1153
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9922516664112352e-05,
+      "loss": 0.673,
+      "step": 1154
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9922381320116918e-05,
+      "loss": 0.6397,
+      "step": 1155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9922245858479104e-05,
+      "loss": 0.6543,
+      "step": 1156
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9922110279200512e-05,
+      "loss": 0.6268,
+      "step": 1157
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9921974582282753e-05,
+      "loss": 0.6548,
+      "step": 1158
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992183876772743e-05,
+      "loss": 0.7301,
+      "step": 1159
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9921702835536164e-05,
+      "loss": 0.5919,
+      "step": 1160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9921566785710557e-05,
+      "loss": 0.6387,
+      "step": 1161
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9921430618252222e-05,
+      "loss": 0.6725,
+      "step": 1162
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992129433316278e-05,
+      "loss": 0.6473,
+      "step": 1163
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9921157930443843e-05,
+      "loss": 0.6224,
+      "step": 1164
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9921021410097024e-05,
+      "loss": 0.7047,
+      "step": 1165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992088477212395e-05,
+      "loss": 0.6174,
+      "step": 1166
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9920748016526238e-05,
+      "loss": 0.69,
+      "step": 1167
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9920611143305507e-05,
+      "loss": 0.6087,
+      "step": 1168
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9920474152463378e-05,
+      "loss": 0.6355,
+      "step": 1169
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9920337044001482e-05,
+      "loss": 0.6484,
+      "step": 1170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992019981792144e-05,
+      "loss": 0.7094,
+      "step": 1171
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.992006247422488e-05,
+      "loss": 0.7021,
+      "step": 1172
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991992501291343e-05,
+      "loss": 0.6468,
+      "step": 1173
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9919787433988725e-05,
+      "loss": 0.6814,
+      "step": 1174
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9919649737452385e-05,
+      "loss": 0.6521,
+      "step": 1175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9919511923306053e-05,
+      "loss": 0.6325,
+      "step": 1176
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9919373991551355e-05,
+      "loss": 0.6266,
+      "step": 1177
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9919235942189933e-05,
+      "loss": 0.7159,
+      "step": 1178
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991909777522342e-05,
+      "loss": 0.6689,
+      "step": 1179
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9918959490653455e-05,
+      "loss": 0.6934,
+      "step": 1180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991882108848168e-05,
+      "loss": 0.6841,
+      "step": 1181
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9918682568709732e-05,
+      "loss": 0.6274,
+      "step": 1182
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9918543931339255e-05,
+      "loss": 0.6321,
+      "step": 1183
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991840517637189e-05,
+      "loss": 0.665,
+      "step": 1184
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991826630380929e-05,
+      "loss": 0.6128,
+      "step": 1185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9918127313653094e-05,
+      "loss": 0.6636,
+      "step": 1186
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991798820590495e-05,
+      "loss": 0.7125,
+      "step": 1187
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9917848980566515e-05,
+      "loss": 0.5833,
+      "step": 1188
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991770963763943e-05,
+      "loss": 0.6172,
+      "step": 1189
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991757017712535e-05,
+      "loss": 0.6579,
+      "step": 1190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9917430599025934e-05,
+      "loss": 0.6499,
+      "step": 1191
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9917290903342826e-05,
+      "loss": 0.649,
+      "step": 1192
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9917151090077694e-05,
+      "loss": 0.673,
+      "step": 1193
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991701115923219e-05,
+      "loss": 0.6416,
+      "step": 1194
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9916871110807972e-05,
+      "loss": 0.6226,
+      "step": 1195
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9916730944806706e-05,
+      "loss": 0.6329,
+      "step": 1196
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9916590661230047e-05,
+      "loss": 0.6454,
+      "step": 1197
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9916450260079662e-05,
+      "loss": 0.6165,
+      "step": 1198
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9916309741357214e-05,
+      "loss": 0.6876,
+      "step": 1199
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991616910506437e-05,
+      "loss": 0.6788,
+      "step": 1200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9916028351202798e-05,
+      "loss": 0.6536,
+      "step": 1201
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9915887479774164e-05,
+      "loss": 0.6663,
+      "step": 1202
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9915746490780145e-05,
+      "loss": 0.5825,
+      "step": 1203
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99156053842224e-05,
+      "loss": 0.7305,
+      "step": 1204
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9915464160102614e-05,
+      "loss": 0.6215,
+      "step": 1205
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991532281842246e-05,
+      "loss": 0.6503,
+      "step": 1206
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991518135918361e-05,
+      "loss": 0.6364,
+      "step": 1207
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991503978238774e-05,
+      "loss": 0.6243,
+      "step": 1208
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991489808803653e-05,
+      "loss": 0.6171,
+      "step": 1209
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9914756276131664e-05,
+      "loss": 0.6022,
+      "step": 1210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991461434667482e-05,
+      "loss": 0.6931,
+      "step": 1211
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991447229966768e-05,
+      "loss": 0.6628,
+      "step": 1212
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9914330135111925e-05,
+      "loss": 0.6439,
+      "step": 1213
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991418785300925e-05,
+      "loss": 0.668,
+      "step": 1214
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9914045453361333e-05,
+      "loss": 0.6198,
+      "step": 1215
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9913902936169868e-05,
+      "loss": 0.6636,
+      "step": 1216
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9913760301436542e-05,
+      "loss": 0.6635,
+      "step": 1217
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9913617549163046e-05,
+      "loss": 0.6834,
+      "step": 1218
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9913474679351073e-05,
+      "loss": 0.5948,
+      "step": 1219
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991333169200232e-05,
+      "loss": 0.6185,
+      "step": 1220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9913188587118477e-05,
+      "loss": 0.6185,
+      "step": 1221
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9913045364701244e-05,
+      "loss": 0.6673,
+      "step": 1222
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9912902024752318e-05,
+      "loss": 0.65,
+      "step": 1223
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9912758567273396e-05,
+      "loss": 0.6195,
+      "step": 1224
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9912614992266183e-05,
+      "loss": 0.688,
+      "step": 1225
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991247129973238e-05,
+      "loss": 0.6656,
+      "step": 1226
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9912327489673694e-05,
+      "loss": 0.6057,
+      "step": 1227
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9912183562091822e-05,
+      "loss": 0.6479,
+      "step": 1228
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9912039516988475e-05,
+      "loss": 0.7484,
+      "step": 1229
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991189535436536e-05,
+      "loss": 0.6628,
+      "step": 1230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9911751074224187e-05,
+      "loss": 0.6569,
+      "step": 1231
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991160667656667e-05,
+      "loss": 0.6042,
+      "step": 1232
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9911462161394515e-05,
+      "loss": 0.6056,
+      "step": 1233
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991131752870944e-05,
+      "loss": 0.6185,
+      "step": 1234
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9911172778513156e-05,
+      "loss": 0.6059,
+      "step": 1235
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9911027910807383e-05,
+      "loss": 0.7182,
+      "step": 1236
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9910882925593834e-05,
+      "loss": 0.6209,
+      "step": 1237
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991073782287423e-05,
+      "loss": 0.688,
+      "step": 1238
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9910592602650295e-05,
+      "loss": 0.616,
+      "step": 1239
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9910447264923747e-05,
+      "loss": 0.629,
+      "step": 1240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.991030180969631e-05,
+      "loss": 0.6231,
+      "step": 1241
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9910156236969706e-05,
+      "loss": 0.65,
+      "step": 1242
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9910010546745668e-05,
+      "loss": 0.622,
+      "step": 1243
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9909864739025915e-05,
+      "loss": 0.6233,
+      "step": 1244
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990971881381218e-05,
+      "loss": 0.6408,
+      "step": 1245
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9909572771106195e-05,
+      "loss": 0.6481,
+      "step": 1246
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9909426610909686e-05,
+      "loss": 0.6068,
+      "step": 1247
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990928033322439e-05,
+      "loss": 0.638,
+      "step": 1248
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990913393805204e-05,
+      "loss": 0.6249,
+      "step": 1249
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9908987425394373e-05,
+      "loss": 0.6598,
+      "step": 1250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9908840795253126e-05,
+      "loss": 0.6811,
+      "step": 1251
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9908694047630033e-05,
+      "loss": 0.665,
+      "step": 1252
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990854718252684e-05,
+      "loss": 0.7012,
+      "step": 1253
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9908400199945286e-05,
+      "loss": 0.6408,
+      "step": 1254
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9908253099887114e-05,
+      "loss": 0.7042,
+      "step": 1255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9908105882354066e-05,
+      "loss": 0.6041,
+      "step": 1256
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990795854734789e-05,
+      "loss": 0.5998,
+      "step": 1257
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9907811094870332e-05,
+      "loss": 0.6611,
+      "step": 1258
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9907663524923137e-05,
+      "loss": 0.6815,
+      "step": 1259
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990751583750806e-05,
+      "loss": 0.6243,
+      "step": 1260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990736803262685e-05,
+      "loss": 0.5945,
+      "step": 1261
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990722011028126e-05,
+      "loss": 0.6729,
+      "step": 1262
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990707207047304e-05,
+      "loss": 0.6078,
+      "step": 1263
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990692391320395e-05,
+      "loss": 0.6977,
+      "step": 1264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9906775638475745e-05,
+      "loss": 0.6339,
+      "step": 1265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9906627246290182e-05,
+      "loss": 0.6844,
+      "step": 1266
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990647873664902e-05,
+      "loss": 0.6349,
+      "step": 1267
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9906330109554023e-05,
+      "loss": 0.5935,
+      "step": 1268
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990618136500695e-05,
+      "loss": 0.6314,
+      "step": 1269
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9906032503009567e-05,
+      "loss": 0.5971,
+      "step": 1270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9905883523563637e-05,
+      "loss": 0.5836,
+      "step": 1271
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9905734426670924e-05,
+      "loss": 0.6617,
+      "step": 1272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99055852123332e-05,
+      "loss": 0.675,
+      "step": 1273
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9905435880552233e-05,
+      "loss": 0.6747,
+      "step": 1274
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9905286431329793e-05,
+      "loss": 0.6227,
+      "step": 1275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9905136864667654e-05,
+      "loss": 0.6117,
+      "step": 1276
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9904987180567585e-05,
+      "loss": 0.6002,
+      "step": 1277
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9904837379031365e-05,
+      "loss": 0.6462,
+      "step": 1278
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9904687460060765e-05,
+      "loss": 0.6675,
+      "step": 1279
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9904537423657566e-05,
+      "loss": 0.6415,
+      "step": 1280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990438726982355e-05,
+      "loss": 0.6785,
+      "step": 1281
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990423699856049e-05,
+      "loss": 0.642,
+      "step": 1282
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9904086609870176e-05,
+      "loss": 0.6322,
+      "step": 1283
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9903936103754383e-05,
+      "loss": 0.7202,
+      "step": 1284
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99037854802149e-05,
+      "loss": 0.6776,
+      "step": 1285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990363473925351e-05,
+      "loss": 0.7208,
+      "step": 1286
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9903483880872005e-05,
+      "loss": 0.6708,
+      "step": 1287
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9903332905072167e-05,
+      "loss": 0.6082,
+      "step": 1288
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9903181811855795e-05,
+      "loss": 0.6612,
+      "step": 1289
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990303060122467e-05,
+      "loss": 0.6859,
+      "step": 1290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9902879273180592e-05,
+      "loss": 0.6536,
+      "step": 1291
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9902727827725355e-05,
+      "loss": 0.6328,
+      "step": 1292
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9902576264860747e-05,
+      "loss": 0.6062,
+      "step": 1293
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9902424584588574e-05,
+      "loss": 0.6671,
+      "step": 1294
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.990227278691063e-05,
+      "loss": 0.7005,
+      "step": 1295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9902120871828717e-05,
+      "loss": 0.6873,
+      "step": 1296
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9901968839344635e-05,
+      "loss": 0.6075,
+      "step": 1297
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9901816689460188e-05,
+      "loss": 0.6664,
+      "step": 1298
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9901664422177177e-05,
+      "loss": 0.6847,
+      "step": 1299
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9901512037497405e-05,
+      "loss": 0.6691,
+      "step": 1300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9901359535422686e-05,
+      "loss": 0.6783,
+      "step": 1301
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9901206915954824e-05,
+      "loss": 0.6504,
+      "step": 1302
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9901054179095626e-05,
+      "loss": 0.6447,
+      "step": 1303
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9900901324846907e-05,
+      "loss": 0.6051,
+      "step": 1304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.990074835321048e-05,
+      "loss": 0.6807,
+      "step": 1305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9900595264188157e-05,
+      "loss": 0.7003,
+      "step": 1306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9900442057781753e-05,
+      "loss": 0.6297,
+      "step": 1307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9900288733993084e-05,
+      "loss": 0.6416,
+      "step": 1308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9900135292823964e-05,
+      "loss": 0.6375,
+      "step": 1309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989998173427622e-05,
+      "loss": 0.7014,
+      "step": 1310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9899828058351668e-05,
+      "loss": 0.6584,
+      "step": 1311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989967426505213e-05,
+      "loss": 0.6365,
+      "step": 1312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9899520354379435e-05,
+      "loss": 0.6424,
+      "step": 1313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98993663263354e-05,
+      "loss": 0.6478,
+      "step": 1314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9899212180921854e-05,
+      "loss": 0.6692,
+      "step": 1315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9899057918140628e-05,
+      "loss": 0.6653,
+      "step": 1316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9898903537993544e-05,
+      "loss": 0.6362,
+      "step": 1317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989874904048244e-05,
+      "loss": 0.6465,
+      "step": 1318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989859442560914e-05,
+      "loss": 0.6551,
+      "step": 1319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989843969337548e-05,
+      "loss": 0.5773,
+      "step": 1320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98982848437833e-05,
+      "loss": 0.6539,
+      "step": 1321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9898129876834434e-05,
+      "loss": 0.6845,
+      "step": 1322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9897974792530714e-05,
+      "loss": 0.7116,
+      "step": 1323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9897819590873985e-05,
+      "loss": 0.6396,
+      "step": 1324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989766427186608e-05,
+      "loss": 0.6699,
+      "step": 1325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9897508835508843e-05,
+      "loss": 0.6493,
+      "step": 1326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9897353281804123e-05,
+      "loss": 0.6554,
+      "step": 1327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9897197610753755e-05,
+      "loss": 0.6815,
+      "step": 1328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989704182235959e-05,
+      "loss": 0.6426,
+      "step": 1329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9896885916623478e-05,
+      "loss": 0.6327,
+      "step": 1330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989672989354726e-05,
+      "loss": 0.6879,
+      "step": 1331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9896573753132794e-05,
+      "loss": 0.6964,
+      "step": 1332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9896417495381925e-05,
+      "loss": 0.6737,
+      "step": 1333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9896261120296505e-05,
+      "loss": 0.6901,
+      "step": 1334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9896104627878393e-05,
+      "loss": 0.6876,
+      "step": 1335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9895948018129444e-05,
+      "loss": 0.6862,
+      "step": 1336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989579129105151e-05,
+      "loss": 0.7041,
+      "step": 1337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9895634446646453e-05,
+      "loss": 0.6725,
+      "step": 1338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989547748491613e-05,
+      "loss": 0.662,
+      "step": 1339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9895320405862403e-05,
+      "loss": 0.663,
+      "step": 1340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989516320948714e-05,
+      "loss": 0.6535,
+      "step": 1341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9895005895792194e-05,
+      "loss": 0.6184,
+      "step": 1342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989484846477944e-05,
+      "loss": 0.63,
+      "step": 1343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9894690916450737e-05,
+      "loss": 0.6764,
+      "step": 1344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9894533250807958e-05,
+      "loss": 0.6407,
+      "step": 1345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9894375467852972e-05,
+      "loss": 0.7016,
+      "step": 1346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9894217567587647e-05,
+      "loss": 0.6326,
+      "step": 1347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9894059550013852e-05,
+      "loss": 0.633,
+      "step": 1348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9893901415133472e-05,
+      "loss": 0.6352,
+      "step": 1349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9893743162948373e-05,
+      "loss": 0.6563,
+      "step": 1350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9893584793460432e-05,
+      "loss": 0.6657,
+      "step": 1351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9893426306671527e-05,
+      "loss": 0.6722,
+      "step": 1352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989326770258354e-05,
+      "loss": 0.6753,
+      "step": 1353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9893108981198348e-05,
+      "loss": 0.6914,
+      "step": 1354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9892950142517834e-05,
+      "loss": 0.669,
+      "step": 1355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9892791186543884e-05,
+      "loss": 0.6292,
+      "step": 1356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989263211327838e-05,
+      "loss": 0.6698,
+      "step": 1357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9892472922723205e-05,
+      "loss": 0.5943,
+      "step": 1358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.989231361488025e-05,
+      "loss": 0.6626,
+      "step": 1359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9892154189751403e-05,
+      "loss": 0.6376,
+      "step": 1360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9891994647338556e-05,
+      "loss": 0.6931,
+      "step": 1361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98918349876436e-05,
+      "loss": 0.6318,
+      "step": 1362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9891675210668423e-05,
+      "loss": 0.659,
+      "step": 1363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9891515316414926e-05,
+      "loss": 0.6579,
+      "step": 1364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9891355304885e-05,
+      "loss": 0.6273,
+      "step": 1365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9891195176080548e-05,
+      "loss": 0.6335,
+      "step": 1366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9891034930003463e-05,
+      "loss": 0.6412,
+      "step": 1367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9890874566655648e-05,
+      "loss": 0.5981,
+      "step": 1368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9890714086039e-05,
+      "loss": 0.6142,
+      "step": 1369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9890553488155425e-05,
+      "loss": 0.6059,
+      "step": 1370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9890392773006828e-05,
+      "loss": 0.6973,
+      "step": 1371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9890231940595112e-05,
+      "loss": 0.619,
+      "step": 1372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9890070990922187e-05,
+      "loss": 0.6237,
+      "step": 1373
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988990992398996e-05,
+      "loss": 0.6392,
+      "step": 1374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9889748739800336e-05,
+      "loss": 0.6901,
+      "step": 1375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9889587438355235e-05,
+      "loss": 0.7011,
+      "step": 1376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9889426019656562e-05,
+      "loss": 0.6798,
+      "step": 1377
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9889264483706233e-05,
+      "loss": 0.6473,
+      "step": 1378
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9889102830506167e-05,
+      "loss": 0.7064,
+      "step": 1379
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9888941060058272e-05,
+      "loss": 0.6719,
+      "step": 1380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9888779172364474e-05,
+      "loss": 0.6724,
+      "step": 1381
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988861716742669e-05,
+      "loss": 0.6901,
+      "step": 1382
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9888455045246838e-05,
+      "loss": 0.6484,
+      "step": 1383
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9888292805826843e-05,
+      "loss": 0.6231,
+      "step": 1384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988813044916863e-05,
+      "loss": 0.6263,
+      "step": 1385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988796797527412e-05,
+      "loss": 0.643,
+      "step": 1386
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9887805384145244e-05,
+      "loss": 0.6264,
+      "step": 1387
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9887642675783926e-05,
+      "loss": 0.669,
+      "step": 1388
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9887479850192094e-05,
+      "loss": 0.6464,
+      "step": 1389
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988731690737168e-05,
+      "loss": 0.6208,
+      "step": 1390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988715384732462e-05,
+      "loss": 0.6935,
+      "step": 1391
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988699067005284e-05,
+      "loss": 0.6288,
+      "step": 1392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9886827375558283e-05,
+      "loss": 0.612,
+      "step": 1393
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988666396384288e-05,
+      "loss": 0.6623,
+      "step": 1394
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9886500434908565e-05,
+      "loss": 0.6891,
+      "step": 1395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9886336788757285e-05,
+      "loss": 0.6628,
+      "step": 1396
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9886173025390977e-05,
+      "loss": 0.6336,
+      "step": 1397
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9886009144811575e-05,
+      "loss": 0.6723,
+      "step": 1398
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9885845147021035e-05,
+      "loss": 0.6729,
+      "step": 1399
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988568103202129e-05,
+      "loss": 0.6553,
+      "step": 1400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9885516799814298e-05,
+      "loss": 0.6421,
+      "step": 1401
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9885352450401994e-05,
+      "loss": 0.6021,
+      "step": 1402
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988518798378633e-05,
+      "loss": 0.6737,
+      "step": 1403
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9885023399969262e-05,
+      "loss": 0.6451,
+      "step": 1404
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9884858698952735e-05,
+      "loss": 0.6484,
+      "step": 1405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9884693880738703e-05,
+      "loss": 0.6756,
+      "step": 1406
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9884528945329124e-05,
+      "loss": 0.6399,
+      "step": 1407
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9884363892725948e-05,
+      "loss": 0.6359,
+      "step": 1408
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9884198722931134e-05,
+      "loss": 0.6248,
+      "step": 1409
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988403343594664e-05,
+      "loss": 0.6312,
+      "step": 1410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9883868031774425e-05,
+      "loss": 0.6324,
+      "step": 1411
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9883702510416456e-05,
+      "loss": 0.7058,
+      "step": 1412
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9883536871874688e-05,
+      "loss": 0.6388,
+      "step": 1413
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9883371116151088e-05,
+      "loss": 0.6312,
+      "step": 1414
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9883205243247622e-05,
+      "loss": 0.6651,
+      "step": 1415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9883039253166255e-05,
+      "loss": 0.652,
+      "step": 1416
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9882873145908955e-05,
+      "loss": 0.6632,
+      "step": 1417
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988270692147769e-05,
+      "loss": 0.6767,
+      "step": 1418
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9882540579874437e-05,
+      "loss": 0.6731,
+      "step": 1419
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988237412110116e-05,
+      "loss": 0.6888,
+      "step": 1420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988220754515984e-05,
+      "loss": 0.6435,
+      "step": 1421
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9882040852052444e-05,
+      "loss": 0.6407,
+      "step": 1422
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9881874041780956e-05,
+      "loss": 0.7255,
+      "step": 1423
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988170711434735e-05,
+      "loss": 0.7067,
+      "step": 1424
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9881540069753606e-05,
+      "loss": 0.6413,
+      "step": 1425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9881372908001704e-05,
+      "loss": 0.6557,
+      "step": 1426
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988120562909363e-05,
+      "loss": 0.6536,
+      "step": 1427
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9881038233031356e-05,
+      "loss": 0.6743,
+      "step": 1428
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9880870719816882e-05,
+      "loss": 0.6251,
+      "step": 1429
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.988070308945218e-05,
+      "loss": 0.6675,
+      "step": 1430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9880535341939245e-05,
+      "loss": 0.6477,
+      "step": 1431
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9880367477280064e-05,
+      "loss": 0.6542,
+      "step": 1432
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9880199495476632e-05,
+      "loss": 0.6795,
+      "step": 1433
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9880031396530933e-05,
+      "loss": 0.6401,
+      "step": 1434
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9879863180444965e-05,
+      "loss": 0.7072,
+      "step": 1435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987969484722072e-05,
+      "loss": 0.6383,
+      "step": 1436
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9879526396860194e-05,
+      "loss": 0.6869,
+      "step": 1437
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9879357829365385e-05,
+      "loss": 0.6173,
+      "step": 1438
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9879189144738294e-05,
+      "loss": 0.6392,
+      "step": 1439
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987902034298092e-05,
+      "loss": 0.6364,
+      "step": 1440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9878851424095254e-05,
+      "loss": 0.6799,
+      "step": 1441
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9878682388083314e-05,
+      "loss": 0.6432,
+      "step": 1442
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9878513234947097e-05,
+      "loss": 0.6622,
+      "step": 1443
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987834396468861e-05,
+      "loss": 0.6867,
+      "step": 1444
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987817457730986e-05,
+      "loss": 0.6514,
+      "step": 1445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9878005072812848e-05,
+      "loss": 0.6882,
+      "step": 1446
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9877835451199598e-05,
+      "loss": 0.7195,
+      "step": 1447
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9877665712472106e-05,
+      "loss": 0.7126,
+      "step": 1448
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98774958566324e-05,
+      "loss": 0.6164,
+      "step": 1449
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987732588368248e-05,
+      "loss": 0.6606,
+      "step": 1450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9877155793624366e-05,
+      "loss": 0.7277,
+      "step": 1451
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9876985586460078e-05,
+      "loss": 0.6283,
+      "step": 1452
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987681526219163e-05,
+      "loss": 0.6189,
+      "step": 1453
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9876644820821044e-05,
+      "loss": 0.7051,
+      "step": 1454
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987647426235034e-05,
+      "loss": 0.6037,
+      "step": 1455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987630358678154e-05,
+      "loss": 0.5958,
+      "step": 1456
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9876132794116666e-05,
+      "loss": 0.6899,
+      "step": 1457
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9875961884357743e-05,
+      "loss": 0.6467,
+      "step": 1458
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9875790857506797e-05,
+      "loss": 0.6572,
+      "step": 1459
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9875619713565863e-05,
+      "loss": 0.6796,
+      "step": 1460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987544845253696e-05,
+      "loss": 0.6563,
+      "step": 1461
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9875277074422127e-05,
+      "loss": 0.6815,
+      "step": 1462
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9875105579223392e-05,
+      "loss": 0.6722,
+      "step": 1463
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9874933966942786e-05,
+      "loss": 0.6836,
+      "step": 1464
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9874762237582348e-05,
+      "loss": 0.6554,
+      "step": 1465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9874590391144108e-05,
+      "loss": 0.6118,
+      "step": 1466
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9874418427630113e-05,
+      "loss": 0.5869,
+      "step": 1467
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9874246347042395e-05,
+      "loss": 0.5536,
+      "step": 1468
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987407414938299e-05,
+      "loss": 0.6216,
+      "step": 1469
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987390183465395e-05,
+      "loss": 0.6683,
+      "step": 1470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9873729402857315e-05,
+      "loss": 0.7107,
+      "step": 1471
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9873556853995123e-05,
+      "loss": 0.6345,
+      "step": 1472
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9873384188069427e-05,
+      "loss": 0.6362,
+      "step": 1473
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987321140508227e-05,
+      "loss": 0.6946,
+      "step": 1474
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9873038505035705e-05,
+      "loss": 0.6397,
+      "step": 1475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9872865487931776e-05,
+      "loss": 0.6781,
+      "step": 1476
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9872692353772538e-05,
+      "loss": 0.6732,
+      "step": 1477
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9872519102560045e-05,
+      "loss": 0.6368,
+      "step": 1478
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9872345734296348e-05,
+      "loss": 0.6446,
+      "step": 1479
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9872172248983505e-05,
+      "loss": 0.6964,
+      "step": 1480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987199864662357e-05,
+      "loss": 0.6075,
+      "step": 1481
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9871824927218603e-05,
+      "loss": 0.6619,
+      "step": 1482
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9871651090770664e-05,
+      "loss": 0.6759,
+      "step": 1483
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9871477137281813e-05,
+      "loss": 0.7027,
+      "step": 1484
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9871303066754113e-05,
+      "loss": 0.6964,
+      "step": 1485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9871128879189628e-05,
+      "loss": 0.6691,
+      "step": 1486
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9870954574590425e-05,
+      "loss": 0.6382,
+      "step": 1487
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9870780152958566e-05,
+      "loss": 0.6271,
+      "step": 1488
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9870605614296123e-05,
+      "loss": 0.6333,
+      "step": 1489
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9870430958605165e-05,
+      "loss": 0.589,
+      "step": 1490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9870256185887758e-05,
+      "loss": 0.6758,
+      "step": 1491
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.987008129614598e-05,
+      "loss": 0.6234,
+      "step": 1492
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9869906289381906e-05,
+      "loss": 0.6719,
+      "step": 1493
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9869731165597603e-05,
+      "loss": 0.6746,
+      "step": 1494
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9869555924795153e-05,
+      "loss": 0.6948,
+      "step": 1495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9869380566976637e-05,
+      "loss": 0.6403,
+      "step": 1496
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9869205092144123e-05,
+      "loss": 0.649,
+      "step": 1497
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98690295002997e-05,
+      "loss": 0.6473,
+      "step": 1498
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9868853791445448e-05,
+      "loss": 0.72,
+      "step": 1499
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986867796558345e-05,
+      "loss": 0.597,
+      "step": 1500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986850202271579e-05,
+      "loss": 0.6726,
+      "step": 1501
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986832596284456e-05,
+      "loss": 0.6268,
+      "step": 1502
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9868149785971837e-05,
+      "loss": 0.5721,
+      "step": 1503
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9867973492099715e-05,
+      "loss": 0.6725,
+      "step": 1504
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9867797081230285e-05,
+      "loss": 0.6535,
+      "step": 1505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986762055336564e-05,
+      "loss": 0.6679,
+      "step": 1506
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9867443908507868e-05,
+      "loss": 0.636,
+      "step": 1507
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9867267146659067e-05,
+      "loss": 0.6293,
+      "step": 1508
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9867090267821333e-05,
+      "loss": 0.694,
+      "step": 1509
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986691327199676e-05,
+      "loss": 0.6236,
+      "step": 1510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986673615918745e-05,
+      "loss": 0.6556,
+      "step": 1511
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98665589293955e-05,
+      "loss": 0.6188,
+      "step": 1512
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9866381582623012e-05,
+      "loss": 0.6471,
+      "step": 1513
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9866204118872093e-05,
+      "loss": 0.6375,
+      "step": 1514
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986602653814484e-05,
+      "loss": 0.6576,
+      "step": 1515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9865848840443363e-05,
+      "loss": 0.6541,
+      "step": 1516
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9865671025769767e-05,
+      "loss": 0.6435,
+      "step": 1517
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986549309412616e-05,
+      "loss": 0.6192,
+      "step": 1518
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9865315045514653e-05,
+      "loss": 0.6498,
+      "step": 1519
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9865136879937357e-05,
+      "loss": 0.6623,
+      "step": 1520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9864958597396385e-05,
+      "loss": 0.6449,
+      "step": 1521
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9864780197893848e-05,
+      "loss": 0.6395,
+      "step": 1522
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9864601681431862e-05,
+      "loss": 0.6753,
+      "step": 1523
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986442304801255e-05,
+      "loss": 0.5958,
+      "step": 1524
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986424429763802e-05,
+      "loss": 0.6662,
+      "step": 1525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9864065430310392e-05,
+      "loss": 0.6639,
+      "step": 1526
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986388644603179e-05,
+      "loss": 0.6664,
+      "step": 1527
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9863707344804345e-05,
+      "loss": 0.6815,
+      "step": 1528
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9863528126630166e-05,
+      "loss": 0.6952,
+      "step": 1529
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9863348791511382e-05,
+      "loss": 0.669,
+      "step": 1530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9863169339450126e-05,
+      "loss": 0.6493,
+      "step": 1531
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9862989770448517e-05,
+      "loss": 0.6283,
+      "step": 1532
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9862810084508687e-05,
+      "loss": 0.66,
+      "step": 1533
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986263028163277e-05,
+      "loss": 0.668,
+      "step": 1534
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986245036182289e-05,
+      "loss": 0.669,
+      "step": 1535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986227032508119e-05,
+      "loss": 0.6989,
+      "step": 1536
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9862090171409795e-05,
+      "loss": 0.6742,
+      "step": 1537
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986190990081085e-05,
+      "loss": 0.6209,
+      "step": 1538
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986172951328648e-05,
+      "loss": 0.6729,
+      "step": 1539
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986154900883884e-05,
+      "loss": 0.6454,
+      "step": 1540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986136838747006e-05,
+      "loss": 0.6318,
+      "step": 1541
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986118764918228e-05,
+      "loss": 0.632,
+      "step": 1542
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9861006793977648e-05,
+      "loss": 0.6394,
+      "step": 1543
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9860825821858304e-05,
+      "loss": 0.6525,
+      "step": 1544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98606447328264e-05,
+      "loss": 0.6132,
+      "step": 1545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9860463526884074e-05,
+      "loss": 0.7398,
+      "step": 1546
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9860282204033484e-05,
+      "loss": 0.6128,
+      "step": 1547
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.986010076427677e-05,
+      "loss": 0.6755,
+      "step": 1548
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985991920761609e-05,
+      "loss": 0.6336,
+      "step": 1549
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98597375340536e-05,
+      "loss": 0.6816,
+      "step": 1550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9859555743591443e-05,
+      "loss": 0.6598,
+      "step": 1551
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985937383623178e-05,
+      "loss": 0.6152,
+      "step": 1552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985919181197677e-05,
+      "loss": 0.654,
+      "step": 1553
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985900967082857e-05,
+      "loss": 0.6337,
+      "step": 1554
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985882741278934e-05,
+      "loss": 0.6617,
+      "step": 1555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9858645037861237e-05,
+      "loss": 0.6571,
+      "step": 1556
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9858462546046426e-05,
+      "loss": 0.6223,
+      "step": 1557
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985827993734707e-05,
+      "loss": 0.628,
+      "step": 1558
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9858097211765336e-05,
+      "loss": 0.64,
+      "step": 1559
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985791436930339e-05,
+      "loss": 0.6326,
+      "step": 1560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9857731409963397e-05,
+      "loss": 0.6731,
+      "step": 1561
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985754833374753e-05,
+      "loss": 0.6663,
+      "step": 1562
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9857365140657954e-05,
+      "loss": 0.6838,
+      "step": 1563
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9857181830696848e-05,
+      "loss": 0.6229,
+      "step": 1564
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9856998403866378e-05,
+      "loss": 0.6276,
+      "step": 1565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985681486016873e-05,
+      "loss": 0.6309,
+      "step": 1566
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9856631199606066e-05,
+      "loss": 0.6796,
+      "step": 1567
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985644742218057e-05,
+      "loss": 0.6719,
+      "step": 1568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9856263527894426e-05,
+      "loss": 0.693,
+      "step": 1569
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9856079516749807e-05,
+      "loss": 0.6737,
+      "step": 1570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98558953887489e-05,
+      "loss": 0.6907,
+      "step": 1571
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985571114389388e-05,
+      "loss": 0.6599,
+      "step": 1572
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9855526782186942e-05,
+      "loss": 0.6608,
+      "step": 1573
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9855342303630264e-05,
+      "loss": 0.628,
+      "step": 1574
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9855157708226035e-05,
+      "loss": 0.6486,
+      "step": 1575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9854972995976446e-05,
+      "loss": 0.6874,
+      "step": 1576
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9854788166883687e-05,
+      "loss": 0.7012,
+      "step": 1577
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9854603220949946e-05,
+      "loss": 0.6682,
+      "step": 1578
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985441815817742e-05,
+      "loss": 0.6766,
+      "step": 1579
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.98542329785683e-05,
+      "loss": 0.6982,
+      "step": 1580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9854047682124782e-05,
+      "loss": 0.6698,
+      "step": 1581
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9853862268849063e-05,
+      "loss": 0.6369,
+      "step": 1582
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985367673874334e-05,
+      "loss": 0.6122,
+      "step": 1583
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9853491091809814e-05,
+      "loss": 0.6798,
+      "step": 1584
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.985330532805069e-05,
+      "loss": 0.6697,
+      "step": 1585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9853119447468167e-05,
+      "loss": 0.6557,
+      "step": 1586
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9852933450064446e-05,
+      "loss": 0.6646,
+      "step": 1587
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9852747335841736e-05,
+      "loss": 0.7025,
+      "step": 1588
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9852561104802238e-05,
+      "loss": 0.6149,
+      "step": 1589
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9852374756948173e-05,
+      "loss": 0.6556,
+      "step": 1590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9852188292281736e-05,
+      "loss": 0.6084,
+      "step": 1591
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9852001710805145e-05,
+      "loss": 0.6217,
+      "step": 1592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9851815012520614e-05,
+      "loss": 0.6503,
+      "step": 1593
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9851628197430347e-05,
+      "loss": 0.6302,
+      "step": 1594
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9851441265536572e-05,
+      "loss": 0.674,
+      "step": 1595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9851254216841493e-05,
+      "loss": 0.6594,
+      "step": 1596
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9851067051347337e-05,
+      "loss": 0.6454,
+      "step": 1597
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9850879769056316e-05,
+      "loss": 0.682,
+      "step": 1598
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.985069236997066e-05,
+      "loss": 0.6718,
+      "step": 1599
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.985050485409258e-05,
+      "loss": 0.7154,
+      "step": 1600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9850317221424308e-05,
+      "loss": 0.6944,
+      "step": 1601
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9850129471968057e-05,
+      "loss": 0.6619,
+      "step": 1602
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9849941605726068e-05,
+      "loss": 0.6105,
+      "step": 1603
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984975362270056e-05,
+      "loss": 0.6755,
+      "step": 1604
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984956552289376e-05,
+      "loss": 0.6717,
+      "step": 1605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.98493773063079e-05,
+      "loss": 0.6058,
+      "step": 1606
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9849188972945215e-05,
+      "loss": 0.6064,
+      "step": 1607
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9849000522807938e-05,
+      "loss": 0.6756,
+      "step": 1608
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9848811955898296e-05,
+      "loss": 0.6622,
+      "step": 1609
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984862327221853e-05,
+      "loss": 0.6317,
+      "step": 1610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9848434471770878e-05,
+      "loss": 0.6827,
+      "step": 1611
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9848245554557577e-05,
+      "loss": 0.619,
+      "step": 1612
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9848056520580866e-05,
+      "loss": 0.651,
+      "step": 1613
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9847867369842987e-05,
+      "loss": 0.622,
+      "step": 1614
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9847678102346188e-05,
+      "loss": 0.7039,
+      "step": 1615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.98474887180927e-05,
+      "loss": 0.6355,
+      "step": 1616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984729921708478e-05,
+      "loss": 0.6322,
+      "step": 1617
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9847109599324674e-05,
+      "loss": 0.677,
+      "step": 1618
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984691986481462e-05,
+      "loss": 0.6476,
+      "step": 1619
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9846730013556877e-05,
+      "loss": 0.6263,
+      "step": 1620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9846540045553696e-05,
+      "loss": 0.6533,
+      "step": 1621
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9846349960807325e-05,
+      "loss": 0.6446,
+      "step": 1622
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9846159759320017e-05,
+      "loss": 0.6793,
+      "step": 1623
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9845969441094035e-05,
+      "loss": 0.7626,
+      "step": 1624
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9845779006131626e-05,
+      "loss": 0.6473,
+      "step": 1625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9845588454435056e-05,
+      "loss": 0.6648,
+      "step": 1626
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9845397786006578e-05,
+      "loss": 0.6301,
+      "step": 1627
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9845207000848453e-05,
+      "loss": 0.7015,
+      "step": 1628
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9845016098962943e-05,
+      "loss": 0.5658,
+      "step": 1629
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9844825080352315e-05,
+      "loss": 0.5699,
+      "step": 1630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9844633945018837e-05,
+      "loss": 0.5991,
+      "step": 1631
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9844442692964763e-05,
+      "loss": 0.6714,
+      "step": 1632
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9844251324192376e-05,
+      "loss": 0.6434,
+      "step": 1633
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9844059838703926e-05,
+      "loss": 0.6246,
+      "step": 1634
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.98438682365017e-05,
+      "loss": 0.6685,
+      "step": 1635
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984367651758796e-05,
+      "loss": 0.6328,
+      "step": 1636
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984348468196499e-05,
+      "loss": 0.6532,
+      "step": 1637
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984329272963505e-05,
+      "loss": 0.6377,
+      "step": 1638
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9843100660600426e-05,
+      "loss": 0.6448,
+      "step": 1639
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984290847486339e-05,
+      "loss": 0.6868,
+      "step": 1640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984271617242622e-05,
+      "loss": 0.6848,
+      "step": 1641
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9842523753291206e-05,
+      "loss": 0.616,
+      "step": 1642
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9842331217460618e-05,
+      "loss": 0.6908,
+      "step": 1643
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9842138564936744e-05,
+      "loss": 0.6943,
+      "step": 1644
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9841945795721863e-05,
+      "loss": 0.6651,
+      "step": 1645
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984175290981827e-05,
+      "loss": 0.7029,
+      "step": 1646
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9841559907228243e-05,
+      "loss": 0.6716,
+      "step": 1647
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9841366787954076e-05,
+      "loss": 0.6629,
+      "step": 1648
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9841173551998055e-05,
+      "loss": 0.6408,
+      "step": 1649
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9840980199362473e-05,
+      "loss": 0.6243,
+      "step": 1650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984078673004962e-05,
+      "loss": 0.6464,
+      "step": 1651
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9840593144061794e-05,
+      "loss": 0.657,
+      "step": 1652
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9840399441401284e-05,
+      "loss": 0.6517,
+      "step": 1653
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9840205622070396e-05,
+      "loss": 0.6366,
+      "step": 1654
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.984001168607142e-05,
+      "loss": 0.6209,
+      "step": 1655
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9839817633406656e-05,
+      "loss": 0.6138,
+      "step": 1656
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9839623464078407e-05,
+      "loss": 0.6765,
+      "step": 1657
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9839429178088976e-05,
+      "loss": 0.6673,
+      "step": 1658
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9839234775440667e-05,
+      "loss": 0.6125,
+      "step": 1659
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983904025613578e-05,
+      "loss": 0.6958,
+      "step": 1660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9838845620176623e-05,
+      "loss": 0.6555,
+      "step": 1661
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9838650867565507e-05,
+      "loss": 0.6502,
+      "step": 1662
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983845599830474e-05,
+      "loss": 0.6443,
+      "step": 1663
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983826101239663e-05,
+      "loss": 0.6601,
+      "step": 1664
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9838065909843493e-05,
+      "loss": 0.6819,
+      "step": 1665
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9837870690647633e-05,
+      "loss": 0.7038,
+      "step": 1666
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9837675354811378e-05,
+      "loss": 0.6084,
+      "step": 1667
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9837479902337033e-05,
+      "loss": 0.5908,
+      "step": 1668
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983728433322692e-05,
+      "loss": 0.6706,
+      "step": 1669
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983708864748336e-05,
+      "loss": 0.7215,
+      "step": 1670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9836892845108666e-05,
+      "loss": 0.6364,
+      "step": 1671
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9836696926105162e-05,
+      "loss": 0.6727,
+      "step": 1672
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9836500890475176e-05,
+      "loss": 0.6614,
+      "step": 1673
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9836304738221027e-05,
+      "loss": 0.6358,
+      "step": 1674
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9836108469345045e-05,
+      "loss": 0.6526,
+      "step": 1675
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9835912083849554e-05,
+      "loss": 0.6642,
+      "step": 1676
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983571558173688e-05,
+      "loss": 0.6378,
+      "step": 1677
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983551896300936e-05,
+      "loss": 0.6549,
+      "step": 1678
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9835322227669317e-05,
+      "loss": 0.6591,
+      "step": 1679
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9835125375719086e-05,
+      "loss": 0.6439,
+      "step": 1680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9834928407161005e-05,
+      "loss": 0.6908,
+      "step": 1681
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9834731321997406e-05,
+      "loss": 0.6865,
+      "step": 1682
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9834534120230626e-05,
+      "loss": 0.6352,
+      "step": 1683
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9834336801863004e-05,
+      "loss": 0.6781,
+      "step": 1684
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983413936689688e-05,
+      "loss": 0.6983,
+      "step": 1685
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9833941815334588e-05,
+      "loss": 0.6352,
+      "step": 1686
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983374414717848e-05,
+      "loss": 0.7022,
+      "step": 1687
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9833546362430892e-05,
+      "loss": 0.6684,
+      "step": 1688
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983334846109418e-05,
+      "loss": 0.6555,
+      "step": 1689
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9833150443170675e-05,
+      "loss": 0.6848,
+      "step": 1690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9832952308662733e-05,
+      "loss": 0.5992,
+      "step": 1691
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9832754057572706e-05,
+      "loss": 0.6612,
+      "step": 1692
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9832555689902942e-05,
+      "loss": 0.6513,
+      "step": 1693
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983235720565579e-05,
+      "loss": 0.6094,
+      "step": 1694
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9832158604833605e-05,
+      "loss": 0.5358,
+      "step": 1695
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9831959887438743e-05,
+      "loss": 0.6332,
+      "step": 1696
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9831761053473558e-05,
+      "loss": 0.6903,
+      "step": 1697
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9831562102940406e-05,
+      "loss": 0.6224,
+      "step": 1698
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9831363035841654e-05,
+      "loss": 0.708,
+      "step": 1699
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.983116385217965e-05,
+      "loss": 0.6803,
+      "step": 1700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9830964551956767e-05,
+      "loss": 0.6729,
+      "step": 1701
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9830765135175363e-05,
+      "loss": 0.6696,
+      "step": 1702
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.98305656018378e-05,
+      "loss": 0.6447,
+      "step": 1703
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9830365951946448e-05,
+      "loss": 0.7428,
+      "step": 1704
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9830166185503674e-05,
+      "loss": 0.6939,
+      "step": 1705
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982996630251184e-05,
+      "loss": 0.6875,
+      "step": 1706
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9829766302973322e-05,
+      "loss": 0.6409,
+      "step": 1707
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982956618689049e-05,
+      "loss": 0.6107,
+      "step": 1708
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9829365954265717e-05,
+      "loss": 0.6183,
+      "step": 1709
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9829165605101375e-05,
+      "loss": 0.6833,
+      "step": 1710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982896513939984e-05,
+      "loss": 0.6267,
+      "step": 1711
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9828764557163494e-05,
+      "loss": 0.6329,
+      "step": 1712
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9828563858394706e-05,
+      "loss": 0.6547,
+      "step": 1713
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9828363043095864e-05,
+      "loss": 0.5749,
+      "step": 1714
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9828162111269345e-05,
+      "loss": 0.6522,
+      "step": 1715
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982796106291753e-05,
+      "loss": 0.604,
+      "step": 1716
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.98277598980428e-05,
+      "loss": 0.6603,
+      "step": 1717
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982755861664755e-05,
+      "loss": 0.6165,
+      "step": 1718
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9827357218734158e-05,
+      "loss": 0.6548,
+      "step": 1719
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9827155704305018e-05,
+      "loss": 0.6317,
+      "step": 1720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9826954073362512e-05,
+      "loss": 0.6283,
+      "step": 1721
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9826752325909035e-05,
+      "loss": 0.6713,
+      "step": 1722
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982655046194698e-05,
+      "loss": 0.6847,
+      "step": 1723
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9826348481478736e-05,
+      "loss": 0.6297,
+      "step": 1724
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9826146384506702e-05,
+      "loss": 0.6341,
+      "step": 1725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9825944171033272e-05,
+      "loss": 0.6742,
+      "step": 1726
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9825741841060844e-05,
+      "loss": 0.6974,
+      "step": 1727
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9825539394591818e-05,
+      "loss": 0.6195,
+      "step": 1728
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9825336831628593e-05,
+      "loss": 0.624,
+      "step": 1729
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982513415217357e-05,
+      "loss": 0.6598,
+      "step": 1730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982493135622915e-05,
+      "loss": 0.6939,
+      "step": 1731
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9824728443797746e-05,
+      "loss": 0.6733,
+      "step": 1732
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9824525414881753e-05,
+      "loss": 0.6946,
+      "step": 1733
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9824322269483586e-05,
+      "loss": 0.6246,
+      "step": 1734
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982411900760565e-05,
+      "loss": 0.672,
+      "step": 1735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9823915629250354e-05,
+      "loss": 0.6647,
+      "step": 1736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982371213442011e-05,
+      "loss": 0.655,
+      "step": 1737
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9823508523117336e-05,
+      "loss": 0.6474,
+      "step": 1738
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982330479534444e-05,
+      "loss": 0.6522,
+      "step": 1739
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9823100951103837e-05,
+      "loss": 0.5937,
+      "step": 1740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9822896990397948e-05,
+      "loss": 0.6453,
+      "step": 1741
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982269291322919e-05,
+      "loss": 0.6907,
+      "step": 1742
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982248871959998e-05,
+      "loss": 0.6383,
+      "step": 1743
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982228440951274e-05,
+      "loss": 0.6409,
+      "step": 1744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9822079982969894e-05,
+      "loss": 0.6679,
+      "step": 1745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9821875439973865e-05,
+      "loss": 0.6567,
+      "step": 1746
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9821670780527076e-05,
+      "loss": 0.6972,
+      "step": 1747
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982146600463196e-05,
+      "loss": 0.682,
+      "step": 1748
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9821261112290937e-05,
+      "loss": 0.6318,
+      "step": 1749
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982105610350644e-05,
+      "loss": 0.6345,
+      "step": 1750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9820850978280897e-05,
+      "loss": 0.6752,
+      "step": 1751
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9820645736616745e-05,
+      "loss": 0.6217,
+      "step": 1752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9820440378516415e-05,
+      "loss": 0.621,
+      "step": 1753
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.982023490398234e-05,
+      "loss": 0.6486,
+      "step": 1754
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9820029313016956e-05,
+      "loss": 0.6138,
+      "step": 1755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9819823605622705e-05,
+      "loss": 0.6629,
+      "step": 1756
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9819617781802024e-05,
+      "loss": 0.6557,
+      "step": 1757
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981941184155735e-05,
+      "loss": 0.6627,
+      "step": 1758
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9819205784891126e-05,
+      "loss": 0.6343,
+      "step": 1759
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9818999611805798e-05,
+      "loss": 0.6815,
+      "step": 1760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9818793322303807e-05,
+      "loss": 0.6758,
+      "step": 1761
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.98185869163876e-05,
+      "loss": 0.6688,
+      "step": 1762
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9818380394059627e-05,
+      "loss": 0.6287,
+      "step": 1763
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9818173755322334e-05,
+      "loss": 0.6486,
+      "step": 1764
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981796700017817e-05,
+      "loss": 0.6651,
+      "step": 1765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981776012862959e-05,
+      "loss": 0.6541,
+      "step": 1766
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981755314067904e-05,
+      "loss": 0.6983,
+      "step": 1767
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981734603632898e-05,
+      "loss": 0.691,
+      "step": 1768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9817138815581863e-05,
+      "loss": 0.7108,
+      "step": 1769
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9816931478440148e-05,
+      "loss": 0.6643,
+      "step": 1770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981672402490629e-05,
+      "loss": 0.6701,
+      "step": 1771
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981651645498275e-05,
+      "loss": 0.6634,
+      "step": 1772
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9816308768671992e-05,
+      "loss": 0.6075,
+      "step": 1773
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9816100965976473e-05,
+      "loss": 0.6565,
+      "step": 1774
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9815893046898665e-05,
+      "loss": 0.6068,
+      "step": 1775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9815685011441023e-05,
+      "loss": 0.636,
+      "step": 1776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9815476859606018e-05,
+      "loss": 0.6016,
+      "step": 1777
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9815268591396123e-05,
+      "loss": 0.6395,
+      "step": 1778
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9815060206813797e-05,
+      "loss": 0.6853,
+      "step": 1779
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981485170586152e-05,
+      "loss": 0.6496,
+      "step": 1780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9814643088541756e-05,
+      "loss": 0.6125,
+      "step": 1781
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9814434354856983e-05,
+      "loss": 0.698,
+      "step": 1782
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9814225504809677e-05,
+      "loss": 0.7012,
+      "step": 1783
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9814016538402314e-05,
+      "loss": 0.6877,
+      "step": 1784
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981380745563737e-05,
+      "loss": 0.6483,
+      "step": 1785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9813598256517322e-05,
+      "loss": 0.692,
+      "step": 1786
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9813388941044656e-05,
+      "loss": 0.6808,
+      "step": 1787
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9813179509221847e-05,
+      "loss": 0.6846,
+      "step": 1788
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981296996105138e-05,
+      "loss": 0.669,
+      "step": 1789
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9812760296535744e-05,
+      "loss": 0.6212,
+      "step": 1790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981255051567742e-05,
+      "loss": 0.6957,
+      "step": 1791
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9812340618478897e-05,
+      "loss": 0.6847,
+      "step": 1792
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9812130604942662e-05,
+      "loss": 0.6481,
+      "step": 1793
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9811920475071206e-05,
+      "loss": 0.6662,
+      "step": 1794
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981171022886702e-05,
+      "loss": 0.6987,
+      "step": 1795
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9811499866332605e-05,
+      "loss": 0.6446,
+      "step": 1796
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981128938747044e-05,
+      "loss": 0.6416,
+      "step": 1797
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981107879228303e-05,
+      "loss": 0.627,
+      "step": 1798
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9810868080772868e-05,
+      "loss": 0.6362,
+      "step": 1799
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9810657252942458e-05,
+      "loss": 0.6943,
+      "step": 1800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9810446308794292e-05,
+      "loss": 0.6134,
+      "step": 1801
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9810235248330874e-05,
+      "loss": 0.6157,
+      "step": 1802
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.981002407155471e-05,
+      "loss": 0.6391,
+      "step": 1803
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.98098127784683e-05,
+      "loss": 0.6102,
+      "step": 1804
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9809601369074153e-05,
+      "loss": 0.6443,
+      "step": 1805
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9809389843374768e-05,
+      "loss": 0.669,
+      "step": 1806
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980917820137266e-05,
+      "loss": 0.673,
+      "step": 1807
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9808966443070334e-05,
+      "loss": 0.6509,
+      "step": 1808
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9808754568470303e-05,
+      "loss": 0.6335,
+      "step": 1809
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9808542577575077e-05,
+      "loss": 0.5754,
+      "step": 1810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9808330470387177e-05,
+      "loss": 0.6934,
+      "step": 1811
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9808118246909105e-05,
+      "loss": 0.6484,
+      "step": 1812
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9807905907143387e-05,
+      "loss": 0.6079,
+      "step": 1813
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9807693451092537e-05,
+      "loss": 0.6721,
+      "step": 1814
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9807480878759073e-05,
+      "loss": 0.6284,
+      "step": 1815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9807268190145515e-05,
+      "loss": 0.6307,
+      "step": 1816
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980705538525439e-05,
+      "loss": 0.656,
+      "step": 1817
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9806842464088216e-05,
+      "loss": 0.63,
+      "step": 1818
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980662942664952e-05,
+      "loss": 0.6083,
+      "step": 1819
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9806416272940824e-05,
+      "loss": 0.6887,
+      "step": 1820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980620300296466e-05,
+      "loss": 0.6545,
+      "step": 1821
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9805989616723557e-05,
+      "loss": 0.7075,
+      "step": 1822
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980577611422004e-05,
+      "loss": 0.7079,
+      "step": 1823
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9805562495456644e-05,
+      "loss": 0.6022,
+      "step": 1824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9805348760435896e-05,
+      "loss": 0.608,
+      "step": 1825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9805134909160338e-05,
+      "loss": 0.657,
+      "step": 1826
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9804920941632503e-05,
+      "loss": 0.6542,
+      "step": 1827
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9804706857854925e-05,
+      "loss": 0.6589,
+      "step": 1828
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9804492657830145e-05,
+      "loss": 0.6494,
+      "step": 1829
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9804278341560702e-05,
+      "loss": 0.6328,
+      "step": 1830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9804063909049133e-05,
+      "loss": 0.6306,
+      "step": 1831
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9803849360297992e-05,
+      "loss": 0.6331,
+      "step": 1832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9803634695309808e-05,
+      "loss": 0.6643,
+      "step": 1833
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9803419914087138e-05,
+      "loss": 0.6672,
+      "step": 1834
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980320501663252e-05,
+      "loss": 0.6517,
+      "step": 1835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9802990002948504e-05,
+      "loss": 0.6593,
+      "step": 1836
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9802774873037643e-05,
+      "loss": 0.6169,
+      "step": 1837
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9802559626902488e-05,
+      "loss": 0.6247,
+      "step": 1838
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9802344264545583e-05,
+      "loss": 0.6639,
+      "step": 1839
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980212878596949e-05,
+      "loss": 0.6209,
+      "step": 1840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980191319117676e-05,
+      "loss": 0.6136,
+      "step": 1841
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980169748016995e-05,
+      "loss": 0.5871,
+      "step": 1842
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980148165295162e-05,
+      "loss": 0.592,
+      "step": 1843
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980126570952432e-05,
+      "loss": 0.5993,
+      "step": 1844
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.980104964989062e-05,
+      "loss": 0.6877,
+      "step": 1845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9800833474053077e-05,
+      "loss": 0.6383,
+      "step": 1846
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9800617182014256e-05,
+      "loss": 0.5955,
+      "step": 1847
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9800400773776718e-05,
+      "loss": 0.649,
+      "step": 1848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9800184249343034e-05,
+      "loss": 0.6669,
+      "step": 1849
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9799967608715765e-05,
+      "loss": 0.6896,
+      "step": 1850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9799750851897488e-05,
+      "loss": 0.6312,
+      "step": 1851
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9799533978890762e-05,
+      "loss": 0.6943,
+      "step": 1852
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.979931698969817e-05,
+      "loss": 0.6839,
+      "step": 1853
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9799099884322274e-05,
+      "loss": 0.6102,
+      "step": 1854
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9798882662765658e-05,
+      "loss": 0.6801,
+      "step": 1855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9798665325030887e-05,
+      "loss": 0.6297,
+      "step": 1856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9798447871120547e-05,
+      "loss": 0.5974,
+      "step": 1857
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9798230301037212e-05,
+      "loss": 0.6171,
+      "step": 1858
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.979801261478346e-05,
+      "loss": 0.6694,
+      "step": 1859
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9797794812361875e-05,
+      "loss": 0.5956,
+      "step": 1860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.979757689377504e-05,
+      "loss": 0.639,
+      "step": 1861
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9797358859025534e-05,
+      "loss": 0.6525,
+      "step": 1862
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9797140708115946e-05,
+      "loss": 0.6563,
+      "step": 1863
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9796922441048863e-05,
+      "loss": 0.6101,
+      "step": 1864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.979670405782687e-05,
+      "loss": 0.6311,
+      "step": 1865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9796485558452557e-05,
+      "loss": 0.6436,
+      "step": 1866
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9796266942928514e-05,
+      "loss": 0.6838,
+      "step": 1867
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9796048211257335e-05,
+      "loss": 0.5818,
+      "step": 1868
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9795829363441614e-05,
+      "loss": 0.6486,
+      "step": 1869
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9795610399483945e-05,
+      "loss": 0.653,
+      "step": 1870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.979539131938692e-05,
+      "loss": 0.6771,
+      "step": 1871
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9795172123153142e-05,
+      "loss": 0.6628,
+      "step": 1872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9794952810785207e-05,
+      "loss": 0.6289,
+      "step": 1873
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9794733382285714e-05,
+      "loss": 0.639,
+      "step": 1874
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.979451383765727e-05,
+      "loss": 0.692,
+      "step": 1875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.979429417690247e-05,
+      "loss": 0.6382,
+      "step": 1876
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9794074400023923e-05,
+      "loss": 0.6595,
+      "step": 1877
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9793854507024237e-05,
+      "loss": 0.5874,
+      "step": 1878
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9793634497906017e-05,
+      "loss": 0.6658,
+      "step": 1879
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.979341437267187e-05,
+      "loss": 0.6495,
+      "step": 1880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9793194131324404e-05,
+      "loss": 0.6673,
+      "step": 1881
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.979297377386624e-05,
+      "loss": 0.6384,
+      "step": 1882
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9792753300299976e-05,
+      "loss": 0.727,
+      "step": 1883
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9792532710628237e-05,
+      "loss": 0.584,
+      "step": 1884
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9792312004853634e-05,
+      "loss": 0.6538,
+      "step": 1885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9792091182978787e-05,
+      "loss": 0.6244,
+      "step": 1886
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.979187024500631e-05,
+      "loss": 0.6891,
+      "step": 1887
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9791649190938827e-05,
+      "loss": 0.6249,
+      "step": 1888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9791428020778953e-05,
+      "loss": 0.6699,
+      "step": 1889
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9791206734529317e-05,
+      "loss": 0.6565,
+      "step": 1890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9790985332192535e-05,
+      "loss": 0.6404,
+      "step": 1891
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.979076381377124e-05,
+      "loss": 0.7028,
+      "step": 1892
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9790542179268053e-05,
+      "loss": 0.5966,
+      "step": 1893
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9790320428685605e-05,
+      "loss": 0.6158,
+      "step": 1894
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.979009856202652e-05,
+      "loss": 0.6724,
+      "step": 1895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9789876579293435e-05,
+      "loss": 0.6689,
+      "step": 1896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978965448048898e-05,
+      "loss": 0.5927,
+      "step": 1897
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9789432265615785e-05,
+      "loss": 0.5469,
+      "step": 1898
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9789209934676488e-05,
+      "loss": 0.6034,
+      "step": 1899
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9788987487673725e-05,
+      "loss": 0.6556,
+      "step": 1900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9788764924610128e-05,
+      "loss": 0.6426,
+      "step": 1901
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978854224548834e-05,
+      "loss": 0.6258,
+      "step": 1902
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9788319450311006e-05,
+      "loss": 0.6146,
+      "step": 1903
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978809653908076e-05,
+      "loss": 0.6152,
+      "step": 1904
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9787873511800245e-05,
+      "loss": 0.6898,
+      "step": 1905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978765036847211e-05,
+      "loss": 0.6333,
+      "step": 1906
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9787427109099e-05,
+      "loss": 0.64,
+      "step": 1907
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9787203733683558e-05,
+      "loss": 0.6443,
+      "step": 1908
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9786980242228435e-05,
+      "loss": 0.6558,
+      "step": 1909
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978675663473628e-05,
+      "loss": 0.6615,
+      "step": 1910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978653291120975e-05,
+      "loss": 0.6221,
+      "step": 1911
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9786309071651488e-05,
+      "loss": 0.62,
+      "step": 1912
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978608511606415e-05,
+      "loss": 0.6715,
+      "step": 1913
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9785861044450398e-05,
+      "loss": 0.6753,
+      "step": 1914
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9785636856812882e-05,
+      "loss": 0.6459,
+      "step": 1915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9785412553154265e-05,
+      "loss": 0.6416,
+      "step": 1916
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.97851881334772e-05,
+      "loss": 0.7021,
+      "step": 1917
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978496359778435e-05,
+      "loss": 0.6721,
+      "step": 1918
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9784738946078385e-05,
+      "loss": 0.6911,
+      "step": 1919
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978451417836196e-05,
+      "loss": 0.6565,
+      "step": 1920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9784289294637737e-05,
+      "loss": 0.6327,
+      "step": 1921
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978406429490839e-05,
+      "loss": 0.6332,
+      "step": 1922
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9783839179176584e-05,
+      "loss": 0.6335,
+      "step": 1923
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978361394744499e-05,
+      "loss": 0.6471,
+      "step": 1924
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9783388599716272e-05,
+      "loss": 0.6429,
+      "step": 1925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9783163135993108e-05,
+      "loss": 0.6452,
+      "step": 1926
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9782937556278173e-05,
+      "loss": 0.6815,
+      "step": 1927
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978271186057413e-05,
+      "loss": 0.6608,
+      "step": 1928
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9782486048883668e-05,
+      "loss": 0.6352,
+      "step": 1929
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.978226012120946e-05,
+      "loss": 0.5929,
+      "step": 1930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9782034077554182e-05,
+      "loss": 0.6814,
+      "step": 1931
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9781807917920515e-05,
+      "loss": 0.6718,
+      "step": 1932
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9781581642311144e-05,
+      "loss": 0.67,
+      "step": 1933
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9781355250728746e-05,
+      "loss": 0.6711,
+      "step": 1934
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9781128743176004e-05,
+      "loss": 0.621,
+      "step": 1935
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9780902119655616e-05,
+      "loss": 0.6603,
+      "step": 1936
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9780675380170257e-05,
+      "loss": 0.6392,
+      "step": 1937
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9780448524722615e-05,
+      "loss": 0.7097,
+      "step": 1938
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9780221553315387e-05,
+      "loss": 0.6351,
+      "step": 1939
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977999446595126e-05,
+      "loss": 0.6301,
+      "step": 1940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9779767262632926e-05,
+      "loss": 0.6342,
+      "step": 1941
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9779539943363083e-05,
+      "loss": 0.6734,
+      "step": 1942
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977931250814442e-05,
+      "loss": 0.6906,
+      "step": 1943
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9779084956979634e-05,
+      "loss": 0.6511,
+      "step": 1944
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977885728987143e-05,
+      "loss": 0.5811,
+      "step": 1945
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.97786295068225e-05,
+      "loss": 0.598,
+      "step": 1946
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977840160783555e-05,
+      "loss": 0.6097,
+      "step": 1947
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9778173592913274e-05,
+      "loss": 0.6418,
+      "step": 1948
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9777945462058383e-05,
+      "loss": 0.6943,
+      "step": 1949
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977771721527358e-05,
+      "loss": 0.6618,
+      "step": 1950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977748885256157e-05,
+      "loss": 0.606,
+      "step": 1951
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977726037392506e-05,
+      "loss": 0.5941,
+      "step": 1952
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9777031779366763e-05,
+      "loss": 0.6305,
+      "step": 1953
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9776803068889388e-05,
+      "loss": 0.6461,
+      "step": 1954
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9776574242495638e-05,
+      "loss": 0.6318,
+      "step": 1955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9776345300188238e-05,
+      "loss": 0.6588,
+      "step": 1956
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9776116241969895e-05,
+      "loss": 0.642,
+      "step": 1957
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977588706784333e-05,
+      "loss": 0.6645,
+      "step": 1958
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9775657777811253e-05,
+      "loss": 0.6195,
+      "step": 1959
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9775428371876392e-05,
+      "loss": 0.6399,
+      "step": 1960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9775198850041457e-05,
+      "loss": 0.6571,
+      "step": 1961
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9774969212309176e-05,
+      "loss": 0.7086,
+      "step": 1962
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977473945868227e-05,
+      "loss": 0.681,
+      "step": 1963
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977450958916346e-05,
+      "loss": 0.6816,
+      "step": 1964
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9774279603755477e-05,
+      "loss": 0.6755,
+      "step": 1965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977404950246104e-05,
+      "loss": 0.6826,
+      "step": 1966
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9773819285282888e-05,
+      "loss": 0.6688,
+      "step": 1967
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977358895222374e-05,
+      "loss": 0.6819,
+      "step": 1968
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9773358503286335e-05,
+      "loss": 0.6577,
+      "step": 1969
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.97731279384734e-05,
+      "loss": 0.5945,
+      "step": 1970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9772897257787668e-05,
+      "loss": 0.6507,
+      "step": 1971
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9772666461231878e-05,
+      "loss": 0.6488,
+      "step": 1972
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9772435548808767e-05,
+      "loss": 0.6168,
+      "step": 1973
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9772204520521066e-05,
+      "loss": 0.665,
+      "step": 1974
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977197337637152e-05,
+      "loss": 0.6329,
+      "step": 1975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977174211636287e-05,
+      "loss": 0.6436,
+      "step": 1976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9771510740497855e-05,
+      "loss": 0.6117,
+      "step": 1977
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.977127924877922e-05,
+      "loss": 0.6598,
+      "step": 1978
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9771047641209704e-05,
+      "loss": 0.6906,
+      "step": 1979
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9770815917792062e-05,
+      "loss": 0.646,
+      "step": 1980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9770584078529037e-05,
+      "loss": 0.6403,
+      "step": 1981
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9770352123423378e-05,
+      "loss": 0.6413,
+      "step": 1982
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9770120052477835e-05,
+      "loss": 0.6187,
+      "step": 1983
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976988786569516e-05,
+      "loss": 0.6322,
+      "step": 1984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.97696555630781e-05,
+      "loss": 0.6453,
+      "step": 1985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976942314462942e-05,
+      "loss": 0.6194,
+      "step": 1986
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976919061035187e-05,
+      "loss": 0.6643,
+      "step": 1987
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9768957960248203e-05,
+      "loss": 0.6507,
+      "step": 1988
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9768725194321187e-05,
+      "loss": 0.6055,
+      "step": 1989
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9768492312573576e-05,
+      "loss": 0.7409,
+      "step": 1990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976825931500813e-05,
+      "loss": 0.6888,
+      "step": 1991
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976802620162761e-05,
+      "loss": 0.6312,
+      "step": 1992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9767792972434785e-05,
+      "loss": 0.6059,
+      "step": 1993
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976755962743242e-05,
+      "loss": 0.6474,
+      "step": 1994
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9767326166623277e-05,
+      "loss": 0.6474,
+      "step": 1995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9767092590010128e-05,
+      "loss": 0.655,
+      "step": 1996
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976685889759574e-05,
+      "loss": 0.6314,
+      "step": 1997
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9766625089382884e-05,
+      "loss": 0.6017,
+      "step": 1998
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9766391165374333e-05,
+      "loss": 0.6662,
+      "step": 1999
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976615712557286e-05,
+      "loss": 0.6728,
+      "step": 2000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976592296998124e-05,
+      "loss": 0.6309,
+      "step": 2001
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976568869860225e-05,
+      "loss": 0.5995,
+      "step": 2002
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976545431143867e-05,
+      "loss": 0.6118,
+      "step": 2003
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9765219808493274e-05,
+      "loss": 0.6591,
+      "step": 2004
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976498518976884e-05,
+      "loss": 0.6851,
+      "step": 2005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976475045526816e-05,
+      "loss": 0.6285,
+      "step": 2006
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9764515604994006e-05,
+      "loss": 0.6283,
+      "step": 2007
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9764280638949168e-05,
+      "loss": 0.6426,
+      "step": 2008
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976404555713643e-05,
+      "loss": 0.6341,
+      "step": 2009
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9763810359558584e-05,
+      "loss": 0.6295,
+      "step": 2010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9763575046218414e-05,
+      "loss": 0.7087,
+      "step": 2011
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976333961711871e-05,
+      "loss": 0.6454,
+      "step": 2012
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9763104072262266e-05,
+      "loss": 0.6878,
+      "step": 2013
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976286841165187e-05,
+      "loss": 0.6402,
+      "step": 2014
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976263263529032e-05,
+      "loss": 0.6457,
+      "step": 2015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9762396743180407e-05,
+      "loss": 0.6479,
+      "step": 2016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9762160735324937e-05,
+      "loss": 0.671,
+      "step": 2017
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.97619246117267e-05,
+      "loss": 0.6706,
+      "step": 2018
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9761688372388496e-05,
+      "loss": 0.681,
+      "step": 2019
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976145201731313e-05,
+      "loss": 0.6944,
+      "step": 2020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9761215546503402e-05,
+      "loss": 0.6079,
+      "step": 2021
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9760978959962113e-05,
+      "loss": 0.6505,
+      "step": 2022
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9760742257692075e-05,
+      "loss": 0.6477,
+      "step": 2023
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9760505439696086e-05,
+      "loss": 0.5803,
+      "step": 2024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.976026850597696e-05,
+      "loss": 0.6362,
+      "step": 2025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9760031456537506e-05,
+      "loss": 0.6331,
+      "step": 2026
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975979429138053e-05,
+      "loss": 0.7165,
+      "step": 2027
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9759557010508845e-05,
+      "loss": 0.5964,
+      "step": 2028
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9759319613925266e-05,
+      "loss": 0.6509,
+      "step": 2029
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9759082101632613e-05,
+      "loss": 0.6478,
+      "step": 2030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975884447363369e-05,
+      "loss": 0.6993,
+      "step": 2031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9758606729931324e-05,
+      "loss": 0.6475,
+      "step": 2032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975836887052833e-05,
+      "loss": 0.6435,
+      "step": 2033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9758130895427526e-05,
+      "loss": 0.6801,
+      "step": 2034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975789280463174e-05,
+      "loss": 0.6954,
+      "step": 2035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9757654598143788e-05,
+      "loss": 0.6895,
+      "step": 2036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9757416275966497e-05,
+      "loss": 0.7047,
+      "step": 2037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9757177838102697e-05,
+      "loss": 0.6224,
+      "step": 2038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9756939284555207e-05,
+      "loss": 0.6586,
+      "step": 2039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975670061532686e-05,
+      "loss": 0.6825,
+      "step": 2040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9756461830420486e-05,
+      "loss": 0.6305,
+      "step": 2041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9756222929838914e-05,
+      "loss": 0.6808,
+      "step": 2042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975598391358498e-05,
+      "loss": 0.6754,
+      "step": 2043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975574478166151e-05,
+      "loss": 0.676,
+      "step": 2044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975550553407135e-05,
+      "loss": 0.6589,
+      "step": 2045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975526617081733e-05,
+      "loss": 0.643,
+      "step": 2046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9755026691902286e-05,
+      "loss": 0.653,
+      "step": 2047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9754787097329063e-05,
+      "loss": 0.6209,
+      "step": 2048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.97545473871005e-05,
+      "loss": 0.6699,
+      "step": 2049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9754307561219434e-05,
+      "loss": 0.6348,
+      "step": 2050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9754067619688716e-05,
+      "loss": 0.7143,
+      "step": 2051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9753827562511186e-05,
+      "loss": 0.6228,
+      "step": 2052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9753587389689692e-05,
+      "loss": 0.6521,
+      "step": 2053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975334710122708e-05,
+      "loss": 0.6984,
+      "step": 2054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9753106697126204e-05,
+      "loss": 0.7173,
+      "step": 2055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9752866177389908e-05,
+      "loss": 0.6519,
+      "step": 2056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9752625542021045e-05,
+      "loss": 0.6973,
+      "step": 2057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975238479102247e-05,
+      "loss": 0.6537,
+      "step": 2058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9752143924397035e-05,
+      "loss": 0.6217,
+      "step": 2059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9751902942147597e-05,
+      "loss": 0.6881,
+      "step": 2060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9751661844277016e-05,
+      "loss": 0.6414,
+      "step": 2061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9751420630788146e-05,
+      "loss": 0.6493,
+      "step": 2062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9751179301683853e-05,
+      "loss": 0.613,
+      "step": 2063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975093785696699e-05,
+      "loss": 0.5898,
+      "step": 2064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9750696296640424e-05,
+      "loss": 0.662,
+      "step": 2065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975045462070702e-05,
+      "loss": 0.6276,
+      "step": 2066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.975021282916964e-05,
+      "loss": 0.6918,
+      "step": 2067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9749970922031156e-05,
+      "loss": 0.6403,
+      "step": 2068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9749728899294428e-05,
+      "loss": 0.6793,
+      "step": 2069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9749486760962334e-05,
+      "loss": 0.6071,
+      "step": 2070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974924450703774e-05,
+      "loss": 0.662,
+      "step": 2071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9749002137523523e-05,
+      "loss": 0.6306,
+      "step": 2072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9748759652422548e-05,
+      "loss": 0.5944,
+      "step": 2073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.97485170517377e-05,
+      "loss": 0.6209,
+      "step": 2074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9748274335471845e-05,
+      "loss": 0.6691,
+      "step": 2075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974803150362787e-05,
+      "loss": 0.6359,
+      "step": 2076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974778855620865e-05,
+      "loss": 0.7043,
+      "step": 2077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9747545493217063e-05,
+      "loss": 0.6533,
+      "step": 2078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9747302314655998e-05,
+      "loss": 0.6483,
+      "step": 2079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974705902052833e-05,
+      "loss": 0.6081,
+      "step": 2080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9746815610836946e-05,
+      "loss": 0.6496,
+      "step": 2081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9746572085584736e-05,
+      "loss": 0.5861,
+      "step": 2082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9746328444774586e-05,
+      "loss": 0.6522,
+      "step": 2083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974608468840938e-05,
+      "loss": 0.5862,
+      "step": 2084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974584081649201e-05,
+      "loss": 0.6676,
+      "step": 2085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974559682902537e-05,
+      "loss": 0.6554,
+      "step": 2086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9745352726012354e-05,
+      "loss": 0.6628,
+      "step": 2087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9745108507455853e-05,
+      "loss": 0.6908,
+      "step": 2088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974486417335876e-05,
+      "loss": 0.673,
+      "step": 2089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9744619723723977e-05,
+      "loss": 0.6773,
+      "step": 2090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.97443751585544e-05,
+      "loss": 0.597,
+      "step": 2091
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9744130477852924e-05,
+      "loss": 0.5708,
+      "step": 2092
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974388568162246e-05,
+      "loss": 0.6644,
+      "step": 2093
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9743640769865906e-05,
+      "loss": 0.6914,
+      "step": 2094
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9743395742586166e-05,
+      "loss": 0.638,
+      "step": 2095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9743150599786138e-05,
+      "loss": 0.7225,
+      "step": 2096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974290534146874e-05,
+      "loss": 0.634,
+      "step": 2097
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974265996763687e-05,
+      "loss": 0.6732,
+      "step": 2098
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9742414478293445e-05,
+      "loss": 0.6805,
+      "step": 2099
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974216887344137e-05,
+      "loss": 0.6495,
+      "step": 2100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.974192315308356e-05,
+      "loss": 0.6675,
+      "step": 2101
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9741677317222927e-05,
+      "loss": 0.5874,
+      "step": 2102
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9741431365862385e-05,
+      "loss": 0.6805,
+      "step": 2103
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9741185299004852e-05,
+      "loss": 0.6416,
+      "step": 2104
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9740939116653245e-05,
+      "loss": 0.6179,
+      "step": 2105
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9740692818810484e-05,
+      "loss": 0.7175,
+      "step": 2106
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9740446405479487e-05,
+      "loss": 0.6455,
+      "step": 2107
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9740199876663175e-05,
+      "loss": 0.6981,
+      "step": 2108
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9739953232364474e-05,
+      "loss": 0.6375,
+      "step": 2109
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9739706472586305e-05,
+      "loss": 0.6463,
+      "step": 2110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9739459597331593e-05,
+      "loss": 0.6573,
+      "step": 2111
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9739212606603267e-05,
+      "loss": 0.6311,
+      "step": 2112
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9738965500404256e-05,
+      "loss": 0.6602,
+      "step": 2113
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9738718278737493e-05,
+      "loss": 0.6505,
+      "step": 2114
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9738470941605903e-05,
+      "loss": 0.6822,
+      "step": 2115
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9738223489012422e-05,
+      "loss": 0.6629,
+      "step": 2116
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973797592095998e-05,
+      "loss": 0.6268,
+      "step": 2117
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9737728237451513e-05,
+      "loss": 0.7028,
+      "step": 2118
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9737480438489964e-05,
+      "loss": 0.6609,
+      "step": 2119
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9737232524078264e-05,
+      "loss": 0.6327,
+      "step": 2120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9736984494219354e-05,
+      "loss": 0.6604,
+      "step": 2121
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9736736348916177e-05,
+      "loss": 0.6645,
+      "step": 2122
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9736488088171675e-05,
+      "loss": 0.6796,
+      "step": 2123
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9736239711988786e-05,
+      "loss": 0.68,
+      "step": 2124
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9735991220370462e-05,
+      "loss": 0.6625,
+      "step": 2125
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9735742613319643e-05,
+      "loss": 0.679,
+      "step": 2126
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9735493890839284e-05,
+      "loss": 0.6868,
+      "step": 2127
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9735245052932326e-05,
+      "loss": 0.648,
+      "step": 2128
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9734996099601726e-05,
+      "loss": 0.6652,
+      "step": 2129
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973474703085043e-05,
+      "loss": 0.6854,
+      "step": 2130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9734497846681392e-05,
+      "loss": 0.6477,
+      "step": 2131
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973424854709757e-05,
+      "loss": 0.6479,
+      "step": 2132
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973399913210192e-05,
+      "loss": 0.6387,
+      "step": 2133
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973374960169739e-05,
+      "loss": 0.5708,
+      "step": 2134
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973349995588695e-05,
+      "loss": 0.6536,
+      "step": 2135
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9733250194673555e-05,
+      "loss": 0.6217,
+      "step": 2136
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9733000318060166e-05,
+      "loss": 0.6324,
+      "step": 2137
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9732750326049746e-05,
+      "loss": 0.6407,
+      "step": 2138
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9732500218645256e-05,
+      "loss": 0.6235,
+      "step": 2139
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9732249995849665e-05,
+      "loss": 0.6167,
+      "step": 2140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9731999657665942e-05,
+      "loss": 0.6248,
+      "step": 2141
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973174920409705e-05,
+      "loss": 0.596,
+      "step": 2142
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973149863514596e-05,
+      "loss": 0.5955,
+      "step": 2143
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9731247950815646e-05,
+      "loss": 0.6754,
+      "step": 2144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9730997151109075e-05,
+      "loss": 0.5907,
+      "step": 2145
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9730746236029224e-05,
+      "loss": 0.6457,
+      "step": 2146
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9730495205579067e-05,
+      "loss": 0.6364,
+      "step": 2147
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.973024405976158e-05,
+      "loss": 0.6198,
+      "step": 2148
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972999279857974e-05,
+      "loss": 0.646,
+      "step": 2149
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9729741422036528e-05,
+      "loss": 0.6881,
+      "step": 2150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972948993013492e-05,
+      "loss": 0.6325,
+      "step": 2151
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9729238322877905e-05,
+      "loss": 0.6259,
+      "step": 2152
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972898660026846e-05,
+      "loss": 0.6526,
+      "step": 2153
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9728734762309574e-05,
+      "loss": 0.6604,
+      "step": 2154
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972848280900423e-05,
+      "loss": 0.6643,
+      "step": 2155
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9728230740355413e-05,
+      "loss": 0.6364,
+      "step": 2156
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9727978556366115e-05,
+      "loss": 0.607,
+      "step": 2157
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972772625703933e-05,
+      "loss": 0.6759,
+      "step": 2158
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9727473842378037e-05,
+      "loss": 0.6367,
+      "step": 2159
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972722131238524e-05,
+      "loss": 0.5954,
+      "step": 2160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972696866706393e-05,
+      "loss": 0.6104,
+      "step": 2161
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9726715906417102e-05,
+      "loss": 0.6591,
+      "step": 2162
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972646303044775e-05,
+      "loss": 0.6548,
+      "step": 2163
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.972621003915888e-05,
+      "loss": 0.6675,
+      "step": 2164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972595693255348e-05,
+      "loss": 0.6534,
+      "step": 2165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972570371063456e-05,
+      "loss": 0.6971,
+      "step": 2166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972545037340512e-05,
+      "loss": 0.6474,
+      "step": 2167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972519692086816e-05,
+      "loss": 0.5973,
+      "step": 2168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9724943353026692e-05,
+      "loss": 0.6762,
+      "step": 2169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972468966988372e-05,
+      "loss": 0.6234,
+      "step": 2170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9724435871442244e-05,
+      "loss": 0.6374,
+      "step": 2171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972418195770528e-05,
+      "loss": 0.5963,
+      "step": 2172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9723927928675844e-05,
+      "loss": 0.678,
+      "step": 2173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972367378435694e-05,
+      "loss": 0.6222,
+      "step": 2174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9723419524751578e-05,
+      "loss": 0.6535,
+      "step": 2175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972316514986278e-05,
+      "loss": 0.6727,
+      "step": 2176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9722910659693557e-05,
+      "loss": 0.7144,
+      "step": 2177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972265605424693e-05,
+      "loss": 0.641,
+      "step": 2178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972240133352592e-05,
+      "loss": 0.6344,
+      "step": 2179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9722146497533538e-05,
+      "loss": 0.6611,
+      "step": 2180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9721891546272812e-05,
+      "loss": 0.666,
+      "step": 2181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9721636479746765e-05,
+      "loss": 0.6884,
+      "step": 2182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.972138129795842e-05,
+      "loss": 0.6944,
+      "step": 2183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9721126000910802e-05,
+      "loss": 0.7121,
+      "step": 2184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9720870588606936e-05,
+      "loss": 0.6755,
+      "step": 2185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9720615061049852e-05,
+      "loss": 0.6368,
+      "step": 2186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9720359418242582e-05,
+      "loss": 0.676,
+      "step": 2187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9720103660188154e-05,
+      "loss": 0.7082,
+      "step": 2188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9719847786889602e-05,
+      "loss": 0.6201,
+      "step": 2189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9719591798349954e-05,
+      "loss": 0.7122,
+      "step": 2190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9719335694572253e-05,
+      "loss": 0.6616,
+      "step": 2191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9719079475559538e-05,
+      "loss": 0.6671,
+      "step": 2192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9718823141314834e-05,
+      "loss": 0.6449,
+      "step": 2193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971856669184119e-05,
+      "loss": 0.621,
+      "step": 2194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9718310127141642e-05,
+      "loss": 0.6277,
+      "step": 2195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9718053447219235e-05,
+      "loss": 0.6049,
+      "step": 2196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9717796652077014e-05,
+      "loss": 0.6165,
+      "step": 2197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9717539741718016e-05,
+      "loss": 0.615,
+      "step": 2198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9717282716145292e-05,
+      "loss": 0.663,
+      "step": 2199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971702557536189e-05,
+      "loss": 0.7006,
+      "step": 2200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971676831937086e-05,
+      "loss": 0.6513,
+      "step": 2201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9716510948175245e-05,
+      "loss": 0.6503,
+      "step": 2202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9716253461778105e-05,
+      "loss": 0.5981,
+      "step": 2203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9715995860182485e-05,
+      "loss": 0.6222,
+      "step": 2204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971573814339145e-05,
+      "loss": 0.6454,
+      "step": 2205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971548031140804e-05,
+      "loss": 0.6357,
+      "step": 2206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971522236423533e-05,
+      "loss": 0.6123,
+      "step": 2207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9714964301876364e-05,
+      "loss": 0.6826,
+      "step": 2208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9714706124334207e-05,
+      "loss": 0.6515,
+      "step": 2209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971444783161192e-05,
+      "loss": 0.6262,
+      "step": 2210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9714189423712567e-05,
+      "loss": 0.6912,
+      "step": 2211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9713930900639206e-05,
+      "loss": 0.6184,
+      "step": 2212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971367226239491e-05,
+      "loss": 0.6372,
+      "step": 2213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971341350898274e-05,
+      "loss": 0.7006,
+      "step": 2214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9713154640405765e-05,
+      "loss": 0.62,
+      "step": 2215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971289565666705e-05,
+      "loss": 0.6528,
+      "step": 2216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9712636557769678e-05,
+      "loss": 0.6698,
+      "step": 2217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9712377343716712e-05,
+      "loss": 0.6152,
+      "step": 2218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9712118014511223e-05,
+      "loss": 0.6091,
+      "step": 2219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971185857015629e-05,
+      "loss": 0.6574,
+      "step": 2220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9711599010654987e-05,
+      "loss": 0.6124,
+      "step": 2221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9711339336010393e-05,
+      "loss": 0.6586,
+      "step": 2222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971107954622559e-05,
+      "loss": 0.5973,
+      "step": 2223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971081964130365e-05,
+      "loss": 0.6331,
+      "step": 2224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971055962124766e-05,
+      "loss": 0.6098,
+      "step": 2225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9710299486060703e-05,
+      "loss": 0.6335,
+      "step": 2226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.971003923574586e-05,
+      "loss": 0.6483,
+      "step": 2227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970977887030622e-05,
+      "loss": 0.6137,
+      "step": 2228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9709518389744868e-05,
+      "loss": 0.6718,
+      "step": 2229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970925779406489e-05,
+      "loss": 0.6727,
+      "step": 2230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9708997083269382e-05,
+      "loss": 0.6501,
+      "step": 2231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9708736257361435e-05,
+      "loss": 0.6003,
+      "step": 2232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9708475316344133e-05,
+      "loss": 0.6659,
+      "step": 2233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9708214260220577e-05,
+      "loss": 0.5728,
+      "step": 2234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970795308899386e-05,
+      "loss": 0.636,
+      "step": 2235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9707691802667076e-05,
+      "loss": 0.6266,
+      "step": 2236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9707430401243325e-05,
+      "loss": 0.684,
+      "step": 2237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9707168884725713e-05,
+      "loss": 0.6765,
+      "step": 2238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970690725311733e-05,
+      "loss": 0.6843,
+      "step": 2239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970664550642128e-05,
+      "loss": 0.6361,
+      "step": 2240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970638364464067e-05,
+      "loss": 0.6403,
+      "step": 2241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9706121667778605e-05,
+      "loss": 0.6601,
+      "step": 2242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970585957583819e-05,
+      "loss": 0.6392,
+      "step": 2243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9705597368822528e-05,
+      "loss": 0.6143,
+      "step": 2244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9705335046734736e-05,
+      "loss": 0.6056,
+      "step": 2245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9705072609577915e-05,
+      "loss": 0.6505,
+      "step": 2246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9704810057355185e-05,
+      "loss": 0.6225,
+      "step": 2247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9704547390069653e-05,
+      "loss": 0.5879,
+      "step": 2248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9704284607724435e-05,
+      "loss": 0.6255,
+      "step": 2249
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970402171032265e-05,
+      "loss": 0.6619,
+      "step": 2250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9703758697867407e-05,
+      "loss": 0.6645,
+      "step": 2251
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9703495570361834e-05,
+      "loss": 0.6372,
+      "step": 2252
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9703232327809044e-05,
+      "loss": 0.6941,
+      "step": 2253
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9702968970212158e-05,
+      "loss": 0.6171,
+      "step": 2254
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9702705497574306e-05,
+      "loss": 0.7018,
+      "step": 2255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9702441909898603e-05,
+      "loss": 0.6013,
+      "step": 2256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9702178207188178e-05,
+      "loss": 0.6586,
+      "step": 2257
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9701914389446156e-05,
+      "loss": 0.696,
+      "step": 2258
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9701650456675663e-05,
+      "loss": 0.6375,
+      "step": 2259
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.970138640887984e-05,
+      "loss": 0.6344,
+      "step": 2260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9701122246061804e-05,
+      "loss": 0.6603,
+      "step": 2261
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9700857968224692e-05,
+      "loss": 0.6233,
+      "step": 2262
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9700593575371636e-05,
+      "loss": 0.6682,
+      "step": 2263
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9700329067505772e-05,
+      "loss": 0.5949,
+      "step": 2264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9700064444630236e-05,
+      "loss": 0.6485,
+      "step": 2265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9699799706748165e-05,
+      "loss": 0.6615,
+      "step": 2266
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.96995348538627e-05,
+      "loss": 0.6824,
+      "step": 2267
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9699269885976976e-05,
+      "loss": 0.6091,
+      "step": 2268
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9699004803094143e-05,
+      "loss": 0.6602,
+      "step": 2269
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9698739605217335e-05,
+      "loss": 0.6149,
+      "step": 2270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.96984742923497e-05,
+      "loss": 0.6128,
+      "step": 2271
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9698208864494385e-05,
+      "loss": 0.6864,
+      "step": 2272
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9697943321654535e-05,
+      "loss": 0.5946,
+      "step": 2273
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.96976776638333e-05,
+      "loss": 0.6462,
+      "step": 2274
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9697411891033828e-05,
+      "loss": 0.7159,
+      "step": 2275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969714600325927e-05,
+      "loss": 0.5784,
+      "step": 2276
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969688000051278e-05,
+      "loss": 0.6593,
+      "step": 2277
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969661388279751e-05,
+      "loss": 0.6379,
+      "step": 2278
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969634765011662e-05,
+      "loss": 0.6413,
+      "step": 2279
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9696081302473257e-05,
+      "loss": 0.6062,
+      "step": 2280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969581483987059e-05,
+      "loss": 0.6211,
+      "step": 2281
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9695548262311765e-05,
+      "loss": 0.6658,
+      "step": 2282
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9695281569799957e-05,
+      "loss": 0.6595,
+      "step": 2283
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969501476233832e-05,
+      "loss": 0.6879,
+      "step": 2284
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969474783993002e-05,
+      "loss": 0.568,
+      "step": 2285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9694480802578217e-05,
+      "loss": 0.6625,
+      "step": 2286
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9694213650286083e-05,
+      "loss": 0.6992,
+      "step": 2287
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969394638305678e-05,
+      "loss": 0.6755,
+      "step": 2288
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9693679000893484e-05,
+      "loss": 0.6506,
+      "step": 2289
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9693411503799358e-05,
+      "loss": 0.6419,
+      "step": 2290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969314389177758e-05,
+      "loss": 0.6772,
+      "step": 2291
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9692876164831314e-05,
+      "loss": 0.6185,
+      "step": 2292
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9692608322963743e-05,
+      "loss": 0.5849,
+      "step": 2293
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969234036617804e-05,
+      "loss": 0.6105,
+      "step": 2294
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9692072294477378e-05,
+      "loss": 0.6211,
+      "step": 2295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969180410786494e-05,
+      "loss": 0.6509,
+      "step": 2296
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9691535806343905e-05,
+      "loss": 0.6455,
+      "step": 2297
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969126738991745e-05,
+      "loss": 0.6573,
+      "step": 2298
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969099885858876e-05,
+      "loss": 0.6929,
+      "step": 2299
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9690730212361026e-05,
+      "loss": 0.6488,
+      "step": 2300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9690461451237423e-05,
+      "loss": 0.5831,
+      "step": 2301
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.969019257522114e-05,
+      "loss": 0.6445,
+      "step": 2302
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9689923584315364e-05,
+      "loss": 0.6763,
+      "step": 2303
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968965447852329e-05,
+      "loss": 0.63,
+      "step": 2304
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9689385257848102e-05,
+      "loss": 0.6482,
+      "step": 2305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9689115922292996e-05,
+      "loss": 0.691,
+      "step": 2306
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9688846471861163e-05,
+      "loss": 0.624,
+      "step": 2307
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.96885769065558e-05,
+      "loss": 0.6146,
+      "step": 2308
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.96883072263801e-05,
+      "loss": 0.5542,
+      "step": 2309
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968803743133726e-05,
+      "loss": 0.6864,
+      "step": 2310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9687767521430485e-05,
+      "loss": 0.6508,
+      "step": 2311
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9687497496662972e-05,
+      "loss": 0.6303,
+      "step": 2312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968722735703792e-05,
+      "loss": 0.6589,
+      "step": 2313
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968695710255853e-05,
+      "loss": 0.6575,
+      "step": 2314
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9686686733228014e-05,
+      "loss": 0.6417,
+      "step": 2315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968641624904957e-05,
+      "loss": 0.7045,
+      "step": 2316
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968614565002641e-05,
+      "loss": 0.6264,
+      "step": 2317
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9685874936161738e-05,
+      "loss": 0.6893,
+      "step": 2318
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9685604107458768e-05,
+      "loss": 0.6542,
+      "step": 2319
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968533316392071e-05,
+      "loss": 0.6461,
+      "step": 2320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9685062105550773e-05,
+      "loss": 0.645,
+      "step": 2321
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968479093235217e-05,
+      "loss": 0.6441,
+      "step": 2322
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9684519644328123e-05,
+      "loss": 0.6407,
+      "step": 2323
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9684248241481844e-05,
+      "loss": 0.6348,
+      "step": 2324
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9683976723816553e-05,
+      "loss": 0.6855,
+      "step": 2325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9683705091335467e-05,
+      "loss": 0.6835,
+      "step": 2326
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9683433344041804e-05,
+      "loss": 0.6439,
+      "step": 2327
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9683161481938794e-05,
+      "loss": 0.6287,
+      "step": 2328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9682889505029653e-05,
+      "loss": 0.6405,
+      "step": 2329
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9682617413317607e-05,
+      "loss": 0.6536,
+      "step": 2330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9682345206805885e-05,
+      "loss": 0.6686,
+      "step": 2331
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9682072885497712e-05,
+      "loss": 0.5975,
+      "step": 2332
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9681800449396317e-05,
+      "loss": 0.6634,
+      "step": 2333
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968152789850493e-05,
+      "loss": 0.6204,
+      "step": 2334
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968125523282678e-05,
+      "loss": 0.6376,
+      "step": 2335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9680982452365105e-05,
+      "loss": 0.6625,
+      "step": 2336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9680709557123135e-05,
+      "loss": 0.6287,
+      "step": 2337
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968043654710411e-05,
+      "loss": 0.6437,
+      "step": 2338
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.968016342231126e-05,
+      "loss": 0.6965,
+      "step": 2339
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967989018274783e-05,
+      "loss": 0.6205,
+      "step": 2340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9679616828417058e-05,
+      "loss": 0.6088,
+      "step": 2341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967934335932218e-05,
+      "loss": 0.6458,
+      "step": 2342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9679069775466444e-05,
+      "loss": 0.6516,
+      "step": 2343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967879607685309e-05,
+      "loss": 0.6275,
+      "step": 2344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9678522263485368e-05,
+      "loss": 0.6128,
+      "step": 2345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967824833536652e-05,
+      "loss": 0.6036,
+      "step": 2346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967797429249979e-05,
+      "loss": 0.6548,
+      "step": 2347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9677700134888438e-05,
+      "loss": 0.6545,
+      "step": 2348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9677425862535708e-05,
+      "loss": 0.639,
+      "step": 2349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967715147544485e-05,
+      "loss": 0.6142,
+      "step": 2350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967687697361912e-05,
+      "loss": 0.6152,
+      "step": 2351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9676602357061775e-05,
+      "loss": 0.6211,
+      "step": 2352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9676327625776067e-05,
+      "loss": 0.6446,
+      "step": 2353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9676052779765255e-05,
+      "loss": 0.7061,
+      "step": 2354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9675777819032592e-05,
+      "loss": 0.6398,
+      "step": 2355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9675502743581348e-05,
+      "loss": 0.6813,
+      "step": 2356
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9675227553414778e-05,
+      "loss": 0.7004,
+      "step": 2357
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9674952248536146e-05,
+      "loss": 0.6159,
+      "step": 2358
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9674676828948717e-05,
+      "loss": 0.6823,
+      "step": 2359
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9674401294655758e-05,
+      "loss": 0.6689,
+      "step": 2360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9674125645660527e-05,
+      "loss": 0.6213,
+      "step": 2361
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9673849881966302e-05,
+      "loss": 0.6579,
+      "step": 2362
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967357400357635e-05,
+      "loss": 0.6596,
+      "step": 2363
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967329801049394e-05,
+      "loss": 0.5946,
+      "step": 2364
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967302190272235e-05,
+      "loss": 0.59,
+      "step": 2365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967274568026484e-05,
+      "loss": 0.6563,
+      "step": 2366
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.96724693431247e-05,
+      "loss": 0.6346,
+      "step": 2367
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9672192891305197e-05,
+      "loss": 0.6637,
+      "step": 2368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967191632480961e-05,
+      "loss": 0.652,
+      "step": 2369
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9671639643641227e-05,
+      "loss": 0.659,
+      "step": 2370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9671362847803316e-05,
+      "loss": 0.6262,
+      "step": 2371
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9671085937299166e-05,
+      "loss": 0.6316,
+      "step": 2372
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.967080891213206e-05,
+      "loss": 0.6868,
+      "step": 2373
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9670531772305277e-05,
+      "loss": 0.5558,
+      "step": 2374
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9670254517822108e-05,
+      "loss": 0.6651,
+      "step": 2375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9669977148685843e-05,
+      "loss": 0.7398,
+      "step": 2376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966969966489976e-05,
+      "loss": 0.679,
+      "step": 2377
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966942206646716e-05,
+      "loss": 0.6307,
+      "step": 2378
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9669144353391326e-05,
+      "loss": 0.6484,
+      "step": 2379
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9668866525675553e-05,
+      "loss": 0.7205,
+      "step": 2380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966858858332314e-05,
+      "loss": 0.6429,
+      "step": 2381
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9668310526337376e-05,
+      "loss": 0.6772,
+      "step": 2382
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9668032354721562e-05,
+      "loss": 0.6608,
+      "step": 2383
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966775406847899e-05,
+      "loss": 0.576,
+      "step": 2384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966747566761297e-05,
+      "loss": 0.6594,
+      "step": 2385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9667197152126794e-05,
+      "loss": 0.6165,
+      "step": 2386
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966691852202377e-05,
+      "loss": 0.6754,
+      "step": 2387
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9666639777307193e-05,
+      "loss": 0.6696,
+      "step": 2388
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9666360917980375e-05,
+      "loss": 0.6064,
+      "step": 2389
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966608194404662e-05,
+      "loss": 0.7055,
+      "step": 2390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9665802855509236e-05,
+      "loss": 0.6496,
+      "step": 2391
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966552365237153e-05,
+      "loss": 0.6624,
+      "step": 2392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9665244334636818e-05,
+      "loss": 0.6558,
+      "step": 2393
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9664964902308404e-05,
+      "loss": 0.6365,
+      "step": 2394
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9664685355389607e-05,
+      "loss": 0.6595,
+      "step": 2395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966440569388374e-05,
+      "loss": 0.6099,
+      "step": 2396
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9664125917794116e-05,
+      "loss": 0.6556,
+      "step": 2397
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9663846027124057e-05,
+      "loss": 0.6563,
+      "step": 2398
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9663566021876875e-05,
+      "loss": 0.6486,
+      "step": 2399
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9663285902055893e-05,
+      "loss": 0.6313,
+      "step": 2400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9663005667664435e-05,
+      "loss": 0.7002,
+      "step": 2401
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966272531870582e-05,
+      "loss": 0.6634,
+      "step": 2402
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9662444855183373e-05,
+      "loss": 0.6136,
+      "step": 2403
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966216427710042e-05,
+      "loss": 0.6273,
+      "step": 2404
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9661883584460288e-05,
+      "loss": 0.6624,
+      "step": 2405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9661602777266303e-05,
+      "loss": 0.6488,
+      "step": 2406
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966132185552179e-05,
+      "loss": 0.6148,
+      "step": 2407
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966104081923009e-05,
+      "loss": 0.6318,
+      "step": 2408
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966075966839453e-05,
+      "loss": 0.6337,
+      "step": 2409
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.966047840301844e-05,
+      "loss": 0.6249,
+      "step": 2410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9660197023105162e-05,
+      "loss": 0.7171,
+      "step": 2411
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9659915528658025e-05,
+      "loss": 0.5855,
+      "step": 2412
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9659633919680372e-05,
+      "loss": 0.6123,
+      "step": 2413
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9659352196175537e-05,
+      "loss": 0.6755,
+      "step": 2414
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9659070358146863e-05,
+      "loss": 0.6762,
+      "step": 2415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965878840559769e-05,
+      "loss": 0.6306,
+      "step": 2416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9658506338531366e-05,
+      "loss": 0.6536,
+      "step": 2417
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9658224156951227e-05,
+      "loss": 0.6233,
+      "step": 2418
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9657941860860627e-05,
+      "loss": 0.6342,
+      "step": 2419
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9657659450262905e-05,
+      "loss": 0.6431,
+      "step": 2420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9657376925161413e-05,
+      "loss": 0.6225,
+      "step": 2421
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9657094285559505e-05,
+      "loss": 0.6511,
+      "step": 2422
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9656811531460524e-05,
+      "loss": 0.6108,
+      "step": 2423
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965652866286783e-05,
+      "loss": 0.6491,
+      "step": 2424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965624567978477e-05,
+      "loss": 0.6653,
+      "step": 2425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9655962582214702e-05,
+      "loss": 0.6419,
+      "step": 2426
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965567937016099e-05,
+      "loss": 0.6838,
+      "step": 2427
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9655396043626975e-05,
+      "loss": 0.6447,
+      "step": 2428
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9655112602616027e-05,
+      "loss": 0.6391,
+      "step": 2429
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9654829047131508e-05,
+      "loss": 0.7095,
+      "step": 2430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9654545377176778e-05,
+      "loss": 0.6124,
+      "step": 2431
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.96542615927552e-05,
+      "loss": 0.7071,
+      "step": 2432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9653977693870138e-05,
+      "loss": 0.6509,
+      "step": 2433
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9653693680524957e-05,
+      "loss": 0.7269,
+      "step": 2434
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965340955272302e-05,
+      "loss": 0.6307,
+      "step": 2435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965312531046771e-05,
+      "loss": 0.6143,
+      "step": 2436
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9652840953762384e-05,
+      "loss": 0.6112,
+      "step": 2437
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965255648261042e-05,
+      "loss": 0.611,
+      "step": 2438
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9652271897015185e-05,
+      "loss": 0.6468,
+      "step": 2439
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965198719698006e-05,
+      "loss": 0.6241,
+      "step": 2440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965170238250841e-05,
+      "loss": 0.6413,
+      "step": 2441
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9651417453603625e-05,
+      "loss": 0.6714,
+      "step": 2442
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9651132410269075e-05,
+      "loss": 0.6638,
+      "step": 2443
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965084725250814e-05,
+      "loss": 0.6673,
+      "step": 2444
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965056198032421e-05,
+      "loss": 0.6142,
+      "step": 2445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.965027659372065e-05,
+      "loss": 0.6838,
+      "step": 2446
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.964999109270086e-05,
+      "loss": 0.66,
+      "step": 2447
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9649705477268215e-05,
+      "loss": 0.6516,
+      "step": 2448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9649419747426105e-05,
+      "loss": 0.6575,
+      "step": 2449
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9649133903177918e-05,
+      "loss": 0.6742,
+      "step": 2450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9648847944527042e-05,
+      "loss": 0.6536,
+      "step": 2451
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.964856187147687e-05,
+      "loss": 0.7029,
+      "step": 2452
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.964827568403079e-05,
+      "loss": 0.6345,
+      "step": 2453
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9647989382192195e-05,
+      "loss": 0.6208,
+      "step": 2454
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964770296596448e-05,
+      "loss": 0.6461,
+      "step": 2455
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9647416435351047e-05,
+      "loss": 0.6667,
+      "step": 2456
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9647129790355282e-05,
+      "loss": 0.6493,
+      "step": 2457
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9646843030980597e-05,
+      "loss": 0.6895,
+      "step": 2458
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9646556157230378e-05,
+      "loss": 0.6002,
+      "step": 2459
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9646269169108038e-05,
+      "loss": 0.6145,
+      "step": 2460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9645982066616968e-05,
+      "loss": 0.6683,
+      "step": 2461
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9645694849760584e-05,
+      "loss": 0.6422,
+      "step": 2462
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9645407518542284e-05,
+      "loss": 0.6735,
+      "step": 2463
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9645120072965475e-05,
+      "loss": 0.6339,
+      "step": 2464
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9644832513033568e-05,
+      "loss": 0.6536,
+      "step": 2465
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964454483874997e-05,
+      "loss": 0.5935,
+      "step": 2466
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9644257050118092e-05,
+      "loss": 0.6051,
+      "step": 2467
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9643969147141348e-05,
+      "loss": 0.6445,
+      "step": 2468
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9643681129823152e-05,
+      "loss": 0.6039,
+      "step": 2469
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9643392998166912e-05,
+      "loss": 0.632,
+      "step": 2470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964310475217605e-05,
+      "loss": 0.647,
+      "step": 2471
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9642816391853987e-05,
+      "loss": 0.6165,
+      "step": 2472
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9642527917204132e-05,
+      "loss": 0.6499,
+      "step": 2473
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9642239328229915e-05,
+      "loss": 0.6483,
+      "step": 2474
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964195062493475e-05,
+      "loss": 0.6338,
+      "step": 2475
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9641661807322066e-05,
+      "loss": 0.6835,
+      "step": 2476
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9641372875395285e-05,
+      "loss": 0.6518,
+      "step": 2477
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964108382915783e-05,
+      "loss": 0.6701,
+      "step": 2478
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9640794668613128e-05,
+      "loss": 0.604,
+      "step": 2479
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9640505393764614e-05,
+      "loss": 0.6215,
+      "step": 2480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9640216004615712e-05,
+      "loss": 0.624,
+      "step": 2481
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963992650116985e-05,
+      "loss": 0.6407,
+      "step": 2482
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9639636883430467e-05,
+      "loss": 0.6325,
+      "step": 2483
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9639347151401e-05,
+      "loss": 0.635,
+      "step": 2484
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963905730508487e-05,
+      "loss": 0.607,
+      "step": 2485
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963876734448553e-05,
+      "loss": 0.6267,
+      "step": 2486
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9638477269606405e-05,
+      "loss": 0.646,
+      "step": 2487
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963818708045094e-05,
+      "loss": 0.5939,
+      "step": 2488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9637896777022577e-05,
+      "loss": 0.6435,
+      "step": 2489
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9637606359324756e-05,
+      "loss": 0.6321,
+      "step": 2490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9637315827360916e-05,
+      "loss": 0.6389,
+      "step": 2491
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963702518113451e-05,
+      "loss": 0.6848,
+      "step": 2492
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9636734420648977e-05,
+      "loss": 0.6075,
+      "step": 2493
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963644354590777e-05,
+      "loss": 0.6157,
+      "step": 2494
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963615255691433e-05,
+      "loss": 0.6214,
+      "step": 2495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9635861453672113e-05,
+      "loss": 0.6631,
+      "step": 2496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963557023618457e-05,
+      "loss": 0.6313,
+      "step": 2497
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9635278904455152e-05,
+      "loss": 0.6703,
+      "step": 2498
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9634987458487315e-05,
+      "loss": 0.635,
+      "step": 2499
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9634695898284515e-05,
+      "loss": 0.6346,
+      "step": 2500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9634404223850202e-05,
+      "loss": 0.6277,
+      "step": 2501
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9634112435187844e-05,
+      "loss": 0.7033,
+      "step": 2502
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9633820532300894e-05,
+      "loss": 0.6695,
+      "step": 2503
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9633528515192816e-05,
+      "loss": 0.6626,
+      "step": 2504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963323638386707e-05,
+      "loss": 0.6023,
+      "step": 2505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9632944138327118e-05,
+      "loss": 0.6715,
+      "step": 2506
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9632651778576434e-05,
+      "loss": 0.6308,
+      "step": 2507
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9632359304618472e-05,
+      "loss": 0.638,
+      "step": 2508
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9632066716456708e-05,
+      "loss": 0.6387,
+      "step": 2509
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9631774014094607e-05,
+      "loss": 0.6195,
+      "step": 2510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963148119753564e-05,
+      "loss": 0.6731,
+      "step": 2511
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9631188266783282e-05,
+      "loss": 0.6624,
+      "step": 2512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9630895221841004e-05,
+      "loss": 0.7061,
+      "step": 2513
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963060206271228e-05,
+      "loss": 0.6501,
+      "step": 2514
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963030878940058e-05,
+      "loss": 0.6438,
+      "step": 2515
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9630015401909393e-05,
+      "loss": 0.6296,
+      "step": 2516
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962972190024219e-05,
+      "loss": 0.6144,
+      "step": 2517
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962942828440245e-05,
+      "loss": 0.6192,
+      "step": 2518
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962913455439366e-05,
+      "loss": 0.6291,
+      "step": 2519
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9628840710219298e-05,
+      "loss": 0.663,
+      "step": 2520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9628546751882846e-05,
+      "loss": 0.6496,
+      "step": 2521
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9628252679387795e-05,
+      "loss": 0.6621,
+      "step": 2522
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962795849273763e-05,
+      "loss": 0.6489,
+      "step": 2523
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9627664191935833e-05,
+      "loss": 0.6585,
+      "step": 2524
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96273697769859e-05,
+      "loss": 0.6489,
+      "step": 2525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962707524789132e-05,
+      "loss": 0.6375,
+      "step": 2526
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9626780604655583e-05,
+      "loss": 0.6967,
+      "step": 2527
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962648584728219e-05,
+      "loss": 0.6458,
+      "step": 2528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9626190975774625e-05,
+      "loss": 0.7115,
+      "step": 2529
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9625895990136386e-05,
+      "loss": 0.6716,
+      "step": 2530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962560089037098e-05,
+      "loss": 0.6522,
+      "step": 2531
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962530567648189e-05,
+      "loss": 0.6635,
+      "step": 2532
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9625010348472634e-05,
+      "loss": 0.6238,
+      "step": 2533
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96247149063467e-05,
+      "loss": 0.6241,
+      "step": 2534
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9624419350107598e-05,
+      "loss": 0.632,
+      "step": 2535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9624123679758828e-05,
+      "loss": 0.5853,
+      "step": 2536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9623827895303895e-05,
+      "loss": 0.6243,
+      "step": 2537
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9623531996746313e-05,
+      "loss": 0.5766,
+      "step": 2538
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962323598408958e-05,
+      "loss": 0.701,
+      "step": 2539
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9622939857337216e-05,
+      "loss": 0.6765,
+      "step": 2540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9622643616492726e-05,
+      "loss": 0.6712,
+      "step": 2541
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962234726155962e-05,
+      "loss": 0.6085,
+      "step": 2542
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9622050792541417e-05,
+      "loss": 0.6496,
+      "step": 2543
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962175420944163e-05,
+      "loss": 0.6531,
+      "step": 2544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9621457512263776e-05,
+      "loss": 0.6274,
+      "step": 2545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962116070101137e-05,
+      "loss": 0.5921,
+      "step": 2546
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9620863775687934e-05,
+      "loss": 0.6233,
+      "step": 2547
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.962056673629699e-05,
+      "loss": 0.6562,
+      "step": 2548
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9620269582842054e-05,
+      "loss": 0.6512,
+      "step": 2549
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961997231532665e-05,
+      "loss": 0.6094,
+      "step": 2550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9619674933754308e-05,
+      "loss": 0.7113,
+      "step": 2551
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9619377438128552e-05,
+      "loss": 0.6229,
+      "step": 2552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9619079828452906e-05,
+      "loss": 0.6474,
+      "step": 2553
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96187821047309e-05,
+      "loss": 0.6051,
+      "step": 2554
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9618484266966067e-05,
+      "loss": 0.6143,
+      "step": 2555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9618186315161935e-05,
+      "loss": 0.6597,
+      "step": 2556
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9617888249322035e-05,
+      "loss": 0.668,
+      "step": 2557
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9617590069449904e-05,
+      "loss": 0.6574,
+      "step": 2558
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9617291775549077e-05,
+      "loss": 0.7242,
+      "step": 2559
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961699336762309e-05,
+      "loss": 0.6199,
+      "step": 2560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961669484567548e-05,
+      "loss": 0.6537,
+      "step": 2561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9616396209709786e-05,
+      "loss": 0.6306,
+      "step": 2562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9616097459729553e-05,
+      "loss": 0.6368,
+      "step": 2563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9615798595738323e-05,
+      "loss": 0.5995,
+      "step": 2564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961549961773963e-05,
+      "loss": 0.6665,
+      "step": 2565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961520052573703e-05,
+      "loss": 0.6442,
+      "step": 2566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961490131973406e-05,
+      "loss": 0.6282,
+      "step": 2567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9614601999734278e-05,
+      "loss": 0.654,
+      "step": 2568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9614302565741222e-05,
+      "loss": 0.6534,
+      "step": 2569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961400301775845e-05,
+      "loss": 0.6323,
+      "step": 2570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9613703355789508e-05,
+      "loss": 0.6136,
+      "step": 2571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961340357983795e-05,
+      "loss": 0.6763,
+      "step": 2572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9613103689907338e-05,
+      "loss": 0.6529,
+      "step": 2573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9612803686001218e-05,
+      "loss": 0.6389,
+      "step": 2574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961250356812315e-05,
+      "loss": 0.629,
+      "step": 2575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9612203336276688e-05,
+      "loss": 0.6543,
+      "step": 2576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96119029904654e-05,
+      "loss": 0.6359,
+      "step": 2577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9611602530692843e-05,
+      "loss": 0.6652,
+      "step": 2578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961130195696258e-05,
+      "loss": 0.6808,
+      "step": 2579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961100126927817e-05,
+      "loss": 0.6116,
+      "step": 2580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9610700467643182e-05,
+      "loss": 0.6244,
+      "step": 2581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.961039955206119e-05,
+      "loss": 0.6722,
+      "step": 2582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9610098522535745e-05,
+      "loss": 0.642,
+      "step": 2583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9609797379070427e-05,
+      "loss": 0.6394,
+      "step": 2584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9609496121668807e-05,
+      "loss": 0.6991,
+      "step": 2585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960919475033446e-05,
+      "loss": 0.5908,
+      "step": 2586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9608893265070943e-05,
+      "loss": 0.6498,
+      "step": 2587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9608591665881845e-05,
+      "loss": 0.6697,
+      "step": 2588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960828995277074e-05,
+      "loss": 0.5942,
+      "step": 2589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96079881257412e-05,
+      "loss": 0.6464,
+      "step": 2590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960768618479681e-05,
+      "loss": 0.588,
+      "step": 2591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9607384129941143e-05,
+      "loss": 0.6606,
+      "step": 2592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9607081961177785e-05,
+      "loss": 0.6773,
+      "step": 2593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960677967851032e-05,
+      "loss": 0.6848,
+      "step": 2594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9606477281942323e-05,
+      "loss": 0.6029,
+      "step": 2595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9606174771477392e-05,
+      "loss": 0.6399,
+      "step": 2596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9605872147119104e-05,
+      "loss": 0.6517,
+      "step": 2597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9605569408871053e-05,
+      "loss": 0.601,
+      "step": 2598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9605266556736823e-05,
+      "loss": 0.6026,
+      "step": 2599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9604963590720007e-05,
+      "loss": 0.6839,
+      "step": 2600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9604660510824198e-05,
+      "loss": 0.6355,
+      "step": 2601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9604357317052988e-05,
+      "loss": 0.7125,
+      "step": 2602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9604054009409975e-05,
+      "loss": 0.6094,
+      "step": 2603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960375058789875e-05,
+      "loss": 0.6309,
+      "step": 2604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9603447052522915e-05,
+      "loss": 0.6325,
+      "step": 2605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9603143403286064e-05,
+      "loss": 0.6619,
+      "step": 2606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9602839640191803e-05,
+      "loss": 0.6805,
+      "step": 2607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9602535763243726e-05,
+      "loss": 0.5466,
+      "step": 2608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9602231772445443e-05,
+      "loss": 0.6244,
+      "step": 2609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960192766780056e-05,
+      "loss": 0.7076,
+      "step": 2610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960162344931267e-05,
+      "loss": 0.5925,
+      "step": 2611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960131911698539e-05,
+      "loss": 0.6809,
+      "step": 2612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9601014670822328e-05,
+      "loss": 0.6642,
+      "step": 2613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9600710110827093e-05,
+      "loss": 0.6569,
+      "step": 2614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960040543700329e-05,
+      "loss": 0.7129,
+      "step": 2615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.960010064935454e-05,
+      "loss": 0.6078,
+      "step": 2616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9599795747884452e-05,
+      "loss": 0.6601,
+      "step": 2617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959949073259664e-05,
+      "loss": 0.6384,
+      "step": 2618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9599185603494725e-05,
+      "loss": 0.6872,
+      "step": 2619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959888036058232e-05,
+      "loss": 0.6482,
+      "step": 2620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9598575003863043e-05,
+      "loss": 0.5938,
+      "step": 2621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959826953334052e-05,
+      "loss": 0.629,
+      "step": 2622
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959796394901837e-05,
+      "loss": 0.6484,
+      "step": 2623
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9597658250900214e-05,
+      "loss": 0.6328,
+      "step": 2624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959735243898968e-05,
+      "loss": 0.6658,
+      "step": 2625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9597046513290394e-05,
+      "loss": 0.6356,
+      "step": 2626
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9596740473805976e-05,
+      "loss": 0.5979,
+      "step": 2627
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9596434320540063e-05,
+      "loss": 0.6407,
+      "step": 2628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959612805349628e-05,
+      "loss": 0.6532,
+      "step": 2629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9595821672678263e-05,
+      "loss": 0.665,
+      "step": 2630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9595515178089636e-05,
+      "loss": 0.6219,
+      "step": 2631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9595208569734042e-05,
+      "loss": 0.6687,
+      "step": 2632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9594901847615113e-05,
+      "loss": 0.6375,
+      "step": 2633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9594595011736485e-05,
+      "loss": 0.5864,
+      "step": 2634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9594288062101793e-05,
+      "loss": 0.6177,
+      "step": 2635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9593980998714682e-05,
+      "loss": 0.6532,
+      "step": 2636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959367382157879e-05,
+      "loss": 0.6385,
+      "step": 2637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9593366530697758e-05,
+      "loss": 0.6591,
+      "step": 2638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9593059126075228e-05,
+      "loss": 0.6708,
+      "step": 2639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9592751607714846e-05,
+      "loss": 0.5967,
+      "step": 2640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959244397562026e-05,
+      "loss": 0.6308,
+      "step": 2641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959213622979512e-05,
+      "loss": 0.6708,
+      "step": 2642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9591828370243066e-05,
+      "loss": 0.5922,
+      "step": 2643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959152039696776e-05,
+      "loss": 0.6431,
+      "step": 2644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9591212309972838e-05,
+      "loss": 0.656,
+      "step": 2645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9590904109261965e-05,
+      "loss": 0.6308,
+      "step": 2646
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9590595794838792e-05,
+      "loss": 0.6837,
+      "step": 2647
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.959028736670697e-05,
+      "loss": 0.6679,
+      "step": 2648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9589978824870164e-05,
+      "loss": 0.6359,
+      "step": 2649
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9589670169332025e-05,
+      "loss": 0.6262,
+      "step": 2650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9589361400096215e-05,
+      "loss": 0.6332,
+      "step": 2651
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.95890525171664e-05,
+      "loss": 0.6248,
+      "step": 2652
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958874352054623e-05,
+      "loss": 0.6816,
+      "step": 2653
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9588434410239378e-05,
+      "loss": 0.646,
+      "step": 2654
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9588125186249506e-05,
+      "loss": 0.688,
+      "step": 2655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9587815848580283e-05,
+      "loss": 0.6323,
+      "step": 2656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9587506397235374e-05,
+      "loss": 0.7012,
+      "step": 2657
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958719683221845e-05,
+      "loss": 0.6704,
+      "step": 2658
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9586887153533177e-05,
+      "loss": 0.6393,
+      "step": 2659
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958657736118323e-05,
+      "loss": 0.6744,
+      "step": 2660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9586267455172282e-05,
+      "loss": 0.6782,
+      "step": 2661
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9585957435504003e-05,
+      "loss": 0.6519,
+      "step": 2662
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9585647302182077e-05,
+      "loss": 0.6364,
+      "step": 2663
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9585337055210175e-05,
+      "loss": 0.6572,
+      "step": 2664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9585026694591978e-05,
+      "loss": 0.6484,
+      "step": 2665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9584716220331162e-05,
+      "loss": 0.6856,
+      "step": 2666
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9584405632431413e-05,
+      "loss": 0.6852,
+      "step": 2667
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9584094930896407e-05,
+      "loss": 0.6405,
+      "step": 2668
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958378411572983e-05,
+      "loss": 0.6483,
+      "step": 2669
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9583473186935376e-05,
+      "loss": 0.6019,
+      "step": 2670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958316214451672e-05,
+      "loss": 0.6268,
+      "step": 2671
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9582850988477554e-05,
+      "loss": 0.6486,
+      "step": 2672
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958253971882157e-05,
+      "loss": 0.5842,
+      "step": 2673
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958222833555245e-05,
+      "loss": 0.6209,
+      "step": 2674
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9581916838673898e-05,
+      "loss": 0.677,
+      "step": 2675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9581605228189598e-05,
+      "loss": 0.6585,
+      "step": 2676
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9581293504103245e-05,
+      "loss": 0.6389,
+      "step": 2677
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.958098166641854e-05,
+      "loss": 0.6431,
+      "step": 2678
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9580669715139176e-05,
+      "loss": 0.7201,
+      "step": 2679
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9580357650268853e-05,
+      "loss": 0.7021,
+      "step": 2680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9580045471811272e-05,
+      "loss": 0.6655,
+      "step": 2681
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9579733179770134e-05,
+      "loss": 0.6091,
+      "step": 2682
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957942077414914e-05,
+      "loss": 0.6935,
+      "step": 2683
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957910825495199e-05,
+      "loss": 0.6861,
+      "step": 2684
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.95787956221824e-05,
+      "loss": 0.6154,
+      "step": 2685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9578482875844068e-05,
+      "loss": 0.6505,
+      "step": 2686
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9578170015940706e-05,
+      "loss": 0.6189,
+      "step": 2687
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957785704247602e-05,
+      "loss": 0.5898,
+      "step": 2688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9577543955453725e-05,
+      "loss": 0.6318,
+      "step": 2689
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9577230754877526e-05,
+      "loss": 0.6819,
+      "step": 2690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9576917440751147e-05,
+      "loss": 0.6584,
+      "step": 2691
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9576604013078296e-05,
+      "loss": 0.6755,
+      "step": 2692
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9576290471862685e-05,
+      "loss": 0.6354,
+      "step": 2693
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9575976817108044e-05,
+      "loss": 0.6703,
+      "step": 2694
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9575663048818078e-05,
+      "loss": 0.6471,
+      "step": 2695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9575349166996516e-05,
+      "loss": 0.6442,
+      "step": 2696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9575035171647077e-05,
+      "loss": 0.6109,
+      "step": 2697
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957472106277348e-05,
+      "loss": 0.6067,
+      "step": 2698
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9574406840379456e-05,
+      "loss": 0.712,
+      "step": 2699
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9574092504468728e-05,
+      "loss": 0.6462,
+      "step": 2700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957377805504502e-05,
+      "loss": 0.6295,
+      "step": 2701
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957346349211206e-05,
+      "loss": 0.6998,
+      "step": 2702
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9573148815673587e-05,
+      "loss": 0.6651,
+      "step": 2703
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9572834025733318e-05,
+      "loss": 0.6651,
+      "step": 2704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9572519122294994e-05,
+      "loss": 0.6137,
+      "step": 2705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957220410536235e-05,
+      "loss": 0.6569,
+      "step": 2706
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9571888974939112e-05,
+      "loss": 0.6337,
+      "step": 2707
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9571573731029024e-05,
+      "loss": 0.6113,
+      "step": 2708
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9571258373635823e-05,
+      "loss": 0.5898,
+      "step": 2709
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957094290276324e-05,
+      "loss": 0.6605,
+      "step": 2710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.957062731841503e-05,
+      "loss": 0.6398,
+      "step": 2711
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9570311620594923e-05,
+      "loss": 0.6479,
+      "step": 2712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9569995809306663e-05,
+      "loss": 0.5832,
+      "step": 2713
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9569679884554e-05,
+      "loss": 0.6633,
+      "step": 2714
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9569363846340672e-05,
+      "loss": 0.5929,
+      "step": 2715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9569047694670433e-05,
+      "loss": 0.6923,
+      "step": 2716
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956873142954703e-05,
+      "loss": 0.6365,
+      "step": 2717
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9568415050974207e-05,
+      "loss": 0.6549,
+      "step": 2718
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9568098558955725e-05,
+      "loss": 0.6108,
+      "step": 2719
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9567781953495326e-05,
+      "loss": 0.6289,
+      "step": 2720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9567465234596772e-05,
+      "loss": 0.584,
+      "step": 2721
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9567148402263813e-05,
+      "loss": 0.66,
+      "step": 2722
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956683145650021e-05,
+      "loss": 0.6591,
+      "step": 2723
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9566514397309713e-05,
+      "loss": 0.6642,
+      "step": 2724
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956619722469609e-05,
+      "loss": 0.6585,
+      "step": 2725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9565879938663096e-05,
+      "loss": 0.6457,
+      "step": 2726
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9565562539214497e-05,
+      "loss": 0.6501,
+      "step": 2727
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9565245026354048e-05,
+      "loss": 0.6403,
+      "step": 2728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9564927400085523e-05,
+      "loss": 0.6898,
+      "step": 2729
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9564609660412684e-05,
+      "loss": 0.6217,
+      "step": 2730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9564291807339296e-05,
+      "loss": 0.5731,
+      "step": 2731
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9563973840869136e-05,
+      "loss": 0.619,
+      "step": 2732
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956365576100596e-05,
+      "loss": 0.6802,
+      "step": 2733
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956333756775355e-05,
+      "loss": 0.6326,
+      "step": 2734
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9563019261115674e-05,
+      "loss": 0.6635,
+      "step": 2735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956270084109611e-05,
+      "loss": 0.6747,
+      "step": 2736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9562382307698628e-05,
+      "loss": 0.6393,
+      "step": 2737
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956206366092701e-05,
+      "loss": 0.6167,
+      "step": 2738
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956174490078503e-05,
+      "loss": 0.6309,
+      "step": 2739
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956142602727647e-05,
+      "loss": 0.5782,
+      "step": 2740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.956110704040511e-05,
+      "loss": 0.6235,
+      "step": 2741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9560787940174725e-05,
+      "loss": 0.6865,
+      "step": 2742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9560468726589107e-05,
+      "loss": 0.6216,
+      "step": 2743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9560149399652042e-05,
+      "loss": 0.6084,
+      "step": 2744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955982995936731e-05,
+      "loss": 0.6352,
+      "step": 2745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95595104057387e-05,
+      "loss": 0.6544,
+      "step": 2746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955919073877e-05,
+      "loss": 0.6926,
+      "step": 2747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9558870958465005e-05,
+      "loss": 0.7066,
+      "step": 2748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95585510648275e-05,
+      "loss": 0.6209,
+      "step": 2749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955823105786128e-05,
+      "loss": 0.6425,
+      "step": 2750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9557910937570143e-05,
+      "loss": 0.6129,
+      "step": 2751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9557590703957874e-05,
+      "loss": 0.719,
+      "step": 2752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9557270357028285e-05,
+      "loss": 0.6424,
+      "step": 2753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9556949896785162e-05,
+      "loss": 0.6297,
+      "step": 2754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955662932323231e-05,
+      "loss": 0.6891,
+      "step": 2755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9556308636373525e-05,
+      "loss": 0.63,
+      "step": 2756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9555987836212615e-05,
+      "loss": 0.6266,
+      "step": 2757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955566692275338e-05,
+      "loss": 0.6845,
+      "step": 2758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9555345895999626e-05,
+      "loss": 0.6472,
+      "step": 2759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9555024755955158e-05,
+      "loss": 0.6683,
+      "step": 2760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9554703502623783e-05,
+      "loss": 0.5818,
+      "step": 2761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9554382136009315e-05,
+      "loss": 0.617,
+      "step": 2762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955406065611556e-05,
+      "loss": 0.6291,
+      "step": 2763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9553739062946327e-05,
+      "loss": 0.6285,
+      "step": 2764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9553417356505435e-05,
+      "loss": 0.6387,
+      "step": 2765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9553095536796692e-05,
+      "loss": 0.7261,
+      "step": 2766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955277360382392e-05,
+      "loss": 0.6218,
+      "step": 2767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955245155759093e-05,
+      "loss": 0.6988,
+      "step": 2768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9552129398101543e-05,
+      "loss": 0.597,
+      "step": 2769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955180712535958e-05,
+      "loss": 0.6237,
+      "step": 2770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9551484739368856e-05,
+      "loss": 0.6066,
+      "step": 2771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9551162240133203e-05,
+      "loss": 0.6079,
+      "step": 2772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9550839627656438e-05,
+      "loss": 0.6488,
+      "step": 2773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955051690194239e-05,
+      "loss": 0.6178,
+      "step": 2774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955019406299488e-05,
+      "loss": 0.5575,
+      "step": 2775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954987111081774e-05,
+      "loss": 0.6515,
+      "step": 2776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9549548045414794e-05,
+      "loss": 0.644,
+      "step": 2777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9549224866789874e-05,
+      "loss": 0.6518,
+      "step": 2778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9548901574946816e-05,
+      "loss": 0.6289,
+      "step": 2779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954857816988945e-05,
+      "loss": 0.6311,
+      "step": 2780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9548254651621614e-05,
+      "loss": 0.6149,
+      "step": 2781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9547931020147137e-05,
+      "loss": 0.623,
+      "step": 2782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954760727546986e-05,
+      "loss": 0.6721,
+      "step": 2783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9547283417593625e-05,
+      "loss": 0.6467,
+      "step": 2784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9546959446522263e-05,
+      "loss": 0.6447,
+      "step": 2785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954663536225962e-05,
+      "loss": 0.632,
+      "step": 2786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9546311164809538e-05,
+      "loss": 0.6277,
+      "step": 2787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954598685417586e-05,
+      "loss": 0.6762,
+      "step": 2788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954566243036244e-05,
+      "loss": 0.597,
+      "step": 2789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954533789337311e-05,
+      "loss": 0.6238,
+      "step": 2790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9545013243211723e-05,
+      "loss": 0.6793,
+      "step": 2791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9544688479882135e-05,
+      "loss": 0.6244,
+      "step": 2792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9544363603388187e-05,
+      "loss": 0.6815,
+      "step": 2793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9544038613733733e-05,
+      "loss": 0.5783,
+      "step": 2794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9543713510922633e-05,
+      "loss": 0.6111,
+      "step": 2795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9543388294958733e-05,
+      "loss": 0.6687,
+      "step": 2796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9543062965845895e-05,
+      "loss": 0.6304,
+      "step": 2797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954273752358797e-05,
+      "loss": 0.6509,
+      "step": 2798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9542411968188824e-05,
+      "loss": 0.6234,
+      "step": 2799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954208629965231e-05,
+      "loss": 0.6132,
+      "step": 2800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9541760517982294e-05,
+      "loss": 0.6042,
+      "step": 2801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9541434623182636e-05,
+      "loss": 0.6635,
+      "step": 2802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9541108615257198e-05,
+      "loss": 0.6475,
+      "step": 2803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954078249420985e-05,
+      "loss": 0.6577,
+      "step": 2804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9540456260044454e-05,
+      "loss": 0.6458,
+      "step": 2805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9540129912764888e-05,
+      "loss": 0.6328,
+      "step": 2806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9539803452375005e-05,
+      "loss": 0.6905,
+      "step": 2807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953947687887869e-05,
+      "loss": 0.6322,
+      "step": 2808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9539150192279807e-05,
+      "loss": 0.6796,
+      "step": 2809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953882339258223e-05,
+      "loss": 0.6208,
+      "step": 2810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9538496479789834e-05,
+      "loss": 0.655,
+      "step": 2811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95381694539065e-05,
+      "loss": 0.6464,
+      "step": 2812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95378423149361e-05,
+      "loss": 0.6313,
+      "step": 2813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953751506288251e-05,
+      "loss": 0.6682,
+      "step": 2814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9537187697749615e-05,
+      "loss": 0.6181,
+      "step": 2815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95368602195413e-05,
+      "loss": 0.6531,
+      "step": 2816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953653262826144e-05,
+      "loss": 0.6319,
+      "step": 2817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9536204923913922e-05,
+      "loss": 0.6597,
+      "step": 2818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953587710650263e-05,
+      "loss": 0.6444,
+      "step": 2819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9535549176031455e-05,
+      "loss": 0.598,
+      "step": 2820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9535221132504277e-05,
+      "loss": 0.6913,
+      "step": 2821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9534892975924995e-05,
+      "loss": 0.6836,
+      "step": 2822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9534564706297494e-05,
+      "loss": 0.6893,
+      "step": 2823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9534236323625667e-05,
+      "loss": 0.6624,
+      "step": 2824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9533907827913406e-05,
+      "loss": 0.6024,
+      "step": 2825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953357921916461e-05,
+      "loss": 0.6448,
+      "step": 2826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9533250497383172e-05,
+      "loss": 0.6559,
+      "step": 2827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9532921662572988e-05,
+      "loss": 0.6198,
+      "step": 2828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9532592714737956e-05,
+      "loss": 0.6704,
+      "step": 2829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9532263653881983e-05,
+      "loss": 0.6756,
+      "step": 2830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9531934480008968e-05,
+      "loss": 0.724,
+      "step": 2831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9531605193122805e-05,
+      "loss": 0.6679,
+      "step": 2832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9531275793227407e-05,
+      "loss": 0.6936,
+      "step": 2833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953094628032668e-05,
+      "loss": 0.573,
+      "step": 2834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953061665442453e-05,
+      "loss": 0.6127,
+      "step": 2835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9530286915524858e-05,
+      "loss": 0.6848,
+      "step": 2836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9529957063631578e-05,
+      "loss": 0.6984,
+      "step": 2837
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9529627098748606e-05,
+      "loss": 0.6324,
+      "step": 2838
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9529297020879848e-05,
+      "loss": 0.6237,
+      "step": 2839
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9528966830029217e-05,
+      "loss": 0.6467,
+      "step": 2840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9528636526200634e-05,
+      "loss": 0.6685,
+      "step": 2841
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9528306109398012e-05,
+      "loss": 0.6563,
+      "step": 2842
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9527975579625264e-05,
+      "loss": 0.731,
+      "step": 2843
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952764493688631e-05,
+      "loss": 0.6585,
+      "step": 2844
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952731418118508e-05,
+      "loss": 0.696,
+      "step": 2845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9526983312525486e-05,
+      "loss": 0.7009,
+      "step": 2846
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9526652330911452e-05,
+      "loss": 0.6506,
+      "step": 2847
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9526321236346902e-05,
+      "loss": 0.6041,
+      "step": 2848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9525990028835765e-05,
+      "loss": 0.6767,
+      "step": 2849
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9525658708381962e-05,
+      "loss": 0.6112,
+      "step": 2850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952532727498943e-05,
+      "loss": 0.62,
+      "step": 2851
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9524995728662092e-05,
+      "loss": 0.6768,
+      "step": 2852
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9524664069403882e-05,
+      "loss": 0.6304,
+      "step": 2853
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9524332297218724e-05,
+      "loss": 0.682,
+      "step": 2854
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9524000412110563e-05,
+      "loss": 0.6622,
+      "step": 2855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952366841408333e-05,
+      "loss": 0.6274,
+      "step": 2856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952333630314096e-05,
+      "loss": 0.6717,
+      "step": 2857
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952300407928739e-05,
+      "loss": 0.6291,
+      "step": 2858
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952267174252656e-05,
+      "loss": 0.6333,
+      "step": 2859
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9522339292862412e-05,
+      "loss": 0.6463,
+      "step": 2860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9522006730298885e-05,
+      "loss": 0.6926,
+      "step": 2861
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952167405483992e-05,
+      "loss": 0.6034,
+      "step": 2862
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9521341266489465e-05,
+      "loss": 0.6216,
+      "step": 2863
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952100836525147e-05,
+      "loss": 0.6206,
+      "step": 2864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952067535112987e-05,
+      "loss": 0.6007,
+      "step": 2865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9520342224128623e-05,
+      "loss": 0.6337,
+      "step": 2866
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9520008984251676e-05,
+      "loss": 0.632,
+      "step": 2867
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9519675631502978e-05,
+      "loss": 0.6735,
+      "step": 2868
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9519342165886486e-05,
+      "loss": 0.6077,
+      "step": 2869
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.951900858740615e-05,
+      "loss": 0.591,
+      "step": 2870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9518674896065925e-05,
+      "loss": 0.5599,
+      "step": 2871
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9518341091869767e-05,
+      "loss": 0.6671,
+      "step": 2872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9518007174821635e-05,
+      "loss": 0.6408,
+      "step": 2873
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9517673144925485e-05,
+      "loss": 0.6097,
+      "step": 2874
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9517339002185284e-05,
+      "loss": 0.6445,
+      "step": 2875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9517004746604988e-05,
+      "loss": 0.6357,
+      "step": 2876
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.951667037818856e-05,
+      "loss": 0.6399,
+      "step": 2877
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.951633589693997e-05,
+      "loss": 0.6776,
+      "step": 2878
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9516001302863173e-05,
+      "loss": 0.6565,
+      "step": 2879
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9515666595962147e-05,
+      "loss": 0.6183,
+      "step": 2880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9515331776240858e-05,
+      "loss": 0.6782,
+      "step": 2881
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9514996843703272e-05,
+      "loss": 0.6518,
+      "step": 2882
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9514661798353362e-05,
+      "loss": 0.6313,
+      "step": 2883
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9514326640195103e-05,
+      "loss": 0.6545,
+      "step": 2884
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.951399136923246e-05,
+      "loss": 0.6188,
+      "step": 2885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.951365598546942e-05,
+      "loss": 0.6785,
+      "step": 2886
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9513320488909948e-05,
+      "loss": 0.6182,
+      "step": 2887
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.951298487955803e-05,
+      "loss": 0.6473,
+      "step": 2888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9512649157417645e-05,
+      "loss": 0.6272,
+      "step": 2889
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9512313322492767e-05,
+      "loss": 0.5942,
+      "step": 2890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9511977374787382e-05,
+      "loss": 0.6619,
+      "step": 2891
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9511641314305473e-05,
+      "loss": 0.6752,
+      "step": 2892
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.951130514105102e-05,
+      "loss": 0.6497,
+      "step": 2893
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.951096885502802e-05,
+      "loss": 0.6138,
+      "step": 2894
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9510632456240446e-05,
+      "loss": 0.6298,
+      "step": 2895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95102959446923e-05,
+      "loss": 0.6372,
+      "step": 2896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950995932038756e-05,
+      "loss": 0.6444,
+      "step": 2897
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9509622583330222e-05,
+      "loss": 0.6476,
+      "step": 2898
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950928573352428e-05,
+      "loss": 0.648,
+      "step": 2899
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950894877097373e-05,
+      "loss": 0.6717,
+      "step": 2900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9508611695682557e-05,
+      "loss": 0.6462,
+      "step": 2901
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9508274507654767e-05,
+      "loss": 0.6252,
+      "step": 2902
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9507937206894355e-05,
+      "loss": 0.6384,
+      "step": 2903
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950759979340532e-05,
+      "loss": 0.6555,
+      "step": 2904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9507262267191662e-05,
+      "loss": 0.6148,
+      "step": 2905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950692462825738e-05,
+      "loss": 0.6039,
+      "step": 2906
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9506586876606485e-05,
+      "loss": 0.5911,
+      "step": 2907
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9506249012242974e-05,
+      "loss": 0.6076,
+      "step": 2908
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9505911035170854e-05,
+      "loss": 0.6519,
+      "step": 2909
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950557294539414e-05,
+      "loss": 0.6645,
+      "step": 2910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9505234742916827e-05,
+      "loss": 0.6329,
+      "step": 2911
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9504896427742934e-05,
+      "loss": 0.6529,
+      "step": 2912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9504557999876473e-05,
+      "loss": 0.682,
+      "step": 2913
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9504219459321448e-05,
+      "loss": 0.6165,
+      "step": 2914
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9503880806081882e-05,
+      "loss": 0.6951,
+      "step": 2915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9503542040161788e-05,
+      "loss": 0.6626,
+      "step": 2916
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950320316156518e-05,
+      "loss": 0.642,
+      "step": 2917
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9502864170296073e-05,
+      "loss": 0.6248,
+      "step": 2918
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9502525066358496e-05,
+      "loss": 0.6212,
+      "step": 2919
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950218584975646e-05,
+      "loss": 0.6539,
+      "step": 2920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9501846520493993e-05,
+      "loss": 0.6314,
+      "step": 2921
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.950150707857511e-05,
+      "loss": 0.5868,
+      "step": 2922
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9501167524003846e-05,
+      "loss": 0.6768,
+      "step": 2923
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9500827856784223e-05,
+      "loss": 0.6489,
+      "step": 2924
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9500488076920266e-05,
+      "loss": 0.6414,
+      "step": 2925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9500148184416004e-05,
+      "loss": 0.6657,
+      "step": 2926
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9499808179275466e-05,
+      "loss": 0.5883,
+      "step": 2927
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9499468061502687e-05,
+      "loss": 0.6605,
+      "step": 2928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9499127831101696e-05,
+      "loss": 0.6914,
+      "step": 2929
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9498787488076527e-05,
+      "loss": 0.649,
+      "step": 2930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.949844703243122e-05,
+      "loss": 0.6413,
+      "step": 2931
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9498106464169805e-05,
+      "loss": 0.6226,
+      "step": 2932
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9497765783296323e-05,
+      "loss": 0.6956,
+      "step": 2933
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9497424989814814e-05,
+      "loss": 0.6488,
+      "step": 2934
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9497084083729315e-05,
+      "loss": 0.6563,
+      "step": 2935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.949674306504387e-05,
+      "loss": 0.6346,
+      "step": 2936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9496401933762526e-05,
+      "loss": 0.6351,
+      "step": 2937
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9496060689889323e-05,
+      "loss": 0.6696,
+      "step": 2938
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9495719333428308e-05,
+      "loss": 0.6481,
+      "step": 2939
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9495377864383526e-05,
+      "loss": 0.637,
+      "step": 2940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.949503628275903e-05,
+      "loss": 0.6508,
+      "step": 2941
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9494694588558863e-05,
+      "loss": 0.6682,
+      "step": 2942
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.949435278178709e-05,
+      "loss": 0.65,
+      "step": 2943
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9494010862447745e-05,
+      "loss": 0.6324,
+      "step": 2944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9493668830544897e-05,
+      "loss": 0.6406,
+      "step": 2945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9493326686082592e-05,
+      "loss": 0.6587,
+      "step": 2946
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9492984429064893e-05,
+      "loss": 0.6883,
+      "step": 2947
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9492642059495854e-05,
+      "loss": 0.6353,
+      "step": 2948
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9492299577379534e-05,
+      "loss": 0.6333,
+      "step": 2949
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9491956982719994e-05,
+      "loss": 0.6484,
+      "step": 2950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.94916142755213e-05,
+      "loss": 0.5977,
+      "step": 2951
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.949127145578751e-05,
+      "loss": 0.679,
+      "step": 2952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9490928523522692e-05,
+      "loss": 0.7095,
+      "step": 2953
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9490585478730904e-05,
+      "loss": 0.6252,
+      "step": 2954
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9490242321416227e-05,
+      "loss": 0.6887,
+      "step": 2955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948989905158272e-05,
+      "loss": 0.6389,
+      "step": 2956
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948955566923445e-05,
+      "loss": 0.6223,
+      "step": 2957
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.94892121743755e-05,
+      "loss": 0.6817,
+      "step": 2958
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9488868567009935e-05,
+      "loss": 0.6624,
+      "step": 2959
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9488524847141825e-05,
+      "loss": 0.6844,
+      "step": 2960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9488181014775255e-05,
+      "loss": 0.6694,
+      "step": 2961
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9487837069914294e-05,
+      "loss": 0.6936,
+      "step": 2962
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948749301256302e-05,
+      "loss": 0.6617,
+      "step": 2963
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9487148842725517e-05,
+      "loss": 0.6485,
+      "step": 2964
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9486804560405863e-05,
+      "loss": 0.6164,
+      "step": 2965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948646016560814e-05,
+      "loss": 0.6194,
+      "step": 2966
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9486115658336433e-05,
+      "loss": 0.7146,
+      "step": 2967
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9485771038594822e-05,
+      "loss": 0.5839,
+      "step": 2968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.94854263063874e-05,
+      "loss": 0.6575,
+      "step": 2969
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9485081461718245e-05,
+      "loss": 0.5793,
+      "step": 2970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9484736504591454e-05,
+      "loss": 0.6651,
+      "step": 2971
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948439143501111e-05,
+      "loss": 0.7193,
+      "step": 2972
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948404625298131e-05,
+      "loss": 0.6561,
+      "step": 2973
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9483700958506143e-05,
+      "loss": 0.6226,
+      "step": 2974
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9483355551589706e-05,
+      "loss": 0.645,
+      "step": 2975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9483010032236092e-05,
+      "loss": 0.6416,
+      "step": 2976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.94826644004494e-05,
+      "loss": 0.6335,
+      "step": 2977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948231865623372e-05,
+      "loss": 0.628,
+      "step": 2978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9481972799593164e-05,
+      "loss": 0.6177,
+      "step": 2979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948162683053182e-05,
+      "loss": 0.6396,
+      "step": 2980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9481280749053798e-05,
+      "loss": 0.6138,
+      "step": 2981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.94809345551632e-05,
+      "loss": 0.6541,
+      "step": 2982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9480588248864126e-05,
+      "loss": 0.6375,
+      "step": 2983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.948024183016069e-05,
+      "loss": 0.6352,
+      "step": 2984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9479895299056992e-05,
+      "loss": 0.6467,
+      "step": 2985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9479548655557145e-05,
+      "loss": 0.6427,
+      "step": 2986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9479201899665255e-05,
+      "loss": 0.7015,
+      "step": 2987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9478855031385435e-05,
+      "loss": 0.6062,
+      "step": 2988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.94785080507218e-05,
+      "loss": 0.6458,
+      "step": 2989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.947816095767846e-05,
+      "loss": 0.6326,
+      "step": 2990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9477813752259533e-05,
+      "loss": 0.6358,
+      "step": 2991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.947746643446913e-05,
+      "loss": 0.6853,
+      "step": 2992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9477119004311376e-05,
+      "loss": 0.674,
+      "step": 2993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.947677146179039e-05,
+      "loss": 0.6589,
+      "step": 2994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9476423806910286e-05,
+      "loss": 0.6283,
+      "step": 2995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9476076039675195e-05,
+      "loss": 0.6294,
+      "step": 2996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9475728160089234e-05,
+      "loss": 0.6335,
+      "step": 2997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9475380168156527e-05,
+      "loss": 0.6314,
+      "step": 2998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9475032063881202e-05,
+      "loss": 0.6668,
+      "step": 2999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9474683847267384e-05,
+      "loss": 0.5716,
+      "step": 3000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9474335518319206e-05,
+      "loss": 0.5958,
+      "step": 3001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9473987077040795e-05,
+      "loss": 0.6297,
+      "step": 3002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9473638523436285e-05,
+      "loss": 0.6791,
+      "step": 3003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.94732898575098e-05,
+      "loss": 0.6715,
+      "step": 3004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9472941079265486e-05,
+      "loss": 0.6935,
+      "step": 3005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9472592188707467e-05,
+      "loss": 0.7065,
+      "step": 3006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.947224318583989e-05,
+      "loss": 0.667,
+      "step": 3007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9471894070666884e-05,
+      "loss": 0.6577,
+      "step": 3008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9471544843192593e-05,
+      "loss": 0.6942,
+      "step": 3009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9471195503421158e-05,
+      "loss": 0.6628,
+      "step": 3010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9470846051356715e-05,
+      "loss": 0.6345,
+      "step": 3011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9470496487003414e-05,
+      "loss": 0.6421,
+      "step": 3012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9470146810365397e-05,
+      "loss": 0.6375,
+      "step": 3013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.946979702144681e-05,
+      "loss": 0.6728,
+      "step": 3014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.94694471202518e-05,
+      "loss": 0.698,
+      "step": 3015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9469097106784513e-05,
+      "loss": 0.6699,
+      "step": 3016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9468746981049102e-05,
+      "loss": 0.6403,
+      "step": 3017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.946839674304972e-05,
+      "loss": 0.6336,
+      "step": 3018
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9468046392790516e-05,
+      "loss": 0.6554,
+      "step": 3019
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9467695930275642e-05,
+      "loss": 0.6653,
+      "step": 3020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.946734535550926e-05,
+      "loss": 0.6416,
+      "step": 3021
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.946699466849552e-05,
+      "loss": 0.593,
+      "step": 3022
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9466643869238585e-05,
+      "loss": 0.6224,
+      "step": 3023
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9466292957742608e-05,
+      "loss": 0.5788,
+      "step": 3024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9465941934011757e-05,
+      "loss": 0.6412,
+      "step": 3025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.946559079805019e-05,
+      "loss": 0.5931,
+      "step": 3026
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9465239549862068e-05,
+      "loss": 0.6723,
+      "step": 3027
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9464888189451558e-05,
+      "loss": 0.6886,
+      "step": 3028
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9464536716822823e-05,
+      "loss": 0.5871,
+      "step": 3029
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9464185131980034e-05,
+      "loss": 0.6522,
+      "step": 3030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9463833434927358e-05,
+      "loss": 0.6113,
+      "step": 3031
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9463481625668966e-05,
+      "loss": 0.6701,
+      "step": 3032
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9463129704209026e-05,
+      "loss": 0.6714,
+      "step": 3033
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9462777670551715e-05,
+      "loss": 0.6528,
+      "step": 3034
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9462425524701204e-05,
+      "loss": 0.6543,
+      "step": 3035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9462073266661666e-05,
+      "loss": 0.5957,
+      "step": 3036
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9461720896437284e-05,
+      "loss": 0.638,
+      "step": 3037
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9461368414032228e-05,
+      "loss": 0.6218,
+      "step": 3038
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9461015819450678e-05,
+      "loss": 0.6295,
+      "step": 3039
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9460663112696823e-05,
+      "loss": 0.6267,
+      "step": 3040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.946031029377484e-05,
+      "loss": 0.6694,
+      "step": 3041
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9459957362688902e-05,
+      "loss": 0.6847,
+      "step": 3042
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9459604319443213e-05,
+      "loss": 0.6347,
+      "step": 3043
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9459251164041945e-05,
+      "loss": 0.6352,
+      "step": 3044
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9458897896489283e-05,
+      "loss": 0.5886,
+      "step": 3045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9458544516789425e-05,
+      "loss": 0.6914,
+      "step": 3046
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945819102494656e-05,
+      "loss": 0.6448,
+      "step": 3047
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945783742096487e-05,
+      "loss": 0.6415,
+      "step": 3048
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9457483704848555e-05,
+      "loss": 0.6407,
+      "step": 3049
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9457129876601807e-05,
+      "loss": 0.6142,
+      "step": 3050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9456775936228825e-05,
+      "loss": 0.5882,
+      "step": 3051
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9456421883733797e-05,
+      "loss": 0.6539,
+      "step": 3052
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9456067719120924e-05,
+      "loss": 0.6023,
+      "step": 3053
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945571344239441e-05,
+      "loss": 0.6728,
+      "step": 3054
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945535905355845e-05,
+      "loss": 0.6222,
+      "step": 3055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9455004552617248e-05,
+      "loss": 0.6741,
+      "step": 3056
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9454649939575002e-05,
+      "loss": 0.661,
+      "step": 3057
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945429521443593e-05,
+      "loss": 0.6496,
+      "step": 3058
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945394037720422e-05,
+      "loss": 0.6755,
+      "step": 3059
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945358542788409e-05,
+      "loss": 0.6241,
+      "step": 3060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9453230366479746e-05,
+      "loss": 0.6432,
+      "step": 3061
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94528751929954e-05,
+      "loss": 0.6157,
+      "step": 3062
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9452519907435256e-05,
+      "loss": 0.6187,
+      "step": 3063
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9452164509803535e-05,
+      "loss": 0.6488,
+      "step": 3064
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9451809000104446e-05,
+      "loss": 0.6612,
+      "step": 3065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94514533783422e-05,
+      "loss": 0.6838,
+      "step": 3066
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9451097644521023e-05,
+      "loss": 0.634,
+      "step": 3067
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945074179864513e-05,
+      "loss": 0.5979,
+      "step": 3068
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945038584071873e-05,
+      "loss": 0.6853,
+      "step": 3069
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9450029770746058e-05,
+      "loss": 0.651,
+      "step": 3070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9449673588731325e-05,
+      "loss": 0.6535,
+      "step": 3071
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944931729467876e-05,
+      "loss": 0.6404,
+      "step": 3072
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9448960888592582e-05,
+      "loss": 0.6389,
+      "step": 3073
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9448604370477022e-05,
+      "loss": 0.6477,
+      "step": 3074
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9448247740336303e-05,
+      "loss": 0.6357,
+      "step": 3075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9447890998174657e-05,
+      "loss": 0.6639,
+      "step": 3076
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9447534143996312e-05,
+      "loss": 0.5919,
+      "step": 3077
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9447177177805497e-05,
+      "loss": 0.617,
+      "step": 3078
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9446820099606446e-05,
+      "loss": 0.6443,
+      "step": 3079
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9446462909403392e-05,
+      "loss": 0.6581,
+      "step": 3080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9446105607200573e-05,
+      "loss": 0.6251,
+      "step": 3081
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9445748193002217e-05,
+      "loss": 0.6437,
+      "step": 3082
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9445390666812574e-05,
+      "loss": 0.6047,
+      "step": 3083
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944503302863587e-05,
+      "loss": 0.6008,
+      "step": 3084
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9444675278476354e-05,
+      "loss": 0.6747,
+      "step": 3085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944431741633827e-05,
+      "loss": 0.6692,
+      "step": 3086
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944395944222585e-05,
+      "loss": 0.6599,
+      "step": 3087
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944360135614334e-05,
+      "loss": 0.6431,
+      "step": 3088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9443243158095e-05,
+      "loss": 0.6313,
+      "step": 3089
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944288484808506e-05,
+      "loss": 0.6328,
+      "step": 3090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944252642611778e-05,
+      "loss": 0.6597,
+      "step": 3091
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94421678921974e-05,
+      "loss": 0.7125,
+      "step": 3092
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9441809246328174e-05,
+      "loss": 0.6483,
+      "step": 3093
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9441450488514357e-05,
+      "loss": 0.6538,
+      "step": 3094
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9441091618760202e-05,
+      "loss": 0.6629,
+      "step": 3095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9440732637069962e-05,
+      "loss": 0.651,
+      "step": 3096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9440373543447896e-05,
+      "loss": 0.6338,
+      "step": 3097
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9440014337898256e-05,
+      "loss": 0.6601,
+      "step": 3098
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9439655020425305e-05,
+      "loss": 0.6394,
+      "step": 3099
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9439295591033304e-05,
+      "loss": 0.6131,
+      "step": 3100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9438936049726512e-05,
+      "loss": 0.6369,
+      "step": 3101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9438576396509196e-05,
+      "loss": 0.6314,
+      "step": 3102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9438216631385613e-05,
+      "loss": 0.6551,
+      "step": 3103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9437856754360033e-05,
+      "loss": 0.6567,
+      "step": 3104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9437496765436722e-05,
+      "loss": 0.6271,
+      "step": 3105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9437136664619953e-05,
+      "loss": 0.5671,
+      "step": 3106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9436776451913986e-05,
+      "loss": 0.6519,
+      "step": 3107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9436416127323095e-05,
+      "loss": 0.5807,
+      "step": 3108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943605569085156e-05,
+      "loss": 0.6495,
+      "step": 3109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9435695142503645e-05,
+      "loss": 0.6915,
+      "step": 3110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9435334482283626e-05,
+      "loss": 0.6565,
+      "step": 3111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943497371019578e-05,
+      "loss": 0.6466,
+      "step": 3112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943461282624439e-05,
+      "loss": 0.6629,
+      "step": 3113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9434251830433728e-05,
+      "loss": 0.6518,
+      "step": 3114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9433890722768077e-05,
+      "loss": 0.6832,
+      "step": 3115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9433529503251716e-05,
+      "loss": 0.6896,
+      "step": 3116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943316817188893e-05,
+      "loss": 0.6675,
+      "step": 3117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9432806728684e-05,
+      "loss": 0.6286,
+      "step": 3118
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9432445173641216e-05,
+      "loss": 0.6674,
+      "step": 3119
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9432083506764864e-05,
+      "loss": 0.6316,
+      "step": 3120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9431721728059228e-05,
+      "loss": 0.6473,
+      "step": 3121
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94313598375286e-05,
+      "loss": 0.6282,
+      "step": 3122
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943099783517727e-05,
+      "loss": 0.6587,
+      "step": 3123
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943063572100953e-05,
+      "loss": 0.645,
+      "step": 3124
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9430273495029675e-05,
+      "loss": 0.6508,
+      "step": 3125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9429911157242e-05,
+      "loss": 0.6414,
+      "step": 3126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9429548707650795e-05,
+      "loss": 0.6116,
+      "step": 3127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9429186146260366e-05,
+      "loss": 0.6113,
+      "step": 3128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9428823473075008e-05,
+      "loss": 0.638,
+      "step": 3129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9428460688099015e-05,
+      "loss": 0.6345,
+      "step": 3130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94280977913367e-05,
+      "loss": 0.5896,
+      "step": 3131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9427734782792352e-05,
+      "loss": 0.6514,
+      "step": 3132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9427371662470286e-05,
+      "loss": 0.7311,
+      "step": 3133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9427008430374802e-05,
+      "loss": 0.6387,
+      "step": 3134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9426645086510207e-05,
+      "loss": 0.6737,
+      "step": 3135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.942628163088081e-05,
+      "loss": 0.6173,
+      "step": 3136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9425918063490922e-05,
+      "loss": 0.6257,
+      "step": 3137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.942555438434485e-05,
+      "loss": 0.6192,
+      "step": 3138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9425190593446906e-05,
+      "loss": 0.6783,
+      "step": 3139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.942482669080141e-05,
+      "loss": 0.6251,
+      "step": 3140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9424462676412664e-05,
+      "loss": 0.6449,
+      "step": 3141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9424098550284997e-05,
+      "loss": 0.6743,
+      "step": 3142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9423734312422717e-05,
+      "loss": 0.6623,
+      "step": 3143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9423369962830147e-05,
+      "loss": 0.6346,
+      "step": 3144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9423005501511603e-05,
+      "loss": 0.6562,
+      "step": 3145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.942264092847141e-05,
+      "loss": 0.6426,
+      "step": 3146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.942227624371389e-05,
+      "loss": 0.6003,
+      "step": 3147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9421911447243365e-05,
+      "loss": 0.6501,
+      "step": 3148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.942154653906416e-05,
+      "loss": 0.6335,
+      "step": 3149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9421181519180606e-05,
+      "loss": 0.6445,
+      "step": 3150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9420816387597024e-05,
+      "loss": 0.6382,
+      "step": 3151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9420451144317745e-05,
+      "loss": 0.6755,
+      "step": 3152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9420085789347106e-05,
+      "loss": 0.6448,
+      "step": 3153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.941972032268943e-05,
+      "loss": 0.619,
+      "step": 3154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9419354744349054e-05,
+      "loss": 0.6582,
+      "step": 3155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9418989054330314e-05,
+      "loss": 0.6275,
+      "step": 3156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.941862325263754e-05,
+      "loss": 0.6265,
+      "step": 3157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9418257339275075e-05,
+      "loss": 0.5854,
+      "step": 3158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9417891314247256e-05,
+      "loss": 0.698,
+      "step": 3159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9417525177558418e-05,
+      "loss": 0.6824,
+      "step": 3160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9417158929212907e-05,
+      "loss": 0.6686,
+      "step": 3161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9416792569215067e-05,
+      "loss": 0.6721,
+      "step": 3162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9416426097569238e-05,
+      "loss": 0.5964,
+      "step": 3163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9416059514279764e-05,
+      "loss": 0.6744,
+      "step": 3164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.941569281935099e-05,
+      "loss": 0.6743,
+      "step": 3165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9415326012787275e-05,
+      "loss": 0.678,
+      "step": 3166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9414959094592954e-05,
+      "loss": 0.6114,
+      "step": 3167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9414592064772386e-05,
+      "loss": 0.6124,
+      "step": 3168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9414224923329918e-05,
+      "loss": 0.6545,
+      "step": 3169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9413857670269903e-05,
+      "loss": 0.6617,
+      "step": 3170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94134903055967e-05,
+      "loss": 0.6378,
+      "step": 3171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9413122829314657e-05,
+      "loss": 0.6551,
+      "step": 3172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.941275524142814e-05,
+      "loss": 0.6292,
+      "step": 3173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9412387541941497e-05,
+      "loss": 0.6375,
+      "step": 3174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9412019730859097e-05,
+      "loss": 0.6796,
+      "step": 3175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9411651808185294e-05,
+      "loss": 0.6432,
+      "step": 3176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9411283773924455e-05,
+      "loss": 0.5815,
+      "step": 3177
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9410915628080943e-05,
+      "loss": 0.5916,
+      "step": 3178
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.941054737065912e-05,
+      "loss": 0.6141,
+      "step": 3179
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.941017900166335e-05,
+      "loss": 0.5911,
+      "step": 3180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.940981052109801e-05,
+      "loss": 0.6455,
+      "step": 3181
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9409441928967457e-05,
+      "loss": 0.6442,
+      "step": 3182
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9409073225276073e-05,
+      "loss": 0.6025,
+      "step": 3183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.940870441002822e-05,
+      "loss": 0.6165,
+      "step": 3184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9408335483228272e-05,
+      "loss": 0.6409,
+      "step": 3185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9407966444880608e-05,
+      "loss": 0.6806,
+      "step": 3186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94075972949896e-05,
+      "loss": 0.5893,
+      "step": 3187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9407228033559628e-05,
+      "loss": 0.5674,
+      "step": 3188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9406858660595064e-05,
+      "loss": 0.6596,
+      "step": 3189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9406489176100292e-05,
+      "loss": 0.6762,
+      "step": 3190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.940611958007969e-05,
+      "loss": 0.6767,
+      "step": 3191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9405749872537645e-05,
+      "loss": 0.6326,
+      "step": 3192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9405380053478534e-05,
+      "loss": 0.5791,
+      "step": 3193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9405010122906746e-05,
+      "loss": 0.5877,
+      "step": 3194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9404640080826665e-05,
+      "loss": 0.6365,
+      "step": 3195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.940426992724268e-05,
+      "loss": 0.6996,
+      "step": 3196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9403899662159177e-05,
+      "loss": 0.6283,
+      "step": 3197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9403529285580547e-05,
+      "loss": 0.6357,
+      "step": 3198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9403158797511182e-05,
+      "loss": 0.6243,
+      "step": 3199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9402788197955476e-05,
+      "loss": 0.62,
+      "step": 3200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.940241748691782e-05,
+      "loss": 0.6419,
+      "step": 3201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9402046664402612e-05,
+      "loss": 0.6246,
+      "step": 3202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9401675730414246e-05,
+      "loss": 0.59,
+      "step": 3203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.940130468495712e-05,
+      "loss": 0.6738,
+      "step": 3204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.940093352803564e-05,
+      "loss": 0.6733,
+      "step": 3205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9400562259654194e-05,
+      "loss": 0.6233,
+      "step": 3206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9400190879817192e-05,
+      "loss": 0.5503,
+      "step": 3207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9399819388529038e-05,
+      "loss": 0.6588,
+      "step": 3208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9399447785794133e-05,
+      "loss": 0.6376,
+      "step": 3209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9399076071616886e-05,
+      "loss": 0.6789,
+      "step": 3210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9398704246001697e-05,
+      "loss": 0.6527,
+      "step": 3211
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.939833230895299e-05,
+      "loss": 0.6706,
+      "step": 3212
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9397960260475154e-05,
+      "loss": 0.6729,
+      "step": 3213
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9397588100572612e-05,
+      "loss": 0.6613,
+      "step": 3214
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9397215829249777e-05,
+      "loss": 0.6424,
+      "step": 3215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9396843446511065e-05,
+      "loss": 0.6657,
+      "step": 3216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9396470952360882e-05,
+      "loss": 0.644,
+      "step": 3217
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.939609834680365e-05,
+      "loss": 0.5904,
+      "step": 3218
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9395725629843786e-05,
+      "loss": 0.6298,
+      "step": 3219
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9395352801485708e-05,
+      "loss": 0.6621,
+      "step": 3220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9394979861733843e-05,
+      "loss": 0.665,
+      "step": 3221
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9394606810592603e-05,
+      "loss": 0.6663,
+      "step": 3222
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.939423364806641e-05,
+      "loss": 0.6324,
+      "step": 3223
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9393860374159703e-05,
+      "loss": 0.7051,
+      "step": 3224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.939348698887689e-05,
+      "loss": 0.6807,
+      "step": 3225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.939311349222241e-05,
+      "loss": 0.6581,
+      "step": 3226
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9392739884200687e-05,
+      "loss": 0.627,
+      "step": 3227
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.939236616481615e-05,
+      "loss": 0.6897,
+      "step": 3228
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.939199233407323e-05,
+      "loss": 0.6474,
+      "step": 3229
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9391618391976363e-05,
+      "loss": 0.645,
+      "step": 3230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9391244338529977e-05,
+      "loss": 0.6322,
+      "step": 3231
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9390870173738508e-05,
+      "loss": 0.617,
+      "step": 3232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9390495897606398e-05,
+      "loss": 0.6526,
+      "step": 3233
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9390121510138074e-05,
+      "loss": 0.6691,
+      "step": 3234
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9389747011337987e-05,
+      "loss": 0.6229,
+      "step": 3235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9389372401210567e-05,
+      "loss": 0.6199,
+      "step": 3236
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9388997679760263e-05,
+      "loss": 0.5942,
+      "step": 3237
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938862284699151e-05,
+      "loss": 0.6264,
+      "step": 3238
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938824790290876e-05,
+      "loss": 0.6392,
+      "step": 3239
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9387872847516452e-05,
+      "loss": 0.5896,
+      "step": 3240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9387497680819037e-05,
+      "loss": 0.6363,
+      "step": 3241
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9387122402820965e-05,
+      "loss": 0.7295,
+      "step": 3242
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938674701352668e-05,
+      "loss": 0.6591,
+      "step": 3243
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938637151294063e-05,
+      "loss": 0.637,
+      "step": 3244
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9385995901067276e-05,
+      "loss": 0.6126,
+      "step": 3245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938562017791107e-05,
+      "loss": 0.6596,
+      "step": 3246
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938524434347646e-05,
+      "loss": 0.6384,
+      "step": 3247
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9384868397767906e-05,
+      "loss": 0.6283,
+      "step": 3248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9384492340789868e-05,
+      "loss": 0.6618,
+      "step": 3249
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9384116172546798e-05,
+      "loss": 0.6021,
+      "step": 3250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9383739893043162e-05,
+      "loss": 0.6354,
+      "step": 3251
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938336350228342e-05,
+      "loss": 0.6577,
+      "step": 3252
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9382987000272035e-05,
+      "loss": 0.5928,
+      "step": 3253
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9382610387013464e-05,
+      "loss": 0.613,
+      "step": 3254
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9382233662512182e-05,
+      "loss": 0.5817,
+      "step": 3255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938185682677265e-05,
+      "loss": 0.6563,
+      "step": 3256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938147987979934e-05,
+      "loss": 0.6562,
+      "step": 3257
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9381102821596712e-05,
+      "loss": 0.6055,
+      "step": 3258
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.938072565216925e-05,
+      "loss": 0.606,
+      "step": 3259
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9380348371521417e-05,
+      "loss": 0.5991,
+      "step": 3260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9379970979657682e-05,
+      "loss": 0.5861,
+      "step": 3261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9379593476582533e-05,
+      "loss": 0.5849,
+      "step": 3262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9379215862300435e-05,
+      "loss": 0.6591,
+      "step": 3263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9378838136815864e-05,
+      "loss": 0.6198,
+      "step": 3264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937846030013331e-05,
+      "loss": 0.579,
+      "step": 3265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9378082352257243e-05,
+      "loss": 0.6293,
+      "step": 3266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9377704293192144e-05,
+      "loss": 0.64,
+      "step": 3267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9377326122942498e-05,
+      "loss": 0.6227,
+      "step": 3268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937694784151279e-05,
+      "loss": 0.6723,
+      "step": 3269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9376569448907505e-05,
+      "loss": 0.5972,
+      "step": 3270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9376190945131125e-05,
+      "loss": 0.6178,
+      "step": 3271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937581233018814e-05,
+      "loss": 0.5915,
+      "step": 3272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9375433604083042e-05,
+      "loss": 0.633,
+      "step": 3273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9375054766820317e-05,
+      "loss": 0.6703,
+      "step": 3274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9374675818404454e-05,
+      "loss": 0.5989,
+      "step": 3275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937429675883996e-05,
+      "loss": 0.6374,
+      "step": 3276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937391758813131e-05,
+      "loss": 0.6051,
+      "step": 3277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937353830628301e-05,
+      "loss": 0.6732,
+      "step": 3278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937315891329956e-05,
+      "loss": 0.5908,
+      "step": 3279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9372779409185453e-05,
+      "loss": 0.6661,
+      "step": 3280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9372399793945193e-05,
+      "loss": 0.6091,
+      "step": 3281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937202006758327e-05,
+      "loss": 0.6655,
+      "step": 3282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9371640230104197e-05,
+      "loss": 0.664,
+      "step": 3283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9371260281512473e-05,
+      "loss": 0.6435,
+      "step": 3284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9370880221812606e-05,
+      "loss": 0.6417,
+      "step": 3285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9370500051009097e-05,
+      "loss": 0.6282,
+      "step": 3286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.937011976910646e-05,
+      "loss": 0.6341,
+      "step": 3287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9369739376109197e-05,
+      "loss": 0.6755,
+      "step": 3288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9369358872021818e-05,
+      "loss": 0.6856,
+      "step": 3289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9368978256848843e-05,
+      "loss": 0.6295,
+      "step": 3290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9368597530594775e-05,
+      "loss": 0.6399,
+      "step": 3291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9368216693264133e-05,
+      "loss": 0.6023,
+      "step": 3292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936783574486143e-05,
+      "loss": 0.669,
+      "step": 3293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9367454685391188e-05,
+      "loss": 0.6061,
+      "step": 3294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9367073514857917e-05,
+      "loss": 0.6133,
+      "step": 3295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936669223326614e-05,
+      "loss": 0.6474,
+      "step": 3296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936631084062038e-05,
+      "loss": 0.5922,
+      "step": 3297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9365929336925156e-05,
+      "loss": 0.6521,
+      "step": 3298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936554772218499e-05,
+      "loss": 0.6581,
+      "step": 3299
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9365165996404412e-05,
+      "loss": 0.6236,
+      "step": 3300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936478415958794e-05,
+      "loss": 0.62,
+      "step": 3301
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9364402211740106e-05,
+      "loss": 0.638,
+      "step": 3302
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936402015286544e-05,
+      "loss": 0.6888,
+      "step": 3303
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9363637982968467e-05,
+      "loss": 0.6667,
+      "step": 3304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936325570205372e-05,
+      "loss": 0.6613,
+      "step": 3305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9362873310125735e-05,
+      "loss": 0.6506,
+      "step": 3306
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936249080718904e-05,
+      "loss": 0.7132,
+      "step": 3307
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9362108193248176e-05,
+      "loss": 0.6212,
+      "step": 3308
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9361725468307674e-05,
+      "loss": 0.6384,
+      "step": 3309
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9361342632372076e-05,
+      "loss": 0.6507,
+      "step": 3310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9360959685445916e-05,
+      "loss": 0.6545,
+      "step": 3311
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.936057662753374e-05,
+      "loss": 0.6433,
+      "step": 3312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9360193458640086e-05,
+      "loss": 0.6219,
+      "step": 3313
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.93598101787695e-05,
+      "loss": 0.6703,
+      "step": 3314
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9359426787926518e-05,
+      "loss": 0.6763,
+      "step": 3315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9359043286115695e-05,
+      "loss": 0.6481,
+      "step": 3316
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9358659673341577e-05,
+      "loss": 0.682,
+      "step": 3317
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9358275949608707e-05,
+      "loss": 0.6546,
+      "step": 3318
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9357892114921637e-05,
+      "loss": 0.6033,
+      "step": 3319
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9357508169284922e-05,
+      "loss": 0.6829,
+      "step": 3320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9357124112703104e-05,
+      "loss": 0.5963,
+      "step": 3321
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9356739945180746e-05,
+      "loss": 0.6401,
+      "step": 3322
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9356355666722405e-05,
+      "loss": 0.7062,
+      "step": 3323
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9355971277332624e-05,
+      "loss": 0.6604,
+      "step": 3324
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.935558677701597e-05,
+      "loss": 0.6527,
+      "step": 3325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9355202165777002e-05,
+      "loss": 0.6968,
+      "step": 3326
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9354817443620275e-05,
+      "loss": 0.6418,
+      "step": 3327
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9354432610550357e-05,
+      "loss": 0.5997,
+      "step": 3328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9354047666571805e-05,
+      "loss": 0.6172,
+      "step": 3329
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9353662611689183e-05,
+      "loss": 0.634,
+      "step": 3330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.935327744590706e-05,
+      "loss": 0.6374,
+      "step": 3331
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.935289216923e-05,
+      "loss": 0.6322,
+      "step": 3332
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9352506781662578e-05,
+      "loss": 0.6685,
+      "step": 3333
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.935212128320935e-05,
+      "loss": 0.6815,
+      "step": 3334
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9351735673874895e-05,
+      "loss": 0.619,
+      "step": 3335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9351349953663787e-05,
+      "loss": 0.6465,
+      "step": 3336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9350964122580592e-05,
+      "loss": 0.6616,
+      "step": 3337
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9350578180629892e-05,
+      "loss": 0.6217,
+      "step": 3338
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9350192127816253e-05,
+      "loss": 0.6642,
+      "step": 3339
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9349805964144264e-05,
+      "loss": 0.6855,
+      "step": 3340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9349419689618494e-05,
+      "loss": 0.6196,
+      "step": 3341
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934903330424353e-05,
+      "loss": 0.6437,
+      "step": 3342
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9348646808023948e-05,
+      "loss": 0.5858,
+      "step": 3343
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934826020096433e-05,
+      "loss": 0.6265,
+      "step": 3344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9347873483069264e-05,
+      "loss": 0.6286,
+      "step": 3345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9347486654343332e-05,
+      "loss": 0.6339,
+      "step": 3346
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934709971479112e-05,
+      "loss": 0.6554,
+      "step": 3347
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934671266441722e-05,
+      "loss": 0.6675,
+      "step": 3348
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9346325503226214e-05,
+      "loss": 0.6689,
+      "step": 3349
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93459382312227e-05,
+      "loss": 0.6458,
+      "step": 3350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9345550848411263e-05,
+      "loss": 0.6135,
+      "step": 3351
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93451633547965e-05,
+      "loss": 0.6529,
+      "step": 3352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9344775750383e-05,
+      "loss": 0.6056,
+      "step": 3353
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9344388035175365e-05,
+      "loss": 0.597,
+      "step": 3354
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9344000209178188e-05,
+      "loss": 0.6397,
+      "step": 3355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9343612272396067e-05,
+      "loss": 0.6755,
+      "step": 3356
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9343224224833607e-05,
+      "loss": 0.6014,
+      "step": 3357
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9342836066495402e-05,
+      "loss": 0.7102,
+      "step": 3358
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934244779738606e-05,
+      "loss": 0.6745,
+      "step": 3359
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9342059417510177e-05,
+      "loss": 0.6311,
+      "step": 3360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934167092687236e-05,
+      "loss": 0.6236,
+      "step": 3361
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9341282325477223e-05,
+      "loss": 0.632,
+      "step": 3362
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9340893613329365e-05,
+      "loss": 0.6323,
+      "step": 3363
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9340504790433396e-05,
+      "loss": 0.6377,
+      "step": 3364
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9340115856793927e-05,
+      "loss": 0.6846,
+      "step": 3365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933972681241557e-05,
+      "loss": 0.6436,
+      "step": 3366
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9339337657302938e-05,
+      "loss": 0.6397,
+      "step": 3367
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933894839146064e-05,
+      "loss": 0.6935,
+      "step": 3368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9338559014893298e-05,
+      "loss": 0.7015,
+      "step": 3369
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9338169527605523e-05,
+      "loss": 0.617,
+      "step": 3370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9337779929601938e-05,
+      "loss": 0.6717,
+      "step": 3371
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9337390220887156e-05,
+      "loss": 0.6833,
+      "step": 3372
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9337000401465805e-05,
+      "loss": 0.6686,
+      "step": 3373
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9336610471342503e-05,
+      "loss": 0.6675,
+      "step": 3374
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933622043052187e-05,
+      "loss": 0.6235,
+      "step": 3375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9335830279008534e-05,
+      "loss": 0.6326,
+      "step": 3376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933544001680712e-05,
+      "loss": 0.6081,
+      "step": 3377
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933504964392226e-05,
+      "loss": 0.6052,
+      "step": 3378
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933465916035857e-05,
+      "loss": 0.5826,
+      "step": 3379
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9334268566120692e-05,
+      "loss": 0.6249,
+      "step": 3380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9333877861213253e-05,
+      "loss": 0.6743,
+      "step": 3381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933348704564088e-05,
+      "loss": 0.638,
+      "step": 3382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9333096119408215e-05,
+      "loss": 0.6389,
+      "step": 3383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933270508251989e-05,
+      "loss": 0.6544,
+      "step": 3384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9332313934980538e-05,
+      "loss": 0.6731,
+      "step": 3385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93319226767948e-05,
+      "loss": 0.6993,
+      "step": 3386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9331531307967313e-05,
+      "loss": 0.6707,
+      "step": 3387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9331139828502718e-05,
+      "loss": 0.6287,
+      "step": 3388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9330748238405655e-05,
+      "loss": 0.6412,
+      "step": 3389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933035653768077e-05,
+      "loss": 0.6444,
+      "step": 3390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9329964726332705e-05,
+      "loss": 0.6195,
+      "step": 3391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9329572804366104e-05,
+      "loss": 0.6171,
+      "step": 3392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932918077178562e-05,
+      "loss": 0.5822,
+      "step": 3393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932878862859589e-05,
+      "loss": 0.6316,
+      "step": 3394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932839637480157e-05,
+      "loss": 0.6493,
+      "step": 3395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9328004010407317e-05,
+      "loss": 0.6224,
+      "step": 3396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9327611535417774e-05,
+      "loss": 0.6474,
+      "step": 3397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9327218949837596e-05,
+      "loss": 0.6281,
+      "step": 3398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9326826253671437e-05,
+      "loss": 0.6359,
+      "step": 3399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9326433446923957e-05,
+      "loss": 0.6705,
+      "step": 3400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932604052959981e-05,
+      "loss": 0.6422,
+      "step": 3401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9325647501703654e-05,
+      "loss": 0.6011,
+      "step": 3402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9325254363240146e-05,
+      "loss": 0.6095,
+      "step": 3403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932486111421396e-05,
+      "loss": 0.5911,
+      "step": 3404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932446775462974e-05,
+      "loss": 0.6536,
+      "step": 3405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9324074284492162e-05,
+      "loss": 0.6033,
+      "step": 3406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932368070380589e-05,
+      "loss": 0.586,
+      "step": 3407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9323287012575593e-05,
+      "loss": 0.6715,
+      "step": 3408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932289321080593e-05,
+      "loss": 0.652,
+      "step": 3409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932249929850157e-05,
+      "loss": 0.6563,
+      "step": 3410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9322105275667194e-05,
+      "loss": 0.6465,
+      "step": 3411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9321711142307466e-05,
+      "loss": 0.668,
+      "step": 3412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932131689842706e-05,
+      "loss": 0.6796,
+      "step": 3413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.932092254403065e-05,
+      "loss": 0.6049,
+      "step": 3414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9320528079122915e-05,
+      "loss": 0.5967,
+      "step": 3415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9320133503708527e-05,
+      "loss": 0.6737,
+      "step": 3416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9319738817792165e-05,
+      "loss": 0.6311,
+      "step": 3417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9319344021378512e-05,
+      "loss": 0.6811,
+      "step": 3418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9318949114472247e-05,
+      "loss": 0.6369,
+      "step": 3419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.931855409707805e-05,
+      "loss": 0.6605,
+      "step": 3420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9318158969200607e-05,
+      "loss": 0.6505,
+      "step": 3421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93177637308446e-05,
+      "loss": 0.6729,
+      "step": 3422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9317368382014717e-05,
+      "loss": 0.6551,
+      "step": 3423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9316972922715648e-05,
+      "loss": 0.6233,
+      "step": 3424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.931657735295208e-05,
+      "loss": 0.6027,
+      "step": 3425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93161816727287e-05,
+      "loss": 0.6027,
+      "step": 3426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93157858820502e-05,
+      "loss": 0.6537,
+      "step": 3427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9315389980921275e-05,
+      "loss": 0.6645,
+      "step": 3428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9314993969346617e-05,
+      "loss": 0.6535,
+      "step": 3429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.931459784733092e-05,
+      "loss": 0.6463,
+      "step": 3430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9314201614878887e-05,
+      "loss": 0.6478,
+      "step": 3431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.931380527199521e-05,
+      "loss": 0.6534,
+      "step": 3432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.931340881868459e-05,
+      "loss": 0.6188,
+      "step": 3433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9313012254951726e-05,
+      "loss": 0.6882,
+      "step": 3434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9312615580801317e-05,
+      "loss": 0.6225,
+      "step": 3435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9312218796238073e-05,
+      "loss": 0.6897,
+      "step": 3436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9311821901266696e-05,
+      "loss": 0.6053,
+      "step": 3437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.931142489589189e-05,
+      "loss": 0.5912,
+      "step": 3438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9311027780118363e-05,
+      "loss": 0.668,
+      "step": 3439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9310630553950823e-05,
+      "loss": 0.6329,
+      "step": 3440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.931023321739398e-05,
+      "loss": 0.6015,
+      "step": 3441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9309835770452544e-05,
+      "loss": 0.6658,
+      "step": 3442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9309438213131232e-05,
+      "loss": 0.5799,
+      "step": 3443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9309040545434747e-05,
+      "loss": 0.6378,
+      "step": 3444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9308642767367813e-05,
+      "loss": 0.672,
+      "step": 3445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9308244878935145e-05,
+      "loss": 0.6353,
+      "step": 3446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.930784688014146e-05,
+      "loss": 0.6407,
+      "step": 3447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.930744877099147e-05,
+      "loss": 0.6239,
+      "step": 3448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9307050551489904e-05,
+      "loss": 0.6383,
+      "step": 3449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9306652221641483e-05,
+      "loss": 0.6459,
+      "step": 3450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9306253781450923e-05,
+      "loss": 0.6462,
+      "step": 3451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9305855230922955e-05,
+      "loss": 0.6353,
+      "step": 3452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93054565700623e-05,
+      "loss": 0.6448,
+      "step": 3453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9305057798873684e-05,
+      "loss": 0.6127,
+      "step": 3454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.930465891736184e-05,
+      "loss": 0.6316,
+      "step": 3455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9304259925531494e-05,
+      "loss": 0.6279,
+      "step": 3456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9303860823387373e-05,
+      "loss": 0.6389,
+      "step": 3457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9303461610934215e-05,
+      "loss": 0.6613,
+      "step": 3458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.930306228817675e-05,
+      "loss": 0.6455,
+      "step": 3459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9302662855119717e-05,
+      "loss": 0.6429,
+      "step": 3460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9302263311767842e-05,
+      "loss": 0.6442,
+      "step": 3461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.930186365812587e-05,
+      "loss": 0.6459,
+      "step": 3462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.930146389419854e-05,
+      "loss": 0.6174,
+      "step": 3463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9301064019990586e-05,
+      "loss": 0.6398,
+      "step": 3464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.930066403550675e-05,
+      "loss": 0.6347,
+      "step": 3465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9300263940751784e-05,
+      "loss": 0.6361,
+      "step": 3466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929986373573042e-05,
+      "loss": 0.6777,
+      "step": 3467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9299463420447405e-05,
+      "loss": 0.6388,
+      "step": 3468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929906299490749e-05,
+      "loss": 0.6093,
+      "step": 3469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9298662459115417e-05,
+      "loss": 0.6863,
+      "step": 3470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9298261813075943e-05,
+      "loss": 0.6607,
+      "step": 3471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9297861056793807e-05,
+      "loss": 0.6256,
+      "step": 3472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9297460190273768e-05,
+      "loss": 0.6855,
+      "step": 3473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929705921352058e-05,
+      "loss": 0.6737,
+      "step": 3474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929665812653899e-05,
+      "loss": 0.6624,
+      "step": 3475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9296256929333758e-05,
+      "loss": 0.6209,
+      "step": 3476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9295855621909645e-05,
+      "loss": 0.6104,
+      "step": 3477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92954542042714e-05,
+      "loss": 0.6308,
+      "step": 3478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9295052676423788e-05,
+      "loss": 0.6467,
+      "step": 3479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9294651038371568e-05,
+      "loss": 0.617,
+      "step": 3480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92942492901195e-05,
+      "loss": 0.645,
+      "step": 3481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929384743167235e-05,
+      "loss": 0.6796,
+      "step": 3482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9293445463034888e-05,
+      "loss": 0.6553,
+      "step": 3483
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929304338421187e-05,
+      "loss": 0.5978,
+      "step": 3484
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9292641195208065e-05,
+      "loss": 0.6432,
+      "step": 3485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9292238896028242e-05,
+      "loss": 0.6035,
+      "step": 3486
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9291836486677175e-05,
+      "loss": 0.6622,
+      "step": 3487
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929143396715963e-05,
+      "loss": 0.6101,
+      "step": 3488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9291031337480386e-05,
+      "loss": 0.6265,
+      "step": 3489
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929062859764421e-05,
+      "loss": 0.6967,
+      "step": 3490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.929022574765588e-05,
+      "loss": 0.6182,
+      "step": 3491
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9289822787520168e-05,
+      "loss": 0.6165,
+      "step": 3492
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.928941971724186e-05,
+      "loss": 0.66,
+      "step": 3493
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9289016536825726e-05,
+      "loss": 0.5948,
+      "step": 3494
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.928861324627655e-05,
+      "loss": 0.6492,
+      "step": 3495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.928820984559912e-05,
+      "loss": 0.6539,
+      "step": 3496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9287806334798207e-05,
+      "loss": 0.7062,
+      "step": 3497
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9287402713878602e-05,
+      "loss": 0.6414,
+      "step": 3498
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.928699898284509e-05,
+      "loss": 0.701,
+      "step": 3499
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9286595141702455e-05,
+      "loss": 0.6014,
+      "step": 3500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.928619119045549e-05,
+      "loss": 0.679,
+      "step": 3501
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.928578712910898e-05,
+      "loss": 0.6202,
+      "step": 3502
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9285382957667716e-05,
+      "loss": 0.6077,
+      "step": 3503
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9284978676136494e-05,
+      "loss": 0.6577,
+      "step": 3504
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92845742845201e-05,
+      "loss": 0.6834,
+      "step": 3505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9284169782823333e-05,
+      "loss": 0.5994,
+      "step": 3506
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.928376517105099e-05,
+      "loss": 0.6636,
+      "step": 3507
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9283360449207867e-05,
+      "loss": 0.7007,
+      "step": 3508
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9282955617298766e-05,
+      "loss": 0.6246,
+      "step": 3509
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9282550675328477e-05,
+      "loss": 0.5976,
+      "step": 3510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.928214562330181e-05,
+      "loss": 0.6229,
+      "step": 3511
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9281740461223566e-05,
+      "loss": 0.5942,
+      "step": 3512
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9281335189098547e-05,
+      "loss": 0.6169,
+      "step": 3513
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9280929806931558e-05,
+      "loss": 0.6122,
+      "step": 3514
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9280524314727406e-05,
+      "loss": 0.6919,
+      "step": 3515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9280118712490896e-05,
+      "loss": 0.6176,
+      "step": 3516
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9279713000226842e-05,
+      "loss": 0.6365,
+      "step": 3517
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9279307177940046e-05,
+      "loss": 0.6951,
+      "step": 3518
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9278901245635332e-05,
+      "loss": 0.674,
+      "step": 3519
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92784952033175e-05,
+      "loss": 0.6064,
+      "step": 3520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9278089050991374e-05,
+      "loss": 0.7077,
+      "step": 3521
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9277682788661763e-05,
+      "loss": 0.6504,
+      "step": 3522
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.927727641633349e-05,
+      "loss": 0.6712,
+      "step": 3523
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9276869934011366e-05,
+      "loss": 0.5404,
+      "step": 3524
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.927646334170022e-05,
+      "loss": 0.659,
+      "step": 3525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9276056639404856e-05,
+      "loss": 0.6261,
+      "step": 3526
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9275649827130115e-05,
+      "loss": 0.6584,
+      "step": 3527
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9275242904880804e-05,
+      "loss": 0.6334,
+      "step": 3528
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9274835872661758e-05,
+      "loss": 0.6577,
+      "step": 3529
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9274428730477805e-05,
+      "loss": 0.6606,
+      "step": 3530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9274021478333762e-05,
+      "loss": 0.6789,
+      "step": 3531
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9273614116234465e-05,
+      "loss": 0.659,
+      "step": 3532
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9273206644184738e-05,
+      "loss": 0.6356,
+      "step": 3533
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9272799062189417e-05,
+      "loss": 0.6559,
+      "step": 3534
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9272391370253335e-05,
+      "loss": 0.591,
+      "step": 3535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.927198356838132e-05,
+      "loss": 0.6612,
+      "step": 3536
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9271575656578214e-05,
+      "loss": 0.6269,
+      "step": 3537
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9271167634848852e-05,
+      "loss": 0.6208,
+      "step": 3538
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9270759503198065e-05,
+      "loss": 0.6615,
+      "step": 3539
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92703512616307e-05,
+      "loss": 0.6508,
+      "step": 3540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.926994291015159e-05,
+      "loss": 0.622,
+      "step": 3541
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9269534448765583e-05,
+      "loss": 0.6332,
+      "step": 3542
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9269125877477517e-05,
+      "loss": 0.6467,
+      "step": 3543
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9268717196292242e-05,
+      "loss": 0.6294,
+      "step": 3544
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9268308405214594e-05,
+      "loss": 0.6296,
+      "step": 3545
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9267899504249433e-05,
+      "loss": 0.6464,
+      "step": 3546
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9267490493401592e-05,
+      "loss": 0.6748,
+      "step": 3547
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9267081372675932e-05,
+      "loss": 0.6313,
+      "step": 3548
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9266672142077295e-05,
+      "loss": 0.5971,
+      "step": 3549
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9266262801610543e-05,
+      "loss": 0.6157,
+      "step": 3550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.926585335128052e-05,
+      "loss": 0.6312,
+      "step": 3551
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9265443791092087e-05,
+      "loss": 0.6147,
+      "step": 3552
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9265034121050094e-05,
+      "loss": 0.6934,
+      "step": 3553
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9264624341159403e-05,
+      "loss": 0.6358,
+      "step": 3554
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.926421445142487e-05,
+      "loss": 0.6143,
+      "step": 3555
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9263804451851356e-05,
+      "loss": 0.6727,
+      "step": 3556
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.926339434244372e-05,
+      "loss": 0.621,
+      "step": 3557
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.926298412320683e-05,
+      "loss": 0.6666,
+      "step": 3558
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.926257379414554e-05,
+      "loss": 0.599,
+      "step": 3559
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9262163355264723e-05,
+      "loss": 0.6553,
+      "step": 3560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9261752806569243e-05,
+      "loss": 0.6134,
+      "step": 3561
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9261342148063968e-05,
+      "loss": 0.6209,
+      "step": 3562
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9260931379753763e-05,
+      "loss": 0.6152,
+      "step": 3563
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9260520501643504e-05,
+      "loss": 0.6339,
+      "step": 3564
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.926010951373806e-05,
+      "loss": 0.7084,
+      "step": 3565
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92596984160423e-05,
+      "loss": 0.6156,
+      "step": 3566
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9259287208561102e-05,
+      "loss": 0.6346,
+      "step": 3567
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9258875891299348e-05,
+      "loss": 0.6646,
+      "step": 3568
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9258464464261903e-05,
+      "loss": 0.5862,
+      "step": 3569
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925805292745365e-05,
+      "loss": 0.6484,
+      "step": 3570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9257641280879468e-05,
+      "loss": 0.6522,
+      "step": 3571
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9257229524544236e-05,
+      "loss": 0.6649,
+      "step": 3572
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9256817658452837e-05,
+      "loss": 0.6234,
+      "step": 3573
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925640568261016e-05,
+      "loss": 0.6634,
+      "step": 3574
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9255993597021077e-05,
+      "loss": 0.627,
+      "step": 3575
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9255581401690488e-05,
+      "loss": 0.6454,
+      "step": 3576
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925516909662327e-05,
+      "loss": 0.7241,
+      "step": 3577
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9254756681824314e-05,
+      "loss": 0.6622,
+      "step": 3578
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925434415729851e-05,
+      "loss": 0.6459,
+      "step": 3579
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9253931523050753e-05,
+      "loss": 0.6871,
+      "step": 3580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9253518779085926e-05,
+      "loss": 0.6547,
+      "step": 3581
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9253105925408934e-05,
+      "loss": 0.6655,
+      "step": 3582
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925269296202466e-05,
+      "loss": 0.6344,
+      "step": 3583
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9252279888938013e-05,
+      "loss": 0.6313,
+      "step": 3584
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925186670615388e-05,
+      "loss": 0.6583,
+      "step": 3585
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925145341367716e-05,
+      "loss": 0.681,
+      "step": 3586
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9251040011512762e-05,
+      "loss": 0.6666,
+      "step": 3587
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925062649966558e-05,
+      "loss": 0.6688,
+      "step": 3588
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925021287814052e-05,
+      "loss": 0.6795,
+      "step": 3589
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924979914694248e-05,
+      "loss": 0.617,
+      "step": 3590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9249385306076377e-05,
+      "loss": 0.6466,
+      "step": 3591
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9248971355547103e-05,
+      "loss": 0.6773,
+      "step": 3592
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924855729535958e-05,
+      "loss": 0.67,
+      "step": 3593
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924814312551871e-05,
+      "loss": 0.6373,
+      "step": 3594
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92477288460294e-05,
+      "loss": 0.6707,
+      "step": 3595
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9247314456896568e-05,
+      "loss": 0.6144,
+      "step": 3596
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924689995812513e-05,
+      "loss": 0.6219,
+      "step": 3597
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924648534971999e-05,
+      "loss": 0.6173,
+      "step": 3598
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9246070631686072e-05,
+      "loss": 0.5784,
+      "step": 3599
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924565580402829e-05,
+      "loss": 0.561,
+      "step": 3600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924524086675156e-05,
+      "loss": 0.5706,
+      "step": 3601
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9244825819860805e-05,
+      "loss": 0.6255,
+      "step": 3602
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924441066336095e-05,
+      "loss": 0.6934,
+      "step": 3603
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9243995397256908e-05,
+      "loss": 0.6467,
+      "step": 3604
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9243580021553606e-05,
+      "loss": 0.6096,
+      "step": 3605
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924316453625597e-05,
+      "loss": 0.6346,
+      "step": 3606
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9242748941368927e-05,
+      "loss": 0.639,
+      "step": 3607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9242333236897406e-05,
+      "loss": 0.6182,
+      "step": 3608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.924191742284633e-05,
+      "loss": 0.6393,
+      "step": 3609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.924150149922063e-05,
+      "loss": 0.6537,
+      "step": 3610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9241085466025242e-05,
+      "loss": 0.6607,
+      "step": 3611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9240669323265092e-05,
+      "loss": 0.6256,
+      "step": 3612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9240253070945122e-05,
+      "loss": 0.615,
+      "step": 3613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9239836709070264e-05,
+      "loss": 0.646,
+      "step": 3614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923942023764545e-05,
+      "loss": 0.5895,
+      "step": 3615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9239003656675623e-05,
+      "loss": 0.5995,
+      "step": 3616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9238586966165716e-05,
+      "loss": 0.6305,
+      "step": 3617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9238170166120677e-05,
+      "loss": 0.6609,
+      "step": 3618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9237753256545445e-05,
+      "loss": 0.6777,
+      "step": 3619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923733623744496e-05,
+      "loss": 0.6328,
+      "step": 3620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923691910882417e-05,
+      "loss": 0.6708,
+      "step": 3621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923650187068802e-05,
+      "loss": 0.5624,
+      "step": 3622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9236084523041456e-05,
+      "loss": 0.6226,
+      "step": 3623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9235667065889425e-05,
+      "loss": 0.6176,
+      "step": 3624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9235249499236878e-05,
+      "loss": 0.6382,
+      "step": 3625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9234831823088768e-05,
+      "loss": 0.6266,
+      "step": 3626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923441403745004e-05,
+      "loss": 0.5934,
+      "step": 3627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9233996142325653e-05,
+      "loss": 0.6023,
+      "step": 3628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9233578137720562e-05,
+      "loss": 0.6225,
+      "step": 3629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9233160023639722e-05,
+      "loss": 0.6817,
+      "step": 3630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923274180008809e-05,
+      "loss": 0.6336,
+      "step": 3631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9232323467070625e-05,
+      "loss": 0.6227,
+      "step": 3632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9231905024592285e-05,
+      "loss": 0.6769,
+      "step": 3633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9231486472658034e-05,
+      "loss": 0.7383,
+      "step": 3634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923106781127283e-05,
+      "loss": 0.6784,
+      "step": 3635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923064904044164e-05,
+      "loss": 0.6241,
+      "step": 3636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9230230160169434e-05,
+      "loss": 0.6516,
+      "step": 3637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922981117046117e-05,
+      "loss": 0.6549,
+      "step": 3638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9229392071321818e-05,
+      "loss": 0.6759,
+      "step": 3639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9228972862756347e-05,
+      "loss": 0.5907,
+      "step": 3640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922855354476973e-05,
+      "loss": 0.6055,
+      "step": 3641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9228134117366938e-05,
+      "loss": 0.62,
+      "step": 3642
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922771458055294e-05,
+      "loss": 0.66,
+      "step": 3643
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922729493433271e-05,
+      "loss": 0.637,
+      "step": 3644
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922687517871123e-05,
+      "loss": 0.6915,
+      "step": 3645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922645531369347e-05,
+      "loss": 0.6701,
+      "step": 3646
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922603533928441e-05,
+      "loss": 0.6512,
+      "step": 3647
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9225615255489035e-05,
+      "loss": 0.6439,
+      "step": 3648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9225195062312317e-05,
+      "loss": 0.6646,
+      "step": 3649
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9224774759759245e-05,
+      "loss": 0.6171,
+      "step": 3650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9224354347834794e-05,
+      "loss": 0.5985,
+      "step": 3651
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9223933826543956e-05,
+      "loss": 0.6831,
+      "step": 3652
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9223513195891713e-05,
+      "loss": 0.5749,
+      "step": 3653
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9223092455883056e-05,
+      "loss": 0.5947,
+      "step": 3654
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9222671606522965e-05,
+      "loss": 0.6524,
+      "step": 3655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9222250647816443e-05,
+      "loss": 0.6512,
+      "step": 3656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9221829579768467e-05,
+      "loss": 0.5757,
+      "step": 3657
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9221408402384042e-05,
+      "loss": 0.6958,
+      "step": 3658
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922098711566815e-05,
+      "loss": 0.591,
+      "step": 3659
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9220565719625793e-05,
+      "loss": 0.672,
+      "step": 3660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922014421426196e-05,
+      "loss": 0.6865,
+      "step": 3661
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9219722599581662e-05,
+      "loss": 0.6643,
+      "step": 3662
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9219300875589885e-05,
+      "loss": 0.6537,
+      "step": 3663
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9218879042291636e-05,
+      "loss": 0.6543,
+      "step": 3664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9218457099691913e-05,
+      "loss": 0.6541,
+      "step": 3665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9218035047795718e-05,
+      "loss": 0.611,
+      "step": 3666
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9217612886608058e-05,
+      "loss": 0.65,
+      "step": 3667
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.921719061613394e-05,
+      "loss": 0.6005,
+      "step": 3668
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9216768236378362e-05,
+      "loss": 0.6494,
+      "step": 3669
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.921634574734634e-05,
+      "loss": 0.6575,
+      "step": 3670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.921592314904288e-05,
+      "loss": 0.5845,
+      "step": 3671
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9215500441472992e-05,
+      "loss": 0.6844,
+      "step": 3672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9215077624641688e-05,
+      "loss": 0.7066,
+      "step": 3673
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9214654698553986e-05,
+      "loss": 0.6956,
+      "step": 3674
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9214231663214893e-05,
+      "loss": 0.641,
+      "step": 3675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9213808518629428e-05,
+      "loss": 0.6717,
+      "step": 3676
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9213385264802605e-05,
+      "loss": 0.6116,
+      "step": 3677
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9212961901739448e-05,
+      "loss": 0.6594,
+      "step": 3678
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9212538429444972e-05,
+      "loss": 0.6755,
+      "step": 3679
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9212114847924202e-05,
+      "loss": 0.6518,
+      "step": 3680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9211691157182152e-05,
+      "loss": 0.6125,
+      "step": 3681
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9211267357223853e-05,
+      "loss": 0.6842,
+      "step": 3682
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9210843448054327e-05,
+      "loss": 0.5969,
+      "step": 3683
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9210419429678598e-05,
+      "loss": 0.7185,
+      "step": 3684
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9209995302101697e-05,
+      "loss": 0.6229,
+      "step": 3685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9209571065328653e-05,
+      "loss": 0.6258,
+      "step": 3686
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9209146719364494e-05,
+      "loss": 0.597,
+      "step": 3687
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9208722264214247e-05,
+      "loss": 0.6166,
+      "step": 3688
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.920829769988295e-05,
+      "loss": 0.7169,
+      "step": 3689
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.920787302637564e-05,
+      "loss": 0.6631,
+      "step": 3690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.920744824369734e-05,
+      "loss": 0.6677,
+      "step": 3691
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.92070233518531e-05,
+      "loss": 0.5927,
+      "step": 3692
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9206598350847946e-05,
+      "loss": 0.6265,
+      "step": 3693
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9206173240686925e-05,
+      "loss": 0.6505,
+      "step": 3694
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9205748021375072e-05,
+      "loss": 0.6285,
+      "step": 3695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9205322692917434e-05,
+      "loss": 0.6863,
+      "step": 3696
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.920489725531905e-05,
+      "loss": 0.6177,
+      "step": 3697
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.920447170858496e-05,
+      "loss": 0.6005,
+      "step": 3698
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.920404605272022e-05,
+      "loss": 0.5945,
+      "step": 3699
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9203620287729867e-05,
+      "loss": 0.6326,
+      "step": 3700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9203194413618958e-05,
+      "loss": 0.6354,
+      "step": 3701
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9202768430392532e-05,
+      "loss": 0.6226,
+      "step": 3702
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9202342338055646e-05,
+      "loss": 0.6588,
+      "step": 3703
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9201916136613353e-05,
+      "loss": 0.5926,
+      "step": 3704
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.92014898260707e-05,
+      "loss": 0.613,
+      "step": 3705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9201063406432748e-05,
+      "loss": 0.6444,
+      "step": 3706
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9200636877704548e-05,
+      "loss": 0.6631,
+      "step": 3707
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.920021023989116e-05,
+      "loss": 0.6598,
+      "step": 3708
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.919978349299764e-05,
+      "loss": 0.6985,
+      "step": 3709
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9199356637029052e-05,
+      "loss": 0.6473,
+      "step": 3710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.919892967199045e-05,
+      "loss": 0.6514,
+      "step": 3711
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91985025978869e-05,
+      "loss": 0.7051,
+      "step": 3712
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9198075414723472e-05,
+      "loss": 0.5915,
+      "step": 3713
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.919764812250522e-05,
+      "loss": 0.6206,
+      "step": 3714
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9197220721237212e-05,
+      "loss": 0.6574,
+      "step": 3715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.919679321092452e-05,
+      "loss": 0.6789,
+      "step": 3716
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.919636559157221e-05,
+      "loss": 0.6544,
+      "step": 3717
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9195937863185352e-05,
+      "loss": 0.6609,
+      "step": 3718
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9195510025769023e-05,
+      "loss": 0.6577,
+      "step": 3719
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9195082079328284e-05,
+      "loss": 0.666,
+      "step": 3720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9194654023868215e-05,
+      "loss": 0.6149,
+      "step": 3721
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9194225859393897e-05,
+      "loss": 0.6568,
+      "step": 3722
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9193797585910394e-05,
+      "loss": 0.6317,
+      "step": 3723
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9193369203422792e-05,
+      "loss": 0.6275,
+      "step": 3724
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9192940711936173e-05,
+      "loss": 0.5831,
+      "step": 3725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9192512111455606e-05,
+      "loss": 0.6363,
+      "step": 3726
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9192083401986186e-05,
+      "loss": 0.6325,
+      "step": 3727
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9191654583532987e-05,
+      "loss": 0.6149,
+      "step": 3728
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9191225656101094e-05,
+      "loss": 0.6465,
+      "step": 3729
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9190796619695594e-05,
+      "loss": 0.6257,
+      "step": 3730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9190367474321572e-05,
+      "loss": 0.5962,
+      "step": 3731
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918993821998412e-05,
+      "loss": 0.5845,
+      "step": 3732
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9189508856688325e-05,
+      "loss": 0.6539,
+      "step": 3733
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9189079384439278e-05,
+      "loss": 0.6777,
+      "step": 3734
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918864980324207e-05,
+      "loss": 0.6349,
+      "step": 3735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9188220113101794e-05,
+      "loss": 0.6206,
+      "step": 3736
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9187790314023546e-05,
+      "loss": 0.6574,
+      "step": 3737
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9187360406012422e-05,
+      "loss": 0.6468,
+      "step": 3738
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9186930389073518e-05,
+      "loss": 0.6193,
+      "step": 3739
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9186500263211934e-05,
+      "loss": 0.6254,
+      "step": 3740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9186070028432765e-05,
+      "loss": 0.5945,
+      "step": 3741
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918563968474112e-05,
+      "loss": 0.6568,
+      "step": 3742
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9185209232142093e-05,
+      "loss": 0.6227,
+      "step": 3743
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9184778670640792e-05,
+      "loss": 0.6731,
+      "step": 3744
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9184348000242322e-05,
+      "loss": 0.6525,
+      "step": 3745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9183917220951788e-05,
+      "loss": 0.7041,
+      "step": 3746
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9183486332774297e-05,
+      "loss": 0.656,
+      "step": 3747
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918305533571496e-05,
+      "loss": 0.6159,
+      "step": 3748
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918262422977888e-05,
+      "loss": 0.6323,
+      "step": 3749
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9182193014971177e-05,
+      "loss": 0.6577,
+      "step": 3750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918176169129696e-05,
+      "loss": 0.6238,
+      "step": 3751
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918133025876135e-05,
+      "loss": 0.6313,
+      "step": 3752
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9180898717369448e-05,
+      "loss": 0.6812,
+      "step": 3753
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918046706712638e-05,
+      "loss": 0.6005,
+      "step": 3754
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918003530803726e-05,
+      "loss": 0.6429,
+      "step": 3755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917960344010721e-05,
+      "loss": 0.6391,
+      "step": 3756
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9179171463341345e-05,
+      "loss": 0.6729,
+      "step": 3757
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9178739377744792e-05,
+      "loss": 0.6574,
+      "step": 3758
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9178307183322675e-05,
+      "loss": 0.6385,
+      "step": 3759
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9177874880080118e-05,
+      "loss": 0.594,
+      "step": 3760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917744246802224e-05,
+      "loss": 0.5916,
+      "step": 3761
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9177009947154174e-05,
+      "loss": 0.6113,
+      "step": 3762
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9176577317481043e-05,
+      "loss": 0.7035,
+      "step": 3763
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9176144579007983e-05,
+      "loss": 0.6481,
+      "step": 3764
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917571173174012e-05,
+      "loss": 0.6794,
+      "step": 3765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9175278775682587e-05,
+      "loss": 0.6077,
+      "step": 3766
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9174845710840517e-05,
+      "loss": 0.6436,
+      "step": 3767
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9174412537219046e-05,
+      "loss": 0.5757,
+      "step": 3768
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917397925482331e-05,
+      "loss": 0.6204,
+      "step": 3769
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917354586365844e-05,
+      "loss": 0.6332,
+      "step": 3770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9173112363729585e-05,
+      "loss": 0.6573,
+      "step": 3771
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9172678755041873e-05,
+      "loss": 0.689,
+      "step": 3772
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9172245037600456e-05,
+      "loss": 0.6012,
+      "step": 3773
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9171811211410467e-05,
+      "loss": 0.6595,
+      "step": 3774
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9171377276477058e-05,
+      "loss": 0.7174,
+      "step": 3775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9170943232805366e-05,
+      "loss": 0.6624,
+      "step": 3776
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917050908040054e-05,
+      "loss": 0.6229,
+      "step": 3777
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9170074819267733e-05,
+      "loss": 0.5394,
+      "step": 3778
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9169640449412083e-05,
+      "loss": 0.6392,
+      "step": 3779
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916920597083875e-05,
+      "loss": 0.6307,
+      "step": 3780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916877138355288e-05,
+      "loss": 0.6088,
+      "step": 3781
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9168336687559624e-05,
+      "loss": 0.6509,
+      "step": 3782
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916790188286414e-05,
+      "loss": 0.6344,
+      "step": 3783
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9167466969471583e-05,
+      "loss": 0.6266,
+      "step": 3784
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916703194738711e-05,
+      "loss": 0.62,
+      "step": 3785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9166596816615875e-05,
+      "loss": 0.6925,
+      "step": 3786
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9166161577163035e-05,
+      "loss": 0.6077,
+      "step": 3787
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916572622903376e-05,
+      "loss": 0.6358,
+      "step": 3788
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9165290772233206e-05,
+      "loss": 0.6389,
+      "step": 3789
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916485520676653e-05,
+      "loss": 0.6659,
+      "step": 3790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9164419532638906e-05,
+      "loss": 0.6682,
+      "step": 3791
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9163983749855495e-05,
+      "loss": 0.6526,
+      "step": 3792
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9163547858421463e-05,
+      "loss": 0.5878,
+      "step": 3793
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9163111858341982e-05,
+      "loss": 0.6109,
+      "step": 3794
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9162675749622214e-05,
+      "loss": 0.6518,
+      "step": 3795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9162239532267336e-05,
+      "loss": 0.6496,
+      "step": 3796
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916180320628252e-05,
+      "loss": 0.6427,
+      "step": 3797
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9161366771672937e-05,
+      "loss": 0.5929,
+      "step": 3798
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916093022844376e-05,
+      "loss": 0.6421,
+      "step": 3799
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9160493576600167e-05,
+      "loss": 0.6179,
+      "step": 3800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9160056816147334e-05,
+      "loss": 0.5858,
+      "step": 3801
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9159619947090443e-05,
+      "loss": 0.6815,
+      "step": 3802
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915918296943467e-05,
+      "loss": 0.6215,
+      "step": 3803
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915874588318519e-05,
+      "loss": 0.6062,
+      "step": 3804
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91583086883472e-05,
+      "loss": 0.6598,
+      "step": 3805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9157871384925872e-05,
+      "loss": 0.6362,
+      "step": 3806
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9157433972926395e-05,
+      "loss": 0.6623,
+      "step": 3807
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915699645235395e-05,
+      "loss": 0.6719,
+      "step": 3808
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915655882321373e-05,
+      "loss": 0.5919,
+      "step": 3809
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9156121085510922e-05,
+      "loss": 0.6205,
+      "step": 3810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915568323925072e-05,
+      "loss": 0.5998,
+      "step": 3811
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9155245284438308e-05,
+      "loss": 0.6651,
+      "step": 3812
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9154807221078883e-05,
+      "loss": 0.6624,
+      "step": 3813
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9154369049177634e-05,
+      "loss": 0.6807,
+      "step": 3814
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9153930768739762e-05,
+      "loss": 0.6779,
+      "step": 3815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915349237977046e-05,
+      "loss": 0.6564,
+      "step": 3816
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915305388227493e-05,
+      "loss": 0.6271,
+      "step": 3817
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9152615276258364e-05,
+      "loss": 0.6339,
+      "step": 3818
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9152176561725966e-05,
+      "loss": 0.5899,
+      "step": 3819
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915173773868294e-05,
+      "loss": 0.6577,
+      "step": 3820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9151298807134485e-05,
+      "loss": 0.6209,
+      "step": 3821
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9150859767085805e-05,
+      "loss": 0.6298,
+      "step": 3822
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9150420618542103e-05,
+      "loss": 0.6075,
+      "step": 3823
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91499813615086e-05,
+      "loss": 0.6377,
+      "step": 3824
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9149541995990483e-05,
+      "loss": 0.6406,
+      "step": 3825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9149102521992976e-05,
+      "loss": 0.6463,
+      "step": 3826
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9148662939521282e-05,
+      "loss": 0.6505,
+      "step": 3827
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914822324858062e-05,
+      "loss": 0.6287,
+      "step": 3828
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9147783449176195e-05,
+      "loss": 0.6351,
+      "step": 3829
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9147343541313227e-05,
+      "loss": 0.6339,
+      "step": 3830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914690352499693e-05,
+      "loss": 0.6259,
+      "step": 3831
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9146463400232516e-05,
+      "loss": 0.6544,
+      "step": 3832
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9146023167025216e-05,
+      "loss": 0.6233,
+      "step": 3833
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9145582825380236e-05,
+      "loss": 0.6616,
+      "step": 3834
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9145142375302805e-05,
+      "loss": 0.6211,
+      "step": 3835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9144701816798143e-05,
+      "loss": 0.6586,
+      "step": 3836
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914426114987147e-05,
+      "loss": 0.6269,
+      "step": 3837
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9143820374528017e-05,
+      "loss": 0.6643,
+      "step": 3838
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9143379490773003e-05,
+      "loss": 0.5995,
+      "step": 3839
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914293849861166e-05,
+      "loss": 0.6285,
+      "step": 3840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9142497398049215e-05,
+      "loss": 0.5983,
+      "step": 3841
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9142056189090897e-05,
+      "loss": 0.5782,
+      "step": 3842
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9141614871741943e-05,
+      "loss": 0.6357,
+      "step": 3843
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9141173446007575e-05,
+      "loss": 0.5978,
+      "step": 3844
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9140731911893034e-05,
+      "loss": 0.6542,
+      "step": 3845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9140290269403555e-05,
+      "loss": 0.6591,
+      "step": 3846
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9139848518544372e-05,
+      "loss": 0.6603,
+      "step": 3847
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913940665932072e-05,
+      "loss": 0.6546,
+      "step": 3848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913896469173784e-05,
+      "loss": 0.6439,
+      "step": 3849
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9138522615800976e-05,
+      "loss": 0.6116,
+      "step": 3850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9138080431515366e-05,
+      "loss": 0.6876,
+      "step": 3851
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9137638138886253e-05,
+      "loss": 0.5811,
+      "step": 3852
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9137195737918878e-05,
+      "loss": 0.707,
+      "step": 3853
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913675322861849e-05,
+      "loss": 0.6908,
+      "step": 3854
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9136310610990336e-05,
+      "loss": 0.6895,
+      "step": 3855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9135867885039662e-05,
+      "loss": 0.5792,
+      "step": 3856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9135425050771715e-05,
+      "loss": 0.5842,
+      "step": 3857
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913498210819175e-05,
+      "loss": 0.627,
+      "step": 3858
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9134539057305016e-05,
+      "loss": 0.6787,
+      "step": 3859
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9134095898116766e-05,
+      "loss": 0.5951,
+      "step": 3860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9133652630632253e-05,
+      "loss": 0.676,
+      "step": 3861
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9133209254856733e-05,
+      "loss": 0.6486,
+      "step": 3862
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9132765770795467e-05,
+      "loss": 0.651,
+      "step": 3863
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9132322178453708e-05,
+      "loss": 0.6353,
+      "step": 3864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9131878477836716e-05,
+      "loss": 0.5975,
+      "step": 3865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9131434668949756e-05,
+      "loss": 0.6854,
+      "step": 3866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9130990751798085e-05,
+      "loss": 0.6725,
+      "step": 3867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9130546726386967e-05,
+      "loss": 0.6279,
+      "step": 3868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913010259272167e-05,
+      "loss": 0.6576,
+      "step": 3869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9129658350807453e-05,
+      "loss": 0.6419,
+      "step": 3870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912921400064959e-05,
+      "loss": 0.5909,
+      "step": 3871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9128769542253346e-05,
+      "loss": 0.6865,
+      "step": 3872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9128324975623994e-05,
+      "loss": 0.6023,
+      "step": 3873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9127880300766796e-05,
+      "loss": 0.6546,
+      "step": 3874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9127435517687034e-05,
+      "loss": 0.6876,
+      "step": 3875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9126990626389976e-05,
+      "loss": 0.6535,
+      "step": 3876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91265456268809e-05,
+      "loss": 0.698,
+      "step": 3877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912610051916508e-05,
+      "loss": 0.6202,
+      "step": 3878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9125655303247793e-05,
+      "loss": 0.6237,
+      "step": 3879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9125209979134318e-05,
+      "loss": 0.6398,
+      "step": 3880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9124764546829937e-05,
+      "loss": 0.604,
+      "step": 3881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912431900633993e-05,
+      "loss": 0.6902,
+      "step": 3882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9123873357669577e-05,
+      "loss": 0.5979,
+      "step": 3883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9123427600824164e-05,
+      "loss": 0.5798,
+      "step": 3884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9122981735808977e-05,
+      "loss": 0.6988,
+      "step": 3885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9122535762629297e-05,
+      "loss": 0.6156,
+      "step": 3886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912208968129042e-05,
+      "loss": 0.6435,
+      "step": 3887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9121643491797627e-05,
+      "loss": 0.6971,
+      "step": 3888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9121197194156212e-05,
+      "loss": 0.6296,
+      "step": 3889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912075078837147e-05,
+      "loss": 0.6518,
+      "step": 3890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9120304274448685e-05,
+      "loss": 0.6229,
+      "step": 3891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9119857652393157e-05,
+      "loss": 0.6008,
+      "step": 3892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9119410922210177e-05,
+      "loss": 0.6446,
+      "step": 3893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911896408390505e-05,
+      "loss": 0.5959,
+      "step": 3894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9118517137483063e-05,
+      "loss": 0.6638,
+      "step": 3895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9118070082949524e-05,
+      "loss": 0.6597,
+      "step": 3896
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911762292030973e-05,
+      "loss": 0.5972,
+      "step": 3897
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911717564956898e-05,
+      "loss": 0.6057,
+      "step": 3898
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911672827073258e-05,
+      "loss": 0.6261,
+      "step": 3899
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9116280783805835e-05,
+      "loss": 0.6355,
+      "step": 3900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9115833188794046e-05,
+      "loss": 0.6133,
+      "step": 3901
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9115385485702528e-05,
+      "loss": 0.7078,
+      "step": 3902
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9114937674536578e-05,
+      "loss": 0.6106,
+      "step": 3903
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9114489755301517e-05,
+      "loss": 0.6625,
+      "step": 3904
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9114041728002645e-05,
+      "loss": 0.6528,
+      "step": 3905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911359359264528e-05,
+      "loss": 0.6805,
+      "step": 3906
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911314534923474e-05,
+      "loss": 0.6697,
+      "step": 3907
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9112696997776328e-05,
+      "loss": 0.6653,
+      "step": 3908
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9112248538275365e-05,
+      "loss": 0.6068,
+      "step": 3909
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9111799970737172e-05,
+      "loss": 0.6357,
+      "step": 3910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9111351295167063e-05,
+      "loss": 0.6725,
+      "step": 3911
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911090251157036e-05,
+      "loss": 0.6202,
+      "step": 3912
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9110453619952377e-05,
+      "loss": 0.695,
+      "step": 3913
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9110004620318446e-05,
+      "loss": 0.6066,
+      "step": 3914
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9109555512673882e-05,
+      "loss": 0.6183,
+      "step": 3915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910910629702402e-05,
+      "loss": 0.6301,
+      "step": 3916
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9108656973374174e-05,
+      "loss": 0.663,
+      "step": 3917
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910820754172968e-05,
+      "loss": 0.5878,
+      "step": 3918
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9107758002095863e-05,
+      "loss": 0.6255,
+      "step": 3919
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9107308354478055e-05,
+      "loss": 0.7236,
+      "step": 3920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910685859888158e-05,
+      "loss": 0.6553,
+      "step": 3921
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910640873531178e-05,
+      "loss": 0.633,
+      "step": 3922
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9105958763773988e-05,
+      "loss": 0.6552,
+      "step": 3923
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910550868427353e-05,
+      "loss": 0.6341,
+      "step": 3924
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910505849681575e-05,
+      "loss": 0.6355,
+      "step": 3925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910460820140599e-05,
+      "loss": 0.6183,
+      "step": 3926
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9104157798049576e-05,
+      "loss": 0.6458,
+      "step": 3927
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910370728675185e-05,
+      "loss": 0.5981,
+      "step": 3928
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9103256667518165e-05,
+      "loss": 0.6502,
+      "step": 3929
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9102805940353857e-05,
+      "loss": 0.6209,
+      "step": 3930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9102355105264266e-05,
+      "loss": 0.6309,
+      "step": 3931
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910190416225474e-05,
+      "loss": 0.6314,
+      "step": 3932
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910145311133063e-05,
+      "loss": 0.633,
+      "step": 3933
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9101001952497275e-05,
+      "loss": 0.6243,
+      "step": 3934
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9100550685760034e-05,
+      "loss": 0.6358,
+      "step": 3935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9100099311124252e-05,
+      "loss": 0.6968,
+      "step": 3936
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9099647828595277e-05,
+      "loss": 0.6636,
+      "step": 3937
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909919623817847e-05,
+      "loss": 0.6561,
+      "step": 3938
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9098744539879178e-05,
+      "loss": 0.6008,
+      "step": 3939
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9098292733702762e-05,
+      "loss": 0.6923,
+      "step": 3940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9097840819654574e-05,
+      "loss": 0.6311,
+      "step": 3941
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909738879773997e-05,
+      "loss": 0.6134,
+      "step": 3942
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9096936667964323e-05,
+      "loss": 0.5899,
+      "step": 3943
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909648443033298e-05,
+      "loss": 0.6275,
+      "step": 3944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9096032084851308e-05,
+      "loss": 0.668,
+      "step": 3945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909557963152467e-05,
+      "loss": 0.6349,
+      "step": 3946
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9095127070358422e-05,
+      "loss": 0.5979,
+      "step": 3947
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9094674401357944e-05,
+      "loss": 0.7024,
+      "step": 3948
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9094221624528595e-05,
+      "loss": 0.6902,
+      "step": 3949
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9093768739875743e-05,
+      "loss": 0.5834,
+      "step": 3950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9093315747404763e-05,
+      "loss": 0.6319,
+      "step": 3951
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909286264712102e-05,
+      "loss": 0.6641,
+      "step": 3952
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9092409439029884e-05,
+      "loss": 0.6019,
+      "step": 3953
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9091956123136736e-05,
+      "loss": 0.582,
+      "step": 3954
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9091502699446944e-05,
+      "loss": 0.6662,
+      "step": 3955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9091049167965886e-05,
+      "loss": 0.7051,
+      "step": 3956
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9090595528698944e-05,
+      "loss": 0.5981,
+      "step": 3957
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909014178165149e-05,
+      "loss": 0.6675,
+      "step": 3958
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9089687926828908e-05,
+      "loss": 0.6593,
+      "step": 3959
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9089233964236575e-05,
+      "loss": 0.6613,
+      "step": 3960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9088779893879875e-05,
+      "loss": 0.6395,
+      "step": 3961
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9088325715764192e-05,
+      "loss": 0.6406,
+      "step": 3962
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9087871429894914e-05,
+      "loss": 0.6465,
+      "step": 3963
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908741703627742e-05,
+      "loss": 0.6108,
+      "step": 3964
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9086962534917103e-05,
+      "loss": 0.6598,
+      "step": 3965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908650792581935e-05,
+      "loss": 0.6917,
+      "step": 3966
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9086053208989552e-05,
+      "loss": 0.6389,
+      "step": 3967
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90855983844331e-05,
+      "loss": 0.6618,
+      "step": 3968
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9085143452155385e-05,
+      "loss": 0.6187,
+      "step": 3969
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90846884121618e-05,
+      "loss": 0.6455,
+      "step": 3970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908423326445774e-05,
+      "loss": 0.6189,
+      "step": 3971
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9083778009048607e-05,
+      "loss": 0.611,
+      "step": 3972
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9083322645939795e-05,
+      "loss": 0.6764,
+      "step": 3973
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90828671751367e-05,
+      "loss": 0.6589,
+      "step": 3974
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9082411596644728e-05,
+      "loss": 0.6173,
+      "step": 3975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9081955910469272e-05,
+      "loss": 0.6596,
+      "step": 3976
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9081500116615746e-05,
+      "loss": 0.6464,
+      "step": 3977
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9081044215089542e-05,
+      "loss": 0.6103,
+      "step": 3978
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9080588205896076e-05,
+      "loss": 0.6375,
+      "step": 3979
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9080132089040745e-05,
+      "loss": 0.6073,
+      "step": 3980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9079675864528966e-05,
+      "loss": 0.6651,
+      "step": 3981
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9079219532366144e-05,
+      "loss": 0.6647,
+      "step": 3982
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9078763092557685e-05,
+      "loss": 0.6983,
+      "step": 3983
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907830654510901e-05,
+      "loss": 0.6581,
+      "step": 3984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9077849890025524e-05,
+      "loss": 0.7043,
+      "step": 3985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9077393127312646e-05,
+      "loss": 0.6311,
+      "step": 3986
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076936256975785e-05,
+      "loss": 0.6409,
+      "step": 3987
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076479279020366e-05,
+      "loss": 0.6163,
+      "step": 3988
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076022193451803e-05,
+      "loss": 0.6504,
+      "step": 3989
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9075565000275514e-05,
+      "loss": 0.6449,
+      "step": 3990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907510769949692e-05,
+      "loss": 0.6375,
+      "step": 3991
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907465029112145e-05,
+      "loss": 0.6007,
+      "step": 3992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9074192775154517e-05,
+      "loss": 0.5889,
+      "step": 3993
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907373515160155e-05,
+      "loss": 0.7103,
+      "step": 3994
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9073277420467975e-05,
+      "loss": 0.6192,
+      "step": 3995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9072819581759216e-05,
+      "loss": 0.611,
+      "step": 3996
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9072361635480706e-05,
+      "loss": 0.642,
+      "step": 3997
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907190358163787e-05,
+      "loss": 0.6667,
+      "step": 3998
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9071445420236147e-05,
+      "loss": 0.6411,
+      "step": 3999
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907098715128096e-05,
+      "loss": 0.608,
+      "step": 4000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9070528774777743e-05,
+      "loss": 0.6342,
+      "step": 4001
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9070070290731933e-05,
+      "loss": 0.6088,
+      "step": 4002
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906961169914897e-05,
+      "loss": 0.672,
+      "step": 4003
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9069153000034286e-05,
+      "loss": 0.6732,
+      "step": 4004
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906869419339332e-05,
+      "loss": 0.6268,
+      "step": 4005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9068235279231513e-05,
+      "loss": 0.6503,
+      "step": 4006
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90677762575543e-05,
+      "loss": 0.6347,
+      "step": 4007
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9067317128367134e-05,
+      "loss": 0.6503,
+      "step": 4008
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9066857891675454e-05,
+      "loss": 0.6599,
+      "step": 4009
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90663985474847e-05,
+      "loss": 0.6487,
+      "step": 4010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9065939095800327e-05,
+      "loss": 0.6538,
+      "step": 4011
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9065479536627774e-05,
+      "loss": 0.6295,
+      "step": 4012
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9065019869972497e-05,
+      "loss": 0.6518,
+      "step": 4013
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906456009583994e-05,
+      "loss": 0.5956,
+      "step": 4014
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9064100214235553e-05,
+      "loss": 0.7012,
+      "step": 4015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9063640225164794e-05,
+      "loss": 0.7303,
+      "step": 4016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9063180128633116e-05,
+      "loss": 0.6168,
+      "step": 4017
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906271992464597e-05,
+      "loss": 0.6944,
+      "step": 4018
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906225961320882e-05,
+      "loss": 0.6517,
+      "step": 4019
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9061799194327114e-05,
+      "loss": 0.6199,
+      "step": 4020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9061338668006317e-05,
+      "loss": 0.6153,
+      "step": 4021
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9060878034251887e-05,
+      "loss": 0.6122,
+      "step": 4022
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9060417293069288e-05,
+      "loss": 0.654,
+      "step": 4023
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9059956444463975e-05,
+      "loss": 0.6551,
+      "step": 4024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9059495488441423e-05,
+      "loss": 0.621,
+      "step": 4025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905903442500709e-05,
+      "loss": 0.5993,
+      "step": 4026
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9058573254166443e-05,
+      "loss": 0.633,
+      "step": 4027
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9058111975924948e-05,
+      "loss": 0.6867,
+      "step": 4028
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9057650590288085e-05,
+      "loss": 0.6468,
+      "step": 4029
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9057189097261308e-05,
+      "loss": 0.6375,
+      "step": 4030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9056727496850103e-05,
+      "loss": 0.6471,
+      "step": 4031
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9056265789059932e-05,
+      "loss": 0.6697,
+      "step": 4032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9055803973896277e-05,
+      "loss": 0.6398,
+      "step": 4033
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905534205136461e-05,
+      "loss": 0.6665,
+      "step": 4034
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9054880021470406e-05,
+      "loss": 0.62,
+      "step": 4035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905441788421915e-05,
+      "loss": 0.667,
+      "step": 4036
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9053955639616308e-05,
+      "loss": 0.6272,
+      "step": 4037
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9053493287667372e-05,
+      "loss": 0.5821,
+      "step": 4038
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9053030828377823e-05,
+      "loss": 0.6451,
+      "step": 4039
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905256826175314e-05,
+      "loss": 0.5887,
+      "step": 4040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9052105587798806e-05,
+      "loss": 0.5971,
+      "step": 4041
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905164280652031e-05,
+      "loss": 0.6256,
+      "step": 4042
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9051179917923144e-05,
+      "loss": 0.5889,
+      "step": 4043
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9050716922012786e-05,
+      "loss": 0.6163,
+      "step": 4044
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905025381879473e-05,
+      "loss": 0.6403,
+      "step": 4045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9049790608274462e-05,
+      "loss": 0.6922,
+      "step": 4046
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9049327290457485e-05,
+      "loss": 0.62,
+      "step": 4047
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904886386534928e-05,
+      "loss": 0.6614,
+      "step": 4048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904840033295535e-05,
+      "loss": 0.6164,
+      "step": 4049
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904793669328119e-05,
+      "loss": 0.6431,
+      "step": 4050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904747294633229e-05,
+      "loss": 0.638,
+      "step": 4051
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9047009092114156e-05,
+      "loss": 0.6435,
+      "step": 4052
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9046545130632282e-05,
+      "loss": 0.6341,
+      "step": 4053
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9046081061892178e-05,
+      "loss": 0.6041,
+      "step": 4054
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9045616885899335e-05,
+      "loss": 0.5987,
+      "step": 4055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9045152602659257e-05,
+      "loss": 0.6747,
+      "step": 4056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904468821217746e-05,
+      "loss": 0.6419,
+      "step": 4057
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904422371445944e-05,
+      "loss": 0.6232,
+      "step": 4058
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9043759109510708e-05,
+      "loss": 0.6513,
+      "step": 4059
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904329439733677e-05,
+      "loss": 0.5927,
+      "step": 4060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9042829577943136e-05,
+      "loss": 0.6291,
+      "step": 4061
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904236465133532e-05,
+      "loss": 0.6292,
+      "step": 4062
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904189961751883e-05,
+      "loss": 0.6559,
+      "step": 4063
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9041434476499184e-05,
+      "loss": 0.6931,
+      "step": 4064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9040969228281893e-05,
+      "loss": 0.6202,
+      "step": 4065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9040503872872477e-05,
+      "loss": 0.6605,
+      "step": 4066
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904003841027645e-05,
+      "loss": 0.6155,
+      "step": 4067
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903957284049933e-05,
+      "loss": 0.661,
+      "step": 4068
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9039107163546644e-05,
+      "loss": 0.61,
+      "step": 4069
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9038641379423906e-05,
+      "loss": 0.6403,
+      "step": 4070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903817548813664e-05,
+      "loss": 0.639,
+      "step": 4071
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903770948969037e-05,
+      "loss": 0.6393,
+      "step": 4072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903724338409062e-05,
+      "loss": 0.6198,
+      "step": 4073
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9036777171342918e-05,
+      "loss": 0.6075,
+      "step": 4074
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9036310851452792e-05,
+      "loss": 0.6036,
+      "step": 4075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903584442442577e-05,
+      "loss": 0.6659,
+      "step": 4076
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9035377890267384e-05,
+      "loss": 0.6759,
+      "step": 4077
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903491124898316e-05,
+      "loss": 0.6318,
+      "step": 4078
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9034444500578636e-05,
+      "loss": 0.6493,
+      "step": 4079
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9033977645059342e-05,
+      "loss": 0.6346,
+      "step": 4080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9033510682430814e-05,
+      "loss": 0.6803,
+      "step": 4081
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9033043612698592e-05,
+      "loss": 0.673,
+      "step": 4082
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903257643586821e-05,
+      "loss": 0.6357,
+      "step": 4083
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9032109151945207e-05,
+      "loss": 0.6437,
+      "step": 4084
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9031641760935126e-05,
+      "loss": 0.5928,
+      "step": 4085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9031174262843506e-05,
+      "loss": 0.6547,
+      "step": 4086
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9030706657675893e-05,
+      "loss": 0.6781,
+      "step": 4087
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9030238945437823e-05,
+      "loss": 0.6277,
+      "step": 4088
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9029771126134854e-05,
+      "loss": 0.6325,
+      "step": 4089
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9029303199772523e-05,
+      "loss": 0.713,
+      "step": 4090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902883516635638e-05,
+      "loss": 0.655,
+      "step": 4091
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9028367025891973e-05,
+      "loss": 0.6807,
+      "step": 4092
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9027898778384856e-05,
+      "loss": 0.6796,
+      "step": 4093
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9027430423840578e-05,
+      "loss": 0.6629,
+      "step": 4094
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902696196226469e-05,
+      "loss": 0.7129,
+      "step": 4095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9026493393662752e-05,
+      "loss": 0.6411,
+      "step": 4096
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902602471804032e-05,
+      "loss": 0.6645,
+      "step": 4097
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902555593540294e-05,
+      "loss": 0.6381,
+      "step": 4098
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9025087045756177e-05,
+      "loss": 0.6052,
+      "step": 4099
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9024618049105593e-05,
+      "loss": 0.6606,
+      "step": 4100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9024148945456747e-05,
+      "loss": 0.6137,
+      "step": 4101
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9023679734815198e-05,
+      "loss": 0.6335,
+      "step": 4102
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902321041718651e-05,
+      "loss": 0.6881,
+      "step": 4103
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902274099257625e-05,
+      "loss": 0.658,
+      "step": 4104
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902227146098998e-05,
+      "loss": 0.592,
+      "step": 4105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902180182243327e-05,
+      "loss": 0.6976,
+      "step": 4106
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9021332076911684e-05,
+      "loss": 0.6376,
+      "step": 4107
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9020862224430797e-05,
+      "loss": 0.6188,
+      "step": 4108
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.902039226499618e-05,
+      "loss": 0.6052,
+      "step": 4109
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9019922198613394e-05,
+      "loss": 0.5609,
+      "step": 4110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9019452025288025e-05,
+      "loss": 0.6355,
+      "step": 4111
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9018981745025644e-05,
+      "loss": 0.6591,
+      "step": 4112
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.901851135783182e-05,
+      "loss": 0.6937,
+      "step": 4113
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.901804086371214e-05,
+      "loss": 0.6168,
+      "step": 4114
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9017570262672175e-05,
+      "loss": 0.656,
+      "step": 4115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9017099554717512e-05,
+      "loss": 0.6188,
+      "step": 4116
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.901662873985372e-05,
+      "loss": 0.6604,
+      "step": 4117
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.901615781808639e-05,
+      "loss": 0.6229,
+      "step": 4118
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9015686789421107e-05,
+      "loss": 0.6513,
+      "step": 4119
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.901521565386345e-05,
+      "loss": 0.6697,
+      "step": 4120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.901474441141901e-05,
+      "loss": 0.6556,
+      "step": 4121
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9014273062093368e-05,
+      "loss": 0.656,
+      "step": 4122
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9013801605892113e-05,
+      "loss": 0.6601,
+      "step": 4123
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9013330042820844e-05,
+      "loss": 0.6682,
+      "step": 4124
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.901285837288514e-05,
+      "loss": 0.6552,
+      "step": 4125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9012386596090603e-05,
+      "loss": 0.6507,
+      "step": 4126
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9011914712442822e-05,
+      "loss": 0.6359,
+      "step": 4127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9011442721947388e-05,
+      "loss": 0.6273,
+      "step": 4128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9010970624609904e-05,
+      "loss": 0.6152,
+      "step": 4129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9010498420435962e-05,
+      "loss": 0.6587,
+      "step": 4130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9010026109431165e-05,
+      "loss": 0.6307,
+      "step": 4131
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.900955369160111e-05,
+      "loss": 0.6808,
+      "step": 4132
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90090811669514e-05,
+      "loss": 0.6513,
+      "step": 4133
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9008608535487634e-05,
+      "loss": 0.6641,
+      "step": 4134
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.900813579721542e-05,
+      "loss": 0.6858,
+      "step": 4135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.900766295214036e-05,
+      "loss": 0.6285,
+      "step": 4136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9007190000268058e-05,
+      "loss": 0.6544,
+      "step": 4137
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9006716941604128e-05,
+      "loss": 0.6587,
+      "step": 4138
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9006243776154174e-05,
+      "loss": 0.6401,
+      "step": 4139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.900577050392381e-05,
+      "loss": 0.601,
+      "step": 4140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.900529712491864e-05,
+      "loss": 0.6526,
+      "step": 4141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9004823639144283e-05,
+      "loss": 0.6794,
+      "step": 4142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9004350046606352e-05,
+      "loss": 0.6188,
+      "step": 4143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9003876347310455e-05,
+      "loss": 0.573,
+      "step": 4144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.900340254126222e-05,
+      "loss": 0.6221,
+      "step": 4145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9002928628467256e-05,
+      "loss": 0.6362,
+      "step": 4146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9002454608931183e-05,
+      "loss": 0.6268,
+      "step": 4147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9001980482659625e-05,
+      "loss": 0.6159,
+      "step": 4148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9001506249658198e-05,
+      "loss": 0.5981,
+      "step": 4149
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.900103190993253e-05,
+      "loss": 0.6814,
+      "step": 4150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9000557463488244e-05,
+      "loss": 0.564,
+      "step": 4151
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.900008291033096e-05,
+      "loss": 0.6632,
+      "step": 4152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899960825046631e-05,
+      "loss": 0.6754,
+      "step": 4153
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8999133483899917e-05,
+      "loss": 0.6119,
+      "step": 4154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8998658610637415e-05,
+      "loss": 0.6749,
+      "step": 4155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899818363068443e-05,
+      "loss": 0.7039,
+      "step": 4156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8997708544046597e-05,
+      "loss": 0.6297,
+      "step": 4157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899723335072955e-05,
+      "loss": 0.6651,
+      "step": 4158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899675805073891e-05,
+      "loss": 0.6856,
+      "step": 4159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899628264408033e-05,
+      "loss": 0.5886,
+      "step": 4160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899580713075944e-05,
+      "loss": 0.6282,
+      "step": 4161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8995331510781873e-05,
+      "loss": 0.6643,
+      "step": 4162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8994855784153274e-05,
+      "loss": 0.6436,
+      "step": 4163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899437995087928e-05,
+      "loss": 0.6595,
+      "step": 4164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8993904010965535e-05,
+      "loss": 0.6276,
+      "step": 4165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8993427964417677e-05,
+      "loss": 0.6201,
+      "step": 4166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899295181124136e-05,
+      "loss": 0.6319,
+      "step": 4167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8992475551442216e-05,
+      "loss": 0.6688,
+      "step": 4168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8991999185025902e-05,
+      "loss": 0.5875,
+      "step": 4169
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8991522711998063e-05,
+      "loss": 0.6285,
+      "step": 4170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.899104613236435e-05,
+      "loss": 0.692,
+      "step": 4171
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8990569446130406e-05,
+      "loss": 0.6631,
+      "step": 4172
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8990092653301894e-05,
+      "loss": 0.6737,
+      "step": 4173
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.898961575388446e-05,
+      "loss": 0.6217,
+      "step": 4174
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8989138747883754e-05,
+      "loss": 0.6948,
+      "step": 4175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8988661635305444e-05,
+      "loss": 0.6258,
+      "step": 4176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8988184416155173e-05,
+      "loss": 0.6474,
+      "step": 4177
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.898770709043861e-05,
+      "loss": 0.6622,
+      "step": 4178
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8987229658161408e-05,
+      "loss": 0.619,
+      "step": 4179
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.898675211932923e-05,
+      "loss": 0.6656,
+      "step": 4180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8986274473947738e-05,
+      "loss": 0.6032,
+      "step": 4181
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8985796722022596e-05,
+      "loss": 0.6324,
+      "step": 4182
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8985318863559463e-05,
+      "loss": 0.681,
+      "step": 4183
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.898484089856401e-05,
+      "loss": 0.6984,
+      "step": 4184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8984362827041903e-05,
+      "loss": 0.6549,
+      "step": 4185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898388464899881e-05,
+      "loss": 0.6517,
+      "step": 4186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8983406364440396e-05,
+      "loss": 0.6544,
+      "step": 4187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8982927973372338e-05,
+      "loss": 0.6675,
+      "step": 4188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8982449475800308e-05,
+      "loss": 0.6113,
+      "step": 4189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8981970871729972e-05,
+      "loss": 0.5703,
+      "step": 4190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898149216116701e-05,
+      "loss": 0.6622,
+      "step": 4191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89810133441171e-05,
+      "loss": 0.6143,
+      "step": 4192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898053442058591e-05,
+      "loss": 0.592,
+      "step": 4193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898005539057913e-05,
+      "loss": 0.5827,
+      "step": 4194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897957625410243e-05,
+      "loss": 0.6293,
+      "step": 4195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8979097011161497e-05,
+      "loss": 0.6419,
+      "step": 4196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897861766176201e-05,
+      "loss": 0.6656,
+      "step": 4197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897813820590965e-05,
+      "loss": 0.6738,
+      "step": 4198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8977658643610106e-05,
+      "loss": 0.6589,
+      "step": 4199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8977178974869065e-05,
+      "loss": 0.5942,
+      "step": 4200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8976699199692206e-05,
+      "loss": 0.6655,
+      "step": 4201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8976219318085224e-05,
+      "loss": 0.6473,
+      "step": 4202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8975739330053807e-05,
+      "loss": 0.618,
+      "step": 4203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8975259235603646e-05,
+      "loss": 0.6405,
+      "step": 4204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8974779034740432e-05,
+      "loss": 0.6672,
+      "step": 4205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8974298727469863e-05,
+      "loss": 0.5964,
+      "step": 4206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8973818313797627e-05,
+      "loss": 0.6645,
+      "step": 4207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897333779372942e-05,
+      "loss": 0.6293,
+      "step": 4208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897285716727095e-05,
+      "loss": 0.7272,
+      "step": 4209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89723764344279e-05,
+      "loss": 0.6197,
+      "step": 4210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897189559520598e-05,
+      "loss": 0.6793,
+      "step": 4211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897141464961089e-05,
+      "loss": 0.6443,
+      "step": 4212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8970933597648326e-05,
+      "loss": 0.6052,
+      "step": 4213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8970452439324002e-05,
+      "loss": 0.7044,
+      "step": 4214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8969971174643615e-05,
+      "loss": 0.6483,
+      "step": 4215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8969489803612868e-05,
+      "loss": 0.6647,
+      "step": 4216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8969008326237473e-05,
+      "loss": 0.6118,
+      "step": 4217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8968526742523147e-05,
+      "loss": 0.6658,
+      "step": 4218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8968045052475583e-05,
+      "loss": 0.6035,
+      "step": 4219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8967563256100503e-05,
+      "loss": 0.6004,
+      "step": 4220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8967081353403614e-05,
+      "loss": 0.6045,
+      "step": 4221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8966599344390636e-05,
+      "loss": 0.6089,
+      "step": 4222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.896611722906728e-05,
+      "loss": 0.6888,
+      "step": 4223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.896563500743926e-05,
+      "loss": 0.6392,
+      "step": 4224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89651526795123e-05,
+      "loss": 0.7264,
+      "step": 4225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8964670245292108e-05,
+      "loss": 0.6189,
+      "step": 4226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.896418770478441e-05,
+      "loss": 0.6222,
+      "step": 4227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.896370505799493e-05,
+      "loss": 0.6814,
+      "step": 4228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8963222304929386e-05,
+      "loss": 0.622,
+      "step": 4229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8962739445593504e-05,
+      "loss": 0.6694,
+      "step": 4230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8962256479993008e-05,
+      "loss": 0.6516,
+      "step": 4231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8961773408133625e-05,
+      "loss": 0.653,
+      "step": 4232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8961290230021077e-05,
+      "loss": 0.6049,
+      "step": 4233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89608069456611e-05,
+      "loss": 0.5989,
+      "step": 4234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.896032355505942e-05,
+      "loss": 0.6146,
+      "step": 4235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8959840058221772e-05,
+      "loss": 0.6547,
+      "step": 4236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8959356455153886e-05,
+      "loss": 0.6499,
+      "step": 4237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8958872745861495e-05,
+      "loss": 0.7389,
+      "step": 4238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8958388930350332e-05,
+      "loss": 0.6316,
+      "step": 4239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.895790500862614e-05,
+      "loss": 0.6885,
+      "step": 4240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.895742098069465e-05,
+      "loss": 0.6124,
+      "step": 4241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89569368465616e-05,
+      "loss": 0.6298,
+      "step": 4242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.895645260623274e-05,
+      "loss": 0.6333,
+      "step": 4243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89559682597138e-05,
+      "loss": 0.656,
+      "step": 4244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8955483807010524e-05,
+      "loss": 0.6397,
+      "step": 4245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8954999248128664e-05,
+      "loss": 0.5894,
+      "step": 4246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8954514583073958e-05,
+      "loss": 0.6656,
+      "step": 4247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8954029811852153e-05,
+      "loss": 0.5952,
+      "step": 4248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8953544934469e-05,
+      "loss": 0.6878,
+      "step": 4249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8953059950930243e-05,
+      "loss": 0.7441,
+      "step": 4250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8952574861241638e-05,
+      "loss": 0.6165,
+      "step": 4251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8952089665408932e-05,
+      "loss": 0.6481,
+      "step": 4252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8951604363437874e-05,
+      "loss": 0.6031,
+      "step": 4253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8951118955334226e-05,
+      "loss": 0.677,
+      "step": 4254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8950633441103742e-05,
+      "loss": 0.6364,
+      "step": 4255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.895014782075217e-05,
+      "loss": 0.6256,
+      "step": 4256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8949662094285276e-05,
+      "loss": 0.6145,
+      "step": 4257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.894917626170882e-05,
+      "loss": 0.6472,
+      "step": 4258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8948690323028557e-05,
+      "loss": 0.681,
+      "step": 4259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8948204278250245e-05,
+      "loss": 0.6159,
+      "step": 4260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8947718127379658e-05,
+      "loss": 0.677,
+      "step": 4261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.894723187042255e-05,
+      "loss": 0.6159,
+      "step": 4262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8946745507384695e-05,
+      "loss": 0.6841,
+      "step": 4263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.894625903827185e-05,
+      "loss": 0.6628,
+      "step": 4264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8945772463089786e-05,
+      "loss": 0.6127,
+      "step": 4265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8945285781844274e-05,
+      "loss": 0.6559,
+      "step": 4266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8944798994541085e-05,
+      "loss": 0.6338,
+      "step": 4267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8944312101185987e-05,
+      "loss": 0.6606,
+      "step": 4268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8943825101784752e-05,
+      "loss": 0.6752,
+      "step": 4269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.894333799634316e-05,
+      "loss": 0.6199,
+      "step": 4270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8942850784866988e-05,
+      "loss": 0.5722,
+      "step": 4271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8942363467361998e-05,
+      "loss": 0.7209,
+      "step": 4272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.894187604383398e-05,
+      "loss": 0.6353,
+      "step": 4273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.894138851428871e-05,
+      "loss": 0.5975,
+      "step": 4274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.894090087873197e-05,
+      "loss": 0.5835,
+      "step": 4275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8940413137169537e-05,
+      "loss": 0.618,
+      "step": 4276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89399252896072e-05,
+      "loss": 0.6829,
+      "step": 4277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8939437336050735e-05,
+      "loss": 0.6055,
+      "step": 4278
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8938949276505933e-05,
+      "loss": 0.5941,
+      "step": 4279
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.893846111097858e-05,
+      "loss": 0.6655,
+      "step": 4280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8937972839474466e-05,
+      "loss": 0.6436,
+      "step": 4281
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8937484461999375e-05,
+      "loss": 0.6575,
+      "step": 4282
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89369959785591e-05,
+      "loss": 0.6134,
+      "step": 4283
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8936507389159435e-05,
+      "loss": 0.6771,
+      "step": 4284
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.893601869380617e-05,
+      "loss": 0.6386,
+      "step": 4285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8935529892505092e-05,
+      "loss": 0.617,
+      "step": 4286
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8935040985262013e-05,
+      "loss": 0.6482,
+      "step": 4287
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8934551972082714e-05,
+      "loss": 0.6253,
+      "step": 4288
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8934062852973e-05,
+      "loss": 0.6816,
+      "step": 4289
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.893357362793867e-05,
+      "loss": 0.6939,
+      "step": 4290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8933084296985526e-05,
+      "loss": 0.631,
+      "step": 4291
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8932594860119365e-05,
+      "loss": 0.6405,
+      "step": 4292
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.893210531734599e-05,
+      "loss": 0.6074,
+      "step": 4293
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.893161566867121e-05,
+      "loss": 0.6463,
+      "step": 4294
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8931125914100827e-05,
+      "loss": 0.6453,
+      "step": 4295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8930636053640648e-05,
+      "loss": 0.6458,
+      "step": 4296
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8930146087296482e-05,
+      "loss": 0.6064,
+      "step": 4297
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.892965601507414e-05,
+      "loss": 0.6245,
+      "step": 4298
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8929165836979428e-05,
+      "loss": 0.6141,
+      "step": 4299
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8928675553018157e-05,
+      "loss": 0.6584,
+      "step": 4300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8928185163196146e-05,
+      "loss": 0.6928,
+      "step": 4301
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8927694667519204e-05,
+      "loss": 0.6837,
+      "step": 4302
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8927204065993145e-05,
+      "loss": 0.5881,
+      "step": 4303
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8926713358623794e-05,
+      "loss": 0.6035,
+      "step": 4304
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.892622254541696e-05,
+      "loss": 0.6291,
+      "step": 4305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8925731626378468e-05,
+      "loss": 0.6199,
+      "step": 4306
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8925240601514138e-05,
+      "loss": 0.6259,
+      "step": 4307
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8924749470829785e-05,
+      "loss": 0.6371,
+      "step": 4308
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.892425823433124e-05,
+      "loss": 0.6955,
+      "step": 4309
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8923766892024326e-05,
+      "loss": 0.6374,
+      "step": 4310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8923275443914867e-05,
+      "loss": 0.6175,
+      "step": 4311
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8922783890008686e-05,
+      "loss": 0.6878,
+      "step": 4312
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8922292230311618e-05,
+      "loss": 0.6107,
+      "step": 4313
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8921800464829488e-05,
+      "loss": 0.638,
+      "step": 4314
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8921308593568126e-05,
+      "loss": 0.6852,
+      "step": 4315
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.892081661653337e-05,
+      "loss": 0.665,
+      "step": 4316
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.892032453373104e-05,
+      "loss": 0.6673,
+      "step": 4317
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8919832345166985e-05,
+      "loss": 0.6803,
+      "step": 4318
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8919340050847032e-05,
+      "loss": 0.6089,
+      "step": 4319
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.891884765077702e-05,
+      "loss": 0.5858,
+      "step": 4320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8918355144962782e-05,
+      "loss": 0.5712,
+      "step": 4321
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8917862533410166e-05,
+      "loss": 0.6607,
+      "step": 4322
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.891736981612501e-05,
+      "loss": 0.6221,
+      "step": 4323
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8916876993113153e-05,
+      "loss": 0.6366,
+      "step": 4324
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8916384064380437e-05,
+      "loss": 0.6863,
+      "step": 4325
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8915891029932713e-05,
+      "loss": 0.6455,
+      "step": 4326
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.891539788977582e-05,
+      "loss": 0.6925,
+      "step": 4327
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8914904643915607e-05,
+      "loss": 0.6613,
+      "step": 4328
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.891441129235792e-05,
+      "loss": 0.6849,
+      "step": 4329
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8913917835108612e-05,
+      "loss": 0.6883,
+      "step": 4330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8913424272173536e-05,
+      "loss": 0.6859,
+      "step": 4331
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8912930603558536e-05,
+      "loss": 0.6543,
+      "step": 4332
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8912436829269464e-05,
+      "loss": 0.6374,
+      "step": 4333
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.891194294931219e-05,
+      "loss": 0.6377,
+      "step": 4334
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8911448963692547e-05,
+      "loss": 0.5917,
+      "step": 4335
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.891095487241641e-05,
+      "loss": 0.7067,
+      "step": 4336
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8910460675489628e-05,
+      "loss": 0.6268,
+      "step": 4337
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8909966372918067e-05,
+      "loss": 0.6289,
+      "step": 4338
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.890947196470758e-05,
+      "loss": 0.6673,
+      "step": 4339
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8908977450864032e-05,
+      "loss": 0.673,
+      "step": 4340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8908482831393288e-05,
+      "loss": 0.6226,
+      "step": 4341
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.890798810630121e-05,
+      "loss": 0.5689,
+      "step": 4342
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8907493275593664e-05,
+      "loss": 0.6906,
+      "step": 4343
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.890699833927652e-05,
+      "loss": 0.6734,
+      "step": 4344
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.890650329735564e-05,
+      "loss": 0.6492,
+      "step": 4345
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89060081498369e-05,
+      "loss": 0.6499,
+      "step": 4346
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8905512896726165e-05,
+      "loss": 0.623,
+      "step": 4347
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.890501753802931e-05,
+      "loss": 0.5874,
+      "step": 4348
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8904522073752204e-05,
+      "loss": 0.6331,
+      "step": 4349
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8904026503900725e-05,
+      "loss": 0.6395,
+      "step": 4350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8903530828480752e-05,
+      "loss": 0.6712,
+      "step": 4351
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8903035047498156e-05,
+      "loss": 0.6248,
+      "step": 4352
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8902539160958815e-05,
+      "loss": 0.6903,
+      "step": 4353
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8902043168868612e-05,
+      "loss": 0.6346,
+      "step": 4354
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.890154707123343e-05,
+      "loss": 0.6945,
+      "step": 4355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8901050868059142e-05,
+      "loss": 0.5838,
+      "step": 4356
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8900554559351638e-05,
+      "loss": 0.6672,
+      "step": 4357
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89000581451168e-05,
+      "loss": 0.6411,
+      "step": 4358
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8899561625360513e-05,
+      "loss": 0.6598,
+      "step": 4359
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.889906500008867e-05,
+      "loss": 0.6287,
+      "step": 4360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8898568269307147e-05,
+      "loss": 0.6903,
+      "step": 4361
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8898071433021842e-05,
+      "loss": 0.6893,
+      "step": 4362
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8897574491238647e-05,
+      "loss": 0.6423,
+      "step": 4363
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.889707744396345e-05,
+      "loss": 0.6528,
+      "step": 4364
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8896580291202145e-05,
+      "loss": 0.6276,
+      "step": 4365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8896083032960627e-05,
+      "loss": 0.6621,
+      "step": 4366
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8895585669244785e-05,
+      "loss": 0.6076,
+      "step": 4367
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.889508820006053e-05,
+      "loss": 0.6158,
+      "step": 4368
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.889459062541375e-05,
+      "loss": 0.6009,
+      "step": 4369
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8894092945310345e-05,
+      "loss": 0.6723,
+      "step": 4370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.889359515975622e-05,
+      "loss": 0.6458,
+      "step": 4371
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8893097268757268e-05,
+      "loss": 0.7086,
+      "step": 4372
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8892599272319402e-05,
+      "loss": 0.6264,
+      "step": 4373
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.889210117044852e-05,
+      "loss": 0.6897,
+      "step": 4374
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.889160296315053e-05,
+      "loss": 0.6235,
+      "step": 4375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8891104650431346e-05,
+      "loss": 0.6086,
+      "step": 4376
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8890606232296863e-05,
+      "loss": 0.6784,
+      "step": 4377
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8890107708752996e-05,
+      "loss": 0.6273,
+      "step": 4378
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8889609079805656e-05,
+      "loss": 0.6547,
+      "step": 4379
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8889110345460757e-05,
+      "loss": 0.6157,
+      "step": 4380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.888861150572421e-05,
+      "loss": 0.6341,
+      "step": 4381
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8888112560601928e-05,
+      "loss": 0.671,
+      "step": 4382
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.888761351009983e-05,
+      "loss": 0.6093,
+      "step": 4383
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.888711435422383e-05,
+      "loss": 0.6383,
+      "step": 4384
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8886615092979848e-05,
+      "loss": 0.6059,
+      "step": 4385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8886115726373798e-05,
+      "loss": 0.6007,
+      "step": 4386
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.888561625441161e-05,
+      "loss": 0.6803,
+      "step": 4387
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.88851166770992e-05,
+      "loss": 0.6458,
+      "step": 4388
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8884616994442493e-05,
+      "loss": 0.674,
+      "step": 4389
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.888411720644741e-05,
+      "loss": 0.6404,
+      "step": 4390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8883617313119878e-05,
+      "loss": 0.6791,
+      "step": 4391
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.888311731446583e-05,
+      "loss": 0.6104,
+      "step": 4392
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.888261721049118e-05,
+      "loss": 0.6435,
+      "step": 4393
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8882117001201875e-05,
+      "loss": 0.5999,
+      "step": 4394
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8881616686603836e-05,
+      "loss": 0.671,
+      "step": 4395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8881116266702992e-05,
+      "loss": 0.7054,
+      "step": 4396
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8880615741505284e-05,
+      "loss": 0.6924,
+      "step": 4397
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8880115111016638e-05,
+      "loss": 0.6334,
+      "step": 4398
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8879614375242998e-05,
+      "loss": 0.5868,
+      "step": 4399
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8879113534190296e-05,
+      "loss": 0.6344,
+      "step": 4400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8878612587864473e-05,
+      "loss": 0.623,
+      "step": 4401
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8878111536271458e-05,
+      "loss": 0.628,
+      "step": 4402
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.887761037941721e-05,
+      "loss": 0.5686,
+      "step": 4403
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8877109117307653e-05,
+      "loss": 0.5985,
+      "step": 4404
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8876607749948742e-05,
+      "loss": 0.6493,
+      "step": 4405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8876106277346414e-05,
+      "loss": 0.6479,
+      "step": 4406
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8875604699506622e-05,
+      "loss": 0.613,
+      "step": 4407
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8875103016435305e-05,
+      "loss": 0.6591,
+      "step": 4408
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8874601228138417e-05,
+      "loss": 0.6735,
+      "step": 4409
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8874099334621904e-05,
+      "loss": 0.5728,
+      "step": 4410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.887359733589172e-05,
+      "loss": 0.6601,
+      "step": 4411
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.887309523195381e-05,
+      "loss": 0.6768,
+      "step": 4412
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.887259302281413e-05,
+      "loss": 0.6327,
+      "step": 4413
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.887209070847864e-05,
+      "loss": 0.6291,
+      "step": 4414
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8871588288953294e-05,
+      "loss": 0.6538,
+      "step": 4415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8871085764244044e-05,
+      "loss": 0.6636,
+      "step": 4416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.887058313435685e-05,
+      "loss": 0.6006,
+      "step": 4417
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.887008039929767e-05,
+      "loss": 0.6389,
+      "step": 4418
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8869577559072466e-05,
+      "loss": 0.6179,
+      "step": 4419
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8869074613687203e-05,
+      "loss": 0.5969,
+      "step": 4420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8868571563147834e-05,
+      "loss": 0.6172,
+      "step": 4421
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.886806840746034e-05,
+      "loss": 0.6062,
+      "step": 4422
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.886756514663067e-05,
+      "loss": 0.6441,
+      "step": 4423
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8867061780664803e-05,
+      "loss": 0.621,
+      "step": 4424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8866558309568694e-05,
+      "loss": 0.6762,
+      "step": 4425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.886605473334833e-05,
+      "loss": 0.6155,
+      "step": 4426
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8865551052009663e-05,
+      "loss": 0.6206,
+      "step": 4427
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8865047265558675e-05,
+      "loss": 0.6616,
+      "step": 4428
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.886454337400134e-05,
+      "loss": 0.6188,
+      "step": 4429
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8864039377343628e-05,
+      "loss": 0.608,
+      "step": 4430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8863535275591516e-05,
+      "loss": 0.642,
+      "step": 4431
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.886303106875098e-05,
+      "loss": 0.6063,
+      "step": 4432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8862526756828e-05,
+      "loss": 0.6434,
+      "step": 4433
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8862022339828555e-05,
+      "loss": 0.6001,
+      "step": 4434
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8861517817758623e-05,
+      "loss": 0.636,
+      "step": 4435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.886101319062419e-05,
+      "loss": 0.6476,
+      "step": 4436
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8860508458431235e-05,
+      "loss": 0.699,
+      "step": 4437
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8860003621185744e-05,
+      "loss": 0.657,
+      "step": 4438
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.88594986788937e-05,
+      "loss": 0.6434,
+      "step": 4439
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8858993631561092e-05,
+      "loss": 0.6641,
+      "step": 4440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.885848847919391e-05,
+      "loss": 0.597,
+      "step": 4441
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.885798322179814e-05,
+      "loss": 0.5973,
+      "step": 4442
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8857477859379775e-05,
+      "loss": 0.5977,
+      "step": 4443
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8856972391944804e-05,
+      "loss": 0.6285,
+      "step": 4444
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8856466819499218e-05,
+      "loss": 0.6459,
+      "step": 4445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8855961142049018e-05,
+      "loss": 0.6266,
+      "step": 4446
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8855455359600197e-05,
+      "loss": 0.6671,
+      "step": 4447
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.885494947215875e-05,
+      "loss": 0.6866,
+      "step": 4448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8854443479730674e-05,
+      "loss": 0.6193,
+      "step": 4449
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.885393738232197e-05,
+      "loss": 0.6525,
+      "step": 4450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8853431179938634e-05,
+      "loss": 0.652,
+      "step": 4451
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8852924872586677e-05,
+      "loss": 0.5822,
+      "step": 4452
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8852418460272095e-05,
+      "loss": 0.6193,
+      "step": 4453
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8851911943000894e-05,
+      "loss": 0.6075,
+      "step": 4454
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.885140532077908e-05,
+      "loss": 0.6229,
+      "step": 4455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8850898593612657e-05,
+      "loss": 0.6761,
+      "step": 4456
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8850391761507638e-05,
+      "loss": 0.6135,
+      "step": 4457
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8849884824470023e-05,
+      "loss": 0.6637,
+      "step": 4458
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8849377782505835e-05,
+      "loss": 0.6291,
+      "step": 4459
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8848870635621075e-05,
+      "loss": 0.5853,
+      "step": 4460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8848363383821758e-05,
+      "loss": 0.7012,
+      "step": 4461
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8847856027113903e-05,
+      "loss": 0.6112,
+      "step": 4462
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8847348565503522e-05,
+      "loss": 0.5669,
+      "step": 4463
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.884684099899663e-05,
+      "loss": 0.6863,
+      "step": 4464
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8846333327599252e-05,
+      "loss": 0.6849,
+      "step": 4465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8845825551317394e-05,
+      "loss": 0.6629,
+      "step": 4466
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.884531767015709e-05,
+      "loss": 0.6243,
+      "step": 4467
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.884480968412435e-05,
+      "loss": 0.6144,
+      "step": 4468
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8844301593225208e-05,
+      "loss": 0.6015,
+      "step": 4469
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.884379339746568e-05,
+      "loss": 0.5922,
+      "step": 4470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8843285096851796e-05,
+      "loss": 0.6388,
+      "step": 4471
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8842776691389576e-05,
+      "loss": 0.6173,
+      "step": 4472
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8842268181085056e-05,
+      "loss": 0.6087,
+      "step": 4473
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8841759565944258e-05,
+      "loss": 0.6534,
+      "step": 4474
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.884125084597322e-05,
+      "loss": 0.6406,
+      "step": 4475
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8840742021177964e-05,
+      "loss": 0.7058,
+      "step": 4476
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.884023309156453e-05,
+      "loss": 0.6814,
+      "step": 4477
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883972405713895e-05,
+      "loss": 0.6511,
+      "step": 4478
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883921491790726e-05,
+      "loss": 0.6577,
+      "step": 4479
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8838705673875496e-05,
+      "loss": 0.6537,
+      "step": 4480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883819632504969e-05,
+      "loss": 0.6587,
+      "step": 4481
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883768687143589e-05,
+      "loss": 0.6319,
+      "step": 4482
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8837177313040135e-05,
+      "loss": 0.6394,
+      "step": 4483
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883666764986846e-05,
+      "loss": 0.6163,
+      "step": 4484
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883615788192691e-05,
+      "loss": 0.6488,
+      "step": 4485
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8835648009221534e-05,
+      "loss": 0.6313,
+      "step": 4486
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8835138031758375e-05,
+      "loss": 0.6001,
+      "step": 4487
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8834627949543475e-05,
+      "loss": 0.6135,
+      "step": 4488
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8834117762582885e-05,
+      "loss": 0.6066,
+      "step": 4489
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8833607470882652e-05,
+      "loss": 0.6166,
+      "step": 4490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883309707444883e-05,
+      "loss": 0.6219,
+      "step": 4491
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883258657328747e-05,
+      "loss": 0.6499,
+      "step": 4492
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883207596740462e-05,
+      "loss": 0.6376,
+      "step": 4493
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8831565256806338e-05,
+      "loss": 0.6352,
+      "step": 4494
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.883105444149868e-05,
+      "loss": 0.6154,
+      "step": 4495
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.88305435214877e-05,
+      "loss": 0.6241,
+      "step": 4496
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8830032496779452e-05,
+      "loss": 0.6692,
+      "step": 4497
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882952136738e-05,
+      "loss": 0.6401,
+      "step": 4498
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8829010133295403e-05,
+      "loss": 0.6696,
+      "step": 4499
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8828498794531723e-05,
+      "loss": 0.7034,
+      "step": 4500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8827987351095022e-05,
+      "loss": 0.699,
+      "step": 4501
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882747580299136e-05,
+      "loss": 0.5925,
+      "step": 4502
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882696415022681e-05,
+      "loss": 0.6076,
+      "step": 4503
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8826452392807432e-05,
+      "loss": 0.6311,
+      "step": 4504
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8825940530739296e-05,
+      "loss": 0.6197,
+      "step": 4505
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882542856402847e-05,
+      "loss": 0.649,
+      "step": 4506
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8824916492681024e-05,
+      "loss": 0.6053,
+      "step": 4507
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882440431670303e-05,
+      "loss": 0.6143,
+      "step": 4508
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882389203610056e-05,
+      "loss": 0.687,
+      "step": 4509
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8823379650879686e-05,
+      "loss": 0.6606,
+      "step": 4510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882286716104648e-05,
+      "loss": 0.6303,
+      "step": 4511
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8822354566607032e-05,
+      "loss": 0.6182,
+      "step": 4512
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8821841867567406e-05,
+      "loss": 0.6272,
+      "step": 4513
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8821329063933686e-05,
+      "loss": 0.6385,
+      "step": 4514
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882081615571195e-05,
+      "loss": 0.6478,
+      "step": 4515
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.882030314290828e-05,
+      "loss": 0.6246,
+      "step": 4516
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.881979002552876e-05,
+      "loss": 0.6197,
+      "step": 4517
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.881927680357947e-05,
+      "loss": 0.6118,
+      "step": 4518
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.88187634770665e-05,
+      "loss": 0.6615,
+      "step": 4519
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.881825004599593e-05,
+      "loss": 0.6255,
+      "step": 4520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8817736510373854e-05,
+      "loss": 0.6337,
+      "step": 4521
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8817222870206353e-05,
+      "loss": 0.6347,
+      "step": 4522
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8816709125499522e-05,
+      "loss": 0.6617,
+      "step": 4523
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8816195276259454e-05,
+      "loss": 0.6327,
+      "step": 4524
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8815681322492238e-05,
+      "loss": 0.6751,
+      "step": 4525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8815167264203968e-05,
+      "loss": 0.635,
+      "step": 4526
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8814653101400736e-05,
+      "loss": 0.6233,
+      "step": 4527
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.881413883408864e-05,
+      "loss": 0.6467,
+      "step": 4528
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8813624462273782e-05,
+      "loss": 0.6309,
+      "step": 4529
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8813109985962254e-05,
+      "loss": 0.6967,
+      "step": 4530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.881259540516016e-05,
+      "loss": 0.6916,
+      "step": 4531
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.88120807198736e-05,
+      "loss": 0.6608,
+      "step": 4532
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8811565930108674e-05,
+      "loss": 0.6772,
+      "step": 4533
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.881105103587149e-05,
+      "loss": 0.5767,
+      "step": 4534
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.881053603716815e-05,
+      "loss": 0.6495,
+      "step": 4535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8810020934004755e-05,
+      "loss": 0.5983,
+      "step": 4536
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8809505726387423e-05,
+      "loss": 0.6248,
+      "step": 4537
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.880899041432225e-05,
+      "loss": 0.6307,
+      "step": 4538
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.880847499781536e-05,
+      "loss": 0.6595,
+      "step": 4539
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.880795947687285e-05,
+      "loss": 0.6481,
+      "step": 4540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.880744385150084e-05,
+      "loss": 0.6185,
+      "step": 4541
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8806928121705442e-05,
+      "loss": 0.6153,
+      "step": 4542
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.880641228749277e-05,
+      "loss": 0.5983,
+      "step": 4543
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8805896348868942e-05,
+      "loss": 0.6498,
+      "step": 4544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.880538030584007e-05,
+      "loss": 0.6225,
+      "step": 4545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8804864158412276e-05,
+      "loss": 0.613,
+      "step": 4546
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8804347906591683e-05,
+      "loss": 0.595,
+      "step": 4547
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8803831550384407e-05,
+      "loss": 0.6815,
+      "step": 4548
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.880331508979657e-05,
+      "loss": 0.6927,
+      "step": 4549
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8802798524834293e-05,
+      "loss": 0.6391,
+      "step": 4550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8802281855503706e-05,
+      "loss": 0.6836,
+      "step": 4551
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8801765081810936e-05,
+      "loss": 0.6605,
+      "step": 4552
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.88012482037621e-05,
+      "loss": 0.6957,
+      "step": 4553
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.880073122136334e-05,
+      "loss": 0.5803,
+      "step": 4554
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8800214134620774e-05,
+      "loss": 0.6001,
+      "step": 4555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.879969694354054e-05,
+      "loss": 0.6333,
+      "step": 4556
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8799179648128768e-05,
+      "loss": 0.6637,
+      "step": 4557
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8798662248391585e-05,
+      "loss": 0.6181,
+      "step": 4558
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8798144744335135e-05,
+      "loss": 0.5929,
+      "step": 4559
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8797627135965547e-05,
+      "loss": 0.6561,
+      "step": 4560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8797109423288962e-05,
+      "loss": 0.6157,
+      "step": 4561
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8796591606311514e-05,
+      "loss": 0.6055,
+      "step": 4562
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8796073685039347e-05,
+      "loss": 0.6434,
+      "step": 4563
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8795555659478596e-05,
+      "loss": 0.6937,
+      "step": 4564
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.879503752963541e-05,
+      "loss": 0.6251,
+      "step": 4565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8794519295515927e-05,
+      "loss": 0.6459,
+      "step": 4566
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8794000957126295e-05,
+      "loss": 0.6614,
+      "step": 4567
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8793482514472653e-05,
+      "loss": 0.5986,
+      "step": 4568
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.879296396756115e-05,
+      "loss": 0.6259,
+      "step": 4569
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8792445316397946e-05,
+      "loss": 0.619,
+      "step": 4570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8791926560989172e-05,
+      "loss": 0.6704,
+      "step": 4571
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.879140770134099e-05,
+      "loss": 0.6914,
+      "step": 4572
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8790888737459546e-05,
+      "loss": 0.603,
+      "step": 4573
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8790369669351e-05,
+      "loss": 0.5983,
+      "step": 4574
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.87898504970215e-05,
+      "loss": 0.6937,
+      "step": 4575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.87893312204772e-05,
+      "loss": 0.6285,
+      "step": 4576
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.878881183972426e-05,
+      "loss": 0.6473,
+      "step": 4577
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.878829235476884e-05,
+      "loss": 0.6365,
+      "step": 4578
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8787772765617096e-05,
+      "loss": 0.6719,
+      "step": 4579
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.878725307227519e-05,
+      "loss": 0.616,
+      "step": 4580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8786733274749283e-05,
+      "loss": 0.6429,
+      "step": 4581
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8786213373045536e-05,
+      "loss": 0.6197,
+      "step": 4582
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8785693367170114e-05,
+      "loss": 0.6349,
+      "step": 4583
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8785173257129188e-05,
+      "loss": 0.6803,
+      "step": 4584
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.878465304292892e-05,
+      "loss": 0.6506,
+      "step": 4585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.878413272457547e-05,
+      "loss": 0.6678,
+      "step": 4586
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.878361230207502e-05,
+      "loss": 0.6773,
+      "step": 4587
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8783091775433732e-05,
+      "loss": 0.6854,
+      "step": 4588
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8782571144657784e-05,
+      "loss": 0.6919,
+      "step": 4589
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8782050409753344e-05,
+      "loss": 0.6507,
+      "step": 4590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8781529570726587e-05,
+      "loss": 0.6797,
+      "step": 4591
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.878100862758369e-05,
+      "loss": 0.6184,
+      "step": 4592
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8780487580330823e-05,
+      "loss": 0.6315,
+      "step": 4593
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8779966428974168e-05,
+      "loss": 0.6429,
+      "step": 4594
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.877944517351991e-05,
+      "loss": 0.6331,
+      "step": 4595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.877892381397422e-05,
+      "loss": 0.6128,
+      "step": 4596
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.877840235034328e-05,
+      "loss": 0.637,
+      "step": 4597
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.877788078263328e-05,
+      "loss": 0.6693,
+      "step": 4598
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8777359110850397e-05,
+      "loss": 0.5934,
+      "step": 4599
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.877683733500082e-05,
+      "loss": 0.6201,
+      "step": 4600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.877631545509073e-05,
+      "loss": 0.6459,
+      "step": 4601
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8775793471126324e-05,
+      "loss": 0.6777,
+      "step": 4602
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.877527138311378e-05,
+      "loss": 0.6089,
+      "step": 4603
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8774749191059295e-05,
+      "loss": 0.6307,
+      "step": 4604
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8774226894969057e-05,
+      "loss": 0.6375,
+      "step": 4605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8773704494849262e-05,
+      "loss": 0.5783,
+      "step": 4606
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8773181990706103e-05,
+      "loss": 0.5945,
+      "step": 4607
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.877265938254577e-05,
+      "loss": 0.6209,
+      "step": 4608
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8772136670374463e-05,
+      "loss": 0.6447,
+      "step": 4609
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8771613854198376e-05,
+      "loss": 0.6229,
+      "step": 4610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8771090934023715e-05,
+      "loss": 0.6713,
+      "step": 4611
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8770567909856675e-05,
+      "loss": 0.6003,
+      "step": 4612
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8770044781703456e-05,
+      "loss": 0.6639,
+      "step": 4613
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.876952154957026e-05,
+      "loss": 0.6266,
+      "step": 4614
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8768998213463298e-05,
+      "loss": 0.6224,
+      "step": 4615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8768474773388766e-05,
+      "loss": 0.6273,
+      "step": 4616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8767951229352873e-05,
+      "loss": 0.6622,
+      "step": 4617
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.876742758136183e-05,
+      "loss": 0.6282,
+      "step": 4618
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8766903829421838e-05,
+      "loss": 0.6765,
+      "step": 4619
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8766379973539113e-05,
+      "loss": 0.6341,
+      "step": 4620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8765856013719863e-05,
+      "loss": 0.6545,
+      "step": 4621
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8765331949970303e-05,
+      "loss": 0.6753,
+      "step": 4622
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8764807782296644e-05,
+      "loss": 0.7012,
+      "step": 4623
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8764283510705098e-05,
+      "loss": 0.6463,
+      "step": 4624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8763759135201887e-05,
+      "loss": 0.6487,
+      "step": 4625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8763234655793226e-05,
+      "loss": 0.5993,
+      "step": 4626
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8762710072485332e-05,
+      "loss": 0.6291,
+      "step": 4627
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8762185385284422e-05,
+      "loss": 0.6347,
+      "step": 4628
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8761660594196724e-05,
+      "loss": 0.6497,
+      "step": 4629
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8761135699228454e-05,
+      "loss": 0.6642,
+      "step": 4630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.876061070038584e-05,
+      "loss": 0.6679,
+      "step": 4631
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.87600855976751e-05,
+      "loss": 0.6631,
+      "step": 4632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8759560391102467e-05,
+      "loss": 0.6467,
+      "step": 4633
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8759035080674162e-05,
+      "loss": 0.631,
+      "step": 4634
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8758509666396417e-05,
+      "loss": 0.6175,
+      "step": 4635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.875798414827546e-05,
+      "loss": 0.6333,
+      "step": 4636
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.875745852631752e-05,
+      "loss": 0.6398,
+      "step": 4637
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8756932800528837e-05,
+      "loss": 0.6353,
+      "step": 4638
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8756406970915634e-05,
+      "loss": 0.6882,
+      "step": 4639
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8755881037484148e-05,
+      "loss": 0.6597,
+      "step": 4640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8755355000240622e-05,
+      "loss": 0.5771,
+      "step": 4641
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.875482885919128e-05,
+      "loss": 0.6046,
+      "step": 4642
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.875430261434237e-05,
+      "loss": 0.6418,
+      "step": 4643
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.875377626570013e-05,
+      "loss": 0.6215,
+      "step": 4644
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8753249813270796e-05,
+      "loss": 0.6367,
+      "step": 4645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8752723257060612e-05,
+      "loss": 0.6234,
+      "step": 4646
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8752196597075824e-05,
+      "loss": 0.6305,
+      "step": 4647
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8751669833322673e-05,
+      "loss": 0.532,
+      "step": 4648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8751142965807406e-05,
+      "loss": 0.6598,
+      "step": 4649
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8750615994536268e-05,
+      "loss": 0.6816,
+      "step": 4650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8750088919515507e-05,
+      "loss": 0.6687,
+      "step": 4651
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874956174075137e-05,
+      "loss": 0.6969,
+      "step": 4652
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8749034458250116e-05,
+      "loss": 0.6736,
+      "step": 4653
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8748507072017988e-05,
+      "loss": 0.6353,
+      "step": 4654
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8747979582061244e-05,
+      "loss": 0.6241,
+      "step": 4655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8747451988386136e-05,
+      "loss": 0.6942,
+      "step": 4656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874692429099892e-05,
+      "loss": 0.6654,
+      "step": 4657
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874639648990585e-05,
+      "loss": 0.6547,
+      "step": 4658
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874586858511318e-05,
+      "loss": 0.6351,
+      "step": 4659
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874534057662718e-05,
+      "loss": 0.6086,
+      "step": 4660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8744812464454105e-05,
+      "loss": 0.6218,
+      "step": 4661
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874428424860021e-05,
+      "loss": 0.663,
+      "step": 4662
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874375592907177e-05,
+      "loss": 0.5909,
+      "step": 4663
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874322750587504e-05,
+      "loss": 0.6045,
+      "step": 4664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8742698979016287e-05,
+      "loss": 0.6865,
+      "step": 4665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8742170348501782e-05,
+      "loss": 0.6368,
+      "step": 4666
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8741641614337783e-05,
+      "loss": 0.6492,
+      "step": 4667
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.874111277653057e-05,
+      "loss": 0.6659,
+      "step": 4668
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8740583835086403e-05,
+      "loss": 0.6404,
+      "step": 4669
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8740054790011558e-05,
+      "loss": 0.6611,
+      "step": 4670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.873952564131231e-05,
+      "loss": 0.6248,
+      "step": 4671
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.873899638899493e-05,
+      "loss": 0.6344,
+      "step": 4672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.873846703306569e-05,
+      "loss": 0.6667,
+      "step": 4673
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.873793757353087e-05,
+      "loss": 0.6642,
+      "step": 4674
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.873740801039675e-05,
+      "loss": 0.6537,
+      "step": 4675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.87368783436696e-05,
+      "loss": 0.6235,
+      "step": 4676
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8736348573355708e-05,
+      "loss": 0.5861,
+      "step": 4677
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8735818699461353e-05,
+      "loss": 0.6135,
+      "step": 4678
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8735288721992816e-05,
+      "loss": 0.6752,
+      "step": 4679
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8734758640956378e-05,
+      "loss": 0.6307,
+      "step": 4680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8734228456358327e-05,
+      "loss": 0.6368,
+      "step": 4681
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8733698168204952e-05,
+      "loss": 0.6201,
+      "step": 4682
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8733167776502535e-05,
+      "loss": 0.6876,
+      "step": 4683
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.873263728125737e-05,
+      "loss": 0.6247,
+      "step": 4684
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.873210668247574e-05,
+      "loss": 0.6363,
+      "step": 4685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8731575980163938e-05,
+      "loss": 0.6401,
+      "step": 4686
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8731045174328258e-05,
+      "loss": 0.6085,
+      "step": 4687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8730514264974994e-05,
+      "loss": 0.6299,
+      "step": 4688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.872998325211044e-05,
+      "loss": 0.7117,
+      "step": 4689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.872945213574089e-05,
+      "loss": 0.6276,
+      "step": 4690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.872892091587264e-05,
+      "loss": 0.6059,
+      "step": 4691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8728389592511993e-05,
+      "loss": 0.6398,
+      "step": 4692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8727858165665244e-05,
+      "loss": 0.6197,
+      "step": 4693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8727326635338696e-05,
+      "loss": 0.6469,
+      "step": 4694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8726795001538654e-05,
+      "loss": 0.606,
+      "step": 4695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8726263264271413e-05,
+      "loss": 0.6706,
+      "step": 4696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8725731423543285e-05,
+      "loss": 0.593,
+      "step": 4697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8725199479360572e-05,
+      "loss": 0.6001,
+      "step": 4698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8724667431729583e-05,
+      "loss": 0.6003,
+      "step": 4699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.872413528065662e-05,
+      "loss": 0.6181,
+      "step": 4700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8723603026148e-05,
+      "loss": 0.6552,
+      "step": 4701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8723070668210036e-05,
+      "loss": 0.6534,
+      "step": 4702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8722538206849027e-05,
+      "loss": 0.6742,
+      "step": 4703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8722005642071298e-05,
+      "loss": 0.7013,
+      "step": 4704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.872147297388316e-05,
+      "loss": 0.6227,
+      "step": 4705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.872094020229092e-05,
+      "loss": 0.7003,
+      "step": 4706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8720407327300907e-05,
+      "loss": 0.6637,
+      "step": 4707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8719874348919433e-05,
+      "loss": 0.6504,
+      "step": 4708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.871934126715282e-05,
+      "loss": 0.6045,
+      "step": 4709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8718808082007385e-05,
+      "loss": 0.6592,
+      "step": 4710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.871827479348945e-05,
+      "loss": 0.6477,
+      "step": 4711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8717741401605338e-05,
+      "loss": 0.6106,
+      "step": 4712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8717207906361375e-05,
+      "loss": 0.6168,
+      "step": 4713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8716674307763884e-05,
+      "loss": 0.6469,
+      "step": 4714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8716140605819192e-05,
+      "loss": 0.6426,
+      "step": 4715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8715606800533627e-05,
+      "loss": 0.611,
+      "step": 4716
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.871507289191352e-05,
+      "loss": 0.663,
+      "step": 4717
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8714538879965198e-05,
+      "loss": 0.6219,
+      "step": 4718
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8714004764694994e-05,
+      "loss": 0.6435,
+      "step": 4719
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.871347054610924e-05,
+      "loss": 0.5847,
+      "step": 4720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.871293622421427e-05,
+      "loss": 0.6238,
+      "step": 4721
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8712401799016418e-05,
+      "loss": 0.6204,
+      "step": 4722
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8711867270522024e-05,
+      "loss": 0.6781,
+      "step": 4723
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.871133263873742e-05,
+      "loss": 0.614,
+      "step": 4724
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.871079790366895e-05,
+      "loss": 0.5951,
+      "step": 4725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8710263065322952e-05,
+      "loss": 0.6438,
+      "step": 4726
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8709728123705765e-05,
+      "loss": 0.6295,
+      "step": 4727
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8709193078823735e-05,
+      "loss": 0.5874,
+      "step": 4728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8708657930683202e-05,
+      "loss": 0.6966,
+      "step": 4729
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8708122679290515e-05,
+      "loss": 0.6314,
+      "step": 4730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8707587324652016e-05,
+      "loss": 0.6398,
+      "step": 4731
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.870705186677405e-05,
+      "loss": 0.6707,
+      "step": 4732
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8706516305662977e-05,
+      "loss": 0.7148,
+      "step": 4733
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.870598064132514e-05,
+      "loss": 0.6196,
+      "step": 4734
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8705444873766882e-05,
+      "loss": 0.6429,
+      "step": 4735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.870490900299457e-05,
+      "loss": 0.5796,
+      "step": 4736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8704373029014547e-05,
+      "loss": 0.6406,
+      "step": 4737
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8703836951833166e-05,
+      "loss": 0.612,
+      "step": 4738
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8703300771456793e-05,
+      "loss": 0.6977,
+      "step": 4739
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.870276448789178e-05,
+      "loss": 0.6248,
+      "step": 4740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.870222810114448e-05,
+      "loss": 0.6579,
+      "step": 4741
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.870169161122126e-05,
+      "loss": 0.643,
+      "step": 4742
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8701155018128484e-05,
+      "loss": 0.6169,
+      "step": 4743
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.87006183218725e-05,
+      "loss": 0.6065,
+      "step": 4744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8700081522459682e-05,
+      "loss": 0.6399,
+      "step": 4745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8699544619896393e-05,
+      "loss": 0.6522,
+      "step": 4746
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8699007614189e-05,
+      "loss": 0.6503,
+      "step": 4747
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8698470505343863e-05,
+      "loss": 0.5965,
+      "step": 4748
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8697933293367357e-05,
+      "loss": 0.6617,
+      "step": 4749
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.869739597826585e-05,
+      "loss": 0.6681,
+      "step": 4750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.869685856004571e-05,
+      "loss": 0.6624,
+      "step": 4751
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.869632103871331e-05,
+      "loss": 0.7247,
+      "step": 4752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.869578341427502e-05,
+      "loss": 0.6279,
+      "step": 4753
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.869524568673722e-05,
+      "loss": 0.7001,
+      "step": 4754
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8694707856106284e-05,
+      "loss": 0.6289,
+      "step": 4755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8694169922388586e-05,
+      "loss": 0.5541,
+      "step": 4756
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8693631885590505e-05,
+      "loss": 0.6316,
+      "step": 4757
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.869309374571842e-05,
+      "loss": 0.6254,
+      "step": 4758
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8692555502778715e-05,
+      "loss": 0.7052,
+      "step": 4759
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8692017156777765e-05,
+      "loss": 0.6439,
+      "step": 4760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.869147870772196e-05,
+      "loss": 0.6367,
+      "step": 4761
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8690940155617676e-05,
+      "loss": 0.6039,
+      "step": 4762
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8690401500471303e-05,
+      "loss": 0.6797,
+      "step": 4763
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.868986274228923e-05,
+      "loss": 0.6068,
+      "step": 4764
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8689323881077838e-05,
+      "loss": 0.6014,
+      "step": 4765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.868878491684352e-05,
+      "loss": 0.6244,
+      "step": 4766
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.868824584959267e-05,
+      "loss": 0.5863,
+      "step": 4767
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8687706679331673e-05,
+      "loss": 0.7018,
+      "step": 4768
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8687167406066917e-05,
+      "loss": 0.6306,
+      "step": 4769
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8686628029804807e-05,
+      "loss": 0.6188,
+      "step": 4770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8686088550551735e-05,
+      "loss": 0.655,
+      "step": 4771
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.868554896831409e-05,
+      "loss": 0.6787,
+      "step": 4772
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.868500928309828e-05,
+      "loss": 0.6495,
+      "step": 4773
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8684469494910694e-05,
+      "loss": 0.651,
+      "step": 4774
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8683929603757738e-05,
+      "loss": 0.618,
+      "step": 4775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8683389609645812e-05,
+      "loss": 0.6496,
+      "step": 4776
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8682849512581315e-05,
+      "loss": 0.609,
+      "step": 4777
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.868230931257066e-05,
+      "loss": 0.6372,
+      "step": 4778
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8681769009620237e-05,
+      "loss": 0.6401,
+      "step": 4779
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8681228603736462e-05,
+      "loss": 0.6011,
+      "step": 4780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8680688094925742e-05,
+      "loss": 0.6102,
+      "step": 4781
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8680147483194484e-05,
+      "loss": 0.6712,
+      "step": 4782
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8679606768549093e-05,
+      "loss": 0.6547,
+      "step": 4783
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.867906595099599e-05,
+      "loss": 0.6247,
+      "step": 4784
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8678525030541578e-05,
+      "loss": 0.6567,
+      "step": 4785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.867798400719227e-05,
+      "loss": 0.6171,
+      "step": 4786
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.867744288095449e-05,
+      "loss": 0.5696,
+      "step": 4787
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8676901651834645e-05,
+      "loss": 0.6543,
+      "step": 4788
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8676360319839153e-05,
+      "loss": 0.6076,
+      "step": 4789
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8675818884974434e-05,
+      "loss": 0.6193,
+      "step": 4790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.867527734724691e-05,
+      "loss": 0.6167,
+      "step": 4791
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8674735706662998e-05,
+      "loss": 0.6661,
+      "step": 4792
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8674193963229116e-05,
+      "loss": 0.6836,
+      "step": 4793
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8673652116951698e-05,
+      "loss": 0.6105,
+      "step": 4794
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.867311016783716e-05,
+      "loss": 0.6166,
+      "step": 4795
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8672568115891926e-05,
+      "loss": 0.6272,
+      "step": 4796
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.867202596112243e-05,
+      "loss": 0.673,
+      "step": 4797
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8671483703535092e-05,
+      "loss": 0.6683,
+      "step": 4798
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8670941343136347e-05,
+      "loss": 0.6155,
+      "step": 4799
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8670398879932626e-05,
+      "loss": 0.6422,
+      "step": 4800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8669856313930358e-05,
+      "loss": 0.6163,
+      "step": 4801
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8669313645135973e-05,
+      "loss": 0.6298,
+      "step": 4802
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.866877087355591e-05,
+      "loss": 0.6707,
+      "step": 4803
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8668227999196605e-05,
+      "loss": 0.6075,
+      "step": 4804
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8667685022064486e-05,
+      "loss": 0.6265,
+      "step": 4805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8667141942165998e-05,
+      "loss": 0.6372,
+      "step": 4806
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8666598759507582e-05,
+      "loss": 0.6325,
+      "step": 4807
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.866605547409567e-05,
+      "loss": 0.7276,
+      "step": 4808
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.866551208593671e-05,
+      "loss": 0.6292,
+      "step": 4809
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8664968595037143e-05,
+      "loss": 0.7128,
+      "step": 4810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.866442500140341e-05,
+      "loss": 0.6886,
+      "step": 4811
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.866388130504196e-05,
+      "loss": 0.6813,
+      "step": 4812
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8663337505959233e-05,
+      "loss": 0.6535,
+      "step": 4813
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8662793604161687e-05,
+      "loss": 0.5794,
+      "step": 4814
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.866224959965576e-05,
+      "loss": 0.6694,
+      "step": 4815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8661705492447905e-05,
+      "loss": 0.6601,
+      "step": 4816
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8661161282544577e-05,
+      "loss": 0.6957,
+      "step": 4817
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8660616969952227e-05,
+      "loss": 0.6236,
+      "step": 4818
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8660072554677304e-05,
+      "loss": 0.7016,
+      "step": 4819
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8659528036726268e-05,
+      "loss": 0.5994,
+      "step": 4820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.865898341610557e-05,
+      "loss": 0.6827,
+      "step": 4821
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8658438692821673e-05,
+      "loss": 0.6862,
+      "step": 4822
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.865789386688103e-05,
+      "loss": 0.6264,
+      "step": 4823
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8657348938290105e-05,
+      "loss": 0.6515,
+      "step": 4824
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8656803907055358e-05,
+      "loss": 0.5885,
+      "step": 4825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8656258773183246e-05,
+      "loss": 0.6854,
+      "step": 4826
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8655713536680235e-05,
+      "loss": 0.6263,
+      "step": 4827
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8655168197552796e-05,
+      "loss": 0.612,
+      "step": 4828
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8654622755807386e-05,
+      "loss": 0.631,
+      "step": 4829
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8654077211450474e-05,
+      "loss": 0.6431,
+      "step": 4830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8653531564488528e-05,
+      "loss": 0.6326,
+      "step": 4831
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8652985814928025e-05,
+      "loss": 0.5824,
+      "step": 4832
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8652439962775423e-05,
+      "loss": 0.6111,
+      "step": 4833
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.86518940080372e-05,
+      "loss": 0.6223,
+      "step": 4834
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8651347950719833e-05,
+      "loss": 0.63,
+      "step": 4835
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.865080179082979e-05,
+      "loss": 0.6183,
+      "step": 4836
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.865025552837355e-05,
+      "loss": 0.6273,
+      "step": 4837
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8649709163357585e-05,
+      "loss": 0.6433,
+      "step": 4838
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8649162695788378e-05,
+      "loss": 0.5928,
+      "step": 4839
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8648616125672405e-05,
+      "loss": 0.638,
+      "step": 4840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8648069453016148e-05,
+      "loss": 0.6406,
+      "step": 4841
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8647522677826086e-05,
+      "loss": 0.6403,
+      "step": 4842
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8646975800108705e-05,
+      "loss": 0.6449,
+      "step": 4843
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.864642881987049e-05,
+      "loss": 0.6871,
+      "step": 4844
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8645881737117915e-05,
+      "loss": 0.6721,
+      "step": 4845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8645334551857485e-05,
+      "loss": 0.6574,
+      "step": 4846
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.864478726409567e-05,
+      "loss": 0.6076,
+      "step": 4847
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.864423987383897e-05,
+      "loss": 0.6159,
+      "step": 4848
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8643692381093873e-05,
+      "loss": 0.5847,
+      "step": 4849
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8643144785866863e-05,
+      "loss": 0.628,
+      "step": 4850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.864259708816444e-05,
+      "loss": 0.6153,
+      "step": 4851
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.86420492879931e-05,
+      "loss": 0.677,
+      "step": 4852
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8641501385359327e-05,
+      "loss": 0.7157,
+      "step": 4853
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.864095338026963e-05,
+      "loss": 0.6272,
+      "step": 4854
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8640405272730496e-05,
+      "loss": 0.6213,
+      "step": 4855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8639857062748428e-05,
+      "loss": 0.6728,
+      "step": 4856
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8639308750329925e-05,
+      "loss": 0.6185,
+      "step": 4857
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8638760335481486e-05,
+      "loss": 0.6255,
+      "step": 4858
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8638211818209616e-05,
+      "loss": 0.6245,
+      "step": 4859
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8637663198520818e-05,
+      "loss": 0.6785,
+      "step": 4860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.86371144764216e-05,
+      "loss": 0.6226,
+      "step": 4861
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8636565651918455e-05,
+      "loss": 0.5986,
+      "step": 4862
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.863601672501791e-05,
+      "loss": 0.6621,
+      "step": 4863
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8635467695726454e-05,
+      "loss": 0.6364,
+      "step": 4864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8634918564050606e-05,
+      "loss": 0.6405,
+      "step": 4865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8634369329996875e-05,
+      "loss": 0.6418,
+      "step": 4866
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8633819993571775e-05,
+      "loss": 0.6417,
+      "step": 4867
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8633270554781818e-05,
+      "loss": 0.6937,
+      "step": 4868
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8632721013633513e-05,
+      "loss": 0.6393,
+      "step": 4869
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8632171370133386e-05,
+      "loss": 0.6328,
+      "step": 4870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8631621624287945e-05,
+      "loss": 0.6575,
+      "step": 4871
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.863107177610371e-05,
+      "loss": 0.5902,
+      "step": 4872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.86305218255872e-05,
+      "loss": 0.6626,
+      "step": 4873
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8629971772744936e-05,
+      "loss": 0.6392,
+      "step": 4874
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.862942161758344e-05,
+      "loss": 0.6641,
+      "step": 4875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.862887136010923e-05,
+      "loss": 0.6244,
+      "step": 4876
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8628321000328845e-05,
+      "loss": 0.6355,
+      "step": 4877
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8627770538248793e-05,
+      "loss": 0.6341,
+      "step": 4878
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8627219973875608e-05,
+      "loss": 0.6511,
+      "step": 4879
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8626669307215815e-05,
+      "loss": 0.6921,
+      "step": 4880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8626118538275947e-05,
+      "loss": 0.6519,
+      "step": 4881
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.862556766706253e-05,
+      "loss": 0.6392,
+      "step": 4882
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.86250166935821e-05,
+      "loss": 0.6559,
+      "step": 4883
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8624465617841182e-05,
+      "loss": 0.6266,
+      "step": 4884
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8623914439846318e-05,
+      "loss": 0.6388,
+      "step": 4885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8623363159604037e-05,
+      "loss": 0.5824,
+      "step": 4886
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8622811777120878e-05,
+      "loss": 0.6654,
+      "step": 4887
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8622260292403377e-05,
+      "loss": 0.6111,
+      "step": 4888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.862170870545807e-05,
+      "loss": 0.6833,
+      "step": 4889
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8621157016291506e-05,
+      "loss": 0.6175,
+      "step": 4890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8620605224910215e-05,
+      "loss": 0.607,
+      "step": 4891
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8620053331320745e-05,
+      "loss": 0.6195,
+      "step": 4892
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.861950133552964e-05,
+      "loss": 0.6027,
+      "step": 4893
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.861894923754344e-05,
+      "loss": 0.608,
+      "step": 4894
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8618397037368698e-05,
+      "loss": 0.623,
+      "step": 4895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8617844735011953e-05,
+      "loss": 0.6617,
+      "step": 4896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8617292330479757e-05,
+      "loss": 0.6446,
+      "step": 4897
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.861673982377866e-05,
+      "loss": 0.6466,
+      "step": 4898
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8616187214915214e-05,
+      "loss": 0.6343,
+      "step": 4899
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8615634503895968e-05,
+      "loss": 0.6784,
+      "step": 4900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8615081690727476e-05,
+      "loss": 0.6467,
+      "step": 4901
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.861452877541629e-05,
+      "loss": 0.5737,
+      "step": 4902
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8613975757968974e-05,
+      "loss": 0.6416,
+      "step": 4903
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8613422638392077e-05,
+      "loss": 0.6959,
+      "step": 4904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8612869416692154e-05,
+      "loss": 0.6301,
+      "step": 4905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8612316092875772e-05,
+      "loss": 0.6246,
+      "step": 4906
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.861176266694949e-05,
+      "loss": 0.6276,
+      "step": 4907
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8611209138919865e-05,
+      "loss": 0.6958,
+      "step": 4908
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8610655508793462e-05,
+      "loss": 0.6177,
+      "step": 4909
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8610101776576844e-05,
+      "loss": 0.6507,
+      "step": 4910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.860954794227658e-05,
+      "loss": 0.6072,
+      "step": 4911
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8608994005899233e-05,
+      "loss": 0.6369,
+      "step": 4912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8608439967451372e-05,
+      "loss": 0.6859,
+      "step": 4913
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8607885826939568e-05,
+      "loss": 0.6586,
+      "step": 4914
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8607331584370384e-05,
+      "loss": 0.6591,
+      "step": 4915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.86067772397504e-05,
+      "loss": 0.576,
+      "step": 4916
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.860622279308618e-05,
+      "loss": 0.6781,
+      "step": 4917
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8605668244384306e-05,
+      "loss": 0.6332,
+      "step": 4918
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8605113593651343e-05,
+      "loss": 0.6963,
+      "step": 4919
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.860455884089388e-05,
+      "loss": 0.6394,
+      "step": 4920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8604003986118483e-05,
+      "loss": 0.6333,
+      "step": 4921
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8603449029331734e-05,
+      "loss": 0.5868,
+      "step": 4922
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8602893970540216e-05,
+      "loss": 0.6599,
+      "step": 4923
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8602338809750504e-05,
+      "loss": 0.6761,
+      "step": 4924
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8601783546969187e-05,
+      "loss": 0.6136,
+      "step": 4925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8601228182202842e-05,
+      "loss": 0.6627,
+      "step": 4926
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8600672715458057e-05,
+      "loss": 0.6485,
+      "step": 4927
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.860011714674142e-05,
+      "loss": 0.6324,
+      "step": 4928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859956147605951e-05,
+      "loss": 0.6264,
+      "step": 4929
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859900570341892e-05,
+      "loss": 0.6234,
+      "step": 4930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8598449828826244e-05,
+      "loss": 0.6551,
+      "step": 4931
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8597893852288066e-05,
+      "loss": 0.6321,
+      "step": 4932
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859733777381098e-05,
+      "loss": 0.6324,
+      "step": 4933
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859678159340158e-05,
+      "loss": 0.6407,
+      "step": 4934
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8596225311066458e-05,
+      "loss": 0.6994,
+      "step": 4935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859566892681221e-05,
+      "loss": 0.6406,
+      "step": 4936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859511244064544e-05,
+      "loss": 0.6509,
+      "step": 4937
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859455585257273e-05,
+      "loss": 0.6904,
+      "step": 4938
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8593999162600692e-05,
+      "loss": 0.5893,
+      "step": 4939
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859344237073592e-05,
+      "loss": 0.6898,
+      "step": 4940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.859288547698502e-05,
+      "loss": 0.645,
+      "step": 4941
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8592328481354592e-05,
+      "loss": 0.6249,
+      "step": 4942
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8591771383851243e-05,
+      "loss": 0.6808,
+      "step": 4943
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8591214184481574e-05,
+      "loss": 0.6308,
+      "step": 4944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8590656883252192e-05,
+      "loss": 0.6344,
+      "step": 4945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8590099480169707e-05,
+      "loss": 0.6396,
+      "step": 4946
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8589541975240726e-05,
+      "loss": 0.6251,
+      "step": 4947
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.858898436847186e-05,
+      "loss": 0.6302,
+      "step": 4948
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8588426659869716e-05,
+      "loss": 0.6778,
+      "step": 4949
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.858786884944091e-05,
+      "loss": 0.5824,
+      "step": 4950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8587310937192057e-05,
+      "loss": 0.6394,
+      "step": 4951
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.858675292312977e-05,
+      "loss": 0.66,
+      "step": 4952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8586194807260663e-05,
+      "loss": 0.64,
+      "step": 4953
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8585636589591356e-05,
+      "loss": 0.6473,
+      "step": 4954
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8585078270128466e-05,
+      "loss": 0.6018,
+      "step": 4955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8584519848878612e-05,
+      "loss": 0.6786,
+      "step": 4956
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.858396132584842e-05,
+      "loss": 0.6455,
+      "step": 4957
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8583402701044502e-05,
+      "loss": 0.6583,
+      "step": 4958
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.858284397447349e-05,
+      "loss": 0.6634,
+      "step": 4959
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8582285146142005e-05,
+      "loss": 0.5984,
+      "step": 4960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.858172621605667e-05,
+      "loss": 0.6435,
+      "step": 4961
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.858116718422412e-05,
+      "loss": 0.6318,
+      "step": 4962
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8580608050650976e-05,
+      "loss": 0.6207,
+      "step": 4963
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.858004881534387e-05,
+      "loss": 0.6283,
+      "step": 4964
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8579489478309426e-05,
+      "loss": 0.6391,
+      "step": 4965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8578930039554285e-05,
+      "loss": 0.6847,
+      "step": 4966
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8578370499085077e-05,
+      "loss": 0.5995,
+      "step": 4967
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8577810856908434e-05,
+      "loss": 0.6289,
+      "step": 4968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8577251113030992e-05,
+      "loss": 0.6787,
+      "step": 4969
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8576691267459388e-05,
+      "loss": 0.6348,
+      "step": 4970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8576131320200258e-05,
+      "loss": 0.6417,
+      "step": 4971
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8575571271260244e-05,
+      "loss": 0.6903,
+      "step": 4972
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8575011120645987e-05,
+      "loss": 0.6281,
+      "step": 4973
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8574450868364123e-05,
+      "loss": 0.6278,
+      "step": 4974
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8573890514421294e-05,
+      "loss": 0.6262,
+      "step": 4975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8573330058824154e-05,
+      "loss": 0.644,
+      "step": 4976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8572769501579336e-05,
+      "loss": 0.6526,
+      "step": 4977
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8572208842693492e-05,
+      "loss": 0.6892,
+      "step": 4978
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8571648082173266e-05,
+      "loss": 0.6126,
+      "step": 4979
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.857108722002531e-05,
+      "loss": 0.6129,
+      "step": 4980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8570526256256275e-05,
+      "loss": 0.678,
+      "step": 4981
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.856996519087281e-05,
+      "loss": 0.6563,
+      "step": 4982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8569404023881562e-05,
+      "loss": 0.6785,
+      "step": 4983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.856884275528919e-05,
+      "loss": 0.6541,
+      "step": 4984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.856828138510235e-05,
+      "loss": 0.6504,
+      "step": 4985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8567719913327696e-05,
+      "loss": 0.6369,
+      "step": 4986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8567158339971882e-05,
+      "loss": 0.5985,
+      "step": 4987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.856659666504157e-05,
+      "loss": 0.637,
+      "step": 4988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8566034888543414e-05,
+      "loss": 0.6113,
+      "step": 4989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8565473010484082e-05,
+      "loss": 0.6329,
+      "step": 4990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8564911030870234e-05,
+      "loss": 0.6476,
+      "step": 4991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8564348949708525e-05,
+      "loss": 0.6659,
+      "step": 4992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.856378676700563e-05,
+      "loss": 0.6513,
+      "step": 4993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8563224482768205e-05,
+      "loss": 0.641,
+      "step": 4994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8562662097002925e-05,
+      "loss": 0.6503,
+      "step": 4995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8562099609716454e-05,
+      "loss": 0.6154,
+      "step": 4996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.856153702091546e-05,
+      "loss": 0.6116,
+      "step": 4997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8560974330606615e-05,
+      "loss": 0.6905,
+      "step": 4998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8560411538796587e-05,
+      "loss": 0.5852,
+      "step": 4999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.855984864549205e-05,
+      "loss": 0.6535,
+      "step": 5000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8559285650699685e-05,
+      "loss": 0.6599,
+      "step": 5001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8558722554426157e-05,
+      "loss": 0.6151,
+      "step": 5002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8558159356678146e-05,
+      "loss": 0.6342,
+      "step": 5003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.855759605746233e-05,
+      "loss": 0.6576,
+      "step": 5004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8557032656785393e-05,
+      "loss": 0.6348,
+      "step": 5005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8556469154654006e-05,
+      "loss": 0.6403,
+      "step": 5006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8555905551074848e-05,
+      "loss": 0.6104,
+      "step": 5007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8555341846054615e-05,
+      "loss": 0.6184,
+      "step": 5008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8554778039599973e-05,
+      "loss": 0.6524,
+      "step": 5009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.855421413171762e-05,
+      "loss": 0.6368,
+      "step": 5010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8553650122414236e-05,
+      "loss": 0.5921,
+      "step": 5011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8553086011696514e-05,
+      "loss": 0.6647,
+      "step": 5012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8552521799571133e-05,
+      "loss": 0.5958,
+      "step": 5013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8551957486044788e-05,
+      "loss": 0.6236,
+      "step": 5014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8551393071124168e-05,
+      "loss": 0.6452,
+      "step": 5015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8550828554815968e-05,
+      "loss": 0.6482,
+      "step": 5016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8550263937126875e-05,
+      "loss": 0.6734,
+      "step": 5017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.854969921806359e-05,
+      "loss": 0.6055,
+      "step": 5018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8549134397632805e-05,
+      "loss": 0.6271,
+      "step": 5019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8548569475841215e-05,
+      "loss": 0.6439,
+      "step": 5020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.854800445269552e-05,
+      "loss": 0.6649,
+      "step": 5021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8547439328202422e-05,
+      "loss": 0.6743,
+      "step": 5022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8546874102368613e-05,
+      "loss": 0.6517,
+      "step": 5023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8546308775200807e-05,
+      "loss": 0.6482,
+      "step": 5024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.854574334670569e-05,
+      "loss": 0.6178,
+      "step": 5025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8545177816889983e-05,
+      "loss": 0.6224,
+      "step": 5026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.854461218576038e-05,
+      "loss": 0.628,
+      "step": 5027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8544046453323593e-05,
+      "loss": 0.6526,
+      "step": 5028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8543480619586324e-05,
+      "loss": 0.6347,
+      "step": 5029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8542914684555287e-05,
+      "loss": 0.6512,
+      "step": 5030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8542348648237186e-05,
+      "loss": 0.6266,
+      "step": 5031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.854178251063874e-05,
+      "loss": 0.6673,
+      "step": 5032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8541216271766657e-05,
+      "loss": 0.6002,
+      "step": 5033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8540649931627646e-05,
+      "loss": 0.6051,
+      "step": 5034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.854008349022843e-05,
+      "loss": 0.6765,
+      "step": 5035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.853951694757572e-05,
+      "loss": 0.6413,
+      "step": 5036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8538950303676233e-05,
+      "loss": 0.6338,
+      "step": 5037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.853838355853669e-05,
+      "loss": 0.6795,
+      "step": 5038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8537816712163805e-05,
+      "loss": 0.6271,
+      "step": 5039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.853724976456431e-05,
+      "loss": 0.7059,
+      "step": 5040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.853668271574491e-05,
+      "loss": 0.6024,
+      "step": 5041
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.853611556571234e-05,
+      "loss": 0.5582,
+      "step": 5042
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8535548314473325e-05,
+      "loss": 0.62,
+      "step": 5043
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8534980962034587e-05,
+      "loss": 0.6341,
+      "step": 5044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.853441350840285e-05,
+      "loss": 0.6586,
+      "step": 5045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.853384595358485e-05,
+      "loss": 0.6135,
+      "step": 5046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8533278297587303e-05,
+      "loss": 0.6441,
+      "step": 5047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8532710540416955e-05,
+      "loss": 0.677,
+      "step": 5048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8532142682080524e-05,
+      "loss": 0.6522,
+      "step": 5049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8531574722584752e-05,
+      "loss": 0.6021,
+      "step": 5050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.853100666193637e-05,
+      "loss": 0.6689,
+      "step": 5051
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.853043850014211e-05,
+      "loss": 0.6417,
+      "step": 5052
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8529870237208705e-05,
+      "loss": 0.6356,
+      "step": 5053
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8529301873142906e-05,
+      "loss": 0.5657,
+      "step": 5054
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.852873340795144e-05,
+      "loss": 0.6434,
+      "step": 5055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8528164841641052e-05,
+      "loss": 0.6285,
+      "step": 5056
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8527596174218483e-05,
+      "loss": 0.6835,
+      "step": 5057
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.852702740569047e-05,
+      "loss": 0.6155,
+      "step": 5058
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8526458536063762e-05,
+      "loss": 0.5861,
+      "step": 5059
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.85258895653451e-05,
+      "loss": 0.6335,
+      "step": 5060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8525320493541238e-05,
+      "loss": 0.6317,
+      "step": 5061
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8524751320658913e-05,
+      "loss": 0.5732,
+      "step": 5062
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.852418204670488e-05,
+      "loss": 0.65,
+      "step": 5063
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.852361267168588e-05,
+      "loss": 0.6622,
+      "step": 5064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8523043195608676e-05,
+      "loss": 0.6054,
+      "step": 5065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.852247361848001e-05,
+      "loss": 0.599,
+      "step": 5066
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8521903940306634e-05,
+      "loss": 0.604,
+      "step": 5067
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8521334161095314e-05,
+      "loss": 0.6226,
+      "step": 5068
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8520764280852793e-05,
+      "loss": 0.6235,
+      "step": 5069
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8520194299585832e-05,
+      "loss": 0.653,
+      "step": 5070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8519624217301194e-05,
+      "loss": 0.6452,
+      "step": 5071
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8519054034005633e-05,
+      "loss": 0.6288,
+      "step": 5072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8518483749705906e-05,
+      "loss": 0.6194,
+      "step": 5073
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.851791336440878e-05,
+      "loss": 0.6708,
+      "step": 5074
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8517342878121016e-05,
+      "loss": 0.61,
+      "step": 5075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8516772290849378e-05,
+      "loss": 0.6361,
+      "step": 5076
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.851620160260063e-05,
+      "loss": 0.6389,
+      "step": 5077
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.851563081338154e-05,
+      "loss": 0.6148,
+      "step": 5078
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.851505992319887e-05,
+      "loss": 0.62,
+      "step": 5079
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.85144889320594e-05,
+      "loss": 0.589,
+      "step": 5080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.851391783996989e-05,
+      "loss": 0.6127,
+      "step": 5081
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8513346646937114e-05,
+      "loss": 0.6169,
+      "step": 5082
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8512775352967843e-05,
+      "loss": 0.6628,
+      "step": 5083
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8512203958068853e-05,
+      "loss": 0.6337,
+      "step": 5084
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8511632462246914e-05,
+      "loss": 0.6706,
+      "step": 5085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8511060865508806e-05,
+      "loss": 0.6683,
+      "step": 5086
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8510489167861308e-05,
+      "loss": 0.6436,
+      "step": 5087
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8509917369311193e-05,
+      "loss": 0.6487,
+      "step": 5088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8509345469865245e-05,
+      "loss": 0.6121,
+      "step": 5089
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.850877346953024e-05,
+      "loss": 0.6836,
+      "step": 5090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8508201368312957e-05,
+      "loss": 0.608,
+      "step": 5091
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.850762916622019e-05,
+      "loss": 0.6609,
+      "step": 5092
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8507056863258714e-05,
+      "loss": 0.6125,
+      "step": 5093
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.850648445943532e-05,
+      "loss": 0.6084,
+      "step": 5094
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.850591195475679e-05,
+      "loss": 0.6206,
+      "step": 5095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8505339349229913e-05,
+      "loss": 0.6178,
+      "step": 5096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.850476664286148e-05,
+      "loss": 0.5591,
+      "step": 5097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8504193835658283e-05,
+      "loss": 0.6332,
+      "step": 5098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8503620927627107e-05,
+      "loss": 0.6145,
+      "step": 5099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8503047918774745e-05,
+      "loss": 0.6896,
+      "step": 5100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8502474809108e-05,
+      "loss": 0.6257,
+      "step": 5101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8501901598633653e-05,
+      "loss": 0.6531,
+      "step": 5102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8501328287358514e-05,
+      "loss": 0.6335,
+      "step": 5103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.850075487528937e-05,
+      "loss": 0.6241,
+      "step": 5104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8500181362433026e-05,
+      "loss": 0.6448,
+      "step": 5105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8499607748796276e-05,
+      "loss": 0.6397,
+      "step": 5106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.849903403438593e-05,
+      "loss": 0.6156,
+      "step": 5107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8498460219208776e-05,
+      "loss": 0.6609,
+      "step": 5108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8497886303271632e-05,
+      "loss": 0.65,
+      "step": 5109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.849731228658129e-05,
+      "loss": 0.6586,
+      "step": 5110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8496738169144568e-05,
+      "loss": 0.652,
+      "step": 5111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8496163950968264e-05,
+      "loss": 0.6973,
+      "step": 5112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8495589632059187e-05,
+      "loss": 0.6392,
+      "step": 5113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8495015212424153e-05,
+      "loss": 0.6167,
+      "step": 5114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.849444069206996e-05,
+      "loss": 0.6068,
+      "step": 5115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.849386607100343e-05,
+      "loss": 0.6233,
+      "step": 5116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8493291349231376e-05,
+      "loss": 0.5714,
+      "step": 5117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.849271652676061e-05,
+      "loss": 0.6973,
+      "step": 5118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8492141603597942e-05,
+      "loss": 0.6571,
+      "step": 5119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8491566579750194e-05,
+      "loss": 0.5916,
+      "step": 5120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.849099145522418e-05,
+      "loss": 0.6601,
+      "step": 5121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8490416230026724e-05,
+      "loss": 0.6694,
+      "step": 5122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8489840904164645e-05,
+      "loss": 0.6444,
+      "step": 5123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.848926547764476e-05,
+      "loss": 0.6017,
+      "step": 5124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.848868995047389e-05,
+      "loss": 0.5687,
+      "step": 5125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8488114322658866e-05,
+      "loss": 0.6634,
+      "step": 5126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.848753859420651e-05,
+      "loss": 0.6491,
+      "step": 5127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8486962765123645e-05,
+      "loss": 0.6484,
+      "step": 5128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.84863868354171e-05,
+      "loss": 0.6399,
+      "step": 5129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.848581080509371e-05,
+      "loss": 0.6515,
+      "step": 5130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8485234674160292e-05,
+      "loss": 0.6454,
+      "step": 5131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8484658442623685e-05,
+      "loss": 0.6725,
+      "step": 5132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.848408211049072e-05,
+      "loss": 0.633,
+      "step": 5133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.848350567776823e-05,
+      "loss": 0.6044,
+      "step": 5134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8482929144463043e-05,
+      "loss": 0.6112,
+      "step": 5135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8482352510582004e-05,
+      "loss": 0.6521,
+      "step": 5136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8481775776131944e-05,
+      "loss": 0.6052,
+      "step": 5137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8481198941119703e-05,
+      "loss": 0.5778,
+      "step": 5138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8480622005552126e-05,
+      "loss": 0.6449,
+      "step": 5139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8480044969436038e-05,
+      "loss": 0.5935,
+      "step": 5140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.84794678327783e-05,
+      "loss": 0.6354,
+      "step": 5141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.847889059558573e-05,
+      "loss": 0.6621,
+      "step": 5142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8478313257865196e-05,
+      "loss": 0.6497,
+      "step": 5143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8477735819623534e-05,
+      "loss": 0.6684,
+      "step": 5144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8477158280867586e-05,
+      "loss": 0.6679,
+      "step": 5145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8476580641604202e-05,
+      "loss": 0.6654,
+      "step": 5146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8476002901840234e-05,
+      "loss": 0.6525,
+      "step": 5147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8475425061582525e-05,
+      "loss": 0.6045,
+      "step": 5148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8474847120837937e-05,
+      "loss": 0.6455,
+      "step": 5149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8474269079613313e-05,
+      "loss": 0.5865,
+      "step": 5150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8473690937915508e-05,
+      "loss": 0.6515,
+      "step": 5151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.847311269575138e-05,
+      "loss": 0.6764,
+      "step": 5152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8472534353127778e-05,
+      "loss": 0.6533,
+      "step": 5153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8471955910051567e-05,
+      "loss": 0.6341,
+      "step": 5154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.84713773665296e-05,
+      "loss": 0.5798,
+      "step": 5155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.847079872256874e-05,
+      "loss": 0.6157,
+      "step": 5156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8470219978175843e-05,
+      "loss": 0.6162,
+      "step": 5157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.846964113335777e-05,
+      "loss": 0.6663,
+      "step": 5158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8469062188121395e-05,
+      "loss": 0.6063,
+      "step": 5159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8468483142473572e-05,
+      "loss": 0.6261,
+      "step": 5160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8467903996421168e-05,
+      "loss": 0.6164,
+      "step": 5161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8467324749971046e-05,
+      "loss": 0.6112,
+      "step": 5162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8466745403130085e-05,
+      "loss": 0.6183,
+      "step": 5163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8466165955905145e-05,
+      "loss": 0.5857,
+      "step": 5164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8465586408303094e-05,
+      "loss": 0.5694,
+      "step": 5165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.846500676033081e-05,
+      "loss": 0.6465,
+      "step": 5166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.846442701199516e-05,
+      "loss": 0.6296,
+      "step": 5167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8463847163303022e-05,
+      "loss": 0.6506,
+      "step": 5168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.846326721426127e-05,
+      "loss": 0.5876,
+      "step": 5169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.846268716487678e-05,
+      "loss": 0.6111,
+      "step": 5170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8462107015156424e-05,
+      "loss": 0.6232,
+      "step": 5171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.846152676510709e-05,
+      "loss": 0.5749,
+      "step": 5172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.846094641473565e-05,
+      "loss": 0.6499,
+      "step": 5173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8460365964048985e-05,
+      "loss": 0.5915,
+      "step": 5174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.845978541305398e-05,
+      "loss": 0.6209,
+      "step": 5175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8459204761757518e-05,
+      "loss": 0.6339,
+      "step": 5176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8458624010166485e-05,
+      "loss": 0.6393,
+      "step": 5177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.845804315828776e-05,
+      "loss": 0.6321,
+      "step": 5178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8457462206128236e-05,
+      "loss": 0.6312,
+      "step": 5179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.84568811536948e-05,
+      "loss": 0.6091,
+      "step": 5180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8456300000994337e-05,
+      "loss": 0.6902,
+      "step": 5181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.845571874803374e-05,
+      "loss": 0.5848,
+      "step": 5182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8455137394819905e-05,
+      "loss": 0.6137,
+      "step": 5183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8454555941359717e-05,
+      "loss": 0.6372,
+      "step": 5184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8453974387660073e-05,
+      "loss": 0.6288,
+      "step": 5185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8453392733727867e-05,
+      "loss": 0.614,
+      "step": 5186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8452810979570003e-05,
+      "loss": 0.6513,
+      "step": 5187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8452229125193366e-05,
+      "loss": 0.6328,
+      "step": 5188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8451647170604863e-05,
+      "loss": 0.6564,
+      "step": 5189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.845106511581139e-05,
+      "loss": 0.6346,
+      "step": 5190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.845048296081985e-05,
+      "loss": 0.6593,
+      "step": 5191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8449900705637145e-05,
+      "loss": 0.6498,
+      "step": 5192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.844931835027018e-05,
+      "loss": 0.6864,
+      "step": 5193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8448735894725853e-05,
+      "loss": 0.5876,
+      "step": 5194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.844815333901108e-05,
+      "loss": 0.6697,
+      "step": 5195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8447570683132757e-05,
+      "loss": 0.6027,
+      "step": 5196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8446987927097797e-05,
+      "loss": 0.6443,
+      "step": 5197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.844640507091311e-05,
+      "loss": 0.6046,
+      "step": 5198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.844582211458561e-05,
+      "loss": 0.568,
+      "step": 5199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8445239058122202e-05,
+      "loss": 0.6224,
+      "step": 5200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8444655901529804e-05,
+      "loss": 0.5947,
+      "step": 5201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8444072644815322e-05,
+      "loss": 0.5833,
+      "step": 5202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8443489287985683e-05,
+      "loss": 0.6598,
+      "step": 5203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.84429058310478e-05,
+      "loss": 0.6007,
+      "step": 5204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.844232227400858e-05,
+      "loss": 0.681,
+      "step": 5205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8441738616874953e-05,
+      "loss": 0.6806,
+      "step": 5206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8441154859653836e-05,
+      "loss": 0.6228,
+      "step": 5207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.844057100235215e-05,
+      "loss": 0.6561,
+      "step": 5208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8439987044976817e-05,
+      "loss": 0.6179,
+      "step": 5209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.843940298753476e-05,
+      "loss": 0.6608,
+      "step": 5210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8438818830032908e-05,
+      "loss": 0.6272,
+      "step": 5211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8438234572478184e-05,
+      "loss": 0.6894,
+      "step": 5212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8437650214877508e-05,
+      "loss": 0.6399,
+      "step": 5213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8437065757237823e-05,
+      "loss": 0.6325,
+      "step": 5214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8436481199566045e-05,
+      "loss": 0.6544,
+      "step": 5215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.843589654186911e-05,
+      "loss": 0.6256,
+      "step": 5216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8435311784153954e-05,
+      "loss": 0.6227,
+      "step": 5217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8434726926427506e-05,
+      "loss": 0.6938,
+      "step": 5218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8434141968696696e-05,
+      "loss": 0.592,
+      "step": 5219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.843355691096847e-05,
+      "loss": 0.5702,
+      "step": 5220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8432971753249753e-05,
+      "loss": 0.629,
+      "step": 5221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8432386495547486e-05,
+      "loss": 0.6424,
+      "step": 5222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8431801137868617e-05,
+      "loss": 0.65,
+      "step": 5223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8431215680220076e-05,
+      "loss": 0.6578,
+      "step": 5224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8430630122608807e-05,
+      "loss": 0.6707,
+      "step": 5225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8430044465041754e-05,
+      "loss": 0.5872,
+      "step": 5226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.842945870752586e-05,
+      "loss": 0.6775,
+      "step": 5227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.842887285006807e-05,
+      "loss": 0.5764,
+      "step": 5228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8428286892675327e-05,
+      "loss": 0.6497,
+      "step": 5229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8427700835354585e-05,
+      "loss": 0.6664,
+      "step": 5230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8427114678112785e-05,
+      "loss": 0.6248,
+      "step": 5231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8426528420956883e-05,
+      "loss": 0.5956,
+      "step": 5232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8425942063893824e-05,
+      "loss": 0.6241,
+      "step": 5233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8425355606930564e-05,
+      "loss": 0.636,
+      "step": 5234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8424769050074056e-05,
+      "loss": 0.5955,
+      "step": 5235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8424182393331255e-05,
+      "loss": 0.6467,
+      "step": 5236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.842359563670911e-05,
+      "loss": 0.6269,
+      "step": 5237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.842300878021459e-05,
+      "loss": 0.6287,
+      "step": 5238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.842242182385464e-05,
+      "loss": 0.5745,
+      "step": 5239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8421834767636226e-05,
+      "loss": 0.6379,
+      "step": 5240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.842124761156631e-05,
+      "loss": 0.6701,
+      "step": 5241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8420660355651847e-05,
+      "loss": 0.6352,
+      "step": 5242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8420072999899806e-05,
+      "loss": 0.6335,
+      "step": 5243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.841948554431715e-05,
+      "loss": 0.6113,
+      "step": 5244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8418897988910836e-05,
+      "loss": 0.5564,
+      "step": 5245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.841831033368784e-05,
+      "loss": 0.6172,
+      "step": 5246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8417722578655123e-05,
+      "loss": 0.6325,
+      "step": 5247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8417134723819662e-05,
+      "loss": 0.6347,
+      "step": 5248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.841654676918842e-05,
+      "loss": 0.6454,
+      "step": 5249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8415958714768363e-05,
+      "loss": 0.6612,
+      "step": 5250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8415370560566473e-05,
+      "loss": 0.652,
+      "step": 5251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.841478230658972e-05,
+      "loss": 0.6302,
+      "step": 5252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8414193952845077e-05,
+      "loss": 0.6295,
+      "step": 5253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8413605499339522e-05,
+      "loss": 0.5553,
+      "step": 5254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.841301694608003e-05,
+      "loss": 0.629,
+      "step": 5255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.841242829307358e-05,
+      "loss": 0.6141,
+      "step": 5256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8411839540327147e-05,
+      "loss": 0.6072,
+      "step": 5257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.841125068784772e-05,
+      "loss": 0.6766,
+      "step": 5258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8410661735642273e-05,
+      "loss": 0.6383,
+      "step": 5259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8410072683717794e-05,
+      "loss": 0.6479,
+      "step": 5260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8409483532081263e-05,
+      "loss": 0.7031,
+      "step": 5261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8408894280739668e-05,
+      "loss": 0.6543,
+      "step": 5262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.840830492969999e-05,
+      "loss": 0.6176,
+      "step": 5263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8407715478969225e-05,
+      "loss": 0.6847,
+      "step": 5264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8407125928554356e-05,
+      "loss": 0.6233,
+      "step": 5265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8406536278462375e-05,
+      "loss": 0.6128,
+      "step": 5266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8405946528700268e-05,
+      "loss": 0.6528,
+      "step": 5267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8405356679275034e-05,
+      "loss": 0.6934,
+      "step": 5268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8404766730193664e-05,
+      "loss": 0.6482,
+      "step": 5269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8404176681463153e-05,
+      "loss": 0.6541,
+      "step": 5270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8403586533090495e-05,
+      "loss": 0.6486,
+      "step": 5271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.840299628508269e-05,
+      "loss": 0.6694,
+      "step": 5272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8402405937446734e-05,
+      "loss": 0.6169,
+      "step": 5273
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8401815490189623e-05,
+      "loss": 0.6511,
+      "step": 5274
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8401224943318362e-05,
+      "loss": 0.648,
+      "step": 5275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8400634296839953e-05,
+      "loss": 0.653,
+      "step": 5276
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8400043550761396e-05,
+      "loss": 0.6628,
+      "step": 5277
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.83994527050897e-05,
+      "loss": 0.6057,
+      "step": 5278
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8398861759831862e-05,
+      "loss": 0.6145,
+      "step": 5279
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8398270714994898e-05,
+      "loss": 0.6096,
+      "step": 5280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.839767957058581e-05,
+      "loss": 0.6358,
+      "step": 5281
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8397088326611606e-05,
+      "loss": 0.6298,
+      "step": 5282
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8396496983079295e-05,
+      "loss": 0.6825,
+      "step": 5283
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8395905539995896e-05,
+      "loss": 0.6321,
+      "step": 5284
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8395313997368412e-05,
+      "loss": 0.5516,
+      "step": 5285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.839472235520386e-05,
+      "loss": 0.6402,
+      "step": 5286
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8394130613509256e-05,
+      "loss": 0.6358,
+      "step": 5287
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.839353877229162e-05,
+      "loss": 0.6274,
+      "step": 5288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8392946831557957e-05,
+      "loss": 0.6851,
+      "step": 5289
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8392354791315297e-05,
+      "loss": 0.6237,
+      "step": 5290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8391762651570654e-05,
+      "loss": 0.6342,
+      "step": 5291
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8391170412331047e-05,
+      "loss": 0.6866,
+      "step": 5292
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.83905780736035e-05,
+      "loss": 0.5956,
+      "step": 5293
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8389985635395037e-05,
+      "loss": 0.6685,
+      "step": 5294
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.838939309771268e-05,
+      "loss": 0.5411,
+      "step": 5295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8388800460563458e-05,
+      "loss": 0.62,
+      "step": 5296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.838820772395439e-05,
+      "loss": 0.6467,
+      "step": 5297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8387614887892507e-05,
+      "loss": 0.6472,
+      "step": 5298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8387021952384848e-05,
+      "loss": 0.6617,
+      "step": 5299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8386428917438425e-05,
+      "loss": 0.6391,
+      "step": 5300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8385835783060284e-05,
+      "loss": 0.621,
+      "step": 5301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8385242549257447e-05,
+      "loss": 0.6018,
+      "step": 5302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8384649216036957e-05,
+      "loss": 0.6267,
+      "step": 5303
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.838405578340584e-05,
+      "loss": 0.6975,
+      "step": 5304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.838346225137114e-05,
+      "loss": 0.6732,
+      "step": 5305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8382868619939886e-05,
+      "loss": 0.6426,
+      "step": 5306
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.838227488911912e-05,
+      "loss": 0.6068,
+      "step": 5307
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8381681058915887e-05,
+      "loss": 0.7103,
+      "step": 5308
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8381087129337215e-05,
+      "loss": 0.6001,
+      "step": 5309
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8380493100390157e-05,
+      "loss": 0.6526,
+      "step": 5310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8379898972081754e-05,
+      "loss": 0.6783,
+      "step": 5311
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8379304744419042e-05,
+      "loss": 0.6001,
+      "step": 5312
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8378710417409076e-05,
+      "loss": 0.6538,
+      "step": 5313
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.83781159910589e-05,
+      "loss": 0.6162,
+      "step": 5314
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8377521465375557e-05,
+      "loss": 0.5948,
+      "step": 5315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.83769268403661e-05,
+      "loss": 0.6637,
+      "step": 5316
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.837633211603758e-05,
+      "loss": 0.669,
+      "step": 5317
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8375737292397046e-05,
+      "loss": 0.6244,
+      "step": 5318
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8375142369451554e-05,
+      "loss": 0.6745,
+      "step": 5319
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.837454734720815e-05,
+      "loss": 0.6252,
+      "step": 5320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.83739522256739e-05,
+      "loss": 0.6304,
+      "step": 5321
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.837335700485585e-05,
+      "loss": 0.6036,
+      "step": 5322
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8372761684761057e-05,
+      "loss": 0.6431,
+      "step": 5323
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.837216626539658e-05,
+      "loss": 0.662,
+      "step": 5324
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8371570746769488e-05,
+      "loss": 0.6586,
+      "step": 5325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8370975128886833e-05,
+      "loss": 0.6235,
+      "step": 5326
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8370379411755677e-05,
+      "loss": 0.6707,
+      "step": 5327
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.836978359538309e-05,
+      "loss": 0.6383,
+      "step": 5328
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8369187679776125e-05,
+      "loss": 0.6676,
+      "step": 5329
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8368591664941854e-05,
+      "loss": 0.6325,
+      "step": 5330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.836799555088734e-05,
+      "loss": 0.6878,
+      "step": 5331
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.836739933761966e-05,
+      "loss": 0.6528,
+      "step": 5332
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8366803025145867e-05,
+      "loss": 0.6124,
+      "step": 5333
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8366206613473045e-05,
+      "loss": 0.6244,
+      "step": 5334
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8365610102608263e-05,
+      "loss": 0.69,
+      "step": 5335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8365013492558587e-05,
+      "loss": 0.6223,
+      "step": 5336
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8364416783331093e-05,
+      "loss": 0.6159,
+      "step": 5337
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8363819974932863e-05,
+      "loss": 0.6099,
+      "step": 5338
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.836322306737096e-05,
+      "loss": 0.6502,
+      "step": 5339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8362626060652472e-05,
+      "loss": 0.5467,
+      "step": 5340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.836202895478447e-05,
+      "loss": 0.6283,
+      "step": 5341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.836143174977404e-05,
+      "loss": 0.6195,
+      "step": 5342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.836083444562826e-05,
+      "loss": 0.6126,
+      "step": 5343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8360237042354207e-05,
+      "loss": 0.7186,
+      "step": 5344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.835963953995897e-05,
+      "loss": 0.5851,
+      "step": 5345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8359041938449636e-05,
+      "loss": 0.5955,
+      "step": 5346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.835844423783328e-05,
+      "loss": 0.6371,
+      "step": 5347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8357846438116995e-05,
+      "loss": 0.6714,
+      "step": 5348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.835724853930787e-05,
+      "loss": 0.6596,
+      "step": 5349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8356650541412987e-05,
+      "loss": 0.6761,
+      "step": 5350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8356052444439445e-05,
+      "loss": 0.6088,
+      "step": 5351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8355454248394328e-05,
+      "loss": 0.6246,
+      "step": 5352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8354855953284733e-05,
+      "loss": 0.6227,
+      "step": 5353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.835425755911775e-05,
+      "loss": 0.6622,
+      "step": 5354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.835365906590048e-05,
+      "loss": 0.6466,
+      "step": 5355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.835306047364001e-05,
+      "loss": 0.6467,
+      "step": 5356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.835246178234344e-05,
+      "loss": 0.5771,
+      "step": 5357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.835186299201787e-05,
+      "loss": 0.6301,
+      "step": 5358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8351264102670405e-05,
+      "loss": 0.6632,
+      "step": 5359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8350665114308138e-05,
+      "loss": 0.649,
+      "step": 5360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8350066026938168e-05,
+      "loss": 0.6598,
+      "step": 5361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8349466840567604e-05,
+      "loss": 0.6782,
+      "step": 5362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.834886755520355e-05,
+      "loss": 0.5737,
+      "step": 5363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8348268170853106e-05,
+      "loss": 0.6499,
+      "step": 5364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.834766868752339e-05,
+      "loss": 0.6001,
+      "step": 5365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8347069105221493e-05,
+      "loss": 0.6542,
+      "step": 5366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8346469423954535e-05,
+      "loss": 0.6349,
+      "step": 5367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.834586964372963e-05,
+      "loss": 0.6157,
+      "step": 5368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8345269764553877e-05,
+      "loss": 0.6233,
+      "step": 5369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8344669786434393e-05,
+      "loss": 0.6544,
+      "step": 5370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8344069709378294e-05,
+      "loss": 0.6545,
+      "step": 5371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8343469533392696e-05,
+      "loss": 0.599,
+      "step": 5372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.834286925848471e-05,
+      "loss": 0.7031,
+      "step": 5373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8342268884661452e-05,
+      "loss": 0.6785,
+      "step": 5374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.834166841193005e-05,
+      "loss": 0.6267,
+      "step": 5375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.834106784029761e-05,
+      "loss": 0.636,
+      "step": 5376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8340467169771266e-05,
+      "loss": 0.6181,
+      "step": 5377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.833986640035813e-05,
+      "loss": 0.6342,
+      "step": 5378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8339265532065328e-05,
+      "loss": 0.6083,
+      "step": 5379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.833866456489998e-05,
+      "loss": 0.6292,
+      "step": 5380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8338063498869218e-05,
+      "loss": 0.6312,
+      "step": 5381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8337462333980166e-05,
+      "loss": 0.6259,
+      "step": 5382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8336861070239954e-05,
+      "loss": 0.6167,
+      "step": 5383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8336259707655702e-05,
+      "loss": 0.696,
+      "step": 5384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.833565824623455e-05,
+      "loss": 0.6203,
+      "step": 5385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.833505668598362e-05,
+      "loss": 0.6843,
+      "step": 5386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8334455026910056e-05,
+      "loss": 0.6276,
+      "step": 5387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.833385326902098e-05,
+      "loss": 0.577,
+      "step": 5388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8333251412323535e-05,
+      "loss": 0.6386,
+      "step": 5389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8332649456824847e-05,
+      "loss": 0.6484,
+      "step": 5390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8332047402532064e-05,
+      "loss": 0.6522,
+      "step": 5391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8331445249452316e-05,
+      "loss": 0.6399,
+      "step": 5392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8330842997592748e-05,
+      "loss": 0.5963,
+      "step": 5393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8330240646960494e-05,
+      "loss": 0.6274,
+      "step": 5394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.83296381975627e-05,
+      "loss": 0.6197,
+      "step": 5395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.832903564940651e-05,
+      "loss": 0.5964,
+      "step": 5396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8328433002499065e-05,
+      "loss": 0.6141,
+      "step": 5397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.832783025684751e-05,
+      "loss": 0.6741,
+      "step": 5398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8327227412458994e-05,
+      "loss": 0.6568,
+      "step": 5399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8326624469340662e-05,
+      "loss": 0.6958,
+      "step": 5400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8326021427499668e-05,
+      "loss": 0.6332,
+      "step": 5401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8325418286943153e-05,
+      "loss": 0.664,
+      "step": 5402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8324815047678274e-05,
+      "loss": 0.643,
+      "step": 5403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.832421170971218e-05,
+      "loss": 0.6761,
+      "step": 5404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8323608273052028e-05,
+      "loss": 0.6394,
+      "step": 5405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.832300473770497e-05,
+      "loss": 0.6174,
+      "step": 5406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8322401103678163e-05,
+      "loss": 0.6466,
+      "step": 5407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8321797370978764e-05,
+      "loss": 0.6128,
+      "step": 5408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8321193539613924e-05,
+      "loss": 0.6315,
+      "step": 5409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8320589609590813e-05,
+      "loss": 0.6512,
+      "step": 5410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8319985580916588e-05,
+      "loss": 0.5846,
+      "step": 5411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8319381453598406e-05,
+      "loss": 0.6276,
+      "step": 5412
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.831877722764343e-05,
+      "loss": 0.5678,
+      "step": 5413
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8318172903058834e-05,
+      "loss": 0.6427,
+      "step": 5414
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8317568479851772e-05,
+      "loss": 0.6722,
+      "step": 5415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.831696395802941e-05,
+      "loss": 0.6378,
+      "step": 5416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8316359337598928e-05,
+      "loss": 0.6209,
+      "step": 5417
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.831575461856748e-05,
+      "loss": 0.6602,
+      "step": 5418
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8315149800942243e-05,
+      "loss": 0.6578,
+      "step": 5419
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8314544884730385e-05,
+      "loss": 0.6519,
+      "step": 5420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8313939869939084e-05,
+      "loss": 0.636,
+      "step": 5421
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8313334756575503e-05,
+      "loss": 0.6368,
+      "step": 5422
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8312729544646824e-05,
+      "loss": 0.638,
+      "step": 5423
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.831212423416022e-05,
+      "loss": 0.6661,
+      "step": 5424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8311518825122873e-05,
+      "loss": 0.666,
+      "step": 5425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.831091331754195e-05,
+      "loss": 0.6202,
+      "step": 5426
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.831030771142464e-05,
+      "loss": 0.6605,
+      "step": 5427
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8309702006778118e-05,
+      "loss": 0.7051,
+      "step": 5428
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8309096203609566e-05,
+      "loss": 0.5795,
+      "step": 5429
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8308490301926167e-05,
+      "loss": 0.652,
+      "step": 5430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.830788430173511e-05,
+      "loss": 0.6556,
+      "step": 5431
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.830727820304357e-05,
+      "loss": 0.6892,
+      "step": 5432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.830667200585874e-05,
+      "loss": 0.6471,
+      "step": 5433
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8306065710187804e-05,
+      "loss": 0.6777,
+      "step": 5434
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.830545931603795e-05,
+      "loss": 0.6327,
+      "step": 5435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.830485282341637e-05,
+      "loss": 0.67,
+      "step": 5436
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.830424623233026e-05,
+      "loss": 0.6638,
+      "step": 5437
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8303639542786806e-05,
+      "loss": 0.6431,
+      "step": 5438
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8303032754793192e-05,
+      "loss": 0.6163,
+      "step": 5439
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.830242586835663e-05,
+      "loss": 0.6604,
+      "step": 5440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8301818883484303e-05,
+      "loss": 0.6012,
+      "step": 5441
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.830121180018341e-05,
+      "loss": 0.6359,
+      "step": 5442
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8300604618461154e-05,
+      "loss": 0.6157,
+      "step": 5443
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8299997338324725e-05,
+      "loss": 0.6677,
+      "step": 5444
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8299389959781333e-05,
+      "loss": 0.6564,
+      "step": 5445
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.829878248283817e-05,
+      "loss": 0.6511,
+      "step": 5446
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8298174907502442e-05,
+      "loss": 0.6636,
+      "step": 5447
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.829756723378136e-05,
+      "loss": 0.5845,
+      "step": 5448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8296959461682117e-05,
+      "loss": 0.6389,
+      "step": 5449
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8296351591211924e-05,
+      "loss": 0.6301,
+      "step": 5450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.829574362237799e-05,
+      "loss": 0.6651,
+      "step": 5451
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.829513555518752e-05,
+      "loss": 0.6036,
+      "step": 5452
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8294527389647722e-05,
+      "loss": 0.6285,
+      "step": 5453
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8293919125765812e-05,
+      "loss": 0.6885,
+      "step": 5454
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8293310763548998e-05,
+      "loss": 0.6097,
+      "step": 5455
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8292702303004496e-05,
+      "loss": 0.6041,
+      "step": 5456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8292093744139513e-05,
+      "loss": 0.5934,
+      "step": 5457
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8291485086961273e-05,
+      "loss": 0.6735,
+      "step": 5458
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8290876331476988e-05,
+      "loss": 0.6487,
+      "step": 5459
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8290267477693876e-05,
+      "loss": 0.6333,
+      "step": 5460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8289658525619157e-05,
+      "loss": 0.6538,
+      "step": 5461
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8289049475260044e-05,
+      "loss": 0.6631,
+      "step": 5462
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828844032662377e-05,
+      "loss": 0.6722,
+      "step": 5463
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828783107971755e-05,
+      "loss": 0.6439,
+      "step": 5464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828722173454861e-05,
+      "loss": 0.6113,
+      "step": 5465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828661229112417e-05,
+      "loss": 0.6688,
+      "step": 5466
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8286002749451458e-05,
+      "loss": 0.5986,
+      "step": 5467
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.82853931095377e-05,
+      "loss": 0.648,
+      "step": 5468
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828478337139013e-05,
+      "loss": 0.6518,
+      "step": 5469
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828417353501597e-05,
+      "loss": 0.6205,
+      "step": 5470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8283563600422453e-05,
+      "loss": 0.6601,
+      "step": 5471
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8282953567616814e-05,
+      "loss": 0.6496,
+      "step": 5472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828234343660628e-05,
+      "loss": 0.5873,
+      "step": 5473
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8281733207398086e-05,
+      "loss": 0.6425,
+      "step": 5474
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828112287999947e-05,
+      "loss": 0.627,
+      "step": 5475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.828051245441767e-05,
+      "loss": 0.6386,
+      "step": 5476
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8279901930659914e-05,
+      "loss": 0.6222,
+      "step": 5477
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.827929130873345e-05,
+      "loss": 0.6334,
+      "step": 5478
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8278680588645513e-05,
+      "loss": 0.6453,
+      "step": 5479
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.827806977040335e-05,
+      "loss": 0.6294,
+      "step": 5480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.827745885401419e-05,
+      "loss": 0.6221,
+      "step": 5481
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8276847839485287e-05,
+      "loss": 0.6733,
+      "step": 5482
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8276236726823886e-05,
+      "loss": 0.6213,
+      "step": 5483
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8275625516037225e-05,
+      "loss": 0.634,
+      "step": 5484
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8275014207132555e-05,
+      "loss": 0.6766,
+      "step": 5485
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8274402800117125e-05,
+      "loss": 0.643,
+      "step": 5486
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8273791294998184e-05,
+      "loss": 0.6178,
+      "step": 5487
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8273179691782978e-05,
+      "loss": 0.6375,
+      "step": 5488
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.827256799047876e-05,
+      "loss": 0.6277,
+      "step": 5489
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8271956191092788e-05,
+      "loss": 0.6202,
+      "step": 5490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.827134429363231e-05,
+      "loss": 0.604,
+      "step": 5491
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8270732298104584e-05,
+      "loss": 0.6512,
+      "step": 5492
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.827012020451686e-05,
+      "loss": 0.6152,
+      "step": 5493
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.82695080128764e-05,
+      "loss": 0.6622,
+      "step": 5494
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.826889572319046e-05,
+      "loss": 0.5792,
+      "step": 5495
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8268283335466304e-05,
+      "loss": 0.7125,
+      "step": 5496
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.826767084971119e-05,
+      "loss": 0.6577,
+      "step": 5497
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.826705826593238e-05,
+      "loss": 0.6308,
+      "step": 5498
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.826644558413713e-05,
+      "loss": 0.5865,
+      "step": 5499
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8265832804332717e-05,
+      "loss": 0.6244,
+      "step": 5500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8265219926526398e-05,
+      "loss": 0.6203,
+      "step": 5501
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.826460695072544e-05,
+      "loss": 0.6474,
+      "step": 5502
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8263993876937114e-05,
+      "loss": 0.6529,
+      "step": 5503
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.826338070516869e-05,
+      "loss": 0.6141,
+      "step": 5504
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8262767435427428e-05,
+      "loss": 0.64,
+      "step": 5505
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8262154067720606e-05,
+      "loss": 0.6458,
+      "step": 5506
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.82615406020555e-05,
+      "loss": 0.6538,
+      "step": 5507
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8260927038439375e-05,
+      "loss": 0.6771,
+      "step": 5508
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.826031337687951e-05,
+      "loss": 0.5957,
+      "step": 5509
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.825969961738318e-05,
+      "loss": 0.5753,
+      "step": 5510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8259085759957667e-05,
+      "loss": 0.6326,
+      "step": 5511
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.825847180461024e-05,
+      "loss": 0.6056,
+      "step": 5512
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8257857751348188e-05,
+      "loss": 0.6173,
+      "step": 5513
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.825724360017878e-05,
+      "loss": 0.6031,
+      "step": 5514
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.825662935110931e-05,
+      "loss": 0.6386,
+      "step": 5515
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.825601500414705e-05,
+      "loss": 0.6342,
+      "step": 5516
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.825540055929929e-05,
+      "loss": 0.5765,
+      "step": 5517
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8254786016573313e-05,
+      "loss": 0.6222,
+      "step": 5518
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8254171375976403e-05,
+      "loss": 0.6798,
+      "step": 5519
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8253556637515852e-05,
+      "loss": 0.5926,
+      "step": 5520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8252941801198946e-05,
+      "loss": 0.6688,
+      "step": 5521
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8252326867032973e-05,
+      "loss": 0.6777,
+      "step": 5522
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.825171183502523e-05,
+      "loss": 0.6393,
+      "step": 5523
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8251096705183e-05,
+      "loss": 0.6157,
+      "step": 5524
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8250481477513583e-05,
+      "loss": 0.6484,
+      "step": 5525
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.824986615202427e-05,
+      "loss": 0.6149,
+      "step": 5526
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.824925072872236e-05,
+      "loss": 0.6351,
+      "step": 5527
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8248635207615147e-05,
+      "loss": 0.6013,
+      "step": 5528
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.824801958870993e-05,
+      "loss": 0.6152,
+      "step": 5529
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8247403872014e-05,
+      "loss": 0.6026,
+      "step": 5530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8246788057534672e-05,
+      "loss": 0.6801,
+      "step": 5531
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8246172145279237e-05,
+      "loss": 0.6636,
+      "step": 5532
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8245556135255e-05,
+      "loss": 0.6327,
+      "step": 5533
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8244940027469263e-05,
+      "loss": 0.6373,
+      "step": 5534
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8244323821929334e-05,
+      "loss": 0.6658,
+      "step": 5535
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8243707518642514e-05,
+      "loss": 0.6185,
+      "step": 5536
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8243091117616115e-05,
+      "loss": 0.5985,
+      "step": 5537
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8242474618857442e-05,
+      "loss": 0.6476,
+      "step": 5538
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8241858022373807e-05,
+      "loss": 0.6566,
+      "step": 5539
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.824124132817252e-05,
+      "loss": 0.6255,
+      "step": 5540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.824062453626089e-05,
+      "loss": 0.6315,
+      "step": 5541
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.824000764664623e-05,
+      "loss": 0.6915,
+      "step": 5542
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.823939065933586e-05,
+      "loss": 0.6583,
+      "step": 5543
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.823877357433709e-05,
+      "loss": 0.7038,
+      "step": 5544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8238156391657234e-05,
+      "loss": 0.6627,
+      "step": 5545
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8237539111303615e-05,
+      "loss": 0.6746,
+      "step": 5546
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8236921733283547e-05,
+      "loss": 0.6617,
+      "step": 5547
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8236304257604357e-05,
+      "loss": 0.6575,
+      "step": 5548
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8235686684273358e-05,
+      "loss": 0.6495,
+      "step": 5549
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8235069013297875e-05,
+      "loss": 0.6438,
+      "step": 5550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8234451244685232e-05,
+      "loss": 0.677,
+      "step": 5551
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8233833378442752e-05,
+      "loss": 0.6836,
+      "step": 5552
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.823321541457776e-05,
+      "loss": 0.6971,
+      "step": 5553
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8232597353097587e-05,
+      "loss": 0.6726,
+      "step": 5554
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8231979194009556e-05,
+      "loss": 0.6364,
+      "step": 5555
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8231360937321e-05,
+      "loss": 0.6137,
+      "step": 5556
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8230742583039245e-05,
+      "loss": 0.6449,
+      "step": 5557
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8230124131171626e-05,
+      "loss": 0.6651,
+      "step": 5558
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8229505581725476e-05,
+      "loss": 0.6005,
+      "step": 5559
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8228886934708127e-05,
+      "loss": 0.6244,
+      "step": 5560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.822826819012691e-05,
+      "loss": 0.669,
+      "step": 5561
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.822764934798917e-05,
+      "loss": 0.653,
+      "step": 5562
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8227030408302236e-05,
+      "loss": 0.643,
+      "step": 5563
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.822641137107345e-05,
+      "loss": 0.6253,
+      "step": 5564
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.822579223631015e-05,
+      "loss": 0.6401,
+      "step": 5565
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8225173004019675e-05,
+      "loss": 0.6873,
+      "step": 5566
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8224553674209372e-05,
+      "loss": 0.6521,
+      "step": 5567
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.822393424688658e-05,
+      "loss": 0.6154,
+      "step": 5568
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8223314722058647e-05,
+      "loss": 0.6674,
+      "step": 5569
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8222695099732912e-05,
+      "loss": 0.6535,
+      "step": 5570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8222075379916728e-05,
+      "loss": 0.6556,
+      "step": 5571
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.822145556261744e-05,
+      "loss": 0.5962,
+      "step": 5572
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.822083564784239e-05,
+      "loss": 0.6828,
+      "step": 5573
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8220215635598938e-05,
+      "loss": 0.6538,
+      "step": 5574
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.821959552589443e-05,
+      "loss": 0.6265,
+      "step": 5575
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.821897531873622e-05,
+      "loss": 0.659,
+      "step": 5576
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8218355014131662e-05,
+      "loss": 0.6303,
+      "step": 5577
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8217734612088108e-05,
+      "loss": 0.6235,
+      "step": 5578
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8217114112612912e-05,
+      "loss": 0.6286,
+      "step": 5579
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8216493515713437e-05,
+      "loss": 0.5933,
+      "step": 5580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8215872821397034e-05,
+      "loss": 0.648,
+      "step": 5581
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.821525202967107e-05,
+      "loss": 0.6492,
+      "step": 5582
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8214631140542896e-05,
+      "loss": 0.6445,
+      "step": 5583
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.821401015401988e-05,
+      "loss": 0.6458,
+      "step": 5584
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8213389070109385e-05,
+      "loss": 0.657,
+      "step": 5585
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8212767888818772e-05,
+      "loss": 0.6497,
+      "step": 5586
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8212146610155407e-05,
+      "loss": 0.6569,
+      "step": 5587
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.821152523412665e-05,
+      "loss": 0.5769,
+      "step": 5588
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.821090376073988e-05,
+      "loss": 0.6527,
+      "step": 5589
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8210282190002458e-05,
+      "loss": 0.6758,
+      "step": 5590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8209660521921753e-05,
+      "loss": 0.6408,
+      "step": 5591
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8209038756505137e-05,
+      "loss": 0.6058,
+      "step": 5592
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8208416893759984e-05,
+      "loss": 0.705,
+      "step": 5593
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8207794933693663e-05,
+      "loss": 0.6479,
+      "step": 5594
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.820717287631355e-05,
+      "loss": 0.6443,
+      "step": 5595
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8206550721627022e-05,
+      "loss": 0.6557,
+      "step": 5596
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8205928469641454e-05,
+      "loss": 0.6964,
+      "step": 5597
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8205306120364222e-05,
+      "loss": 0.6388,
+      "step": 5598
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8204683673802708e-05,
+      "loss": 0.6353,
+      "step": 5599
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8204061129964288e-05,
+      "loss": 0.6237,
+      "step": 5600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8203438488856348e-05,
+      "loss": 0.6607,
+      "step": 5601
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8202815750486264e-05,
+      "loss": 0.6388,
+      "step": 5602
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8202192914861423e-05,
+      "loss": 0.6502,
+      "step": 5603
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.820156998198921e-05,
+      "loss": 0.635,
+      "step": 5604
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.820094695187701e-05,
+      "loss": 0.6079,
+      "step": 5605
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.820032382453221e-05,
+      "loss": 0.6103,
+      "step": 5606
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.81997005999622e-05,
+      "loss": 0.6061,
+      "step": 5607
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.819907727817436e-05,
+      "loss": 0.6136,
+      "step": 5608
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8198453859176093e-05,
+      "loss": 0.6008,
+      "step": 5609
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8197830342974783e-05,
+      "loss": 0.6617,
+      "step": 5610
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.819720672957782e-05,
+      "loss": 0.6672,
+      "step": 5611
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8196583018992608e-05,
+      "loss": 0.6834,
+      "step": 5612
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.819595921122653e-05,
+      "loss": 0.6469,
+      "step": 5613
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8195335306286993e-05,
+      "loss": 0.6241,
+      "step": 5614
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8194711304181384e-05,
+      "loss": 0.5881,
+      "step": 5615
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.819408720491711e-05,
+      "loss": 0.6256,
+      "step": 5616
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8193463008501565e-05,
+      "loss": 0.6255,
+      "step": 5617
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.819283871494215e-05,
+      "loss": 0.6257,
+      "step": 5618
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.819221432424627e-05,
+      "loss": 0.5905,
+      "step": 5619
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8191589836421322e-05,
+      "loss": 0.6378,
+      "step": 5620
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.819096525147472e-05,
+      "loss": 0.6903,
+      "step": 5621
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.819034056941386e-05,
+      "loss": 0.6382,
+      "step": 5622
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.818971579024615e-05,
+      "loss": 0.6435,
+      "step": 5623
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8189090913979e-05,
+      "loss": 0.6316,
+      "step": 5624
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8188465940619823e-05,
+      "loss": 0.6905,
+      "step": 5625
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.818784087017602e-05,
+      "loss": 0.7193,
+      "step": 5626
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8187215702655004e-05,
+      "loss": 0.6179,
+      "step": 5627
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8186590438064192e-05,
+      "loss": 0.6333,
+      "step": 5628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8185965076410994e-05,
+      "loss": 0.7022,
+      "step": 5629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8185339617702822e-05,
+      "loss": 0.673,
+      "step": 5630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8184714061947098e-05,
+      "loss": 0.6772,
+      "step": 5631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8184088409151233e-05,
+      "loss": 0.6865,
+      "step": 5632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.818346265932265e-05,
+      "loss": 0.6913,
+      "step": 5633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.818283681246876e-05,
+      "loss": 0.6006,
+      "step": 5634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.818221086859699e-05,
+      "loss": 0.5941,
+      "step": 5635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.818158482771476e-05,
+      "loss": 0.6682,
+      "step": 5636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8180958689829493e-05,
+      "loss": 0.6147,
+      "step": 5637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.818033245494861e-05,
+      "loss": 0.5888,
+      "step": 5638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.817970612307954e-05,
+      "loss": 0.6053,
+      "step": 5639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8179079694229704e-05,
+      "loss": 0.6374,
+      "step": 5640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8178453168406535e-05,
+      "loss": 0.5984,
+      "step": 5641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8177826545617454e-05,
+      "loss": 0.6772,
+      "step": 5642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8177199825869897e-05,
+      "loss": 0.6536,
+      "step": 5643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.817657300917129e-05,
+      "loss": 0.6478,
+      "step": 5644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8175946095529066e-05,
+      "loss": 0.5984,
+      "step": 5645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8175319084950663e-05,
+      "loss": 0.6705,
+      "step": 5646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8174691977443505e-05,
+      "loss": 0.6436,
+      "step": 5647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8174064773015035e-05,
+      "loss": 0.6075,
+      "step": 5648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8173437471672688e-05,
+      "loss": 0.6667,
+      "step": 5649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.81728100734239e-05,
+      "loss": 0.572,
+      "step": 5650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8172182578276104e-05,
+      "loss": 0.6472,
+      "step": 5651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8171554986236756e-05,
+      "loss": 0.6339,
+      "step": 5652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8170927297313277e-05,
+      "loss": 0.6363,
+      "step": 5653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8170299511513122e-05,
+      "loss": 0.6336,
+      "step": 5654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8169671628843732e-05,
+      "loss": 0.6678,
+      "step": 5655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8169043649312553e-05,
+      "loss": 0.616,
+      "step": 5656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8168415572927022e-05,
+      "loss": 0.6423,
+      "step": 5657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8167787399694594e-05,
+      "loss": 0.6564,
+      "step": 5658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8167159129622715e-05,
+      "loss": 0.6705,
+      "step": 5659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.816653076271883e-05,
+      "loss": 0.6319,
+      "step": 5660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8165902298990393e-05,
+      "loss": 0.6624,
+      "step": 5661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8165273738444855e-05,
+      "loss": 0.6135,
+      "step": 5662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8164645081089668e-05,
+      "loss": 0.6441,
+      "step": 5663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8164016326932286e-05,
+      "loss": 0.6864,
+      "step": 5664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.816338747598016e-05,
+      "loss": 0.6524,
+      "step": 5665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.816275852824075e-05,
+      "loss": 0.6269,
+      "step": 5666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8162129483721515e-05,
+      "loss": 0.6116,
+      "step": 5667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8161500342429907e-05,
+      "loss": 0.6031,
+      "step": 5668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.816087110437339e-05,
+      "loss": 0.643,
+      "step": 5669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8160241769559422e-05,
+      "loss": 0.5983,
+      "step": 5670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.815961233799546e-05,
+      "loss": 0.641,
+      "step": 5671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.815898280968898e-05,
+      "loss": 0.637,
+      "step": 5672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.815835318464743e-05,
+      "loss": 0.6298,
+      "step": 5673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8157723462878288e-05,
+      "loss": 0.7178,
+      "step": 5674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.815709364438901e-05,
+      "loss": 0.668,
+      "step": 5675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8156463729187073e-05,
+      "loss": 0.616,
+      "step": 5676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8155833717279938e-05,
+      "loss": 0.6202,
+      "step": 5677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8155203608675077e-05,
+      "loss": 0.6328,
+      "step": 5678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.815457340337996e-05,
+      "loss": 0.6795,
+      "step": 5679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.815394310140206e-05,
+      "loss": 0.5134,
+      "step": 5680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8153312702748854e-05,
+      "loss": 0.593,
+      "step": 5681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.815268220742781e-05,
+      "loss": 0.6712,
+      "step": 5682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8152051615446402e-05,
+      "loss": 0.6376,
+      "step": 5683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8151420926812113e-05,
+      "loss": 0.6416,
+      "step": 5684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8150790141532414e-05,
+      "loss": 0.6131,
+      "step": 5685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8150159259614788e-05,
+      "loss": 0.6331,
+      "step": 5686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8149528281066714e-05,
+      "loss": 0.6571,
+      "step": 5687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8148897205895675e-05,
+      "loss": 0.661,
+      "step": 5688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.814826603410915e-05,
+      "loss": 0.6878,
+      "step": 5689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8147634765714625e-05,
+      "loss": 0.609,
+      "step": 5690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.814700340071958e-05,
+      "loss": 0.6181,
+      "step": 5691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8146371939131504e-05,
+      "loss": 0.6766,
+      "step": 5692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8145740380957887e-05,
+      "loss": 0.6108,
+      "step": 5693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.814510872620621e-05,
+      "loss": 0.6255,
+      "step": 5694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8144476974883968e-05,
+      "loss": 0.6319,
+      "step": 5695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8143845126998645e-05,
+      "loss": 0.6559,
+      "step": 5696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.814321318255774e-05,
+      "loss": 0.6219,
+      "step": 5697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.814258114156874e-05,
+      "loss": 0.673,
+      "step": 5698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8141949004039138e-05,
+      "loss": 0.6578,
+      "step": 5699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8141316769976435e-05,
+      "loss": 0.637,
+      "step": 5700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.814068443938812e-05,
+      "loss": 0.6665,
+      "step": 5701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.814005201228169e-05,
+      "loss": 0.6677,
+      "step": 5702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8139419488664654e-05,
+      "loss": 0.6171,
+      "step": 5703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.81387868685445e-05,
+      "loss": 0.6328,
+      "step": 5704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.813815415192873e-05,
+      "loss": 0.6259,
+      "step": 5705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.813752133882485e-05,
+      "loss": 0.6429,
+      "step": 5706
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.813688842924036e-05,
+      "loss": 0.5939,
+      "step": 5707
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.813625542318276e-05,
+      "loss": 0.6557,
+      "step": 5708
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.813562232065957e-05,
+      "loss": 0.61,
+      "step": 5709
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8134989121678277e-05,
+      "loss": 0.6052,
+      "step": 5710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.81343558262464e-05,
+      "loss": 0.6129,
+      "step": 5711
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8133722434371444e-05,
+      "loss": 0.641,
+      "step": 5712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8133088946060916e-05,
+      "loss": 0.6223,
+      "step": 5713
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8132455361322337e-05,
+      "loss": 0.5894,
+      "step": 5714
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.813182168016321e-05,
+      "loss": 0.6749,
+      "step": 5715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.813118790259105e-05,
+      "loss": 0.648,
+      "step": 5716
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.813055402861337e-05,
+      "loss": 0.6845,
+      "step": 5717
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8129920058237685e-05,
+      "loss": 0.6029,
+      "step": 5718
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8129285991471514e-05,
+      "loss": 0.6168,
+      "step": 5719
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8128651828322376e-05,
+      "loss": 0.6008,
+      "step": 5720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.812801756879779e-05,
+      "loss": 0.6453,
+      "step": 5721
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8127383212905267e-05,
+      "loss": 0.6513,
+      "step": 5722
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8126748760652343e-05,
+      "loss": 0.5887,
+      "step": 5723
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8126114212046527e-05,
+      "loss": 0.5915,
+      "step": 5724
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.812547956709535e-05,
+      "loss": 0.6419,
+      "step": 5725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.812484482580633e-05,
+      "loss": 0.626,
+      "step": 5726
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8124209988187e-05,
+      "loss": 0.6393,
+      "step": 5727
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.812357505424488e-05,
+      "loss": 0.5963,
+      "step": 5728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8122940023987504e-05,
+      "loss": 0.6992,
+      "step": 5729
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8122304897422397e-05,
+      "loss": 0.6667,
+      "step": 5730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.812166967455709e-05,
+      "loss": 0.6285,
+      "step": 5731
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.812103435539912e-05,
+      "loss": 0.6022,
+      "step": 5732
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.812039893995601e-05,
+      "loss": 0.6563,
+      "step": 5733
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.81197634282353e-05,
+      "loss": 0.6084,
+      "step": 5734
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8119127820244522e-05,
+      "loss": 0.633,
+      "step": 5735
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8118492115991212e-05,
+      "loss": 0.6605,
+      "step": 5736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.811785631548291e-05,
+      "loss": 0.6291,
+      "step": 5737
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8117220418727154e-05,
+      "loss": 0.5887,
+      "step": 5738
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8116584425731477e-05,
+      "loss": 0.6236,
+      "step": 5739
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8115948336503427e-05,
+      "loss": 0.6938,
+      "step": 5740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.811531215105054e-05,
+      "loss": 0.6681,
+      "step": 5741
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8114675869380366e-05,
+      "loss": 0.6302,
+      "step": 5742
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8114039491500442e-05,
+      "loss": 0.6404,
+      "step": 5743
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8113403017418318e-05,
+      "loss": 0.6375,
+      "step": 5744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.811276644714153e-05,
+      "loss": 0.6292,
+      "step": 5745
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.811212978067764e-05,
+      "loss": 0.6025,
+      "step": 5746
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8111493018034187e-05,
+      "loss": 0.6345,
+      "step": 5747
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8110856159218727e-05,
+      "loss": 0.6681,
+      "step": 5748
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8110219204238804e-05,
+      "loss": 0.6284,
+      "step": 5749
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8109582153101972e-05,
+      "loss": 0.7254,
+      "step": 5750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8108945005815784e-05,
+      "loss": 0.6527,
+      "step": 5751
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.81083077623878e-05,
+      "loss": 0.6655,
+      "step": 5752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8107670422825567e-05,
+      "loss": 0.6253,
+      "step": 5753
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8107032987136644e-05,
+      "loss": 0.5821,
+      "step": 5754
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.810639545532859e-05,
+      "loss": 0.5844,
+      "step": 5755
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8105757827408964e-05,
+      "loss": 0.644,
+      "step": 5756
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8105120103385327e-05,
+      "loss": 0.5669,
+      "step": 5757
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8104482283265236e-05,
+      "loss": 0.6485,
+      "step": 5758
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8103844367056257e-05,
+      "loss": 0.6757,
+      "step": 5759
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8103206354765946e-05,
+      "loss": 0.6231,
+      "step": 5760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.810256824640188e-05,
+      "loss": 0.6973,
+      "step": 5761
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8101930041971613e-05,
+      "loss": 0.6406,
+      "step": 5762
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.810129174148272e-05,
+      "loss": 0.6171,
+      "step": 5763
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8100653344942762e-05,
+      "loss": 0.5669,
+      "step": 5764
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8100014852359312e-05,
+      "loss": 0.6671,
+      "step": 5765
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.809937626373994e-05,
+      "loss": 0.6568,
+      "step": 5766
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.809873757909222e-05,
+      "loss": 0.6303,
+      "step": 5767
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8098098798423715e-05,
+      "loss": 0.6279,
+      "step": 5768
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8097459921742007e-05,
+      "loss": 0.5785,
+      "step": 5769
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8096820949054668e-05,
+      "loss": 0.6485,
+      "step": 5770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8096181880369273e-05,
+      "loss": 0.6257,
+      "step": 5771
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8095542715693402e-05,
+      "loss": 0.629,
+      "step": 5772
+    },
+    {
+      "epoch": 0.2,
+      "eval_loss": 0.580484926700592,
+      "eval_runtime": 4925.8584,
+      "eval_samples_per_second": 111.989,
+      "eval_steps_per_second": 37.33,
+      "step": 5772
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8094903455034634e-05,
+      "loss": 0.6303,
+      "step": 5773
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.809426409840054e-05,
+      "loss": 0.6349,
+      "step": 5774
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.809362464579871e-05,
+      "loss": 0.6056,
+      "step": 5775
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.809298509723672e-05,
+      "loss": 0.6424,
+      "step": 5776
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8092345452722152e-05,
+      "loss": 0.6212,
+      "step": 5777
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8091705712262594e-05,
+      "loss": 0.6177,
+      "step": 5778
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.809106587586563e-05,
+      "loss": 0.6557,
+      "step": 5779
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.809042594353884e-05,
+      "loss": 0.6419,
+      "step": 5780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.808978591528982e-05,
+      "loss": 0.652,
+      "step": 5781
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8089145791126156e-05,
+      "loss": 0.5965,
+      "step": 5782
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.808850557105543e-05,
+      "loss": 0.5975,
+      "step": 5783
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.808786525508524e-05,
+      "loss": 0.6019,
+      "step": 5784
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.808722484322318e-05,
+      "loss": 0.6316,
+      "step": 5785
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8086584335476835e-05,
+      "loss": 0.6188,
+      "step": 5786
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8085943731853807e-05,
+      "loss": 0.6465,
+      "step": 5787
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8085303032361686e-05,
+      "loss": 0.6494,
+      "step": 5788
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8084662237008065e-05,
+      "loss": 0.6187,
+      "step": 5789
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.808402134580055e-05,
+      "loss": 0.6777,
+      "step": 5790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8083380358746736e-05,
+      "loss": 0.592,
+      "step": 5791
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.808273927585422e-05,
+      "loss": 0.675,
+      "step": 5792
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.808209809713061e-05,
+      "loss": 0.6473,
+      "step": 5793
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.80814568225835e-05,
+      "loss": 0.6049,
+      "step": 5794
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8080815452220495e-05,
+      "loss": 0.6355,
+      "step": 5795
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8080173986049204e-05,
+      "loss": 0.6665,
+      "step": 5796
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8079532424077222e-05,
+      "loss": 0.6552,
+      "step": 5797
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.807889076631217e-05,
+      "loss": 0.5823,
+      "step": 5798
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8078249012761646e-05,
+      "loss": 0.6406,
+      "step": 5799
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.807760716343326e-05,
+      "loss": 0.5953,
+      "step": 5800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8076965218334622e-05,
+      "loss": 0.6558,
+      "step": 5801
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8076323177473346e-05,
+      "loss": 0.6123,
+      "step": 5802
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.807568104085704e-05,
+      "loss": 0.6494,
+      "step": 5803
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.807503880849332e-05,
+      "loss": 0.6335,
+      "step": 5804
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.80743964803898e-05,
+      "loss": 0.6866,
+      "step": 5805
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8073754056554097e-05,
+      "loss": 0.652,
+      "step": 5806
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8073111536993825e-05,
+      "loss": 0.6657,
+      "step": 5807
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8072468921716604e-05,
+      "loss": 0.675,
+      "step": 5808
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8071826210730054e-05,
+      "loss": 0.6383,
+      "step": 5809
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8071183404041788e-05,
+      "loss": 0.6181,
+      "step": 5810
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8070540501659436e-05,
+      "loss": 0.6427,
+      "step": 5811
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8069897503590618e-05,
+      "loss": 0.6114,
+      "step": 5812
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8069254409842957e-05,
+      "loss": 0.6551,
+      "step": 5813
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8068611220424074e-05,
+      "loss": 0.6589,
+      "step": 5814
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.80679679353416e-05,
+      "loss": 0.6764,
+      "step": 5815
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8067324554603162e-05,
+      "loss": 0.6316,
+      "step": 5816
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.806668107821638e-05,
+      "loss": 0.6187,
+      "step": 5817
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8066037506188896e-05,
+      "loss": 0.6209,
+      "step": 5818
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8065393838528332e-05,
+      "loss": 0.6849,
+      "step": 5819
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.806475007524232e-05,
+      "loss": 0.6178,
+      "step": 5820
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8064106216338495e-05,
+      "loss": 0.6413,
+      "step": 5821
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8063462261824488e-05,
+      "loss": 0.6736,
+      "step": 5822
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8062818211707935e-05,
+      "loss": 0.6468,
+      "step": 5823
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8062174065996477e-05,
+      "loss": 0.6868,
+      "step": 5824
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.806152982469774e-05,
+      "loss": 0.647,
+      "step": 5825
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8060885487819374e-05,
+      "loss": 0.5567,
+      "step": 5826
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8060241055369015e-05,
+      "loss": 0.6259,
+      "step": 5827
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8059596527354298e-05,
+      "loss": 0.6644,
+      "step": 5828
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.805895190378287e-05,
+      "loss": 0.6277,
+      "step": 5829
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8058307184662377e-05,
+      "loss": 0.6441,
+      "step": 5830
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8057662370000452e-05,
+      "loss": 0.6719,
+      "step": 5831
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8057017459804747e-05,
+      "loss": 0.6236,
+      "step": 5832
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.805637245408291e-05,
+      "loss": 0.6351,
+      "step": 5833
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8055727352842587e-05,
+      "loss": 0.6931,
+      "step": 5834
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8055082156091425e-05,
+      "loss": 0.5984,
+      "step": 5835
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8054436863837074e-05,
+      "loss": 0.6663,
+      "step": 5836
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8053791476087186e-05,
+      "loss": 0.6301,
+      "step": 5837
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.805314599284941e-05,
+      "loss": 0.6531,
+      "step": 5838
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8052500414131403e-05,
+      "loss": 0.5841,
+      "step": 5839
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8051854739940817e-05,
+      "loss": 0.649,
+      "step": 5840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8051208970285304e-05,
+      "loss": 0.676,
+      "step": 5841
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.805056310517253e-05,
+      "loss": 0.6285,
+      "step": 5842
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8049917144610138e-05,
+      "loss": 0.6587,
+      "step": 5843
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8049271088605803e-05,
+      "loss": 0.6402,
+      "step": 5844
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8048624937167174e-05,
+      "loss": 0.6738,
+      "step": 5845
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.804797869030191e-05,
+      "loss": 0.6242,
+      "step": 5846
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8047332348017686e-05,
+      "loss": 0.6853,
+      "step": 5847
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.804668591032215e-05,
+      "loss": 0.6479,
+      "step": 5848
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.804603937722298e-05,
+      "loss": 0.6085,
+      "step": 5849
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.804539274872783e-05,
+      "loss": 0.5331,
+      "step": 5850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8044746024844374e-05,
+      "loss": 0.6677,
+      "step": 5851
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8044099205580276e-05,
+      "loss": 0.6521,
+      "step": 5852
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8043452290943206e-05,
+      "loss": 0.6397,
+      "step": 5853
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8042805280940835e-05,
+      "loss": 0.6639,
+      "step": 5854
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8042158175580833e-05,
+      "loss": 0.6779,
+      "step": 5855
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8041510974870874e-05,
+      "loss": 0.612,
+      "step": 5856
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.804086367881863e-05,
+      "loss": 0.6291,
+      "step": 5857
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.804021628743177e-05,
+      "loss": 0.6329,
+      "step": 5858
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8039568800717977e-05,
+      "loss": 0.6307,
+      "step": 5859
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8038921218684932e-05,
+      "loss": 0.6249,
+      "step": 5860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8038273541340297e-05,
+      "loss": 0.6098,
+      "step": 5861
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8037625768691768e-05,
+      "loss": 0.6316,
+      "step": 5862
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8036977900747014e-05,
+      "loss": 0.6839,
+      "step": 5863
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8036329937513723e-05,
+      "loss": 0.5936,
+      "step": 5864
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8035681878999573e-05,
+      "loss": 0.6425,
+      "step": 5865
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8035033725212248e-05,
+      "loss": 0.6753,
+      "step": 5866
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8034385476159436e-05,
+      "loss": 0.6094,
+      "step": 5867
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8033737131848818e-05,
+      "loss": 0.588,
+      "step": 5868
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8033088692288084e-05,
+      "loss": 0.6909,
+      "step": 5869
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8032440157484926e-05,
+      "loss": 0.6004,
+      "step": 5870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8031791527447024e-05,
+      "loss": 0.6062,
+      "step": 5871
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8031142802182074e-05,
+      "loss": 0.6192,
+      "step": 5872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.803049398169777e-05,
+      "loss": 0.6976,
+      "step": 5873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8029845066001798e-05,
+      "loss": 0.6532,
+      "step": 5874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8029196055101857e-05,
+      "loss": 0.6215,
+      "step": 5875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.802854694900564e-05,
+      "loss": 0.6501,
+      "step": 5876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8027897747720837e-05,
+      "loss": 0.6465,
+      "step": 5877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8027248451255156e-05,
+      "loss": 0.6206,
+      "step": 5878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.802659905961629e-05,
+      "loss": 0.6583,
+      "step": 5879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.802594957281194e-05,
+      "loss": 0.6112,
+      "step": 5880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.80252999908498e-05,
+      "loss": 0.6552,
+      "step": 5881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.802465031373758e-05,
+      "loss": 0.6148,
+      "step": 5882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8024000541482978e-05,
+      "loss": 0.6512,
+      "step": 5883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.80233506740937e-05,
+      "loss": 0.6459,
+      "step": 5884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8022700711577447e-05,
+      "loss": 0.6096,
+      "step": 5885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8022050653941932e-05,
+      "loss": 0.6018,
+      "step": 5886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.802140050119486e-05,
+      "loss": 0.6165,
+      "step": 5887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.802075025334393e-05,
+      "loss": 0.6897,
+      "step": 5888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8020099910396867e-05,
+      "loss": 0.6179,
+      "step": 5889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8019449472361374e-05,
+      "loss": 0.626,
+      "step": 5890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.801879893924516e-05,
+      "loss": 0.6289,
+      "step": 5891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8018148311055944e-05,
+      "loss": 0.6233,
+      "step": 5892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8017497587801434e-05,
+      "loss": 0.6527,
+      "step": 5893
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8016846769489347e-05,
+      "loss": 0.6625,
+      "step": 5894
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8016195856127403e-05,
+      "loss": 0.6024,
+      "step": 5895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8015544847723314e-05,
+      "loss": 0.6915,
+      "step": 5896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.80148937442848e-05,
+      "loss": 0.6812,
+      "step": 5897
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8014242545819587e-05,
+      "loss": 0.6524,
+      "step": 5898
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8013591252335386e-05,
+      "loss": 0.6321,
+      "step": 5899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8012939863839926e-05,
+      "loss": 0.6187,
+      "step": 5900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8012288380340925e-05,
+      "loss": 0.6792,
+      "step": 5901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8011636801846112e-05,
+      "loss": 0.61,
+      "step": 5902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.801098512836321e-05,
+      "loss": 0.6305,
+      "step": 5903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8010333359899945e-05,
+      "loss": 0.6004,
+      "step": 5904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8009681496464045e-05,
+      "loss": 0.6494,
+      "step": 5905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8009029538063235e-05,
+      "loss": 0.647,
+      "step": 5906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8008377484705252e-05,
+      "loss": 0.6361,
+      "step": 5907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8007725336397823e-05,
+      "loss": 0.6355,
+      "step": 5908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8007073093148683e-05,
+      "loss": 0.6973,
+      "step": 5909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8006420754965557e-05,
+      "loss": 0.6393,
+      "step": 5910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8005768321856188e-05,
+      "loss": 0.6004,
+      "step": 5911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8005115793828307e-05,
+      "loss": 0.6088,
+      "step": 5912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.800446317088965e-05,
+      "loss": 0.6198,
+      "step": 5913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8003810453047958e-05,
+      "loss": 0.634,
+      "step": 5914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.800315764031097e-05,
+      "loss": 0.6331,
+      "step": 5915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8002504732686423e-05,
+      "loss": 0.66,
+      "step": 5916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.800185173018206e-05,
+      "loss": 0.6476,
+      "step": 5917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8001198632805617e-05,
+      "loss": 0.5753,
+      "step": 5918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.800054544056485e-05,
+      "loss": 0.5939,
+      "step": 5919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.799989215346749e-05,
+      "loss": 0.6588,
+      "step": 5920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7999238771521294e-05,
+      "loss": 0.6462,
+      "step": 5921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7998585294733997e-05,
+      "loss": 0.6502,
+      "step": 5922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.799793172311336e-05,
+      "loss": 0.668,
+      "step": 5923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7997278056667122e-05,
+      "loss": 0.5748,
+      "step": 5924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7996624295403037e-05,
+      "loss": 0.6025,
+      "step": 5925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7995970439328854e-05,
+      "loss": 0.6748,
+      "step": 5926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7995316488452324e-05,
+      "loss": 0.6589,
+      "step": 5927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7994662442781206e-05,
+      "loss": 0.6584,
+      "step": 5928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7994008302323253e-05,
+      "loss": 0.5808,
+      "step": 5929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7993354067086216e-05,
+      "loss": 0.6148,
+      "step": 5930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7992699737077857e-05,
+      "loss": 0.6087,
+      "step": 5931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.799204531230593e-05,
+      "loss": 0.6354,
+      "step": 5932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7991390792778198e-05,
+      "loss": 0.5697,
+      "step": 5933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7990736178502418e-05,
+      "loss": 0.602,
+      "step": 5934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7990081469486348e-05,
+      "loss": 0.6029,
+      "step": 5935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.798942666573776e-05,
+      "loss": 0.6255,
+      "step": 5936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7988771767264413e-05,
+      "loss": 0.6387,
+      "step": 5937
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7988116774074068e-05,
+      "loss": 0.6293,
+      "step": 5938
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7987461686174496e-05,
+      "loss": 0.6464,
+      "step": 5939
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.798680650357346e-05,
+      "loss": 0.5912,
+      "step": 5940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7986151226278726e-05,
+      "loss": 0.6229,
+      "step": 5941
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.798549585429807e-05,
+      "loss": 0.6987,
+      "step": 5942
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.798484038763926e-05,
+      "loss": 0.6536,
+      "step": 5943
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7984184826310062e-05,
+      "loss": 0.5952,
+      "step": 5944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.798352917031826e-05,
+      "loss": 0.6453,
+      "step": 5945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7982873419671615e-05,
+      "loss": 0.5765,
+      "step": 5946
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7982217574377905e-05,
+      "loss": 0.6254,
+      "step": 5947
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7981561634444914e-05,
+      "loss": 0.6603,
+      "step": 5948
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7980905599880412e-05,
+      "loss": 0.6411,
+      "step": 5949
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.798024947069218e-05,
+      "loss": 0.6461,
+      "step": 5950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.797959324688799e-05,
+      "loss": 0.603,
+      "step": 5951
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.797893692847563e-05,
+      "loss": 0.6164,
+      "step": 5952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7978280515462878e-05,
+      "loss": 0.6803,
+      "step": 5953
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7977624007857523e-05,
+      "loss": 0.6689,
+      "step": 5954
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.797696740566734e-05,
+      "loss": 0.6412,
+      "step": 5955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7976310708900115e-05,
+      "loss": 0.6871,
+      "step": 5956
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.797565391756364e-05,
+      "loss": 0.6548,
+      "step": 5957
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7974997031665698e-05,
+      "loss": 0.6671,
+      "step": 5958
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.797434005121408e-05,
+      "loss": 0.5975,
+      "step": 5959
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.797368297621657e-05,
+      "loss": 0.622,
+      "step": 5960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7973025806680962e-05,
+      "loss": 0.606,
+      "step": 5961
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7972368542615046e-05,
+      "loss": 0.6799,
+      "step": 5962
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7971711184026618e-05,
+      "loss": 0.6124,
+      "step": 5963
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7971053730923466e-05,
+      "loss": 0.6949,
+      "step": 5964
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7970396183313396e-05,
+      "loss": 0.6079,
+      "step": 5965
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.796973854120419e-05,
+      "loss": 0.6227,
+      "step": 5966
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7969080804603653e-05,
+      "loss": 0.6879,
+      "step": 5967
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7968422973519584e-05,
+      "loss": 0.6429,
+      "step": 5968
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.796776504795978e-05,
+      "loss": 0.6307,
+      "step": 5969
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.796710702793204e-05,
+      "loss": 0.5842,
+      "step": 5970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.796644891344417e-05,
+      "loss": 0.5808,
+      "step": 5971
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.796579070450397e-05,
+      "loss": 0.6307,
+      "step": 5972
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.796513240111924e-05,
+      "loss": 0.6474,
+      "step": 5973
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7964474003297797e-05,
+      "loss": 0.6618,
+      "step": 5974
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7963815511047435e-05,
+      "loss": 0.6607,
+      "step": 5975
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.796315692437597e-05,
+      "loss": 0.6114,
+      "step": 5976
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.79624982432912e-05,
+      "loss": 0.5861,
+      "step": 5977
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7961839467800947e-05,
+      "loss": 0.7041,
+      "step": 5978
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7961180597913014e-05,
+      "loss": 0.5995,
+      "step": 5979
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.796052163363521e-05,
+      "loss": 0.644,
+      "step": 5980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7959862574975356e-05,
+      "loss": 0.6936,
+      "step": 5981
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7959203421941257e-05,
+      "loss": 0.6726,
+      "step": 5982
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.795854417454074e-05,
+      "loss": 0.6409,
+      "step": 5983
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7957884832781607e-05,
+      "loss": 0.6081,
+      "step": 5984
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7957225396671685e-05,
+      "loss": 0.7048,
+      "step": 5985
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7956565866218793e-05,
+      "loss": 0.6027,
+      "step": 5986
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7955906241430745e-05,
+      "loss": 0.5902,
+      "step": 5987
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7955246522315365e-05,
+      "loss": 0.6269,
+      "step": 5988
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7954586708880475e-05,
+      "loss": 0.6093,
+      "step": 5989
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7953926801133894e-05,
+      "loss": 0.5901,
+      "step": 5990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7953266799083452e-05,
+      "loss": 0.6239,
+      "step": 5991
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.795260670273697e-05,
+      "loss": 0.6087,
+      "step": 5992
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7951946512102275e-05,
+      "loss": 0.6619,
+      "step": 5993
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7951286227187193e-05,
+      "loss": 0.6683,
+      "step": 5994
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.795062584799956e-05,
+      "loss": 0.6281,
+      "step": 5995
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7949965374547195e-05,
+      "loss": 0.6256,
+      "step": 5996
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7949304806837936e-05,
+      "loss": 0.678,
+      "step": 5997
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.794864414487961e-05,
+      "loss": 0.6704,
+      "step": 5998
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7947983388680053e-05,
+      "loss": 0.6584,
+      "step": 5999
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7947322538247104e-05,
+      "loss": 0.6378,
+      "step": 6000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7946661593588586e-05,
+      "loss": 0.6467,
+      "step": 6001
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7946000554712347e-05,
+      "loss": 0.6209,
+      "step": 6002
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7945339421626217e-05,
+      "loss": 0.6077,
+      "step": 6003
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.794467819433804e-05,
+      "loss": 0.6547,
+      "step": 6004
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.794401687285565e-05,
+      "loss": 0.6817,
+      "step": 6005
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7943355457186892e-05,
+      "loss": 0.6022,
+      "step": 6006
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7942693947339606e-05,
+      "loss": 0.62,
+      "step": 6007
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7942032343321637e-05,
+      "loss": 0.6712,
+      "step": 6008
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7941370645140827e-05,
+      "loss": 0.64,
+      "step": 6009
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.794070885280502e-05,
+      "loss": 0.5979,
+      "step": 6010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7940046966322068e-05,
+      "loss": 0.657,
+      "step": 6011
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7939384985699812e-05,
+      "loss": 0.6252,
+      "step": 6012
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7938722910946108e-05,
+      "loss": 0.6284,
+      "step": 6013
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7938060742068798e-05,
+      "loss": 0.6244,
+      "step": 6014
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7937398479075736e-05,
+      "loss": 0.6034,
+      "step": 6015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7936736121974776e-05,
+      "loss": 0.6191,
+      "step": 6016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.793607367077377e-05,
+      "loss": 0.661,
+      "step": 6017
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7935411125480567e-05,
+      "loss": 0.6471,
+      "step": 6018
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7934748486103027e-05,
+      "loss": 0.6087,
+      "step": 6019
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7934085752649012e-05,
+      "loss": 0.7051,
+      "step": 6020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.793342292512637e-05,
+      "loss": 0.6202,
+      "step": 6021
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7932760003542965e-05,
+      "loss": 0.6091,
+      "step": 6022
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7932096987906652e-05,
+      "loss": 0.6299,
+      "step": 6023
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.79314338782253e-05,
+      "loss": 0.6402,
+      "step": 6024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.793077067450676e-05,
+      "loss": 0.6261,
+      "step": 6025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7930107376758905e-05,
+      "loss": 0.6514,
+      "step": 6026
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7929443984989595e-05,
+      "loss": 0.6106,
+      "step": 6027
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7928780499206696e-05,
+      "loss": 0.6187,
+      "step": 6028
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.792811691941807e-05,
+      "loss": 0.6798,
+      "step": 6029
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7927453245631595e-05,
+      "loss": 0.6504,
+      "step": 6030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.792678947785513e-05,
+      "loss": 0.6717,
+      "step": 6031
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.792612561609655e-05,
+      "loss": 0.5778,
+      "step": 6032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7925461660363724e-05,
+      "loss": 0.6299,
+      "step": 6033
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7924797610664524e-05,
+      "loss": 0.6197,
+      "step": 6034
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7924133467006825e-05,
+      "loss": 0.5911,
+      "step": 6035
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7923469229398495e-05,
+      "loss": 0.6851,
+      "step": 6036
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7922804897847418e-05,
+      "loss": 0.5903,
+      "step": 6037
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7922140472361464e-05,
+      "loss": 0.6348,
+      "step": 6038
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.792147595294851e-05,
+      "loss": 0.5905,
+      "step": 6039
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7920811339616447e-05,
+      "loss": 0.6915,
+      "step": 6040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7920146632373142e-05,
+      "loss": 0.6434,
+      "step": 6041
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7919481831226477e-05,
+      "loss": 0.6245,
+      "step": 6042
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7918816936184336e-05,
+      "loss": 0.6414,
+      "step": 6043
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7918151947254607e-05,
+      "loss": 0.6844,
+      "step": 6044
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.791748686444517e-05,
+      "loss": 0.6872,
+      "step": 6045
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7916821687763907e-05,
+      "loss": 0.6197,
+      "step": 6046
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7916156417218712e-05,
+      "loss": 0.5847,
+      "step": 6047
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7915491052817465e-05,
+      "loss": 0.671,
+      "step": 6048
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.791482559456806e-05,
+      "loss": 0.6758,
+      "step": 6049
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7914160042478386e-05,
+      "loss": 0.6553,
+      "step": 6050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7913494396556334e-05,
+      "loss": 0.6147,
+      "step": 6051
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7912828656809794e-05,
+      "loss": 0.6582,
+      "step": 6052
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.791216282324666e-05,
+      "loss": 0.66,
+      "step": 6053
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7911496895874827e-05,
+      "loss": 0.6284,
+      "step": 6054
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.791083087470219e-05,
+      "loss": 0.6522,
+      "step": 6055
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7910164759736646e-05,
+      "loss": 0.6486,
+      "step": 6056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7909498550986092e-05,
+      "loss": 0.6004,
+      "step": 6057
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7908832248458427e-05,
+      "loss": 0.6341,
+      "step": 6058
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7908165852161552e-05,
+      "loss": 0.6242,
+      "step": 6059
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7907499362103364e-05,
+      "loss": 0.6163,
+      "step": 6060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7906832778291772e-05,
+      "loss": 0.6706,
+      "step": 6061
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7906166100734676e-05,
+      "loss": 0.6161,
+      "step": 6062
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7905499329439977e-05,
+      "loss": 0.6132,
+      "step": 6063
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.790483246441558e-05,
+      "loss": 0.6337,
+      "step": 6064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7904165505669398e-05,
+      "loss": 0.6938,
+      "step": 6065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.790349845320934e-05,
+      "loss": 0.6556,
+      "step": 6066
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7902831307043303e-05,
+      "loss": 0.6465,
+      "step": 6067
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7902164067179204e-05,
+      "loss": 0.6381,
+      "step": 6068
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7901496733624953e-05,
+      "loss": 0.5602,
+      "step": 6069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7900829306388465e-05,
+      "loss": 0.6365,
+      "step": 6070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7900161785477652e-05,
+      "loss": 0.6503,
+      "step": 6071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7899494170900427e-05,
+      "loss": 0.6849,
+      "step": 6072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7898826462664703e-05,
+      "loss": 0.677,
+      "step": 6073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.78981586607784e-05,
+      "loss": 0.6381,
+      "step": 6074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7897490765249438e-05,
+      "loss": 0.6655,
+      "step": 6075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.789682277608573e-05,
+      "loss": 0.6685,
+      "step": 6076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7896154693295197e-05,
+      "loss": 0.6341,
+      "step": 6077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7895486516885766e-05,
+      "loss": 0.6617,
+      "step": 6078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.789481824686535e-05,
+      "loss": 0.6477,
+      "step": 6079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7894149883241877e-05,
+      "loss": 0.6438,
+      "step": 6080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7893481426023273e-05,
+      "loss": 0.5927,
+      "step": 6081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7892812875217462e-05,
+      "loss": 0.6468,
+      "step": 6082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.789214423083237e-05,
+      "loss": 0.6987,
+      "step": 6083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7891475492875924e-05,
+      "loss": 0.6358,
+      "step": 6084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.789080666135605e-05,
+      "loss": 0.6404,
+      "step": 6085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7890137736280687e-05,
+      "loss": 0.6292,
+      "step": 6086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7889468717657753e-05,
+      "loss": 0.6323,
+      "step": 6087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.788879960549519e-05,
+      "loss": 0.6202,
+      "step": 6088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.788813039980093e-05,
+      "loss": 0.6311,
+      "step": 6089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7887461100582907e-05,
+      "loss": 0.5722,
+      "step": 6090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.788679170784905e-05,
+      "loss": 0.6735,
+      "step": 6091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7886122221607304e-05,
+      "loss": 0.5911,
+      "step": 6092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7885452641865605e-05,
+      "loss": 0.6326,
+      "step": 6093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7884782968631885e-05,
+      "loss": 0.6038,
+      "step": 6094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.788411320191409e-05,
+      "loss": 0.6091,
+      "step": 6095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7883443341720162e-05,
+      "loss": 0.6572,
+      "step": 6096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.788277338805804e-05,
+      "loss": 0.6086,
+      "step": 6097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7882103340935665e-05,
+      "loss": 0.6069,
+      "step": 6098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7881433200360983e-05,
+      "loss": 0.6172,
+      "step": 6099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7880762966341945e-05,
+      "loss": 0.6956,
+      "step": 6100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.788009263888649e-05,
+      "loss": 0.5686,
+      "step": 6101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7879422218002568e-05,
+      "loss": 0.628,
+      "step": 6102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.787875170369813e-05,
+      "loss": 0.6615,
+      "step": 6103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.787808109598112e-05,
+      "loss": 0.6263,
+      "step": 6104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.78774103948595e-05,
+      "loss": 0.629,
+      "step": 6105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.787673960034121e-05,
+      "loss": 0.5994,
+      "step": 6106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7876068712434207e-05,
+      "loss": 0.6324,
+      "step": 6107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7875397731146446e-05,
+      "loss": 0.6764,
+      "step": 6108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7874726656485887e-05,
+      "loss": 0.6457,
+      "step": 6109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7874055488460478e-05,
+      "loss": 0.673,
+      "step": 6110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.787338422707818e-05,
+      "loss": 0.5814,
+      "step": 6111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7872712872346952e-05,
+      "loss": 0.6448,
+      "step": 6112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7872041424274758e-05,
+      "loss": 0.6421,
+      "step": 6113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7871369882869554e-05,
+      "loss": 0.6754,
+      "step": 6114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.78706982481393e-05,
+      "loss": 0.6385,
+      "step": 6115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.787002652009196e-05,
+      "loss": 0.6419,
+      "step": 6116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7869354698735506e-05,
+      "loss": 0.6331,
+      "step": 6117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7868682784077892e-05,
+      "loss": 0.65,
+      "step": 6118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.786801077612709e-05,
+      "loss": 0.6352,
+      "step": 6119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.786733867489107e-05,
+      "loss": 0.5773,
+      "step": 6120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7866666480377795e-05,
+      "loss": 0.6385,
+      "step": 6121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.786599419259524e-05,
+      "loss": 0.6192,
+      "step": 6122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.786532181155137e-05,
+      "loss": 0.667,
+      "step": 6123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7864649337254163e-05,
+      "loss": 0.6562,
+      "step": 6124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.786397676971159e-05,
+      "loss": 0.5974,
+      "step": 6125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.786330410893162e-05,
+      "loss": 0.6057,
+      "step": 6126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.786263135492224e-05,
+      "loss": 0.5961,
+      "step": 6127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7861958507691414e-05,
+      "loss": 0.639,
+      "step": 6128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7861285567247123e-05,
+      "loss": 0.6695,
+      "step": 6129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7860612533597352e-05,
+      "loss": 0.6303,
+      "step": 6130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7859939406750075e-05,
+      "loss": 0.6925,
+      "step": 6131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.785926618671327e-05,
+      "loss": 0.6145,
+      "step": 6132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.785859287349492e-05,
+      "loss": 0.6683,
+      "step": 6133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7857919467103016e-05,
+      "loss": 0.5773,
+      "step": 6134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7857245967545534e-05,
+      "loss": 0.5925,
+      "step": 6135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7856572374830464e-05,
+      "loss": 0.6471,
+      "step": 6136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.785589868896579e-05,
+      "loss": 0.6532,
+      "step": 6137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7855224909959493e-05,
+      "loss": 0.6343,
+      "step": 6138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7854551037819573e-05,
+      "loss": 0.6929,
+      "step": 6139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7853877072554012e-05,
+      "loss": 0.626,
+      "step": 6140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.785320301417081e-05,
+      "loss": 0.6681,
+      "step": 6141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7852528862677943e-05,
+      "loss": 0.6007,
+      "step": 6142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7851854618083416e-05,
+      "loss": 0.5735,
+      "step": 6143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.785118028039522e-05,
+      "loss": 0.6079,
+      "step": 6144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.785050584962135e-05,
+      "loss": 0.6605,
+      "step": 6145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.78498313257698e-05,
+      "loss": 0.6478,
+      "step": 6146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7849156708848573e-05,
+      "loss": 0.6279,
+      "step": 6147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7848481998865665e-05,
+      "loss": 0.6629,
+      "step": 6148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7847807195829072e-05,
+      "loss": 0.623,
+      "step": 6149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7847132299746798e-05,
+      "loss": 0.6877,
+      "step": 6150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.784645731062684e-05,
+      "loss": 0.6204,
+      "step": 6151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7845782228477212e-05,
+      "loss": 0.6084,
+      "step": 6152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7845107053305906e-05,
+      "loss": 0.6416,
+      "step": 6153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7844431785120934e-05,
+      "loss": 0.6478,
+      "step": 6154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7843756423930296e-05,
+      "loss": 0.5702,
+      "step": 6155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7843080969742004e-05,
+      "loss": 0.6489,
+      "step": 6156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.784240542256407e-05,
+      "loss": 0.644,
+      "step": 6157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7841729782404494e-05,
+      "loss": 0.6528,
+      "step": 6158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7841054049271295e-05,
+      "loss": 0.6442,
+      "step": 6159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.784037822317248e-05,
+      "loss": 0.6461,
+      "step": 6160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7839702304116063e-05,
+      "loss": 0.6876,
+      "step": 6161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.783902629211006e-05,
+      "loss": 0.6156,
+      "step": 6162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7838350187162482e-05,
+      "loss": 0.5995,
+      "step": 6163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7837673989281346e-05,
+      "loss": 0.6038,
+      "step": 6164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7836997698474672e-05,
+      "loss": 0.6558,
+      "step": 6165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7836321314750472e-05,
+      "loss": 0.6285,
+      "step": 6166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7835644838116776e-05,
+      "loss": 0.6205,
+      "step": 6167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7834968268581595e-05,
+      "loss": 0.6318,
+      "step": 6168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7834291606152957e-05,
+      "loss": 0.702,
+      "step": 6169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7833614850838876e-05,
+      "loss": 0.6477,
+      "step": 6170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7832938002647385e-05,
+      "loss": 0.6632,
+      "step": 6171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7832261061586504e-05,
+      "loss": 0.5773,
+      "step": 6172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7831584027664262e-05,
+      "loss": 0.6615,
+      "step": 6173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7830906900888687e-05,
+      "loss": 0.629,
+      "step": 6174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7830229681267797e-05,
+      "loss": 0.6246,
+      "step": 6175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7829552368809633e-05,
+      "loss": 0.6372,
+      "step": 6176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.782887496352222e-05,
+      "loss": 0.5479,
+      "step": 6177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7828197465413595e-05,
+      "loss": 0.6194,
+      "step": 6178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7827519874491784e-05,
+      "loss": 0.6537,
+      "step": 6179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7826842190764822e-05,
+      "loss": 0.6131,
+      "step": 6180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.782616441424074e-05,
+      "loss": 0.6647,
+      "step": 6181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7825486544927587e-05,
+      "loss": 0.6643,
+      "step": 6182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7824808582833387e-05,
+      "loss": 0.622,
+      "step": 6183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7824130527966183e-05,
+      "loss": 0.6473,
+      "step": 6184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7823452380334017e-05,
+      "loss": 0.6291,
+      "step": 6185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7822774139944924e-05,
+      "loss": 0.6328,
+      "step": 6186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7822095806806946e-05,
+      "loss": 0.617,
+      "step": 6187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7821417380928133e-05,
+      "loss": 0.6025,
+      "step": 6188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.782073886231652e-05,
+      "loss": 0.5982,
+      "step": 6189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7820060250980155e-05,
+      "loss": 0.6339,
+      "step": 6190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.781938154692708e-05,
+      "loss": 0.593,
+      "step": 6191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7818702750165346e-05,
+      "loss": 0.6542,
+      "step": 6192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7818023860703003e-05,
+      "loss": 0.6551,
+      "step": 6193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7817344878548092e-05,
+      "loss": 0.6236,
+      "step": 6194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7816665803708675e-05,
+      "loss": 0.6038,
+      "step": 6195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7815986636192794e-05,
+      "loss": 0.6341,
+      "step": 6196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7815307376008504e-05,
+      "loss": 0.6515,
+      "step": 6197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7814628023163858e-05,
+      "loss": 0.6307,
+      "step": 6198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7813948577666912e-05,
+      "loss": 0.6308,
+      "step": 6199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.781326903952572e-05,
+      "loss": 0.6149,
+      "step": 6200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.781258940874834e-05,
+      "loss": 0.6056,
+      "step": 6201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.781190968534283e-05,
+      "loss": 0.5867,
+      "step": 6202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.781122986931725e-05,
+      "loss": 0.6762,
+      "step": 6203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.781054996067966e-05,
+      "loss": 0.6363,
+      "step": 6204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.7809869959438115e-05,
+      "loss": 0.6786,
+      "step": 6205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.780918986560068e-05,
+      "loss": 0.6414,
+      "step": 6206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7808509679175427e-05,
+      "loss": 0.6723,
+      "step": 6207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7807829400170414e-05,
+      "loss": 0.6461,
+      "step": 6208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7807149028593703e-05,
+      "loss": 0.6517,
+      "step": 6209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7806468564453366e-05,
+      "loss": 0.6254,
+      "step": 6210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.780578800775747e-05,
+      "loss": 0.6789,
+      "step": 6211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.780510735851408e-05,
+      "loss": 0.6261,
+      "step": 6212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7804426616731268e-05,
+      "loss": 0.6679,
+      "step": 6213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7803745782417107e-05,
+      "loss": 0.6047,
+      "step": 6214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.780306485557967e-05,
+      "loss": 0.6531,
+      "step": 6215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7802383836227027e-05,
+      "loss": 0.6411,
+      "step": 6216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7801702724367253e-05,
+      "loss": 0.6697,
+      "step": 6217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7801021520008423e-05,
+      "loss": 0.6687,
+      "step": 6218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7800340223158615e-05,
+      "loss": 0.6406,
+      "step": 6219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.779965883382591e-05,
+      "loss": 0.6704,
+      "step": 6220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.779897735201838e-05,
+      "loss": 0.6434,
+      "step": 6221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7798295777744105e-05,
+      "loss": 0.6008,
+      "step": 6222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7797614111011175e-05,
+      "loss": 0.6264,
+      "step": 6223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.779693235182766e-05,
+      "loss": 0.6435,
+      "step": 6224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.779625050020165e-05,
+      "loss": 0.6582,
+      "step": 6225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.779556855614123e-05,
+      "loss": 0.6374,
+      "step": 6226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7794886519654482e-05,
+      "loss": 0.6624,
+      "step": 6227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7794204390749493e-05,
+      "loss": 0.6432,
+      "step": 6228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.779352216943435e-05,
+      "loss": 0.5938,
+      "step": 6229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.779283985571715e-05,
+      "loss": 0.6533,
+      "step": 6230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7792157449605967e-05,
+      "loss": 0.6807,
+      "step": 6231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7791474951108905e-05,
+      "loss": 0.6599,
+      "step": 6232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.779079236023405e-05,
+      "loss": 0.6318,
+      "step": 6233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7790109676989496e-05,
+      "loss": 0.6176,
+      "step": 6234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7789426901383335e-05,
+      "loss": 0.6133,
+      "step": 6235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7788744033423668e-05,
+      "loss": 0.649,
+      "step": 6236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7788061073118584e-05,
+      "loss": 0.6679,
+      "step": 6237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7787378020476193e-05,
+      "loss": 0.6258,
+      "step": 6238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7786694875504575e-05,
+      "loss": 0.5994,
+      "step": 6239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.778601163821184e-05,
+      "loss": 0.6717,
+      "step": 6240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.778532830860609e-05,
+      "loss": 0.6276,
+      "step": 6241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7784644886695427e-05,
+      "loss": 0.5737,
+      "step": 6242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7783961372487947e-05,
+      "loss": 0.6178,
+      "step": 6243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.778327776599176e-05,
+      "loss": 0.6661,
+      "step": 6244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.778259406721497e-05,
+      "loss": 0.603,
+      "step": 6245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.778191027616568e-05,
+      "loss": 0.6039,
+      "step": 6246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7781226392851997e-05,
+      "loss": 0.6035,
+      "step": 6247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.778054241728204e-05,
+      "loss": 0.6409,
+      "step": 6248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7779858349463904e-05,
+      "loss": 0.6917,
+      "step": 6249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7779174189405708e-05,
+      "loss": 0.6479,
+      "step": 6250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.777848993711556e-05,
+      "loss": 0.6668,
+      "step": 6251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7777805592601575e-05,
+      "loss": 0.6926,
+      "step": 6252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7777121155871864e-05,
+      "loss": 0.638,
+      "step": 6253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7776436626934543e-05,
+      "loss": 0.6134,
+      "step": 6254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.777575200579773e-05,
+      "loss": 0.6788,
+      "step": 6255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7775067292469545e-05,
+      "loss": 0.641,
+      "step": 6256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7774382486958094e-05,
+      "loss": 0.6338,
+      "step": 6257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7773697589271508e-05,
+      "loss": 0.6775,
+      "step": 6258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7773012599417902e-05,
+      "loss": 0.6311,
+      "step": 6259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.77723275174054e-05,
+      "loss": 0.6247,
+      "step": 6260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.777164234324212e-05,
+      "loss": 0.6408,
+      "step": 6261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7770957076936195e-05,
+      "loss": 0.6323,
+      "step": 6262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.777027171849574e-05,
+      "loss": 0.6206,
+      "step": 6263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7769586267928883e-05,
+      "loss": 0.6192,
+      "step": 6264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7768900725243753e-05,
+      "loss": 0.6557,
+      "step": 6265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.776821509044848e-05,
+      "loss": 0.5892,
+      "step": 6266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7767529363551188e-05,
+      "loss": 0.5731,
+      "step": 6267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.776684354456001e-05,
+      "loss": 0.6482,
+      "step": 6268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7766157633483075e-05,
+      "loss": 0.6386,
+      "step": 6269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7765471630328518e-05,
+      "loss": 0.6781,
+      "step": 6270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7764785535104474e-05,
+      "loss": 0.626,
+      "step": 6271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7764099347819073e-05,
+      "loss": 0.6474,
+      "step": 6272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.776341306848045e-05,
+      "loss": 0.6227,
+      "step": 6273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7762726697096748e-05,
+      "loss": 0.606,
+      "step": 6274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7762040233676102e-05,
+      "loss": 0.689,
+      "step": 6275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7761353678226644e-05,
+      "loss": 0.6072,
+      "step": 6276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7760667030756526e-05,
+      "loss": 0.6837,
+      "step": 6277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.775998029127388e-05,
+      "loss": 0.6248,
+      "step": 6278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.775929345978685e-05,
+      "loss": 0.6791,
+      "step": 6279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7758606536303586e-05,
+      "loss": 0.6327,
+      "step": 6280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.775791952083222e-05,
+      "loss": 0.6113,
+      "step": 6281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.775723241338091e-05,
+      "loss": 0.6094,
+      "step": 6282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7756545213957793e-05,
+      "loss": 0.6435,
+      "step": 6283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7755857922571023e-05,
+      "loss": 0.6563,
+      "step": 6284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7755170539228744e-05,
+      "loss": 0.6195,
+      "step": 6285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.775448306393911e-05,
+      "loss": 0.609,
+      "step": 6286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7753795496710267e-05,
+      "loss": 0.6576,
+      "step": 6287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7753107837550373e-05,
+      "loss": 0.6118,
+      "step": 6288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7752420086467577e-05,
+      "loss": 0.6389,
+      "step": 6289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.775173224347003e-05,
+      "loss": 0.6157,
+      "step": 6290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7751044308565894e-05,
+      "loss": 0.6306,
+      "step": 6291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7750356281763326e-05,
+      "loss": 0.6314,
+      "step": 6292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7749668163070474e-05,
+      "loss": 0.6166,
+      "step": 6293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7748979952495507e-05,
+      "loss": 0.5925,
+      "step": 6294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.774829165004658e-05,
+      "loss": 0.6589,
+      "step": 6295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.774760325573185e-05,
+      "loss": 0.6165,
+      "step": 6296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7746914769559488e-05,
+      "loss": 0.5919,
+      "step": 6297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7746226191537647e-05,
+      "loss": 0.6593,
+      "step": 6298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.77455375216745e-05,
+      "loss": 0.6533,
+      "step": 6299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7744848759978202e-05,
+      "loss": 0.6723,
+      "step": 6300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.774415990645693e-05,
+      "loss": 0.6303,
+      "step": 6301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.774347096111884e-05,
+      "loss": 0.656,
+      "step": 6302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.774278192397211e-05,
+      "loss": 0.6594,
+      "step": 6303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7742092795024906e-05,
+      "loss": 0.6584,
+      "step": 6304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7741403574285398e-05,
+      "loss": 0.5816,
+      "step": 6305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7740714261761756e-05,
+      "loss": 0.6293,
+      "step": 6306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7740024857462158e-05,
+      "loss": 0.6605,
+      "step": 6307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.773933536139477e-05,
+      "loss": 0.6526,
+      "step": 6308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7738645773567774e-05,
+      "loss": 0.6455,
+      "step": 6309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.773795609398934e-05,
+      "loss": 0.6057,
+      "step": 6310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7737266322667647e-05,
+      "loss": 0.6522,
+      "step": 6311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.773657645961088e-05,
+      "loss": 0.6662,
+      "step": 6312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7735886504827207e-05,
+      "loss": 0.5417,
+      "step": 6313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7735196458324816e-05,
+      "loss": 0.7134,
+      "step": 6314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7734506320111885e-05,
+      "loss": 0.678,
+      "step": 6315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7733816090196598e-05,
+      "loss": 0.6079,
+      "step": 6316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7733125768587135e-05,
+      "loss": 0.6184,
+      "step": 6317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7732435355291685e-05,
+      "loss": 0.6572,
+      "step": 6318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7731744850318438e-05,
+      "loss": 0.6413,
+      "step": 6319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7731054253675568e-05,
+      "loss": 0.6775,
+      "step": 6320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7730363565371274e-05,
+      "loss": 0.6046,
+      "step": 6321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7729672785413737e-05,
+      "loss": 0.6735,
+      "step": 6322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7728981913811155e-05,
+      "loss": 0.6362,
+      "step": 6323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7728290950571714e-05,
+      "loss": 0.6304,
+      "step": 6324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7727599895703606e-05,
+      "loss": 0.6117,
+      "step": 6325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7726908749215028e-05,
+      "loss": 0.6547,
+      "step": 6326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7726217511114168e-05,
+      "loss": 0.6525,
+      "step": 6327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7725526181409232e-05,
+      "loss": 0.5855,
+      "step": 6328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7724834760108403e-05,
+      "loss": 0.6075,
+      "step": 6329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.772414324721989e-05,
+      "loss": 0.6118,
+      "step": 6330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.772345164275189e-05,
+      "loss": 0.6547,
+      "step": 6331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7722759946712598e-05,
+      "loss": 0.5965,
+      "step": 6332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7722068159110216e-05,
+      "loss": 0.6437,
+      "step": 6333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7721376279952948e-05,
+      "loss": 0.6355,
+      "step": 6334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7720684309249e-05,
+      "loss": 0.6536,
+      "step": 6335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.771999224700657e-05,
+      "loss": 0.6447,
+      "step": 6336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.771930009323387e-05,
+      "loss": 0.6461,
+      "step": 6337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.77186078479391e-05,
+      "loss": 0.5968,
+      "step": 6338
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7717915511130465e-05,
+      "loss": 0.7068,
+      "step": 6339
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7717223082816183e-05,
+      "loss": 0.6066,
+      "step": 6340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7716530563004458e-05,
+      "loss": 0.622,
+      "step": 6341
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7715837951703504e-05,
+      "loss": 0.655,
+      "step": 6342
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7715145248921525e-05,
+      "loss": 0.6378,
+      "step": 6343
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7714452454666744e-05,
+      "loss": 0.6497,
+      "step": 6344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.771375956894737e-05,
+      "loss": 0.6836,
+      "step": 6345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.771306659177162e-05,
+      "loss": 0.661,
+      "step": 6346
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7712373523147704e-05,
+      "loss": 0.6046,
+      "step": 6347
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7711680363083845e-05,
+      "loss": 0.622,
+      "step": 6348
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.771098711158826e-05,
+      "loss": 0.7245,
+      "step": 6349
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.771029376866917e-05,
+      "loss": 0.6822,
+      "step": 6350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7709600334334794e-05,
+      "loss": 0.6306,
+      "step": 6351
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.770890680859335e-05,
+      "loss": 0.604,
+      "step": 6352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7708213191453065e-05,
+      "loss": 0.6336,
+      "step": 6353
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7707519482922165e-05,
+      "loss": 0.6488,
+      "step": 6354
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.770682568300887e-05,
+      "loss": 0.6491,
+      "step": 6355
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7706131791721406e-05,
+      "loss": 0.6283,
+      "step": 6356
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7705437809068e-05,
+      "loss": 0.6717,
+      "step": 6357
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7704743735056883e-05,
+      "loss": 0.6395,
+      "step": 6358
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7704049569696284e-05,
+      "loss": 0.6365,
+      "step": 6359
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.770335531299443e-05,
+      "loss": 0.6563,
+      "step": 6360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7702660964959557e-05,
+      "loss": 0.637,
+      "step": 6361
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7701966525599893e-05,
+      "loss": 0.6878,
+      "step": 6362
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7701271994923667e-05,
+      "loss": 0.6833,
+      "step": 6363
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.770057737293913e-05,
+      "loss": 0.5652,
+      "step": 6364
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7699882659654498e-05,
+      "loss": 0.6598,
+      "step": 6365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7699187855078018e-05,
+      "loss": 0.5691,
+      "step": 6366
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.769849295921793e-05,
+      "loss": 0.5826,
+      "step": 6367
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7697797972082465e-05,
+      "loss": 0.5988,
+      "step": 6368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.769710289367987e-05,
+      "loss": 0.6696,
+      "step": 6369
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7696407724018384e-05,
+      "loss": 0.657,
+      "step": 6370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7695712463106245e-05,
+      "loss": 0.6536,
+      "step": 6371
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7695017110951703e-05,
+      "loss": 0.6693,
+      "step": 6372
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7694321667563e-05,
+      "loss": 0.6919,
+      "step": 6373
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7693626132948378e-05,
+      "loss": 0.5516,
+      "step": 6374
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7692930507116084e-05,
+      "loss": 0.6466,
+      "step": 6375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.769223479007437e-05,
+      "loss": 0.6162,
+      "step": 6376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.769153898183148e-05,
+      "loss": 0.598,
+      "step": 6377
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7690843082395666e-05,
+      "loss": 0.6393,
+      "step": 6378
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7690147091775177e-05,
+      "loss": 0.6404,
+      "step": 6379
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.768945100997827e-05,
+      "loss": 0.6197,
+      "step": 6380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.768875483701319e-05,
+      "loss": 0.6315,
+      "step": 6381
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.76880585728882e-05,
+      "loss": 0.6745,
+      "step": 6382
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7687362217611545e-05,
+      "loss": 0.609,
+      "step": 6383
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.768666577119149e-05,
+      "loss": 0.6607,
+      "step": 6384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7685969233636286e-05,
+      "loss": 0.6393,
+      "step": 6385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7685272604954195e-05,
+      "loss": 0.6545,
+      "step": 6386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7684575885153474e-05,
+      "loss": 0.6277,
+      "step": 6387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7683879074242388e-05,
+      "loss": 0.5958,
+      "step": 6388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7683182172229192e-05,
+      "loss": 0.6146,
+      "step": 6389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7682485179122157e-05,
+      "loss": 0.6337,
+      "step": 6390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7681788094929535e-05,
+      "loss": 0.6648,
+      "step": 6391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7681090919659604e-05,
+      "loss": 0.6125,
+      "step": 6392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.768039365332062e-05,
+      "loss": 0.6527,
+      "step": 6393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7679696295920852e-05,
+      "loss": 0.6653,
+      "step": 6394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7678998847468572e-05,
+      "loss": 0.6559,
+      "step": 6395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7678301307972048e-05,
+      "loss": 0.6422,
+      "step": 6396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7677603677439546e-05,
+      "loss": 0.6412,
+      "step": 6397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.767690595587934e-05,
+      "loss": 0.6402,
+      "step": 6398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7676208143299706e-05,
+      "loss": 0.5953,
+      "step": 6399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7675510239708912e-05,
+      "loss": 0.6393,
+      "step": 6400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7674812245115233e-05,
+      "loss": 0.6043,
+      "step": 6401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.767411415952695e-05,
+      "loss": 0.6521,
+      "step": 6402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7673415982952333e-05,
+      "loss": 0.6097,
+      "step": 6403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7672717715399664e-05,
+      "loss": 0.7108,
+      "step": 6404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.767201935687722e-05,
+      "loss": 0.5815,
+      "step": 6405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.767132090739328e-05,
+      "loss": 0.6251,
+      "step": 6406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7670622366956126e-05,
+      "loss": 0.6521,
+      "step": 6407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7669923735574045e-05,
+      "loss": 0.6249,
+      "step": 6408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7669225013255312e-05,
+      "loss": 0.7024,
+      "step": 6409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7668526200008212e-05,
+      "loss": 0.6622,
+      "step": 6410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7667827295841037e-05,
+      "loss": 0.5927,
+      "step": 6411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7667128300762073e-05,
+      "loss": 0.5946,
+      "step": 6412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7666429214779597e-05,
+      "loss": 0.5833,
+      "step": 6413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7665730037901907e-05,
+      "loss": 0.6634,
+      "step": 6414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.766503077013729e-05,
+      "loss": 0.6608,
+      "step": 6415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.766433141149404e-05,
+      "loss": 0.6102,
+      "step": 6416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7663631961980444e-05,
+      "loss": 0.6371,
+      "step": 6417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7662932421604796e-05,
+      "loss": 0.5829,
+      "step": 6418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.766223279037539e-05,
+      "loss": 0.6739,
+      "step": 6419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.766153306830052e-05,
+      "loss": 0.6459,
+      "step": 6420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7660833255388486e-05,
+      "loss": 0.6202,
+      "step": 6421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7660133351647582e-05,
+      "loss": 0.6812,
+      "step": 6422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7659433357086108e-05,
+      "loss": 0.6098,
+      "step": 6423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.765873327171236e-05,
+      "loss": 0.6911,
+      "step": 6424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7658033095534643e-05,
+      "loss": 0.5746,
+      "step": 6425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7657332828561255e-05,
+      "loss": 0.6476,
+      "step": 6426
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.76566324708005e-05,
+      "loss": 0.6205,
+      "step": 6427
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7655932022260682e-05,
+      "loss": 0.677,
+      "step": 6428
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7655231482950107e-05,
+      "loss": 0.6709,
+      "step": 6429
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7654530852877075e-05,
+      "loss": 0.6492,
+      "step": 6430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7653830132049898e-05,
+      "loss": 0.6276,
+      "step": 6431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7653129320476882e-05,
+      "loss": 0.6605,
+      "step": 6432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7652428418166336e-05,
+      "loss": 0.6458,
+      "step": 6433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7651727425126574e-05,
+      "loss": 0.6309,
+      "step": 6434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.76510263413659e-05,
+      "loss": 0.6636,
+      "step": 6435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.765032516689263e-05,
+      "loss": 0.6406,
+      "step": 6436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7649623901715084e-05,
+      "loss": 0.6791,
+      "step": 6437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7648922545841567e-05,
+      "loss": 0.6589,
+      "step": 6438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7648221099280395e-05,
+      "loss": 0.6564,
+      "step": 6439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7647519562039885e-05,
+      "loss": 0.6246,
+      "step": 6440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.764681793412836e-05,
+      "loss": 0.6476,
+      "step": 6441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7646116215554134e-05,
+      "loss": 0.6927,
+      "step": 6442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7645414406325527e-05,
+      "loss": 0.6752,
+      "step": 6443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7644712506450862e-05,
+      "loss": 0.6573,
+      "step": 6444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.764401051593846e-05,
+      "loss": 0.6245,
+      "step": 6445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.764330843479664e-05,
+      "loss": 0.664,
+      "step": 6446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7642606263033735e-05,
+      "loss": 0.6486,
+      "step": 6447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7641904000658062e-05,
+      "loss": 0.5877,
+      "step": 6448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7641201647677952e-05,
+      "loss": 0.6907,
+      "step": 6449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.764049920410173e-05,
+      "loss": 0.6861,
+      "step": 6450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.763979666993772e-05,
+      "loss": 0.6867,
+      "step": 6451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.763909404519426e-05,
+      "loss": 0.6527,
+      "step": 6452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7638391329879677e-05,
+      "loss": 0.6752,
+      "step": 6453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7637688524002302e-05,
+      "loss": 0.6367,
+      "step": 6454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7636985627570468e-05,
+      "loss": 0.675,
+      "step": 6455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7636282640592505e-05,
+      "loss": 0.6748,
+      "step": 6456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7635579563076753e-05,
+      "loss": 0.5962,
+      "step": 6457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.763487639503155e-05,
+      "loss": 0.585,
+      "step": 6458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7634173136465228e-05,
+      "loss": 0.6575,
+      "step": 6459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7633469787386122e-05,
+      "loss": 0.6399,
+      "step": 6460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7632766347802576e-05,
+      "loss": 0.6907,
+      "step": 6461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7632062817722934e-05,
+      "loss": 0.6834,
+      "step": 6462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7631359197155528e-05,
+      "loss": 0.6881,
+      "step": 6463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.763065548610871e-05,
+      "loss": 0.6117,
+      "step": 6464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7629951684590813e-05,
+      "loss": 0.6307,
+      "step": 6465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.762924779261019e-05,
+      "loss": 0.6526,
+      "step": 6466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7628543810175183e-05,
+      "loss": 0.6274,
+      "step": 6467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7627839737294138e-05,
+      "loss": 0.6567,
+      "step": 6468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7627135573975402e-05,
+      "loss": 0.6109,
+      "step": 6469
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.762643132022733e-05,
+      "loss": 0.6882,
+      "step": 6470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7625726976058265e-05,
+      "loss": 0.6601,
+      "step": 6471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.762502254147656e-05,
+      "loss": 0.6306,
+      "step": 6472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7624318016490565e-05,
+      "loss": 0.6456,
+      "step": 6473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7623613401108638e-05,
+      "loss": 0.6722,
+      "step": 6474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7622908695339127e-05,
+      "loss": 0.5628,
+      "step": 6475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7622203899190394e-05,
+      "loss": 0.6852,
+      "step": 6476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.762149901267079e-05,
+      "loss": 0.6222,
+      "step": 6477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7620794035788673e-05,
+      "loss": 0.6415,
+      "step": 6478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7620088968552405e-05,
+      "loss": 0.6762,
+      "step": 6479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7619383810970343e-05,
+      "loss": 0.6548,
+      "step": 6480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7618678563050845e-05,
+      "loss": 0.6462,
+      "step": 6481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7617973224802276e-05,
+      "loss": 0.6201,
+      "step": 6482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7617267796232998e-05,
+      "loss": 0.6712,
+      "step": 6483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7616562277351373e-05,
+      "loss": 0.6239,
+      "step": 6484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7615856668165767e-05,
+      "loss": 0.611,
+      "step": 6485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7615150968684548e-05,
+      "loss": 0.6264,
+      "step": 6486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.761444517891608e-05,
+      "loss": 0.6335,
+      "step": 6487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7613739298868732e-05,
+      "loss": 0.6559,
+      "step": 6488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7613033328550872e-05,
+      "loss": 0.6428,
+      "step": 6489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.761232726797087e-05,
+      "loss": 0.678,
+      "step": 6490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.7611621117137104e-05,
+      "loss": 0.608,
+      "step": 6491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.761091487605794e-05,
+      "loss": 0.654,
+      "step": 6492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.761020854474175e-05,
+      "loss": 0.6302,
+      "step": 6493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.760950212319691e-05,
+      "loss": 0.6077,
+      "step": 6494
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7608795611431796e-05,
+      "loss": 0.6231,
+      "step": 6495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7608089009454787e-05,
+      "loss": 0.6681,
+      "step": 6496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7607382317274257e-05,
+      "loss": 0.6383,
+      "step": 6497
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7606675534898586e-05,
+      "loss": 0.6991,
+      "step": 6498
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7605968662336156e-05,
+      "loss": 0.6879,
+      "step": 6499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7605261699595345e-05,
+      "loss": 0.6172,
+      "step": 6500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7604554646684537e-05,
+      "loss": 0.6621,
+      "step": 6501
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7603847503612113e-05,
+      "loss": 0.6532,
+      "step": 6502
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.760314027038646e-05,
+      "loss": 0.6292,
+      "step": 6503
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7602432947015958e-05,
+      "loss": 0.6177,
+      "step": 6504
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7601725533508994e-05,
+      "loss": 0.656,
+      "step": 6505
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7601018029873963e-05,
+      "loss": 0.6029,
+      "step": 6506
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7600310436119245e-05,
+      "loss": 0.6543,
+      "step": 6507
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7599602752253232e-05,
+      "loss": 0.6431,
+      "step": 6508
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7598894978284316e-05,
+      "loss": 0.6072,
+      "step": 6509
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7598187114220887e-05,
+      "loss": 0.5566,
+      "step": 6510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.759747916007134e-05,
+      "loss": 0.5789,
+      "step": 6511
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7596771115844063e-05,
+      "loss": 0.6472,
+      "step": 6512
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7596062981547456e-05,
+      "loss": 0.6042,
+      "step": 6513
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.759535475718991e-05,
+      "loss": 0.6776,
+      "step": 6514
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.759464644277983e-05,
+      "loss": 0.6492,
+      "step": 6515
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.759393803832561e-05,
+      "loss": 0.6801,
+      "step": 6516
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7593229543835643e-05,
+      "loss": 0.6543,
+      "step": 6517
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7592520959318338e-05,
+      "loss": 0.6248,
+      "step": 6518
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7591812284782092e-05,
+      "loss": 0.6491,
+      "step": 6519
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.759110352023531e-05,
+      "loss": 0.6699,
+      "step": 6520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.759039466568639e-05,
+      "loss": 0.6308,
+      "step": 6521
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7589685721143737e-05,
+      "loss": 0.5761,
+      "step": 6522
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.758897668661576e-05,
+      "loss": 0.6326,
+      "step": 6523
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.758826756211087e-05,
+      "loss": 0.6412,
+      "step": 6524
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7587558347637462e-05,
+      "loss": 0.6907,
+      "step": 6525
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7586849043203955e-05,
+      "loss": 0.629,
+      "step": 6526
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7586139648818755e-05,
+      "loss": 0.633,
+      "step": 6527
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7585430164490273e-05,
+      "loss": 0.6385,
+      "step": 6528
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.758472059022692e-05,
+      "loss": 0.6714,
+      "step": 6529
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7584010926037107e-05,
+      "loss": 0.6368,
+      "step": 6530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.758330117192925e-05,
+      "loss": 0.6554,
+      "step": 6531
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7582591327911772e-05,
+      "loss": 0.6563,
+      "step": 6532
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.758188139399308e-05,
+      "loss": 0.6228,
+      "step": 6533
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7581171370181588e-05,
+      "loss": 0.6011,
+      "step": 6534
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7580461256485722e-05,
+      "loss": 0.6267,
+      "step": 6535
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7579751052913897e-05,
+      "loss": 0.6103,
+      "step": 6536
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7579040759474537e-05,
+      "loss": 0.6608,
+      "step": 6537
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.757833037617606e-05,
+      "loss": 0.6,
+      "step": 6538
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.757761990302689e-05,
+      "loss": 0.6233,
+      "step": 6539
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.757690934003545e-05,
+      "loss": 0.68,
+      "step": 6540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7576198687210167e-05,
+      "loss": 0.6524,
+      "step": 6541
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7575487944559463e-05,
+      "loss": 0.6324,
+      "step": 6542
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7574777112091767e-05,
+      "loss": 0.6019,
+      "step": 6543
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7574066189815505e-05,
+      "loss": 0.5972,
+      "step": 6544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7573355177739107e-05,
+      "loss": 0.6349,
+      "step": 6545
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7572644075871003e-05,
+      "loss": 0.6115,
+      "step": 6546
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7571932884219626e-05,
+      "loss": 0.6363,
+      "step": 6547
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7571221602793405e-05,
+      "loss": 0.652,
+      "step": 6548
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.757051023160077e-05,
+      "loss": 0.5956,
+      "step": 6549
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7569798770650166e-05,
+      "loss": 0.6463,
+      "step": 6550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.756908721995002e-05,
+      "loss": 0.626,
+      "step": 6551
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.756837557950877e-05,
+      "loss": 0.6216,
+      "step": 6552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.756766384933485e-05,
+      "loss": 0.6405,
+      "step": 6553
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7566952029436707e-05,
+      "loss": 0.6408,
+      "step": 6554
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7566240119822772e-05,
+      "loss": 0.574,
+      "step": 6555
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7565528120501492e-05,
+      "loss": 0.5897,
+      "step": 6556
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7564816031481303e-05,
+      "loss": 0.612,
+      "step": 6557
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7564103852770652e-05,
+      "loss": 0.598,
+      "step": 6558
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.756339158437798e-05,
+      "loss": 0.646,
+      "step": 6559
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7562679226311733e-05,
+      "loss": 0.6237,
+      "step": 6560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.756196677858036e-05,
+      "loss": 0.6396,
+      "step": 6561
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.75612542411923e-05,
+      "loss": 0.6275,
+      "step": 6562
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.756054161415601e-05,
+      "loss": 0.6215,
+      "step": 6563
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.755982889747993e-05,
+      "loss": 0.6609,
+      "step": 6564
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7559116091172522e-05,
+      "loss": 0.6471,
+      "step": 6565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7558403195242225e-05,
+      "loss": 0.5921,
+      "step": 6566
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7557690209697496e-05,
+      "loss": 0.6094,
+      "step": 6567
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7556977134546795e-05,
+      "loss": 0.6785,
+      "step": 6568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7556263969798567e-05,
+      "loss": 0.6889,
+      "step": 6569
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7555550715461268e-05,
+      "loss": 0.6313,
+      "step": 6570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7554837371543363e-05,
+      "loss": 0.6561,
+      "step": 6571
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.75541239380533e-05,
+      "loss": 0.6363,
+      "step": 6572
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7553410414999545e-05,
+      "loss": 0.6707,
+      "step": 6573
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.755269680239055e-05,
+      "loss": 0.6483,
+      "step": 6574
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7551983100234783e-05,
+      "loss": 0.6195,
+      "step": 6575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7551269308540704e-05,
+      "loss": 0.6627,
+      "step": 6576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7550555427316773e-05,
+      "loss": 0.7067,
+      "step": 6577
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7549841456571458e-05,
+      "loss": 0.6018,
+      "step": 6578
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7549127396313223e-05,
+      "loss": 0.6673,
+      "step": 6579
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7548413246550527e-05,
+      "loss": 0.6228,
+      "step": 6580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7547699007291848e-05,
+      "loss": 0.6495,
+      "step": 6581
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.754698467854565e-05,
+      "loss": 0.6689,
+      "step": 6582
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7546270260320397e-05,
+      "loss": 0.6248,
+      "step": 6583
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7545555752624566e-05,
+      "loss": 0.7273,
+      "step": 6584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7544841155466627e-05,
+      "loss": 0.6258,
+      "step": 6585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7544126468855052e-05,
+      "loss": 0.6265,
+      "step": 6586
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7543411692798313e-05,
+      "loss": 0.6296,
+      "step": 6587
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.754269682730489e-05,
+      "loss": 0.6506,
+      "step": 6588
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.754198187238325e-05,
+      "loss": 0.5766,
+      "step": 6589
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7541266828041877e-05,
+      "loss": 0.5604,
+      "step": 6590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7540551694289246e-05,
+      "loss": 0.6479,
+      "step": 6591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7539836471133833e-05,
+      "loss": 0.5989,
+      "step": 6592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7539121158584124e-05,
+      "loss": 0.6472,
+      "step": 6593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7538405756648594e-05,
+      "loss": 0.648,
+      "step": 6594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.753769026533573e-05,
+      "loss": 0.6633,
+      "step": 6595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.753697468465401e-05,
+      "loss": 0.6617,
+      "step": 6596
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7536259014611926e-05,
+      "loss": 0.6036,
+      "step": 6597
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7535543255217952e-05,
+      "loss": 0.6557,
+      "step": 6598
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7534827406480582e-05,
+      "loss": 0.6843,
+      "step": 6599
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.75341114684083e-05,
+      "loss": 0.6294,
+      "step": 6600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7533395441009598e-05,
+      "loss": 0.6637,
+      "step": 6601
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7532679324292966e-05,
+      "loss": 0.6083,
+      "step": 6602
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7531963118266887e-05,
+      "loss": 0.6062,
+      "step": 6603
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.753124682293986e-05,
+      "loss": 0.6649,
+      "step": 6604
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7530530438320373e-05,
+      "loss": 0.6269,
+      "step": 6605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7529813964416923e-05,
+      "loss": 0.6623,
+      "step": 6606
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7529097401238003e-05,
+      "loss": 0.6698,
+      "step": 6607
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7528380748792106e-05,
+      "loss": 0.6123,
+      "step": 6608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7527664007087738e-05,
+      "loss": 0.5865,
+      "step": 6609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7526947176133388e-05,
+      "loss": 0.6372,
+      "step": 6610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7526230255937555e-05,
+      "loss": 0.5839,
+      "step": 6611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.752551324650874e-05,
+      "loss": 0.5986,
+      "step": 6612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.752479614785545e-05,
+      "loss": 0.6381,
+      "step": 6613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.752407895998618e-05,
+      "loss": 0.6409,
+      "step": 6614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.752336168290944e-05,
+      "loss": 0.6156,
+      "step": 6615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7522644316633727e-05,
+      "loss": 0.5945,
+      "step": 6616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.752192686116755e-05,
+      "loss": 0.5627,
+      "step": 6617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7521209316519413e-05,
+      "loss": 0.6541,
+      "step": 6618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7520491682697827e-05,
+      "loss": 0.6903,
+      "step": 6619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7519773959711298e-05,
+      "loss": 0.6634,
+      "step": 6620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7519056147568338e-05,
+      "loss": 0.6273,
+      "step": 6621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7518338246277454e-05,
+      "loss": 0.6537,
+      "step": 6622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.751762025584716e-05,
+      "loss": 0.707,
+      "step": 6623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7516902176285968e-05,
+      "loss": 0.5949,
+      "step": 6624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.751618400760239e-05,
+      "loss": 0.6586,
+      "step": 6625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7515465749804945e-05,
+      "loss": 0.64,
+      "step": 6626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7514747402902147e-05,
+      "loss": 0.6597,
+      "step": 6627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.751402896690251e-05,
+      "loss": 0.5763,
+      "step": 6628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7513310441814555e-05,
+      "loss": 0.6046,
+      "step": 6629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.75125918276468e-05,
+      "loss": 0.6151,
+      "step": 6630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7511873124407768e-05,
+      "loss": 0.6908,
+      "step": 6631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7511154332105972e-05,
+      "loss": 0.6292,
+      "step": 6632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7510435450749945e-05,
+      "loss": 0.6235,
+      "step": 6633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7509716480348203e-05,
+      "loss": 0.6401,
+      "step": 6634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7508997420909272e-05,
+      "loss": 0.5997,
+      "step": 6635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.750827827244168e-05,
+      "loss": 0.65,
+      "step": 6636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7507559034953946e-05,
+      "loss": 0.6,
+      "step": 6637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7506839708454606e-05,
+      "loss": 0.6987,
+      "step": 6638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7506120292952186e-05,
+      "loss": 0.6972,
+      "step": 6639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7505400788455215e-05,
+      "loss": 0.6238,
+      "step": 6640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.750468119497222e-05,
+      "loss": 0.605,
+      "step": 6641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.750396151251174e-05,
+      "loss": 0.597,
+      "step": 6642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7503241741082296e-05,
+      "loss": 0.5905,
+      "step": 6643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7502521880692435e-05,
+      "loss": 0.6056,
+      "step": 6644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7501801931350688e-05,
+      "loss": 0.6601,
+      "step": 6645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.750108189306559e-05,
+      "loss": 0.6572,
+      "step": 6646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7500361765845672e-05,
+      "loss": 0.6582,
+      "step": 6647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.749964154969948e-05,
+      "loss": 0.6464,
+      "step": 6648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7498921244635553e-05,
+      "loss": 0.7006,
+      "step": 6649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7498200850662424e-05,
+      "loss": 0.7104,
+      "step": 6650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.749748036778864e-05,
+      "loss": 0.7081,
+      "step": 6651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7496759796022742e-05,
+      "loss": 0.6217,
+      "step": 6652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7496039135373273e-05,
+      "loss": 0.6178,
+      "step": 6653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7495318385848777e-05,
+      "loss": 0.6351,
+      "step": 6654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7494597547457803e-05,
+      "loss": 0.6739,
+      "step": 6655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.749387662020889e-05,
+      "loss": 0.6695,
+      "step": 6656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7493155604110594e-05,
+      "loss": 0.5767,
+      "step": 6657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.749243449917146e-05,
+      "loss": 0.6349,
+      "step": 6658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7491713305400033e-05,
+      "loss": 0.5978,
+      "step": 6659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.749099202280487e-05,
+      "loss": 0.6239,
+      "step": 6660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7490270651394515e-05,
+      "loss": 0.6827,
+      "step": 6661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748954919117753e-05,
+      "loss": 0.6375,
+      "step": 6662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748882764216247e-05,
+      "loss": 0.6619,
+      "step": 6663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7488106004357882e-05,
+      "loss": 0.637,
+      "step": 6664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748738427777232e-05,
+      "loss": 0.677,
+      "step": 6665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748666246241435e-05,
+      "loss": 0.5787,
+      "step": 6666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748594055829253e-05,
+      "loss": 0.647,
+      "step": 6667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7485218565415406e-05,
+      "loss": 0.6353,
+      "step": 6668
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7484496483791552e-05,
+      "loss": 0.6571,
+      "step": 6669
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748377431342952e-05,
+      "loss": 0.6632,
+      "step": 6670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7483052054337883e-05,
+      "loss": 0.6068,
+      "step": 6671
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748232970652519e-05,
+      "loss": 0.6071,
+      "step": 6672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748160727000002e-05,
+      "loss": 0.6405,
+      "step": 6673
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.748088474477093e-05,
+      "loss": 0.6543,
+      "step": 6674
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7480162130846487e-05,
+      "loss": 0.6251,
+      "step": 6675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.747943942823526e-05,
+      "loss": 0.6537,
+      "step": 6676
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7478716636945815e-05,
+      "loss": 0.5749,
+      "step": 6677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.747799375698673e-05,
+      "loss": 0.6233,
+      "step": 6678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7477270788366562e-05,
+      "loss": 0.6271,
+      "step": 6679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7476547731093893e-05,
+      "loss": 0.6247,
+      "step": 6680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.747582458517729e-05,
+      "loss": 0.6103,
+      "step": 6681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7475101350625335e-05,
+      "loss": 0.7034,
+      "step": 6682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7474378027446595e-05,
+      "loss": 0.644,
+      "step": 6683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7473654615649647e-05,
+      "loss": 0.642,
+      "step": 6684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.747293111524307e-05,
+      "loss": 0.5946,
+      "step": 6685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7472207526235442e-05,
+      "loss": 0.6624,
+      "step": 6686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.747148384863534e-05,
+      "loss": 0.6322,
+      "step": 6687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.747076008245135e-05,
+      "loss": 0.7134,
+      "step": 6688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7470036227692044e-05,
+      "loss": 0.6537,
+      "step": 6689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7469312284366008e-05,
+      "loss": 0.6668,
+      "step": 6690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7468588252481827e-05,
+      "loss": 0.5963,
+      "step": 6691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7467864132048087e-05,
+      "loss": 0.6229,
+      "step": 6692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.746713992307337e-05,
+      "loss": 0.6782,
+      "step": 6693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7466415625566263e-05,
+      "loss": 0.6403,
+      "step": 6694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.746569123953535e-05,
+      "loss": 0.6173,
+      "step": 6695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7464966764989228e-05,
+      "loss": 0.6264,
+      "step": 6696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7464242201936483e-05,
+      "loss": 0.6028,
+      "step": 6697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7463517550385698e-05,
+      "loss": 0.6936,
+      "step": 6698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7462792810345477e-05,
+      "loss": 0.6039,
+      "step": 6699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7462067981824403e-05,
+      "loss": 0.6242,
+      "step": 6700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7461343064831074e-05,
+      "loss": 0.6196,
+      "step": 6701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7460618059374085e-05,
+      "loss": 0.6713,
+      "step": 6702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7459892965462035e-05,
+      "loss": 0.6245,
+      "step": 6703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7459167783103512e-05,
+      "loss": 0.6267,
+      "step": 6704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.745844251230712e-05,
+      "loss": 0.6442,
+      "step": 6705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7457717153081456e-05,
+      "loss": 0.6309,
+      "step": 6706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7456991705435125e-05,
+      "loss": 0.6471,
+      "step": 6707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7456266169376718e-05,
+      "loss": 0.6059,
+      "step": 6708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.745554054491485e-05,
+      "loss": 0.6136,
+      "step": 6709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7454814832058112e-05,
+      "loss": 0.6857,
+      "step": 6710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.745408903081512e-05,
+      "loss": 0.6067,
+      "step": 6711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.745336314119447e-05,
+      "loss": 0.6463,
+      "step": 6712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7452637163204767e-05,
+      "loss": 0.6169,
+      "step": 6713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.745191109685463e-05,
+      "loss": 0.638,
+      "step": 6714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7451184942152656e-05,
+      "loss": 0.646,
+      "step": 6715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.745045869910746e-05,
+      "loss": 0.6044,
+      "step": 6716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.744973236772765e-05,
+      "loss": 0.6452,
+      "step": 6717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.744900594802184e-05,
+      "loss": 0.6111,
+      "step": 6718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.744827943999864e-05,
+      "loss": 0.6416,
+      "step": 6719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.744755284366667e-05,
+      "loss": 0.6212,
+      "step": 6720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7446826159034537e-05,
+      "loss": 0.6357,
+      "step": 6721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7446099386110854e-05,
+      "loss": 0.5872,
+      "step": 6722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7445372524904248e-05,
+      "loss": 0.6537,
+      "step": 6723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7444645575423334e-05,
+      "loss": 0.6271,
+      "step": 6724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7443918537676728e-05,
+      "loss": 0.65,
+      "step": 6725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.744319141167305e-05,
+      "loss": 0.6103,
+      "step": 6726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7442464197420926e-05,
+      "loss": 0.6022,
+      "step": 6727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.744173689492897e-05,
+      "loss": 0.6396,
+      "step": 6728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.744100950420581e-05,
+      "loss": 0.6491,
+      "step": 6729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7440282025260068e-05,
+      "loss": 0.633,
+      "step": 6730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7439554458100374e-05,
+      "loss": 0.6026,
+      "step": 6731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7438826802735348e-05,
+      "loss": 0.607,
+      "step": 6732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7438099059173622e-05,
+      "loss": 0.6487,
+      "step": 6733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7437371227423824e-05,
+      "loss": 0.6469,
+      "step": 6734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7436643307494575e-05,
+      "loss": 0.658,
+      "step": 6735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.743591529939452e-05,
+      "loss": 0.6315,
+      "step": 6736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7435187203132277e-05,
+      "loss": 0.6734,
+      "step": 6737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.743445901871649e-05,
+      "loss": 0.6226,
+      "step": 6738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.743373074615578e-05,
+      "loss": 0.6198,
+      "step": 6739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.743300238545879e-05,
+      "loss": 0.6421,
+      "step": 6740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7432273936634158e-05,
+      "loss": 0.7019,
+      "step": 6741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.743154539969051e-05,
+      "loss": 0.6001,
+      "step": 6742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.74308167746365e-05,
+      "loss": 0.6469,
+      "step": 6743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7430088061480746e-05,
+      "loss": 0.6506,
+      "step": 6744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7429359260231907e-05,
+      "loss": 0.6372,
+      "step": 6745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7428630370898614e-05,
+      "loss": 0.6678,
+      "step": 6746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7427901393489507e-05,
+      "loss": 0.5693,
+      "step": 6747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7427172328013236e-05,
+      "loss": 0.6868,
+      "step": 6748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7426443174478444e-05,
+      "loss": 0.6466,
+      "step": 6749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7425713932893767e-05,
+      "loss": 0.6224,
+      "step": 6750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7424984603267866e-05,
+      "loss": 0.7055,
+      "step": 6751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7424255185609373e-05,
+      "loss": 0.616,
+      "step": 6752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7423525679926945e-05,
+      "loss": 0.6298,
+      "step": 6753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.742279608622923e-05,
+      "loss": 0.6185,
+      "step": 6754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7422066404524874e-05,
+      "loss": 0.5859,
+      "step": 6755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7421336634822538e-05,
+      "loss": 0.5945,
+      "step": 6756
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7420606777130866e-05,
+      "loss": 0.6867,
+      "step": 6757
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7419876831458512e-05,
+      "loss": 0.6982,
+      "step": 6758
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.741914679781413e-05,
+      "loss": 0.6303,
+      "step": 6759
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.741841667620638e-05,
+      "loss": 0.637,
+      "step": 6760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7417686466643918e-05,
+      "loss": 0.6113,
+      "step": 6761
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7416956169135396e-05,
+      "loss": 0.6315,
+      "step": 6762
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7416225783689478e-05,
+      "loss": 0.619,
+      "step": 6763
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.741549531031482e-05,
+      "loss": 0.633,
+      "step": 6764
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7414764749020088e-05,
+      "loss": 0.6452,
+      "step": 6765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7414034099813938e-05,
+      "loss": 0.6571,
+      "step": 6766
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7413303362705034e-05,
+      "loss": 0.6084,
+      "step": 6767
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.741257253770204e-05,
+      "loss": 0.6733,
+      "step": 6768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7411841624813627e-05,
+      "loss": 0.6564,
+      "step": 6769
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.741111062404845e-05,
+      "loss": 0.6414,
+      "step": 6770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7410379535415182e-05,
+      "loss": 0.5946,
+      "step": 6771
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.740964835892249e-05,
+      "loss": 0.6238,
+      "step": 6772
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7408917094579047e-05,
+      "loss": 0.6207,
+      "step": 6773
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.740818574239352e-05,
+      "loss": 0.6265,
+      "step": 6774
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7407454302374577e-05,
+      "loss": 0.6165,
+      "step": 6775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7406722774530892e-05,
+      "loss": 0.6588,
+      "step": 6776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.740599115887114e-05,
+      "loss": 0.6033,
+      "step": 6777
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7405259455403993e-05,
+      "loss": 0.5982,
+      "step": 6778
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.740452766413813e-05,
+      "loss": 0.6315,
+      "step": 6779
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7403795785082222e-05,
+      "loss": 0.625,
+      "step": 6780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7403063818244952e-05,
+      "loss": 0.5925,
+      "step": 6781
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7402331763634994e-05,
+      "loss": 0.6105,
+      "step": 6782
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7401599621261028e-05,
+      "loss": 0.5946,
+      "step": 6783
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.740086739113174e-05,
+      "loss": 0.6859,
+      "step": 6784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7400135073255802e-05,
+      "loss": 0.6187,
+      "step": 6785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7399402667641903e-05,
+      "loss": 0.6162,
+      "step": 6786
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.739867017429872e-05,
+      "loss": 0.6249,
+      "step": 6787
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.739793759323495e-05,
+      "loss": 0.609,
+      "step": 6788
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.739720492445927e-05,
+      "loss": 0.623,
+      "step": 6789
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7396472167980368e-05,
+      "loss": 0.6454,
+      "step": 6790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.739573932380693e-05,
+      "loss": 0.6191,
+      "step": 6791
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.739500639194765e-05,
+      "loss": 0.6999,
+      "step": 6792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.739427337241121e-05,
+      "loss": 0.6477,
+      "step": 6793
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.739354026520631e-05,
+      "loss": 0.6512,
+      "step": 6794
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7392807070341634e-05,
+      "loss": 0.6271,
+      "step": 6795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7392073787825875e-05,
+      "loss": 0.6856,
+      "step": 6796
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.739134041766774e-05,
+      "loss": 0.6433,
+      "step": 6797
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7390606959875905e-05,
+      "loss": 0.6207,
+      "step": 6798
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.738987341445908e-05,
+      "loss": 0.6029,
+      "step": 6799
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7389139781425955e-05,
+      "loss": 0.6481,
+      "step": 6800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7388406060785233e-05,
+      "loss": 0.6695,
+      "step": 6801
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7387672252545607e-05,
+      "loss": 0.6001,
+      "step": 6802
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7386938356715784e-05,
+      "loss": 0.5765,
+      "step": 6803
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7386204373304463e-05,
+      "loss": 0.6723,
+      "step": 6804
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7385470302320343e-05,
+      "loss": 0.5645,
+      "step": 6805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.738473614377213e-05,
+      "loss": 0.6174,
+      "step": 6806
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.738400189766853e-05,
+      "loss": 0.6581,
+      "step": 6807
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7383267564018247e-05,
+      "loss": 0.5981,
+      "step": 6808
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7382533142829987e-05,
+      "loss": 0.5718,
+      "step": 6809
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7381798634112457e-05,
+      "loss": 0.6519,
+      "step": 6810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7381064037874367e-05,
+      "loss": 0.644,
+      "step": 6811
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7380329354124424e-05,
+      "loss": 0.6225,
+      "step": 6812
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7379594582871342e-05,
+      "loss": 0.6521,
+      "step": 6813
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.737885972412383e-05,
+      "loss": 0.6754,
+      "step": 6814
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.73781247778906e-05,
+      "loss": 0.6349,
+      "step": 6815
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.737738974418037e-05,
+      "loss": 0.5538,
+      "step": 6816
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.737665462300185e-05,
+      "loss": 0.6383,
+      "step": 6817
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7375919414363764e-05,
+      "loss": 0.7244,
+      "step": 6818
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7375184118274818e-05,
+      "loss": 0.6671,
+      "step": 6819
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7374448734743733e-05,
+      "loss": 0.5907,
+      "step": 6820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.737371326377923e-05,
+      "loss": 0.6405,
+      "step": 6821
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.737297770539003e-05,
+      "loss": 0.6657,
+      "step": 6822
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7372242059584856e-05,
+      "loss": 0.6453,
+      "step": 6823
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.737150632637242e-05,
+      "loss": 0.6544,
+      "step": 6824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7370770505761453e-05,
+      "loss": 0.5818,
+      "step": 6825
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.737003459776068e-05,
+      "loss": 0.5649,
+      "step": 6826
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7369298602378825e-05,
+      "loss": 0.6453,
+      "step": 6827
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7368562519624607e-05,
+      "loss": 0.6156,
+      "step": 6828
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.736782634950676e-05,
+      "loss": 0.6042,
+      "step": 6829
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7367090092034017e-05,
+      "loss": 0.5912,
+      "step": 6830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7366353747215097e-05,
+      "loss": 0.6403,
+      "step": 6831
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7365617315058735e-05,
+      "loss": 0.6638,
+      "step": 6832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7364880795573665e-05,
+      "loss": 0.6613,
+      "step": 6833
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7364144188768613e-05,
+      "loss": 0.6351,
+      "step": 6834
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.736340749465232e-05,
+      "loss": 0.6348,
+      "step": 6835
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7362670713233512e-05,
+      "loss": 0.5969,
+      "step": 6836
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.736193384452093e-05,
+      "loss": 0.6439,
+      "step": 6837
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7361196888523313e-05,
+      "loss": 0.6714,
+      "step": 6838
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7360459845249393e-05,
+      "loss": 0.627,
+      "step": 6839
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7359722714707908e-05,
+      "loss": 0.6611,
+      "step": 6840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7358985496907603e-05,
+      "loss": 0.6019,
+      "step": 6841
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7358248191857212e-05,
+      "loss": 0.6155,
+      "step": 6842
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7357510799565484e-05,
+      "loss": 0.6334,
+      "step": 6843
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.735677332004116e-05,
+      "loss": 0.6433,
+      "step": 6844
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7356035753292973e-05,
+      "loss": 0.6148,
+      "step": 6845
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7355298099329684e-05,
+      "loss": 0.5963,
+      "step": 6846
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7354560358160034e-05,
+      "loss": 0.6312,
+      "step": 6847
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7353822529792762e-05,
+      "loss": 0.7055,
+      "step": 6848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7353084614236622e-05,
+      "loss": 0.6214,
+      "step": 6849
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7352346611500365e-05,
+      "loss": 0.6055,
+      "step": 6850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7351608521592735e-05,
+      "loss": 0.6584,
+      "step": 6851
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7350870344522487e-05,
+      "loss": 0.6257,
+      "step": 6852
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7350132080298374e-05,
+      "loss": 0.6403,
+      "step": 6853
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7349393728929146e-05,
+      "loss": 0.6088,
+      "step": 6854
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7348655290423558e-05,
+      "loss": 0.6572,
+      "step": 6855
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7347916764790365e-05,
+      "loss": 0.5875,
+      "step": 6856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7347178152038323e-05,
+      "loss": 0.615,
+      "step": 6857
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7346439452176193e-05,
+      "loss": 0.6306,
+      "step": 6858
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7345700665212726e-05,
+      "loss": 0.6788,
+      "step": 6859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.734496179115669e-05,
+      "loss": 0.6644,
+      "step": 6860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.734422283001684e-05,
+      "loss": 0.6157,
+      "step": 6861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.734348378180193e-05,
+      "loss": 0.568,
+      "step": 6862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.734274464652074e-05,
+      "loss": 0.6315,
+      "step": 6863
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.734200542418202e-05,
+      "loss": 0.6535,
+      "step": 6864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7341266114794536e-05,
+      "loss": 0.5823,
+      "step": 6865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7340526718367056e-05,
+      "loss": 0.6544,
+      "step": 6866
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.733978723490835e-05,
+      "loss": 0.6396,
+      "step": 6867
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7339047664427175e-05,
+      "loss": 0.6074,
+      "step": 6868
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7338308006932312e-05,
+      "loss": 0.6772,
+      "step": 6869
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.733756826243252e-05,
+      "loss": 0.5861,
+      "step": 6870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7336828430936577e-05,
+      "loss": 0.6661,
+      "step": 6871
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.733608851245325e-05,
+      "loss": 0.5953,
+      "step": 6872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7335348506991314e-05,
+      "loss": 0.6436,
+      "step": 6873
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7334608414559544e-05,
+      "loss": 0.6079,
+      "step": 6874
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7333868235166713e-05,
+      "loss": 0.6513,
+      "step": 6875
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7333127968821593e-05,
+      "loss": 0.6282,
+      "step": 6876
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7332387615532966e-05,
+      "loss": 0.6171,
+      "step": 6877
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.733164717530961e-05,
+      "loss": 0.6322,
+      "step": 6878
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.73309066481603e-05,
+      "loss": 0.7119,
+      "step": 6879
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7330166034093817e-05,
+      "loss": 0.6067,
+      "step": 6880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7329425333118948e-05,
+      "loss": 0.6384,
+      "step": 6881
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7328684545244468e-05,
+      "loss": 0.6739,
+      "step": 6882
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.732794367047916e-05,
+      "loss": 0.6481,
+      "step": 6883
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7327202708831812e-05,
+      "loss": 0.6381,
+      "step": 6884
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7326461660311206e-05,
+      "loss": 0.6723,
+      "step": 6885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.732572052492613e-05,
+      "loss": 0.651,
+      "step": 6886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7324979302685364e-05,
+      "loss": 0.6283,
+      "step": 6887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7324237993597708e-05,
+      "loss": 0.65,
+      "step": 6888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7323496597671943e-05,
+      "loss": 0.6462,
+      "step": 6889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7322755114916865e-05,
+      "loss": 0.6717,
+      "step": 6890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7322013545341262e-05,
+      "loss": 0.7049,
+      "step": 6891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7321271888953923e-05,
+      "loss": 0.6053,
+      "step": 6892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7320530145763643e-05,
+      "loss": 0.579,
+      "step": 6893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.731978831577922e-05,
+      "loss": 0.6561,
+      "step": 6894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7319046399009446e-05,
+      "loss": 0.6072,
+      "step": 6895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.731830439546312e-05,
+      "loss": 0.6094,
+      "step": 6896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7317562305149035e-05,
+      "loss": 0.5793,
+      "step": 6897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7316820128075998e-05,
+      "loss": 0.5948,
+      "step": 6898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7316077864252797e-05,
+      "loss": 0.6639,
+      "step": 6899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.731533551368824e-05,
+      "loss": 0.6487,
+      "step": 6900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7314593076391123e-05,
+      "loss": 0.6349,
+      "step": 6901
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7313850552370256e-05,
+      "loss": 0.6179,
+      "step": 6902
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7313107941634436e-05,
+      "loss": 0.6028,
+      "step": 6903
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.731236524419247e-05,
+      "loss": 0.6647,
+      "step": 6904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.731162246005317e-05,
+      "loss": 0.6205,
+      "step": 6905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.731087958922533e-05,
+      "loss": 0.6908,
+      "step": 6906
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7310136631717766e-05,
+      "loss": 0.6657,
+      "step": 6907
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7309393587539286e-05,
+      "loss": 0.6275,
+      "step": 6908
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.73086504566987e-05,
+      "loss": 0.662,
+      "step": 6909
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7307907239204815e-05,
+      "loss": 0.6319,
+      "step": 6910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7307163935066445e-05,
+      "loss": 0.6216,
+      "step": 6911
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.73064205442924e-05,
+      "loss": 0.6182,
+      "step": 6912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7305677066891503e-05,
+      "loss": 0.6192,
+      "step": 6913
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.730493350287256e-05,
+      "loss": 0.6308,
+      "step": 6914
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7304189852244388e-05,
+      "loss": 0.635,
+      "step": 6915
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7303446115015806e-05,
+      "loss": 0.6627,
+      "step": 6916
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.730270229119563e-05,
+      "loss": 0.6944,
+      "step": 6917
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7301958380792684e-05,
+      "loss": 0.6359,
+      "step": 6918
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7301214383815783e-05,
+      "loss": 0.6364,
+      "step": 6919
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.730047030027375e-05,
+      "loss": 0.6147,
+      "step": 6920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7299726130175406e-05,
+      "loss": 0.6108,
+      "step": 6921
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.729898187352957e-05,
+      "loss": 0.6873,
+      "step": 6922
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7298237530345076e-05,
+      "loss": 0.6483,
+      "step": 6923
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7297493100630746e-05,
+      "loss": 0.6569,
+      "step": 6924
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7296748584395397e-05,
+      "loss": 0.6068,
+      "step": 6925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7296003981647865e-05,
+      "loss": 0.6434,
+      "step": 6926
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.729525929239698e-05,
+      "loss": 0.6264,
+      "step": 6927
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7294514516651565e-05,
+      "loss": 0.6193,
+      "step": 6928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7293769654420452e-05,
+      "loss": 0.6554,
+      "step": 6929
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7293024705712477e-05,
+      "loss": 0.6278,
+      "step": 6930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.729227967053646e-05,
+      "loss": 0.5601,
+      "step": 6931
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.729153454890125e-05,
+      "loss": 0.6853,
+      "step": 6932
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7290789340815674e-05,
+      "loss": 0.651,
+      "step": 6933
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7290044046288564e-05,
+      "loss": 0.6347,
+      "step": 6934
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7289298665328763e-05,
+      "loss": 0.6564,
+      "step": 6935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7288553197945105e-05,
+      "loss": 0.6649,
+      "step": 6936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.728780764414643e-05,
+      "loss": 0.6278,
+      "step": 6937
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.728706200394157e-05,
+      "loss": 0.5986,
+      "step": 6938
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7286316277339378e-05,
+      "loss": 0.6638,
+      "step": 6939
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7285570464348687e-05,
+      "loss": 0.6764,
+      "step": 6940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.728482456497834e-05,
+      "loss": 0.5743,
+      "step": 6941
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.728407857923719e-05,
+      "loss": 0.5939,
+      "step": 6942
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7283332507134064e-05,
+      "loss": 0.6059,
+      "step": 6943
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7282586348677824e-05,
+      "loss": 0.6017,
+      "step": 6944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7281840103877306e-05,
+      "loss": 0.6258,
+      "step": 6945
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7281093772741365e-05,
+      "loss": 0.593,
+      "step": 6946
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7280347355278845e-05,
+      "loss": 0.6945,
+      "step": 6947
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.72796008514986e-05,
+      "loss": 0.6077,
+      "step": 6948
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7278854261409478e-05,
+      "loss": 0.615,
+      "step": 6949
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7278107585020326e-05,
+      "loss": 0.6148,
+      "step": 6950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7277360822340008e-05,
+      "loss": 0.6681,
+      "step": 6951
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7276613973377366e-05,
+      "loss": 0.6641,
+      "step": 6952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7275867038141265e-05,
+      "loss": 0.631,
+      "step": 6953
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7275120016640553e-05,
+      "loss": 0.6825,
+      "step": 6954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.727437290888409e-05,
+      "loss": 0.6334,
+      "step": 6955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.727362571488074e-05,
+      "loss": 0.6486,
+      "step": 6956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.727287843463935e-05,
+      "loss": 0.5864,
+      "step": 6957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7272131068168786e-05,
+      "loss": 0.639,
+      "step": 6958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7271383615477913e-05,
+      "loss": 0.6301,
+      "step": 6959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7270636076575585e-05,
+      "loss": 0.5978,
+      "step": 6960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.726988845147067e-05,
+      "loss": 0.6067,
+      "step": 6961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.726914074017203e-05,
+      "loss": 0.6769,
+      "step": 6962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7268392942688534e-05,
+      "loss": 0.6115,
+      "step": 6963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7267645059029042e-05,
+      "loss": 0.6074,
+      "step": 6964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7266897089202426e-05,
+      "loss": 0.6171,
+      "step": 6965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7266149033217554e-05,
+      "loss": 0.6428,
+      "step": 6966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7265400891083288e-05,
+      "loss": 0.6598,
+      "step": 6967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.726465266280851e-05,
+      "loss": 0.6384,
+      "step": 6968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.726390434840208e-05,
+      "loss": 0.7376,
+      "step": 6969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7263155947872875e-05,
+      "loss": 0.6122,
+      "step": 6970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.726240746122977e-05,
+      "loss": 0.6478,
+      "step": 6971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.726165888848164e-05,
+      "loss": 0.614,
+      "step": 6972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7260910229637354e-05,
+      "loss": 0.6423,
+      "step": 6973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7260161484705796e-05,
+      "loss": 0.6707,
+      "step": 6974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7259412653695836e-05,
+      "loss": 0.6231,
+      "step": 6975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7258663736616358e-05,
+      "loss": 0.5897,
+      "step": 6976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.725791473347624e-05,
+      "loss": 0.6641,
+      "step": 6977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7257165644284358e-05,
+      "loss": 0.643,
+      "step": 6978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7256416469049597e-05,
+      "loss": 0.6509,
+      "step": 6979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7255667207780842e-05,
+      "loss": 0.6698,
+      "step": 6980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7254917860486974e-05,
+      "loss": 0.6137,
+      "step": 6981
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7254168427176873e-05,
+      "loss": 0.6175,
+      "step": 6982
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7253418907859433e-05,
+      "loss": 0.5888,
+      "step": 6983
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7252669302543533e-05,
+      "loss": 0.5839,
+      "step": 6984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7251919611238065e-05,
+      "loss": 0.6059,
+      "step": 6985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.725116983395192e-05,
+      "loss": 0.6618,
+      "step": 6986
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.725041997069398e-05,
+      "loss": 0.6582,
+      "step": 6987
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.724967002147314e-05,
+      "loss": 0.5658,
+      "step": 6988
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.724891998629829e-05,
+      "loss": 0.671,
+      "step": 6989
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7248169865178325e-05,
+      "loss": 0.6272,
+      "step": 6990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.724741965812214e-05,
+      "loss": 0.6298,
+      "step": 6991
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.724666936513862e-05,
+      "loss": 0.5934,
+      "step": 6992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7245918986236672e-05,
+      "loss": 0.6672,
+      "step": 6993
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.724516852142519e-05,
+      "loss": 0.6537,
+      "step": 6994
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7244417970713066e-05,
+      "loss": 0.6168,
+      "step": 6995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7243667334109204e-05,
+      "loss": 0.6464,
+      "step": 6996
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7242916611622506e-05,
+      "loss": 0.6394,
+      "step": 6997
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7242165803261866e-05,
+      "loss": 0.661,
+      "step": 6998
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7241414909036188e-05,
+      "loss": 0.6316,
+      "step": 6999
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7240663928954377e-05,
+      "loss": 0.6344,
+      "step": 7000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7239912863025336e-05,
+      "loss": 0.5946,
+      "step": 7001
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.723916171125797e-05,
+      "loss": 0.6378,
+      "step": 7002
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.723841047366118e-05,
+      "loss": 0.6469,
+      "step": 7003
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7237659150243882e-05,
+      "loss": 0.6304,
+      "step": 7004
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.723690774101498e-05,
+      "loss": 0.6455,
+      "step": 7005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7236156245983378e-05,
+      "loss": 0.6328,
+      "step": 7006
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7235404665157994e-05,
+      "loss": 0.5842,
+      "step": 7007
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7234652998547732e-05,
+      "loss": 0.6784,
+      "step": 7008
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7233901246161507e-05,
+      "loss": 0.6678,
+      "step": 7009
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7233149408008232e-05,
+      "loss": 0.657,
+      "step": 7010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7232397484096822e-05,
+      "loss": 0.6292,
+      "step": 7011
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7231645474436187e-05,
+      "loss": 0.6667,
+      "step": 7012
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7230893379035252e-05,
+      "loss": 0.6068,
+      "step": 7013
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7230141197902924e-05,
+      "loss": 0.6535,
+      "step": 7014
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.722938893104813e-05,
+      "loss": 0.7011,
+      "step": 7015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7228636578479784e-05,
+      "loss": 0.6284,
+      "step": 7016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7227884140206807e-05,
+      "loss": 0.6575,
+      "step": 7017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7227131616238117e-05,
+      "loss": 0.5928,
+      "step": 7018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7226379006582643e-05,
+      "loss": 0.64,
+      "step": 7019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7225626311249305e-05,
+      "loss": 0.6157,
+      "step": 7020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.722487353024702e-05,
+      "loss": 0.6619,
+      "step": 7021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7224120663584728e-05,
+      "loss": 0.5835,
+      "step": 7022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7223367711271345e-05,
+      "loss": 0.6873,
+      "step": 7023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.72226146733158e-05,
+      "loss": 0.6788,
+      "step": 7024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7221861549727017e-05,
+      "loss": 0.6171,
+      "step": 7025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7221108340513935e-05,
+      "loss": 0.5934,
+      "step": 7026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7220355045685476e-05,
+      "loss": 0.6629,
+      "step": 7027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7219601665250576e-05,
+      "loss": 0.6945,
+      "step": 7028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7218848199218165e-05,
+      "loss": 0.6407,
+      "step": 7029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7218094647597174e-05,
+      "loss": 0.6337,
+      "step": 7030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.721734101039654e-05,
+      "loss": 0.6256,
+      "step": 7031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7216587287625204e-05,
+      "loss": 0.6273,
+      "step": 7032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7215833479292095e-05,
+      "loss": 0.6622,
+      "step": 7033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7215079585406147e-05,
+      "loss": 0.6437,
+      "step": 7034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.721432560597631e-05,
+      "loss": 0.6611,
+      "step": 7035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7213571541011512e-05,
+      "loss": 0.5795,
+      "step": 7036
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7212817390520697e-05,
+      "loss": 0.6562,
+      "step": 7037
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7212063154512812e-05,
+      "loss": 0.6013,
+      "step": 7038
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7211308832996792e-05,
+      "loss": 0.5932,
+      "step": 7039
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7210554425981585e-05,
+      "loss": 0.5979,
+      "step": 7040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7209799933476135e-05,
+      "loss": 0.6197,
+      "step": 7041
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.720904535548938e-05,
+      "loss": 0.6662,
+      "step": 7042
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.720829069203028e-05,
+      "loss": 0.6921,
+      "step": 7043
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7207535943107774e-05,
+      "loss": 0.6181,
+      "step": 7044
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.720678110873081e-05,
+      "loss": 0.6114,
+      "step": 7045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.720602618890834e-05,
+      "loss": 0.6669,
+      "step": 7046
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7205271183649313e-05,
+      "loss": 0.6351,
+      "step": 7047
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7204516092962684e-05,
+      "loss": 0.6177,
+      "step": 7048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.72037609168574e-05,
+      "loss": 0.6093,
+      "step": 7049
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.720300565534242e-05,
+      "loss": 0.5606,
+      "step": 7050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7202250308426695e-05,
+      "loss": 0.6213,
+      "step": 7051
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.720149487611918e-05,
+      "loss": 0.6739,
+      "step": 7052
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7200739358428835e-05,
+      "loss": 0.6829,
+      "step": 7053
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7199983755364614e-05,
+      "loss": 0.6354,
+      "step": 7054
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.719922806693548e-05,
+      "loss": 0.5687,
+      "step": 7055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7198472293150393e-05,
+      "loss": 0.6239,
+      "step": 7056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7197716434018307e-05,
+      "loss": 0.6505,
+      "step": 7057
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7196960489548186e-05,
+      "loss": 0.6178,
+      "step": 7058
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7196204459748996e-05,
+      "loss": 0.7034,
+      "step": 7059
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.71954483446297e-05,
+      "loss": 0.6357,
+      "step": 7060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.719469214419926e-05,
+      "loss": 0.6111,
+      "step": 7061
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7193935858466642e-05,
+      "loss": 0.6344,
+      "step": 7062
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.719317948744082e-05,
+      "loss": 0.6561,
+      "step": 7063
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.719242303113075e-05,
+      "loss": 0.6659,
+      "step": 7064
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.719166648954541e-05,
+      "loss": 0.6382,
+      "step": 7065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.719090986269376e-05,
+      "loss": 0.6096,
+      "step": 7066
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7190153150584784e-05,
+      "loss": 0.6564,
+      "step": 7067
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7189396353227444e-05,
+      "loss": 0.6307,
+      "step": 7068
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7188639470630716e-05,
+      "loss": 0.5997,
+      "step": 7069
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7187882502803575e-05,
+      "loss": 0.6543,
+      "step": 7070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.718712544975499e-05,
+      "loss": 0.6414,
+      "step": 7071
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7186368311493945e-05,
+      "loss": 0.6906,
+      "step": 7072
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.718561108802941e-05,
+      "loss": 0.6863,
+      "step": 7073
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7184853779370367e-05,
+      "loss": 0.6222,
+      "step": 7074
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7184096385525792e-05,
+      "loss": 0.5705,
+      "step": 7075
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.718333890650467e-05,
+      "loss": 0.6609,
+      "step": 7076
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7182581342315977e-05,
+      "loss": 0.6778,
+      "step": 7077
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7181823692968694e-05,
+      "loss": 0.6335,
+      "step": 7078
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.718106595847181e-05,
+      "loss": 0.6375,
+      "step": 7079
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.71803081388343e-05,
+      "loss": 0.6369,
+      "step": 7080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.717955023406516e-05,
+      "loss": 0.6502,
+      "step": 7081
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7178792244173365e-05,
+      "loss": 0.6374,
+      "step": 7082
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7178034169167906e-05,
+      "loss": 0.6791,
+      "step": 7083
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7177276009057774e-05,
+      "loss": 0.6623,
+      "step": 7084
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7176517763851954e-05,
+      "loss": 0.651,
+      "step": 7085
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.717575943355944e-05,
+      "loss": 0.6669,
+      "step": 7086
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7175001018189223e-05,
+      "loss": 0.6912,
+      "step": 7087
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7174242517750285e-05,
+      "loss": 0.6329,
+      "step": 7088
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7173483932251632e-05,
+      "loss": 0.5797,
+      "step": 7089
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.717272526170225e-05,
+      "loss": 0.5742,
+      "step": 7090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7171966506111135e-05,
+      "loss": 0.6223,
+      "step": 7091
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.717120766548729e-05,
+      "loss": 0.5795,
+      "step": 7092
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7170448739839703e-05,
+      "loss": 0.6894,
+      "step": 7093
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7169689729177375e-05,
+      "loss": 0.6554,
+      "step": 7094
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7168930633509307e-05,
+      "loss": 0.6732,
+      "step": 7095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7168171452844498e-05,
+      "loss": 0.6471,
+      "step": 7096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7167412187191944e-05,
+      "loss": 0.6029,
+      "step": 7097
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7166652836560657e-05,
+      "loss": 0.5949,
+      "step": 7098
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7165893400959636e-05,
+      "loss": 0.6052,
+      "step": 7099
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7165133880397876e-05,
+      "loss": 0.6365,
+      "step": 7100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7164374274884397e-05,
+      "loss": 0.6492,
+      "step": 7101
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7163614584428194e-05,
+      "loss": 0.6299,
+      "step": 7102
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.716285480903828e-05,
+      "loss": 0.5664,
+      "step": 7103
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7162094948723664e-05,
+      "loss": 0.6242,
+      "step": 7104
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7161335003493347e-05,
+      "loss": 0.6231,
+      "step": 7105
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7160574973356346e-05,
+      "loss": 0.6273,
+      "step": 7106
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7159814858321673e-05,
+      "loss": 0.6583,
+      "step": 7107
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7159054658398334e-05,
+      "loss": 0.6034,
+      "step": 7108
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.715829437359535e-05,
+      "loss": 0.6757,
+      "step": 7109
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7157534003921728e-05,
+      "loss": 0.6061,
+      "step": 7110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.715677354938649e-05,
+      "loss": 0.6398,
+      "step": 7111
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7156013009998643e-05,
+      "loss": 0.6032,
+      "step": 7112
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.715525238576721e-05,
+      "loss": 0.5969,
+      "step": 7113
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7154491676701207e-05,
+      "loss": 0.6044,
+      "step": 7114
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7153730882809657e-05,
+      "loss": 0.5904,
+      "step": 7115
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7152970004101582e-05,
+      "loss": 0.6128,
+      "step": 7116
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7152209040585993e-05,
+      "loss": 0.6268,
+      "step": 7117
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.715144799227192e-05,
+      "loss": 0.6444,
+      "step": 7118
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7150686859168385e-05,
+      "loss": 0.6038,
+      "step": 7119
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7149925641284414e-05,
+      "loss": 0.6378,
+      "step": 7120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7149164338629027e-05,
+      "loss": 0.5909,
+      "step": 7121
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7148402951211253e-05,
+      "loss": 0.6211,
+      "step": 7122
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.714764147904012e-05,
+      "loss": 0.6685,
+      "step": 7123
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7146879922124654e-05,
+      "loss": 0.5957,
+      "step": 7124
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7146118280473887e-05,
+      "loss": 0.6094,
+      "step": 7125
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7145356554096848e-05,
+      "loss": 0.6411,
+      "step": 7126
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.714459474300257e-05,
+      "loss": 0.5905,
+      "step": 7127
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7143832847200078e-05,
+      "loss": 0.6147,
+      "step": 7128
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7143070866698416e-05,
+      "loss": 0.6282,
+      "step": 7129
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7142308801506612e-05,
+      "loss": 0.6253,
+      "step": 7130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.71415466516337e-05,
+      "loss": 0.6601,
+      "step": 7131
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7140784417088717e-05,
+      "loss": 0.6211,
+      "step": 7132
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7140022097880704e-05,
+      "loss": 0.7028,
+      "step": 7133
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7139259694018697e-05,
+      "loss": 0.68,
+      "step": 7134
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7138497205511732e-05,
+      "loss": 0.6434,
+      "step": 7135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7137734632368855e-05,
+      "loss": 0.632,
+      "step": 7136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7136971974599107e-05,
+      "loss": 0.6176,
+      "step": 7137
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7136209232211526e-05,
+      "loss": 0.6703,
+      "step": 7138
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7135446405215156e-05,
+      "loss": 0.6412,
+      "step": 7139
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7134683493619042e-05,
+      "loss": 0.6427,
+      "step": 7140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.713392049743223e-05,
+      "loss": 0.6519,
+      "step": 7141
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.713315741666377e-05,
+      "loss": 0.6481,
+      "step": 7142
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.71323942513227e-05,
+      "loss": 0.5857,
+      "step": 7143
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7131631001418077e-05,
+      "loss": 0.6075,
+      "step": 7144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7130867666958945e-05,
+      "loss": 0.6417,
+      "step": 7145
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.713010424795436e-05,
+      "loss": 0.6392,
+      "step": 7146
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7129340744413366e-05,
+      "loss": 0.6638,
+      "step": 7147
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7128577156345017e-05,
+      "loss": 0.6127,
+      "step": 7148
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7127813483758373e-05,
+      "loss": 0.6706,
+      "step": 7149
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7127049726662482e-05,
+      "loss": 0.6709,
+      "step": 7150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7126285885066398e-05,
+      "loss": 0.6109,
+      "step": 7151
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.712552195897918e-05,
+      "loss": 0.6753,
+      "step": 7152
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.712475794840989e-05,
+      "loss": 0.6299,
+      "step": 7153
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7123993853367576e-05,
+      "loss": 0.6945,
+      "step": 7154
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7123229673861308e-05,
+      "loss": 0.6333,
+      "step": 7155
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7122465409900136e-05,
+      "loss": 0.6473,
+      "step": 7156
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7121701061493134e-05,
+      "loss": 0.6674,
+      "step": 7157
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.712093662864935e-05,
+      "loss": 0.6157,
+      "step": 7158
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.712017211137786e-05,
+      "loss": 0.7164,
+      "step": 7159
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7119407509687717e-05,
+      "loss": 0.6115,
+      "step": 7160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7118642823587994e-05,
+      "loss": 0.6253,
+      "step": 7161
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7117878053087757e-05,
+      "loss": 0.6583,
+      "step": 7162
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.711711319819607e-05,
+      "loss": 0.6231,
+      "step": 7163
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7116348258922002e-05,
+      "loss": 0.5644,
+      "step": 7164
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7115583235274625e-05,
+      "loss": 0.6451,
+      "step": 7165
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7114818127263007e-05,
+      "loss": 0.6186,
+      "step": 7166
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.711405293489622e-05,
+      "loss": 0.6263,
+      "step": 7167
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7113287658183332e-05,
+      "loss": 0.6513,
+      "step": 7168
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7112522297133425e-05,
+      "loss": 0.6422,
+      "step": 7169
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7111756851755567e-05,
+      "loss": 0.6996,
+      "step": 7170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7110991322058835e-05,
+      "loss": 0.6154,
+      "step": 7171
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7110225708052306e-05,
+      "loss": 0.6399,
+      "step": 7172
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7109460009745056e-05,
+      "loss": 0.595,
+      "step": 7173
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7108694227146167e-05,
+      "loss": 0.6339,
+      "step": 7174
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7107928360264714e-05,
+      "loss": 0.6493,
+      "step": 7175
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7107162409109775e-05,
+      "loss": 0.5986,
+      "step": 7176
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.710639637369044e-05,
+      "loss": 0.6685,
+      "step": 7177
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7105630254015785e-05,
+      "loss": 0.6891,
+      "step": 7178
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7104864050094895e-05,
+      "loss": 0.6484,
+      "step": 7179
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7104097761936852e-05,
+      "loss": 0.5876,
+      "step": 7180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7103331389550745e-05,
+      "loss": 0.6628,
+      "step": 7181
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.710256493294566e-05,
+      "loss": 0.6561,
+      "step": 7182
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7101798392130678e-05,
+      "loss": 0.6768,
+      "step": 7183
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7101031767114897e-05,
+      "loss": 0.5898,
+      "step": 7184
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7100265057907398e-05,
+      "loss": 0.632,
+      "step": 7185
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7099498264517275e-05,
+      "loss": 0.6579,
+      "step": 7186
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7098731386953623e-05,
+      "loss": 0.6018,
+      "step": 7187
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7097964425225528e-05,
+      "loss": 0.5938,
+      "step": 7188
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7097197379342084e-05,
+      "loss": 0.6963,
+      "step": 7189
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.709643024931239e-05,
+      "loss": 0.6188,
+      "step": 7190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7095663035145537e-05,
+      "loss": 0.6098,
+      "step": 7191
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7094895736850625e-05,
+      "loss": 0.6948,
+      "step": 7192
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7094128354436745e-05,
+      "loss": 0.669,
+      "step": 7193
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7093360887913002e-05,
+      "loss": 0.6653,
+      "step": 7194
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7092593337288493e-05,
+      "loss": 0.6697,
+      "step": 7195
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7091825702572315e-05,
+      "loss": 0.5986,
+      "step": 7196
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7091057983773574e-05,
+      "loss": 0.6774,
+      "step": 7197
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7090290180901373e-05,
+      "loss": 0.6736,
+      "step": 7198
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7089522293964807e-05,
+      "loss": 0.6267,
+      "step": 7199
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7088754322972987e-05,
+      "loss": 0.6084,
+      "step": 7200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.708798626793502e-05,
+      "loss": 0.5697,
+      "step": 7201
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.708721812886001e-05,
+      "loss": 0.603,
+      "step": 7202
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.708644990575706e-05,
+      "loss": 0.7185,
+      "step": 7203
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7085681598635283e-05,
+      "loss": 0.6192,
+      "step": 7204
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.708491320750379e-05,
+      "loss": 0.613,
+      "step": 7205
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7084144732371685e-05,
+      "loss": 0.6589,
+      "step": 7206
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7083376173248085e-05,
+      "loss": 0.6781,
+      "step": 7207
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.70826075301421e-05,
+      "loss": 0.5949,
+      "step": 7208
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7081838803062844e-05,
+      "loss": 0.6666,
+      "step": 7209
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7081069992019424e-05,
+      "loss": 0.6469,
+      "step": 7210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.708030109702097e-05,
+      "loss": 0.6149,
+      "step": 7211
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7079532118076585e-05,
+      "loss": 0.6274,
+      "step": 7212
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7078763055195393e-05,
+      "loss": 0.6552,
+      "step": 7213
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.707799390838651e-05,
+      "loss": 0.6516,
+      "step": 7214
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7077224677659055e-05,
+      "loss": 0.5999,
+      "step": 7215
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.707645536302215e-05,
+      "loss": 0.6535,
+      "step": 7216
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7075685964484913e-05,
+      "loss": 0.656,
+      "step": 7217
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7074916482056468e-05,
+      "loss": 0.6211,
+      "step": 7218
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7074146915745943e-05,
+      "loss": 0.6363,
+      "step": 7219
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.707337726556245e-05,
+      "loss": 0.6607,
+      "step": 7220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7072607531515125e-05,
+      "loss": 0.6326,
+      "step": 7221
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7071837713613094e-05,
+      "loss": 0.5892,
+      "step": 7222
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7071067811865477e-05,
+      "loss": 0.6658,
+      "step": 7223
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.707029782628141e-05,
+      "loss": 0.6128,
+      "step": 7224
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7069527756870014e-05,
+      "loss": 0.631,
+      "step": 7225
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7068757603640423e-05,
+      "loss": 0.6278,
+      "step": 7226
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.706798736660177e-05,
+      "loss": 0.5707,
+      "step": 7227
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7067217045763186e-05,
+      "loss": 0.6106,
+      "step": 7228
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7066446641133806e-05,
+      "loss": 0.6783,
+      "step": 7229
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.706567615272276e-05,
+      "loss": 0.624,
+      "step": 7230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7064905580539184e-05,
+      "loss": 0.6327,
+      "step": 7231
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7064134924592218e-05,
+      "loss": 0.6593,
+      "step": 7232
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7063364184890994e-05,
+      "loss": 0.639,
+      "step": 7233
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7062593361444655e-05,
+      "loss": 0.6177,
+      "step": 7234
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7061822454262333e-05,
+      "loss": 0.668,
+      "step": 7235
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7061051463353176e-05,
+      "loss": 0.5952,
+      "step": 7236
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7060280388726322e-05,
+      "loss": 0.6316,
+      "step": 7237
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.705950923039091e-05,
+      "loss": 0.6098,
+      "step": 7238
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.705873798835609e-05,
+      "loss": 0.578,
+      "step": 7239
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7057966662630997e-05,
+      "loss": 0.6958,
+      "step": 7240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7057195253224782e-05,
+      "loss": 0.6674,
+      "step": 7241
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7056423760146592e-05,
+      "loss": 0.6025,
+      "step": 7242
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.705565218340557e-05,
+      "loss": 0.7283,
+      "step": 7243
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7054880523010865e-05,
+      "loss": 0.65,
+      "step": 7244
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.705410877897163e-05,
+      "loss": 0.6569,
+      "step": 7245
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.705333695129701e-05,
+      "loss": 0.6201,
+      "step": 7246
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7052565039996158e-05,
+      "loss": 0.6257,
+      "step": 7247
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.705179304507823e-05,
+      "loss": 0.6353,
+      "step": 7248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.705102096655237e-05,
+      "loss": 0.5971,
+      "step": 7249
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7050248804427735e-05,
+      "loss": 0.6157,
+      "step": 7250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7049476558713487e-05,
+      "loss": 0.6184,
+      "step": 7251
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7048704229418775e-05,
+      "loss": 0.6451,
+      "step": 7252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7047931816552753e-05,
+      "loss": 0.6183,
+      "step": 7253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7047159320124592e-05,
+      "loss": 0.5745,
+      "step": 7254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7046386740143435e-05,
+      "loss": 0.6655,
+      "step": 7255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7045614076618457e-05,
+      "loss": 0.6387,
+      "step": 7256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7044841329558804e-05,
+      "loss": 0.6728,
+      "step": 7257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7044068498973652e-05,
+      "loss": 0.6436,
+      "step": 7258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7043295584872153e-05,
+      "loss": 0.6377,
+      "step": 7259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7042522587263475e-05,
+      "loss": 0.616,
+      "step": 7260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7041749506156783e-05,
+      "loss": 0.654,
+      "step": 7261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.704097634156124e-05,
+      "loss": 0.637,
+      "step": 7262
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.704020309348602e-05,
+      "loss": 0.6368,
+      "step": 7263
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.703942976194029e-05,
+      "loss": 0.6298,
+      "step": 7264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7038656346933206e-05,
+      "loss": 0.637,
+      "step": 7265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7037882848473956e-05,
+      "loss": 0.693,
+      "step": 7266
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7037109266571695e-05,
+      "loss": 0.6053,
+      "step": 7267
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7036335601235604e-05,
+      "loss": 0.623,
+      "step": 7268
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7035561852474853e-05,
+      "loss": 0.6727,
+      "step": 7269
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7034788020298618e-05,
+      "loss": 0.6631,
+      "step": 7270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7034014104716073e-05,
+      "loss": 0.5819,
+      "step": 7271
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7033240105736388e-05,
+      "loss": 0.6455,
+      "step": 7272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7032466023368748e-05,
+      "loss": 0.6517,
+      "step": 7273
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7031691857622326e-05,
+      "loss": 0.6571,
+      "step": 7274
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.70309176085063e-05,
+      "loss": 0.6488,
+      "step": 7275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7030143276029856e-05,
+      "loss": 0.6405,
+      "step": 7276
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7029368860202167e-05,
+      "loss": 0.6108,
+      "step": 7277
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7028594361032417e-05,
+      "loss": 0.6288,
+      "step": 7278
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7027819778529796e-05,
+      "loss": 0.623,
+      "step": 7279
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7027045112703474e-05,
+      "loss": 0.6164,
+      "step": 7280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7026270363562646e-05,
+      "loss": 0.6471,
+      "step": 7281
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7025495531116497e-05,
+      "loss": 0.6126,
+      "step": 7282
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7024720615374208e-05,
+      "loss": 0.6374,
+      "step": 7283
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.702394561634497e-05,
+      "loss": 0.6152,
+      "step": 7284
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.702317053403797e-05,
+      "loss": 0.64,
+      "step": 7285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.70223953684624e-05,
+      "loss": 0.6337,
+      "step": 7286
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.702162011962745e-05,
+      "loss": 0.6074,
+      "step": 7287
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.702084478754231e-05,
+      "loss": 0.6261,
+      "step": 7288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7020069372216176e-05,
+      "loss": 0.6281,
+      "step": 7289
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7019293873658238e-05,
+      "loss": 0.6633,
+      "step": 7290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7018518291877693e-05,
+      "loss": 0.6457,
+      "step": 7291
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7017742626883734e-05,
+      "loss": 0.6095,
+      "step": 7292
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7016966878685556e-05,
+      "loss": 0.5947,
+      "step": 7293
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7016191047292364e-05,
+      "loss": 0.652,
+      "step": 7294
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.701541513271335e-05,
+      "loss": 0.6476,
+      "step": 7295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7014639134957713e-05,
+      "loss": 0.6219,
+      "step": 7296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.701386305403466e-05,
+      "loss": 0.6376,
+      "step": 7297
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7013086889953384e-05,
+      "loss": 0.6355,
+      "step": 7298
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.701231064272309e-05,
+      "loss": 0.6442,
+      "step": 7299
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7011534312352985e-05,
+      "loss": 0.6562,
+      "step": 7300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.701075789885227e-05,
+      "loss": 0.6008,
+      "step": 7301
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7009981402230152e-05,
+      "loss": 0.6091,
+      "step": 7302
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.700920482249584e-05,
+      "loss": 0.648,
+      "step": 7303
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7008428159658536e-05,
+      "loss": 0.6102,
+      "step": 7304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.700765141372745e-05,
+      "loss": 0.6142,
+      "step": 7305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7006874584711787e-05,
+      "loss": 0.6118,
+      "step": 7306
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.700609767262077e-05,
+      "loss": 0.6632,
+      "step": 7307
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7005320677463598e-05,
+      "loss": 0.6528,
+      "step": 7308
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.700454359924949e-05,
+      "loss": 0.5996,
+      "step": 7309
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7003766437987654e-05,
+      "loss": 0.6821,
+      "step": 7310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7002989193687307e-05,
+      "loss": 0.589,
+      "step": 7311
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7002211866357667e-05,
+      "loss": 0.6182,
+      "step": 7312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7001434456007945e-05,
+      "loss": 0.5621,
+      "step": 7313
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7000656962647362e-05,
+      "loss": 0.6091,
+      "step": 7314
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6999879386285133e-05,
+      "loss": 0.6873,
+      "step": 7315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6999101726930483e-05,
+      "loss": 0.6215,
+      "step": 7316
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6998323984592625e-05,
+      "loss": 0.6241,
+      "step": 7317
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6997546159280785e-05,
+      "loss": 0.638,
+      "step": 7318
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6996768251004182e-05,
+      "loss": 0.662,
+      "step": 7319
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6995990259772042e-05,
+      "loss": 0.6223,
+      "step": 7320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6995212185593585e-05,
+      "loss": 0.6307,
+      "step": 7321
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6994434028478038e-05,
+      "loss": 0.6611,
+      "step": 7322
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6993655788434632e-05,
+      "loss": 0.6568,
+      "step": 7323
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6992877465472585e-05,
+      "loss": 0.6576,
+      "step": 7324
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.699209905960113e-05,
+      "loss": 0.5888,
+      "step": 7325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6991320570829502e-05,
+      "loss": 0.5923,
+      "step": 7326
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6990541999166917e-05,
+      "loss": 0.6525,
+      "step": 7327
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6989763344622615e-05,
+      "loss": 0.5845,
+      "step": 7328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.698898460720583e-05,
+      "loss": 0.6414,
+      "step": 7329
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.698820578692579e-05,
+      "loss": 0.6175,
+      "step": 7330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6987426883791728e-05,
+      "loss": 0.6449,
+      "step": 7331
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6986647897812884e-05,
+      "loss": 0.6431,
+      "step": 7332
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.698586882899849e-05,
+      "loss": 0.6615,
+      "step": 7333
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.698508967735778e-05,
+      "loss": 0.5945,
+      "step": 7334
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.69843104429e-05,
+      "loss": 0.6761,
+      "step": 7335
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6983531125634386e-05,
+      "loss": 0.6744,
+      "step": 7336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6982751725570173e-05,
+      "loss": 0.6188,
+      "step": 7337
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6981972242716604e-05,
+      "loss": 0.6159,
+      "step": 7338
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6981192677082926e-05,
+      "loss": 0.6443,
+      "step": 7339
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6980413028678374e-05,
+      "loss": 0.6435,
+      "step": 7340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6979633297512197e-05,
+      "loss": 0.6636,
+      "step": 7341
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6978853483593636e-05,
+      "loss": 0.6257,
+      "step": 7342
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.697807358693194e-05,
+      "loss": 0.7056,
+      "step": 7343
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6977293607536355e-05,
+      "loss": 0.63,
+      "step": 7344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6976513545416127e-05,
+      "loss": 0.6317,
+      "step": 7345
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6975733400580506e-05,
+      "loss": 0.6227,
+      "step": 7346
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6974953173038745e-05,
+      "loss": 0.6214,
+      "step": 7347
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6974172862800084e-05,
+      "loss": 0.6449,
+      "step": 7348
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6973392469873788e-05,
+      "loss": 0.6523,
+      "step": 7349
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6972611994269097e-05,
+      "loss": 0.6759,
+      "step": 7350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6971831435995275e-05,
+      "loss": 0.635,
+      "step": 7351
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6971050795061564e-05,
+      "loss": 0.6138,
+      "step": 7352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6970270071477238e-05,
+      "loss": 0.6123,
+      "step": 7353
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6969489265251536e-05,
+      "loss": 0.5839,
+      "step": 7354
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6968708376393724e-05,
+      "loss": 0.6106,
+      "step": 7355
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.696792740491306e-05,
+      "loss": 0.6357,
+      "step": 7356
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6967146350818802e-05,
+      "loss": 0.6181,
+      "step": 7357
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.6966365214120207e-05,
+      "loss": 0.6175,
+      "step": 7358
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.696558399482654e-05,
+      "loss": 0.6117,
+      "step": 7359
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.696480269294707e-05,
+      "loss": 0.6425,
+      "step": 7360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.696402130849105e-05,
+      "loss": 0.6314,
+      "step": 7361
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6963239841467745e-05,
+      "loss": 0.6281,
+      "step": 7362
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6962458291886426e-05,
+      "loss": 0.6499,
+      "step": 7363
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6961676659756356e-05,
+      "loss": 0.6684,
+      "step": 7364
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.69608949450868e-05,
+      "loss": 0.6122,
+      "step": 7365
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6960113147887032e-05,
+      "loss": 0.6781,
+      "step": 7366
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6959331268166318e-05,
+      "loss": 0.6128,
+      "step": 7367
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6958549305933926e-05,
+      "loss": 0.6566,
+      "step": 7368
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6957767261199132e-05,
+      "loss": 0.6426,
+      "step": 7369
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6956985133971206e-05,
+      "loss": 0.6444,
+      "step": 7370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.695620292425942e-05,
+      "loss": 0.6836,
+      "step": 7371
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6955420632073046e-05,
+      "loss": 0.6323,
+      "step": 7372
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6954638257421362e-05,
+      "loss": 0.6504,
+      "step": 7373
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6953855800313643e-05,
+      "loss": 0.6173,
+      "step": 7374
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6953073260759172e-05,
+      "loss": 0.6319,
+      "step": 7375
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6952290638767217e-05,
+      "loss": 0.648,
+      "step": 7376
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6951507934347063e-05,
+      "loss": 0.5749,
+      "step": 7377
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.695072514750799e-05,
+      "loss": 0.6145,
+      "step": 7378
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6949942278259278e-05,
+      "loss": 0.676,
+      "step": 7379
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6949159326610204e-05,
+      "loss": 0.619,
+      "step": 7380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6948376292570062e-05,
+      "loss": 0.5778,
+      "step": 7381
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6947593176148126e-05,
+      "loss": 0.6152,
+      "step": 7382
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6946809977353682e-05,
+      "loss": 0.5872,
+      "step": 7383
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.694602669619602e-05,
+      "loss": 0.628,
+      "step": 7384
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6945243332684427e-05,
+      "loss": 0.6489,
+      "step": 7385
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6944459886828186e-05,
+      "loss": 0.7078,
+      "step": 7386
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.694367635863659e-05,
+      "loss": 0.588,
+      "step": 7387
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6942892748118928e-05,
+      "loss": 0.6331,
+      "step": 7388
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6942109055284485e-05,
+      "loss": 0.6284,
+      "step": 7389
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.694132528014256e-05,
+      "loss": 0.6317,
+      "step": 7390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6940541422702445e-05,
+      "loss": 0.5889,
+      "step": 7391
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6939757482973425e-05,
+      "loss": 0.6238,
+      "step": 7392
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6938973460964808e-05,
+      "loss": 0.6512,
+      "step": 7393
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6938189356685882e-05,
+      "loss": 0.6436,
+      "step": 7394
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6937405170145943e-05,
+      "loss": 0.6523,
+      "step": 7395
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.693662090135429e-05,
+      "loss": 0.6494,
+      "step": 7396
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.693583655032022e-05,
+      "loss": 0.6254,
+      "step": 7397
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6935052117053033e-05,
+      "loss": 0.6572,
+      "step": 7398
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.693426760156203e-05,
+      "loss": 0.625,
+      "step": 7399
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6933483003856517e-05,
+      "loss": 0.597,
+      "step": 7400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.693269832394579e-05,
+      "loss": 0.6277,
+      "step": 7401
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.693191356183915e-05,
+      "loss": 0.5998,
+      "step": 7402
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6931128717545914e-05,
+      "loss": 0.6669,
+      "step": 7403
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6930343791075373e-05,
+      "loss": 0.6331,
+      "step": 7404
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6929558782436842e-05,
+      "loss": 0.6884,
+      "step": 7405
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.692877369163963e-05,
+      "loss": 0.6979,
+      "step": 7406
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6927988518693034e-05,
+      "loss": 0.6245,
+      "step": 7407
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.692720326360637e-05,
+      "loss": 0.6556,
+      "step": 7408
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.692641792638895e-05,
+      "loss": 0.6452,
+      "step": 7409
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6925632507050087e-05,
+      "loss": 0.6156,
+      "step": 7410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6924847005599084e-05,
+      "loss": 0.656,
+      "step": 7411
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6924061422045265e-05,
+      "loss": 0.6654,
+      "step": 7412
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6923275756397938e-05,
+      "loss": 0.6352,
+      "step": 7413
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6922490008666416e-05,
+      "loss": 0.669,
+      "step": 7414
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6921704178860018e-05,
+      "loss": 0.6086,
+      "step": 7415
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6920918266988063e-05,
+      "loss": 0.6775,
+      "step": 7416
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6920132273059866e-05,
+      "loss": 0.6257,
+      "step": 7417
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6919346197084748e-05,
+      "loss": 0.5923,
+      "step": 7418
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.691856003907203e-05,
+      "loss": 0.6184,
+      "step": 7419
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6917773799031026e-05,
+      "loss": 0.6582,
+      "step": 7420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.691698747697106e-05,
+      "loss": 0.6897,
+      "step": 7421
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6916201072901465e-05,
+      "loss": 0.6685,
+      "step": 7422
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6915414586831553e-05,
+      "loss": 0.6625,
+      "step": 7423
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6914628018770652e-05,
+      "loss": 0.6749,
+      "step": 7424
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6913841368728095e-05,
+      "loss": 0.6036,
+      "step": 7425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6913054636713196e-05,
+      "loss": 0.6688,
+      "step": 7426
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6912267822735293e-05,
+      "loss": 0.5863,
+      "step": 7427
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.691148092680371e-05,
+      "loss": 0.5703,
+      "step": 7428
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6910693948927777e-05,
+      "loss": 0.6317,
+      "step": 7429
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6909906889116822e-05,
+      "loss": 0.619,
+      "step": 7430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6909119747380184e-05,
+      "loss": 0.6391,
+      "step": 7431
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.690833252372719e-05,
+      "loss": 0.7282,
+      "step": 7432
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6907545218167176e-05,
+      "loss": 0.6539,
+      "step": 7433
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6906757830709475e-05,
+      "loss": 0.6133,
+      "step": 7434
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.690597036136342e-05,
+      "loss": 0.6649,
+      "step": 7435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6905182810138353e-05,
+      "loss": 0.6155,
+      "step": 7436
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6904395177043608e-05,
+      "loss": 0.617,
+      "step": 7437
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6903607462088526e-05,
+      "loss": 0.5995,
+      "step": 7438
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.690281966528244e-05,
+      "loss": 0.6814,
+      "step": 7439
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6902031786634697e-05,
+      "loss": 0.637,
+      "step": 7440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.690124382615464e-05,
+      "loss": 0.6163,
+      "step": 7441
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6900455783851603e-05,
+      "loss": 0.5606,
+      "step": 7442
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.689966765973493e-05,
+      "loss": 0.6134,
+      "step": 7443
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.689887945381398e-05,
+      "loss": 0.6659,
+      "step": 7444
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6898091166098077e-05,
+      "loss": 0.6286,
+      "step": 7445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.689730279659658e-05,
+      "loss": 0.584,
+      "step": 7446
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6896514345318834e-05,
+      "loss": 0.6608,
+      "step": 7447
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6895725812274184e-05,
+      "loss": 0.6723,
+      "step": 7448
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6894937197471987e-05,
+      "loss": 0.6883,
+      "step": 7449
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6894148500921583e-05,
+      "loss": 0.6146,
+      "step": 7450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6893359722632326e-05,
+      "loss": 0.6362,
+      "step": 7451
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6892570862613572e-05,
+      "loss": 0.6392,
+      "step": 7452
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6891781920874672e-05,
+      "loss": 0.682,
+      "step": 7453
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6890992897424973e-05,
+      "loss": 0.6216,
+      "step": 7454
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6890203792273844e-05,
+      "loss": 0.6017,
+      "step": 7455
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.688941460543063e-05,
+      "loss": 0.686,
+      "step": 7456
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.688862533690469e-05,
+      "loss": 0.6104,
+      "step": 7457
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.688783598670538e-05,
+      "loss": 0.6388,
+      "step": 7458
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6887046554842066e-05,
+      "loss": 0.6543,
+      "step": 7459
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.68862570413241e-05,
+      "loss": 0.5787,
+      "step": 7460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6885467446160845e-05,
+      "loss": 0.6616,
+      "step": 7461
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6884677769361667e-05,
+      "loss": 0.6271,
+      "step": 7462
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6883888010935924e-05,
+      "loss": 0.6452,
+      "step": 7463
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6883098170892975e-05,
+      "loss": 0.6672,
+      "step": 7464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6882308249242197e-05,
+      "loss": 0.6159,
+      "step": 7465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6881518245992944e-05,
+      "loss": 0.5754,
+      "step": 7466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.688072816115459e-05,
+      "loss": 0.6591,
+      "step": 7467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6879937994736498e-05,
+      "loss": 0.6613,
+      "step": 7468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.687914774674804e-05,
+      "loss": 0.6145,
+      "step": 7469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.687835741719858e-05,
+      "loss": 0.6292,
+      "step": 7470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6877567006097494e-05,
+      "loss": 0.6315,
+      "step": 7471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.687677651345415e-05,
+      "loss": 0.6473,
+      "step": 7472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.687598593927792e-05,
+      "loss": 0.7176,
+      "step": 7473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6875195283578178e-05,
+      "loss": 0.6524,
+      "step": 7474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.68744045463643e-05,
+      "loss": 0.5979,
+      "step": 7475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6873613727645662e-05,
+      "loss": 0.6477,
+      "step": 7476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6872822827431635e-05,
+      "loss": 0.6444,
+      "step": 7477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.68720318457316e-05,
+      "loss": 0.6494,
+      "step": 7478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6871240782554937e-05,
+      "loss": 0.5808,
+      "step": 7479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.687044963791102e-05,
+      "loss": 0.625,
+      "step": 7480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6869658411809232e-05,
+      "loss": 0.6904,
+      "step": 7481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.686886710425895e-05,
+      "loss": 0.664,
+      "step": 7482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6868075715269566e-05,
+      "loss": 0.6704,
+      "step": 7483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6867284244850452e-05,
+      "loss": 0.6466,
+      "step": 7484
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6866492693010997e-05,
+      "loss": 0.6463,
+      "step": 7485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6865701059760583e-05,
+      "loss": 0.637,
+      "step": 7486
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6864909345108602e-05,
+      "loss": 0.6489,
+      "step": 7487
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6864117549064432e-05,
+      "loss": 0.6508,
+      "step": 7488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.686332567163747e-05,
+      "loss": 0.6511,
+      "step": 7489
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6862533712837097e-05,
+      "loss": 0.6368,
+      "step": 7490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6861741672672706e-05,
+      "loss": 0.6577,
+      "step": 7491
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6860949551153687e-05,
+      "loss": 0.6114,
+      "step": 7492
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6860157348289435e-05,
+      "loss": 0.6154,
+      "step": 7493
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6859365064089336e-05,
+      "loss": 0.6466,
+      "step": 7494
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6858572698562788e-05,
+      "loss": 0.6137,
+      "step": 7495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6857780251719183e-05,
+      "loss": 0.6315,
+      "step": 7496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.685698772356792e-05,
+      "loss": 0.6619,
+      "step": 7497
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6856195114118394e-05,
+      "loss": 0.6531,
+      "step": 7498
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.685540242338e-05,
+      "loss": 0.6301,
+      "step": 7499
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6854609651362137e-05,
+      "loss": 0.6248,
+      "step": 7500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6853816798074207e-05,
+      "loss": 0.6332,
+      "step": 7501
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.685302386352561e-05,
+      "loss": 0.6999,
+      "step": 7502
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6852230847725745e-05,
+      "loss": 0.5905,
+      "step": 7503
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6851437750684014e-05,
+      "loss": 0.6329,
+      "step": 7504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6850644572409824e-05,
+      "loss": 0.664,
+      "step": 7505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6849851312912574e-05,
+      "loss": 0.6062,
+      "step": 7506
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6849057972201672e-05,
+      "loss": 0.6127,
+      "step": 7507
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6848264550286525e-05,
+      "loss": 0.5974,
+      "step": 7508
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.684747104717654e-05,
+      "loss": 0.6171,
+      "step": 7509
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6846677462881116e-05,
+      "loss": 0.6236,
+      "step": 7510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.684588379740968e-05,
+      "loss": 0.5907,
+      "step": 7511
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6845090050771628e-05,
+      "loss": 0.6249,
+      "step": 7512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6844296222976368e-05,
+      "loss": 0.6252,
+      "step": 7513
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6843502314033323e-05,
+      "loss": 0.6609,
+      "step": 7514
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6842708323951904e-05,
+      "loss": 0.6262,
+      "step": 7515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6841914252741516e-05,
+      "loss": 0.5856,
+      "step": 7516
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6841120100411584e-05,
+      "loss": 0.6316,
+      "step": 7517
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6840325866971517e-05,
+      "loss": 0.6284,
+      "step": 7518
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6839531552430736e-05,
+      "loss": 0.661,
+      "step": 7519
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6838737156798653e-05,
+      "loss": 0.6628,
+      "step": 7520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.683794268008469e-05,
+      "loss": 0.6244,
+      "step": 7521
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.683714812229827e-05,
+      "loss": 0.6368,
+      "step": 7522
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6836353483448807e-05,
+      "loss": 0.6858,
+      "step": 7523
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6835558763545728e-05,
+      "loss": 0.6845,
+      "step": 7524
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6834763962598452e-05,
+      "loss": 0.5765,
+      "step": 7525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.68339690806164e-05,
+      "loss": 0.652,
+      "step": 7526
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6833174117609002e-05,
+      "loss": 0.5468,
+      "step": 7527
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.683237907358568e-05,
+      "loss": 0.7273,
+      "step": 7528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6831583948555863e-05,
+      "loss": 0.6211,
+      "step": 7529
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6830788742528975e-05,
+      "loss": 0.6394,
+      "step": 7530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6829993455514445e-05,
+      "loss": 0.6083,
+      "step": 7531
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6829198087521705e-05,
+      "loss": 0.6699,
+      "step": 7532
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.682840263856018e-05,
+      "loss": 0.6207,
+      "step": 7533
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6827607108639304e-05,
+      "loss": 0.6394,
+      "step": 7534
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6826811497768513e-05,
+      "loss": 0.6235,
+      "step": 7535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.682601580595723e-05,
+      "loss": 0.6883,
+      "step": 7536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6825220033214897e-05,
+      "loss": 0.6533,
+      "step": 7537
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6824424179550947e-05,
+      "loss": 0.6835,
+      "step": 7538
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6823628244974816e-05,
+      "loss": 0.6695,
+      "step": 7539
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.682283222949594e-05,
+      "loss": 0.5917,
+      "step": 7540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.682203613312375e-05,
+      "loss": 0.7009,
+      "step": 7541
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.68212399558677e-05,
+      "loss": 0.6497,
+      "step": 7542
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.682044369773722e-05,
+      "loss": 0.6498,
+      "step": 7543
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.681964735874175e-05,
+      "loss": 0.6564,
+      "step": 7544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6818850938890735e-05,
+      "loss": 0.6482,
+      "step": 7545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6818054438193615e-05,
+      "loss": 0.585,
+      "step": 7546
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6817257856659837e-05,
+      "loss": 0.6494,
+      "step": 7547
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6816461194298842e-05,
+      "loss": 0.7678,
+      "step": 7548
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.681566445112008e-05,
+      "loss": 0.5734,
+      "step": 7549
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.681486762713299e-05,
+      "loss": 0.6589,
+      "step": 7550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6814070722347024e-05,
+      "loss": 0.5806,
+      "step": 7551
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.681327373677163e-05,
+      "loss": 0.6014,
+      "step": 7552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6812476670416257e-05,
+      "loss": 0.6026,
+      "step": 7553
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6811679523290358e-05,
+      "loss": 0.6519,
+      "step": 7554
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6810882295403373e-05,
+      "loss": 0.6264,
+      "step": 7555
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.681008498676477e-05,
+      "loss": 0.6839,
+      "step": 7556
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6809287597383992e-05,
+      "loss": 0.6282,
+      "step": 7557
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6808490127270498e-05,
+      "loss": 0.6158,
+      "step": 7558
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.680769257643374e-05,
+      "loss": 0.6586,
+      "step": 7559
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6806894944883173e-05,
+      "loss": 0.6373,
+      "step": 7560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6806097232628258e-05,
+      "loss": 0.6018,
+      "step": 7561
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.680529943967845e-05,
+      "loss": 0.6907,
+      "step": 7562
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.680450156604321e-05,
+      "loss": 0.6197,
+      "step": 7563
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.680370361173199e-05,
+      "loss": 0.59,
+      "step": 7564
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6802905576754263e-05,
+      "loss": 0.6058,
+      "step": 7565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.680210746111948e-05,
+      "loss": 0.6178,
+      "step": 7566
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6801309264837112e-05,
+      "loss": 0.701,
+      "step": 7567
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6800510987916617e-05,
+      "loss": 0.6167,
+      "step": 7568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6799712630367463e-05,
+      "loss": 0.63,
+      "step": 7569
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6798914192199115e-05,
+      "loss": 0.6019,
+      "step": 7570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6798115673421034e-05,
+      "loss": 0.6163,
+      "step": 7571
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.67973170740427e-05,
+      "loss": 0.6617,
+      "step": 7572
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6796518394073568e-05,
+      "loss": 0.6574,
+      "step": 7573
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6795719633523112e-05,
+      "loss": 0.6299,
+      "step": 7574
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6794920792400805e-05,
+      "loss": 0.6297,
+      "step": 7575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6794121870716117e-05,
+      "loss": 0.6475,
+      "step": 7576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6793322868478518e-05,
+      "loss": 0.6566,
+      "step": 7577
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.679252378569748e-05,
+      "loss": 0.6302,
+      "step": 7578
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6791724622382486e-05,
+      "loss": 0.6159,
+      "step": 7579
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6790925378543002e-05,
+      "loss": 0.6636,
+      "step": 7580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6790126054188507e-05,
+      "loss": 0.6183,
+      "step": 7581
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6789326649328478e-05,
+      "loss": 0.619,
+      "step": 7582
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6788527163972396e-05,
+      "loss": 0.6323,
+      "step": 7583
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6787727598129733e-05,
+      "loss": 0.6319,
+      "step": 7584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6786927951809975e-05,
+      "loss": 0.6489,
+      "step": 7585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.67861282250226e-05,
+      "loss": 0.6439,
+      "step": 7586
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.678532841777709e-05,
+      "loss": 0.6295,
+      "step": 7587
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6784528530082927e-05,
+      "loss": 0.6918,
+      "step": 7588
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6783728561949596e-05,
+      "loss": 0.6112,
+      "step": 7589
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6782928513386583e-05,
+      "loss": 0.6791,
+      "step": 7590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6782128384403372e-05,
+      "loss": 0.6296,
+      "step": 7591
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.678132817500945e-05,
+      "loss": 0.6411,
+      "step": 7592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6780527885214303e-05,
+      "loss": 0.6683,
+      "step": 7593
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6779727515027418e-05,
+      "loss": 0.6256,
+      "step": 7594
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.677892706445829e-05,
+      "loss": 0.6708,
+      "step": 7595
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6778126533516406e-05,
+      "loss": 0.6229,
+      "step": 7596
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.677732592221126e-05,
+      "loss": 0.6145,
+      "step": 7597
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.677652523055234e-05,
+      "loss": 0.6392,
+      "step": 7598
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6775724458549138e-05,
+      "loss": 0.6069,
+      "step": 7599
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6774923606211154e-05,
+      "loss": 0.6342,
+      "step": 7600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.677412267354788e-05,
+      "loss": 0.6288,
+      "step": 7601
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6773321660568813e-05,
+      "loss": 0.5799,
+      "step": 7602
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6772520567283447e-05,
+      "loss": 0.6064,
+      "step": 7603
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6771719393701286e-05,
+      "loss": 0.6453,
+      "step": 7604
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6770918139831822e-05,
+      "loss": 0.6659,
+      "step": 7605
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.677011680568456e-05,
+      "loss": 0.6164,
+      "step": 7606
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6769315391268997e-05,
+      "loss": 0.6224,
+      "step": 7607
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.676851389659464e-05,
+      "loss": 0.6352,
+      "step": 7608
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6767712321670985e-05,
+      "loss": 0.6689,
+      "step": 7609
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6766910666507544e-05,
+      "loss": 0.5952,
+      "step": 7610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6766108931113812e-05,
+      "loss": 0.6321,
+      "step": 7611
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6765307115499305e-05,
+      "loss": 0.6475,
+      "step": 7612
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6764505219673518e-05,
+      "loss": 0.6751,
+      "step": 7613
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.676370324364597e-05,
+      "loss": 0.6979,
+      "step": 7614
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.676290118742616e-05,
+      "loss": 0.6563,
+      "step": 7615
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.67620990510236e-05,
+      "loss": 0.6386,
+      "step": 7616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6761296834447804e-05,
+      "loss": 0.6635,
+      "step": 7617
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6760494537708282e-05,
+      "loss": 0.6462,
+      "step": 7618
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6759692160814544e-05,
+      "loss": 0.5732,
+      "step": 7619
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6758889703776104e-05,
+      "loss": 0.6457,
+      "step": 7620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6758087166602476e-05,
+      "loss": 0.6229,
+      "step": 7621
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6757284549303176e-05,
+      "loss": 0.6478,
+      "step": 7622
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6756481851887723e-05,
+      "loss": 0.6374,
+      "step": 7623
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6755679074365626e-05,
+      "loss": 0.6694,
+      "step": 7624
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.675487621674641e-05,
+      "loss": 0.6553,
+      "step": 7625
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.675407327903959e-05,
+      "loss": 0.6589,
+      "step": 7626
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6753270261254687e-05,
+      "loss": 0.6437,
+      "step": 7627
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6752467163401226e-05,
+      "loss": 0.649,
+      "step": 7628
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6751663985488722e-05,
+      "loss": 0.6168,
+      "step": 7629
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.67508607275267e-05,
+      "loss": 0.6582,
+      "step": 7630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6750057389524682e-05,
+      "loss": 0.6226,
+      "step": 7631
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.67492539714922e-05,
+      "loss": 0.6452,
+      "step": 7632
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6748450473438773e-05,
+      "loss": 0.6101,
+      "step": 7633
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.674764689537393e-05,
+      "loss": 0.6024,
+      "step": 7634
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6746843237307195e-05,
+      "loss": 0.6164,
+      "step": 7635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6746039499248096e-05,
+      "loss": 0.6315,
+      "step": 7636
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6745235681206166e-05,
+      "loss": 0.6428,
+      "step": 7637
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.674443178319094e-05,
+      "loss": 0.6774,
+      "step": 7638
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.674362780521194e-05,
+      "loss": 0.6459,
+      "step": 7639
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6742823747278698e-05,
+      "loss": 0.6525,
+      "step": 7640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6742019609400756e-05,
+      "loss": 0.6083,
+      "step": 7641
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.674121539158764e-05,
+      "loss": 0.6771,
+      "step": 7642
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.674041109384889e-05,
+      "loss": 0.6199,
+      "step": 7643
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6739606716194036e-05,
+      "loss": 0.6204,
+      "step": 7644
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6738802258632622e-05,
+      "loss": 0.6013,
+      "step": 7645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6737997721174178e-05,
+      "loss": 0.6486,
+      "step": 7646
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.6737193103828252e-05,
+      "loss": 0.6048,
+      "step": 7647
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.673638840660438e-05,
+      "loss": 0.6451,
+      "step": 7648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.67355836295121e-05,
+      "loss": 0.6486,
+      "step": 7649
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6734778772560954e-05,
+      "loss": 0.668,
+      "step": 7650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.673397383576049e-05,
+      "loss": 0.5881,
+      "step": 7651
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6733168819120244e-05,
+      "loss": 0.6121,
+      "step": 7652
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6732363722649765e-05,
+      "loss": 0.6149,
+      "step": 7653
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.67315585463586e-05,
+      "loss": 0.6227,
+      "step": 7654
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.673075329025629e-05,
+      "loss": 0.5831,
+      "step": 7655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6729947954352387e-05,
+      "loss": 0.6537,
+      "step": 7656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6729142538656434e-05,
+      "loss": 0.6498,
+      "step": 7657
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6728337043177987e-05,
+      "loss": 0.6214,
+      "step": 7658
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.672753146792659e-05,
+      "loss": 0.6434,
+      "step": 7659
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6726725812911803e-05,
+      "loss": 0.591,
+      "step": 7660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.672592007814317e-05,
+      "loss": 0.6201,
+      "step": 7661
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6725114263630245e-05,
+      "loss": 0.6479,
+      "step": 7662
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.672430836938258e-05,
+      "loss": 0.6507,
+      "step": 7663
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6723502395409737e-05,
+      "loss": 0.6556,
+      "step": 7664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6722696341721267e-05,
+      "loss": 0.5714,
+      "step": 7665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.672189020832673e-05,
+      "loss": 0.6272,
+      "step": 7666
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6721083995235678e-05,
+      "loss": 0.6292,
+      "step": 7667
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.672027770245767e-05,
+      "loss": 0.6034,
+      "step": 7668
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6719471330002277e-05,
+      "loss": 0.6665,
+      "step": 7669
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6718664877879046e-05,
+      "loss": 0.6498,
+      "step": 7670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6717858346097543e-05,
+      "loss": 0.6197,
+      "step": 7671
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6717051734667334e-05,
+      "loss": 0.5976,
+      "step": 7672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6716245043597976e-05,
+      "loss": 0.6317,
+      "step": 7673
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.671543827289904e-05,
+      "loss": 0.6078,
+      "step": 7674
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6714631422580088e-05,
+      "loss": 0.6317,
+      "step": 7675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6713824492650684e-05,
+      "loss": 0.6429,
+      "step": 7676
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.67130174831204e-05,
+      "loss": 0.6714,
+      "step": 7677
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.67122103939988e-05,
+      "loss": 0.6499,
+      "step": 7678
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.671140322529546e-05,
+      "loss": 0.6318,
+      "step": 7679
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.671059597701994e-05,
+      "loss": 0.5943,
+      "step": 7680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6709788649181817e-05,
+      "loss": 0.6454,
+      "step": 7681
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.670898124179066e-05,
+      "loss": 0.6095,
+      "step": 7682
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6708173754856044e-05,
+      "loss": 0.6791,
+      "step": 7683
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6707366188387543e-05,
+      "loss": 0.6097,
+      "step": 7684
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.670655854239473e-05,
+      "loss": 0.655,
+      "step": 7685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6705750816887182e-05,
+      "loss": 0.6669,
+      "step": 7686
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6704943011874478e-05,
+      "loss": 0.6122,
+      "step": 7687
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6704135127366187e-05,
+      "loss": 0.6989,
+      "step": 7688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.67033271633719e-05,
+      "loss": 0.6387,
+      "step": 7689
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6702519119901183e-05,
+      "loss": 0.642,
+      "step": 7690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6701710996963627e-05,
+      "loss": 0.619,
+      "step": 7691
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.670090279456881e-05,
+      "loss": 0.6384,
+      "step": 7692
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6700094512726312e-05,
+      "loss": 0.6127,
+      "step": 7693
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.669928615144572e-05,
+      "loss": 0.6568,
+      "step": 7694
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6698477710736615e-05,
+      "loss": 0.6714,
+      "step": 7695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.669766919060858e-05,
+      "loss": 0.6279,
+      "step": 7696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6696860591071205e-05,
+      "loss": 0.6449,
+      "step": 7697
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.669605191213408e-05,
+      "loss": 0.6193,
+      "step": 7698
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6695243153806787e-05,
+      "loss": 0.5965,
+      "step": 7699
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6694434316098917e-05,
+      "loss": 0.6316,
+      "step": 7700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6693625399020058e-05,
+      "loss": 0.5819,
+      "step": 7701
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6692816402579806e-05,
+      "loss": 0.6918,
+      "step": 7702
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.669200732678775e-05,
+      "loss": 0.5988,
+      "step": 7703
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6691198171653475e-05,
+      "loss": 0.6192,
+      "step": 7704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6690388937186586e-05,
+      "loss": 0.6333,
+      "step": 7705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.668957962339667e-05,
+      "loss": 0.6409,
+      "step": 7706
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.668877023029333e-05,
+      "loss": 0.6259,
+      "step": 7707
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6687960757886156e-05,
+      "loss": 0.6048,
+      "step": 7708
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.668715120618474e-05,
+      "loss": 0.6374,
+      "step": 7709
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6686341575198694e-05,
+      "loss": 0.7457,
+      "step": 7710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.668553186493761e-05,
+      "loss": 0.6033,
+      "step": 7711
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.668472207541109e-05,
+      "loss": 0.6525,
+      "step": 7712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6683912206628728e-05,
+      "loss": 0.5833,
+      "step": 7713
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6683102258600133e-05,
+      "loss": 0.6839,
+      "step": 7714
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6682292231334912e-05,
+      "loss": 0.6265,
+      "step": 7715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6681482124842658e-05,
+      "loss": 0.6093,
+      "step": 7716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6680671939132983e-05,
+      "loss": 0.7036,
+      "step": 7717
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.667986167421549e-05,
+      "loss": 0.6379,
+      "step": 7718
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.667905133009979e-05,
+      "loss": 0.672,
+      "step": 7719
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6678240906795484e-05,
+      "loss": 0.6148,
+      "step": 7720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6677430404312186e-05,
+      "loss": 0.5953,
+      "step": 7721
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6676619822659504e-05,
+      "loss": 0.6249,
+      "step": 7722
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6675809161847046e-05,
+      "loss": 0.6359,
+      "step": 7723
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6674998421884424e-05,
+      "loss": 0.6871,
+      "step": 7724
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6674187602781254e-05,
+      "loss": 0.6104,
+      "step": 7725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6673376704547148e-05,
+      "loss": 0.6286,
+      "step": 7726
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.667256572719172e-05,
+      "loss": 0.6596,
+      "step": 7727
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.667175467072458e-05,
+      "loss": 0.607,
+      "step": 7728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6670943535155353e-05,
+      "loss": 0.6645,
+      "step": 7729
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.667013232049365e-05,
+      "loss": 0.6464,
+      "step": 7730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.666932102674909e-05,
+      "loss": 0.5897,
+      "step": 7731
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.666850965393129e-05,
+      "loss": 0.6346,
+      "step": 7732
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.666769820204988e-05,
+      "loss": 0.6064,
+      "step": 7733
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.666688667111447e-05,
+      "loss": 0.6003,
+      "step": 7734
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6666075061134682e-05,
+      "loss": 0.6139,
+      "step": 7735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6665263372120144e-05,
+      "loss": 0.6037,
+      "step": 7736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6664451604080475e-05,
+      "loss": 0.6523,
+      "step": 7737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6663639757025306e-05,
+      "loss": 0.6384,
+      "step": 7738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6662827830964254e-05,
+      "loss": 0.6765,
+      "step": 7739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.666201582590695e-05,
+      "loss": 0.5676,
+      "step": 7740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.666120374186302e-05,
+      "loss": 0.6612,
+      "step": 7741
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6660391578842095e-05,
+      "loss": 0.662,
+      "step": 7742
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.66595793368538e-05,
+      "loss": 0.6481,
+      "step": 7743
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.665876701590777e-05,
+      "loss": 0.6164,
+      "step": 7744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6657954616013635e-05,
+      "loss": 0.6137,
+      "step": 7745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6657142137181023e-05,
+      "loss": 0.5964,
+      "step": 7746
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.665632957941957e-05,
+      "loss": 0.6588,
+      "step": 7747
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6655516942738912e-05,
+      "loss": 0.6223,
+      "step": 7748
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6654704227148676e-05,
+      "loss": 0.6722,
+      "step": 7749
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.665389143265851e-05,
+      "loss": 0.6562,
+      "step": 7750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6653078559278038e-05,
+      "loss": 0.6743,
+      "step": 7751
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6652265607016906e-05,
+      "loss": 0.653,
+      "step": 7752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.665145257588475e-05,
+      "loss": 0.6384,
+      "step": 7753
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.665063946589121e-05,
+      "loss": 0.6315,
+      "step": 7754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6649826277045926e-05,
+      "loss": 0.6089,
+      "step": 7755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.664901300935854e-05,
+      "loss": 0.5965,
+      "step": 7756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6648199662838692e-05,
+      "loss": 0.6674,
+      "step": 7757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6647386237496025e-05,
+      "loss": 0.6195,
+      "step": 7758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.664657273334019e-05,
+      "loss": 0.6307,
+      "step": 7759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6645759150380824e-05,
+      "loss": 0.6885,
+      "step": 7760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6644945488627578e-05,
+      "loss": 0.6738,
+      "step": 7761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6644131748090097e-05,
+      "loss": 0.5674,
+      "step": 7762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.664331792877803e-05,
+      "loss": 0.6348,
+      "step": 7763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6642504030701024e-05,
+      "loss": 0.6172,
+      "step": 7764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.664169005386873e-05,
+      "loss": 0.6958,
+      "step": 7765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.66408759982908e-05,
+      "loss": 0.6656,
+      "step": 7766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.664006186397688e-05,
+      "loss": 0.6277,
+      "step": 7767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.663924765093663e-05,
+      "loss": 0.6193,
+      "step": 7768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6638433359179703e-05,
+      "loss": 0.6386,
+      "step": 7769
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6637618988715746e-05,
+      "loss": 0.5694,
+      "step": 7770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.663680453955442e-05,
+      "loss": 0.6247,
+      "step": 7771
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6635990011705386e-05,
+      "loss": 0.6747,
+      "step": 7772
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.663517540517829e-05,
+      "loss": 0.6445,
+      "step": 7773
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6634360719982798e-05,
+      "loss": 0.6315,
+      "step": 7774
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6633545956128564e-05,
+      "loss": 0.667,
+      "step": 7775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6632731113625257e-05,
+      "loss": 0.5812,
+      "step": 7776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6631916192482528e-05,
+      "loss": 0.673,
+      "step": 7777
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6631101192710045e-05,
+      "loss": 0.6085,
+      "step": 7778
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6630286114317468e-05,
+      "loss": 0.6176,
+      "step": 7779
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6629470957314464e-05,
+      "loss": 0.63,
+      "step": 7780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6628655721710695e-05,
+      "loss": 0.6555,
+      "step": 7781
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6627840407515824e-05,
+      "loss": 0.6037,
+      "step": 7782
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6627025014739522e-05,
+      "loss": 0.6599,
+      "step": 7783
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6626209543391458e-05,
+      "loss": 0.6258,
+      "step": 7784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6625393993481294e-05,
+      "loss": 0.6276,
+      "step": 7785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6624578365018704e-05,
+      "loss": 0.5492,
+      "step": 7786
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.662376265801336e-05,
+      "loss": 0.575,
+      "step": 7787
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6622946872474926e-05,
+      "loss": 0.5956,
+      "step": 7788
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6622131008413083e-05,
+      "loss": 0.6005,
+      "step": 7789
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6621315065837497e-05,
+      "loss": 0.6094,
+      "step": 7790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6620499044757846e-05,
+      "loss": 0.6437,
+      "step": 7791
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6619682945183805e-05,
+      "loss": 0.6222,
+      "step": 7792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6618866767125046e-05,
+      "loss": 0.6464,
+      "step": 7793
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.661805051059125e-05,
+      "loss": 0.6552,
+      "step": 7794
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6617234175592095e-05,
+      "loss": 0.6483,
+      "step": 7795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6616417762137253e-05,
+      "loss": 0.6252,
+      "step": 7796
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6615601270236416e-05,
+      "loss": 0.5911,
+      "step": 7797
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.661478469989925e-05,
+      "loss": 0.6174,
+      "step": 7798
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6613968051135446e-05,
+      "loss": 0.6333,
+      "step": 7799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6613151323954688e-05,
+      "loss": 0.6151,
+      "step": 7800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6612334518366654e-05,
+      "loss": 0.6279,
+      "step": 7801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6611517634381025e-05,
+      "loss": 0.6625,
+      "step": 7802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6610700672007494e-05,
+      "loss": 0.6214,
+      "step": 7803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6609883631255744e-05,
+      "loss": 0.6427,
+      "step": 7804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.660906651213546e-05,
+      "loss": 0.6828,
+      "step": 7805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6608249314656337e-05,
+      "loss": 0.6497,
+      "step": 7806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6607432038828052e-05,
+      "loss": 0.6679,
+      "step": 7807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.660661468466031e-05,
+      "loss": 0.6315,
+      "step": 7808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6605797252162784e-05,
+      "loss": 0.6594,
+      "step": 7809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.660497974134518e-05,
+      "loss": 0.6017,
+      "step": 7810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6604162152217186e-05,
+      "loss": 0.6074,
+      "step": 7811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6603344484788494e-05,
+      "loss": 0.6308,
+      "step": 7812
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6602526739068803e-05,
+      "loss": 0.6208,
+      "step": 7813
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6601708915067805e-05,
+      "loss": 0.637,
+      "step": 7814
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6600891012795193e-05,
+      "loss": 0.6344,
+      "step": 7815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.660007303226067e-05,
+      "loss": 0.5901,
+      "step": 7816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.659925497347393e-05,
+      "loss": 0.6437,
+      "step": 7817
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6598436836444678e-05,
+      "loss": 0.6312,
+      "step": 7818
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.659761862118261e-05,
+      "loss": 0.6622,
+      "step": 7819
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6596800327697423e-05,
+      "loss": 0.6692,
+      "step": 7820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6595981955998826e-05,
+      "loss": 0.6164,
+      "step": 7821
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.659516350609652e-05,
+      "loss": 0.6089,
+      "step": 7822
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6594344978000204e-05,
+      "loss": 0.648,
+      "step": 7823
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.659352637171959e-05,
+      "loss": 0.5927,
+      "step": 7824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.659270768726438e-05,
+      "loss": 0.6513,
+      "step": 7825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6591888924644276e-05,
+      "loss": 0.6063,
+      "step": 7826
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6591070083868993e-05,
+      "loss": 0.6912,
+      "step": 7827
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.659025116494824e-05,
+      "loss": 0.6305,
+      "step": 7828
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6589432167891717e-05,
+      "loss": 0.6166,
+      "step": 7829
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6588613092709143e-05,
+      "loss": 0.6202,
+      "step": 7830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6587793939410226e-05,
+      "loss": 0.5826,
+      "step": 7831
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6586974708004677e-05,
+      "loss": 0.6218,
+      "step": 7832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6586155398502213e-05,
+      "loss": 0.6031,
+      "step": 7833
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.658533601091254e-05,
+      "loss": 0.6426,
+      "step": 7834
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6584516545245385e-05,
+      "loss": 0.6553,
+      "step": 7835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6583697001510453e-05,
+      "loss": 0.6593,
+      "step": 7836
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6582877379717464e-05,
+      "loss": 0.6104,
+      "step": 7837
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.658205767987614e-05,
+      "loss": 0.6599,
+      "step": 7838
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.658123790199619e-05,
+      "loss": 0.683,
+      "step": 7839
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6580418046087344e-05,
+      "loss": 0.6497,
+      "step": 7840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6579598112159313e-05,
+      "loss": 0.6041,
+      "step": 7841
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6578778100221827e-05,
+      "loss": 0.6833,
+      "step": 7842
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6577958010284604e-05,
+      "loss": 0.6789,
+      "step": 7843
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6577137842357368e-05,
+      "loss": 0.5845,
+      "step": 7844
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6576317596449843e-05,
+      "loss": 0.6212,
+      "step": 7845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.657549727257175e-05,
+      "loss": 0.6256,
+      "step": 7846
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6574676870732824e-05,
+      "loss": 0.6167,
+      "step": 7847
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6573856390942782e-05,
+      "loss": 0.6683,
+      "step": 7848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6573035833211363e-05,
+      "loss": 0.6322,
+      "step": 7849
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6572215197548285e-05,
+      "loss": 0.595,
+      "step": 7850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6571394483963278e-05,
+      "loss": 0.6118,
+      "step": 7851
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.657057369246608e-05,
+      "loss": 0.6383,
+      "step": 7852
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.656975282306642e-05,
+      "loss": 0.5938,
+      "step": 7853
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.656893187577403e-05,
+      "loss": 0.7081,
+      "step": 7854
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6568110850598643e-05,
+      "loss": 0.6119,
+      "step": 7855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.656728974754999e-05,
+      "loss": 0.6356,
+      "step": 7856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6566468566637812e-05,
+      "loss": 0.6757,
+      "step": 7857
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6565647307871843e-05,
+      "loss": 0.6284,
+      "step": 7858
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.656482597126182e-05,
+      "loss": 0.6083,
+      "step": 7859
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6564004556817478e-05,
+      "loss": 0.6336,
+      "step": 7860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.656318306454856e-05,
+      "loss": 0.6323,
+      "step": 7861
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6562361494464806e-05,
+      "loss": 0.6295,
+      "step": 7862
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6561539846575953e-05,
+      "loss": 0.5708,
+      "step": 7863
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6560718120891742e-05,
+      "loss": 0.6139,
+      "step": 7864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6559896317421924e-05,
+      "loss": 0.5982,
+      "step": 7865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6559074436176236e-05,
+      "loss": 0.6454,
+      "step": 7866
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.655825247716442e-05,
+      "loss": 0.6047,
+      "step": 7867
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6557430440396225e-05,
+      "loss": 0.6217,
+      "step": 7868
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6556608325881397e-05,
+      "loss": 0.5812,
+      "step": 7869
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6555786133629683e-05,
+      "loss": 0.6048,
+      "step": 7870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6554963863650834e-05,
+      "loss": 0.6626,
+      "step": 7871
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6554141515954592e-05,
+      "loss": 0.6731,
+      "step": 7872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6553319090550717e-05,
+      "loss": 0.6185,
+      "step": 7873
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.655249658744895e-05,
+      "loss": 0.647,
+      "step": 7874
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.655167400665905e-05,
+      "loss": 0.6475,
+      "step": 7875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6550851348190762e-05,
+      "loss": 0.6396,
+      "step": 7876
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.655002861205385e-05,
+      "loss": 0.6052,
+      "step": 7877
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.654920579825806e-05,
+      "loss": 0.693,
+      "step": 7878
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.654838290681315e-05,
+      "loss": 0.5504,
+      "step": 7879
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.654755993772888e-05,
+      "loss": 0.6233,
+      "step": 7880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6546736891014998e-05,
+      "loss": 0.6338,
+      "step": 7881
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6545913766681273e-05,
+      "loss": 0.6177,
+      "step": 7882
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6545090564737457e-05,
+      "loss": 0.629,
+      "step": 7883
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6544267285193317e-05,
+      "loss": 0.6147,
+      "step": 7884
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6543443928058607e-05,
+      "loss": 0.6221,
+      "step": 7885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6542620493343093e-05,
+      "loss": 0.607,
+      "step": 7886
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6541796981056535e-05,
+      "loss": 0.6246,
+      "step": 7887
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6540973391208698e-05,
+      "loss": 0.641,
+      "step": 7888
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.654014972380935e-05,
+      "loss": 0.6546,
+      "step": 7889
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6539325978868253e-05,
+      "loss": 0.6291,
+      "step": 7890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.653850215639517e-05,
+      "loss": 0.6041,
+      "step": 7891
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6537678256399877e-05,
+      "loss": 0.6542,
+      "step": 7892
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.653685427889214e-05,
+      "loss": 0.6243,
+      "step": 7893
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6536030223881722e-05,
+      "loss": 0.5855,
+      "step": 7894
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6535206091378396e-05,
+      "loss": 0.626,
+      "step": 7895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.653438188139194e-05,
+      "loss": 0.6278,
+      "step": 7896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.653355759393212e-05,
+      "loss": 0.658,
+      "step": 7897
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6532733229008705e-05,
+      "loss": 0.6138,
+      "step": 7898
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6531908786631477e-05,
+      "loss": 0.6011,
+      "step": 7899
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.653108426681021e-05,
+      "loss": 0.6283,
+      "step": 7900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6530259669554674e-05,
+      "loss": 0.6035,
+      "step": 7901
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6529434994874648e-05,
+      "loss": 0.6471,
+      "step": 7902
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.652861024277991e-05,
+      "loss": 0.6544,
+      "step": 7903
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6527785413280243e-05,
+      "loss": 0.6361,
+      "step": 7904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.652696050638542e-05,
+      "loss": 0.64,
+      "step": 7905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6526135522105223e-05,
+      "loss": 0.6237,
+      "step": 7906
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6525310460449434e-05,
+      "loss": 0.6452,
+      "step": 7907
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6524485321427834e-05,
+      "loss": 0.6615,
+      "step": 7908
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6523660105050208e-05,
+      "loss": 0.6474,
+      "step": 7909
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6522834811326338e-05,
+      "loss": 0.6503,
+      "step": 7910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.652200944026601e-05,
+      "loss": 0.5794,
+      "step": 7911
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.652118399187901e-05,
+      "loss": 0.6297,
+      "step": 7912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6520358466175127e-05,
+      "loss": 0.698,
+      "step": 7913
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6519532863164142e-05,
+      "loss": 0.6603,
+      "step": 7914
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.651870718285585e-05,
+      "loss": 0.6367,
+      "step": 7915
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6517881425260038e-05,
+      "loss": 0.6009,
+      "step": 7916
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6517055590386496e-05,
+      "loss": 0.6663,
+      "step": 7917
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6516229678245016e-05,
+      "loss": 0.5758,
+      "step": 7918
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.651540368884539e-05,
+      "loss": 0.6794,
+      "step": 7919
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6514577622197413e-05,
+      "loss": 0.6177,
+      "step": 7920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6513751478310873e-05,
+      "loss": 0.6127,
+      "step": 7921
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6512925257195572e-05,
+      "loss": 0.5998,
+      "step": 7922
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6512098958861305e-05,
+      "loss": 0.6112,
+      "step": 7923
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6511272583317864e-05,
+      "loss": 0.6855,
+      "step": 7924
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.651044613057505e-05,
+      "loss": 0.6003,
+      "step": 7925
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6509619600642664e-05,
+      "loss": 0.5903,
+      "step": 7926
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6508792993530502e-05,
+      "loss": 0.6443,
+      "step": 7927
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6507966309248363e-05,
+      "loss": 0.6382,
+      "step": 7928
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6507139547806053e-05,
+      "loss": 0.6247,
+      "step": 7929
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6506312709213373e-05,
+      "loss": 0.6137,
+      "step": 7930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.650548579348012e-05,
+      "loss": 0.6014,
+      "step": 7931
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6504658800616107e-05,
+      "loss": 0.6581,
+      "step": 7932
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.650383173063114e-05,
+      "loss": 0.6092,
+      "step": 7933
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6503004583535016e-05,
+      "loss": 0.6469,
+      "step": 7934
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6502177359337546e-05,
+      "loss": 0.6202,
+      "step": 7935
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6501350058048538e-05,
+      "loss": 0.6317,
+      "step": 7936
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.6500522679677803e-05,
+      "loss": 0.6248,
+      "step": 7937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6499695224235144e-05,
+      "loss": 0.646,
+      "step": 7938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6498867691730377e-05,
+      "loss": 0.6699,
+      "step": 7939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6498040082173315e-05,
+      "loss": 0.6395,
+      "step": 7940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6497212395573766e-05,
+      "loss": 0.646,
+      "step": 7941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6496384631941547e-05,
+      "loss": 0.6802,
+      "step": 7942
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.649555679128647e-05,
+      "loss": 0.6028,
+      "step": 7943
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.649472887361835e-05,
+      "loss": 0.6653,
+      "step": 7944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6493900878947002e-05,
+      "loss": 0.6289,
+      "step": 7945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6493072807282246e-05,
+      "loss": 0.6502,
+      "step": 7946
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6492244658633895e-05,
+      "loss": 0.6247,
+      "step": 7947
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6491416433011774e-05,
+      "loss": 0.6685,
+      "step": 7948
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.64905881304257e-05,
+      "loss": 0.5878,
+      "step": 7949
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6489759750885492e-05,
+      "loss": 0.6586,
+      "step": 7950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.648893129440097e-05,
+      "loss": 0.6386,
+      "step": 7951
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.648810276098196e-05,
+      "loss": 0.6332,
+      "step": 7952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6487274150638287e-05,
+      "loss": 0.6368,
+      "step": 7953
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6486445463379774e-05,
+      "loss": 0.651,
+      "step": 7954
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.648561669921624e-05,
+      "loss": 0.6504,
+      "step": 7955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6484787858157515e-05,
+      "loss": 0.6203,
+      "step": 7956
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.648395894021343e-05,
+      "loss": 0.6677,
+      "step": 7957
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.648312994539381e-05,
+      "loss": 0.619,
+      "step": 7958
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.648230087370848e-05,
+      "loss": 0.6378,
+      "step": 7959
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6481471725167277e-05,
+      "loss": 0.6519,
+      "step": 7960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.648064249978002e-05,
+      "loss": 0.625,
+      "step": 7961
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6479813197556557e-05,
+      "loss": 0.6508,
+      "step": 7962
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.647898381850671e-05,
+      "loss": 0.6144,
+      "step": 7963
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.647815436264031e-05,
+      "loss": 0.6874,
+      "step": 7964
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6477324829967196e-05,
+      "loss": 0.6561,
+      "step": 7965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6476495220497205e-05,
+      "loss": 0.6343,
+      "step": 7966
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6475665534240166e-05,
+      "loss": 0.6284,
+      "step": 7967
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6474835771205927e-05,
+      "loss": 0.6205,
+      "step": 7968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6474005931404312e-05,
+      "loss": 0.6276,
+      "step": 7969
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.647317601484517e-05,
+      "loss": 0.641,
+      "step": 7970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6472346021538338e-05,
+      "loss": 0.6405,
+      "step": 7971
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6471515951493656e-05,
+      "loss": 0.6308,
+      "step": 7972
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.647068580472097e-05,
+      "loss": 0.6501,
+      "step": 7973
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.646985558123011e-05,
+      "loss": 0.665,
+      "step": 7974
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6469025281030932e-05,
+      "loss": 0.5859,
+      "step": 7975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6468194904133278e-05,
+      "loss": 0.6746,
+      "step": 7976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6467364450546988e-05,
+      "loss": 0.619,
+      "step": 7977
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.646653392028191e-05,
+      "loss": 0.6483,
+      "step": 7978
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6465703313347894e-05,
+      "loss": 0.6663,
+      "step": 7979
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.646487262975479e-05,
+      "loss": 0.6478,
+      "step": 7980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.646404186951244e-05,
+      "loss": 0.6613,
+      "step": 7981
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6463211032630693e-05,
+      "loss": 0.6027,
+      "step": 7982
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6462380119119408e-05,
+      "loss": 0.6143,
+      "step": 7983
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.646154912898843e-05,
+      "loss": 0.6567,
+      "step": 7984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.646071806224761e-05,
+      "loss": 0.6534,
+      "step": 7985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.645988691890681e-05,
+      "loss": 0.6163,
+      "step": 7986
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6459055698975877e-05,
+      "loss": 0.6674,
+      "step": 7987
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6458224402464666e-05,
+      "loss": 0.5981,
+      "step": 7988
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6457393029383038e-05,
+      "loss": 0.647,
+      "step": 7989
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6456561579740843e-05,
+      "loss": 0.5749,
+      "step": 7990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6455730053547947e-05,
+      "loss": 0.6143,
+      "step": 7991
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6454898450814203e-05,
+      "loss": 0.6245,
+      "step": 7992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6454066771549473e-05,
+      "loss": 0.6115,
+      "step": 7993
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6453235015763615e-05,
+      "loss": 0.5967,
+      "step": 7994
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6452403183466493e-05,
+      "loss": 0.6433,
+      "step": 7995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.645157127466797e-05,
+      "loss": 0.6455,
+      "step": 7996
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.645073928937791e-05,
+      "loss": 0.6114,
+      "step": 7997
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.644990722760617e-05,
+      "loss": 0.5969,
+      "step": 7998
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6449075089362626e-05,
+      "loss": 0.625,
+      "step": 7999
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6448242874657136e-05,
+      "loss": 0.6566,
+      "step": 8000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6447410583499573e-05,
+      "loss": 0.5944,
+      "step": 8001
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6446578215899797e-05,
+      "loss": 0.6102,
+      "step": 8002
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.644574577186768e-05,
+      "loss": 0.6233,
+      "step": 8003
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.64449132514131e-05,
+      "loss": 0.6547,
+      "step": 8004
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6444080654545915e-05,
+      "loss": 0.6376,
+      "step": 8005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6443247981276005e-05,
+      "loss": 0.6446,
+      "step": 8006
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.644241523161324e-05,
+      "loss": 0.6292,
+      "step": 8007
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.644158240556749e-05,
+      "loss": 0.6598,
+      "step": 8008
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6440749503148634e-05,
+      "loss": 0.6112,
+      "step": 8009
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6439916524366543e-05,
+      "loss": 0.6426,
+      "step": 8010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6439083469231096e-05,
+      "loss": 0.5837,
+      "step": 8011
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6438250337752166e-05,
+      "loss": 0.6745,
+      "step": 8012
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6437417129939638e-05,
+      "loss": 0.6378,
+      "step": 8013
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6436583845803385e-05,
+      "loss": 0.6576,
+      "step": 8014
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.643575048535329e-05,
+      "loss": 0.6171,
+      "step": 8015
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.643491704859923e-05,
+      "loss": 0.6731,
+      "step": 8016
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6434083535551088e-05,
+      "loss": 0.6229,
+      "step": 8017
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6433249946218748e-05,
+      "loss": 0.6528,
+      "step": 8018
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6432416280612087e-05,
+      "loss": 0.6031,
+      "step": 8019
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6431582538741e-05,
+      "loss": 0.6512,
+      "step": 8020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6430748720615363e-05,
+      "loss": 0.6566,
+      "step": 8021
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6429914826245066e-05,
+      "loss": 0.5827,
+      "step": 8022
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6429080855639995e-05,
+      "loss": 0.658,
+      "step": 8023
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.642824680881004e-05,
+      "loss": 0.6372,
+      "step": 8024
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.642741268576508e-05,
+      "loss": 0.674,
+      "step": 8025
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.642657848651502e-05,
+      "loss": 0.6408,
+      "step": 8026
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6425744211069736e-05,
+      "loss": 0.6458,
+      "step": 8027
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6424909859439128e-05,
+      "loss": 0.6762,
+      "step": 8028
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6424075431633088e-05,
+      "loss": 0.6486,
+      "step": 8029
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6423240927661507e-05,
+      "loss": 0.6647,
+      "step": 8030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6422406347534278e-05,
+      "loss": 0.617,
+      "step": 8031
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6421571691261297e-05,
+      "loss": 0.6249,
+      "step": 8032
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.642073695885246e-05,
+      "loss": 0.6239,
+      "step": 8033
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6419902150317666e-05,
+      "loss": 0.6321,
+      "step": 8034
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6419067265666808e-05,
+      "loss": 0.5991,
+      "step": 8035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.641823230490979e-05,
+      "loss": 0.6039,
+      "step": 8036
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6417397268056507e-05,
+      "loss": 0.6359,
+      "step": 8037
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6416562155116863e-05,
+      "loss": 0.6382,
+      "step": 8038
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.641572696610076e-05,
+      "loss": 0.6272,
+      "step": 8039
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.641489170101809e-05,
+      "loss": 0.6732,
+      "step": 8040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.641405635987877e-05,
+      "loss": 0.6332,
+      "step": 8041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6413220942692697e-05,
+      "loss": 0.6668,
+      "step": 8042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6412385449469778e-05,
+      "loss": 0.69,
+      "step": 8043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6411549880219916e-05,
+      "loss": 0.6704,
+      "step": 8044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.641071423495302e-05,
+      "loss": 0.6307,
+      "step": 8045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6409878513678997e-05,
+      "loss": 0.655,
+      "step": 8046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6409042716407758e-05,
+      "loss": 0.659,
+      "step": 8047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6408206843149206e-05,
+      "loss": 0.6386,
+      "step": 8048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.640737089391326e-05,
+      "loss": 0.6765,
+      "step": 8049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6406534868709824e-05,
+      "loss": 0.63,
+      "step": 8050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.640569876754881e-05,
+      "loss": 0.6166,
+      "step": 8051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.640486259044014e-05,
+      "loss": 0.6767,
+      "step": 8052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6404026337393716e-05,
+      "loss": 0.6523,
+      "step": 8053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6403190008419462e-05,
+      "loss": 0.6328,
+      "step": 8054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.640235360352729e-05,
+      "loss": 0.6181,
+      "step": 8055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6401517122727117e-05,
+      "loss": 0.5848,
+      "step": 8056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6400680566028858e-05,
+      "loss": 0.599,
+      "step": 8057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.639984393344244e-05,
+      "loss": 0.6471,
+      "step": 8058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.639900722497777e-05,
+      "loss": 0.6742,
+      "step": 8059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6398170440644777e-05,
+      "loss": 0.6534,
+      "step": 8060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.639733358045338e-05,
+      "loss": 0.607,
+      "step": 8061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6396496644413502e-05,
+      "loss": 0.5887,
+      "step": 8062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6395659632535064e-05,
+      "loss": 0.6035,
+      "step": 8063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.639482254482799e-05,
+      "loss": 0.5801,
+      "step": 8064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6393985381302203e-05,
+      "loss": 0.6195,
+      "step": 8065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6393148141967633e-05,
+      "loss": 0.693,
+      "step": 8066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6392310826834206e-05,
+      "loss": 0.6161,
+      "step": 8067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6391473435911845e-05,
+      "loss": 0.6177,
+      "step": 8068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6390635969210485e-05,
+      "loss": 0.6563,
+      "step": 8069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6389798426740045e-05,
+      "loss": 0.6332,
+      "step": 8070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6388960808510467e-05,
+      "loss": 0.6309,
+      "step": 8071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6388123114531675e-05,
+      "loss": 0.5357,
+      "step": 8072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.63872853448136e-05,
+      "loss": 0.6483,
+      "step": 8073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.638644749936618e-05,
+      "loss": 0.6951,
+      "step": 8074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6385609578199346e-05,
+      "loss": 0.7481,
+      "step": 8075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6384771581323035e-05,
+      "loss": 0.6424,
+      "step": 8076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6383933508747173e-05,
+      "loss": 0.6279,
+      "step": 8077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.638309536048171e-05,
+      "loss": 0.6761,
+      "step": 8078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6382257136536574e-05,
+      "loss": 0.6329,
+      "step": 8079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6381418836921706e-05,
+      "loss": 0.6714,
+      "step": 8080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6380580461647046e-05,
+      "loss": 0.6604,
+      "step": 8081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6379742010722536e-05,
+      "loss": 0.5978,
+      "step": 8082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6378903484158112e-05,
+      "loss": 0.6504,
+      "step": 8083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6378064881963715e-05,
+      "loss": 0.6263,
+      "step": 8084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.637722620414929e-05,
+      "loss": 0.6673,
+      "step": 8085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6376387450724784e-05,
+      "loss": 0.6432,
+      "step": 8086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.637554862170014e-05,
+      "loss": 0.6625,
+      "step": 8087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.63747097170853e-05,
+      "loss": 0.6586,
+      "step": 8088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.637387073689021e-05,
+      "loss": 0.5652,
+      "step": 8089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6373031681124823e-05,
+      "loss": 0.6221,
+      "step": 8090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.637219254979908e-05,
+      "loss": 0.606,
+      "step": 8091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6371353342922937e-05,
+      "loss": 0.5977,
+      "step": 8092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.637051406050634e-05,
+      "loss": 0.6148,
+      "step": 8093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.636967470255924e-05,
+      "loss": 0.6038,
+      "step": 8094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6368835269091587e-05,
+      "loss": 0.595,
+      "step": 8095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6367995760113333e-05,
+      "loss": 0.6389,
+      "step": 8096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6367156175634438e-05,
+      "loss": 0.6401,
+      "step": 8097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6366316515664852e-05,
+      "loss": 0.6256,
+      "step": 8098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.636547678021453e-05,
+      "loss": 0.6419,
+      "step": 8099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6364636969293426e-05,
+      "loss": 0.5943,
+      "step": 8100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.63637970829115e-05,
+      "loss": 0.6838,
+      "step": 8101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.636295712107871e-05,
+      "loss": 0.5847,
+      "step": 8102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6362117083805017e-05,
+      "loss": 0.5692,
+      "step": 8103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6361276971100373e-05,
+      "loss": 0.6138,
+      "step": 8104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6360436782974748e-05,
+      "loss": 0.6284,
+      "step": 8105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6359596519438094e-05,
+      "loss": 0.609,
+      "step": 8106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6358756180500383e-05,
+      "loss": 0.6093,
+      "step": 8107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.635791576617157e-05,
+      "loss": 0.5721,
+      "step": 8108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6357075276461625e-05,
+      "loss": 0.6667,
+      "step": 8109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6356234711380513e-05,
+      "loss": 0.6835,
+      "step": 8110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6355394070938192e-05,
+      "loss": 0.6158,
+      "step": 8111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6354553355144637e-05,
+      "loss": 0.6131,
+      "step": 8112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6353712564009815e-05,
+      "loss": 0.6103,
+      "step": 8113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6352871697543692e-05,
+      "loss": 0.6127,
+      "step": 8114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6352030755756237e-05,
+      "loss": 0.6372,
+      "step": 8115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6351189738657426e-05,
+      "loss": 0.6116,
+      "step": 8116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6350348646257224e-05,
+      "loss": 0.6367,
+      "step": 8117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6349507478565608e-05,
+      "loss": 0.6094,
+      "step": 8118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6348666235592546e-05,
+      "loss": 0.634,
+      "step": 8119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6347824917348014e-05,
+      "loss": 0.6546,
+      "step": 8120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6346983523841988e-05,
+      "loss": 0.6382,
+      "step": 8121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6346142055084447e-05,
+      "loss": 0.6302,
+      "step": 8122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.634530051108536e-05,
+      "loss": 0.6917,
+      "step": 8123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6344458891854706e-05,
+      "loss": 0.6047,
+      "step": 8124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6343617197402472e-05,
+      "loss": 0.6168,
+      "step": 8125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.634277542773863e-05,
+      "loss": 0.6281,
+      "step": 8126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.634193358287316e-05,
+      "loss": 0.6512,
+      "step": 8127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6341091662816048e-05,
+      "loss": 0.6393,
+      "step": 8128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6340249667577268e-05,
+      "loss": 0.6632,
+      "step": 8129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.633940759716681e-05,
+      "loss": 0.6592,
+      "step": 8130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6338565451594654e-05,
+      "loss": 0.6596,
+      "step": 8131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.633772323087079e-05,
+      "loss": 0.6414,
+      "step": 8132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6336880935005198e-05,
+      "loss": 0.5967,
+      "step": 8133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6336038564007864e-05,
+      "loss": 0.6307,
+      "step": 8134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.633519611788878e-05,
+      "loss": 0.6052,
+      "step": 8135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.633435359665793e-05,
+      "loss": 0.622,
+      "step": 8136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.633351100032531e-05,
+      "loss": 0.6712,
+      "step": 8137
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6332668328900904e-05,
+      "loss": 0.6662,
+      "step": 8138
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6331825582394698e-05,
+      "loss": 0.6198,
+      "step": 8139
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6330982760816697e-05,
+      "loss": 0.597,
+      "step": 8140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6330139864176882e-05,
+      "loss": 0.6469,
+      "step": 8141
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6329296892485252e-05,
+      "loss": 0.6647,
+      "step": 8142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6328453845751803e-05,
+      "loss": 0.6141,
+      "step": 8143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6327610723986524e-05,
+      "loss": 0.6282,
+      "step": 8144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.632676752719942e-05,
+      "loss": 0.6089,
+      "step": 8145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.632592425540048e-05,
+      "loss": 0.5493,
+      "step": 8146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.632508090859971e-05,
+      "loss": 0.671,
+      "step": 8147
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.63242374868071e-05,
+      "loss": 0.6613,
+      "step": 8148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6323393990032656e-05,
+      "loss": 0.5778,
+      "step": 8149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.632255041828638e-05,
+      "loss": 0.6235,
+      "step": 8150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6321706771578266e-05,
+      "loss": 0.6955,
+      "step": 8151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6320863049918322e-05,
+      "loss": 0.6161,
+      "step": 8152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.632001925331655e-05,
+      "loss": 0.6184,
+      "step": 8153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.631917538178296e-05,
+      "loss": 0.6645,
+      "step": 8154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.631833143532755e-05,
+      "loss": 0.6574,
+      "step": 8155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6317487413960328e-05,
+      "loss": 0.6558,
+      "step": 8156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6316643317691298e-05,
+      "loss": 0.5963,
+      "step": 8157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.631579914653047e-05,
+      "loss": 0.6218,
+      "step": 8158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6314954900487858e-05,
+      "loss": 0.6447,
+      "step": 8159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6314110579573466e-05,
+      "loss": 0.6744,
+      "step": 8160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6313266183797305e-05,
+      "loss": 0.5793,
+      "step": 8161
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.631242171316939e-05,
+      "loss": 0.6586,
+      "step": 8162
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6311577167699725e-05,
+      "loss": 0.5962,
+      "step": 8163
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6310732547398332e-05,
+      "loss": 0.6702,
+      "step": 8164
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6309887852275223e-05,
+      "loss": 0.6715,
+      "step": 8165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.630904308234041e-05,
+      "loss": 0.5938,
+      "step": 8166
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6308198237603904e-05,
+      "loss": 0.614,
+      "step": 8167
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6307353318075736e-05,
+      "loss": 0.6132,
+      "step": 8168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6306508323765912e-05,
+      "loss": 0.6791,
+      "step": 8169
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6305663254684455e-05,
+      "loss": 0.681,
+      "step": 8170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6304818110841382e-05,
+      "loss": 0.6863,
+      "step": 8171
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6303972892246717e-05,
+      "loss": 0.5871,
+      "step": 8172
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.630312759891048e-05,
+      "loss": 0.6711,
+      "step": 8173
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6302282230842688e-05,
+      "loss": 0.7095,
+      "step": 8174
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.630143678805337e-05,
+      "loss": 0.7288,
+      "step": 8175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.630059127055255e-05,
+      "loss": 0.6837,
+      "step": 8176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6299745678350245e-05,
+      "loss": 0.5905,
+      "step": 8177
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6298900011456487e-05,
+      "loss": 0.6663,
+      "step": 8178
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6298054269881305e-05,
+      "loss": 0.645,
+      "step": 8179
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6297208453634718e-05,
+      "loss": 0.5918,
+      "step": 8180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6296362562726763e-05,
+      "loss": 0.6146,
+      "step": 8181
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.629551659716746e-05,
+      "loss": 0.6172,
+      "step": 8182
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6294670556966846e-05,
+      "loss": 0.6153,
+      "step": 8183
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.629382444213495e-05,
+      "loss": 0.6271,
+      "step": 8184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6292978252681805e-05,
+      "loss": 0.6781,
+      "step": 8185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.629213198861744e-05,
+      "loss": 0.6255,
+      "step": 8186
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.629128564995189e-05,
+      "loss": 0.607,
+      "step": 8187
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.629043923669519e-05,
+      "loss": 0.679,
+      "step": 8188
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.628959274885738e-05,
+      "loss": 0.6551,
+      "step": 8189
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6288746186448487e-05,
+      "loss": 0.6479,
+      "step": 8190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6287899549478554e-05,
+      "loss": 0.6219,
+      "step": 8191
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6287052837957617e-05,
+      "loss": 0.6528,
+      "step": 8192
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6286206051895716e-05,
+      "loss": 0.6181,
+      "step": 8193
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6285359191302892e-05,
+      "loss": 0.6683,
+      "step": 8194
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.628451225618918e-05,
+      "loss": 0.596,
+      "step": 8195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.628366524656463e-05,
+      "loss": 0.71,
+      "step": 8196
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6282818162439274e-05,
+      "loss": 0.6628,
+      "step": 8197
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6281971003823166e-05,
+      "loss": 0.6477,
+      "step": 8198
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6281123770726343e-05,
+      "loss": 0.6322,
+      "step": 8199
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6280276463158854e-05,
+      "loss": 0.6558,
+      "step": 8200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6279429081130738e-05,
+      "loss": 0.6153,
+      "step": 8201
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.627858162465205e-05,
+      "loss": 0.6831,
+      "step": 8202
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6277734093732833e-05,
+      "loss": 0.5911,
+      "step": 8203
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.627688648838314e-05,
+      "loss": 0.5865,
+      "step": 8204
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6276038808613015e-05,
+      "loss": 0.6659,
+      "step": 8205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.627519105443251e-05,
+      "loss": 0.6715,
+      "step": 8206
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6274343225851678e-05,
+      "loss": 0.6811,
+      "step": 8207
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.627349532288057e-05,
+      "loss": 0.6408,
+      "step": 8208
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6272647345529238e-05,
+      "loss": 0.5871,
+      "step": 8209
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6271799293807736e-05,
+      "loss": 0.6408,
+      "step": 8210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6270951167726123e-05,
+      "loss": 0.62,
+      "step": 8211
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.627010296729445e-05,
+      "loss": 0.6397,
+      "step": 8212
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6269254692522775e-05,
+      "loss": 0.6993,
+      "step": 8213
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.626840634342115e-05,
+      "loss": 0.6274,
+      "step": 8214
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6267557919999647e-05,
+      "loss": 0.67,
+      "step": 8215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.626670942226831e-05,
+      "loss": 0.6389,
+      "step": 8216
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6265860850237202e-05,
+      "loss": 0.5989,
+      "step": 8217
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6265012203916396e-05,
+      "loss": 0.5683,
+      "step": 8218
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.626416348331594e-05,
+      "loss": 0.6371,
+      "step": 8219
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.62633146884459e-05,
+      "loss": 0.6564,
+      "step": 8220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6262465819316345e-05,
+      "loss": 0.6231,
+      "step": 8221
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6261616875937337e-05,
+      "loss": 0.6622,
+      "step": 8222
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6260767858318934e-05,
+      "loss": 0.6184,
+      "step": 8223
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6259918766471216e-05,
+      "loss": 0.6559,
+      "step": 8224
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6259069600404234e-05,
+      "loss": 0.6088,
+      "step": 8225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.625822036012807e-05,
+      "loss": 0.607,
+      "step": 8226
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6257371045652784e-05,
+      "loss": 0.6245,
+      "step": 8227
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.625652165698845e-05,
+      "loss": 0.6457,
+      "step": 8228
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6255672194145136e-05,
+      "loss": 0.6252,
+      "step": 8229
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6254822657132917e-05,
+      "loss": 0.6222,
+      "step": 8230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.625397304596186e-05,
+      "loss": 0.6337,
+      "step": 8231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6253123360642043e-05,
+      "loss": 0.6531,
+      "step": 8232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6252273601183538e-05,
+      "loss": 0.6185,
+      "step": 8233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.625142376759642e-05,
+      "loss": 0.6432,
+      "step": 8234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6250573859890768e-05,
+      "loss": 0.5832,
+      "step": 8235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.624972387807665e-05,
+      "loss": 0.6229,
+      "step": 8236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6248873822164154e-05,
+      "loss": 0.5955,
+      "step": 8237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.624802369216335e-05,
+      "loss": 0.6419,
+      "step": 8238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6247173488084327e-05,
+      "loss": 0.6439,
+      "step": 8239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6246323209937157e-05,
+      "loss": 0.6447,
+      "step": 8240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6245472857731924e-05,
+      "loss": 0.6102,
+      "step": 8241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6244622431478706e-05,
+      "loss": 0.6074,
+      "step": 8242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6243771931187594e-05,
+      "loss": 0.6463,
+      "step": 8243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.624292135686867e-05,
+      "loss": 0.617,
+      "step": 8244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.624207070853201e-05,
+      "loss": 0.6578,
+      "step": 8245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6241219986187706e-05,
+      "loss": 0.5577,
+      "step": 8246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6240369189845847e-05,
+      "loss": 0.621,
+      "step": 8247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6239518319516513e-05,
+      "loss": 0.6058,
+      "step": 8248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.62386673752098e-05,
+      "loss": 0.5958,
+      "step": 8249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.623781635693579e-05,
+      "loss": 0.6554,
+      "step": 8250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6236965264704577e-05,
+      "loss": 0.6143,
+      "step": 8251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6236114098526256e-05,
+      "loss": 0.6122,
+      "step": 8252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.623526285841091e-05,
+      "loss": 0.6759,
+      "step": 8253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6234411544368632e-05,
+      "loss": 0.6408,
+      "step": 8254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6233560156409523e-05,
+      "loss": 0.6582,
+      "step": 8255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6232708694543672e-05,
+      "loss": 0.6105,
+      "step": 8256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6231857158781173e-05,
+      "loss": 0.6563,
+      "step": 8257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.623100554913213e-05,
+      "loss": 0.6428,
+      "step": 8258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.623015386560663e-05,
+      "loss": 0.6245,
+      "step": 8259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.622930210821477e-05,
+      "loss": 0.6143,
+      "step": 8260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6228450276966657e-05,
+      "loss": 0.6807,
+      "step": 8261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6227598371872392e-05,
+      "loss": 0.6306,
+      "step": 8262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6226746392942065e-05,
+      "loss": 0.6256,
+      "step": 8263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6225894340185782e-05,
+      "loss": 0.6505,
+      "step": 8264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6225042213613645e-05,
+      "loss": 0.6198,
+      "step": 8265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6224190013235763e-05,
+      "loss": 0.65,
+      "step": 8266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.622333773906223e-05,
+      "loss": 0.7053,
+      "step": 8267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.622248539110316e-05,
+      "loss": 0.6433,
+      "step": 8268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.622163296936865e-05,
+      "loss": 0.6244,
+      "step": 8269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6220780473868812e-05,
+      "loss": 0.6631,
+      "step": 8270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6219927904613753e-05,
+      "loss": 0.6333,
+      "step": 8271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.621907526161358e-05,
+      "loss": 0.6711,
+      "step": 8272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6218222544878402e-05,
+      "loss": 0.621,
+      "step": 8273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.621736975441833e-05,
+      "loss": 0.6296,
+      "step": 8274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6216516890243476e-05,
+      "loss": 0.6538,
+      "step": 8275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.621566395236395e-05,
+      "loss": 0.6755,
+      "step": 8276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6214810940789862e-05,
+      "loss": 0.5693,
+      "step": 8277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6213957855531334e-05,
+      "loss": 0.669,
+      "step": 8278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.621310469659847e-05,
+      "loss": 0.5954,
+      "step": 8279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6212251464001392e-05,
+      "loss": 0.5718,
+      "step": 8280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6211398157750213e-05,
+      "loss": 0.5975,
+      "step": 8281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6210544777855052e-05,
+      "loss": 0.6284,
+      "step": 8282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.620969132432603e-05,
+      "loss": 0.6308,
+      "step": 8283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6208837797173255e-05,
+      "loss": 0.6407,
+      "step": 8284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.620798419640686e-05,
+      "loss": 0.6456,
+      "step": 8285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6207130522036954e-05,
+      "loss": 0.6151,
+      "step": 8286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.620627677407367e-05,
+      "loss": 0.5913,
+      "step": 8287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.620542295252712e-05,
+      "loss": 0.66,
+      "step": 8288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.620456905740743e-05,
+      "loss": 0.616,
+      "step": 8289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.620371508872473e-05,
+      "loss": 0.71,
+      "step": 8290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6202861046489133e-05,
+      "loss": 0.6259,
+      "step": 8291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6202006930710775e-05,
+      "loss": 0.6497,
+      "step": 8292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.620115274139978e-05,
+      "loss": 0.605,
+      "step": 8293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.620029847856627e-05,
+      "loss": 0.6556,
+      "step": 8294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6199444142220382e-05,
+      "loss": 0.5994,
+      "step": 8295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6198589732372243e-05,
+      "loss": 0.6477,
+      "step": 8296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.619773524903198e-05,
+      "loss": 0.6831,
+      "step": 8297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6196880692209723e-05,
+      "loss": 0.6739,
+      "step": 8298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.619602606191561e-05,
+      "loss": 0.5811,
+      "step": 8299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6195171358159768e-05,
+      "loss": 0.5595,
+      "step": 8300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6194316580952335e-05,
+      "loss": 0.6556,
+      "step": 8301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6193461730303442e-05,
+      "loss": 0.6456,
+      "step": 8302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6192606806223223e-05,
+      "loss": 0.6279,
+      "step": 8303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6191751808721815e-05,
+      "loss": 0.6427,
+      "step": 8304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6190896737809363e-05,
+      "loss": 0.6261,
+      "step": 8305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6190041593495998e-05,
+      "loss": 0.6424,
+      "step": 8306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6189186375791855e-05,
+      "loss": 0.6725,
+      "step": 8307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.618833108470708e-05,
+      "loss": 0.6418,
+      "step": 8308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6187475720251812e-05,
+      "loss": 0.6635,
+      "step": 8309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.618662028243619e-05,
+      "loss": 0.6664,
+      "step": 8310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.618576477127036e-05,
+      "loss": 0.6533,
+      "step": 8311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6184909186764466e-05,
+      "loss": 0.6782,
+      "step": 8312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6184053528928648e-05,
+      "loss": 0.6866,
+      "step": 8313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.618319779777305e-05,
+      "loss": 0.6734,
+      "step": 8314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6182341993307824e-05,
+      "loss": 0.668,
+      "step": 8315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.618148611554311e-05,
+      "loss": 0.6288,
+      "step": 8316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6180630164489063e-05,
+      "loss": 0.6247,
+      "step": 8317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6179774140155824e-05,
+      "loss": 0.5951,
+      "step": 8318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.617891804255354e-05,
+      "loss": 0.6301,
+      "step": 8319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6178061871692374e-05,
+      "loss": 0.6889,
+      "step": 8320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6177205627582467e-05,
+      "loss": 0.6315,
+      "step": 8321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6176349310233972e-05,
+      "loss": 0.6185,
+      "step": 8322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6175492919657047e-05,
+      "loss": 0.6645,
+      "step": 8323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6174636455861837e-05,
+      "loss": 0.5885,
+      "step": 8324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.61737799188585e-05,
+      "loss": 0.6144,
+      "step": 8325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6172923308657195e-05,
+      "loss": 0.6207,
+      "step": 8326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6172066625268074e-05,
+      "loss": 0.6642,
+      "step": 8327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6171209868701297e-05,
+      "loss": 0.6125,
+      "step": 8328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6170353038967017e-05,
+      "loss": 0.6316,
+      "step": 8329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6169496136075396e-05,
+      "loss": 0.658,
+      "step": 8330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6168639160036595e-05,
+      "loss": 0.6788,
+      "step": 8331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6167782110860775e-05,
+      "loss": 0.5938,
+      "step": 8332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6166924988558097e-05,
+      "loss": 0.6189,
+      "step": 8333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.616606779313872e-05,
+      "loss": 0.6447,
+      "step": 8334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.616521052461281e-05,
+      "loss": 0.6113,
+      "step": 8335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.616435318299053e-05,
+      "loss": 0.6751,
+      "step": 8336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6163495768282046e-05,
+      "loss": 0.6131,
+      "step": 8337
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.616263828049752e-05,
+      "loss": 0.6102,
+      "step": 8338
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.616178071964712e-05,
+      "loss": 0.6669,
+      "step": 8339
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.616092308574102e-05,
+      "loss": 0.632,
+      "step": 8340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.616006537878938e-05,
+      "loss": 0.6449,
+      "step": 8341
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6159207598802375e-05,
+      "loss": 0.6921,
+      "step": 8342
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.615834974579017e-05,
+      "loss": 0.6544,
+      "step": 8343
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.615749181976294e-05,
+      "loss": 0.6406,
+      "step": 8344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6156633820730854e-05,
+      "loss": 0.6402,
+      "step": 8345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6155775748704088e-05,
+      "loss": 0.6167,
+      "step": 8346
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.615491760369281e-05,
+      "loss": 0.6301,
+      "step": 8347
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6154059385707203e-05,
+      "loss": 0.6058,
+      "step": 8348
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6153201094757433e-05,
+      "loss": 0.6144,
+      "step": 8349
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6152342730853677e-05,
+      "loss": 0.6807,
+      "step": 8350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.615148429400612e-05,
+      "loss": 0.5992,
+      "step": 8351
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6150625784224935e-05,
+      "loss": 0.6073,
+      "step": 8352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6149767201520297e-05,
+      "loss": 0.5958,
+      "step": 8353
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6148908545902396e-05,
+      "loss": 0.6226,
+      "step": 8354
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6148049817381403e-05,
+      "loss": 0.6486,
+      "step": 8355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6147191015967498e-05,
+      "loss": 0.6541,
+      "step": 8356
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6146332141670874e-05,
+      "loss": 0.6543,
+      "step": 8357
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.61454731945017e-05,
+      "loss": 0.6031,
+      "step": 8358
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6144614174470172e-05,
+      "loss": 0.6029,
+      "step": 8359
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6143755081586467e-05,
+      "loss": 0.6308,
+      "step": 8360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.614289591586078e-05,
+      "loss": 0.6094,
+      "step": 8361
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6142036677303283e-05,
+      "loss": 0.6498,
+      "step": 8362
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6141177365924177e-05,
+      "loss": 0.638,
+      "step": 8363
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.614031798173364e-05,
+      "loss": 0.6368,
+      "step": 8364
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.613945852474187e-05,
+      "loss": 0.649,
+      "step": 8365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.613859899495905e-05,
+      "loss": 0.6454,
+      "step": 8366
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6137739392395375e-05,
+      "loss": 0.6224,
+      "step": 8367
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6136879717061034e-05,
+      "loss": 0.5716,
+      "step": 8368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6136019968966225e-05,
+      "loss": 0.6236,
+      "step": 8369
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.613516014812113e-05,
+      "loss": 0.6942,
+      "step": 8370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6134300254535958e-05,
+      "loss": 0.598,
+      "step": 8371
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6133440288220893e-05,
+      "loss": 0.6812,
+      "step": 8372
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6132580249186133e-05,
+      "loss": 0.5564,
+      "step": 8373
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6131720137441878e-05,
+      "loss": 0.6742,
+      "step": 8374
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6130859952998324e-05,
+      "loss": 0.6243,
+      "step": 8375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.612999969586567e-05,
+      "loss": 0.6111,
+      "step": 8376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6129139366054115e-05,
+      "loss": 0.6265,
+      "step": 8377
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6128278963573858e-05,
+      "loss": 0.6676,
+      "step": 8378
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.61274184884351e-05,
+      "loss": 0.6519,
+      "step": 8379
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.612655794064805e-05,
+      "loss": 0.6416,
+      "step": 8380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6125697320222902e-05,
+      "loss": 0.6415,
+      "step": 8381
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6124836627169866e-05,
+      "loss": 0.6477,
+      "step": 8382
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.612397586149914e-05,
+      "loss": 0.6094,
+      "step": 8383
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6123115023220933e-05,
+      "loss": 0.6227,
+      "step": 8384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6122254112345457e-05,
+      "loss": 0.6336,
+      "step": 8385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.612139312888291e-05,
+      "loss": 0.6291,
+      "step": 8386
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6120532072843505e-05,
+      "loss": 0.6688,
+      "step": 8387
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6119670944237448e-05,
+      "loss": 0.6314,
+      "step": 8388
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6118809743074954e-05,
+      "loss": 0.6103,
+      "step": 8389
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6117948469366226e-05,
+      "loss": 0.6504,
+      "step": 8390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6117087123121483e-05,
+      "loss": 0.6423,
+      "step": 8391
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6116225704350933e-05,
+      "loss": 0.6472,
+      "step": 8392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6115364213064788e-05,
+      "loss": 0.656,
+      "step": 8393
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6114502649273263e-05,
+      "loss": 0.6088,
+      "step": 8394
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.611364101298658e-05,
+      "loss": 0.5975,
+      "step": 8395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6112779304214944e-05,
+      "loss": 0.5862,
+      "step": 8396
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.611191752296858e-05,
+      "loss": 0.6508,
+      "step": 8397
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6111055669257702e-05,
+      "loss": 0.6312,
+      "step": 8398
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.611019374309253e-05,
+      "loss": 0.6655,
+      "step": 8399
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6109331744483278e-05,
+      "loss": 0.6267,
+      "step": 8400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.610846967344017e-05,
+      "loss": 0.5996,
+      "step": 8401
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.610760752997343e-05,
+      "loss": 0.6052,
+      "step": 8402
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6106745314093274e-05,
+      "loss": 0.6528,
+      "step": 8403
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6105883025809925e-05,
+      "loss": 0.6144,
+      "step": 8404
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.610502066513361e-05,
+      "loss": 0.6449,
+      "step": 8405
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6104158232074555e-05,
+      "loss": 0.6166,
+      "step": 8406
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.610329572664298e-05,
+      "loss": 0.6202,
+      "step": 8407
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6102433148849117e-05,
+      "loss": 0.6128,
+      "step": 8408
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6101570498703184e-05,
+      "loss": 0.6698,
+      "step": 8409
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.610070777621542e-05,
+      "loss": 0.6275,
+      "step": 8410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6099844981396044e-05,
+      "loss": 0.636,
+      "step": 8411
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.609898211425529e-05,
+      "loss": 0.6345,
+      "step": 8412
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6098119174803385e-05,
+      "loss": 0.6075,
+      "step": 8413
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6097256163050568e-05,
+      "loss": 0.6459,
+      "step": 8414
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6096393079007064e-05,
+      "loss": 0.6596,
+      "step": 8415
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6095529922683108e-05,
+      "loss": 0.5957,
+      "step": 8416
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6094666694088935e-05,
+      "loss": 0.6858,
+      "step": 8417
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6093803393234778e-05,
+      "loss": 0.5846,
+      "step": 8418
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6092940020130872e-05,
+      "loss": 0.6153,
+      "step": 8419
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6092076574787453e-05,
+      "loss": 0.7032,
+      "step": 8420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6091213057214766e-05,
+      "loss": 0.5707,
+      "step": 8421
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6090349467423038e-05,
+      "loss": 0.6473,
+      "step": 8422
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6089485805422513e-05,
+      "loss": 0.627,
+      "step": 8423
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6088622071223433e-05,
+      "loss": 0.6092,
+      "step": 8424
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6087758264836032e-05,
+      "loss": 0.5783,
+      "step": 8425
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6086894386270558e-05,
+      "loss": 0.6421,
+      "step": 8426
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.608603043553725e-05,
+      "loss": 0.6207,
+      "step": 8427
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.608516641264635e-05,
+      "loss": 0.6003,
+      "step": 8428
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.608430231760811e-05,
+      "loss": 0.5959,
+      "step": 8429
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6083438150432767e-05,
+      "loss": 0.655,
+      "step": 8430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6082573911130567e-05,
+      "loss": 0.655,
+      "step": 8431
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6081709599711765e-05,
+      "loss": 0.6236,
+      "step": 8432
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.60808452161866e-05,
+      "loss": 0.6568,
+      "step": 8433
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6079980760565317e-05,
+      "loss": 0.6106,
+      "step": 8434
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6079116232858173e-05,
+      "loss": 0.6334,
+      "step": 8435
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6078251633075417e-05,
+      "loss": 0.6641,
+      "step": 8436
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.60773869612273e-05,
+      "loss": 0.6867,
+      "step": 8437
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6076522217324075e-05,
+      "loss": 0.6155,
+      "step": 8438
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.607565740137599e-05,
+      "loss": 0.6202,
+      "step": 8439
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.60747925133933e-05,
+      "loss": 0.6393,
+      "step": 8440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6073927553386258e-05,
+      "loss": 0.6736,
+      "step": 8441
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6073062521365126e-05,
+      "loss": 0.6268,
+      "step": 8442
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.607219741734015e-05,
+      "loss": 0.6114,
+      "step": 8443
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6071332241321595e-05,
+      "loss": 0.6904,
+      "step": 8444
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6070466993319717e-05,
+      "loss": 0.6587,
+      "step": 8445
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.606960167334477e-05,
+      "loss": 0.6383,
+      "step": 8446
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.606873628140702e-05,
+      "loss": 0.6195,
+      "step": 8447
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.606787081751673e-05,
+      "loss": 0.6349,
+      "step": 8448
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6067005281684145e-05,
+      "loss": 0.6915,
+      "step": 8449
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6066139673919544e-05,
+      "loss": 0.6387,
+      "step": 8450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6065273994233185e-05,
+      "loss": 0.6451,
+      "step": 8451
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6064408242635325e-05,
+      "loss": 0.6732,
+      "step": 8452
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6063542419136237e-05,
+      "loss": 0.6031,
+      "step": 8453
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.606267652374618e-05,
+      "loss": 0.6114,
+      "step": 8454
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6061810556475425e-05,
+      "loss": 0.6244,
+      "step": 8455
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6060944517334237e-05,
+      "loss": 0.5949,
+      "step": 8456
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6060078406332888e-05,
+      "loss": 0.6989,
+      "step": 8457
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.605921222348164e-05,
+      "loss": 0.6243,
+      "step": 8458
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6058345968790767e-05,
+      "loss": 0.6172,
+      "step": 8459
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6057479642270538e-05,
+      "loss": 0.5963,
+      "step": 8460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.605661324393123e-05,
+      "loss": 0.631,
+      "step": 8461
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6055746773783102e-05,
+      "loss": 0.6607,
+      "step": 8462
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6054880231836444e-05,
+      "loss": 0.6065,
+      "step": 8463
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6054013618101515e-05,
+      "loss": 0.6292,
+      "step": 8464
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6053146932588597e-05,
+      "loss": 0.6129,
+      "step": 8465
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6052280175307967e-05,
+      "loss": 0.6217,
+      "step": 8466
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.60514133462699e-05,
+      "loss": 0.6098,
+      "step": 8467
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.605054644548467e-05,
+      "loss": 0.6564,
+      "step": 8468
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.604967947296256e-05,
+      "loss": 0.6022,
+      "step": 8469
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6048812428713848e-05,
+      "loss": 0.6144,
+      "step": 8470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.604794531274881e-05,
+      "loss": 0.6713,
+      "step": 8471
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.604707812507773e-05,
+      "loss": 0.6304,
+      "step": 8472
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.604621086571089e-05,
+      "loss": 0.6301,
+      "step": 8473
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.604534353465857e-05,
+      "loss": 0.5976,
+      "step": 8474
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.604447613193106e-05,
+      "loss": 0.6292,
+      "step": 8475
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6043608657538638e-05,
+      "loss": 0.6138,
+      "step": 8476
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6042741111491583e-05,
+      "loss": 0.6475,
+      "step": 8477
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6041873493800194e-05,
+      "loss": 0.644,
+      "step": 8478
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.604100580447475e-05,
+      "loss": 0.6122,
+      "step": 8479
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6040138043525543e-05,
+      "loss": 0.6361,
+      "step": 8480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6039270210962855e-05,
+      "loss": 0.628,
+      "step": 8481
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.603840230679698e-05,
+      "loss": 0.6848,
+      "step": 8482
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.603753433103821e-05,
+      "loss": 0.6256,
+      "step": 8483
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6036666283696832e-05,
+      "loss": 0.663,
+      "step": 8484
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6035798164783132e-05,
+      "loss": 0.6775,
+      "step": 8485
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6034929974307417e-05,
+      "loss": 0.5744,
+      "step": 8486
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6034061712279973e-05,
+      "loss": 0.6789,
+      "step": 8487
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.603319337871109e-05,
+      "loss": 0.5675,
+      "step": 8488
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6032324973611072e-05,
+      "loss": 0.6305,
+      "step": 8489
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6031456496990206e-05,
+      "loss": 0.6686,
+      "step": 8490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6030587948858796e-05,
+      "loss": 0.6364,
+      "step": 8491
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6029719329227136e-05,
+      "loss": 0.6537,
+      "step": 8492
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6028850638105527e-05,
+      "loss": 0.5896,
+      "step": 8493
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6027981875504267e-05,
+      "loss": 0.633,
+      "step": 8494
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6027113041433657e-05,
+      "loss": 0.6763,
+      "step": 8495
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6026244135903993e-05,
+      "loss": 0.6118,
+      "step": 8496
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6025375158925586e-05,
+      "loss": 0.6585,
+      "step": 8497
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.602450611050873e-05,
+      "loss": 0.6542,
+      "step": 8498
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6023636990663737e-05,
+      "loss": 0.6252,
+      "step": 8499
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6022767799400908e-05,
+      "loss": 0.6174,
+      "step": 8500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6021898536730546e-05,
+      "loss": 0.6147,
+      "step": 8501
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.602102920266296e-05,
+      "loss": 0.6626,
+      "step": 8502
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6020159797208457e-05,
+      "loss": 0.6421,
+      "step": 8503
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6019290320377344e-05,
+      "loss": 0.6215,
+      "step": 8504
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6018420772179928e-05,
+      "loss": 0.6444,
+      "step": 8505
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.601755115262652e-05,
+      "loss": 0.6452,
+      "step": 8506
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6016681461727434e-05,
+      "loss": 0.6855,
+      "step": 8507
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6015811699492974e-05,
+      "loss": 0.6099,
+      "step": 8508
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.601494186593346e-05,
+      "loss": 0.6086,
+      "step": 8509
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.60140719610592e-05,
+      "loss": 0.6399,
+      "step": 8510
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.601320198488051e-05,
+      "loss": 0.618,
+      "step": 8511
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.60123319374077e-05,
+      "loss": 0.6569,
+      "step": 8512
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6011461818651093e-05,
+      "loss": 0.6306,
+      "step": 8513
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6010591628621003e-05,
+      "loss": 0.6703,
+      "step": 8514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.600972136732774e-05,
+      "loss": 0.6712,
+      "step": 8515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.600885103478163e-05,
+      "loss": 0.5944,
+      "step": 8516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6007980630992995e-05,
+      "loss": 0.5941,
+      "step": 8517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6007110155972144e-05,
+      "loss": 0.6588,
+      "step": 8518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6006239609729402e-05,
+      "loss": 0.6818,
+      "step": 8519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6005368992275093e-05,
+      "loss": 0.6255,
+      "step": 8520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.600449830361954e-05,
+      "loss": 0.6112,
+      "step": 8521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6003627543773062e-05,
+      "loss": 0.6146,
+      "step": 8522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6002756712745985e-05,
+      "loss": 0.6184,
+      "step": 8523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6001885810548635e-05,
+      "loss": 0.6405,
+      "step": 8524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6001014837191336e-05,
+      "loss": 0.5897,
+      "step": 8525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6000143792684414e-05,
+      "loss": 0.6467,
+      "step": 8526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5999272677038196e-05,
+      "loss": 0.6132,
+      "step": 8527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.599840149026301e-05,
+      "loss": 0.6617,
+      "step": 8528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.599753023236919e-05,
+      "loss": 0.6143,
+      "step": 8529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5996658903367063e-05,
+      "loss": 0.6864,
+      "step": 8530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5995787503266958e-05,
+      "loss": 0.63,
+      "step": 8531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5994916032079203e-05,
+      "loss": 0.6505,
+      "step": 8532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.599404448981414e-05,
+      "loss": 0.623,
+      "step": 8533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5993172876482095e-05,
+      "loss": 0.6317,
+      "step": 8534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5992301192093403e-05,
+      "loss": 0.6146,
+      "step": 8535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.59914294366584e-05,
+      "loss": 0.582,
+      "step": 8536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5990557610187426e-05,
+      "loss": 0.6596,
+      "step": 8537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.598968571269081e-05,
+      "loss": 0.6601,
+      "step": 8538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5988813744178893e-05,
+      "loss": 0.5259,
+      "step": 8539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.598794170466201e-05,
+      "loss": 0.6102,
+      "step": 8540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5987069594150512e-05,
+      "loss": 0.6778,
+      "step": 8541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5986197412654724e-05,
+      "loss": 0.6166,
+      "step": 8542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5985325160184995e-05,
+      "loss": 0.632,
+      "step": 8543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5984452836751666e-05,
+      "loss": 0.5737,
+      "step": 8544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5983580442365077e-05,
+      "loss": 0.6429,
+      "step": 8545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.598270797703557e-05,
+      "loss": 0.5876,
+      "step": 8546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.598183544077349e-05,
+      "loss": 0.6827,
+      "step": 8547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.598096283358919e-05,
+      "loss": 0.6173,
+      "step": 8548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5980090155493005e-05,
+      "loss": 0.693,
+      "step": 8549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.597921740649529e-05,
+      "loss": 0.6117,
+      "step": 8550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.597834458660639e-05,
+      "loss": 0.6648,
+      "step": 8551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5977471695836646e-05,
+      "loss": 0.6623,
+      "step": 8552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.597659873419642e-05,
+      "loss": 0.6843,
+      "step": 8553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5975725701696055e-05,
+      "loss": 0.6097,
+      "step": 8554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5974852598345897e-05,
+      "loss": 0.6347,
+      "step": 8555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5973979424156304e-05,
+      "loss": 0.6365,
+      "step": 8556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5973106179137634e-05,
+      "loss": 0.6571,
+      "step": 8557
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.597223286330023e-05,
+      "loss": 0.6471,
+      "step": 8558
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.597135947665445e-05,
+      "loss": 0.6337,
+      "step": 8559
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.597048601921065e-05,
+      "loss": 0.6187,
+      "step": 8560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.596961249097919e-05,
+      "loss": 0.5679,
+      "step": 8561
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5968738891970416e-05,
+      "loss": 0.6155,
+      "step": 8562
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5967865222194693e-05,
+      "loss": 0.5902,
+      "step": 8563
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.596699148166238e-05,
+      "loss": 0.6357,
+      "step": 8564
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5966117670383832e-05,
+      "loss": 0.6512,
+      "step": 8565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.596524378836942e-05,
+      "loss": 0.5978,
+      "step": 8566
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5964369835629486e-05,
+      "loss": 0.6965,
+      "step": 8567
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.596349581217441e-05,
+      "loss": 0.5754,
+      "step": 8568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5962621718014545e-05,
+      "loss": 0.683,
+      "step": 8569
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5961747553160255e-05,
+      "loss": 0.5958,
+      "step": 8570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.596087331762191e-05,
+      "loss": 0.6585,
+      "step": 8571
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.595999901140987e-05,
+      "loss": 0.6791,
+      "step": 8572
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.59591246345345e-05,
+      "loss": 0.6666,
+      "step": 8573
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.595825018700617e-05,
+      "loss": 0.6281,
+      "step": 8574
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.595737566883525e-05,
+      "loss": 0.6231,
+      "step": 8575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.59565010800321e-05,
+      "loss": 0.646,
+      "step": 8576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5955626420607096e-05,
+      "loss": 0.6236,
+      "step": 8577
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.595475169057061e-05,
+      "loss": 0.6115,
+      "step": 8578
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.595387688993301e-05,
+      "loss": 0.6647,
+      "step": 8579
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5953002018704666e-05,
+      "loss": 0.6317,
+      "step": 8580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.595212707689595e-05,
+      "loss": 0.6766,
+      "step": 8581
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5951252064517242e-05,
+      "loss": 0.6041,
+      "step": 8582
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5950376981578913e-05,
+      "loss": 0.6954,
+      "step": 8583
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5949501828091335e-05,
+      "loss": 0.6139,
+      "step": 8584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5948626604064887e-05,
+      "loss": 0.613,
+      "step": 8585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5947751309509945e-05,
+      "loss": 0.6989,
+      "step": 8586
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5946875944436887e-05,
+      "loss": 0.5928,
+      "step": 8587
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5946000508856093e-05,
+      "loss": 0.6554,
+      "step": 8588
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5945125002777942e-05,
+      "loss": 0.6468,
+      "step": 8589
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5944249426212813e-05,
+      "loss": 0.6109,
+      "step": 8590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5943373779171087e-05,
+      "loss": 0.5914,
+      "step": 8591
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5942498061663148e-05,
+      "loss": 0.6252,
+      "step": 8592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5941622273699375e-05,
+      "loss": 0.619,
+      "step": 8593
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5940746415290155e-05,
+      "loss": 0.6376,
+      "step": 8594
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.593987048644587e-05,
+      "loss": 0.6099,
+      "step": 8595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5938994487176906e-05,
+      "loss": 0.594,
+      "step": 8596
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.593811841749365e-05,
+      "loss": 0.6313,
+      "step": 8597
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.593724227740649e-05,
+      "loss": 0.5589,
+      "step": 8598
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.593636606692581e-05,
+      "loss": 0.6197,
+      "step": 8599
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5935489786062005e-05,
+      "loss": 0.6529,
+      "step": 8600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5934613434825455e-05,
+      "loss": 0.617,
+      "step": 8601
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.593373701322656e-05,
+      "loss": 0.6696,
+      "step": 8602
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.59328605212757e-05,
+      "loss": 0.607,
+      "step": 8603
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5931983958983275e-05,
+      "loss": 0.698,
+      "step": 8604
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5931107326359684e-05,
+      "loss": 0.6519,
+      "step": 8605
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5930230623415305e-05,
+      "loss": 0.6118,
+      "step": 8606
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5929353850160544e-05,
+      "loss": 0.6438,
+      "step": 8607
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.592847700660579e-05,
+      "loss": 0.6106,
+      "step": 8608
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.592760009276144e-05,
+      "loss": 0.6024,
+      "step": 8609
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5926723108637895e-05,
+      "loss": 0.6005,
+      "step": 8610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.592584605424555e-05,
+      "loss": 0.7332,
+      "step": 8611
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5924968929594804e-05,
+      "loss": 0.6257,
+      "step": 8612
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5924091734696057e-05,
+      "loss": 0.6208,
+      "step": 8613
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5923214469559706e-05,
+      "loss": 0.6376,
+      "step": 8614
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5922337134196158e-05,
+      "loss": 0.6303,
+      "step": 8615
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5921459728615808e-05,
+      "loss": 0.6366,
+      "step": 8616
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5920582252829064e-05,
+      "loss": 0.6524,
+      "step": 8617
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.591970470684633e-05,
+      "loss": 0.643,
+      "step": 8618
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5918827090678006e-05,
+      "loss": 0.6509,
+      "step": 8619
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5917949404334497e-05,
+      "loss": 0.5954,
+      "step": 8620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5917071647826215e-05,
+      "loss": 0.5931,
+      "step": 8621
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.591619382116356e-05,
+      "loss": 0.6175,
+      "step": 8622
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5915315924356946e-05,
+      "loss": 0.6329,
+      "step": 8623
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.591443795741678e-05,
+      "loss": 0.6298,
+      "step": 8624
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.591355992035347e-05,
+      "loss": 0.6398,
+      "step": 8625
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5912681813177422e-05,
+      "loss": 0.6196,
+      "step": 8626
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5911803635899056e-05,
+      "loss": 0.6198,
+      "step": 8627
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5910925388528777e-05,
+      "loss": 0.615,
+      "step": 8628
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5910047071076998e-05,
+      "loss": 0.6401,
+      "step": 8629
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.590916868355414e-05,
+      "loss": 0.613,
+      "step": 8630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.590829022597061e-05,
+      "loss": 0.5991,
+      "step": 8631
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5907411698336828e-05,
+      "loss": 0.6193,
+      "step": 8632
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5906533100663203e-05,
+      "loss": 0.6428,
+      "step": 8633
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.590565443296016e-05,
+      "loss": 0.5701,
+      "step": 8634
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.590477569523811e-05,
+      "loss": 0.6651,
+      "step": 8635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5903896887507476e-05,
+      "loss": 0.6232,
+      "step": 8636
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5903018009778676e-05,
+      "loss": 0.5938,
+      "step": 8637
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5902139062062126e-05,
+      "loss": 0.6866,
+      "step": 8638
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5901260044368257e-05,
+      "loss": 0.6288,
+      "step": 8639
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.590038095670748e-05,
+      "loss": 0.6545,
+      "step": 8640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.589950179909023e-05,
+      "loss": 0.6483,
+      "step": 8641
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5898622571526917e-05,
+      "loss": 0.5837,
+      "step": 8642
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5897743274027972e-05,
+      "loss": 0.6561,
+      "step": 8643
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5896863906603825e-05,
+      "loss": 0.7057,
+      "step": 8644
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.589598446926489e-05,
+      "loss": 0.6458,
+      "step": 8645
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5895104962021607e-05,
+      "loss": 0.6277,
+      "step": 8646
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.589422538488439e-05,
+      "loss": 0.6143,
+      "step": 8647
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.589334573786368e-05,
+      "loss": 0.5908,
+      "step": 8648
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.58924660209699e-05,
+      "loss": 0.664,
+      "step": 8649
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5891586234213484e-05,
+      "loss": 0.6761,
+      "step": 8650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5890706377604855e-05,
+      "loss": 0.6244,
+      "step": 8651
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5889826451154453e-05,
+      "loss": 0.6746,
+      "step": 8652
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5888946454872707e-05,
+      "loss": 0.6106,
+      "step": 8653
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5888066388770052e-05,
+      "loss": 0.63,
+      "step": 8654
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5887186252856924e-05,
+      "loss": 0.5461,
+      "step": 8655
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.588630604714375e-05,
+      "loss": 0.6755,
+      "step": 8656
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5885425771640974e-05,
+      "loss": 0.6881,
+      "step": 8657
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.588454542635903e-05,
+      "loss": 0.6613,
+      "step": 8658
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5883665011308357e-05,
+      "loss": 0.6069,
+      "step": 8659
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5882784526499393e-05,
+      "loss": 0.6373,
+      "step": 8660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5881903971942574e-05,
+      "loss": 0.652,
+      "step": 8661
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5881023347648348e-05,
+      "loss": 0.6753,
+      "step": 8662
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.588014265362715e-05,
+      "loss": 0.7159,
+      "step": 8663
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.587926188988942e-05,
+      "loss": 0.6025,
+      "step": 8664
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.58783810564456e-05,
+      "loss": 0.6146,
+      "step": 8665
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.587750015330614e-05,
+      "loss": 0.6301,
+      "step": 8666
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.587661918048148e-05,
+      "loss": 0.643,
+      "step": 8667
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5875738137982065e-05,
+      "loss": 0.6219,
+      "step": 8668
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5874857025818342e-05,
+      "loss": 0.6265,
+      "step": 8669
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5873975844000758e-05,
+      "loss": 0.6598,
+      "step": 8670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5873094592539756e-05,
+      "loss": 0.6316,
+      "step": 8671
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5872213271445795e-05,
+      "loss": 0.6395,
+      "step": 8672
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5871331880729314e-05,
+      "loss": 0.629,
+      "step": 8673
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.587045042040076e-05,
+      "loss": 0.6711,
+      "step": 8674
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.58695688904706e-05,
+      "loss": 0.6624,
+      "step": 8675
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.586868729094927e-05,
+      "loss": 0.6413,
+      "step": 8676
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.586780562184723e-05,
+      "loss": 0.6206,
+      "step": 8677
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.586692388317493e-05,
+      "loss": 0.6452,
+      "step": 8678
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5866042074942827e-05,
+      "loss": 0.5941,
+      "step": 8679
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5865160197161378e-05,
+      "loss": 0.6719,
+      "step": 8680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.586427824984103e-05,
+      "loss": 0.6269,
+      "step": 8681
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5863396232992245e-05,
+      "loss": 0.6844,
+      "step": 8682
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5862514146625483e-05,
+      "loss": 0.6197,
+      "step": 8683
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.58616319907512e-05,
+      "loss": 0.6275,
+      "step": 8684
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.586074976537985e-05,
+      "loss": 0.6116,
+      "step": 8685
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5859867470521906e-05,
+      "loss": 0.6438,
+      "step": 8686
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5858985106187815e-05,
+      "loss": 0.636,
+      "step": 8687
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5858102672388044e-05,
+      "loss": 0.617,
+      "step": 8688
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5857220169133055e-05,
+      "loss": 0.6658,
+      "step": 8689
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5856337596433314e-05,
+      "loss": 0.6577,
+      "step": 8690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.585545495429928e-05,
+      "loss": 0.6874,
+      "step": 8691
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5854572242741424e-05,
+      "loss": 0.6086,
+      "step": 8692
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5853689461770206e-05,
+      "loss": 0.642,
+      "step": 8693
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.585280661139609e-05,
+      "loss": 0.5947,
+      "step": 8694
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5851923691629553e-05,
+      "loss": 0.691,
+      "step": 8695
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5851040702481056e-05,
+      "loss": 0.6144,
+      "step": 8696
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.585015764396107e-05,
+      "loss": 0.6448,
+      "step": 8697
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5849274516080068e-05,
+      "loss": 0.636,
+      "step": 8698
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.584839131884851e-05,
+      "loss": 0.5914,
+      "step": 8699
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5847508052276884e-05,
+      "loss": 0.609,
+      "step": 8700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5846624716375646e-05,
+      "loss": 0.6179,
+      "step": 8701
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5845741311155278e-05,
+      "loss": 0.638,
+      "step": 8702
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.584485783662625e-05,
+      "loss": 0.6384,
+      "step": 8703
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.584397429279904e-05,
+      "loss": 0.5932,
+      "step": 8704
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5843090679684123e-05,
+      "loss": 0.5766,
+      "step": 8705
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5842206997291976e-05,
+      "loss": 0.6848,
+      "step": 8706
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.584132324563307e-05,
+      "loss": 0.5924,
+      "step": 8707
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.584043942471789e-05,
+      "loss": 0.6284,
+      "step": 8708
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5839555534556913e-05,
+      "loss": 0.6147,
+      "step": 8709
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5838671575160618e-05,
+      "loss": 0.5985,
+      "step": 8710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5837787546539485e-05,
+      "loss": 0.6344,
+      "step": 8711
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5836903448703997e-05,
+      "loss": 0.619,
+      "step": 8712
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5836019281664632e-05,
+      "loss": 0.6738,
+      "step": 8713
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.583513504543188e-05,
+      "loss": 0.6401,
+      "step": 8714
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5834250740016215e-05,
+      "loss": 0.6963,
+      "step": 8715
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.583336636542813e-05,
+      "loss": 0.6185,
+      "step": 8716
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5832481921678108e-05,
+      "loss": 0.5791,
+      "step": 8717
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5831597408776634e-05,
+      "loss": 0.6872,
+      "step": 8718
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5830712826734193e-05,
+      "loss": 0.6593,
+      "step": 8719
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5829828175561282e-05,
+      "loss": 0.668,
+      "step": 8720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.582894345526838e-05,
+      "loss": 0.7414,
+      "step": 8721
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5828058665865976e-05,
+      "loss": 0.6154,
+      "step": 8722
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5827173807364567e-05,
+      "loss": 0.6456,
+      "step": 8723
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5826288879774645e-05,
+      "loss": 0.6198,
+      "step": 8724
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5825403883106692e-05,
+      "loss": 0.6948,
+      "step": 8725
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5824518817371212e-05,
+      "loss": 0.6138,
+      "step": 8726
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.582363368257869e-05,
+      "loss": 0.5739,
+      "step": 8727
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.582274847873962e-05,
+      "loss": 0.5841,
+      "step": 8728
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.582186320586451e-05,
+      "loss": 0.5634,
+      "step": 8729
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5820977863963842e-05,
+      "loss": 0.587,
+      "step": 8730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.582009245304812e-05,
+      "loss": 0.6274,
+      "step": 8731
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5819206973127837e-05,
+      "loss": 0.6916,
+      "step": 8732
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5818321424213495e-05,
+      "loss": 0.6705,
+      "step": 8733
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5817435806315593e-05,
+      "loss": 0.5973,
+      "step": 8734
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.581655011944463e-05,
+      "loss": 0.5865,
+      "step": 8735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5815664363611105e-05,
+      "loss": 0.6643,
+      "step": 8736
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5814778538825525e-05,
+      "loss": 0.6083,
+      "step": 8737
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5813892645098388e-05,
+      "loss": 0.6288,
+      "step": 8738
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.58130066824402e-05,
+      "loss": 0.5992,
+      "step": 8739
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5812120650861464e-05,
+      "loss": 0.6888,
+      "step": 8740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5811234550372683e-05,
+      "loss": 0.646,
+      "step": 8741
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.581034838098437e-05,
+      "loss": 0.6387,
+      "step": 8742
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5809462142707024e-05,
+      "loss": 0.5663,
+      "step": 8743
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5808575835551153e-05,
+      "loss": 0.6733,
+      "step": 8744
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5807689459527274e-05,
+      "loss": 0.6237,
+      "step": 8745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5806803014645885e-05,
+      "loss": 0.6311,
+      "step": 8746
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.58059165009175e-05,
+      "loss": 0.5718,
+      "step": 8747
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.580502991835263e-05,
+      "loss": 0.6615,
+      "step": 8748
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.580414326696179e-05,
+      "loss": 0.6613,
+      "step": 8749
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5803256546755486e-05,
+      "loss": 0.5902,
+      "step": 8750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.580236975774424e-05,
+      "loss": 0.5513,
+      "step": 8751
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5801482899938556e-05,
+      "loss": 0.6556,
+      "step": 8752
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5800595973348954e-05,
+      "loss": 0.6468,
+      "step": 8753
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.579970897798595e-05,
+      "loss": 0.6156,
+      "step": 8754
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5798821913860058e-05,
+      "loss": 0.6162,
+      "step": 8755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.57979347809818e-05,
+      "loss": 0.6559,
+      "step": 8756
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5797047579361686e-05,
+      "loss": 0.6418,
+      "step": 8757
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5796160309010244e-05,
+      "loss": 0.6323,
+      "step": 8758
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5795272969937985e-05,
+      "loss": 0.6762,
+      "step": 8759
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5794385562155436e-05,
+      "loss": 0.604,
+      "step": 8760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5793498085673115e-05,
+      "loss": 0.6536,
+      "step": 8761
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.579261054050155e-05,
+      "loss": 0.6435,
+      "step": 8762
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5791722926651253e-05,
+      "loss": 0.6369,
+      "step": 8763
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5790835244132758e-05,
+      "loss": 0.6605,
+      "step": 8764
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.578994749295659e-05,
+      "loss": 0.6888,
+      "step": 8765
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.578905967313326e-05,
+      "loss": 0.6403,
+      "step": 8766
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.578817178467331e-05,
+      "loss": 0.6039,
+      "step": 8767
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.578728382758726e-05,
+      "loss": 0.677,
+      "step": 8768
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.578639580188564e-05,
+      "loss": 0.6724,
+      "step": 8769
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.578550770757898e-05,
+      "loss": 0.6353,
+      "step": 8770
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5784619544677804e-05,
+      "loss": 0.691,
+      "step": 8771
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5783731313192646e-05,
+      "loss": 0.6791,
+      "step": 8772
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.578284301313404e-05,
+      "loss": 0.6295,
+      "step": 8773
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5781954644512513e-05,
+      "loss": 0.6811,
+      "step": 8774
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.57810662073386e-05,
+      "loss": 0.6333,
+      "step": 8775
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.578017770162283e-05,
+      "loss": 0.6775,
+      "step": 8776
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5779289127375744e-05,
+      "loss": 0.6088,
+      "step": 8777
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5778400484607876e-05,
+      "loss": 0.68,
+      "step": 8778
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.577751177332976e-05,
+      "loss": 0.6197,
+      "step": 8779
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5776622993551935e-05,
+      "loss": 0.619,
+      "step": 8780
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5775734145284933e-05,
+      "loss": 0.5962,
+      "step": 8781
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5774845228539298e-05,
+      "loss": 0.628,
+      "step": 8782
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.577395624332557e-05,
+      "loss": 0.6278,
+      "step": 8783
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.577306718965428e-05,
+      "loss": 0.6022,
+      "step": 8784
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5772178067535984e-05,
+      "loss": 0.5669,
+      "step": 8785
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.577128887698121e-05,
+      "loss": 0.6849,
+      "step": 8786
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5770399618000508e-05,
+      "loss": 0.6074,
+      "step": 8787
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5769510290604417e-05,
+      "loss": 0.6703,
+      "step": 8788
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5768620894803483e-05,
+      "loss": 0.5874,
+      "step": 8789
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5767731430608253e-05,
+      "loss": 0.6427,
+      "step": 8790
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.576684189802927e-05,
+      "loss": 0.7125,
+      "step": 8791
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.576595229707708e-05,
+      "loss": 0.6505,
+      "step": 8792
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.576506262776223e-05,
+      "loss": 0.6272,
+      "step": 8793
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5764172890095273e-05,
+      "loss": 0.6346,
+      "step": 8794
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5763283084086755e-05,
+      "loss": 0.6675,
+      "step": 8795
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5762393209747226e-05,
+      "loss": 0.6318,
+      "step": 8796
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.576150326708723e-05,
+      "loss": 0.6546,
+      "step": 8797
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.576061325611733e-05,
+      "loss": 0.5844,
+      "step": 8798
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5759723176848073e-05,
+      "loss": 0.5911,
+      "step": 8799
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.575883302929001e-05,
+      "loss": 0.6118,
+      "step": 8800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5757942813453693e-05,
+      "loss": 0.6215,
+      "step": 8801
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.5757052529349687e-05,
+      "loss": 0.6754,
+      "step": 8802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5756162176988533e-05,
+      "loss": 0.6408,
+      "step": 8803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.57552717563808e-05,
+      "loss": 0.6272,
+      "step": 8804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.575438126753704e-05,
+      "loss": 0.6272,
+      "step": 8805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5753490710467807e-05,
+      "loss": 0.6325,
+      "step": 8806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5752600085183666e-05,
+      "loss": 0.6382,
+      "step": 8807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5751709391695175e-05,
+      "loss": 0.6404,
+      "step": 8808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5750818630012892e-05,
+      "loss": 0.6909,
+      "step": 8809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.574992780014738e-05,
+      "loss": 0.595,
+      "step": 8810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.57490369021092e-05,
+      "loss": 0.6623,
+      "step": 8811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.574814593590891e-05,
+      "loss": 0.6069,
+      "step": 8812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5747254901557087e-05,
+      "loss": 0.6081,
+      "step": 8813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.574636379906428e-05,
+      "loss": 0.6867,
+      "step": 8814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5745472628441066e-05,
+      "loss": 0.6511,
+      "step": 8815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5744581389698005e-05,
+      "loss": 0.647,
+      "step": 8816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5743690082845662e-05,
+      "loss": 0.6221,
+      "step": 8817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5742798707894608e-05,
+      "loss": 0.6469,
+      "step": 8818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.574190726485541e-05,
+      "loss": 0.6167,
+      "step": 8819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.574101575373864e-05,
+      "loss": 0.6373,
+      "step": 8820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5740124174554863e-05,
+      "loss": 0.6337,
+      "step": 8821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5739232527314657e-05,
+      "loss": 0.5832,
+      "step": 8822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5738340812028586e-05,
+      "loss": 0.6087,
+      "step": 8823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5737449028707224e-05,
+      "loss": 0.6079,
+      "step": 8824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.573655717736115e-05,
+      "loss": 0.6545,
+      "step": 8825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5735665258000933e-05,
+      "loss": 0.6502,
+      "step": 8826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.573477327063715e-05,
+      "loss": 0.601,
+      "step": 8827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5733881215280373e-05,
+      "loss": 0.6633,
+      "step": 8828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5732989091941178e-05,
+      "loss": 0.6072,
+      "step": 8829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.573209690063015e-05,
+      "loss": 0.6105,
+      "step": 8830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5731204641357857e-05,
+      "loss": 0.6476,
+      "step": 8831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5730312314134888e-05,
+      "loss": 0.6675,
+      "step": 8832
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5729419918971815e-05,
+      "loss": 0.6042,
+      "step": 8833
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5728527455879225e-05,
+      "loss": 0.6166,
+      "step": 8834
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.572763492486769e-05,
+      "loss": 0.6422,
+      "step": 8835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.57267423259478e-05,
+      "loss": 0.5889,
+      "step": 8836
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.572584965913013e-05,
+      "loss": 0.632,
+      "step": 8837
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.572495692442528e-05,
+      "loss": 0.6383,
+      "step": 8838
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5724064121843815e-05,
+      "loss": 0.6455,
+      "step": 8839
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5723171251396332e-05,
+      "loss": 0.6279,
+      "step": 8840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5722278313093413e-05,
+      "loss": 0.5854,
+      "step": 8841
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5721385306945645e-05,
+      "loss": 0.6143,
+      "step": 8842
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.572049223296362e-05,
+      "loss": 0.6113,
+      "step": 8843
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.571959909115792e-05,
+      "loss": 0.6469,
+      "step": 8844
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.571870588153914e-05,
+      "loss": 0.5998,
+      "step": 8845
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5717812604117867e-05,
+      "loss": 0.6691,
+      "step": 8846
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.571691925890469e-05,
+      "loss": 0.6825,
+      "step": 8847
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5716025845910207e-05,
+      "loss": 0.633,
+      "step": 8848
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5715132365145e-05,
+      "loss": 0.6573,
+      "step": 8849
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5714238816619677e-05,
+      "loss": 0.6178,
+      "step": 8850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.571334520034482e-05,
+      "loss": 0.6145,
+      "step": 8851
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.571245151633103e-05,
+      "loss": 0.6371,
+      "step": 8852
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.57115577645889e-05,
+      "loss": 0.6464,
+      "step": 8853
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5710663945129026e-05,
+      "loss": 0.6533,
+      "step": 8854
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5709770057962012e-05,
+      "loss": 0.6443,
+      "step": 8855
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5708876103098444e-05,
+      "loss": 0.6006,
+      "step": 8856
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5707982080548934e-05,
+      "loss": 0.5768,
+      "step": 8857
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5707087990324073e-05,
+      "loss": 0.5783,
+      "step": 8858
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5706193832434463e-05,
+      "loss": 0.6239,
+      "step": 8859
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5705299606890712e-05,
+      "loss": 0.7271,
+      "step": 8860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.570440531370341e-05,
+      "loss": 0.6131,
+      "step": 8861
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5703510952883172e-05,
+      "loss": 0.6724,
+      "step": 8862
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5702616524440596e-05,
+      "loss": 0.6381,
+      "step": 8863
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5701722028386286e-05,
+      "loss": 0.6803,
+      "step": 8864
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.570082746473085e-05,
+      "loss": 0.6018,
+      "step": 8865
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.569993283348489e-05,
+      "loss": 0.5792,
+      "step": 8866
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.569903813465902e-05,
+      "loss": 0.6568,
+      "step": 8867
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.569814336826384e-05,
+      "loss": 0.5923,
+      "step": 8868
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5697248534309963e-05,
+      "loss": 0.6365,
+      "step": 8869
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5696353632808e-05,
+      "loss": 0.6496,
+      "step": 8870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5695458663768554e-05,
+      "loss": 0.644,
+      "step": 8871
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5694563627202245e-05,
+      "loss": 0.6524,
+      "step": 8872
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.569366852311968e-05,
+      "loss": 0.6332,
+      "step": 8873
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.569277335153147e-05,
+      "loss": 0.6805,
+      "step": 8874
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5691878112448233e-05,
+      "loss": 0.5999,
+      "step": 8875
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5690982805880577e-05,
+      "loss": 0.6708,
+      "step": 8876
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5690087431839126e-05,
+      "loss": 0.5845,
+      "step": 8877
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5689191990334487e-05,
+      "loss": 0.6219,
+      "step": 8878
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.568829648137728e-05,
+      "loss": 0.6225,
+      "step": 8879
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5687400904978124e-05,
+      "loss": 0.5981,
+      "step": 8880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5686505261147634e-05,
+      "loss": 0.6825,
+      "step": 8881
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.568560954989643e-05,
+      "loss": 0.6751,
+      "step": 8882
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5684713771235135e-05,
+      "loss": 0.6208,
+      "step": 8883
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5683817925174366e-05,
+      "loss": 0.6593,
+      "step": 8884
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5682922011724747e-05,
+      "loss": 0.6601,
+      "step": 8885
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5682026030896897e-05,
+      "loss": 0.5804,
+      "step": 8886
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.568112998270144e-05,
+      "loss": 0.6305,
+      "step": 8887
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5680233867149e-05,
+      "loss": 0.6878,
+      "step": 8888
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5679337684250205e-05,
+      "loss": 0.6912,
+      "step": 8889
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.567844143401567e-05,
+      "loss": 0.6429,
+      "step": 8890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5677545116456036e-05,
+      "loss": 0.6548,
+      "step": 8891
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5676648731581922e-05,
+      "loss": 0.6318,
+      "step": 8892
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5675752279403954e-05,
+      "loss": 0.6024,
+      "step": 8893
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5674855759932764e-05,
+      "loss": 0.6368,
+      "step": 8894
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.567395917317898e-05,
+      "loss": 0.6384,
+      "step": 8895
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5673062519153233e-05,
+      "loss": 0.5812,
+      "step": 8896
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5672165797866154e-05,
+      "loss": 0.6393,
+      "step": 8897
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5671269009328373e-05,
+      "loss": 0.5753,
+      "step": 8898
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.567037215355052e-05,
+      "loss": 0.5944,
+      "step": 8899
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.566947523054324e-05,
+      "loss": 0.6254,
+      "step": 8900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5668578240317153e-05,
+      "loss": 0.6614,
+      "step": 8901
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.56676811828829e-05,
+      "loss": 0.6209,
+      "step": 8902
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.566678405825112e-05,
+      "loss": 0.6352,
+      "step": 8903
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5665886866432445e-05,
+      "loss": 0.6514,
+      "step": 8904
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5664989607437515e-05,
+      "loss": 0.6315,
+      "step": 8905
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5664092281276964e-05,
+      "loss": 0.6231,
+      "step": 8906
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5663194887961437e-05,
+      "loss": 0.6118,
+      "step": 8907
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.566229742750157e-05,
+      "loss": 0.6374,
+      "step": 8908
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5661399899908003e-05,
+      "loss": 0.6627,
+      "step": 8909
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.566050230519138e-05,
+      "loss": 0.6455,
+      "step": 8910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.565960464336234e-05,
+      "loss": 0.5618,
+      "step": 8911
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.565870691443153e-05,
+      "loss": 0.679,
+      "step": 8912
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.565780911840959e-05,
+      "loss": 0.6177,
+      "step": 8913
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5656911255307164e-05,
+      "loss": 0.6907,
+      "step": 8914
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5656013325134897e-05,
+      "loss": 0.6334,
+      "step": 8915
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.565511532790344e-05,
+      "loss": 0.6411,
+      "step": 8916
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.565421726362344e-05,
+      "loss": 0.5995,
+      "step": 8917
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.565331913230554e-05,
+      "loss": 0.6506,
+      "step": 8918
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5652420933960392e-05,
+      "loss": 0.633,
+      "step": 8919
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.565152266859864e-05,
+      "loss": 0.6835,
+      "step": 8920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5650624336230936e-05,
+      "loss": 0.6199,
+      "step": 8921
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5649725936867936e-05,
+      "loss": 0.6083,
+      "step": 8922
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5648827470520286e-05,
+      "loss": 0.5663,
+      "step": 8923
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5647928937198642e-05,
+      "loss": 0.5544,
+      "step": 8924
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5647030336913657e-05,
+      "loss": 0.6479,
+      "step": 8925
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5646131669675982e-05,
+      "loss": 0.6049,
+      "step": 8926
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5645232935496274e-05,
+      "loss": 0.6449,
+      "step": 8927
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5644334134385187e-05,
+      "loss": 0.6602,
+      "step": 8928
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.564343526635338e-05,
+      "loss": 0.5799,
+      "step": 8929
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5642536331411507e-05,
+      "loss": 0.6002,
+      "step": 8930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.564163732957023e-05,
+      "loss": 0.5949,
+      "step": 8931
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5640738260840206e-05,
+      "loss": 0.6793,
+      "step": 8932
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5639839125232096e-05,
+      "loss": 0.5545,
+      "step": 8933
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5638939922756555e-05,
+      "loss": 0.6307,
+      "step": 8934
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.563804065342425e-05,
+      "loss": 0.6568,
+      "step": 8935
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5637141317245838e-05,
+      "loss": 0.614,
+      "step": 8936
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5636241914231988e-05,
+      "loss": 0.6413,
+      "step": 8937
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.563534244439336e-05,
+      "loss": 0.6472,
+      "step": 8938
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5634442907740617e-05,
+      "loss": 0.5852,
+      "step": 8939
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5633543304284426e-05,
+      "loss": 0.6204,
+      "step": 8940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5632643634035455e-05,
+      "loss": 0.6374,
+      "step": 8941
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5631743897004367e-05,
+      "loss": 0.6492,
+      "step": 8942
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5630844093201827e-05,
+      "loss": 0.6314,
+      "step": 8943
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5629944222638512e-05,
+      "loss": 0.6363,
+      "step": 8944
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5629044285325084e-05,
+      "loss": 0.6415,
+      "step": 8945
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.562814428127222e-05,
+      "loss": 0.6458,
+      "step": 8946
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.562724421049058e-05,
+      "loss": 0.648,
+      "step": 8947
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5626344072990842e-05,
+      "loss": 0.5763,
+      "step": 8948
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.562544386878368e-05,
+      "loss": 0.5904,
+      "step": 8949
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.562454359787976e-05,
+      "loss": 0.6567,
+      "step": 8950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.562364326028977e-05,
+      "loss": 0.6079,
+      "step": 8951
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5622742856024368e-05,
+      "loss": 0.6309,
+      "step": 8952
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5621842385094237e-05,
+      "loss": 0.6191,
+      "step": 8953
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5620941847510054e-05,
+      "loss": 0.6293,
+      "step": 8954
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5620041243282494e-05,
+      "loss": 0.6605,
+      "step": 8955
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5619140572422237e-05,
+      "loss": 0.6446,
+      "step": 8956
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5618239834939962e-05,
+      "loss": 0.6179,
+      "step": 8957
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5617339030846345e-05,
+      "loss": 0.5993,
+      "step": 8958
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.561643816015207e-05,
+      "loss": 0.5899,
+      "step": 8959
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.561553722286781e-05,
+      "loss": 0.5926,
+      "step": 8960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5614636219004254e-05,
+      "loss": 0.6315,
+      "step": 8961
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5613735148572088e-05,
+      "loss": 0.5757,
+      "step": 8962
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5612834011581987e-05,
+      "loss": 0.5722,
+      "step": 8963
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5611932808044637e-05,
+      "loss": 0.6541,
+      "step": 8964
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5611031537970728e-05,
+      "loss": 0.6613,
+      "step": 8965
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.561013020137094e-05,
+      "loss": 0.6097,
+      "step": 8966
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5609228798255958e-05,
+      "loss": 0.6276,
+      "step": 8967
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.560832732863648e-05,
+      "loss": 0.6603,
+      "step": 8968
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5607425792523183e-05,
+      "loss": 0.654,
+      "step": 8969
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5606524189926758e-05,
+      "loss": 0.5973,
+      "step": 8970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5605622520857902e-05,
+      "loss": 0.6316,
+      "step": 8971
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5604720785327296e-05,
+      "loss": 0.6524,
+      "step": 8972
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5603818983345635e-05,
+      "loss": 0.658,
+      "step": 8973
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5602917114923613e-05,
+      "loss": 0.6661,
+      "step": 8974
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.560201518007192e-05,
+      "loss": 0.6117,
+      "step": 8975
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.560111317880125e-05,
+      "loss": 0.6355,
+      "step": 8976
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5600211111122297e-05,
+      "loss": 0.6879,
+      "step": 8977
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.559930897704576e-05,
+      "loss": 0.5768,
+      "step": 8978
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5598406776582326e-05,
+      "loss": 0.66,
+      "step": 8979
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.55975045097427e-05,
+      "loss": 0.6819,
+      "step": 8980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.559660217653758e-05,
+      "loss": 0.6336,
+      "step": 8981
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5595699776977662e-05,
+      "loss": 0.6161,
+      "step": 8982
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5594797311073643e-05,
+      "loss": 0.6323,
+      "step": 8983
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5593894778836224e-05,
+      "loss": 0.6817,
+      "step": 8984
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.559299218027611e-05,
+      "loss": 0.6605,
+      "step": 8985
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5592089515403993e-05,
+      "loss": 0.6552,
+      "step": 8986
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.559118678423058e-05,
+      "loss": 0.6244,
+      "step": 8987
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5590283986766582e-05,
+      "loss": 0.6201,
+      "step": 8988
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5589381123022687e-05,
+      "loss": 0.65,
+      "step": 8989
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5588478193009615e-05,
+      "loss": 0.6646,
+      "step": 8990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.558757519673806e-05,
+      "loss": 0.653,
+      "step": 8991
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5586672134218735e-05,
+      "loss": 0.6515,
+      "step": 8992
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5585769005462344e-05,
+      "loss": 0.6735,
+      "step": 8993
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5584865810479594e-05,
+      "loss": 0.5845,
+      "step": 8994
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5583962549281196e-05,
+      "loss": 0.6413,
+      "step": 8995
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5583059221877863e-05,
+      "loss": 0.6474,
+      "step": 8996
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5582155828280293e-05,
+      "loss": 0.5978,
+      "step": 8997
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.558125236849921e-05,
+      "loss": 0.6067,
+      "step": 8998
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5580348842545314e-05,
+      "loss": 0.6326,
+      "step": 8999
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5579445250429328e-05,
+      "loss": 0.6415,
+      "step": 9000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.557854159216196e-05,
+      "loss": 0.6146,
+      "step": 9001
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.557763786775392e-05,
+      "loss": 0.5799,
+      "step": 9002
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5576734077215933e-05,
+      "loss": 0.6095,
+      "step": 9003
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5575830220558702e-05,
+      "loss": 0.6138,
+      "step": 9004
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.557492629779296e-05,
+      "loss": 0.6461,
+      "step": 9005
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5574022308929406e-05,
+      "loss": 0.6589,
+      "step": 9006
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.557311825397877e-05,
+      "loss": 0.6264,
+      "step": 9007
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5572214132951767e-05,
+      "loss": 0.6638,
+      "step": 9008
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5571309945859116e-05,
+      "loss": 0.6347,
+      "step": 9009
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.557040569271154e-05,
+      "loss": 0.5865,
+      "step": 9010
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5569501373519754e-05,
+      "loss": 0.6467,
+      "step": 9011
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5568596988294486e-05,
+      "loss": 0.6476,
+      "step": 9012
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.556769253704646e-05,
+      "loss": 0.7084,
+      "step": 9013
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.556678801978639e-05,
+      "loss": 0.6249,
+      "step": 9014
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5565883436525005e-05,
+      "loss": 0.6765,
+      "step": 9015
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.556497878727304e-05,
+      "loss": 0.6444,
+      "step": 9016
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5564074072041204e-05,
+      "loss": 0.6426,
+      "step": 9017
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.556316929084023e-05,
+      "loss": 0.6479,
+      "step": 9018
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5562264443680852e-05,
+      "loss": 0.6413,
+      "step": 9019
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.556135953057379e-05,
+      "loss": 0.604,
+      "step": 9020
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5560454551529775e-05,
+      "loss": 0.572,
+      "step": 9021
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.555954950655954e-05,
+      "loss": 0.6181,
+      "step": 9022
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.555864439567381e-05,
+      "loss": 0.6478,
+      "step": 9023
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5557739218883317e-05,
+      "loss": 0.6062,
+      "step": 9024
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.55568339761988e-05,
+      "loss": 0.6262,
+      "step": 9025
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5555928667630984e-05,
+      "loss": 0.6546,
+      "step": 9026
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5555023293190608e-05,
+      "loss": 0.6776,
+      "step": 9027
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5554117852888402e-05,
+      "loss": 0.6202,
+      "step": 9028
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.55532123467351e-05,
+      "loss": 0.6612,
+      "step": 9029
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5552306774741443e-05,
+      "loss": 0.643,
+      "step": 9030
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5551401136918167e-05,
+      "loss": 0.6383,
+      "step": 9031
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5550495433276007e-05,
+      "loss": 0.6424,
+      "step": 9032
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5549589663825702e-05,
+      "loss": 0.6011,
+      "step": 9033
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5548683828577993e-05,
+      "loss": 0.6408,
+      "step": 9034
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5547777927543616e-05,
+      "loss": 0.6453,
+      "step": 9035
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5546871960733313e-05,
+      "loss": 0.6841,
+      "step": 9036
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.554596592815783e-05,
+      "loss": 0.6146,
+      "step": 9037
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.55450598298279e-05,
+      "loss": 0.6198,
+      "step": 9038
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5544153665754277e-05,
+      "loss": 0.6379,
+      "step": 9039
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5543247435947694e-05,
+      "loss": 0.6631,
+      "step": 9040
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.55423411404189e-05,
+      "loss": 0.6342,
+      "step": 9041
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5541434779178645e-05,
+      "loss": 0.6571,
+      "step": 9042
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5540528352237666e-05,
+      "loss": 0.6542,
+      "step": 9043
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5539621859606718e-05,
+      "loss": 0.6657,
+      "step": 9044
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5538715301296547e-05,
+      "loss": 0.6223,
+      "step": 9045
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5537808677317893e-05,
+      "loss": 0.5735,
+      "step": 9046
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5536901987681518e-05,
+      "loss": 0.6402,
+      "step": 9047
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.553599523239816e-05,
+      "loss": 0.5785,
+      "step": 9048
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.553508841147858e-05,
+      "loss": 0.646,
+      "step": 9049
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5534181524933523e-05,
+      "loss": 0.6175,
+      "step": 9050
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5533274572773743e-05,
+      "loss": 0.7057,
+      "step": 9051
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5532367555009994e-05,
+      "loss": 0.6005,
+      "step": 9052
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5531460471653033e-05,
+      "loss": 0.66,
+      "step": 9053
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5530553322713605e-05,
+      "loss": 0.6592,
+      "step": 9054
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5529646108202473e-05,
+      "loss": 0.618,
+      "step": 9055
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.552873882813039e-05,
+      "loss": 0.664,
+      "step": 9056
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.552783148250812e-05,
+      "loss": 0.651,
+      "step": 9057
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5526924071346406e-05,
+      "loss": 0.6043,
+      "step": 9058
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5526016594656025e-05,
+      "loss": 0.6294,
+      "step": 9059
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.552510905244772e-05,
+      "loss": 0.6318,
+      "step": 9060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.552420144473226e-05,
+      "loss": 0.606,
+      "step": 9061
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5523293771520406e-05,
+      "loss": 0.6333,
+      "step": 9062
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5522386032822914e-05,
+      "loss": 0.601,
+      "step": 9063
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.552147822865055e-05,
+      "loss": 0.6382,
+      "step": 9064
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5520570359014078e-05,
+      "loss": 0.641,
+      "step": 9065
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.551966242392426e-05,
+      "loss": 0.6304,
+      "step": 9066
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5518754423391864e-05,
+      "loss": 0.6186,
+      "step": 9067
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5517846357427654e-05,
+      "loss": 0.591,
+      "step": 9068
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5516938226042388e-05,
+      "loss": 0.6369,
+      "step": 9069
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5516030029246845e-05,
+      "loss": 0.5944,
+      "step": 9070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.551512176705179e-05,
+      "loss": 0.6455,
+      "step": 9071
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5514213439467988e-05,
+      "loss": 0.6206,
+      "step": 9072
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.551330504650621e-05,
+      "loss": 0.617,
+      "step": 9073
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5512396588177225e-05,
+      "loss": 0.6208,
+      "step": 9074
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5511488064491807e-05,
+      "loss": 0.6396,
+      "step": 9075
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5510579475460724e-05,
+      "loss": 0.6092,
+      "step": 9076
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5509670821094755e-05,
+      "loss": 0.6375,
+      "step": 9077
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5508762101404666e-05,
+      "loss": 0.6081,
+      "step": 9078
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5507853316401235e-05,
+      "loss": 0.6306,
+      "step": 9079
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5506944466095233e-05,
+      "loss": 0.663,
+      "step": 9080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.550603555049744e-05,
+      "loss": 0.6635,
+      "step": 9081
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5505126569618628e-05,
+      "loss": 0.6064,
+      "step": 9082
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5504217523469576e-05,
+      "loss": 0.6549,
+      "step": 9083
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5503308412061064e-05,
+      "loss": 0.6173,
+      "step": 9084
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.550239923540387e-05,
+      "loss": 0.6257,
+      "step": 9085
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5501489993508773e-05,
+      "loss": 0.6059,
+      "step": 9086
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.550058068638655e-05,
+      "loss": 0.6051,
+      "step": 9087
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5499671314047983e-05,
+      "loss": 0.6417,
+      "step": 9088
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.549876187650386e-05,
+      "loss": 0.5869,
+      "step": 9089
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5497852373764958e-05,
+      "loss": 0.6417,
+      "step": 9090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.5496942805842063e-05,
+      "loss": 0.6179,
+      "step": 9091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5496033172745955e-05,
+      "loss": 0.627,
+      "step": 9092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.549512347448742e-05,
+      "loss": 0.6005,
+      "step": 9093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5494213711077244e-05,
+      "loss": 0.6038,
+      "step": 9094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5493303882526217e-05,
+      "loss": 0.5947,
+      "step": 9095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5492393988845126e-05,
+      "loss": 0.6148,
+      "step": 9096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.549148403004475e-05,
+      "loss": 0.613,
+      "step": 9097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.549057400613589e-05,
+      "loss": 0.6539,
+      "step": 9098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5489663917129325e-05,
+      "loss": 0.6152,
+      "step": 9099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5488753763035852e-05,
+      "loss": 0.6374,
+      "step": 9100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5487843543866257e-05,
+      "loss": 0.6437,
+      "step": 9101
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.548693325963134e-05,
+      "loss": 0.6097,
+      "step": 9102
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5486022910341884e-05,
+      "loss": 0.6485,
+      "step": 9103
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.548511249600869e-05,
+      "loss": 0.6745,
+      "step": 9104
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5484202016642546e-05,
+      "loss": 0.651,
+      "step": 9105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5483291472254248e-05,
+      "loss": 0.5727,
+      "step": 9106
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5482380862854596e-05,
+      "loss": 0.6204,
+      "step": 9107
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5481470188454384e-05,
+      "loss": 0.6566,
+      "step": 9108
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5480559449064404e-05,
+      "loss": 0.5982,
+      "step": 9109
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5479648644695463e-05,
+      "loss": 0.6564,
+      "step": 9110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5478737775358357e-05,
+      "loss": 0.6047,
+      "step": 9111
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5477826841063883e-05,
+      "loss": 0.6421,
+      "step": 9112
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.547691584182284e-05,
+      "loss": 0.6014,
+      "step": 9113
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.547600477764603e-05,
+      "loss": 0.662,
+      "step": 9114
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5475093648544262e-05,
+      "loss": 0.6635,
+      "step": 9115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5474182454528328e-05,
+      "loss": 0.6397,
+      "step": 9116
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.547327119560904e-05,
+      "loss": 0.6617,
+      "step": 9117
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5472359871797194e-05,
+      "loss": 0.6081,
+      "step": 9118
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5471448483103602e-05,
+      "loss": 0.6691,
+      "step": 9119
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5470537029539066e-05,
+      "loss": 0.6868,
+      "step": 9120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5469625511114395e-05,
+      "loss": 0.6485,
+      "step": 9121
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.546871392784039e-05,
+      "loss": 0.6274,
+      "step": 9122
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5467802279727867e-05,
+      "loss": 0.5905,
+      "step": 9123
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.546689056678763e-05,
+      "loss": 0.6225,
+      "step": 9124
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.546597878903049e-05,
+      "loss": 0.6731,
+      "step": 9125
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5465066946467255e-05,
+      "loss": 0.5876,
+      "step": 9126
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.546415503910874e-05,
+      "loss": 0.641,
+      "step": 9127
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5463243066965756e-05,
+      "loss": 0.6217,
+      "step": 9128
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5462331030049115e-05,
+      "loss": 0.5912,
+      "step": 9129
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5461418928369628e-05,
+      "loss": 0.6161,
+      "step": 9130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.546050676193811e-05,
+      "loss": 0.6481,
+      "step": 9131
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5459594530765377e-05,
+      "loss": 0.632,
+      "step": 9132
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5458682234862245e-05,
+      "loss": 0.611,
+      "step": 9133
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.545776987423953e-05,
+      "loss": 0.6602,
+      "step": 9134
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5456857448908048e-05,
+      "loss": 0.6169,
+      "step": 9135
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.545594495887862e-05,
+      "loss": 0.664,
+      "step": 9136
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5455032404162063e-05,
+      "loss": 0.6383,
+      "step": 9137
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5454119784769192e-05,
+      "loss": 0.6271,
+      "step": 9138
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5453207100710834e-05,
+      "loss": 0.6333,
+      "step": 9139
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5452294351997807e-05,
+      "loss": 0.606,
+      "step": 9140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5451381538640933e-05,
+      "loss": 0.6053,
+      "step": 9141
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5450468660651038e-05,
+      "loss": 0.6557,
+      "step": 9142
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.544955571803894e-05,
+      "loss": 0.6667,
+      "step": 9143
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.544864271081546e-05,
+      "loss": 0.5926,
+      "step": 9144
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5447729638991437e-05,
+      "loss": 0.593,
+      "step": 9145
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5446816502577685e-05,
+      "loss": 0.6389,
+      "step": 9146
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.544590330158503e-05,
+      "loss": 0.6862,
+      "step": 9147
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5444990036024306e-05,
+      "loss": 0.5867,
+      "step": 9148
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5444076705906336e-05,
+      "loss": 0.6665,
+      "step": 9149
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5443163311241945e-05,
+      "loss": 0.6467,
+      "step": 9150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5442249852041975e-05,
+      "loss": 0.6248,
+      "step": 9151
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5441336328317244e-05,
+      "loss": 0.6923,
+      "step": 9152
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5440422740078587e-05,
+      "loss": 0.662,
+      "step": 9153
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.543950908733684e-05,
+      "loss": 0.6051,
+      "step": 9154
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5438595370102828e-05,
+      "loss": 0.6581,
+      "step": 9155
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.543768158838739e-05,
+      "loss": 0.5961,
+      "step": 9156
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.543676774220136e-05,
+      "loss": 0.6077,
+      "step": 9157
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.543585383155557e-05,
+      "loss": 0.6622,
+      "step": 9158
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.543493985646085e-05,
+      "loss": 0.6022,
+      "step": 9159
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.543402581692805e-05,
+      "loss": 0.6445,
+      "step": 9160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5433111712967998e-05,
+      "loss": 0.6549,
+      "step": 9161
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5432197544591532e-05,
+      "loss": 0.6064,
+      "step": 9162
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5431283311809497e-05,
+      "loss": 0.6452,
+      "step": 9163
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5430369014632724e-05,
+      "loss": 0.5994,
+      "step": 9164
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5429454653072055e-05,
+      "loss": 0.59,
+      "step": 9165
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5428540227138335e-05,
+      "loss": 0.6181,
+      "step": 9166
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5427625736842404e-05,
+      "loss": 0.6275,
+      "step": 9167
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5426711182195106e-05,
+      "loss": 0.67,
+      "step": 9168
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5425796563207277e-05,
+      "loss": 0.6268,
+      "step": 9169
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5424881879889772e-05,
+      "loss": 0.6392,
+      "step": 9170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5423967132253424e-05,
+      "loss": 0.6388,
+      "step": 9171
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5423052320309087e-05,
+      "loss": 0.6161,
+      "step": 9172
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.54221374440676e-05,
+      "loss": 0.6416,
+      "step": 9173
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.542122250353982e-05,
+      "loss": 0.6276,
+      "step": 9174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.542030749873659e-05,
+      "loss": 0.6375,
+      "step": 9175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5419392429668757e-05,
+      "loss": 0.6341,
+      "step": 9176
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.541847729634717e-05,
+      "loss": 0.6094,
+      "step": 9177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5417562098782682e-05,
+      "loss": 0.6376,
+      "step": 9178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5416646836986137e-05,
+      "loss": 0.6039,
+      "step": 9179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5415731510968397e-05,
+      "loss": 0.6229,
+      "step": 9180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.541481612074031e-05,
+      "loss": 0.6606,
+      "step": 9181
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5413900666312723e-05,
+      "loss": 0.6227,
+      "step": 9182
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5412985147696497e-05,
+      "loss": 0.6199,
+      "step": 9183
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5412069564902484e-05,
+      "loss": 0.6208,
+      "step": 9184
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5411153917941542e-05,
+      "loss": 0.6064,
+      "step": 9185
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5410238206824525e-05,
+      "loss": 0.709,
+      "step": 9186
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.540932243156229e-05,
+      "loss": 0.6131,
+      "step": 9187
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5408406592165693e-05,
+      "loss": 0.657,
+      "step": 9188
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5407490688645593e-05,
+      "loss": 0.6773,
+      "step": 9189
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5406574721012854e-05,
+      "loss": 0.6014,
+      "step": 9190
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.540565868927833e-05,
+      "loss": 0.651,
+      "step": 9191
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5404742593452886e-05,
+      "loss": 0.6056,
+      "step": 9192
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5403826433547378e-05,
+      "loss": 0.5953,
+      "step": 9193
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5402910209572673e-05,
+      "loss": 0.6549,
+      "step": 9194
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.540199392153964e-05,
+      "loss": 0.6328,
+      "step": 9195
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5401077569459123e-05,
+      "loss": 0.6188,
+      "step": 9196
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.540016115334201e-05,
+      "loss": 0.6127,
+      "step": 9197
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.539924467319915e-05,
+      "loss": 0.6185,
+      "step": 9198
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5398328129041415e-05,
+      "loss": 0.6161,
+      "step": 9199
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.539741152087967e-05,
+      "loss": 0.6006,
+      "step": 9200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5396494848724785e-05,
+      "loss": 0.6231,
+      "step": 9201
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5395578112587628e-05,
+      "loss": 0.6741,
+      "step": 9202
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5394661312479064e-05,
+      "loss": 0.5954,
+      "step": 9203
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5393744448409967e-05,
+      "loss": 0.6292,
+      "step": 9204
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5392827520391208e-05,
+      "loss": 0.6527,
+      "step": 9205
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5391910528433655e-05,
+      "loss": 0.6166,
+      "step": 9206
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5390993472548185e-05,
+      "loss": 0.6189,
+      "step": 9207
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5390076352745663e-05,
+      "loss": 0.5953,
+      "step": 9208
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5389159169036967e-05,
+      "loss": 0.5873,
+      "step": 9209
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.538824192143298e-05,
+      "loss": 0.5989,
+      "step": 9210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.538732460994456e-05,
+      "loss": 0.6531,
+      "step": 9211
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5386407234582598e-05,
+      "loss": 0.7412,
+      "step": 9212
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.538548979535796e-05,
+      "loss": 0.6161,
+      "step": 9213
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5384572292281527e-05,
+      "loss": 0.6057,
+      "step": 9214
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5383654725364183e-05,
+      "loss": 0.5928,
+      "step": 9215
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.53827370946168e-05,
+      "loss": 0.6399,
+      "step": 9216
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5381819400050257e-05,
+      "loss": 0.6473,
+      "step": 9217
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.538090164167544e-05,
+      "loss": 0.6004,
+      "step": 9218
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5379983819503226e-05,
+      "loss": 0.6921,
+      "step": 9219
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5379065933544495e-05,
+      "loss": 0.6278,
+      "step": 9220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5378147983810136e-05,
+      "loss": 0.6243,
+      "step": 9221
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5377229970311032e-05,
+      "loss": 0.6204,
+      "step": 9222
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.537631189305806e-05,
+      "loss": 0.6686,
+      "step": 9223
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5375393752062108e-05,
+      "loss": 0.6589,
+      "step": 9224
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.537447554733407e-05,
+      "loss": 0.6431,
+      "step": 9225
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5373557278884822e-05,
+      "loss": 0.6665,
+      "step": 9226
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5372638946725252e-05,
+      "loss": 0.6325,
+      "step": 9227
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5371720550866256e-05,
+      "loss": 0.6434,
+      "step": 9228
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5370802091318714e-05,
+      "loss": 0.6248,
+      "step": 9229
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.536988356809352e-05,
+      "loss": 0.6254,
+      "step": 9230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5368964981201563e-05,
+      "loss": 0.5734,
+      "step": 9231
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5368046330653736e-05,
+      "loss": 0.6119,
+      "step": 9232
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5367127616460927e-05,
+      "loss": 0.6842,
+      "step": 9233
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.536620883863403e-05,
+      "loss": 0.5954,
+      "step": 9234
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5365289997183943e-05,
+      "loss": 0.6619,
+      "step": 9235
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5364371092121554e-05,
+      "loss": 0.6261,
+      "step": 9236
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.536345212345776e-05,
+      "loss": 0.6088,
+      "step": 9237
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5362533091203454e-05,
+      "loss": 0.642,
+      "step": 9238
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.536161399536954e-05,
+      "loss": 0.6076,
+      "step": 9239
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5360694835966903e-05,
+      "loss": 0.6196,
+      "step": 9240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5359775613006448e-05,
+      "loss": 0.6702,
+      "step": 9241
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5358856326499078e-05,
+      "loss": 0.5974,
+      "step": 9242
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5357936976455684e-05,
+      "loss": 0.5836,
+      "step": 9243
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.535701756288717e-05,
+      "loss": 0.6618,
+      "step": 9244
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5356098085804436e-05,
+      "loss": 0.5984,
+      "step": 9245
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5355178545218382e-05,
+      "loss": 0.6838,
+      "step": 9246
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5354258941139915e-05,
+      "loss": 0.6293,
+      "step": 9247
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5353339273579933e-05,
+      "loss": 0.7126,
+      "step": 9248
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.535241954254934e-05,
+      "loss": 0.6282,
+      "step": 9249
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5351499748059046e-05,
+      "loss": 0.6147,
+      "step": 9250
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5350579890119952e-05,
+      "loss": 0.6189,
+      "step": 9251
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5349659968742963e-05,
+      "loss": 0.6814,
+      "step": 9252
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.534873998393899e-05,
+      "loss": 0.6546,
+      "step": 9253
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.534781993571894e-05,
+      "loss": 0.6613,
+      "step": 9254
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5346899824093717e-05,
+      "loss": 0.6403,
+      "step": 9255
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5345979649074233e-05,
+      "loss": 0.6334,
+      "step": 9256
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5345059410671396e-05,
+      "loss": 0.5932,
+      "step": 9257
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5344139108896122e-05,
+      "loss": 0.618,
+      "step": 9258
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5343218743759317e-05,
+      "loss": 0.6503,
+      "step": 9259
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5342298315271892e-05,
+      "loss": 0.6024,
+      "step": 9260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5341377823444762e-05,
+      "loss": 0.6124,
+      "step": 9261
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5340457268288845e-05,
+      "loss": 0.6848,
+      "step": 9262
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.533953664981505e-05,
+      "loss": 0.5913,
+      "step": 9263
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5338615968034294e-05,
+      "loss": 0.6256,
+      "step": 9264
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.533769522295749e-05,
+      "loss": 0.6263,
+      "step": 9265
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.533677441459556e-05,
+      "loss": 0.5887,
+      "step": 9266
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5335853542959417e-05,
+      "loss": 0.6686,
+      "step": 9267
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5334932608059978e-05,
+      "loss": 0.6905,
+      "step": 9268
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5334011609908167e-05,
+      "loss": 0.5804,
+      "step": 9269
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.53330905485149e-05,
+      "loss": 0.6002,
+      "step": 9270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.53321694238911e-05,
+      "loss": 0.5858,
+      "step": 9271
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5331248236047684e-05,
+      "loss": 0.6541,
+      "step": 9272
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5330326984995575e-05,
+      "loss": 0.6335,
+      "step": 9273
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5329405670745695e-05,
+      "loss": 0.6221,
+      "step": 9274
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5328484293308977e-05,
+      "loss": 0.6625,
+      "step": 9275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.532756285269633e-05,
+      "loss": 0.6233,
+      "step": 9276
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.532664134891869e-05,
+      "loss": 0.6902,
+      "step": 9277
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5325719781986976e-05,
+      "loss": 0.6785,
+      "step": 9278
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.532479815191212e-05,
+      "loss": 0.6719,
+      "step": 9279
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5323876458705043e-05,
+      "loss": 0.6357,
+      "step": 9280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5322954702376677e-05,
+      "loss": 0.6311,
+      "step": 9281
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5322032882937946e-05,
+      "loss": 0.6503,
+      "step": 9282
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5321111000399788e-05,
+      "loss": 0.6684,
+      "step": 9283
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5320189054773127e-05,
+      "loss": 0.6735,
+      "step": 9284
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5319267046068893e-05,
+      "loss": 0.6982,
+      "step": 9285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5318344974298017e-05,
+      "loss": 0.6692,
+      "step": 9286
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5317422839471437e-05,
+      "loss": 0.6402,
+      "step": 9287
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5316500641600083e-05,
+      "loss": 0.6444,
+      "step": 9288
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5315578380694885e-05,
+      "loss": 0.67,
+      "step": 9289
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5314656056766787e-05,
+      "loss": 0.668,
+      "step": 9290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5313733669826715e-05,
+      "loss": 0.6583,
+      "step": 9291
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5312811219885606e-05,
+      "loss": 0.6381,
+      "step": 9292
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.53118887069544e-05,
+      "loss": 0.6969,
+      "step": 9293
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5310966131044037e-05,
+      "loss": 0.6568,
+      "step": 9294
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5310043492165453e-05,
+      "loss": 0.6118,
+      "step": 9295
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.530912079032958e-05,
+      "loss": 0.6508,
+      "step": 9296
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5308198025547366e-05,
+      "loss": 0.6233,
+      "step": 9297
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5307275197829756e-05,
+      "loss": 0.6386,
+      "step": 9298
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.530635230718768e-05,
+      "loss": 0.6541,
+      "step": 9299
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.530542935363208e-05,
+      "loss": 0.6315,
+      "step": 9300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.530450633717391e-05,
+      "loss": 0.6135,
+      "step": 9301
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5303583257824107e-05,
+      "loss": 0.6732,
+      "step": 9302
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5302660115593613e-05,
+      "loss": 0.6116,
+      "step": 9303
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5301736910493375e-05,
+      "loss": 0.6312,
+      "step": 9304
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.530081364253434e-05,
+      "loss": 0.615,
+      "step": 9305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5299890311727456e-05,
+      "loss": 0.649,
+      "step": 9306
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5298966918083667e-05,
+      "loss": 0.628,
+      "step": 9307
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.529804346161392e-05,
+      "loss": 0.6774,
+      "step": 9308
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5297119942329166e-05,
+      "loss": 0.6119,
+      "step": 9309
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5296196360240355e-05,
+      "loss": 0.6406,
+      "step": 9310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5295272715358436e-05,
+      "loss": 0.6469,
+      "step": 9311
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5294349007694364e-05,
+      "loss": 0.6174,
+      "step": 9312
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5293425237259082e-05,
+      "loss": 0.635,
+      "step": 9313
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5292501404063548e-05,
+      "loss": 0.6237,
+      "step": 9314
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5291577508118715e-05,
+      "loss": 0.5885,
+      "step": 9315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5290653549435542e-05,
+      "loss": 0.625,
+      "step": 9316
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5289729528024974e-05,
+      "loss": 0.6494,
+      "step": 9317
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.528880544389797e-05,
+      "loss": 0.6219,
+      "step": 9318
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.528788129706549e-05,
+      "loss": 0.6403,
+      "step": 9319
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5286957087538487e-05,
+      "loss": 0.6056,
+      "step": 9320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5286032815327918e-05,
+      "loss": 0.6929,
+      "step": 9321
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5285108480444747e-05,
+      "loss": 0.5985,
+      "step": 9322
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5284184082899926e-05,
+      "loss": 0.6785,
+      "step": 9323
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5283259622704417e-05,
+      "loss": 0.6178,
+      "step": 9324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5282335099869183e-05,
+      "loss": 0.6164,
+      "step": 9325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.528141051440519e-05,
+      "loss": 0.6311,
+      "step": 9326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5280485866323383e-05,
+      "loss": 0.5928,
+      "step": 9327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5279561155634747e-05,
+      "loss": 0.63,
+      "step": 9328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.527863638235023e-05,
+      "loss": 0.6409,
+      "step": 9329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5277711546480797e-05,
+      "loss": 0.6277,
+      "step": 9330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5276786648037424e-05,
+      "loss": 0.6225,
+      "step": 9331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5275861687031067e-05,
+      "loss": 0.5904,
+      "step": 9332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.52749366634727e-05,
+      "loss": 0.6653,
+      "step": 9333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.527401157737328e-05,
+      "loss": 0.6118,
+      "step": 9334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5273086428743785e-05,
+      "loss": 0.6659,
+      "step": 9335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5272161217595174e-05,
+      "loss": 0.6243,
+      "step": 9336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.527123594393843e-05,
+      "loss": 0.6369,
+      "step": 9337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5270310607784513e-05,
+      "loss": 0.629,
+      "step": 9338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5269385209144394e-05,
+      "loss": 0.6539,
+      "step": 9339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.526845974802905e-05,
+      "loss": 0.6481,
+      "step": 9340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5267534224449452e-05,
+      "loss": 0.6171,
+      "step": 9341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.526660863841657e-05,
+      "loss": 0.5997,
+      "step": 9342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.526568298994138e-05,
+      "loss": 0.6346,
+      "step": 9343
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5264757279034864e-05,
+      "loss": 0.6294,
+      "step": 9344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5263831505707983e-05,
+      "loss": 0.6448,
+      "step": 9345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5262905669971724e-05,
+      "loss": 0.6814,
+      "step": 9346
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.526197977183706e-05,
+      "loss": 0.6075,
+      "step": 9347
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5261053811314967e-05,
+      "loss": 0.6397,
+      "step": 9348
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5260127788416428e-05,
+      "loss": 0.6576,
+      "step": 9349
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.525920170315242e-05,
+      "loss": 0.6592,
+      "step": 9350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5258275555533923e-05,
+      "loss": 0.6046,
+      "step": 9351
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5257349345571917e-05,
+      "loss": 0.6372,
+      "step": 9352
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5256423073277384e-05,
+      "loss": 0.6185,
+      "step": 9353
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5255496738661305e-05,
+      "loss": 0.6268,
+      "step": 9354
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5254570341734664e-05,
+      "loss": 0.6159,
+      "step": 9355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5253643882508447e-05,
+      "loss": 0.595,
+      "step": 9356
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5252717360993635e-05,
+      "loss": 0.6025,
+      "step": 9357
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5251790777201211e-05,
+      "loss": 0.5851,
+      "step": 9358
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5250864131142169e-05,
+      "loss": 0.6327,
+      "step": 9359
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5249937422827486e-05,
+      "loss": 0.6213,
+      "step": 9360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5249010652268155e-05,
+      "loss": 0.6101,
+      "step": 9361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5248083819475163e-05,
+      "loss": 0.5986,
+      "step": 9362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5247156924459499e-05,
+      "loss": 0.6173,
+      "step": 9363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5246229967232151e-05,
+      "loss": 0.6795,
+      "step": 9364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5245302947804113e-05,
+      "loss": 0.6485,
+      "step": 9365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5244375866186367e-05,
+      "loss": 0.6434,
+      "step": 9366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5243448722389915e-05,
+      "loss": 0.6291,
+      "step": 9367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5242521516425748e-05,
+      "loss": 0.6168,
+      "step": 9368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5241594248304852e-05,
+      "loss": 0.6151,
+      "step": 9369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5240666918038227e-05,
+      "loss": 0.6205,
+      "step": 9370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5239739525636866e-05,
+      "loss": 0.6525,
+      "step": 9371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5238812071111767e-05,
+      "loss": 0.6977,
+      "step": 9372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5237884554473918e-05,
+      "loss": 0.6289,
+      "step": 9373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5236956975734328e-05,
+      "loss": 0.5968,
+      "step": 9374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5236029334903983e-05,
+      "loss": 0.6393,
+      "step": 9375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5235101631993891e-05,
+      "loss": 0.6216,
+      "step": 9376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5234173867015046e-05,
+      "loss": 0.6956,
+      "step": 9377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5233246039978447e-05,
+      "loss": 0.6204,
+      "step": 9378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5232318150895095e-05,
+      "loss": 0.5522,
+      "step": 9379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5231390199775997e-05,
+      "loss": 0.6534,
+      "step": 9380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5230462186632147e-05,
+      "loss": 0.5782,
+      "step": 9381
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.522953411147455e-05,
+      "loss": 0.6396,
+      "step": 9382
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5228605974314214e-05,
+      "loss": 0.5976,
+      "step": 9383
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5227677775162138e-05,
+      "loss": 0.6411,
+      "step": 9384
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5226749514029329e-05,
+      "loss": 0.6108,
+      "step": 9385
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5225821190926793e-05,
+      "loss": 0.6492,
+      "step": 9386
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5224892805865534e-05,
+      "loss": 0.6589,
+      "step": 9387
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5223964358856567e-05,
+      "loss": 0.7096,
+      "step": 9388
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.522303584991089e-05,
+      "loss": 0.6469,
+      "step": 9389
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5222107279039513e-05,
+      "loss": 0.6965,
+      "step": 9390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5221178646253454e-05,
+      "loss": 0.6008,
+      "step": 9391
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5220249951563714e-05,
+      "loss": 0.6579,
+      "step": 9392
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5219321194981307e-05,
+      "loss": 0.6628,
+      "step": 9393
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5218392376517248e-05,
+      "loss": 0.6208,
+      "step": 9394
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5217463496182543e-05,
+      "loss": 0.6451,
+      "step": 9395
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5216534553988206e-05,
+      "loss": 0.6245,
+      "step": 9396
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5215605549945261e-05,
+      "loss": 0.6604,
+      "step": 9397
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5214676484064707e-05,
+      "loss": 0.6367,
+      "step": 9398
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5213747356357569e-05,
+      "loss": 0.5502,
+      "step": 9399
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5212818166834859e-05,
+      "loss": 0.6618,
+      "step": 9400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.52118889155076e-05,
+      "loss": 0.6653,
+      "step": 9401
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5210959602386801e-05,
+      "loss": 0.6098,
+      "step": 9402
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5210030227483485e-05,
+      "loss": 0.7012,
+      "step": 9403
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.520910079080867e-05,
+      "loss": 0.6329,
+      "step": 9404
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5208171292373375e-05,
+      "loss": 0.6477,
+      "step": 9405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.520724173218862e-05,
+      "loss": 0.6379,
+      "step": 9406
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.520631211026543e-05,
+      "loss": 0.6963,
+      "step": 9407
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5205382426614821e-05,
+      "loss": 0.5887,
+      "step": 9408
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.520445268124782e-05,
+      "loss": 0.6322,
+      "step": 9409
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5203522874175448e-05,
+      "loss": 0.5572,
+      "step": 9410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.520259300540873e-05,
+      "loss": 0.6034,
+      "step": 9411
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5201663074958695e-05,
+      "loss": 0.6132,
+      "step": 9412
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5200733082836356e-05,
+      "loss": 0.6829,
+      "step": 9413
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.519980302905275e-05,
+      "loss": 0.6199,
+      "step": 9414
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5198872913618905e-05,
+      "loss": 0.6534,
+      "step": 9415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5197942736545845e-05,
+      "loss": 0.6109,
+      "step": 9416
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5197012497844594e-05,
+      "loss": 0.67,
+      "step": 9417
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.519608219752619e-05,
+      "loss": 0.652,
+      "step": 9418
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5195151835601656e-05,
+      "loss": 0.5577,
+      "step": 9419
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5194221412082024e-05,
+      "loss": 0.6446,
+      "step": 9420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5193290926978331e-05,
+      "loss": 0.632,
+      "step": 9421
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.51923603803016e-05,
+      "loss": 0.6234,
+      "step": 9422
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5191429772062867e-05,
+      "loss": 0.6538,
+      "step": 9423
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.519049910227317e-05,
+      "loss": 0.65,
+      "step": 9424
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5189568370943541e-05,
+      "loss": 0.6546,
+      "step": 9425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5188637578085011e-05,
+      "loss": 0.6284,
+      "step": 9426
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5187706723708621e-05,
+      "loss": 0.6734,
+      "step": 9427
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5186775807825407e-05,
+      "loss": 0.6786,
+      "step": 9428
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.51858448304464e-05,
+      "loss": 0.6837,
+      "step": 9429
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5184913791582645e-05,
+      "loss": 0.6118,
+      "step": 9430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.518398269124518e-05,
+      "loss": 0.6043,
+      "step": 9431
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5183051529445038e-05,
+      "loss": 0.5735,
+      "step": 9432
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5182120306193269e-05,
+      "loss": 0.6286,
+      "step": 9433
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5181189021500906e-05,
+      "loss": 0.6817,
+      "step": 9434
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5180257675378988e-05,
+      "loss": 0.6365,
+      "step": 9435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5179326267838567e-05,
+      "loss": 0.6547,
+      "step": 9436
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5178394798890682e-05,
+      "loss": 0.6574,
+      "step": 9437
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5177463268546374e-05,
+      "loss": 0.6089,
+      "step": 9438
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5176531676816688e-05,
+      "loss": 0.58,
+      "step": 9439
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5175600023712675e-05,
+      "loss": 0.6802,
+      "step": 9440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5174668309245371e-05,
+      "loss": 0.6403,
+      "step": 9441
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5173736533425834e-05,
+      "loss": 0.6552,
+      "step": 9442
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5172804696265102e-05,
+      "loss": 0.6161,
+      "step": 9443
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5171872797774225e-05,
+      "loss": 0.6483,
+      "step": 9444
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5170940837964259e-05,
+      "loss": 0.6428,
+      "step": 9445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5170008816846247e-05,
+      "loss": 0.6841,
+      "step": 9446
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5169076734431238e-05,
+      "loss": 0.6262,
+      "step": 9447
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5168144590730285e-05,
+      "loss": 0.6321,
+      "step": 9448
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5167212385754441e-05,
+      "loss": 0.6262,
+      "step": 9449
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5166280119514758e-05,
+      "loss": 0.6549,
+      "step": 9450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5165347792022286e-05,
+      "loss": 0.6615,
+      "step": 9451
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5164415403288088e-05,
+      "loss": 0.6546,
+      "step": 9452
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5163482953323208e-05,
+      "loss": 0.6482,
+      "step": 9453
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5162550442138706e-05,
+      "loss": 0.6069,
+      "step": 9454
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5161617869745636e-05,
+      "loss": 0.6831,
+      "step": 9455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.516068523615506e-05,
+      "loss": 0.6085,
+      "step": 9456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.515975254137803e-05,
+      "loss": 0.68,
+      "step": 9457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5158819785425612e-05,
+      "loss": 0.6499,
+      "step": 9458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5157886968308852e-05,
+      "loss": 0.6163,
+      "step": 9459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.515695409003882e-05,
+      "loss": 0.6289,
+      "step": 9460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5156021150626576e-05,
+      "loss": 0.6077,
+      "step": 9461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5155088150083175e-05,
+      "loss": 0.5708,
+      "step": 9462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5154155088419684e-05,
+      "loss": 0.6571,
+      "step": 9463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5153221965647167e-05,
+      "loss": 0.6003,
+      "step": 9464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5152288781776681e-05,
+      "loss": 0.663,
+      "step": 9465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5151355536819296e-05,
+      "loss": 0.5954,
+      "step": 9466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5150422230786075e-05,
+      "loss": 0.6301,
+      "step": 9467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5149488863688082e-05,
+      "loss": 0.6541,
+      "step": 9468
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5148555435536386e-05,
+      "loss": 0.6727,
+      "step": 9469
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.514762194634205e-05,
+      "loss": 0.6582,
+      "step": 9470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5146688396116144e-05,
+      "loss": 0.666,
+      "step": 9471
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5145754784869737e-05,
+      "loss": 0.6291,
+      "step": 9472
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5144821112613901e-05,
+      "loss": 0.5997,
+      "step": 9473
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5143887379359696e-05,
+      "loss": 0.6027,
+      "step": 9474
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5142953585118205e-05,
+      "loss": 0.6381,
+      "step": 9475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.514201972990049e-05,
+      "loss": 0.5898,
+      "step": 9476
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5141085813717629e-05,
+      "loss": 0.6413,
+      "step": 9477
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.514015183658069e-05,
+      "loss": 0.6826,
+      "step": 9478
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.513921779850075e-05,
+      "loss": 0.6223,
+      "step": 9479
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.513828369948888e-05,
+      "loss": 0.6809,
+      "step": 9480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.513734953955616e-05,
+      "loss": 0.5912,
+      "step": 9481
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.513641531871366e-05,
+      "loss": 0.7068,
+      "step": 9482
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5135481036972457e-05,
+      "loss": 0.6273,
+      "step": 9483
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5134546694343635e-05,
+      "loss": 0.6226,
+      "step": 9484
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5133612290838262e-05,
+      "loss": 0.6416,
+      "step": 9485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5132677826467421e-05,
+      "loss": 0.6602,
+      "step": 9486
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5131743301242196e-05,
+      "loss": 0.622,
+      "step": 9487
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5130808715173659e-05,
+      "loss": 0.6701,
+      "step": 9488
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5129874068272896e-05,
+      "loss": 0.6654,
+      "step": 9489
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5128939360550985e-05,
+      "loss": 0.6009,
+      "step": 9490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.512800459201901e-05,
+      "loss": 0.6294,
+      "step": 9491
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5127069762688056e-05,
+      "loss": 0.6223,
+      "step": 9492
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5126134872569202e-05,
+      "loss": 0.6903,
+      "step": 9493
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5125199921673534e-05,
+      "loss": 0.6854,
+      "step": 9494
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5124264910012139e-05,
+      "loss": 0.6295,
+      "step": 9495
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5123329837596101e-05,
+      "loss": 0.6235,
+      "step": 9496
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5122394704436505e-05,
+      "loss": 0.6625,
+      "step": 9497
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.512145951054444e-05,
+      "loss": 0.5883,
+      "step": 9498
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5120524255930998e-05,
+      "loss": 0.6126,
+      "step": 9499
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.511958894060726e-05,
+      "loss": 0.6061,
+      "step": 9500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5118653564584318e-05,
+      "loss": 0.6304,
+      "step": 9501
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5117718127873264e-05,
+      "loss": 0.6322,
+      "step": 9502
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.511678263048519e-05,
+      "loss": 0.617,
+      "step": 9503
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.511584707243118e-05,
+      "loss": 0.6636,
+      "step": 9504
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5114911453722336e-05,
+      "loss": 0.6395,
+      "step": 9505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5113975774369744e-05,
+      "loss": 0.6013,
+      "step": 9506
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5113040034384501e-05,
+      "loss": 0.6378,
+      "step": 9507
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.51121042337777e-05,
+      "loss": 0.6162,
+      "step": 9508
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5111168372560436e-05,
+      "loss": 0.6945,
+      "step": 9509
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5110232450743806e-05,
+      "loss": 0.6893,
+      "step": 9510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5109296468338906e-05,
+      "loss": 0.5834,
+      "step": 9511
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5108360425356834e-05,
+      "loss": 0.6633,
+      "step": 9512
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5107424321808684e-05,
+      "loss": 0.6432,
+      "step": 9513
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5106488157705558e-05,
+      "loss": 0.5838,
+      "step": 9514
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5105551933058558e-05,
+      "loss": 0.5578,
+      "step": 9515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5104615647878778e-05,
+      "loss": 0.6147,
+      "step": 9516
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5103679302177327e-05,
+      "loss": 0.6041,
+      "step": 9517
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5102742895965299e-05,
+      "loss": 0.6021,
+      "step": 9518
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5101806429253795e-05,
+      "loss": 0.6538,
+      "step": 9519
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5100869902053927e-05,
+      "loss": 0.5898,
+      "step": 9520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5099933314376793e-05,
+      "loss": 0.6778,
+      "step": 9521
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5098996666233497e-05,
+      "loss": 0.5752,
+      "step": 9522
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5098059957635145e-05,
+      "loss": 0.6617,
+      "step": 9523
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5097123188592844e-05,
+      "loss": 0.6522,
+      "step": 9524
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.50961863591177e-05,
+      "loss": 0.6211,
+      "step": 9525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5095249469220821e-05,
+      "loss": 0.638,
+      "step": 9526
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5094312518913312e-05,
+      "loss": 0.6425,
+      "step": 9527
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5093375508206282e-05,
+      "loss": 0.6217,
+      "step": 9528
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5092438437110846e-05,
+      "loss": 0.6881,
+      "step": 9529
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5091501305638107e-05,
+      "loss": 0.6465,
+      "step": 9530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.509056411379918e-05,
+      "loss": 0.6618,
+      "step": 9531
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.508962686160518e-05,
+      "loss": 0.6242,
+      "step": 9532
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.508868954906721e-05,
+      "loss": 0.5907,
+      "step": 9533
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.508775217619639e-05,
+      "loss": 0.6012,
+      "step": 9534
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5086814743003833e-05,
+      "loss": 0.65,
+      "step": 9535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.508587724950065e-05,
+      "loss": 0.6322,
+      "step": 9536
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5084939695697962e-05,
+      "loss": 0.5988,
+      "step": 9537
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5084002081606878e-05,
+      "loss": 0.6992,
+      "step": 9538
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5083064407238524e-05,
+      "loss": 0.679,
+      "step": 9539
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5082126672604003e-05,
+      "loss": 0.6157,
+      "step": 9540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5081188877714445e-05,
+      "loss": 0.6228,
+      "step": 9541
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5080251022580968e-05,
+      "loss": 0.6548,
+      "step": 9542
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5079313107214683e-05,
+      "loss": 0.6741,
+      "step": 9543
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5078375131626719e-05,
+      "loss": 0.6326,
+      "step": 9544
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5077437095828193e-05,
+      "loss": 0.6151,
+      "step": 9545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5076498999830229e-05,
+      "loss": 0.6264,
+      "step": 9546
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5075560843643944e-05,
+      "loss": 0.5832,
+      "step": 9547
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.507462262728047e-05,
+      "loss": 0.6465,
+      "step": 9548
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5073684350750921e-05,
+      "loss": 0.6397,
+      "step": 9549
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5072746014066427e-05,
+      "loss": 0.6626,
+      "step": 9550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5071807617238112e-05,
+      "loss": 0.5935,
+      "step": 9551
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5070869160277099e-05,
+      "loss": 0.6904,
+      "step": 9552
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5069930643194521e-05,
+      "loss": 0.6426,
+      "step": 9553
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5068992066001503e-05,
+      "loss": 0.6975,
+      "step": 9554
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5068053428709165e-05,
+      "loss": 0.6426,
+      "step": 9555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5067114731328647e-05,
+      "loss": 0.6234,
+      "step": 9556
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5066175973871076e-05,
+      "loss": 0.5986,
+      "step": 9557
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5065237156347578e-05,
+      "loss": 0.6045,
+      "step": 9558
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5064298278769285e-05,
+      "loss": 0.6424,
+      "step": 9559
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5063359341147332e-05,
+      "loss": 0.6383,
+      "step": 9560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5062420343492846e-05,
+      "loss": 0.6186,
+      "step": 9561
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5061481285816963e-05,
+      "loss": 0.6245,
+      "step": 9562
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5060542168130818e-05,
+      "loss": 0.6472,
+      "step": 9563
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5059602990445543e-05,
+      "loss": 0.5976,
+      "step": 9564
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5058663752772274e-05,
+      "loss": 0.6281,
+      "step": 9565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.505772445512215e-05,
+      "loss": 0.6061,
+      "step": 9566
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5056785097506301e-05,
+      "loss": 0.646,
+      "step": 9567
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5055845679935868e-05,
+      "loss": 0.682,
+      "step": 9568
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5054906202421991e-05,
+      "loss": 0.69,
+      "step": 9569
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5053966664975806e-05,
+      "loss": 0.6716,
+      "step": 9570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5053027067608452e-05,
+      "loss": 0.6794,
+      "step": 9571
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5052087410331071e-05,
+      "loss": 0.6297,
+      "step": 9572
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.50511476931548e-05,
+      "loss": 0.5939,
+      "step": 9573
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5050207916090787e-05,
+      "loss": 0.6007,
+      "step": 9574
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5049268079150168e-05,
+      "loss": 0.6615,
+      "step": 9575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5048328182344087e-05,
+      "loss": 0.6007,
+      "step": 9576
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5047388225683694e-05,
+      "loss": 0.6485,
+      "step": 9577
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5046448209180126e-05,
+      "loss": 0.5741,
+      "step": 9578
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5045508132844531e-05,
+      "loss": 0.6506,
+      "step": 9579
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5044567996688056e-05,
+      "loss": 0.6531,
+      "step": 9580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5043627800721843e-05,
+      "loss": 0.7069,
+      "step": 9581
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5042687544957044e-05,
+      "loss": 0.6328,
+      "step": 9582
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5041747229404803e-05,
+      "loss": 0.6804,
+      "step": 9583
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5040806854076273e-05,
+      "loss": 0.6554,
+      "step": 9584
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5039866418982603e-05,
+      "loss": 0.6476,
+      "step": 9585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5038925924134939e-05,
+      "loss": 0.6512,
+      "step": 9586
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5037985369544432e-05,
+      "loss": 0.6192,
+      "step": 9587
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5037044755222237e-05,
+      "loss": 0.6341,
+      "step": 9588
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5036104081179506e-05,
+      "loss": 0.627,
+      "step": 9589
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5035163347427389e-05,
+      "loss": 0.6454,
+      "step": 9590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.503422255397704e-05,
+      "loss": 0.6309,
+      "step": 9591
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5033281700839614e-05,
+      "loss": 0.6569,
+      "step": 9592
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5032340788026267e-05,
+      "loss": 0.6544,
+      "step": 9593
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5031399815548155e-05,
+      "loss": 0.6702,
+      "step": 9594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5030458783416436e-05,
+      "loss": 0.6308,
+      "step": 9595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.502951769164226e-05,
+      "loss": 0.6047,
+      "step": 9596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.502857654023679e-05,
+      "loss": 0.5766,
+      "step": 9597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5027635329211187e-05,
+      "loss": 0.5964,
+      "step": 9598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5026694058576602e-05,
+      "loss": 0.5834,
+      "step": 9599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5025752728344204e-05,
+      "loss": 0.6379,
+      "step": 9600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.502481133852515e-05,
+      "loss": 0.6613,
+      "step": 9601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.50238698891306e-05,
+      "loss": 0.6698,
+      "step": 9602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5022928380171715e-05,
+      "loss": 0.5862,
+      "step": 9603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5021986811659665e-05,
+      "loss": 0.608,
+      "step": 9604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5021045183605605e-05,
+      "loss": 0.6368,
+      "step": 9605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5020103496020704e-05,
+      "loss": 0.6252,
+      "step": 9606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5019161748916125e-05,
+      "loss": 0.6265,
+      "step": 9607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5018219942303036e-05,
+      "loss": 0.6213,
+      "step": 9608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5017278076192598e-05,
+      "loss": 0.6622,
+      "step": 9609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5016336150595984e-05,
+      "loss": 0.6479,
+      "step": 9610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.501539416552436e-05,
+      "loss": 0.6642,
+      "step": 9611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5014452120988892e-05,
+      "loss": 0.6376,
+      "step": 9612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.501351001700075e-05,
+      "loss": 0.6492,
+      "step": 9613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5012567853571107e-05,
+      "loss": 0.625,
+      "step": 9614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5011625630711128e-05,
+      "loss": 0.6264,
+      "step": 9615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5010683348431992e-05,
+      "loss": 0.6841,
+      "step": 9616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5009741006744862e-05,
+      "loss": 0.6185,
+      "step": 9617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5008798605660916e-05,
+      "loss": 0.6979,
+      "step": 9618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5007856145191329e-05,
+      "loss": 0.5906,
+      "step": 9619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.500691362534727e-05,
+      "loss": 0.6479,
+      "step": 9620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5005971046139915e-05,
+      "loss": 0.6666,
+      "step": 9621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5005028407580443e-05,
+      "loss": 0.5868,
+      "step": 9622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5004085709680027e-05,
+      "loss": 0.6342,
+      "step": 9623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5003142952449845e-05,
+      "loss": 0.6242,
+      "step": 9624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5002200135901075e-05,
+      "loss": 0.6748,
+      "step": 9625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5001257260044892e-05,
+      "loss": 0.606,
+      "step": 9626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.500031432489248e-05,
+      "loss": 0.6427,
+      "step": 9627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4999371330455017e-05,
+      "loss": 0.6953,
+      "step": 9628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4998428276743679e-05,
+      "loss": 0.6525,
+      "step": 9629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4997485163769652e-05,
+      "loss": 0.6556,
+      "step": 9630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.499654199154412e-05,
+      "loss": 0.6183,
+      "step": 9631
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4995598760078258e-05,
+      "loss": 0.6257,
+      "step": 9632
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4994655469383256e-05,
+      "loss": 0.648,
+      "step": 9633
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4993712119470294e-05,
+      "loss": 0.6158,
+      "step": 9634
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4992768710350558e-05,
+      "loss": 0.6619,
+      "step": 9635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4991825242035232e-05,
+      "loss": 0.6417,
+      "step": 9636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4990881714535504e-05,
+      "loss": 0.6677,
+      "step": 9637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4989938127862558e-05,
+      "loss": 0.6462,
+      "step": 9638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4988994482027587e-05,
+      "loss": 0.6453,
+      "step": 9639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.498805077704177e-05,
+      "loss": 0.611,
+      "step": 9640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4987107012916307e-05,
+      "loss": 0.5948,
+      "step": 9641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4986163189662377e-05,
+      "loss": 0.6242,
+      "step": 9642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4985219307291177e-05,
+      "loss": 0.7034,
+      "step": 9643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4984275365813895e-05,
+      "loss": 0.6165,
+      "step": 9644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4983331365241724e-05,
+      "loss": 0.673,
+      "step": 9645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4982387305585855e-05,
+      "loss": 0.7053,
+      "step": 9646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4981443186857482e-05,
+      "loss": 0.6397,
+      "step": 9647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.49804990090678e-05,
+      "loss": 0.6804,
+      "step": 9648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4979554772228e-05,
+      "loss": 0.6417,
+      "step": 9649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.497861047634928e-05,
+      "loss": 0.6332,
+      "step": 9650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4977666121442835e-05,
+      "loss": 0.6448,
+      "step": 9651
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.497672170751986e-05,
+      "loss": 0.6741,
+      "step": 9652
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4975777234591553e-05,
+      "loss": 0.6571,
+      "step": 9653
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4974832702669116e-05,
+      "loss": 0.5846,
+      "step": 9654
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4973888111763742e-05,
+      "loss": 0.6227,
+      "step": 9655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.497294346188663e-05,
+      "loss": 0.6647,
+      "step": 9656
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4971998753048986e-05,
+      "loss": 0.6125,
+      "step": 9657
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4971053985262005e-05,
+      "loss": 0.6173,
+      "step": 9658
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4970109158536892e-05,
+      "loss": 0.659,
+      "step": 9659
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.496916427288485e-05,
+      "loss": 0.6388,
+      "step": 9660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4968219328317076e-05,
+      "loss": 0.6203,
+      "step": 9661
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4967274324844776e-05,
+      "loss": 0.6646,
+      "step": 9662
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4966329262479158e-05,
+      "loss": 0.6426,
+      "step": 9663
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4965384141231425e-05,
+      "loss": 0.6573,
+      "step": 9664
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4964438961112777e-05,
+      "loss": 0.61,
+      "step": 9665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.496349372213443e-05,
+      "loss": 0.6441,
+      "step": 9666
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4962548424307586e-05,
+      "loss": 0.6465,
+      "step": 9667
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.4961603067643449e-05,
+      "loss": 0.6254,
+      "step": 9668
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4960657652153234e-05,
+      "loss": 0.6888,
+      "step": 9669
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4959712177848146e-05,
+      "loss": 0.5963,
+      "step": 9670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4958766644739396e-05,
+      "loss": 0.6593,
+      "step": 9671
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4957821052838195e-05,
+      "loss": 0.5854,
+      "step": 9672
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4956875402155755e-05,
+      "loss": 0.6429,
+      "step": 9673
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4955929692703283e-05,
+      "loss": 0.5908,
+      "step": 9674
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4954983924492e-05,
+      "loss": 0.6236,
+      "step": 9675
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4954038097533113e-05,
+      "loss": 0.6449,
+      "step": 9676
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4953092211837831e-05,
+      "loss": 0.6587,
+      "step": 9677
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4952146267417385e-05,
+      "loss": 0.586,
+      "step": 9678
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4951200264282977e-05,
+      "loss": 0.668,
+      "step": 9679
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4950254202445822e-05,
+      "loss": 0.6099,
+      "step": 9680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4949308081917147e-05,
+      "loss": 0.6622,
+      "step": 9681
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4948361902708161e-05,
+      "loss": 0.669,
+      "step": 9682
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4947415664830085e-05,
+      "loss": 0.6325,
+      "step": 9683
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.494646936829414e-05,
+      "loss": 0.6315,
+      "step": 9684
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4945523013111542e-05,
+      "loss": 0.6173,
+      "step": 9685
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.494457659929351e-05,
+      "loss": 0.6118,
+      "step": 9686
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4943630126851272e-05,
+      "loss": 0.6288,
+      "step": 9687
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4942683595796043e-05,
+      "loss": 0.6556,
+      "step": 9688
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4941737006139045e-05,
+      "loss": 0.6252,
+      "step": 9689
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4940790357891509e-05,
+      "loss": 0.661,
+      "step": 9690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.493984365106465e-05,
+      "loss": 0.6013,
+      "step": 9691
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4938896885669694e-05,
+      "loss": 0.6687,
+      "step": 9692
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.493795006171787e-05,
+      "loss": 0.5984,
+      "step": 9693
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4937003179220402e-05,
+      "loss": 0.6283,
+      "step": 9694
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4936056238188515e-05,
+      "loss": 0.6164,
+      "step": 9695
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4935109238633437e-05,
+      "loss": 0.5823,
+      "step": 9696
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4934162180566395e-05,
+      "loss": 0.6244,
+      "step": 9697
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.493321506399862e-05,
+      "loss": 0.7202,
+      "step": 9698
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4932267888941338e-05,
+      "loss": 0.7007,
+      "step": 9699
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4931320655405785e-05,
+      "loss": 0.5885,
+      "step": 9700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4930373363403184e-05,
+      "loss": 0.6518,
+      "step": 9701
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.492942601294477e-05,
+      "loss": 0.5849,
+      "step": 9702
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4928478604041776e-05,
+      "loss": 0.6209,
+      "step": 9703
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4927531136705433e-05,
+      "loss": 0.6623,
+      "step": 9704
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4926583610946974e-05,
+      "loss": 0.6603,
+      "step": 9705
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4925636026777637e-05,
+      "loss": 0.6063,
+      "step": 9706
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.492468838420865e-05,
+      "loss": 0.6411,
+      "step": 9707
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4923740683251255e-05,
+      "loss": 0.6238,
+      "step": 9708
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4922792923916687e-05,
+      "loss": 0.6547,
+      "step": 9709
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.492184510621618e-05,
+      "loss": 0.6175,
+      "step": 9710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4920897230160971e-05,
+      "loss": 0.639,
+      "step": 9711
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4919949295762304e-05,
+      "loss": 0.6006,
+      "step": 9712
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.491900130303141e-05,
+      "loss": 0.623,
+      "step": 9713
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4918053251979535e-05,
+      "loss": 0.6377,
+      "step": 9714
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.491710514261792e-05,
+      "loss": 0.6461,
+      "step": 9715
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.49161569749578e-05,
+      "loss": 0.7113,
+      "step": 9716
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4915208749010421e-05,
+      "loss": 0.635,
+      "step": 9717
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4914260464787027e-05,
+      "loss": 0.6375,
+      "step": 9718
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4913312122298855e-05,
+      "loss": 0.5736,
+      "step": 9719
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4912363721557154e-05,
+      "loss": 0.7045,
+      "step": 9720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4911415262573168e-05,
+      "loss": 0.6477,
+      "step": 9721
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4910466745358139e-05,
+      "loss": 0.653,
+      "step": 9722
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4909518169923316e-05,
+      "loss": 0.6612,
+      "step": 9723
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4908569536279944e-05,
+      "loss": 0.6164,
+      "step": 9724
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4907620844439272e-05,
+      "loss": 0.6305,
+      "step": 9725
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4906672094412546e-05,
+      "loss": 0.6005,
+      "step": 9726
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4905723286211013e-05,
+      "loss": 0.6395,
+      "step": 9727
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4904774419845928e-05,
+      "loss": 0.598,
+      "step": 9728
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4903825495328538e-05,
+      "loss": 0.6672,
+      "step": 9729
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.490287651267009e-05,
+      "loss": 0.688,
+      "step": 9730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.490192747188184e-05,
+      "loss": 0.5966,
+      "step": 9731
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.490097837297504e-05,
+      "loss": 0.6172,
+      "step": 9732
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4900029215960942e-05,
+      "loss": 0.6336,
+      "step": 9733
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4899080000850799e-05,
+      "loss": 0.6319,
+      "step": 9734
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4898130727655863e-05,
+      "loss": 0.5606,
+      "step": 9735
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4897181396387392e-05,
+      "loss": 0.6385,
+      "step": 9736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.489623200705664e-05,
+      "loss": 0.5649,
+      "step": 9737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4895282559674869e-05,
+      "loss": 0.6452,
+      "step": 9738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4894333054253325e-05,
+      "loss": 0.592,
+      "step": 9739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4893383490803272e-05,
+      "loss": 0.6259,
+      "step": 9740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4892433869335969e-05,
+      "loss": 0.6197,
+      "step": 9741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4891484189862673e-05,
+      "loss": 0.5842,
+      "step": 9742
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4890534452394643e-05,
+      "loss": 0.6536,
+      "step": 9743
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4889584656943143e-05,
+      "loss": 0.6295,
+      "step": 9744
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4888634803519431e-05,
+      "loss": 0.6113,
+      "step": 9745
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4887684892134766e-05,
+      "loss": 0.5939,
+      "step": 9746
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4886734922800416e-05,
+      "loss": 0.6028,
+      "step": 9747
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4885784895527642e-05,
+      "loss": 0.6551,
+      "step": 9748
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4884834810327706e-05,
+      "loss": 0.6407,
+      "step": 9749
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4883884667211879e-05,
+      "loss": 0.6516,
+      "step": 9750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4882934466191414e-05,
+      "loss": 0.619,
+      "step": 9751
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4881984207277589e-05,
+      "loss": 0.6203,
+      "step": 9752
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4881033890481664e-05,
+      "loss": 0.6866,
+      "step": 9753
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4880083515814906e-05,
+      "loss": 0.6385,
+      "step": 9754
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4879133083288587e-05,
+      "loss": 0.67,
+      "step": 9755
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4878182592913973e-05,
+      "loss": 0.586,
+      "step": 9756
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4877232044702335e-05,
+      "loss": 0.6252,
+      "step": 9757
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4876281438664934e-05,
+      "loss": 0.6189,
+      "step": 9758
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4875330774813055e-05,
+      "loss": 0.6101,
+      "step": 9759
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4874380053157961e-05,
+      "loss": 0.6112,
+      "step": 9760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4873429273710924e-05,
+      "loss": 0.6159,
+      "step": 9761
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.487247843648322e-05,
+      "loss": 0.5884,
+      "step": 9762
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4871527541486122e-05,
+      "loss": 0.6222,
+      "step": 9763
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4870576588730896e-05,
+      "loss": 0.6415,
+      "step": 9764
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4869625578228828e-05,
+      "loss": 0.6589,
+      "step": 9765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4868674509991188e-05,
+      "loss": 0.6415,
+      "step": 9766
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.486772338402925e-05,
+      "loss": 0.6508,
+      "step": 9767
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4866772200354298e-05,
+      "loss": 0.6447,
+      "step": 9768
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4865820958977603e-05,
+      "loss": 0.6224,
+      "step": 9769
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4864869659910444e-05,
+      "loss": 0.5935,
+      "step": 9770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4863918303164103e-05,
+      "loss": 0.6199,
+      "step": 9771
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4862966888749859e-05,
+      "loss": 0.6077,
+      "step": 9772
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4862015416678985e-05,
+      "loss": 0.6437,
+      "step": 9773
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4861063886962774e-05,
+      "loss": 0.6134,
+      "step": 9774
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4860112299612499e-05,
+      "loss": 0.6437,
+      "step": 9775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4859160654639441e-05,
+      "loss": 0.5943,
+      "step": 9776
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4858208952054893e-05,
+      "loss": 0.6009,
+      "step": 9777
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4857257191870132e-05,
+      "loss": 0.7042,
+      "step": 9778
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4856305374096435e-05,
+      "loss": 0.6642,
+      "step": 9779
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4855353498745102e-05,
+      "loss": 0.6209,
+      "step": 9780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.485440156582741e-05,
+      "loss": 0.5917,
+      "step": 9781
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.485344957535464e-05,
+      "loss": 0.5827,
+      "step": 9782
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4852497527338094e-05,
+      "loss": 0.6099,
+      "step": 9783
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4851545421789048e-05,
+      "loss": 0.6215,
+      "step": 9784
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4850593258718793e-05,
+      "loss": 0.5637,
+      "step": 9785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.484964103813862e-05,
+      "loss": 0.6192,
+      "step": 9786
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4848688760059819e-05,
+      "loss": 0.6386,
+      "step": 9787
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4847736424493677e-05,
+      "loss": 0.6581,
+      "step": 9788
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4846784031451486e-05,
+      "loss": 0.6456,
+      "step": 9789
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4845831580944543e-05,
+      "loss": 0.6489,
+      "step": 9790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4844879072984135e-05,
+      "loss": 0.6163,
+      "step": 9791
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4843926507581556e-05,
+      "loss": 0.6782,
+      "step": 9792
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4842973884748102e-05,
+      "loss": 0.631,
+      "step": 9793
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4842021204495065e-05,
+      "loss": 0.6368,
+      "step": 9794
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4841068466833742e-05,
+      "loss": 0.6468,
+      "step": 9795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4840115671775428e-05,
+      "loss": 0.6078,
+      "step": 9796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.483916281933142e-05,
+      "loss": 0.6731,
+      "step": 9797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4838209909513012e-05,
+      "loss": 0.6269,
+      "step": 9798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4837256942331514e-05,
+      "loss": 0.6163,
+      "step": 9799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4836303917798207e-05,
+      "loss": 0.6307,
+      "step": 9800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.48353508359244e-05,
+      "loss": 0.5507,
+      "step": 9801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4834397696721396e-05,
+      "loss": 0.6993,
+      "step": 9802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.483344450020049e-05,
+      "loss": 0.6368,
+      "step": 9803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4832491246372985e-05,
+      "loss": 0.5999,
+      "step": 9804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4831537935250185e-05,
+      "loss": 0.6203,
+      "step": 9805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4830584566843387e-05,
+      "loss": 0.6309,
+      "step": 9806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.48296311411639e-05,
+      "loss": 0.6172,
+      "step": 9807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4828677658223028e-05,
+      "loss": 0.6496,
+      "step": 9808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4827724118032074e-05,
+      "loss": 0.6364,
+      "step": 9809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4826770520602341e-05,
+      "loss": 0.6546,
+      "step": 9810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.482581686594514e-05,
+      "loss": 0.6819,
+      "step": 9811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4824863154071774e-05,
+      "loss": 0.6588,
+      "step": 9812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.482390938499355e-05,
+      "loss": 0.6,
+      "step": 9813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.482295555872178e-05,
+      "loss": 0.6456,
+      "step": 9814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4822001675267772e-05,
+      "loss": 0.6362,
+      "step": 9815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4821047734642833e-05,
+      "loss": 0.6577,
+      "step": 9816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4820093736858274e-05,
+      "loss": 0.6391,
+      "step": 9817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4819139681925408e-05,
+      "loss": 0.6909,
+      "step": 9818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4818185569855541e-05,
+      "loss": 0.6151,
+      "step": 9819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4817231400659995e-05,
+      "loss": 0.6735,
+      "step": 9820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4816277174350071e-05,
+      "loss": 0.5816,
+      "step": 9821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4815322890937091e-05,
+      "loss": 0.6338,
+      "step": 9822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4814368550432368e-05,
+      "loss": 0.6367,
+      "step": 9823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4813414152847213e-05,
+      "loss": 0.6253,
+      "step": 9824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4812459698192947e-05,
+      "loss": 0.6251,
+      "step": 9825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.481150518648088e-05,
+      "loss": 0.6261,
+      "step": 9826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4810550617722336e-05,
+      "loss": 0.6246,
+      "step": 9827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4809595991928628e-05,
+      "loss": 0.6254,
+      "step": 9828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4808641309111074e-05,
+      "loss": 0.6407,
+      "step": 9829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4807686569280995e-05,
+      "loss": 0.6551,
+      "step": 9830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4806731772449712e-05,
+      "loss": 0.6378,
+      "step": 9831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.480577691862854e-05,
+      "loss": 0.6582,
+      "step": 9832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4804822007828805e-05,
+      "loss": 0.6875,
+      "step": 9833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4803867040061828e-05,
+      "loss": 0.6292,
+      "step": 9834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4802912015338928e-05,
+      "loss": 0.553,
+      "step": 9835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.480195693367143e-05,
+      "loss": 0.5683,
+      "step": 9836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4801001795070661e-05,
+      "loss": 0.614,
+      "step": 9837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4800046599547941e-05,
+      "loss": 0.548,
+      "step": 9838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4799091347114595e-05,
+      "loss": 0.6669,
+      "step": 9839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4798136037781954e-05,
+      "loss": 0.6467,
+      "step": 9840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4797180671561337e-05,
+      "loss": 0.6347,
+      "step": 9841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4796225248464073e-05,
+      "loss": 0.6694,
+      "step": 9842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4795269768501496e-05,
+      "loss": 0.6744,
+      "step": 9843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4794314231684927e-05,
+      "loss": 0.6322,
+      "step": 9844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4793358638025699e-05,
+      "loss": 0.6414,
+      "step": 9845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4792402987535144e-05,
+      "loss": 0.6129,
+      "step": 9846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4791447280224585e-05,
+      "loss": 0.6299,
+      "step": 9847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4790491516105356e-05,
+      "loss": 0.6229,
+      "step": 9848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4789535695188795e-05,
+      "loss": 0.6395,
+      "step": 9849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4788579817486226e-05,
+      "loss": 0.6161,
+      "step": 9850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4787623883008986e-05,
+      "loss": 0.6327,
+      "step": 9851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.478666789176841e-05,
+      "loss": 0.62,
+      "step": 9852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4785711843775828e-05,
+      "loss": 0.6543,
+      "step": 9853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.478475573904258e-05,
+      "loss": 0.6756,
+      "step": 9854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4783799577580001e-05,
+      "loss": 0.5924,
+      "step": 9855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4782843359399426e-05,
+      "loss": 0.5746,
+      "step": 9856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.478188708451219e-05,
+      "loss": 0.612,
+      "step": 9857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4780930752929638e-05,
+      "loss": 0.634,
+      "step": 9858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4779974364663101e-05,
+      "loss": 0.6128,
+      "step": 9859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4779017919723919e-05,
+      "loss": 0.7026,
+      "step": 9860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.477806141812344e-05,
+      "loss": 0.6368,
+      "step": 9861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4777104859872995e-05,
+      "loss": 0.6115,
+      "step": 9862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4776148244983928e-05,
+      "loss": 0.6303,
+      "step": 9863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4775191573467584e-05,
+      "loss": 0.6847,
+      "step": 9864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4774234845335303e-05,
+      "loss": 0.5724,
+      "step": 9865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4773278060598426e-05,
+      "loss": 0.6563,
+      "step": 9866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4772321219268304e-05,
+      "loss": 0.6775,
+      "step": 9867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4771364321356272e-05,
+      "loss": 0.6743,
+      "step": 9868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4770407366873682e-05,
+      "loss": 0.6289,
+      "step": 9869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4769450355831878e-05,
+      "loss": 0.6661,
+      "step": 9870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4768493288242207e-05,
+      "loss": 0.6976,
+      "step": 9871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4767536164116017e-05,
+      "loss": 0.6584,
+      "step": 9872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4766578983464654e-05,
+      "loss": 0.6189,
+      "step": 9873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4765621746299465e-05,
+      "loss": 0.6222,
+      "step": 9874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4764664452631806e-05,
+      "loss": 0.6383,
+      "step": 9875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4763707102473019e-05,
+      "loss": 0.6349,
+      "step": 9876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.476274969583446e-05,
+      "loss": 0.6371,
+      "step": 9877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4761792232727477e-05,
+      "loss": 0.6011,
+      "step": 9878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4760834713163426e-05,
+      "loss": 0.6435,
+      "step": 9879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4759877137153653e-05,
+      "loss": 0.6341,
+      "step": 9880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4758919504709517e-05,
+      "loss": 0.6621,
+      "step": 9881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4757961815842372e-05,
+      "loss": 0.5841,
+      "step": 9882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.475700407056357e-05,
+      "loss": 0.6329,
+      "step": 9883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4756046268884468e-05,
+      "loss": 0.6323,
+      "step": 9884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4755088410816418e-05,
+      "loss": 0.5998,
+      "step": 9885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4754130496370782e-05,
+      "loss": 0.6351,
+      "step": 9886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4753172525558914e-05,
+      "loss": 0.6256,
+      "step": 9887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4752214498392174e-05,
+      "loss": 0.6203,
+      "step": 9888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4751256414881918e-05,
+      "loss": 0.6279,
+      "step": 9889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4750298275039509e-05,
+      "loss": 0.6158,
+      "step": 9890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4749340078876302e-05,
+      "loss": 0.6584,
+      "step": 9891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4748381826403662e-05,
+      "loss": 0.5687,
+      "step": 9892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.474742351763295e-05,
+      "loss": 0.6637,
+      "step": 9893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4746465152575524e-05,
+      "loss": 0.6316,
+      "step": 9894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4745506731242748e-05,
+      "loss": 0.6731,
+      "step": 9895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4744548253645992e-05,
+      "loss": 0.6013,
+      "step": 9896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4743589719796612e-05,
+      "loss": 0.5539,
+      "step": 9897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4742631129705974e-05,
+      "loss": 0.6044,
+      "step": 9898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4741672483385445e-05,
+      "loss": 0.6121,
+      "step": 9899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4740713780846392e-05,
+      "loss": 0.641,
+      "step": 9900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4739755022100176e-05,
+      "loss": 0.7026,
+      "step": 9901
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4738796207158175e-05,
+      "loss": 0.6224,
+      "step": 9902
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4737837336031744e-05,
+      "loss": 0.5836,
+      "step": 9903
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4736878408732263e-05,
+      "loss": 0.6135,
+      "step": 9904
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4735919425271093e-05,
+      "loss": 0.633,
+      "step": 9905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4734960385659608e-05,
+      "loss": 0.6398,
+      "step": 9906
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4734001289909179e-05,
+      "loss": 0.6633,
+      "step": 9907
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4733042138031175e-05,
+      "loss": 0.6397,
+      "step": 9908
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4732082930036968e-05,
+      "loss": 0.6258,
+      "step": 9909
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4731123665937937e-05,
+      "loss": 0.619,
+      "step": 9910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4730164345745444e-05,
+      "loss": 0.6204,
+      "step": 9911
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4729204969470869e-05,
+      "loss": 0.6627,
+      "step": 9912
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4728245537125591e-05,
+      "loss": 0.5899,
+      "step": 9913
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.472728604872098e-05,
+      "loss": 0.6381,
+      "step": 9914
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4726326504268412e-05,
+      "loss": 0.6489,
+      "step": 9915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4725366903779264e-05,
+      "loss": 0.6137,
+      "step": 9916
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4724407247264911e-05,
+      "loss": 0.6524,
+      "step": 9917
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4723447534736738e-05,
+      "loss": 0.6513,
+      "step": 9918
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4722487766206117e-05,
+      "loss": 0.6484,
+      "step": 9919
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.472152794168443e-05,
+      "loss": 0.6329,
+      "step": 9920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4720568061183058e-05,
+      "loss": 0.5981,
+      "step": 9921
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4719608124713378e-05,
+      "loss": 0.6175,
+      "step": 9922
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4718648132286773e-05,
+      "loss": 0.6434,
+      "step": 9923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4717688083914628e-05,
+      "loss": 0.659,
+      "step": 9924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4716727979608321e-05,
+      "loss": 0.6441,
+      "step": 9925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4715767819379236e-05,
+      "loss": 0.6059,
+      "step": 9926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4714807603238762e-05,
+      "loss": 0.6485,
+      "step": 9927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4713847331198276e-05,
+      "loss": 0.6839,
+      "step": 9928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4712887003269165e-05,
+      "loss": 0.6023,
+      "step": 9929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4711926619462821e-05,
+      "loss": 0.6645,
+      "step": 9930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4710966179790624e-05,
+      "loss": 0.6435,
+      "step": 9931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4710005684263966e-05,
+      "loss": 0.6357,
+      "step": 9932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.470904513289423e-05,
+      "loss": 0.5675,
+      "step": 9933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4708084525692808e-05,
+      "loss": 0.6611,
+      "step": 9934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.470712386267109e-05,
+      "loss": 0.6291,
+      "step": 9935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4706163143840462e-05,
+      "loss": 0.6493,
+      "step": 9936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4705202369212315e-05,
+      "loss": 0.6555,
+      "step": 9937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4704241538798043e-05,
+      "loss": 0.6278,
+      "step": 9938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4703280652609037e-05,
+      "loss": 0.6531,
+      "step": 9939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4702319710656689e-05,
+      "loss": 0.6191,
+      "step": 9940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4701358712952393e-05,
+      "loss": 0.6742,
+      "step": 9941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4700397659507543e-05,
+      "loss": 0.6177,
+      "step": 9942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4699436550333532e-05,
+      "loss": 0.5482,
+      "step": 9943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4698475385441755e-05,
+      "loss": 0.6526,
+      "step": 9944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4697514164843609e-05,
+      "loss": 0.641,
+      "step": 9945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.469655288855049e-05,
+      "loss": 0.6531,
+      "step": 9946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4695591556573797e-05,
+      "loss": 0.6713,
+      "step": 9947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4694630168924926e-05,
+      "loss": 0.626,
+      "step": 9948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4693668725615272e-05,
+      "loss": 0.639,
+      "step": 9949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.469270722665624e-05,
+      "loss": 0.7042,
+      "step": 9950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.469174567205923e-05,
+      "loss": 0.6308,
+      "step": 9951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4690784061835638e-05,
+      "loss": 0.6198,
+      "step": 9952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4689822395996864e-05,
+      "loss": 0.6339,
+      "step": 9953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4688860674554317e-05,
+      "loss": 0.6036,
+      "step": 9954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4687898897519394e-05,
+      "loss": 0.6941,
+      "step": 9955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.46869370649035e-05,
+      "loss": 0.6411,
+      "step": 9956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4685975176718037e-05,
+      "loss": 0.6254,
+      "step": 9957
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.468501323297441e-05,
+      "loss": 0.5907,
+      "step": 9958
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4684051233684025e-05,
+      "loss": 0.6272,
+      "step": 9959
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4683089178858288e-05,
+      "loss": 0.6849,
+      "step": 9960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4682127068508604e-05,
+      "loss": 0.6365,
+      "step": 9961
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.468116490264638e-05,
+      "loss": 0.7223,
+      "step": 9962
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4680202681283027e-05,
+      "loss": 0.6494,
+      "step": 9963
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.467924040442995e-05,
+      "loss": 0.6024,
+      "step": 9964
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4678278072098555e-05,
+      "loss": 0.648,
+      "step": 9965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.467731568430026e-05,
+      "loss": 0.6236,
+      "step": 9966
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4676353241046468e-05,
+      "loss": 0.6064,
+      "step": 9967
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4675390742348592e-05,
+      "loss": 0.6481,
+      "step": 9968
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4674428188218049e-05,
+      "loss": 0.5998,
+      "step": 9969
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4673465578666242e-05,
+      "loss": 0.678,
+      "step": 9970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4672502913704589e-05,
+      "loss": 0.6292,
+      "step": 9971
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4671540193344507e-05,
+      "loss": 0.6227,
+      "step": 9972
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4670577417597403e-05,
+      "loss": 0.6468,
+      "step": 9973
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4669614586474694e-05,
+      "loss": 0.6949,
+      "step": 9974
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4668651699987802e-05,
+      "loss": 0.6386,
+      "step": 9975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4667688758148134e-05,
+      "loss": 0.6379,
+      "step": 9976
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4666725760967115e-05,
+      "loss": 0.5971,
+      "step": 9977
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4665762708456155e-05,
+      "loss": 0.6611,
+      "step": 9978
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4664799600626678e-05,
+      "loss": 0.6519,
+      "step": 9979
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.46638364374901e-05,
+      "loss": 0.613,
+      "step": 9980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.466287321905784e-05,
+      "loss": 0.6184,
+      "step": 9981
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4661909945341321e-05,
+      "loss": 0.6473,
+      "step": 9982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4660946616351964e-05,
+      "loss": 0.6248,
+      "step": 9983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4659983232101185e-05,
+      "loss": 0.6371,
+      "step": 9984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4659019792600411e-05,
+      "loss": 0.6531,
+      "step": 9985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4658056297861065e-05,
+      "loss": 0.6715,
+      "step": 9986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4657092747894568e-05,
+      "loss": 0.6352,
+      "step": 9987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4656129142712346e-05,
+      "loss": 0.5826,
+      "step": 9988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4655165482325821e-05,
+      "loss": 0.6761,
+      "step": 9989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4654201766746422e-05,
+      "loss": 0.7039,
+      "step": 9990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4653237995985572e-05,
+      "loss": 0.6376,
+      "step": 9991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4652274170054703e-05,
+      "loss": 0.6207,
+      "step": 9992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4651310288965235e-05,
+      "loss": 0.5697,
+      "step": 9993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4650346352728599e-05,
+      "loss": 0.6231,
+      "step": 9994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4649382361356229e-05,
+      "loss": 0.6809,
+      "step": 9995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4648418314859546e-05,
+      "loss": 0.611,
+      "step": 9996
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4647454213249983e-05,
+      "loss": 0.6764,
+      "step": 9997
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4646490056538974e-05,
+      "loss": 0.7005,
+      "step": 9998
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4645525844737944e-05,
+      "loss": 0.6246,
+      "step": 9999
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.464456157785833e-05,
+      "loss": 0.6288,
+      "step": 10000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4643597255911564e-05,
+      "loss": 0.6672,
+      "step": 10001
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4642632878909078e-05,
+      "loss": 0.623,
+      "step": 10002
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4641668446862304e-05,
+      "loss": 0.6369,
+      "step": 10003
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4640703959782682e-05,
+      "loss": 0.6539,
+      "step": 10004
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4639739417681643e-05,
+      "loss": 0.6407,
+      "step": 10005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4638774820570625e-05,
+      "loss": 0.5602,
+      "step": 10006
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4637810168461062e-05,
+      "loss": 0.7274,
+      "step": 10007
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4636845461364396e-05,
+      "loss": 0.6346,
+      "step": 10008
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4635880699292053e-05,
+      "loss": 0.6309,
+      "step": 10009
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4634915882255486e-05,
+      "loss": 0.6171,
+      "step": 10010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.463395101026613e-05,
+      "loss": 0.584,
+      "step": 10011
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4632986083335417e-05,
+      "loss": 0.6473,
+      "step": 10012
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4632021101474798e-05,
+      "loss": 0.6206,
+      "step": 10013
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4631056064695708e-05,
+      "loss": 0.6511,
+      "step": 10014
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4630090973009588e-05,
+      "loss": 0.6317,
+      "step": 10015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4629125826427883e-05,
+      "loss": 0.599,
+      "step": 10016
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.462816062496204e-05,
+      "loss": 0.6705,
+      "step": 10017
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.462719536862349e-05,
+      "loss": 0.6891,
+      "step": 10018
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4626230057423692e-05,
+      "loss": 0.6033,
+      "step": 10019
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4625264691374084e-05,
+      "loss": 0.6575,
+      "step": 10020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4624299270486106e-05,
+      "loss": 0.5686,
+      "step": 10021
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4623333794771217e-05,
+      "loss": 0.6304,
+      "step": 10022
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4622368264240859e-05,
+      "loss": 0.6323,
+      "step": 10023
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.462140267890647e-05,
+      "loss": 0.6793,
+      "step": 10024
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.462043703877951e-05,
+      "loss": 0.6523,
+      "step": 10025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4619471343871428e-05,
+      "loss": 0.6321,
+      "step": 10026
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4618505594193661e-05,
+      "loss": 0.6581,
+      "step": 10027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4617539789757676e-05,
+      "loss": 0.6506,
+      "step": 10028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4616573930574913e-05,
+      "loss": 0.5897,
+      "step": 10029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4615608016656823e-05,
+      "loss": 0.7071,
+      "step": 10030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4614642048014864e-05,
+      "loss": 0.6471,
+      "step": 10031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4613676024660488e-05,
+      "loss": 0.6842,
+      "step": 10032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4612709946605139e-05,
+      "loss": 0.6356,
+      "step": 10033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4611743813860285e-05,
+      "loss": 0.6502,
+      "step": 10034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4610777626437373e-05,
+      "loss": 0.6471,
+      "step": 10035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4609811384347858e-05,
+      "loss": 0.6052,
+      "step": 10036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4608845087603198e-05,
+      "loss": 0.6033,
+      "step": 10037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4607878736214854e-05,
+      "loss": 0.6504,
+      "step": 10038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4606912330194272e-05,
+      "loss": 0.5833,
+      "step": 10039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.460594586955292e-05,
+      "loss": 0.5938,
+      "step": 10040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4604979354302251e-05,
+      "loss": 0.6515,
+      "step": 10041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4604012784453727e-05,
+      "loss": 0.638,
+      "step": 10042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4603046160018808e-05,
+      "loss": 0.6554,
+      "step": 10043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4602079481008955e-05,
+      "loss": 0.6333,
+      "step": 10044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4601112747435626e-05,
+      "loss": 0.6262,
+      "step": 10045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4600145959310286e-05,
+      "loss": 0.5912,
+      "step": 10046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4599179116644397e-05,
+      "loss": 0.6217,
+      "step": 10047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4598212219449421e-05,
+      "loss": 0.6036,
+      "step": 10048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4597245267736821e-05,
+      "loss": 0.6861,
+      "step": 10049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4596278261518066e-05,
+      "loss": 0.6415,
+      "step": 10050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4595311200804616e-05,
+      "loss": 0.6703,
+      "step": 10051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4594344085607937e-05,
+      "loss": 0.5677,
+      "step": 10052
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4593376915939499e-05,
+      "loss": 0.6312,
+      "step": 10053
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4592409691810767e-05,
+      "loss": 0.5867,
+      "step": 10054
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4591442413233208e-05,
+      "loss": 0.6792,
+      "step": 10055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4590475080218291e-05,
+      "loss": 0.6338,
+      "step": 10056
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4589507692777484e-05,
+      "loss": 0.6363,
+      "step": 10057
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4588540250922259e-05,
+      "loss": 0.6719,
+      "step": 10058
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4587572754664088e-05,
+      "loss": 0.589,
+      "step": 10059
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4586605204014432e-05,
+      "loss": 0.6572,
+      "step": 10060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4585637598984772e-05,
+      "loss": 0.6083,
+      "step": 10061
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.458466993958658e-05,
+      "loss": 0.5982,
+      "step": 10062
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4583702225831324e-05,
+      "loss": 0.6355,
+      "step": 10063
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4582734457730479e-05,
+      "loss": 0.6915,
+      "step": 10064
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4581766635295522e-05,
+      "loss": 0.6941,
+      "step": 10065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4580798758537922e-05,
+      "loss": 0.6812,
+      "step": 10066
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4579830827469162e-05,
+      "loss": 0.6195,
+      "step": 10067
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4578862842100711e-05,
+      "loss": 0.6344,
+      "step": 10068
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.457789480244405e-05,
+      "loss": 0.6497,
+      "step": 10069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4576926708510657e-05,
+      "loss": 0.6534,
+      "step": 10070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4575958560312004e-05,
+      "loss": 0.6944,
+      "step": 10071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4574990357859575e-05,
+      "loss": 0.6053,
+      "step": 10072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4574022101164848e-05,
+      "loss": 0.5922,
+      "step": 10073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4573053790239304e-05,
+      "loss": 0.6624,
+      "step": 10074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4572085425094423e-05,
+      "loss": 0.6518,
+      "step": 10075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4571117005741684e-05,
+      "loss": 0.7102,
+      "step": 10076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.457014853219257e-05,
+      "loss": 0.6667,
+      "step": 10077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4569180004458563e-05,
+      "loss": 0.6418,
+      "step": 10078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.456821142255115e-05,
+      "loss": 0.6763,
+      "step": 10079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.456724278648181e-05,
+      "loss": 0.649,
+      "step": 10080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.456627409626203e-05,
+      "loss": 0.5909,
+      "step": 10081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4565305351903293e-05,
+      "loss": 0.658,
+      "step": 10082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4564336553417086e-05,
+      "loss": 0.5867,
+      "step": 10083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4563367700814896e-05,
+      "loss": 0.6569,
+      "step": 10084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.456239879410821e-05,
+      "loss": 0.6494,
+      "step": 10085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4561429833308513e-05,
+      "loss": 0.6581,
+      "step": 10086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4560460818427297e-05,
+      "loss": 0.6472,
+      "step": 10087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4559491749476048e-05,
+      "loss": 0.6553,
+      "step": 10088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4558522626466255e-05,
+      "loss": 0.6463,
+      "step": 10089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4557553449409411e-05,
+      "loss": 0.6478,
+      "step": 10090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.455658421831701e-05,
+      "loss": 0.5779,
+      "step": 10091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4555614933200533e-05,
+      "loss": 0.6444,
+      "step": 10092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.455464559407148e-05,
+      "loss": 0.6379,
+      "step": 10093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4553676200941342e-05,
+      "loss": 0.6499,
+      "step": 10094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.455270675382161e-05,
+      "loss": 0.6058,
+      "step": 10095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.455173725272378e-05,
+      "loss": 0.5847,
+      "step": 10096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4550767697659352e-05,
+      "loss": 0.6006,
+      "step": 10097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4549798088639813e-05,
+      "loss": 0.5841,
+      "step": 10098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.454882842567666e-05,
+      "loss": 0.613,
+      "step": 10099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4547858708781397e-05,
+      "loss": 0.624,
+      "step": 10100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4546888937965514e-05,
+      "loss": 0.6499,
+      "step": 10101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4545919113240508e-05,
+      "loss": 0.5806,
+      "step": 10102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4544949234617882e-05,
+      "loss": 0.6062,
+      "step": 10103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4543979302109135e-05,
+      "loss": 0.6135,
+      "step": 10104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.454300931572576e-05,
+      "loss": 0.6584,
+      "step": 10105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4542039275479267e-05,
+      "loss": 0.6912,
+      "step": 10106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4541069181381156e-05,
+      "loss": 0.706,
+      "step": 10107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4540099033442918e-05,
+      "loss": 0.5459,
+      "step": 10108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4539128831676066e-05,
+      "loss": 0.6603,
+      "step": 10109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4538158576092101e-05,
+      "loss": 0.6211,
+      "step": 10110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4537188266702522e-05,
+      "loss": 0.6358,
+      "step": 10111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4536217903518842e-05,
+      "loss": 0.6493,
+      "step": 10112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4535247486552555e-05,
+      "loss": 0.6194,
+      "step": 10113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4534277015815172e-05,
+      "loss": 0.6164,
+      "step": 10114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4533306491318202e-05,
+      "loss": 0.5681,
+      "step": 10115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.453233591307315e-05,
+      "loss": 0.6312,
+      "step": 10116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4531365281091518e-05,
+      "loss": 0.6229,
+      "step": 10117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.453039459538482e-05,
+      "loss": 0.5896,
+      "step": 10118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4529423855964566e-05,
+      "loss": 0.639,
+      "step": 10119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4528453062842258e-05,
+      "loss": 0.6431,
+      "step": 10120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4527482216029414e-05,
+      "loss": 0.6438,
+      "step": 10121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4526511315537541e-05,
+      "loss": 0.6397,
+      "step": 10122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4525540361378145e-05,
+      "loss": 0.6532,
+      "step": 10123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.452456935356275e-05,
+      "loss": 0.6399,
+      "step": 10124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4523598292102858e-05,
+      "loss": 0.6066,
+      "step": 10125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4522627177009985e-05,
+      "loss": 0.6156,
+      "step": 10126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4521656008295647e-05,
+      "loss": 0.7049,
+      "step": 10127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.452068478597136e-05,
+      "loss": 0.6816,
+      "step": 10128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.451971351004863e-05,
+      "loss": 0.6216,
+      "step": 10129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4518742180538985e-05,
+      "loss": 0.6118,
+      "step": 10130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4517770797453933e-05,
+      "loss": 0.6715,
+      "step": 10131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4516799360804992e-05,
+      "loss": 0.6333,
+      "step": 10132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.451582787060368e-05,
+      "loss": 0.606,
+      "step": 10133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4514856326861519e-05,
+      "loss": 0.6584,
+      "step": 10134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4513884729590025e-05,
+      "loss": 0.6787,
+      "step": 10135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4512913078800714e-05,
+      "loss": 0.682,
+      "step": 10136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4511941374505112e-05,
+      "loss": 0.7002,
+      "step": 10137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4510969616714739e-05,
+      "loss": 0.6724,
+      "step": 10138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4509997805441109e-05,
+      "loss": 0.6589,
+      "step": 10139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4509025940695756e-05,
+      "loss": 0.6554,
+      "step": 10140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4508054022490195e-05,
+      "loss": 0.6106,
+      "step": 10141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4507082050835947e-05,
+      "loss": 0.6755,
+      "step": 10142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4506110025744544e-05,
+      "loss": 0.6052,
+      "step": 10143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4505137947227506e-05,
+      "loss": 0.5701,
+      "step": 10144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4504165815296356e-05,
+      "loss": 0.6466,
+      "step": 10145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4503193629962626e-05,
+      "loss": 0.6237,
+      "step": 10146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4502221391237836e-05,
+      "loss": 0.6111,
+      "step": 10147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4501249099133516e-05,
+      "loss": 0.6638,
+      "step": 10148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4500276753661195e-05,
+      "loss": 0.6022,
+      "step": 10149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4499304354832403e-05,
+      "loss": 0.6385,
+      "step": 10150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4498331902658663e-05,
+      "loss": 0.6712,
+      "step": 10151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4497359397151507e-05,
+      "loss": 0.6183,
+      "step": 10152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4496386838322468e-05,
+      "loss": 0.6354,
+      "step": 10153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4495414226183073e-05,
+      "loss": 0.6228,
+      "step": 10154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.449444156074486e-05,
+      "loss": 0.6894,
+      "step": 10155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4493468842019351e-05,
+      "loss": 0.615,
+      "step": 10156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4492496070018088e-05,
+      "loss": 0.6556,
+      "step": 10157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.44915232447526e-05,
+      "loss": 0.6381,
+      "step": 10158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4490550366234424e-05,
+      "loss": 0.6215,
+      "step": 10159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4489577434475091e-05,
+      "loss": 0.65,
+      "step": 10160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.448860444948614e-05,
+      "loss": 0.617,
+      "step": 10161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4487631411279104e-05,
+      "loss": 0.6535,
+      "step": 10162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4486658319865521e-05,
+      "loss": 0.6945,
+      "step": 10163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4485685175256927e-05,
+      "loss": 0.6421,
+      "step": 10164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4484711977464861e-05,
+      "loss": 0.616,
+      "step": 10165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4483738726500862e-05,
+      "loss": 0.5814,
+      "step": 10166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4482765422376468e-05,
+      "loss": 0.6427,
+      "step": 10167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4481792065103219e-05,
+      "loss": 0.5951,
+      "step": 10168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4480818654692657e-05,
+      "loss": 0.6482,
+      "step": 10169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4479845191156318e-05,
+      "loss": 0.6004,
+      "step": 10170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.447887167450575e-05,
+      "loss": 0.6456,
+      "step": 10171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4477898104752495e-05,
+      "loss": 0.6394,
+      "step": 10172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.447692448190809e-05,
+      "loss": 0.6615,
+      "step": 10173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4475950805984082e-05,
+      "loss": 0.6589,
+      "step": 10174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4474977076992017e-05,
+      "loss": 0.5973,
+      "step": 10175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4474003294943436e-05,
+      "loss": 0.6298,
+      "step": 10176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4473029459849885e-05,
+      "loss": 0.6636,
+      "step": 10177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4472055571722914e-05,
+      "loss": 0.6247,
+      "step": 10178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4471081630574066e-05,
+      "loss": 0.6382,
+      "step": 10179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.447010763641489e-05,
+      "loss": 0.6822,
+      "step": 10180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4469133589256932e-05,
+      "loss": 0.6129,
+      "step": 10181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4468159489111741e-05,
+      "loss": 0.5896,
+      "step": 10182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.446718533599087e-05,
+      "loss": 0.6289,
+      "step": 10183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4466211129905864e-05,
+      "loss": 0.6272,
+      "step": 10184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4465236870868275e-05,
+      "loss": 0.6737,
+      "step": 10185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4464262558889654e-05,
+      "loss": 0.6627,
+      "step": 10186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4463288193981558e-05,
+      "loss": 0.6575,
+      "step": 10187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4462313776155528e-05,
+      "loss": 0.5965,
+      "step": 10188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4461339305423125e-05,
+      "loss": 0.7005,
+      "step": 10189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4460364781795904e-05,
+      "loss": 0.6835,
+      "step": 10190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4459390205285413e-05,
+      "loss": 0.6313,
+      "step": 10191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4458415575903211e-05,
+      "loss": 0.6799,
+      "step": 10192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4457440893660852e-05,
+      "loss": 0.6559,
+      "step": 10193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4456466158569892e-05,
+      "loss": 0.6169,
+      "step": 10194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4455491370641888e-05,
+      "loss": 0.5935,
+      "step": 10195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.44545165298884e-05,
+      "loss": 0.5657,
+      "step": 10196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4453541636320983e-05,
+      "loss": 0.5984,
+      "step": 10197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4452566689951194e-05,
+      "loss": 0.5665,
+      "step": 10198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4451591690790596e-05,
+      "loss": 0.5847,
+      "step": 10199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4450616638850745e-05,
+      "loss": 0.6322,
+      "step": 10200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4449641534143207e-05,
+      "loss": 0.6209,
+      "step": 10201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.444866637667954e-05,
+      "loss": 0.5886,
+      "step": 10202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4447691166471303e-05,
+      "loss": 0.6162,
+      "step": 10203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.444671590353006e-05,
+      "loss": 0.6167,
+      "step": 10204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4445740587867377e-05,
+      "loss": 0.648,
+      "step": 10205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4444765219494814e-05,
+      "loss": 0.6566,
+      "step": 10206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4443789798423938e-05,
+      "loss": 0.6185,
+      "step": 10207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4442814324666312e-05,
+      "loss": 0.6628,
+      "step": 10208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.44418387982335e-05,
+      "loss": 0.6841,
+      "step": 10209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4440863219137074e-05,
+      "loss": 0.6444,
+      "step": 10210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4439887587388595e-05,
+      "loss": 0.6475,
+      "step": 10211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4438911902999633e-05,
+      "loss": 0.6006,
+      "step": 10212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4437936165981752e-05,
+      "loss": 0.6265,
+      "step": 10213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4436960376346527e-05,
+      "loss": 0.6156,
+      "step": 10214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4435984534105522e-05,
+      "loss": 0.5972,
+      "step": 10215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4435008639270312e-05,
+      "loss": 0.6759,
+      "step": 10216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4434032691852463e-05,
+      "loss": 0.6811,
+      "step": 10217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4433056691863545e-05,
+      "loss": 0.6585,
+      "step": 10218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4432080639315131e-05,
+      "loss": 0.6364,
+      "step": 10219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.44311045342188e-05,
+      "loss": 0.6815,
+      "step": 10220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4430128376586114e-05,
+      "loss": 0.5986,
+      "step": 10221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4429152166428655e-05,
+      "loss": 0.6235,
+      "step": 10222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4428175903757993e-05,
+      "loss": 0.6271,
+      "step": 10223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4427199588585706e-05,
+      "loss": 0.5914,
+      "step": 10224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4426223220923366e-05,
+      "loss": 0.6809,
+      "step": 10225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4425246800782549e-05,
+      "loss": 0.6691,
+      "step": 10226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4424270328174834e-05,
+      "loss": 0.5852,
+      "step": 10227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4423293803111799e-05,
+      "loss": 0.5891,
+      "step": 10228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4422317225605019e-05,
+      "loss": 0.6848,
+      "step": 10229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4421340595666074e-05,
+      "loss": 0.5899,
+      "step": 10230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4420363913306545e-05,
+      "loss": 0.5838,
+      "step": 10231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4419387178538008e-05,
+      "loss": 0.6072,
+      "step": 10232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4418410391372046e-05,
+      "loss": 0.6771,
+      "step": 10233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4417433551820241e-05,
+      "loss": 0.5519,
+      "step": 10234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4416456659894172e-05,
+      "loss": 0.6159,
+      "step": 10235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.441547971560542e-05,
+      "loss": 0.6803,
+      "step": 10236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4414502718965575e-05,
+      "loss": 0.6687,
+      "step": 10237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4413525669986212e-05,
+      "loss": 0.7016,
+      "step": 10238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4412548568678921e-05,
+      "loss": 0.619,
+      "step": 10239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4411571415055285e-05,
+      "loss": 0.6438,
+      "step": 10240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4410594209126887e-05,
+      "loss": 0.6663,
+      "step": 10241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4409616950905316e-05,
+      "loss": 0.647,
+      "step": 10242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.440863964040216e-05,
+      "loss": 0.6509,
+      "step": 10243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4407662277629003e-05,
+      "loss": 0.65,
+      "step": 10244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4406684862597434e-05,
+      "loss": 0.6534,
+      "step": 10245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4405707395319042e-05,
+      "loss": 0.6421,
+      "step": 10246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4404729875805414e-05,
+      "loss": 0.6838,
+      "step": 10247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4403752304068141e-05,
+      "loss": 0.69,
+      "step": 10248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4402774680118816e-05,
+      "loss": 0.6939,
+      "step": 10249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4401797003969025e-05,
+      "loss": 0.6017,
+      "step": 10250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4400819275630364e-05,
+      "loss": 0.6591,
+      "step": 10251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4399841495114423e-05,
+      "loss": 0.6531,
+      "step": 10252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4398863662432797e-05,
+      "loss": 0.6389,
+      "step": 10253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4397885777597073e-05,
+      "loss": 0.6553,
+      "step": 10254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4396907840618855e-05,
+      "loss": 0.6607,
+      "step": 10255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4395929851509731e-05,
+      "loss": 0.6665,
+      "step": 10256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4394951810281295e-05,
+      "loss": 0.6465,
+      "step": 10257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4393973716945149e-05,
+      "loss": 0.6479,
+      "step": 10258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4392995571512886e-05,
+      "loss": 0.6449,
+      "step": 10259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4392017373996099e-05,
+      "loss": 0.6467,
+      "step": 10260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4391039124406394e-05,
+      "loss": 0.6043,
+      "step": 10261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4390060822755367e-05,
+      "loss": 0.6193,
+      "step": 10262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.438908246905461e-05,
+      "loss": 0.605,
+      "step": 10263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4388104063315733e-05,
+      "loss": 0.5687,
+      "step": 10264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4387125605550329e-05,
+      "loss": 0.6481,
+      "step": 10265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.438614709577e-05,
+      "loss": 0.6256,
+      "step": 10266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4385168533986353e-05,
+      "loss": 0.6267,
+      "step": 10267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4384189920210985e-05,
+      "loss": 0.5785,
+      "step": 10268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4383211254455494e-05,
+      "loss": 0.6261,
+      "step": 10269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4382232536731496e-05,
+      "loss": 0.6631,
+      "step": 10270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4381253767050583e-05,
+      "loss": 0.6445,
+      "step": 10271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4380274945424365e-05,
+      "loss": 0.609,
+      "step": 10272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4379296071864446e-05,
+      "loss": 0.6068,
+      "step": 10273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4378317146382434e-05,
+      "loss": 0.613,
+      "step": 10274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4377338168989933e-05,
+      "loss": 0.6381,
+      "step": 10275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.437635913969855e-05,
+      "loss": 0.6645,
+      "step": 10276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4375380058519895e-05,
+      "loss": 0.651,
+      "step": 10277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4374400925465573e-05,
+      "loss": 0.6405,
+      "step": 10278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4373421740547193e-05,
+      "loss": 0.6217,
+      "step": 10279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4372442503776369e-05,
+      "loss": 0.6217,
+      "step": 10280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4371463215164707e-05,
+      "loss": 0.6481,
+      "step": 10281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4370483874723817e-05,
+      "loss": 0.6079,
+      "step": 10282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4369504482465314e-05,
+      "loss": 0.6557,
+      "step": 10283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4368525038400808e-05,
+      "loss": 0.6195,
+      "step": 10284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4367545542541908e-05,
+      "loss": 0.6377,
+      "step": 10285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4366565994900234e-05,
+      "loss": 0.5845,
+      "step": 10286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4365586395487395e-05,
+      "loss": 0.6108,
+      "step": 10287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4364606744315006e-05,
+      "loss": 0.61,
+      "step": 10288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4363627041394686e-05,
+      "loss": 0.6339,
+      "step": 10289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4362647286738045e-05,
+      "loss": 0.5748,
+      "step": 10290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.43616674803567e-05,
+      "loss": 0.6341,
+      "step": 10291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4360687622262273e-05,
+      "loss": 0.6117,
+      "step": 10292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4359707712466375e-05,
+      "loss": 0.6316,
+      "step": 10293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.435872775098063e-05,
+      "loss": 0.6365,
+      "step": 10294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4357747737816655e-05,
+      "loss": 0.69,
+      "step": 10295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4356767672986063e-05,
+      "loss": 0.6593,
+      "step": 10296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.435578755650048e-05,
+      "loss": 0.6512,
+      "step": 10297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4354807388371528e-05,
+      "loss": 0.6281,
+      "step": 10298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4353827168610823e-05,
+      "loss": 0.6364,
+      "step": 10299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.435284689722999e-05,
+      "loss": 0.623,
+      "step": 10300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4351866574240651e-05,
+      "loss": 0.605,
+      "step": 10301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4350886199654431e-05,
+      "loss": 0.5913,
+      "step": 10302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4349905773482945e-05,
+      "loss": 0.6379,
+      "step": 10303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.434892529573783e-05,
+      "loss": 0.6112,
+      "step": 10304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4347944766430702e-05,
+      "loss": 0.6299,
+      "step": 10305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4346964185573184e-05,
+      "loss": 0.6677,
+      "step": 10306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4345983553176914e-05,
+      "loss": 0.579,
+      "step": 10307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4345002869253506e-05,
+      "loss": 0.5946,
+      "step": 10308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4344022133814594e-05,
+      "loss": 0.6198,
+      "step": 10309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4343041346871807e-05,
+      "loss": 0.6209,
+      "step": 10310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4342060508436768e-05,
+      "loss": 0.6296,
+      "step": 10311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.434107961852111e-05,
+      "loss": 0.6445,
+      "step": 10312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.434009867713646e-05,
+      "loss": 0.6366,
+      "step": 10313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4339117684294453e-05,
+      "loss": 0.663,
+      "step": 10314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4338136640006712e-05,
+      "loss": 0.5345,
+      "step": 10315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.433715554428488e-05,
+      "loss": 0.677,
+      "step": 10316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4336174397140576e-05,
+      "loss": 0.6828,
+      "step": 10317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4335193198585443e-05,
+      "loss": 0.657,
+      "step": 10318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4334211948631111e-05,
+      "loss": 0.653,
+      "step": 10319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4333230647289211e-05,
+      "loss": 0.6265,
+      "step": 10320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4332249294571383e-05,
+      "loss": 0.6272,
+      "step": 10321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4331267890489257e-05,
+      "loss": 0.5772,
+      "step": 10322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4330286435054471e-05,
+      "loss": 0.6236,
+      "step": 10323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4329304928278665e-05,
+      "loss": 0.6419,
+      "step": 10324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4328323370173467e-05,
+      "loss": 0.63,
+      "step": 10325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4327341760750524e-05,
+      "loss": 0.6453,
+      "step": 10326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4326360100021468e-05,
+      "loss": 0.6572,
+      "step": 10327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4325378387997941e-05,
+      "loss": 0.668,
+      "step": 10328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4324396624691581e-05,
+      "loss": 0.6568,
+      "step": 10329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4323414810114031e-05,
+      "loss": 0.6114,
+      "step": 10330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4322432944276927e-05,
+      "loss": 0.6324,
+      "step": 10331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4321451027191913e-05,
+      "loss": 0.6446,
+      "step": 10332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4320469058870633e-05,
+      "loss": 0.7105,
+      "step": 10333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4319487039324723e-05,
+      "loss": 0.6655,
+      "step": 10334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4318504968565832e-05,
+      "loss": 0.6413,
+      "step": 10335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.43175228466056e-05,
+      "loss": 0.6827,
+      "step": 10336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4316540673455675e-05,
+      "loss": 0.6467,
+      "step": 10337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4315558449127698e-05,
+      "loss": 0.6553,
+      "step": 10338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4314576173633319e-05,
+      "loss": 0.6719,
+      "step": 10339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4313593846984179e-05,
+      "loss": 0.6332,
+      "step": 10340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4312611469191928e-05,
+      "loss": 0.6271,
+      "step": 10341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4311629040268214e-05,
+      "loss": 0.671,
+      "step": 10342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4310646560224682e-05,
+      "loss": 0.621,
+      "step": 10343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.430966402907298e-05,
+      "loss": 0.6129,
+      "step": 10344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4308681446824764e-05,
+      "loss": 0.5969,
+      "step": 10345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4307698813491676e-05,
+      "loss": 0.6021,
+      "step": 10346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4306716129085367e-05,
+      "loss": 0.5758,
+      "step": 10347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4305733393617494e-05,
+      "loss": 0.6085,
+      "step": 10348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4304750607099705e-05,
+      "loss": 0.5801,
+      "step": 10349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4303767769543649e-05,
+      "loss": 0.6254,
+      "step": 10350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4302784880960985e-05,
+      "loss": 0.6402,
+      "step": 10351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4301801941363363e-05,
+      "loss": 0.6245,
+      "step": 10352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4300818950762435e-05,
+      "loss": 0.6581,
+      "step": 10353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4299835909169857e-05,
+      "loss": 0.65,
+      "step": 10354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4298852816597288e-05,
+      "loss": 0.5855,
+      "step": 10355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4297869673056377e-05,
+      "loss": 0.6822,
+      "step": 10356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4296886478558789e-05,
+      "loss": 0.6395,
+      "step": 10357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4295903233116172e-05,
+      "loss": 0.5695,
+      "step": 10358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4294919936740189e-05,
+      "loss": 0.6164,
+      "step": 10359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4293936589442499e-05,
+      "loss": 0.5871,
+      "step": 10360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4292953191234759e-05,
+      "loss": 0.6307,
+      "step": 10361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4291969742128625e-05,
+      "loss": 0.6327,
+      "step": 10362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4290986242135763e-05,
+      "loss": 0.6473,
+      "step": 10363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4290002691267831e-05,
+      "loss": 0.6319,
+      "step": 10364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4289019089536487e-05,
+      "loss": 0.6173,
+      "step": 10365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.42880354369534e-05,
+      "loss": 0.6907,
+      "step": 10366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.428705173353023e-05,
+      "loss": 0.6176,
+      "step": 10367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4286067979278631e-05,
+      "loss": 0.6073,
+      "step": 10368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.428508417421028e-05,
+      "loss": 0.6136,
+      "step": 10369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4284100318336836e-05,
+      "loss": 0.5826,
+      "step": 10370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4283116411669956e-05,
+      "loss": 0.6264,
+      "step": 10371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.428213245422132e-05,
+      "loss": 0.5915,
+      "step": 10372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4281148446002589e-05,
+      "loss": 0.6432,
+      "step": 10373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.428016438702542e-05,
+      "loss": 0.5966,
+      "step": 10374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4279180277301493e-05,
+      "loss": 0.7162,
+      "step": 10375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4278196116842468e-05,
+      "loss": 0.6615,
+      "step": 10376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4277211905660015e-05,
+      "loss": 0.6531,
+      "step": 10377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4276227643765807e-05,
+      "loss": 0.6114,
+      "step": 10378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4275243331171513e-05,
+      "loss": 0.6118,
+      "step": 10379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4274258967888797e-05,
+      "loss": 0.6298,
+      "step": 10380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4273274553929335e-05,
+      "loss": 0.6565,
+      "step": 10381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4272290089304796e-05,
+      "loss": 0.6455,
+      "step": 10382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4271305574026856e-05,
+      "loss": 0.6146,
+      "step": 10383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4270321008107185e-05,
+      "loss": 0.6082,
+      "step": 10384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4269336391557456e-05,
+      "loss": 0.6483,
+      "step": 10385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4268351724389342e-05,
+      "loss": 0.6219,
+      "step": 10386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4267367006614519e-05,
+      "loss": 0.6292,
+      "step": 10387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4266382238244664e-05,
+      "loss": 0.5824,
+      "step": 10388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4265397419291449e-05,
+      "loss": 0.6292,
+      "step": 10389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4264412549766551e-05,
+      "loss": 0.6351,
+      "step": 10390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.426342762968165e-05,
+      "loss": 0.6609,
+      "step": 10391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4262442659048419e-05,
+      "loss": 0.6349,
+      "step": 10392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.426145763787854e-05,
+      "loss": 0.6738,
+      "step": 10393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.426047256618369e-05,
+      "loss": 0.6708,
+      "step": 10394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4259487443975546e-05,
+      "loss": 0.6791,
+      "step": 10395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.425850227126579e-05,
+      "loss": 0.6636,
+      "step": 10396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4257517048066108e-05,
+      "loss": 0.5987,
+      "step": 10397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4256531774388169e-05,
+      "loss": 0.6028,
+      "step": 10398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4255546450243665e-05,
+      "loss": 0.6374,
+      "step": 10399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4254561075644273e-05,
+      "loss": 0.6203,
+      "step": 10400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4253575650601677e-05,
+      "loss": 0.6457,
+      "step": 10401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4252590175127562e-05,
+      "loss": 0.613,
+      "step": 10402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4251604649233612e-05,
+      "loss": 0.6504,
+      "step": 10403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.425061907293151e-05,
+      "loss": 0.5912,
+      "step": 10404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4249633446232937e-05,
+      "loss": 0.6707,
+      "step": 10405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.424864776914959e-05,
+      "loss": 0.6375,
+      "step": 10406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4247662041693147e-05,
+      "loss": 0.6312,
+      "step": 10407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4246676263875296e-05,
+      "loss": 0.6015,
+      "step": 10408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4245690435707728e-05,
+      "loss": 0.6278,
+      "step": 10409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4244704557202127e-05,
+      "loss": 0.6799,
+      "step": 10410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4243718628370187e-05,
+      "loss": 0.5768,
+      "step": 10411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4242732649223592e-05,
+      "loss": 0.6157,
+      "step": 10412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4241746619774033e-05,
+      "loss": 0.619,
+      "step": 10413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4240760540033204e-05,
+      "loss": 0.5696,
+      "step": 10414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4239774410012794e-05,
+      "loss": 0.6541,
+      "step": 10415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4238788229724494e-05,
+      "loss": 0.6158,
+      "step": 10416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.423780199918e-05,
+      "loss": 0.6052,
+      "step": 10417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4236815718390999e-05,
+      "loss": 0.6011,
+      "step": 10418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.423582938736919e-05,
+      "loss": 0.648,
+      "step": 10419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4234843006126265e-05,
+      "loss": 0.6317,
+      "step": 10420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4233856574673918e-05,
+      "loss": 0.6141,
+      "step": 10421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4232870093023845e-05,
+      "loss": 0.6667,
+      "step": 10422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4231883561187747e-05,
+      "loss": 0.6248,
+      "step": 10423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.423089697917731e-05,
+      "loss": 0.652,
+      "step": 10424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4229910347004241e-05,
+      "loss": 0.6416,
+      "step": 10425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4228923664680235e-05,
+      "loss": 0.6136,
+      "step": 10426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4227936932216986e-05,
+      "loss": 0.6797,
+      "step": 10427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4226950149626198e-05,
+      "loss": 0.6129,
+      "step": 10428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4225963316919569e-05,
+      "loss": 0.6285,
+      "step": 10429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4224976434108796e-05,
+      "loss": 0.6156,
+      "step": 10430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4223989501205586e-05,
+      "loss": 0.6048,
+      "step": 10431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4223002518221638e-05,
+      "loss": 0.6523,
+      "step": 10432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4222015485168649e-05,
+      "loss": 0.6432,
+      "step": 10433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4221028402058329e-05,
+      "loss": 0.629,
+      "step": 10434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4220041268902378e-05,
+      "loss": 0.6358,
+      "step": 10435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4219054085712495e-05,
+      "loss": 0.6381,
+      "step": 10436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4218066852500391e-05,
+      "loss": 0.6309,
+      "step": 10437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4217079569277771e-05,
+      "loss": 0.6936,
+      "step": 10438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4216092236056338e-05,
+      "loss": 0.6094,
+      "step": 10439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4215104852847794e-05,
+      "loss": 0.679,
+      "step": 10440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4214117419663853e-05,
+      "loss": 0.6018,
+      "step": 10441
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4213129936516219e-05,
+      "loss": 0.6242,
+      "step": 10442
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.42121424034166e-05,
+      "loss": 0.6087,
+      "step": 10443
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4211154820376705e-05,
+      "loss": 0.712,
+      "step": 10444
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.421016718740824e-05,
+      "loss": 0.6286,
+      "step": 10445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4209179504522919e-05,
+      "loss": 0.6323,
+      "step": 10446
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.420819177173245e-05,
+      "loss": 0.6035,
+      "step": 10447
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4207203989048544e-05,
+      "loss": 0.6812,
+      "step": 10448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.420621615648291e-05,
+      "loss": 0.6713,
+      "step": 10449
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4205228274047267e-05,
+      "loss": 0.6315,
+      "step": 10450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4204240341753321e-05,
+      "loss": 0.6407,
+      "step": 10451
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4203252359612787e-05,
+      "loss": 0.5906,
+      "step": 10452
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.420226432763738e-05,
+      "loss": 0.632,
+      "step": 10453
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4201276245838812e-05,
+      "loss": 0.6074,
+      "step": 10454
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4200288114228799e-05,
+      "loss": 0.6328,
+      "step": 10455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4199299932819062e-05,
+      "loss": 0.6733,
+      "step": 10456
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4198311701621308e-05,
+      "loss": 0.6147,
+      "step": 10457
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4197323420647258e-05,
+      "loss": 0.6128,
+      "step": 10458
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4196335089908631e-05,
+      "loss": 0.5887,
+      "step": 10459
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4195346709417141e-05,
+      "loss": 0.6764,
+      "step": 10460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4194358279184508e-05,
+      "loss": 0.65,
+      "step": 10461
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4193369799222456e-05,
+      "loss": 0.6304,
+      "step": 10462
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4192381269542697e-05,
+      "loss": 0.671,
+      "step": 10463
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4191392690156956e-05,
+      "loss": 0.6125,
+      "step": 10464
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4190404061076952e-05,
+      "loss": 0.6411,
+      "step": 10465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4189415382314407e-05,
+      "loss": 0.6789,
+      "step": 10466
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.418842665388104e-05,
+      "loss": 0.6514,
+      "step": 10467
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4187437875788581e-05,
+      "loss": 0.6005,
+      "step": 10468
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.418644904804875e-05,
+      "loss": 0.6062,
+      "step": 10469
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4185460170673263e-05,
+      "loss": 0.6622,
+      "step": 10470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4184471243673859e-05,
+      "loss": 0.6373,
+      "step": 10471
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.418348226706225e-05,
+      "loss": 0.6313,
+      "step": 10472
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4182493240850169e-05,
+      "loss": 0.6579,
+      "step": 10473
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4181504165049336e-05,
+      "loss": 0.6454,
+      "step": 10474
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4180515039671485e-05,
+      "loss": 0.6082,
+      "step": 10475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4179525864728338e-05,
+      "loss": 0.6902,
+      "step": 10476
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4178536640231625e-05,
+      "loss": 0.6198,
+      "step": 10477
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4177547366193076e-05,
+      "loss": 0.5814,
+      "step": 10478
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4176558042624418e-05,
+      "loss": 0.6171,
+      "step": 10479
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4175568669537379e-05,
+      "loss": 0.6463,
+      "step": 10480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4174579246943693e-05,
+      "loss": 0.6133,
+      "step": 10481
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4173589774855087e-05,
+      "loss": 0.6176,
+      "step": 10482
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4172600253283296e-05,
+      "loss": 0.6017,
+      "step": 10483
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4171610682240054e-05,
+      "loss": 0.5933,
+      "step": 10484
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4170621061737084e-05,
+      "loss": 0.6267,
+      "step": 10485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4169631391786129e-05,
+      "loss": 0.6105,
+      "step": 10486
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.416864167239892e-05,
+      "loss": 0.6237,
+      "step": 10487
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.416765190358719e-05,
+      "loss": 0.6021,
+      "step": 10488
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4166662085362673e-05,
+      "loss": 0.6587,
+      "step": 10489
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4165672217737108e-05,
+      "loss": 0.62,
+      "step": 10490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4164682300722228e-05,
+      "loss": 0.5843,
+      "step": 10491
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.416369233432977e-05,
+      "loss": 0.638,
+      "step": 10492
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4162702318571473e-05,
+      "loss": 0.6897,
+      "step": 10493
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4161712253459075e-05,
+      "loss": 0.652,
+      "step": 10494
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4160722139004312e-05,
+      "loss": 0.5983,
+      "step": 10495
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4159731975218928e-05,
+      "loss": 0.6619,
+      "step": 10496
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4158741762114656e-05,
+      "loss": 0.5781,
+      "step": 10497
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4157751499703241e-05,
+      "loss": 0.6553,
+      "step": 10498
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4156761187996425e-05,
+      "loss": 0.6358,
+      "step": 10499
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4155770827005943e-05,
+      "loss": 0.6291,
+      "step": 10500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4154780416743541e-05,
+      "loss": 0.6042,
+      "step": 10501
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4153789957220964e-05,
+      "loss": 0.6911,
+      "step": 10502
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.415279944844995e-05,
+      "loss": 0.6703,
+      "step": 10503
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4151808890442247e-05,
+      "loss": 0.6829,
+      "step": 10504
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4150818283209595e-05,
+      "loss": 0.6876,
+      "step": 10505
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4149827626763744e-05,
+      "loss": 0.6137,
+      "step": 10506
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4148836921116435e-05,
+      "loss": 0.6552,
+      "step": 10507
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4147846166279418e-05,
+      "loss": 0.6458,
+      "step": 10508
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4146855362264435e-05,
+      "loss": 0.6009,
+      "step": 10509
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4145864509083238e-05,
+      "loss": 0.6189,
+      "step": 10510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4144873606747571e-05,
+      "loss": 0.6423,
+      "step": 10511
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4143882655269185e-05,
+      "loss": 0.6179,
+      "step": 10512
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4142891654659829e-05,
+      "loss": 0.5811,
+      "step": 10513
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4141900604931252e-05,
+      "loss": 0.6129,
+      "step": 10514
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4140909506095201e-05,
+      "loss": 0.6075,
+      "step": 10515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4139918358163434e-05,
+      "loss": 0.5947,
+      "step": 10516
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4138927161147693e-05,
+      "loss": 0.601,
+      "step": 10517
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4137935915059737e-05,
+      "loss": 0.6222,
+      "step": 10518
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4136944619911317e-05,
+      "loss": 0.6348,
+      "step": 10519
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4135953275714184e-05,
+      "loss": 0.6178,
+      "step": 10520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4134961882480091e-05,
+      "loss": 0.6376,
+      "step": 10521
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4133970440220799e-05,
+      "loss": 0.609,
+      "step": 10522
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4132978948948055e-05,
+      "loss": 0.6565,
+      "step": 10523
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4131987408673618e-05,
+      "loss": 0.6222,
+      "step": 10524
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4130995819409243e-05,
+      "loss": 0.6863,
+      "step": 10525
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4130004181166686e-05,
+      "loss": 0.5793,
+      "step": 10526
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4129012493957706e-05,
+      "loss": 0.5768,
+      "step": 10527
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.412802075779406e-05,
+      "loss": 0.616,
+      "step": 10528
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4127028972687504e-05,
+      "loss": 0.6982,
+      "step": 10529
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.41260371386498e-05,
+      "loss": 0.6356,
+      "step": 10530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4125045255692707e-05,
+      "loss": 0.7041,
+      "step": 10531
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4124053323827983e-05,
+      "loss": 0.6457,
+      "step": 10532
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.412306134306739e-05,
+      "loss": 0.6115,
+      "step": 10533
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.412206931342269e-05,
+      "loss": 0.6092,
+      "step": 10534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4121077234905643e-05,
+      "loss": 0.6598,
+      "step": 10535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4120085107528012e-05,
+      "loss": 0.6603,
+      "step": 10536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4119092931301563e-05,
+      "loss": 0.6415,
+      "step": 10537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4118100706238051e-05,
+      "loss": 0.673,
+      "step": 10538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.411710843234925e-05,
+      "loss": 0.6301,
+      "step": 10539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.411611610964692e-05,
+      "loss": 0.657,
+      "step": 10540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4115123738142826e-05,
+      "loss": 0.6369,
+      "step": 10541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4114131317848733e-05,
+      "loss": 0.6049,
+      "step": 10542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4113138848776413e-05,
+      "loss": 0.6514,
+      "step": 10543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4112146330937627e-05,
+      "loss": 0.6383,
+      "step": 10544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.411115376434414e-05,
+      "loss": 0.6643,
+      "step": 10545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4110161149007731e-05,
+      "loss": 0.5871,
+      "step": 10546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4109168484940157e-05,
+      "loss": 0.6488,
+      "step": 10547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4108175772153195e-05,
+      "loss": 0.6158,
+      "step": 10548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4107183010658612e-05,
+      "loss": 0.6425,
+      "step": 10549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4106190200468177e-05,
+      "loss": 0.6286,
+      "step": 10550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4105197341593666e-05,
+      "loss": 0.5698,
+      "step": 10551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4104204434046847e-05,
+      "loss": 0.6476,
+      "step": 10552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4103211477839491e-05,
+      "loss": 0.6236,
+      "step": 10553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4102218472983372e-05,
+      "loss": 0.637,
+      "step": 10554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4101225419490265e-05,
+      "loss": 0.5853,
+      "step": 10555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4100232317371943e-05,
+      "loss": 0.6633,
+      "step": 10556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4099239166640177e-05,
+      "loss": 0.6037,
+      "step": 10557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.409824596730675e-05,
+      "loss": 0.5872,
+      "step": 10558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.409725271938343e-05,
+      "loss": 0.6128,
+      "step": 10559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4096259422881992e-05,
+      "loss": 0.6356,
+      "step": 10560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4095266077814224e-05,
+      "loss": 0.6709,
+      "step": 10561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4094272684191891e-05,
+      "loss": 0.5922,
+      "step": 10562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4093279242026778e-05,
+      "loss": 0.6726,
+      "step": 10563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4092285751330663e-05,
+      "loss": 0.6256,
+      "step": 10564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4091292212115322e-05,
+      "loss": 0.6301,
+      "step": 10565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4090298624392538e-05,
+      "loss": 0.5409,
+      "step": 10566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4089304988174086e-05,
+      "loss": 0.6865,
+      "step": 10567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4088311303471751e-05,
+      "loss": 0.5843,
+      "step": 10568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4087317570297319e-05,
+      "loss": 0.6176,
+      "step": 10569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.408632378866256e-05,
+      "loss": 0.6183,
+      "step": 10570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4085329958579266e-05,
+      "loss": 0.608,
+      "step": 10571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.408433608005922e-05,
+      "loss": 0.6016,
+      "step": 10572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.40833421531142e-05,
+      "loss": 0.6069,
+      "step": 10573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4082348177755992e-05,
+      "loss": 0.6427,
+      "step": 10574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4081354153996385e-05,
+      "loss": 0.6425,
+      "step": 10575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.408036008184716e-05,
+      "loss": 0.6472,
+      "step": 10576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4079365961320105e-05,
+      "loss": 0.6319,
+      "step": 10577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4078371792427008e-05,
+      "loss": 0.6475,
+      "step": 10578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4077377575179651e-05,
+      "loss": 0.6302,
+      "step": 10579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4076383309589825e-05,
+      "loss": 0.6499,
+      "step": 10580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4075388995669321e-05,
+      "loss": 0.58,
+      "step": 10581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4074394633429924e-05,
+      "loss": 0.6367,
+      "step": 10582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4073400222883425e-05,
+      "loss": 0.6459,
+      "step": 10583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4072405764041612e-05,
+      "loss": 0.6553,
+      "step": 10584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4071411256916276e-05,
+      "loss": 0.6229,
+      "step": 10585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4070416701519211e-05,
+      "loss": 0.6811,
+      "step": 10586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4069422097862209e-05,
+      "loss": 0.5607,
+      "step": 10587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4068427445957058e-05,
+      "loss": 0.6911,
+      "step": 10588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.406743274581555e-05,
+      "loss": 0.6348,
+      "step": 10589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4066437997449488e-05,
+      "loss": 0.6213,
+      "step": 10590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4065443200870655e-05,
+      "loss": 0.6005,
+      "step": 10591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.406444835609085e-05,
+      "loss": 0.6762,
+      "step": 10592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.406345346312187e-05,
+      "loss": 0.6445,
+      "step": 10593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4062458521975507e-05,
+      "loss": 0.6112,
+      "step": 10594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.406146353266356e-05,
+      "loss": 0.6223,
+      "step": 10595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4060468495197825e-05,
+      "loss": 0.6131,
+      "step": 10596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4059473409590103e-05,
+      "loss": 0.6326,
+      "step": 10597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4058478275852184e-05,
+      "loss": 0.6503,
+      "step": 10598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4057483093995872e-05,
+      "loss": 0.6453,
+      "step": 10599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4056487864032966e-05,
+      "loss": 0.6197,
+      "step": 10600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4055492585975263e-05,
+      "loss": 0.6673,
+      "step": 10601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4054497259834567e-05,
+      "loss": 0.6669,
+      "step": 10602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4053501885622679e-05,
+      "loss": 0.5629,
+      "step": 10603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4052506463351392e-05,
+      "loss": 0.5705,
+      "step": 10604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.405151099303252e-05,
+      "loss": 0.6158,
+      "step": 10605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.405051547467786e-05,
+      "loss": 0.6093,
+      "step": 10606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.404951990829921e-05,
+      "loss": 0.6201,
+      "step": 10607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4048524293908383e-05,
+      "loss": 0.7032,
+      "step": 10608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.404752863151718e-05,
+      "loss": 0.6484,
+      "step": 10609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4046532921137401e-05,
+      "loss": 0.6452,
+      "step": 10610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4045537162780858e-05,
+      "loss": 0.6984,
+      "step": 10611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4044541356459357e-05,
+      "loss": 0.6622,
+      "step": 10612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4043545502184696e-05,
+      "loss": 0.5997,
+      "step": 10613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.404254959996869e-05,
+      "loss": 0.5988,
+      "step": 10614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.404155364982315e-05,
+      "loss": 0.6128,
+      "step": 10615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.404055765175987e-05,
+      "loss": 0.6578,
+      "step": 10616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4039561605790673e-05,
+      "loss": 0.6177,
+      "step": 10617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4038565511927364e-05,
+      "loss": 0.7078,
+      "step": 10618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4037569370181748e-05,
+      "loss": 0.6363,
+      "step": 10619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4036573180565642e-05,
+      "loss": 0.6118,
+      "step": 10620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4035576943090857e-05,
+      "loss": 0.6402,
+      "step": 10621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4034580657769196e-05,
+      "loss": 0.654,
+      "step": 10622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4033584324612482e-05,
+      "loss": 0.6573,
+      "step": 10623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4032587943632523e-05,
+      "loss": 0.6249,
+      "step": 10624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.403159151484113e-05,
+      "loss": 0.6003,
+      "step": 10625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4030595038250123e-05,
+      "loss": 0.6259,
+      "step": 10626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4029598513871314e-05,
+      "loss": 0.6736,
+      "step": 10627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4028601941716514e-05,
+      "loss": 0.5895,
+      "step": 10628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4027605321797542e-05,
+      "loss": 0.6445,
+      "step": 10629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4026608654126217e-05,
+      "loss": 0.5936,
+      "step": 10630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.402561193871435e-05,
+      "loss": 0.6007,
+      "step": 10631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4024615175573759e-05,
+      "loss": 0.6652,
+      "step": 10632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.402361836471627e-05,
+      "loss": 0.5859,
+      "step": 10633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.402262150615369e-05,
+      "loss": 0.5791,
+      "step": 10634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4021624599897844e-05,
+      "loss": 0.5734,
+      "step": 10635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4020627645960554e-05,
+      "loss": 0.6606,
+      "step": 10636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4019630644353634e-05,
+      "loss": 0.6261,
+      "step": 10637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4018633595088908e-05,
+      "loss": 0.6567,
+      "step": 10638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4017636498178199e-05,
+      "loss": 0.5598,
+      "step": 10639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4016639353633327e-05,
+      "loss": 0.6117,
+      "step": 10640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4015642161466115e-05,
+      "loss": 0.7032,
+      "step": 10641
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4014644921688385e-05,
+      "loss": 0.6668,
+      "step": 10642
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4013647634311957e-05,
+      "loss": 0.6348,
+      "step": 10643
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4012650299348662e-05,
+      "loss": 0.6152,
+      "step": 10644
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4011652916810322e-05,
+      "loss": 0.6232,
+      "step": 10645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4010655486708762e-05,
+      "loss": 0.6099,
+      "step": 10646
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4009658009055806e-05,
+      "loss": 0.6179,
+      "step": 10647
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4008660483863286e-05,
+      "loss": 0.6227,
+      "step": 10648
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.400766291114302e-05,
+      "loss": 0.6693,
+      "step": 10649
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4006665290906844e-05,
+      "loss": 0.6273,
+      "step": 10650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4005667623166583e-05,
+      "loss": 0.6809,
+      "step": 10651
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4004669907934062e-05,
+      "loss": 0.6431,
+      "step": 10652
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4003672145221115e-05,
+      "loss": 0.6497,
+      "step": 10653
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.400267433503957e-05,
+      "loss": 0.6235,
+      "step": 10654
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4001676477401257e-05,
+      "loss": 0.6452,
+      "step": 10655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4000678572318007e-05,
+      "loss": 0.5731,
+      "step": 10656
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3999680619801652e-05,
+      "loss": 0.588,
+      "step": 10657
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3998682619864023e-05,
+      "loss": 0.6634,
+      "step": 10658
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3997684572516956e-05,
+      "loss": 0.6303,
+      "step": 10659
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3996686477772277e-05,
+      "loss": 0.6322,
+      "step": 10660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3995688335641826e-05,
+      "loss": 0.6251,
+      "step": 10661
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3994690146137437e-05,
+      "loss": 0.7022,
+      "step": 10662
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3993691909270941e-05,
+      "loss": 0.6436,
+      "step": 10663
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3992693625054175e-05,
+      "loss": 0.7223,
+      "step": 10664
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3991695293498976e-05,
+      "loss": 0.6075,
+      "step": 10665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.399069691461718e-05,
+      "loss": 0.6113,
+      "step": 10666
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3989698488420623e-05,
+      "loss": 0.6464,
+      "step": 10667
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3988700014921146e-05,
+      "loss": 0.63,
+      "step": 10668
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.398770149413058e-05,
+      "loss": 0.6375,
+      "step": 10669
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.398670292606077e-05,
+      "loss": 0.6954,
+      "step": 10670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3985704310723555e-05,
+      "loss": 0.6518,
+      "step": 10671
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3984705648130774e-05,
+      "loss": 0.6637,
+      "step": 10672
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3983706938294263e-05,
+      "loss": 0.6569,
+      "step": 10673
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3982708181225873e-05,
+      "loss": 0.6683,
+      "step": 10674
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3981709376937436e-05,
+      "loss": 0.632,
+      "step": 10675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3980710525440797e-05,
+      "loss": 0.6527,
+      "step": 10676
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.39797116267478e-05,
+      "loss": 0.7086,
+      "step": 10677
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3978712680870286e-05,
+      "loss": 0.6101,
+      "step": 10678
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.39777136878201e-05,
+      "loss": 0.6303,
+      "step": 10679
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3976714647609089e-05,
+      "loss": 0.6416,
+      "step": 10680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3975715560249092e-05,
+      "loss": 0.6425,
+      "step": 10681
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.397471642575196e-05,
+      "loss": 0.6406,
+      "step": 10682
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3973717244129537e-05,
+      "loss": 0.5999,
+      "step": 10683
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3972718015393667e-05,
+      "loss": 0.6219,
+      "step": 10684
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.39717187395562e-05,
+      "loss": 0.6644,
+      "step": 10685
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3970719416628985e-05,
+      "loss": 0.6203,
+      "step": 10686
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3969720046623867e-05,
+      "loss": 0.6873,
+      "step": 10687
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3968720629552697e-05,
+      "loss": 0.6118,
+      "step": 10688
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3967721165427324e-05,
+      "loss": 0.6198,
+      "step": 10689
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3966721654259596e-05,
+      "loss": 0.5749,
+      "step": 10690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3965722096061362e-05,
+      "loss": 0.6331,
+      "step": 10691
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.396472249084448e-05,
+      "loss": 0.6021,
+      "step": 10692
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3963722838620797e-05,
+      "loss": 0.5921,
+      "step": 10693
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3962723139402164e-05,
+      "loss": 0.6337,
+      "step": 10694
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3961723393200438e-05,
+      "loss": 0.6762,
+      "step": 10695
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3960723600027469e-05,
+      "loss": 0.5917,
+      "step": 10696
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3959723759895108e-05,
+      "loss": 0.6377,
+      "step": 10697
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3958723872815214e-05,
+      "loss": 0.5656,
+      "step": 10698
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3957723938799643e-05,
+      "loss": 0.5877,
+      "step": 10699
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3956723957860245e-05,
+      "loss": 0.5981,
+      "step": 10700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3955723930008881e-05,
+      "loss": 0.6549,
+      "step": 10701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3954723855257408e-05,
+      "loss": 0.6955,
+      "step": 10702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3953723733617676e-05,
+      "loss": 0.6556,
+      "step": 10703
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3952723565101552e-05,
+      "loss": 0.6519,
+      "step": 10704
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.395172334972089e-05,
+      "loss": 0.6524,
+      "step": 10705
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3950723087487546e-05,
+      "loss": 0.6283,
+      "step": 10706
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3949722778413383e-05,
+      "loss": 0.6338,
+      "step": 10707
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3948722422510262e-05,
+      "loss": 0.639,
+      "step": 10708
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3947722019790036e-05,
+      "loss": 0.6602,
+      "step": 10709
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3946721570264577e-05,
+      "loss": 0.6613,
+      "step": 10710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3945721073945741e-05,
+      "loss": 0.6336,
+      "step": 10711
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3944720530845386e-05,
+      "loss": 0.5937,
+      "step": 10712
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3943719940975381e-05,
+      "loss": 0.6996,
+      "step": 10713
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3942719304347591e-05,
+      "loss": 0.6301,
+      "step": 10714
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.394171862097387e-05,
+      "loss": 0.6067,
+      "step": 10715
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3940717890866093e-05,
+      "loss": 0.6232,
+      "step": 10716
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3939717114036121e-05,
+      "loss": 0.643,
+      "step": 10717
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3938716290495812e-05,
+      "loss": 0.6514,
+      "step": 10718
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3937715420257048e-05,
+      "loss": 0.6648,
+      "step": 10719
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3936714503331683e-05,
+      "loss": 0.6986,
+      "step": 10720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3935713539731586e-05,
+      "loss": 0.5827,
+      "step": 10721
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3934712529468627e-05,
+      "loss": 0.6449,
+      "step": 10722
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3933711472554674e-05,
+      "loss": 0.607,
+      "step": 10723
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3932710369001594e-05,
+      "loss": 0.6478,
+      "step": 10724
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3931709218821259e-05,
+      "loss": 0.6469,
+      "step": 10725
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3930708022025537e-05,
+      "loss": 0.6241,
+      "step": 10726
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3929706778626297e-05,
+      "loss": 0.6815,
+      "step": 10727
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3928705488635414e-05,
+      "loss": 0.6639,
+      "step": 10728
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.392770415206476e-05,
+      "loss": 0.6217,
+      "step": 10729
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3926702768926202e-05,
+      "loss": 0.6222,
+      "step": 10730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3925701339231614e-05,
+      "loss": 0.6145,
+      "step": 10731
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3924699862992874e-05,
+      "loss": 0.6528,
+      "step": 10732
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3923698340221852e-05,
+      "loss": 0.6294,
+      "step": 10733
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3922696770930419e-05,
+      "loss": 0.6184,
+      "step": 10734
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3921695155130456e-05,
+      "loss": 0.673,
+      "step": 10735
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3920693492833836e-05,
+      "loss": 0.6196,
+      "step": 10736
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3919691784052436e-05,
+      "loss": 0.6464,
+      "step": 10737
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.391869002879813e-05,
+      "loss": 0.6325,
+      "step": 10738
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3917688227082797e-05,
+      "loss": 0.6321,
+      "step": 10739
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3916686378918314e-05,
+      "loss": 0.6758,
+      "step": 10740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3915684484316559e-05,
+      "loss": 0.5876,
+      "step": 10741
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.391468254328941e-05,
+      "loss": 0.6766,
+      "step": 10742
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.391368055584875e-05,
+      "loss": 0.6057,
+      "step": 10743
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3912678522006456e-05,
+      "loss": 0.6097,
+      "step": 10744
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3911676441774406e-05,
+      "loss": 0.6257,
+      "step": 10745
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3910674315164486e-05,
+      "loss": 0.6204,
+      "step": 10746
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3909672142188575e-05,
+      "loss": 0.6241,
+      "step": 10747
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3908669922858554e-05,
+      "loss": 0.6588,
+      "step": 10748
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3907667657186306e-05,
+      "loss": 0.631,
+      "step": 10749
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3906665345183715e-05,
+      "loss": 0.6801,
+      "step": 10750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3905662986862665e-05,
+      "loss": 0.6643,
+      "step": 10751
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3904660582235039e-05,
+      "loss": 0.6662,
+      "step": 10752
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3903658131312723e-05,
+      "loss": 0.649,
+      "step": 10753
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.39026556341076e-05,
+      "loss": 0.6426,
+      "step": 10754
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3901653090631562e-05,
+      "loss": 0.6142,
+      "step": 10755
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3900650500896485e-05,
+      "loss": 0.6513,
+      "step": 10756
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3899647864914265e-05,
+      "loss": 0.5645,
+      "step": 10757
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3898645182696786e-05,
+      "loss": 0.6302,
+      "step": 10758
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3897642454255937e-05,
+      "loss": 0.6431,
+      "step": 10759
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3896639679603606e-05,
+      "loss": 0.6087,
+      "step": 10760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3895636858751682e-05,
+      "loss": 0.7021,
+      "step": 10761
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3894633991712056e-05,
+      "loss": 0.6343,
+      "step": 10762
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3893631078496617e-05,
+      "loss": 0.6328,
+      "step": 10763
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3892628119117255e-05,
+      "loss": 0.6611,
+      "step": 10764
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3891625113585866e-05,
+      "loss": 0.6407,
+      "step": 10765
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3890622061914331e-05,
+      "loss": 0.6211,
+      "step": 10766
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3889618964114557e-05,
+      "loss": 0.6401,
+      "step": 10767
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3888615820198426e-05,
+      "loss": 0.6156,
+      "step": 10768
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3887612630177836e-05,
+      "loss": 0.6239,
+      "step": 10769
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3886609394064682e-05,
+      "loss": 0.619,
+      "step": 10770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3885606111870854e-05,
+      "loss": 0.6031,
+      "step": 10771
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3884602783608255e-05,
+      "loss": 0.6644,
+      "step": 10772
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3883599409288772e-05,
+      "loss": 0.6124,
+      "step": 10773
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3882595988924306e-05,
+      "loss": 0.6318,
+      "step": 10774
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3881592522526752e-05,
+      "loss": 0.6727,
+      "step": 10775
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3880589010108012e-05,
+      "loss": 0.6701,
+      "step": 10776
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3879585451679978e-05,
+      "loss": 0.6366,
+      "step": 10777
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.387858184725455e-05,
+      "loss": 0.6118,
+      "step": 10778
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3877578196843629e-05,
+      "loss": 0.5655,
+      "step": 10779
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3876574500459113e-05,
+      "loss": 0.6884,
+      "step": 10780
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3875570758112902e-05,
+      "loss": 0.6235,
+      "step": 10781
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3874566969816898e-05,
+      "loss": 0.625,
+      "step": 10782
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3873563135583e-05,
+      "loss": 0.6225,
+      "step": 10783
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.387255925542311e-05,
+      "loss": 0.6257,
+      "step": 10784
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3871555329349135e-05,
+      "loss": 0.5832,
+      "step": 10785
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3870551357372972e-05,
+      "loss": 0.6088,
+      "step": 10786
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3869547339506526e-05,
+      "loss": 0.5867,
+      "step": 10787
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3868543275761702e-05,
+      "loss": 0.6246,
+      "step": 10788
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.38675391661504e-05,
+      "loss": 0.6541,
+      "step": 10789
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3866535010684534e-05,
+      "loss": 0.6076,
+      "step": 10790
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3865530809376002e-05,
+      "loss": 0.6169,
+      "step": 10791
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3864526562236713e-05,
+      "loss": 0.6415,
+      "step": 10792
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3863522269278571e-05,
+      "loss": 0.6598,
+      "step": 10793
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3862517930513489e-05,
+      "loss": 0.6274,
+      "step": 10794
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3861513545953367e-05,
+      "loss": 0.6615,
+      "step": 10795
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3860509115610116e-05,
+      "loss": 0.5709,
+      "step": 10796
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3859504639495651e-05,
+      "loss": 0.6785,
+      "step": 10797
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3858500117621873e-05,
+      "loss": 0.6217,
+      "step": 10798
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3857495550000695e-05,
+      "loss": 0.6292,
+      "step": 10799
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.385649093664403e-05,
+      "loss": 0.6719,
+      "step": 10800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3855486277563784e-05,
+      "loss": 0.5342,
+      "step": 10801
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3854481572771872e-05,
+      "loss": 0.6522,
+      "step": 10802
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3853476822280205e-05,
+      "loss": 0.6748,
+      "step": 10803
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3852472026100693e-05,
+      "loss": 0.6603,
+      "step": 10804
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3851467184245256e-05,
+      "loss": 0.6021,
+      "step": 10805
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3850462296725802e-05,
+      "loss": 0.6222,
+      "step": 10806
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3849457363554245e-05,
+      "loss": 0.6407,
+      "step": 10807
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3848452384742502e-05,
+      "loss": 0.6371,
+      "step": 10808
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3847447360302489e-05,
+      "loss": 0.6532,
+      "step": 10809
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3846442290246119e-05,
+      "loss": 0.6602,
+      "step": 10810
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3845437174585307e-05,
+      "loss": 0.6224,
+      "step": 10811
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3844432013331978e-05,
+      "loss": 0.6329,
+      "step": 10812
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3843426806498041e-05,
+      "loss": 0.6421,
+      "step": 10813
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3842421554095422e-05,
+      "loss": 0.6732,
+      "step": 10814
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3841416256136029e-05,
+      "loss": 0.6372,
+      "step": 10815
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.384041091263179e-05,
+      "loss": 0.5767,
+      "step": 10816
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3839405523594623e-05,
+      "loss": 0.6327,
+      "step": 10817
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3838400089036444e-05,
+      "loss": 0.6352,
+      "step": 10818
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3837394608969175e-05,
+      "loss": 0.6257,
+      "step": 10819
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3836389083404739e-05,
+      "loss": 0.6446,
+      "step": 10820
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3835383512355058e-05,
+      "loss": 0.6431,
+      "step": 10821
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3834377895832055e-05,
+      "loss": 0.5698,
+      "step": 10822
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3833372233847652e-05,
+      "loss": 0.6509,
+      "step": 10823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3832366526413772e-05,
+      "loss": 0.5905,
+      "step": 10824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3831360773542334e-05,
+      "loss": 0.6146,
+      "step": 10825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3830354975245275e-05,
+      "loss": 0.5626,
+      "step": 10826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3829349131534507e-05,
+      "loss": 0.639,
+      "step": 10827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3828343242421961e-05,
+      "loss": 0.6106,
+      "step": 10828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3827337307919567e-05,
+      "loss": 0.6243,
+      "step": 10829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3826331328039242e-05,
+      "loss": 0.6616,
+      "step": 10830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3825325302792925e-05,
+      "loss": 0.6218,
+      "step": 10831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3824319232192532e-05,
+      "loss": 0.6324,
+      "step": 10832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.382331311625e-05,
+      "loss": 0.6092,
+      "step": 10833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3822306954977252e-05,
+      "loss": 0.6658,
+      "step": 10834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3821300748386222e-05,
+      "loss": 0.6255,
+      "step": 10835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3820294496488835e-05,
+      "loss": 0.6067,
+      "step": 10836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3819288199297027e-05,
+      "loss": 0.6495,
+      "step": 10837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3818281856822725e-05,
+      "loss": 0.5935,
+      "step": 10838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.381727546907786e-05,
+      "loss": 0.6663,
+      "step": 10839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3816269036074368e-05,
+      "loss": 0.6682,
+      "step": 10840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3815262557824177e-05,
+      "loss": 0.5754,
+      "step": 10841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3814256034339221e-05,
+      "loss": 0.633,
+      "step": 10842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3813249465631437e-05,
+      "loss": 0.656,
+      "step": 10843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3812242851712756e-05,
+      "loss": 0.6144,
+      "step": 10844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3811236192595115e-05,
+      "loss": 0.622,
+      "step": 10845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3810229488290445e-05,
+      "loss": 0.6086,
+      "step": 10846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3809222738810685e-05,
+      "loss": 0.5857,
+      "step": 10847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3808215944167774e-05,
+      "loss": 0.6119,
+      "step": 10848
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3807209104373643e-05,
+      "loss": 0.6826,
+      "step": 10849
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.380620221944023e-05,
+      "loss": 0.6125,
+      "step": 10850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3805195289379478e-05,
+      "loss": 0.6653,
+      "step": 10851
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3804188314203319e-05,
+      "loss": 0.6394,
+      "step": 10852
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3803181293923697e-05,
+      "loss": 0.6372,
+      "step": 10853
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3802174228552552e-05,
+      "loss": 0.6248,
+      "step": 10854
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.380116711810182e-05,
+      "loss": 0.7018,
+      "step": 10855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3800159962583442e-05,
+      "loss": 0.6406,
+      "step": 10856
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3799152762009363e-05,
+      "loss": 0.6479,
+      "step": 10857
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.379814551639152e-05,
+      "loss": 0.6541,
+      "step": 10858
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3797138225741856e-05,
+      "loss": 0.6637,
+      "step": 10859
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.379613089007232e-05,
+      "loss": 0.6274,
+      "step": 10860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3795123509394843e-05,
+      "loss": 0.6858,
+      "step": 10861
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3794116083721381e-05,
+      "loss": 0.6335,
+      "step": 10862
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3793108613063872e-05,
+      "loss": 0.6073,
+      "step": 10863
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.379210109743426e-05,
+      "loss": 0.6604,
+      "step": 10864
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3791093536844495e-05,
+      "loss": 0.6481,
+      "step": 10865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3790085931306521e-05,
+      "loss": 0.66,
+      "step": 10866
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3789078280832283e-05,
+      "loss": 0.6529,
+      "step": 10867
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3788070585433729e-05,
+      "loss": 0.5969,
+      "step": 10868
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3787062845122805e-05,
+      "loss": 0.6111,
+      "step": 10869
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.378605505991146e-05,
+      "loss": 0.5514,
+      "step": 10870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3785047229811644e-05,
+      "loss": 0.597,
+      "step": 10871
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3784039354835306e-05,
+      "loss": 0.6354,
+      "step": 10872
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3783031434994393e-05,
+      "loss": 0.6438,
+      "step": 10873
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3782023470300855e-05,
+      "loss": 0.6749,
+      "step": 10874
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3781015460766648e-05,
+      "loss": 0.6647,
+      "step": 10875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3780007406403719e-05,
+      "loss": 0.6705,
+      "step": 10876
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3778999307224019e-05,
+      "loss": 0.6322,
+      "step": 10877
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3777991163239506e-05,
+      "loss": 0.5414,
+      "step": 10878
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3776982974462123e-05,
+      "loss": 0.6058,
+      "step": 10879
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3775974740903831e-05,
+      "loss": 0.65,
+      "step": 10880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3774966462576585e-05,
+      "loss": 0.5934,
+      "step": 10881
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3773958139492329e-05,
+      "loss": 0.6231,
+      "step": 10882
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3772949771663031e-05,
+      "loss": 0.6942,
+      "step": 10883
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.377194135910064e-05,
+      "loss": 0.6671,
+      "step": 10884
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3770932901817112e-05,
+      "loss": 0.6739,
+      "step": 10885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3769924399824402e-05,
+      "loss": 0.5587,
+      "step": 10886
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3768915853134472e-05,
+      "loss": 0.6185,
+      "step": 10887
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3767907261759276e-05,
+      "loss": 0.6458,
+      "step": 10888
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3766898625710771e-05,
+      "loss": 0.6399,
+      "step": 10889
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3765889945000922e-05,
+      "loss": 0.6712,
+      "step": 10890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3764881219641682e-05,
+      "loss": 0.6243,
+      "step": 10891
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.376387244964501e-05,
+      "loss": 0.6308,
+      "step": 10892
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3762863635022871e-05,
+      "loss": 0.6202,
+      "step": 10893
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3761854775787224e-05,
+      "loss": 0.6175,
+      "step": 10894
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3760845871950028e-05,
+      "loss": 0.581,
+      "step": 10895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3759836923523251e-05,
+      "loss": 0.5653,
+      "step": 10896
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3758827930518846e-05,
+      "loss": 0.6297,
+      "step": 10897
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3757818892948782e-05,
+      "loss": 0.6204,
+      "step": 10898
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3756809810825024e-05,
+      "loss": 0.6246,
+      "step": 10899
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3755800684159531e-05,
+      "loss": 0.6514,
+      "step": 10900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3754791512964269e-05,
+      "loss": 0.6434,
+      "step": 10901
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3753782297251206e-05,
+      "loss": 0.6093,
+      "step": 10902
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3752773037032303e-05,
+      "loss": 0.6106,
+      "step": 10903
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3751763732319529e-05,
+      "loss": 0.6399,
+      "step": 10904
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3750754383124853e-05,
+      "loss": 0.6204,
+      "step": 10905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3749744989460236e-05,
+      "loss": 0.6971,
+      "step": 10906
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3748735551337649e-05,
+      "loss": 0.6708,
+      "step": 10907
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3747726068769063e-05,
+      "loss": 0.6145,
+      "step": 10908
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3746716541766438e-05,
+      "loss": 0.6419,
+      "step": 10909
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3745706970341755e-05,
+      "loss": 0.6306,
+      "step": 10910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3744697354506972e-05,
+      "loss": 0.6429,
+      "step": 10911
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3743687694274068e-05,
+      "loss": 0.593,
+      "step": 10912
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3742677989655012e-05,
+      "loss": 0.6274,
+      "step": 10913
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3741668240661769e-05,
+      "loss": 0.6789,
+      "step": 10914
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3740658447306316e-05,
+      "loss": 0.6102,
+      "step": 10915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3739648609600631e-05,
+      "loss": 0.6118,
+      "step": 10916
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3738638727556676e-05,
+      "loss": 0.7153,
+      "step": 10917
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.373762880118643e-05,
+      "loss": 0.6634,
+      "step": 10918
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3736618830501869e-05,
+      "loss": 0.6387,
+      "step": 10919
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3735608815514963e-05,
+      "loss": 0.6053,
+      "step": 10920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3734598756237687e-05,
+      "loss": 0.5979,
+      "step": 10921
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3733588652682024e-05,
+      "loss": 0.6026,
+      "step": 10922
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3732578504859939e-05,
+      "loss": 0.6293,
+      "step": 10923
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3731568312783416e-05,
+      "loss": 0.6104,
+      "step": 10924
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3730558076464433e-05,
+      "loss": 0.711,
+      "step": 10925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3729547795914961e-05,
+      "loss": 0.6878,
+      "step": 10926
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3728537471146983e-05,
+      "loss": 0.5558,
+      "step": 10927
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3727527102172478e-05,
+      "loss": 0.6408,
+      "step": 10928
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3726516689003422e-05,
+      "loss": 0.6857,
+      "step": 10929
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3725506231651797e-05,
+      "loss": 0.6788,
+      "step": 10930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3724495730129585e-05,
+      "loss": 0.6266,
+      "step": 10931
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3723485184448762e-05,
+      "loss": 0.621,
+      "step": 10932
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3722474594621312e-05,
+      "loss": 0.5708,
+      "step": 10933
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3721463960659218e-05,
+      "loss": 0.6046,
+      "step": 10934
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.372045328257446e-05,
+      "loss": 0.6151,
+      "step": 10935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3719442560379022e-05,
+      "loss": 0.6058,
+      "step": 10936
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.371843179408489e-05,
+      "loss": 0.6278,
+      "step": 10937
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3717420983704042e-05,
+      "loss": 0.6083,
+      "step": 10938
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3716410129248465e-05,
+      "loss": 0.6449,
+      "step": 10939
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3715399230730148e-05,
+      "loss": 0.6582,
+      "step": 10940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3714388288161074e-05,
+      "loss": 0.6113,
+      "step": 10941
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3713377301553221e-05,
+      "loss": 0.6721,
+      "step": 10942
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3712366270918589e-05,
+      "loss": 0.6192,
+      "step": 10943
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3711355196269158e-05,
+      "loss": 0.6192,
+      "step": 10944
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3710344077616912e-05,
+      "loss": 0.6734,
+      "step": 10945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.370933291497385e-05,
+      "loss": 0.5825,
+      "step": 10946
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.370832170835195e-05,
+      "loss": 0.6348,
+      "step": 10947
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3707310457763204e-05,
+      "loss": 0.6139,
+      "step": 10948
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3706299163219605e-05,
+      "loss": 0.6421,
+      "step": 10949
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3705287824733142e-05,
+      "loss": 0.6681,
+      "step": 10950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.37042764423158e-05,
+      "loss": 0.6274,
+      "step": 10951
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3703265015979579e-05,
+      "loss": 0.6561,
+      "step": 10952
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3702253545736466e-05,
+      "loss": 0.6197,
+      "step": 10953
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3701242031598452e-05,
+      "loss": 0.6618,
+      "step": 10954
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3700230473577534e-05,
+      "loss": 0.5926,
+      "step": 10955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3699218871685702e-05,
+      "loss": 0.611,
+      "step": 10956
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3698207225934947e-05,
+      "loss": 0.6388,
+      "step": 10957
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3697195536337274e-05,
+      "loss": 0.6137,
+      "step": 10958
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3696183802904668e-05,
+      "loss": 0.6335,
+      "step": 10959
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3695172025649124e-05,
+      "loss": 0.6222,
+      "step": 10960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3694160204582645e-05,
+      "loss": 0.649,
+      "step": 10961
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3693148339717223e-05,
+      "loss": 0.6256,
+      "step": 10962
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3692136431064853e-05,
+      "loss": 0.6404,
+      "step": 10963
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3691124478637538e-05,
+      "loss": 0.612,
+      "step": 10964
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3690112482447276e-05,
+      "loss": 0.5746,
+      "step": 10965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3689100442506055e-05,
+      "loss": 0.5571,
+      "step": 10966
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3688088358825887e-05,
+      "loss": 0.5486,
+      "step": 10967
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3687076231418765e-05,
+      "loss": 0.6103,
+      "step": 10968
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3686064060296687e-05,
+      "loss": 0.6193,
+      "step": 10969
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.368505184547166e-05,
+      "loss": 0.6186,
+      "step": 10970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3684039586955682e-05,
+      "loss": 0.5786,
+      "step": 10971
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3683027284760752e-05,
+      "loss": 0.6378,
+      "step": 10972
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3682014938898875e-05,
+      "loss": 0.6396,
+      "step": 10973
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3681002549382057e-05,
+      "loss": 0.637,
+      "step": 10974
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3679990116222292e-05,
+      "loss": 0.6643,
+      "step": 10975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3678977639431592e-05,
+      "loss": 0.681,
+      "step": 10976
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3677965119021957e-05,
+      "loss": 0.618,
+      "step": 10977
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3676952555005394e-05,
+      "loss": 0.6945,
+      "step": 10978
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3675939947393904e-05,
+      "loss": 0.649,
+      "step": 10979
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3674927296199501e-05,
+      "loss": 0.6671,
+      "step": 10980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3673914601434182e-05,
+      "loss": 0.6112,
+      "step": 10981
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3672901863109958e-05,
+      "loss": 0.5933,
+      "step": 10982
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3671889081238839e-05,
+      "loss": 0.6189,
+      "step": 10983
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3670876255832826e-05,
+      "loss": 0.6718,
+      "step": 10984
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3669863386903935e-05,
+      "loss": 0.6265,
+      "step": 10985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3668850474464171e-05,
+      "loss": 0.6458,
+      "step": 10986
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.366783751852554e-05,
+      "loss": 0.6235,
+      "step": 10987
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3666824519100057e-05,
+      "loss": 0.5818,
+      "step": 10988
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3665811476199733e-05,
+      "loss": 0.6525,
+      "step": 10989
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3664798389836573e-05,
+      "loss": 0.6264,
+      "step": 10990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3663785260022593e-05,
+      "loss": 0.6525,
+      "step": 10991
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3662772086769807e-05,
+      "loss": 0.6633,
+      "step": 10992
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3661758870090221e-05,
+      "loss": 0.5786,
+      "step": 10993
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3660745609995853e-05,
+      "loss": 0.6299,
+      "step": 10994
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3659732306498716e-05,
+      "loss": 0.6739,
+      "step": 10995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3658718959610822e-05,
+      "loss": 0.6526,
+      "step": 10996
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3657705569344183e-05,
+      "loss": 0.6369,
+      "step": 10997
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3656692135710823e-05,
+      "loss": 0.6775,
+      "step": 10998
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.365567865872275e-05,
+      "loss": 0.6233,
+      "step": 10999
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3654665138391981e-05,
+      "loss": 0.5965,
+      "step": 11000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3653651574730536e-05,
+      "loss": 0.5826,
+      "step": 11001
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3652637967750426e-05,
+      "loss": 0.6339,
+      "step": 11002
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3651624317463677e-05,
+      "loss": 0.6067,
+      "step": 11003
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3650610623882299e-05,
+      "loss": 0.6483,
+      "step": 11004
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3649596887018313e-05,
+      "loss": 0.5611,
+      "step": 11005
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3648583106883742e-05,
+      "loss": 0.6991,
+      "step": 11006
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3647569283490602e-05,
+      "loss": 0.6125,
+      "step": 11007
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3646555416850913e-05,
+      "loss": 0.6989,
+      "step": 11008
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3645541506976699e-05,
+      "loss": 0.6056,
+      "step": 11009
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3644527553879976e-05,
+      "loss": 0.6276,
+      "step": 11010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3643513557572768e-05,
+      "loss": 0.6756,
+      "step": 11011
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.36424995180671e-05,
+      "loss": 0.6402,
+      "step": 11012
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3641485435374993e-05,
+      "loss": 0.6173,
+      "step": 11013
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3640471309508465e-05,
+      "loss": 0.6617,
+      "step": 11014
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.363945714047955e-05,
+      "loss": 0.7219,
+      "step": 11015
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3638442928300264e-05,
+      "loss": 0.6295,
+      "step": 11016
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3637428672982633e-05,
+      "loss": 0.6489,
+      "step": 11017
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3636414374538686e-05,
+      "loss": 0.6164,
+      "step": 11018
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3635400032980446e-05,
+      "loss": 0.6132,
+      "step": 11019
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3634385648319941e-05,
+      "loss": 0.6231,
+      "step": 11020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3633371220569195e-05,
+      "loss": 0.6588,
+      "step": 11021
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3632356749740239e-05,
+      "loss": 0.6483,
+      "step": 11022
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3631342235845095e-05,
+      "loss": 0.582,
+      "step": 11023
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3630327678895799e-05,
+      "loss": 0.696,
+      "step": 11024
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3629313078904374e-05,
+      "loss": 0.6234,
+      "step": 11025
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3628298435882849e-05,
+      "loss": 0.6563,
+      "step": 11026
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3627283749843261e-05,
+      "loss": 0.6424,
+      "step": 11027
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3626269020797634e-05,
+      "loss": 0.6621,
+      "step": 11028
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3625254248757997e-05,
+      "loss": 0.6023,
+      "step": 11029
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3624239433736391e-05,
+      "loss": 0.6616,
+      "step": 11030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3623224575744838e-05,
+      "loss": 0.6121,
+      "step": 11031
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3622209674795373e-05,
+      "loss": 0.6671,
+      "step": 11032
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3621194730900034e-05,
+      "loss": 0.6843,
+      "step": 11033
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3620179744070849e-05,
+      "loss": 0.6141,
+      "step": 11034
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.361916471431985e-05,
+      "loss": 0.6506,
+      "step": 11035
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3618149641659079e-05,
+      "loss": 0.6086,
+      "step": 11036
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3617134526100571e-05,
+      "loss": 0.6528,
+      "step": 11037
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3616119367656348e-05,
+      "loss": 0.6436,
+      "step": 11038
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3615104166338462e-05,
+      "loss": 0.6904,
+      "step": 11039
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3614088922158943e-05,
+      "loss": 0.6683,
+      "step": 11040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3613073635129823e-05,
+      "loss": 0.6589,
+      "step": 11041
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3612058305263149e-05,
+      "loss": 0.6508,
+      "step": 11042
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3611042932570956e-05,
+      "loss": 0.6442,
+      "step": 11043
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3610027517065275e-05,
+      "loss": 0.6615,
+      "step": 11044
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3609012058758155e-05,
+      "loss": 0.6077,
+      "step": 11045
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3607996557661634e-05,
+      "loss": 0.578,
+      "step": 11046
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3606981013787743e-05,
+      "loss": 0.5797,
+      "step": 11047
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3605965427148535e-05,
+      "loss": 0.6638,
+      "step": 11048
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3604949797756045e-05,
+      "loss": 0.6418,
+      "step": 11049
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.360393412562231e-05,
+      "loss": 0.6487,
+      "step": 11050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.360291841075938e-05,
+      "loss": 0.5816,
+      "step": 11051
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3601902653179298e-05,
+      "loss": 0.6514,
+      "step": 11052
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3600886852894096e-05,
+      "loss": 0.6408,
+      "step": 11053
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3599871009915831e-05,
+      "loss": 0.5893,
+      "step": 11054
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.359885512425654e-05,
+      "loss": 0.6125,
+      "step": 11055
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3597839195928266e-05,
+      "loss": 0.5929,
+      "step": 11056
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3596823224943063e-05,
+      "loss": 0.6055,
+      "step": 11057
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3595807211312967e-05,
+      "loss": 0.6133,
+      "step": 11058
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3594791155050028e-05,
+      "loss": 0.666,
+      "step": 11059
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3593775056166292e-05,
+      "loss": 0.587,
+      "step": 11060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3592758914673807e-05,
+      "loss": 0.6117,
+      "step": 11061
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3591742730584618e-05,
+      "loss": 0.5922,
+      "step": 11062
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3590726503910781e-05,
+      "loss": 0.6135,
+      "step": 11063
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3589710234664337e-05,
+      "loss": 0.6598,
+      "step": 11064
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3588693922857334e-05,
+      "loss": 0.688,
+      "step": 11065
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3587677568501825e-05,
+      "loss": 0.6404,
+      "step": 11066
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3586661171609863e-05,
+      "loss": 0.6757,
+      "step": 11067
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3585644732193495e-05,
+      "loss": 0.6632,
+      "step": 11068
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3584628250264772e-05,
+      "loss": 0.627,
+      "step": 11069
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3583611725835746e-05,
+      "loss": 0.6489,
+      "step": 11070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.358259515891847e-05,
+      "loss": 0.6988,
+      "step": 11071
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3581578549524997e-05,
+      "loss": 0.5893,
+      "step": 11072
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.358056189766738e-05,
+      "loss": 0.5825,
+      "step": 11073
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3579545203357671e-05,
+      "loss": 0.6617,
+      "step": 11074
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3578528466607925e-05,
+      "loss": 0.6242,
+      "step": 11075
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.35775116874302e-05,
+      "loss": 0.5905,
+      "step": 11076
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3576494865836547e-05,
+      "loss": 0.5773,
+      "step": 11077
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3575478001839023e-05,
+      "loss": 0.6165,
+      "step": 11078
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3574461095449683e-05,
+      "loss": 0.6661,
+      "step": 11079
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3573444146680585e-05,
+      "loss": 0.6196,
+      "step": 11080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3572427155543789e-05,
+      "loss": 0.6138,
+      "step": 11081
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3571410122051347e-05,
+      "loss": 0.6305,
+      "step": 11082
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.357039304621532e-05,
+      "loss": 0.6037,
+      "step": 11083
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3569375928047768e-05,
+      "loss": 0.6315,
+      "step": 11084
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3568358767560752e-05,
+      "loss": 0.6293,
+      "step": 11085
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3567341564766326e-05,
+      "loss": 0.5708,
+      "step": 11086
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.356632431967655e-05,
+      "loss": 0.602,
+      "step": 11087
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3565307032303492e-05,
+      "loss": 0.6163,
+      "step": 11088
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3564289702659208e-05,
+      "loss": 0.6406,
+      "step": 11089
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3563272330755759e-05,
+      "loss": 0.6047,
+      "step": 11090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3562254916605211e-05,
+      "loss": 0.7117,
+      "step": 11091
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3561237460219625e-05,
+      "loss": 0.6143,
+      "step": 11092
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3560219961611063e-05,
+      "loss": 0.6567,
+      "step": 11093
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.355920242079159e-05,
+      "loss": 0.6228,
+      "step": 11094
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3558184837773269e-05,
+      "loss": 0.6192,
+      "step": 11095
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3557167212568167e-05,
+      "loss": 0.6456,
+      "step": 11096
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3556149545188346e-05,
+      "loss": 0.6265,
+      "step": 11097
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3555131835645873e-05,
+      "loss": 0.6376,
+      "step": 11098
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3554114083952821e-05,
+      "loss": 0.5352,
+      "step": 11099
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3553096290121244e-05,
+      "loss": 0.5879,
+      "step": 11100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3552078454163218e-05,
+      "loss": 0.6404,
+      "step": 11101
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.355106057609081e-05,
+      "loss": 0.622,
+      "step": 11102
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3550042655916084e-05,
+      "loss": 0.6338,
+      "step": 11103
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3549024693651112e-05,
+      "loss": 0.6509,
+      "step": 11104
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3548006689307965e-05,
+      "loss": 0.674,
+      "step": 11105
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3546988642898708e-05,
+      "loss": 0.6208,
+      "step": 11106
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3545970554435412e-05,
+      "loss": 0.6477,
+      "step": 11107
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3544952423930156e-05,
+      "loss": 0.5933,
+      "step": 11108
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3543934251394997e-05,
+      "loss": 0.5918,
+      "step": 11109
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3542916036842017e-05,
+      "loss": 0.5649,
+      "step": 11110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3541897780283288e-05,
+      "loss": 0.6496,
+      "step": 11111
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.354087948173088e-05,
+      "loss": 0.6608,
+      "step": 11112
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3539861141196864e-05,
+      "loss": 0.6169,
+      "step": 11113
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3538842758693318e-05,
+      "loss": 0.6564,
+      "step": 11114
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3537824334232315e-05,
+      "loss": 0.5601,
+      "step": 11115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3536805867825927e-05,
+      "loss": 0.612,
+      "step": 11116
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3535787359486232e-05,
+      "loss": 0.6263,
+      "step": 11117
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3534768809225305e-05,
+      "loss": 0.6496,
+      "step": 11118
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.353375021705522e-05,
+      "loss": 0.6431,
+      "step": 11119
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3532731582988061e-05,
+      "loss": 0.6754,
+      "step": 11120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3531712907035898e-05,
+      "loss": 0.6683,
+      "step": 11121
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3530694189210812e-05,
+      "loss": 0.6277,
+      "step": 11122
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3529675429524878e-05,
+      "loss": 0.6191,
+      "step": 11123
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3528656627990176e-05,
+      "loss": 0.6678,
+      "step": 11124
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3527637784618788e-05,
+      "loss": 0.6053,
+      "step": 11125
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3526618899422792e-05,
+      "loss": 0.6792,
+      "step": 11126
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3525599972414264e-05,
+      "loss": 0.6418,
+      "step": 11127
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3524581003605293e-05,
+      "loss": 0.6242,
+      "step": 11128
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3523561993007955e-05,
+      "loss": 0.6122,
+      "step": 11129
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.352254294063433e-05,
+      "loss": 0.5922,
+      "step": 11130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3521523846496503e-05,
+      "loss": 0.5908,
+      "step": 11131
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3520504710606556e-05,
+      "loss": 0.6297,
+      "step": 11132
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3519485532976573e-05,
+      "loss": 0.6831,
+      "step": 11133
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3518466313618634e-05,
+      "loss": 0.6316,
+      "step": 11134
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.351744705254483e-05,
+      "loss": 0.6703,
+      "step": 11135
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.351642774976724e-05,
+      "loss": 0.6124,
+      "step": 11136
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.351540840529795e-05,
+      "loss": 0.654,
+      "step": 11137
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3514389019149047e-05,
+      "loss": 0.6176,
+      "step": 11138
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3513369591332617e-05,
+      "loss": 0.6678,
+      "step": 11139
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3512350121860744e-05,
+      "loss": 0.6494,
+      "step": 11140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3511330610745522e-05,
+      "loss": 0.6016,
+      "step": 11141
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3510311057999028e-05,
+      "loss": 0.656,
+      "step": 11142
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3509291463633358e-05,
+      "loss": 0.6139,
+      "step": 11143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3508271827660599e-05,
+      "loss": 0.671,
+      "step": 11144
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.350725215009284e-05,
+      "loss": 0.6736,
+      "step": 11145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3506232430942168e-05,
+      "loss": 0.6423,
+      "step": 11146
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3505212670220678e-05,
+      "loss": 0.6472,
+      "step": 11147
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3504192867940455e-05,
+      "loss": 0.6846,
+      "step": 11148
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3503173024113592e-05,
+      "loss": 0.6131,
+      "step": 11149
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3502153138752185e-05,
+      "loss": 0.6471,
+      "step": 11150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3501133211868321e-05,
+      "loss": 0.6575,
+      "step": 11151
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3500113243474092e-05,
+      "loss": 0.6212,
+      "step": 11152
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3499093233581595e-05,
+      "loss": 0.6561,
+      "step": 11153
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.349807318220292e-05,
+      "loss": 0.6614,
+      "step": 11154
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3497053089350161e-05,
+      "loss": 0.6461,
+      "step": 11155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3496032955035416e-05,
+      "loss": 0.6117,
+      "step": 11156
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3495012779270777e-05,
+      "loss": 0.6271,
+      "step": 11157
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3493992562068343e-05,
+      "loss": 0.5667,
+      "step": 11158
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3492972303440203e-05,
+      "loss": 0.6125,
+      "step": 11159
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3491952003398461e-05,
+      "loss": 0.6069,
+      "step": 11160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.349093166195521e-05,
+      "loss": 0.6409,
+      "step": 11161
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3489911279122545e-05,
+      "loss": 0.6305,
+      "step": 11162
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.348889085491257e-05,
+      "loss": 0.6332,
+      "step": 11163
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3487870389337381e-05,
+      "loss": 0.6497,
+      "step": 11164
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3486849882409075e-05,
+      "loss": 0.6584,
+      "step": 11165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.348582933413975e-05,
+      "loss": 0.6934,
+      "step": 11166
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3484808744541514e-05,
+      "loss": 0.6464,
+      "step": 11167
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.348378811362646e-05,
+      "loss": 0.6311,
+      "step": 11168
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3482767441406687e-05,
+      "loss": 0.6134,
+      "step": 11169
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3481746727894307e-05,
+      "loss": 0.6181,
+      "step": 11170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3480725973101412e-05,
+      "loss": 0.6184,
+      "step": 11171
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3479705177040106e-05,
+      "loss": 0.604,
+      "step": 11172
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3478684339722494e-05,
+      "loss": 0.6457,
+      "step": 11173
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.347766346116068e-05,
+      "loss": 0.6027,
+      "step": 11174
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3476642541366765e-05,
+      "loss": 0.6657,
+      "step": 11175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3475621580352856e-05,
+      "loss": 0.6447,
+      "step": 11176
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3474600578131054e-05,
+      "loss": 0.628,
+      "step": 11177
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3473579534713468e-05,
+      "loss": 0.662,
+      "step": 11178
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3472558450112206e-05,
+      "loss": 0.6648,
+      "step": 11179
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3471537324339365e-05,
+      "loss": 0.6155,
+      "step": 11180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3470516157407061e-05,
+      "loss": 0.6406,
+      "step": 11181
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3469494949327397e-05,
+      "loss": 0.7016,
+      "step": 11182
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3468473700112482e-05,
+      "loss": 0.6038,
+      "step": 11183
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3467452409774423e-05,
+      "loss": 0.5973,
+      "step": 11184
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.346643107832533e-05,
+      "loss": 0.6474,
+      "step": 11185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3465409705777311e-05,
+      "loss": 0.6042,
+      "step": 11186
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3464388292142475e-05,
+      "loss": 0.5799,
+      "step": 11187
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3463366837432936e-05,
+      "loss": 0.6212,
+      "step": 11188
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.34623453416608e-05,
+      "loss": 0.6441,
+      "step": 11189
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3461323804838178e-05,
+      "loss": 0.6392,
+      "step": 11190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3460302226977189e-05,
+      "loss": 0.6072,
+      "step": 11191
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3459280608089937e-05,
+      "loss": 0.6765,
+      "step": 11192
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3458258948188534e-05,
+      "loss": 0.6188,
+      "step": 11193
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3457237247285103e-05,
+      "loss": 0.6643,
+      "step": 11194
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3456215505391748e-05,
+      "loss": 0.6702,
+      "step": 11195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3455193722520585e-05,
+      "loss": 0.6217,
+      "step": 11196
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.345417189868373e-05,
+      "loss": 0.6177,
+      "step": 11197
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3453150033893298e-05,
+      "loss": 0.6293,
+      "step": 11198
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3452128128161404e-05,
+      "loss": 0.6536,
+      "step": 11199
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3451106181500164e-05,
+      "loss": 0.6236,
+      "step": 11200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3450084193921695e-05,
+      "loss": 0.6064,
+      "step": 11201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3449062165438114e-05,
+      "loss": 0.6051,
+      "step": 11202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3448040096061536e-05,
+      "loss": 0.6466,
+      "step": 11203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3447017985804083e-05,
+      "loss": 0.5877,
+      "step": 11204
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.344599583467787e-05,
+      "loss": 0.6935,
+      "step": 11205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3444973642695016e-05,
+      "loss": 0.6431,
+      "step": 11206
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3443951409867645e-05,
+      "loss": 0.6,
+      "step": 11207
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3442929136207869e-05,
+      "loss": 0.6776,
+      "step": 11208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3441906821727817e-05,
+      "loss": 0.6274,
+      "step": 11209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3440884466439604e-05,
+      "loss": 0.64,
+      "step": 11210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3439862070355352e-05,
+      "loss": 0.6659,
+      "step": 11211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3438839633487184e-05,
+      "loss": 0.6973,
+      "step": 11212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3437817155847224e-05,
+      "loss": 0.61,
+      "step": 11213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.343679463744759e-05,
+      "loss": 0.6441,
+      "step": 11214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.343577207830041e-05,
+      "loss": 0.6334,
+      "step": 11215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3434749478417806e-05,
+      "loss": 0.6332,
+      "step": 11216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.34337268378119e-05,
+      "loss": 0.6654,
+      "step": 11217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3432704156494818e-05,
+      "loss": 0.6117,
+      "step": 11218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3431681434478688e-05,
+      "loss": 0.6261,
+      "step": 11219
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3430658671775634e-05,
+      "loss": 0.6564,
+      "step": 11220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.342963586839778e-05,
+      "loss": 0.6835,
+      "step": 11221
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3428613024357255e-05,
+      "loss": 0.6453,
+      "step": 11222
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3427590139666185e-05,
+      "loss": 0.6464,
+      "step": 11223
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3426567214336696e-05,
+      "loss": 0.6413,
+      "step": 11224
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3425544248380922e-05,
+      "loss": 0.613,
+      "step": 11225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3424521241810983e-05,
+      "loss": 0.7016,
+      "step": 11226
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3423498194639017e-05,
+      "loss": 0.6152,
+      "step": 11227
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3422475106877149e-05,
+      "loss": 0.6529,
+      "step": 11228
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3421451978537507e-05,
+      "loss": 0.6916,
+      "step": 11229
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3420428809632222e-05,
+      "loss": 0.7086,
+      "step": 11230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3419405600173429e-05,
+      "loss": 0.649,
+      "step": 11231
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3418382350173259e-05,
+      "loss": 0.646,
+      "step": 11232
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.341735905964384e-05,
+      "loss": 0.6866,
+      "step": 11233
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3416335728597307e-05,
+      "loss": 0.6267,
+      "step": 11234
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3415312357045788e-05,
+      "loss": 0.6698,
+      "step": 11235
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3414288945001425e-05,
+      "loss": 0.6171,
+      "step": 11236
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3413265492476348e-05,
+      "loss": 0.638,
+      "step": 11237
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.341224199948269e-05,
+      "loss": 0.5913,
+      "step": 11238
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3411218466032585e-05,
+      "loss": 0.6145,
+      "step": 11239
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3410194892138171e-05,
+      "loss": 0.6223,
+      "step": 11240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3409171277811581e-05,
+      "loss": 0.6075,
+      "step": 11241
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3408147623064954e-05,
+      "loss": 0.6731,
+      "step": 11242
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3407123927910428e-05,
+      "loss": 0.6663,
+      "step": 11243
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3406100192360134e-05,
+      "loss": 0.6284,
+      "step": 11244
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3405076416426214e-05,
+      "loss": 0.6565,
+      "step": 11245
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.340405260012081e-05,
+      "loss": 0.677,
+      "step": 11246
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3403028743456053e-05,
+      "loss": 0.6238,
+      "step": 11247
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3402004846444084e-05,
+      "loss": 0.668,
+      "step": 11248
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3400980909097045e-05,
+      "loss": 0.6019,
+      "step": 11249
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3399956931427077e-05,
+      "loss": 0.6775,
+      "step": 11250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.339893291344632e-05,
+      "loss": 0.6208,
+      "step": 11251
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3397908855166908e-05,
+      "loss": 0.6602,
+      "step": 11252
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3396884756600993e-05,
+      "loss": 0.6197,
+      "step": 11253
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3395860617760709e-05,
+      "loss": 0.6116,
+      "step": 11254
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3394836438658203e-05,
+      "loss": 0.5582,
+      "step": 11255
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3393812219305615e-05,
+      "loss": 0.619,
+      "step": 11256
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3392787959715091e-05,
+      "loss": 0.6143,
+      "step": 11257
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3391763659898773e-05,
+      "loss": 0.6857,
+      "step": 11258
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3390739319868808e-05,
+      "loss": 0.619,
+      "step": 11259
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.338971493963734e-05,
+      "loss": 0.6637,
+      "step": 11260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.338869051921651e-05,
+      "loss": 0.6036,
+      "step": 11261
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3387666058618469e-05,
+      "loss": 0.6297,
+      "step": 11262
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3386641557855362e-05,
+      "loss": 0.6454,
+      "step": 11263
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3385617016939333e-05,
+      "loss": 0.6144,
+      "step": 11264
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3384592435882533e-05,
+      "loss": 0.6963,
+      "step": 11265
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3383567814697109e-05,
+      "loss": 0.6184,
+      "step": 11266
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3382543153395207e-05,
+      "loss": 0.6503,
+      "step": 11267
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3381518451988978e-05,
+      "loss": 0.5793,
+      "step": 11268
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.338049371049057e-05,
+      "loss": 0.6147,
+      "step": 11269
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3379468928912131e-05,
+      "loss": 0.6134,
+      "step": 11270
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3378444107265814e-05,
+      "loss": 0.6318,
+      "step": 11271
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.337741924556377e-05,
+      "loss": 0.617,
+      "step": 11272
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3376394343818144e-05,
+      "loss": 0.603,
+      "step": 11273
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3375369402041094e-05,
+      "loss": 0.555,
+      "step": 11274
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3374344420244774e-05,
+      "loss": 0.6296,
+      "step": 11275
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3373319398441327e-05,
+      "loss": 0.6127,
+      "step": 11276
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3372294336642912e-05,
+      "loss": 0.6456,
+      "step": 11277
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3371269234861686e-05,
+      "loss": 0.6798,
+      "step": 11278
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3370244093109795e-05,
+      "loss": 0.5968,
+      "step": 11279
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3369218911399396e-05,
+      "loss": 0.6387,
+      "step": 11280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3368193689742645e-05,
+      "loss": 0.6207,
+      "step": 11281
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.33671684281517e-05,
+      "loss": 0.6618,
+      "step": 11282
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3366143126638708e-05,
+      "loss": 0.6253,
+      "step": 11283
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3365117785215837e-05,
+      "loss": 0.6123,
+      "step": 11284
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3364092403895236e-05,
+      "loss": 0.6574,
+      "step": 11285
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3363066982689061e-05,
+      "loss": 0.6148,
+      "step": 11286
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3362041521609476e-05,
+      "loss": 0.6415,
+      "step": 11287
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3361016020668637e-05,
+      "loss": 0.6753,
+      "step": 11288
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3359990479878697e-05,
+      "loss": 0.6262,
+      "step": 11289
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3358964899251822e-05,
+      "loss": 0.706,
+      "step": 11290
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3357939278800171e-05,
+      "loss": 0.6456,
+      "step": 11291
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3356913618535896e-05,
+      "loss": 0.6582,
+      "step": 11292
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3355887918471169e-05,
+      "loss": 0.6125,
+      "step": 11293
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3354862178618145e-05,
+      "loss": 0.6294,
+      "step": 11294
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3353836398988982e-05,
+      "loss": 0.6103,
+      "step": 11295
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.335281057959585e-05,
+      "loss": 0.588,
+      "step": 11296
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3351784720450906e-05,
+      "loss": 0.6517,
+      "step": 11297
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3350758821566312e-05,
+      "loss": 0.6764,
+      "step": 11298
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3349732882954238e-05,
+      "loss": 0.6808,
+      "step": 11299
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.334870690462684e-05,
+      "loss": 0.6254,
+      "step": 11300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3347680886596285e-05,
+      "loss": 0.5986,
+      "step": 11301
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.334665482887474e-05,
+      "loss": 0.6869,
+      "step": 11302
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.334562873147437e-05,
+      "loss": 0.6411,
+      "step": 11303
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3344602594407334e-05,
+      "loss": 0.655,
+      "step": 11304
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3343576417685809e-05,
+      "loss": 0.652,
+      "step": 11305
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3342550201321954e-05,
+      "loss": 0.6362,
+      "step": 11306
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3341523945327931e-05,
+      "loss": 0.596,
+      "step": 11307
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3340497649715925e-05,
+      "loss": 0.5683,
+      "step": 11308
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.333947131449809e-05,
+      "loss": 0.6343,
+      "step": 11309
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3338444939686594e-05,
+      "loss": 0.6244,
+      "step": 11310
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3337418525293618e-05,
+      "loss": 0.6315,
+      "step": 11311
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3336392071331319e-05,
+      "loss": 0.6143,
+      "step": 11312
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3335365577811872e-05,
+      "loss": 0.604,
+      "step": 11313
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3334339044747445e-05,
+      "loss": 0.6577,
+      "step": 11314
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3333312472150213e-05,
+      "loss": 0.6433,
+      "step": 11315
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3332285860032343e-05,
+      "loss": 0.6073,
+      "step": 11316
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3331259208406009e-05,
+      "loss": 0.6856,
+      "step": 11317
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3330232517283386e-05,
+      "loss": 0.6049,
+      "step": 11318
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.332920578667664e-05,
+      "loss": 0.591,
+      "step": 11319
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.332817901659795e-05,
+      "loss": 0.6345,
+      "step": 11320
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3327152207059486e-05,
+      "loss": 0.631,
+      "step": 11321
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3326125358073425e-05,
+      "loss": 0.6174,
+      "step": 11322
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3325098469651941e-05,
+      "loss": 0.6442,
+      "step": 11323
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3324071541807207e-05,
+      "loss": 0.596,
+      "step": 11324
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3323044574551401e-05,
+      "loss": 0.6835,
+      "step": 11325
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3322017567896697e-05,
+      "loss": 0.6419,
+      "step": 11326
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3320990521855272e-05,
+      "loss": 0.609,
+      "step": 11327
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3319963436439305e-05,
+      "loss": 0.5858,
+      "step": 11328
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3318936311660968e-05,
+      "loss": 0.5993,
+      "step": 11329
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3317909147532446e-05,
+      "loss": 0.6797,
+      "step": 11330
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3316881944065914e-05,
+      "loss": 0.6164,
+      "step": 11331
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3315854701273549e-05,
+      "loss": 0.5826,
+      "step": 11332
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3314827419167533e-05,
+      "loss": 0.6714,
+      "step": 11333
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3313800097760043e-05,
+      "loss": 0.5963,
+      "step": 11334
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3312772737063262e-05,
+      "loss": 0.6441,
+      "step": 11335
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3311745337089372e-05,
+      "loss": 0.6359,
+      "step": 11336
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.331071789785055e-05,
+      "loss": 0.6779,
+      "step": 11337
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3309690419358977e-05,
+      "loss": 0.6799,
+      "step": 11338
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.330866290162684e-05,
+      "loss": 0.7005,
+      "step": 11339
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3307635344666316e-05,
+      "loss": 0.6287,
+      "step": 11340
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3306607748489593e-05,
+      "loss": 0.6558,
+      "step": 11341
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3305580113108852e-05,
+      "loss": 0.5958,
+      "step": 11342
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3304552438536275e-05,
+      "loss": 0.6805,
+      "step": 11343
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3303524724784054e-05,
+      "loss": 0.638,
+      "step": 11344
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3302496971864364e-05,
+      "loss": 0.6342,
+      "step": 11345
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3301469179789395e-05,
+      "loss": 0.6498,
+      "step": 11346
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3300441348571333e-05,
+      "loss": 0.6878,
+      "step": 11347
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3299413478222363e-05,
+      "loss": 0.6226,
+      "step": 11348
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3298385568754672e-05,
+      "loss": 0.6279,
+      "step": 11349
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3297357620180451e-05,
+      "loss": 0.6199,
+      "step": 11350
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3296329632511881e-05,
+      "loss": 0.6083,
+      "step": 11351
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3295301605761156e-05,
+      "loss": 0.6223,
+      "step": 11352
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3294273539940461e-05,
+      "loss": 0.6415,
+      "step": 11353
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3293245435061985e-05,
+      "loss": 0.6634,
+      "step": 11354
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3292217291137917e-05,
+      "loss": 0.6449,
+      "step": 11355
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3291189108180452e-05,
+      "loss": 0.6425,
+      "step": 11356
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3290160886201773e-05,
+      "loss": 0.6779,
+      "step": 11357
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3289132625214077e-05,
+      "loss": 0.6878,
+      "step": 11358
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3288104325229552e-05,
+      "loss": 0.6138,
+      "step": 11359
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3287075986260392e-05,
+      "loss": 0.6144,
+      "step": 11360
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3286047608318786e-05,
+      "loss": 0.6378,
+      "step": 11361
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.328501919141693e-05,
+      "loss": 0.6362,
+      "step": 11362
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3283990735567016e-05,
+      "loss": 0.6001,
+      "step": 11363
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3282962240781237e-05,
+      "loss": 0.6031,
+      "step": 11364
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3281933707071787e-05,
+      "loss": 0.6765,
+      "step": 11365
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3280905134450863e-05,
+      "loss": 0.5949,
+      "step": 11366
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3279876522930658e-05,
+      "loss": 0.6337,
+      "step": 11367
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3278847872523368e-05,
+      "loss": 0.6214,
+      "step": 11368
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3277819183241186e-05,
+      "loss": 0.6437,
+      "step": 11369
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3276790455096314e-05,
+      "loss": 0.6653,
+      "step": 11370
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3275761688100946e-05,
+      "loss": 0.6002,
+      "step": 11371
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3274732882267276e-05,
+      "loss": 0.5815,
+      "step": 11372
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.327370403760751e-05,
+      "loss": 0.6288,
+      "step": 11373
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3272675154133839e-05,
+      "loss": 0.6304,
+      "step": 11374
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3271646231858462e-05,
+      "loss": 0.6107,
+      "step": 11375
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3270617270793582e-05,
+      "loss": 0.6474,
+      "step": 11376
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.32695882709514e-05,
+      "loss": 0.6129,
+      "step": 11377
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3268559232344107e-05,
+      "loss": 0.6774,
+      "step": 11378
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3267530154983912e-05,
+      "loss": 0.6059,
+      "step": 11379
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3266501038883015e-05,
+      "loss": 0.6154,
+      "step": 11380
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3265471884053616e-05,
+      "loss": 0.6439,
+      "step": 11381
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3264442690507914e-05,
+      "loss": 0.6858,
+      "step": 11382
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3263413458258119e-05,
+      "loss": 0.6211,
+      "step": 11383
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3262384187316427e-05,
+      "loss": 0.6672,
+      "step": 11384
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.326135487769504e-05,
+      "loss": 0.6564,
+      "step": 11385
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.326032552940617e-05,
+      "loss": 0.5673,
+      "step": 11386
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3259296142462015e-05,
+      "loss": 0.5876,
+      "step": 11387
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3258266716874781e-05,
+      "loss": 0.6094,
+      "step": 11388
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3257237252656676e-05,
+      "loss": 0.6523,
+      "step": 11389
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.32562077498199e-05,
+      "loss": 0.6462,
+      "step": 11390
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.325517820837666e-05,
+      "loss": 0.6058,
+      "step": 11391
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.325414862833917e-05,
+      "loss": 0.6136,
+      "step": 11392
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.325311900971963e-05,
+      "loss": 0.6251,
+      "step": 11393
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3252089352530244e-05,
+      "loss": 0.6147,
+      "step": 11394
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3251059656783229e-05,
+      "loss": 0.6485,
+      "step": 11395
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.325002992249079e-05,
+      "loss": 0.6622,
+      "step": 11396
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3249000149665133e-05,
+      "loss": 0.6596,
+      "step": 11397
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.324797033831847e-05,
+      "loss": 0.6307,
+      "step": 11398
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.324694048846301e-05,
+      "loss": 0.6501,
+      "step": 11399
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3245910600110961e-05,
+      "loss": 0.6134,
+      "step": 11400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3244880673274537e-05,
+      "loss": 0.6033,
+      "step": 11401
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.324385070796595e-05,
+      "loss": 0.6246,
+      "step": 11402
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3242820704197406e-05,
+      "loss": 0.6445,
+      "step": 11403
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3241790661981124e-05,
+      "loss": 0.6333,
+      "step": 11404
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3240760581329312e-05,
+      "loss": 0.6512,
+      "step": 11405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.323973046225418e-05,
+      "loss": 0.6994,
+      "step": 11406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3238700304767949e-05,
+      "loss": 0.6595,
+      "step": 11407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.323767010888283e-05,
+      "loss": 0.6444,
+      "step": 11408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.323663987461103e-05,
+      "loss": 0.654,
+      "step": 11409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3235609601964774e-05,
+      "loss": 0.6434,
+      "step": 11410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3234579290956276e-05,
+      "loss": 0.5951,
+      "step": 11411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3233548941597743e-05,
+      "loss": 0.6076,
+      "step": 11412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3232518553901399e-05,
+      "loss": 0.6331,
+      "step": 11413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.323148812787946e-05,
+      "loss": 0.6535,
+      "step": 11414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3230457663544139e-05,
+      "loss": 0.6386,
+      "step": 11415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3229427160907657e-05,
+      "loss": 0.641,
+      "step": 11416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3228396619982231e-05,
+      "loss": 0.622,
+      "step": 11417
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3227366040780078e-05,
+      "loss": 0.6307,
+      "step": 11418
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.322633542331342e-05,
+      "loss": 0.6273,
+      "step": 11419
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3225304767594473e-05,
+      "loss": 0.5944,
+      "step": 11420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3224274073635458e-05,
+      "loss": 0.626,
+      "step": 11421
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3223243341448595e-05,
+      "loss": 0.6669,
+      "step": 11422
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3222212571046104e-05,
+      "loss": 0.6413,
+      "step": 11423
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.322118176244021e-05,
+      "loss": 0.5855,
+      "step": 11424
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3220150915643127e-05,
+      "loss": 0.6438,
+      "step": 11425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3219120030667085e-05,
+      "loss": 0.6484,
+      "step": 11426
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3218089107524298e-05,
+      "loss": 0.6567,
+      "step": 11427
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3217058146226997e-05,
+      "loss": 0.6907,
+      "step": 11428
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3216027146787405e-05,
+      "loss": 0.6269,
+      "step": 11429
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.321499610921774e-05,
+      "loss": 0.6967,
+      "step": 11430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3213965033530228e-05,
+      "loss": 0.6553,
+      "step": 11431
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3212933919737096e-05,
+      "loss": 0.6144,
+      "step": 11432
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3211902767850568e-05,
+      "loss": 0.6044,
+      "step": 11433
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3210871577882868e-05,
+      "loss": 0.6318,
+      "step": 11434
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3209840349846228e-05,
+      "loss": 0.6339,
+      "step": 11435
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3208809083752867e-05,
+      "loss": 0.6713,
+      "step": 11436
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3207777779615016e-05,
+      "loss": 0.6084,
+      "step": 11437
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3206746437444901e-05,
+      "loss": 0.6245,
+      "step": 11438
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3205715057254749e-05,
+      "loss": 0.68,
+      "step": 11439
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3204683639056795e-05,
+      "loss": 0.6471,
+      "step": 11440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3203652182863255e-05,
+      "loss": 0.6221,
+      "step": 11441
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.320262068868637e-05,
+      "loss": 0.6118,
+      "step": 11442
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3201589156538366e-05,
+      "loss": 0.6396,
+      "step": 11443
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3200557586431471e-05,
+      "loss": 0.6025,
+      "step": 11444
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3199525978377916e-05,
+      "loss": 0.6378,
+      "step": 11445
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3198494332389938e-05,
+      "loss": 0.6318,
+      "step": 11446
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3197462648479759e-05,
+      "loss": 0.6632,
+      "step": 11447
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3196430926659616e-05,
+      "loss": 0.6188,
+      "step": 11448
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3195399166941742e-05,
+      "loss": 0.6508,
+      "step": 11449
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3194367369338368e-05,
+      "loss": 0.6007,
+      "step": 11450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.319333553386173e-05,
+      "loss": 0.6306,
+      "step": 11451
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3192303660524057e-05,
+      "loss": 0.6485,
+      "step": 11452
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3191271749337586e-05,
+      "loss": 0.6652,
+      "step": 11453
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3190239800314553e-05,
+      "loss": 0.6828,
+      "step": 11454
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3189207813467191e-05,
+      "loss": 0.6741,
+      "step": 11455
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3188175788807733e-05,
+      "loss": 0.6357,
+      "step": 11456
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3187143726348422e-05,
+      "loss": 0.6065,
+      "step": 11457
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.318611162610149e-05,
+      "loss": 0.6274,
+      "step": 11458
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.318507948807917e-05,
+      "loss": 0.6101,
+      "step": 11459
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3184047312293709e-05,
+      "loss": 0.6269,
+      "step": 11460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3183015098757337e-05,
+      "loss": 0.669,
+      "step": 11461
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3181982847482292e-05,
+      "loss": 0.5392,
+      "step": 11462
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.318095055848082e-05,
+      "loss": 0.631,
+      "step": 11463
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3179918231765152e-05,
+      "loss": 0.6117,
+      "step": 11464
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3178885867347532e-05,
+      "loss": 0.6474,
+      "step": 11465
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3177853465240196e-05,
+      "loss": 0.5907,
+      "step": 11466
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3176821025455393e-05,
+      "loss": 0.638,
+      "step": 11467
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3175788548005352e-05,
+      "loss": 0.6304,
+      "step": 11468
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3174756032902323e-05,
+      "loss": 0.6351,
+      "step": 11469
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3173723480158546e-05,
+      "loss": 0.6295,
+      "step": 11470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3172690889786259e-05,
+      "loss": 0.6532,
+      "step": 11471
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.317165826179771e-05,
+      "loss": 0.6296,
+      "step": 11472
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3170625596205142e-05,
+      "loss": 0.6169,
+      "step": 11473
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3169592893020795e-05,
+      "loss": 0.6607,
+      "step": 11474
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3168560152256915e-05,
+      "loss": 0.6775,
+      "step": 11475
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3167527373925748e-05,
+      "loss": 0.6404,
+      "step": 11476
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3166494558039533e-05,
+      "loss": 0.5484,
+      "step": 11477
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.316546170461052e-05,
+      "loss": 0.5863,
+      "step": 11478
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3164428813650957e-05,
+      "loss": 0.6493,
+      "step": 11479
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3163395885173086e-05,
+      "loss": 0.6341,
+      "step": 11480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3162362919189154e-05,
+      "loss": 0.6561,
+      "step": 11481
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3161329915711412e-05,
+      "loss": 0.653,
+      "step": 11482
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3160296874752101e-05,
+      "loss": 0.6417,
+      "step": 11483
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3159263796323477e-05,
+      "loss": 0.5965,
+      "step": 11484
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3158230680437783e-05,
+      "loss": 0.6734,
+      "step": 11485
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3157197527107267e-05,
+      "loss": 0.6462,
+      "step": 11486
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3156164336344179e-05,
+      "loss": 0.624,
+      "step": 11487
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3155131108160775e-05,
+      "loss": 0.6053,
+      "step": 11488
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3154097842569297e-05,
+      "loss": 0.6055,
+      "step": 11489
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3153064539582e-05,
+      "loss": 0.6615,
+      "step": 11490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3152031199211135e-05,
+      "loss": 0.6677,
+      "step": 11491
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3150997821468951e-05,
+      "loss": 0.6168,
+      "step": 11492
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3149964406367703e-05,
+      "loss": 0.6023,
+      "step": 11493
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3148930953919643e-05,
+      "loss": 0.6487,
+      "step": 11494
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3147897464137023e-05,
+      "loss": 0.6029,
+      "step": 11495
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3146863937032095e-05,
+      "loss": 0.6258,
+      "step": 11496
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3145830372617116e-05,
+      "loss": 0.6423,
+      "step": 11497
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3144796770904337e-05,
+      "loss": 0.6578,
+      "step": 11498
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3143763131906018e-05,
+      "loss": 0.5901,
+      "step": 11499
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3142729455634404e-05,
+      "loss": 0.6407,
+      "step": 11500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3141695742101758e-05,
+      "loss": 0.5907,
+      "step": 11501
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.314066199132034e-05,
+      "loss": 0.6077,
+      "step": 11502
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3139628203302395e-05,
+      "loss": 0.6238,
+      "step": 11503
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3138594378060188e-05,
+      "loss": 0.6149,
+      "step": 11504
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3137560515605975e-05,
+      "loss": 0.6299,
+      "step": 11505
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3136526615952015e-05,
+      "loss": 0.6056,
+      "step": 11506
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3135492679110559e-05,
+      "loss": 0.6135,
+      "step": 11507
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3134458705093876e-05,
+      "loss": 0.6058,
+      "step": 11508
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3133424693914218e-05,
+      "loss": 0.6409,
+      "step": 11509
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3132390645583844e-05,
+      "loss": 0.5969,
+      "step": 11510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3131356560115021e-05,
+      "loss": 0.6855,
+      "step": 11511
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3130322437520002e-05,
+      "loss": 0.5766,
+      "step": 11512
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.312928827781105e-05,
+      "loss": 0.6755,
+      "step": 11513
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.312825408100043e-05,
+      "loss": 0.6023,
+      "step": 11514
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3127219847100397e-05,
+      "loss": 0.5994,
+      "step": 11515
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3126185576123219e-05,
+      "loss": 0.6314,
+      "step": 11516
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3125151268081156e-05,
+      "loss": 0.6975,
+      "step": 11517
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.312411692298647e-05,
+      "loss": 0.605,
+      "step": 11518
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3123082540851427e-05,
+      "loss": 0.5787,
+      "step": 11519
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.312204812168829e-05,
+      "loss": 0.621,
+      "step": 11520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3121013665509324e-05,
+      "loss": 0.631,
+      "step": 11521
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.311997917232679e-05,
+      "loss": 0.6259,
+      "step": 11522
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3118944642152958e-05,
+      "loss": 0.657,
+      "step": 11523
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3117910075000091e-05,
+      "loss": 0.6539,
+      "step": 11524
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3116875470880457e-05,
+      "loss": 0.5853,
+      "step": 11525
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3115840829806322e-05,
+      "loss": 0.6192,
+      "step": 11526
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3114806151789949e-05,
+      "loss": 0.6189,
+      "step": 11527
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3113771436843612e-05,
+      "loss": 0.5707,
+      "step": 11528
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3112736684979577e-05,
+      "loss": 0.6167,
+      "step": 11529
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.311170189621011e-05,
+      "loss": 0.6128,
+      "step": 11530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3110667070547476e-05,
+      "loss": 0.6206,
+      "step": 11531
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3109632208003953e-05,
+      "loss": 0.6994,
+      "step": 11532
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3108597308591808e-05,
+      "loss": 0.6337,
+      "step": 11533
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3107562372323304e-05,
+      "loss": 0.6142,
+      "step": 11534
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3106527399210722e-05,
+      "loss": 0.6335,
+      "step": 11535
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3105492389266327e-05,
+      "loss": 0.6239,
+      "step": 11536
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3104457342502387e-05,
+      "loss": 0.6375,
+      "step": 11537
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3103422258931185e-05,
+      "loss": 0.6374,
+      "step": 11538
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3102387138564983e-05,
+      "loss": 0.6138,
+      "step": 11539
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3101351981416053e-05,
+      "loss": 0.5834,
+      "step": 11540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3100316787496678e-05,
+      "loss": 0.644,
+      "step": 11541
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3099281556819125e-05,
+      "loss": 0.6651,
+      "step": 11542
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3098246289395664e-05,
+      "loss": 0.6324,
+      "step": 11543
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3097210985238578e-05,
+      "loss": 0.6033,
+      "step": 11544
+    },
+    {
+      "epoch": 0.4,
+      "eval_loss": 0.5762600302696228,
+      "eval_runtime": 4922.6934,
+      "eval_samples_per_second": 112.061,
+      "eval_steps_per_second": 37.354,
+      "step": 11544
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3096175644360138e-05,
+      "loss": 0.662,
+      "step": 11545
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3095140266772614e-05,
+      "loss": 0.6567,
+      "step": 11546
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3094104852488292e-05,
+      "loss": 0.6154,
+      "step": 11547
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3093069401519443e-05,
+      "loss": 0.7427,
+      "step": 11548
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3092033913878339e-05,
+      "loss": 0.5697,
+      "step": 11549
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3090998389577264e-05,
+      "loss": 0.6315,
+      "step": 11550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3089962828628498e-05,
+      "loss": 0.5809,
+      "step": 11551
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3088927231044304e-05,
+      "loss": 0.668,
+      "step": 11552
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3087891596836976e-05,
+      "loss": 0.6504,
+      "step": 11553
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3086855926018788e-05,
+      "loss": 0.6427,
+      "step": 11554
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3085820218602015e-05,
+      "loss": 0.6349,
+      "step": 11555
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3084784474598942e-05,
+      "loss": 0.6717,
+      "step": 11556
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3083748694021847e-05,
+      "loss": 0.6583,
+      "step": 11557
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3082712876883005e-05,
+      "loss": 0.616,
+      "step": 11558
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3081677023194707e-05,
+      "loss": 0.6281,
+      "step": 11559
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3080641132969231e-05,
+      "loss": 0.6229,
+      "step": 11560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3079605206218854e-05,
+      "loss": 0.594,
+      "step": 11561
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3078569242955864e-05,
+      "loss": 0.6541,
+      "step": 11562
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3077533243192541e-05,
+      "loss": 0.6036,
+      "step": 11563
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3076497206941165e-05,
+      "loss": 0.6025,
+      "step": 11564
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3075461134214024e-05,
+      "loss": 0.5859,
+      "step": 11565
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3074425025023405e-05,
+      "loss": 0.6884,
+      "step": 11566
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3073388879381582e-05,
+      "loss": 0.6299,
+      "step": 11567
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.307235269730085e-05,
+      "loss": 0.6414,
+      "step": 11568
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.307131647879349e-05,
+      "loss": 0.6114,
+      "step": 11569
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3070280223871787e-05,
+      "loss": 0.6416,
+      "step": 11570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3069243932548026e-05,
+      "loss": 0.6549,
+      "step": 11571
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.30682076048345e-05,
+      "loss": 0.5729,
+      "step": 11572
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3067171240743487e-05,
+      "loss": 0.6443,
+      "step": 11573
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3066134840287282e-05,
+      "loss": 0.6223,
+      "step": 11574
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3065098403478169e-05,
+      "loss": 0.6039,
+      "step": 11575
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3064061930328438e-05,
+      "loss": 0.6499,
+      "step": 11576
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3063025420850375e-05,
+      "loss": 0.6371,
+      "step": 11577
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3061988875056274e-05,
+      "loss": 0.6174,
+      "step": 11578
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3060952292958418e-05,
+      "loss": 0.6105,
+      "step": 11579
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3059915674569099e-05,
+      "loss": 0.5876,
+      "step": 11580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3058879019900614e-05,
+      "loss": 0.6341,
+      "step": 11581
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3057842328965244e-05,
+      "loss": 0.6575,
+      "step": 11582
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3056805601775286e-05,
+      "loss": 0.6011,
+      "step": 11583
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3055768838343035e-05,
+      "loss": 0.6456,
+      "step": 11584
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3054732038680776e-05,
+      "loss": 0.6678,
+      "step": 11585
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3053695202800802e-05,
+      "loss": 0.6379,
+      "step": 11586
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3052658330715411e-05,
+      "loss": 0.6255,
+      "step": 11587
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3051621422436892e-05,
+      "loss": 0.6705,
+      "step": 11588
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3050584477977544e-05,
+      "loss": 0.6223,
+      "step": 11589
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3049547497349657e-05,
+      "loss": 0.6235,
+      "step": 11590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3048510480565525e-05,
+      "loss": 0.6306,
+      "step": 11591
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3047473427637448e-05,
+      "loss": 0.659,
+      "step": 11592
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3046436338577713e-05,
+      "loss": 0.6334,
+      "step": 11593
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3045399213398624e-05,
+      "loss": 0.6441,
+      "step": 11594
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3044362052112478e-05,
+      "loss": 0.5832,
+      "step": 11595
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3043324854731565e-05,
+      "loss": 0.6516,
+      "step": 11596
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3042287621268185e-05,
+      "loss": 0.6603,
+      "step": 11597
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3041250351734643e-05,
+      "loss": 0.6068,
+      "step": 11598
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3040213046143224e-05,
+      "loss": 0.6282,
+      "step": 11599
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3039175704506237e-05,
+      "loss": 0.6688,
+      "step": 11600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.303813832683598e-05,
+      "loss": 0.5768,
+      "step": 11601
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3037100913144744e-05,
+      "loss": 0.6134,
+      "step": 11602
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3036063463444837e-05,
+      "loss": 0.6444,
+      "step": 11603
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3035025977748559e-05,
+      "loss": 0.6491,
+      "step": 11604
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3033988456068205e-05,
+      "loss": 0.5782,
+      "step": 11605
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3032950898416082e-05,
+      "loss": 0.6592,
+      "step": 11606
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.303191330480449e-05,
+      "loss": 0.6457,
+      "step": 11607
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3030875675245728e-05,
+      "loss": 0.6522,
+      "step": 11608
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3029838009752103e-05,
+      "loss": 0.6413,
+      "step": 11609
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3028800308335917e-05,
+      "loss": 0.5872,
+      "step": 11610
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3027762571009467e-05,
+      "loss": 0.6316,
+      "step": 11611
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3026724797785063e-05,
+      "loss": 0.6285,
+      "step": 11612
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3025686988675013e-05,
+      "loss": 0.6206,
+      "step": 11613
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.302464914369161e-05,
+      "loss": 0.638,
+      "step": 11614
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3023611262847168e-05,
+      "loss": 0.6602,
+      "step": 11615
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.302257334615399e-05,
+      "loss": 0.6329,
+      "step": 11616
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.302153539362438e-05,
+      "loss": 0.6314,
+      "step": 11617
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3020497405270647e-05,
+      "loss": 0.645,
+      "step": 11618
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3019459381105098e-05,
+      "loss": 0.6271,
+      "step": 11619
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3018421321140033e-05,
+      "loss": 0.6122,
+      "step": 11620
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3017383225387768e-05,
+      "loss": 0.6368,
+      "step": 11621
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.301634509386061e-05,
+      "loss": 0.6538,
+      "step": 11622
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3015306926570863e-05,
+      "loss": 0.6039,
+      "step": 11623
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3014268723530837e-05,
+      "loss": 0.634,
+      "step": 11624
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3013230484752844e-05,
+      "loss": 0.6561,
+      "step": 11625
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.301219221024919e-05,
+      "loss": 0.6333,
+      "step": 11626
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3011153900032188e-05,
+      "loss": 0.6076,
+      "step": 11627
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3010115554114147e-05,
+      "loss": 0.6804,
+      "step": 11628
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.300907717250738e-05,
+      "loss": 0.6117,
+      "step": 11629
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3008038755224193e-05,
+      "loss": 0.6682,
+      "step": 11630
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3007000302276905e-05,
+      "loss": 0.6513,
+      "step": 11631
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3005961813677824e-05,
+      "loss": 0.6561,
+      "step": 11632
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.300492328943926e-05,
+      "loss": 0.6379,
+      "step": 11633
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3003884729573532e-05,
+      "loss": 0.6216,
+      "step": 11634
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3002846134092952e-05,
+      "loss": 0.632,
+      "step": 11635
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3001807503009829e-05,
+      "loss": 0.6923,
+      "step": 11636
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3000768836336483e-05,
+      "loss": 0.6048,
+      "step": 11637
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2999730134085226e-05,
+      "loss": 0.6215,
+      "step": 11638
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.299869139626837e-05,
+      "loss": 0.667,
+      "step": 11639
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.299765262289824e-05,
+      "loss": 0.594,
+      "step": 11640
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2996613813987145e-05,
+      "loss": 0.6872,
+      "step": 11641
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2995574969547398e-05,
+      "loss": 0.6531,
+      "step": 11642
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2994536089591324e-05,
+      "loss": 0.5986,
+      "step": 11643
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2993497174131238e-05,
+      "loss": 0.6453,
+      "step": 11644
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2992458223179451e-05,
+      "loss": 0.6342,
+      "step": 11645
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.299141923674829e-05,
+      "loss": 0.5949,
+      "step": 11646
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2990380214850072e-05,
+      "loss": 0.627,
+      "step": 11647
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2989341157497108e-05,
+      "loss": 0.6639,
+      "step": 11648
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2988302064701728e-05,
+      "loss": 0.6603,
+      "step": 11649
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2987262936476246e-05,
+      "loss": 0.6471,
+      "step": 11650
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2986223772832977e-05,
+      "loss": 0.6289,
+      "step": 11651
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2985184573784256e-05,
+      "loss": 0.7017,
+      "step": 11652
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2984145339342393e-05,
+      "loss": 0.6367,
+      "step": 11653
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2983106069519708e-05,
+      "loss": 0.6439,
+      "step": 11654
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.298206676432853e-05,
+      "loss": 0.6715,
+      "step": 11655
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2981027423781179e-05,
+      "loss": 0.6626,
+      "step": 11656
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2979988047889976e-05,
+      "loss": 0.6015,
+      "step": 11657
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2978948636667243e-05,
+      "loss": 0.6134,
+      "step": 11658
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2977909190125311e-05,
+      "loss": 0.5886,
+      "step": 11659
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2976869708276492e-05,
+      "loss": 0.6172,
+      "step": 11660
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2975830191133119e-05,
+      "loss": 0.6073,
+      "step": 11661
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2974790638707517e-05,
+      "loss": 0.6445,
+      "step": 11662
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2973751051012007e-05,
+      "loss": 0.6397,
+      "step": 11663
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2972711428058915e-05,
+      "loss": 0.6362,
+      "step": 11664
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2971671769860572e-05,
+      "loss": 0.6573,
+      "step": 11665
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2970632076429299e-05,
+      "loss": 0.5928,
+      "step": 11666
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2969592347777424e-05,
+      "loss": 0.6853,
+      "step": 11667
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2968552583917279e-05,
+      "loss": 0.681,
+      "step": 11668
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2967512784861183e-05,
+      "loss": 0.6349,
+      "step": 11669
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2966472950621469e-05,
+      "loss": 0.5678,
+      "step": 11670
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.296543308121047e-05,
+      "loss": 0.6649,
+      "step": 11671
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2964393176640508e-05,
+      "loss": 0.6165,
+      "step": 11672
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2963353236923914e-05,
+      "loss": 0.5998,
+      "step": 11673
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2962313262073022e-05,
+      "loss": 0.6044,
+      "step": 11674
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2961273252100155e-05,
+      "loss": 0.6804,
+      "step": 11675
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.296023320701765e-05,
+      "loss": 0.6257,
+      "step": 11676
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2959193126837834e-05,
+      "loss": 0.6319,
+      "step": 11677
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2958153011573042e-05,
+      "loss": 0.6167,
+      "step": 11678
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2957112861235604e-05,
+      "loss": 0.5979,
+      "step": 11679
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2956072675837849e-05,
+      "loss": 0.6092,
+      "step": 11680
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2955032455392115e-05,
+      "loss": 0.6428,
+      "step": 11681
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2953992199910734e-05,
+      "loss": 0.6451,
+      "step": 11682
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.295295190940604e-05,
+      "loss": 0.6561,
+      "step": 11683
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2951911583890362e-05,
+      "loss": 0.6482,
+      "step": 11684
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.295087122337604e-05,
+      "loss": 0.6619,
+      "step": 11685
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2949830827875406e-05,
+      "loss": 0.6586,
+      "step": 11686
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2948790397400796e-05,
+      "loss": 0.6436,
+      "step": 11687
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2947749931964548e-05,
+      "loss": 0.5652,
+      "step": 11688
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.2946709431578993e-05,
+      "loss": 0.6562,
+      "step": 11689
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2945668896256469e-05,
+      "loss": 0.6139,
+      "step": 11690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2944628326009317e-05,
+      "loss": 0.6095,
+      "step": 11691
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.294358772084987e-05,
+      "loss": 0.6238,
+      "step": 11692
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2942547080790467e-05,
+      "loss": 0.6166,
+      "step": 11693
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2941506405843444e-05,
+      "loss": 0.6793,
+      "step": 11694
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2940465696021142e-05,
+      "loss": 0.6755,
+      "step": 11695
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.29394249513359e-05,
+      "loss": 0.6327,
+      "step": 11696
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.293838417180006e-05,
+      "loss": 0.5996,
+      "step": 11697
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2937343357425955e-05,
+      "loss": 0.685,
+      "step": 11698
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.293630250822593e-05,
+      "loss": 0.6123,
+      "step": 11699
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2935261624212326e-05,
+      "loss": 0.6636,
+      "step": 11700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.293422070539748e-05,
+      "loss": 0.6659,
+      "step": 11701
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2933179751793734e-05,
+      "loss": 0.6292,
+      "step": 11702
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2932138763413438e-05,
+      "loss": 0.6269,
+      "step": 11703
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.293109774026892e-05,
+      "loss": 0.6308,
+      "step": 11704
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2930056682372534e-05,
+      "loss": 0.6003,
+      "step": 11705
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2929015589736622e-05,
+      "loss": 0.5863,
+      "step": 11706
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.292797446237352e-05,
+      "loss": 0.623,
+      "step": 11707
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2926933300295579e-05,
+      "loss": 0.6669,
+      "step": 11708
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2925892103515143e-05,
+      "loss": 0.6463,
+      "step": 11709
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2924850872044552e-05,
+      "loss": 0.6057,
+      "step": 11710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2923809605896156e-05,
+      "loss": 0.6458,
+      "step": 11711
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2922768305082298e-05,
+      "loss": 0.6402,
+      "step": 11712
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2921726969615325e-05,
+      "loss": 0.6155,
+      "step": 11713
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.292068559950758e-05,
+      "loss": 0.6006,
+      "step": 11714
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2919644194771417e-05,
+      "loss": 0.6395,
+      "step": 11715
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2918602755419174e-05,
+      "loss": 0.6041,
+      "step": 11716
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2917561281463205e-05,
+      "loss": 0.622,
+      "step": 11717
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.291651977291586e-05,
+      "loss": 0.6658,
+      "step": 11718
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2915478229789478e-05,
+      "loss": 0.6388,
+      "step": 11719
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2914436652096413e-05,
+      "loss": 0.6487,
+      "step": 11720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2913395039849021e-05,
+      "loss": 0.6619,
+      "step": 11721
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.291235339305964e-05,
+      "loss": 0.6253,
+      "step": 11722
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2911311711740626e-05,
+      "loss": 0.5713,
+      "step": 11723
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2910269995904332e-05,
+      "loss": 0.6197,
+      "step": 11724
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2909228245563103e-05,
+      "loss": 0.6425,
+      "step": 11725
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2908186460729294e-05,
+      "loss": 0.5766,
+      "step": 11726
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2907144641415256e-05,
+      "loss": 0.6258,
+      "step": 11727
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.290610278763334e-05,
+      "loss": 0.6755,
+      "step": 11728
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2905060899395897e-05,
+      "loss": 0.6255,
+      "step": 11729
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2904018976715287e-05,
+      "loss": 0.6511,
+      "step": 11730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2902977019603855e-05,
+      "loss": 0.633,
+      "step": 11731
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2901935028073957e-05,
+      "loss": 0.6165,
+      "step": 11732
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2900893002137953e-05,
+      "loss": 0.5828,
+      "step": 11733
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.289985094180819e-05,
+      "loss": 0.6591,
+      "step": 11734
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2898808847097025e-05,
+      "loss": 0.5781,
+      "step": 11735
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2897766718016815e-05,
+      "loss": 0.6462,
+      "step": 11736
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2896724554579916e-05,
+      "loss": 0.6387,
+      "step": 11737
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.289568235679868e-05,
+      "loss": 0.6461,
+      "step": 11738
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2894640124685469e-05,
+      "loss": 0.6447,
+      "step": 11739
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2893597858252637e-05,
+      "loss": 0.6014,
+      "step": 11740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2892555557512544e-05,
+      "loss": 0.6021,
+      "step": 11741
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2891513222477543e-05,
+      "loss": 0.6364,
+      "step": 11742
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2890470853159995e-05,
+      "loss": 0.6596,
+      "step": 11743
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2889428449572258e-05,
+      "loss": 0.5865,
+      "step": 11744
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2888386011726695e-05,
+      "loss": 0.5962,
+      "step": 11745
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.288734353963566e-05,
+      "loss": 0.6969,
+      "step": 11746
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2886301033311515e-05,
+      "loss": 0.6304,
+      "step": 11747
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.288525849276662e-05,
+      "loss": 0.6261,
+      "step": 11748
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2884215918013335e-05,
+      "loss": 0.5893,
+      "step": 11749
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2883173309064022e-05,
+      "loss": 0.6418,
+      "step": 11750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2882130665931044e-05,
+      "loss": 0.6183,
+      "step": 11751
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2881087988626756e-05,
+      "loss": 0.6507,
+      "step": 11752
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2880045277163531e-05,
+      "loss": 0.6118,
+      "step": 11753
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2879002531553724e-05,
+      "loss": 0.6316,
+      "step": 11754
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2877959751809699e-05,
+      "loss": 0.6159,
+      "step": 11755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2876916937943822e-05,
+      "loss": 0.6541,
+      "step": 11756
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2875874089968452e-05,
+      "loss": 0.6414,
+      "step": 11757
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2874831207895962e-05,
+      "loss": 0.597,
+      "step": 11758
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2873788291738707e-05,
+      "loss": 0.6786,
+      "step": 11759
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2872745341509059e-05,
+      "loss": 0.6358,
+      "step": 11760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.287170235721938e-05,
+      "loss": 0.6547,
+      "step": 11761
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2870659338882037e-05,
+      "loss": 0.6543,
+      "step": 11762
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2869616286509394e-05,
+      "loss": 0.6043,
+      "step": 11763
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2868573200113823e-05,
+      "loss": 0.6794,
+      "step": 11764
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2867530079707686e-05,
+      "loss": 0.6508,
+      "step": 11765
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2866486925303355e-05,
+      "loss": 0.585,
+      "step": 11766
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2865443736913194e-05,
+      "loss": 0.6419,
+      "step": 11767
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2864400514549571e-05,
+      "loss": 0.6185,
+      "step": 11768
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2863357258224855e-05,
+      "loss": 0.702,
+      "step": 11769
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2862313967951423e-05,
+      "loss": 0.6028,
+      "step": 11770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2861270643741633e-05,
+      "loss": 0.5854,
+      "step": 11771
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.286022728560786e-05,
+      "loss": 0.6297,
+      "step": 11772
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2859183893562479e-05,
+      "loss": 0.6441,
+      "step": 11773
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.285814046761785e-05,
+      "loss": 0.6251,
+      "step": 11774
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2857097007786351e-05,
+      "loss": 0.6314,
+      "step": 11775
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2856053514080357e-05,
+      "loss": 0.6889,
+      "step": 11776
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2855009986512234e-05,
+      "loss": 0.6164,
+      "step": 11777
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2853966425094355e-05,
+      "loss": 0.6779,
+      "step": 11778
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2852922829839094e-05,
+      "loss": 0.7002,
+      "step": 11779
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2851879200758823e-05,
+      "loss": 0.646,
+      "step": 11780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.285083553786592e-05,
+      "loss": 0.6573,
+      "step": 11781
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2849791841172752e-05,
+      "loss": 0.6158,
+      "step": 11782
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2848748110691696e-05,
+      "loss": 0.62,
+      "step": 11783
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.284770434643513e-05,
+      "loss": 0.578,
+      "step": 11784
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2846660548415426e-05,
+      "loss": 0.574,
+      "step": 11785
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2845616716644958e-05,
+      "loss": 0.6502,
+      "step": 11786
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2844572851136108e-05,
+      "loss": 0.6563,
+      "step": 11787
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2843528951901245e-05,
+      "loss": 0.6478,
+      "step": 11788
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2842485018952749e-05,
+      "loss": 0.5999,
+      "step": 11789
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2841441052303001e-05,
+      "loss": 0.6947,
+      "step": 11790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.284039705196437e-05,
+      "loss": 0.6663,
+      "step": 11791
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.283935301794924e-05,
+      "loss": 0.6595,
+      "step": 11792
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2838308950269993e-05,
+      "loss": 0.6273,
+      "step": 11793
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2837264848939e-05,
+      "loss": 0.6173,
+      "step": 11794
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2836220713968639e-05,
+      "loss": 0.6468,
+      "step": 11795
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2835176545371299e-05,
+      "loss": 0.6583,
+      "step": 11796
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2834132343159353e-05,
+      "loss": 0.6092,
+      "step": 11797
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2833088107345181e-05,
+      "loss": 0.6711,
+      "step": 11798
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2832043837941168e-05,
+      "loss": 0.5535,
+      "step": 11799
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2830999534959692e-05,
+      "loss": 0.6278,
+      "step": 11800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2829955198413136e-05,
+      "loss": 0.6406,
+      "step": 11801
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.282891082831388e-05,
+      "loss": 0.6265,
+      "step": 11802
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2827866424674307e-05,
+      "loss": 0.6646,
+      "step": 11803
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2826821987506802e-05,
+      "loss": 0.6351,
+      "step": 11804
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2825777516823745e-05,
+      "loss": 0.6403,
+      "step": 11805
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2824733012637523e-05,
+      "loss": 0.6291,
+      "step": 11806
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2823688474960514e-05,
+      "loss": 0.6516,
+      "step": 11807
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2822643903805113e-05,
+      "loss": 0.6368,
+      "step": 11808
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2821599299183692e-05,
+      "loss": 0.6445,
+      "step": 11809
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2820554661108639e-05,
+      "loss": 0.6193,
+      "step": 11810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2819509989592351e-05,
+      "loss": 0.5917,
+      "step": 11811
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.28184652846472e-05,
+      "loss": 0.6594,
+      "step": 11812
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2817420546285576e-05,
+      "loss": 0.6318,
+      "step": 11813
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2816375774519874e-05,
+      "loss": 0.636,
+      "step": 11814
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.281533096936247e-05,
+      "loss": 0.6827,
+      "step": 11815
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2814286130825755e-05,
+      "loss": 0.6577,
+      "step": 11816
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.281324125892212e-05,
+      "loss": 0.6207,
+      "step": 11817
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.281219635366395e-05,
+      "loss": 0.6206,
+      "step": 11818
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2811151415063633e-05,
+      "loss": 0.6548,
+      "step": 11819
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2810106443133565e-05,
+      "loss": 0.6766,
+      "step": 11820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2809061437886127e-05,
+      "loss": 0.6309,
+      "step": 11821
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2808016399333712e-05,
+      "loss": 0.5911,
+      "step": 11822
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2806971327488713e-05,
+      "loss": 0.6217,
+      "step": 11823
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2805926222363516e-05,
+      "loss": 0.6376,
+      "step": 11824
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2804881083970513e-05,
+      "loss": 0.6343,
+      "step": 11825
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.28038359123221e-05,
+      "loss": 0.6556,
+      "step": 11826
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2802790707430662e-05,
+      "loss": 0.6182,
+      "step": 11827
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2801745469308596e-05,
+      "loss": 0.6836,
+      "step": 11828
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2800700197968294e-05,
+      "loss": 0.6449,
+      "step": 11829
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2799654893422146e-05,
+      "loss": 0.6428,
+      "step": 11830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2798609555682549e-05,
+      "loss": 0.621,
+      "step": 11831
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2797564184761897e-05,
+      "loss": 0.6654,
+      "step": 11832
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.279651878067258e-05,
+      "loss": 0.5915,
+      "step": 11833
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2795473343426994e-05,
+      "loss": 0.602,
+      "step": 11834
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2794427873037541e-05,
+      "loss": 0.6715,
+      "step": 11835
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2793382369516606e-05,
+      "loss": 0.6285,
+      "step": 11836
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.279233683287659e-05,
+      "loss": 0.5942,
+      "step": 11837
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.279129126312989e-05,
+      "loss": 0.6349,
+      "step": 11838
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2790245660288898e-05,
+      "loss": 0.6021,
+      "step": 11839
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.278920002436602e-05,
+      "loss": 0.5839,
+      "step": 11840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.278815435537364e-05,
+      "loss": 0.5869,
+      "step": 11841
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2787108653324168e-05,
+      "loss": 0.6652,
+      "step": 11842
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2786062918229996e-05,
+      "loss": 0.5825,
+      "step": 11843
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2785017150103522e-05,
+      "loss": 0.5937,
+      "step": 11844
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.278397134895715e-05,
+      "loss": 0.6321,
+      "step": 11845
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2782925514803275e-05,
+      "loss": 0.6052,
+      "step": 11846
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2781879647654293e-05,
+      "loss": 0.657,
+      "step": 11847
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2780833747522614e-05,
+      "loss": 0.6195,
+      "step": 11848
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2779787814420632e-05,
+      "loss": 0.6846,
+      "step": 11849
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2778741848360747e-05,
+      "loss": 0.6231,
+      "step": 11850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2777695849355365e-05,
+      "loss": 0.6157,
+      "step": 11851
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2776649817416884e-05,
+      "loss": 0.6495,
+      "step": 11852
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2775603752557706e-05,
+      "loss": 0.6412,
+      "step": 11853
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2774557654790235e-05,
+      "loss": 0.6245,
+      "step": 11854
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2773511524126873e-05,
+      "loss": 0.5888,
+      "step": 11855
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2772465360580024e-05,
+      "loss": 0.6322,
+      "step": 11856
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2771419164162088e-05,
+      "loss": 0.5741,
+      "step": 11857
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2770372934885477e-05,
+      "loss": 0.6675,
+      "step": 11858
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2769326672762587e-05,
+      "loss": 0.6646,
+      "step": 11859
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2768280377805827e-05,
+      "loss": 0.6796,
+      "step": 11860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2767234050027602e-05,
+      "loss": 0.6545,
+      "step": 11861
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2766187689440315e-05,
+      "loss": 0.6305,
+      "step": 11862
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2765141296056373e-05,
+      "loss": 0.6526,
+      "step": 11863
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2764094869888185e-05,
+      "loss": 0.6807,
+      "step": 11864
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2763048410948156e-05,
+      "loss": 0.6601,
+      "step": 11865
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2762001919248693e-05,
+      "loss": 0.6946,
+      "step": 11866
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2760955394802201e-05,
+      "loss": 0.6262,
+      "step": 11867
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2759908837621091e-05,
+      "loss": 0.6233,
+      "step": 11868
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2758862247717772e-05,
+      "loss": 0.6521,
+      "step": 11869
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.275781562510465e-05,
+      "loss": 0.5662,
+      "step": 11870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2756768969794134e-05,
+      "loss": 0.6021,
+      "step": 11871
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2755722281798635e-05,
+      "loss": 0.6774,
+      "step": 11872
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2754675561130566e-05,
+      "loss": 0.6452,
+      "step": 11873
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2753628807802333e-05,
+      "loss": 0.6454,
+      "step": 11874
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2752582021826338e-05,
+      "loss": 0.6288,
+      "step": 11875
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.275153520321501e-05,
+      "loss": 0.649,
+      "step": 11876
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2750488351980751e-05,
+      "loss": 0.6409,
+      "step": 11877
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2749441468135968e-05,
+      "loss": 0.6205,
+      "step": 11878
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2748394551693084e-05,
+      "loss": 0.6285,
+      "step": 11879
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2747347602664504e-05,
+      "loss": 0.6819,
+      "step": 11880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.274630062106264e-05,
+      "loss": 0.6775,
+      "step": 11881
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2745253606899913e-05,
+      "loss": 0.6228,
+      "step": 11882
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.274420656018873e-05,
+      "loss": 0.6235,
+      "step": 11883
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2743159480941502e-05,
+      "loss": 0.6745,
+      "step": 11884
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2742112369170654e-05,
+      "loss": 0.6305,
+      "step": 11885
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2741065224888596e-05,
+      "loss": 0.6367,
+      "step": 11886
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2740018048107734e-05,
+      "loss": 0.6495,
+      "step": 11887
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2738970838840498e-05,
+      "loss": 0.5928,
+      "step": 11888
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2737923597099301e-05,
+      "loss": 0.667,
+      "step": 11889
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.273687632289655e-05,
+      "loss": 0.6198,
+      "step": 11890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2735829016244673e-05,
+      "loss": 0.5857,
+      "step": 11891
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.273478167715608e-05,
+      "loss": 0.673,
+      "step": 11892
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2733734305643189e-05,
+      "loss": 0.6235,
+      "step": 11893
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.273268690171842e-05,
+      "loss": 0.6102,
+      "step": 11894
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2731639465394194e-05,
+      "loss": 0.641,
+      "step": 11895
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2730591996682922e-05,
+      "loss": 0.5735,
+      "step": 11896
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2729544495597032e-05,
+      "loss": 0.6216,
+      "step": 11897
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2728496962148937e-05,
+      "loss": 0.5745,
+      "step": 11898
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2727449396351057e-05,
+      "loss": 0.6524,
+      "step": 11899
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2726401798215818e-05,
+      "loss": 0.6615,
+      "step": 11900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2725354167755637e-05,
+      "loss": 0.6541,
+      "step": 11901
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2724306504982931e-05,
+      "loss": 0.6308,
+      "step": 11902
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2723258809910128e-05,
+      "loss": 0.6195,
+      "step": 11903
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2722211082549647e-05,
+      "loss": 0.6444,
+      "step": 11904
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2721163322913908e-05,
+      "loss": 0.6061,
+      "step": 11905
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2720115531015336e-05,
+      "loss": 0.6176,
+      "step": 11906
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2719067706866355e-05,
+      "loss": 0.6656,
+      "step": 11907
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2718019850479383e-05,
+      "loss": 0.6555,
+      "step": 11908
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2716971961866852e-05,
+      "loss": 0.6285,
+      "step": 11909
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.271592404104118e-05,
+      "loss": 0.7077,
+      "step": 11910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2714876088014792e-05,
+      "loss": 0.6303,
+      "step": 11911
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2713828102800111e-05,
+      "loss": 0.6183,
+      "step": 11912
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2712780085409571e-05,
+      "loss": 0.608,
+      "step": 11913
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2711732035855589e-05,
+      "loss": 0.6418,
+      "step": 11914
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.271068395415059e-05,
+      "loss": 0.6206,
+      "step": 11915
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2709635840307008e-05,
+      "loss": 0.7225,
+      "step": 11916
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2708587694337262e-05,
+      "loss": 0.641,
+      "step": 11917
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2707539516253784e-05,
+      "loss": 0.6347,
+      "step": 11918
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2706491306069003e-05,
+      "loss": 0.6292,
+      "step": 11919
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2705443063795338e-05,
+      "loss": 0.623,
+      "step": 11920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2704394789445225e-05,
+      "loss": 0.6173,
+      "step": 11921
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2703346483031094e-05,
+      "loss": 0.66,
+      "step": 11922
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2702298144565366e-05,
+      "loss": 0.6021,
+      "step": 11923
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2701249774060478e-05,
+      "loss": 0.6635,
+      "step": 11924
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2700201371528858e-05,
+      "loss": 0.6615,
+      "step": 11925
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.269915293698293e-05,
+      "loss": 0.623,
+      "step": 11926
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2698104470435133e-05,
+      "loss": 0.6438,
+      "step": 11927
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2697055971897895e-05,
+      "loss": 0.6791,
+      "step": 11928
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2696007441383644e-05,
+      "loss": 0.6225,
+      "step": 11929
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2694958878904818e-05,
+      "loss": 0.5513,
+      "step": 11930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2693910284473843e-05,
+      "loss": 0.6199,
+      "step": 11931
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2692861658103155e-05,
+      "loss": 0.6527,
+      "step": 11932
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2691812999805184e-05,
+      "loss": 0.6866,
+      "step": 11933
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2690764309592365e-05,
+      "loss": 0.572,
+      "step": 11934
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2689715587477132e-05,
+      "loss": 0.6729,
+      "step": 11935
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2688666833471919e-05,
+      "loss": 0.6806,
+      "step": 11936
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2687618047589157e-05,
+      "loss": 0.6906,
+      "step": 11937
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2686569229841284e-05,
+      "loss": 0.6503,
+      "step": 11938
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2685520380240738e-05,
+      "loss": 0.6499,
+      "step": 11939
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2684471498799946e-05,
+      "loss": 0.6173,
+      "step": 11940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2683422585531348e-05,
+      "loss": 0.6257,
+      "step": 11941
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2682373640447385e-05,
+      "loss": 0.6634,
+      "step": 11942
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2681324663560485e-05,
+      "loss": 0.6615,
+      "step": 11943
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2680275654883092e-05,
+      "loss": 0.6638,
+      "step": 11944
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2679226614427642e-05,
+      "loss": 0.6138,
+      "step": 11945
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2678177542206566e-05,
+      "loss": 0.63,
+      "step": 11946
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2677128438232312e-05,
+      "loss": 0.6126,
+      "step": 11947
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2676079302517312e-05,
+      "loss": 0.6154,
+      "step": 11948
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2675030135074006e-05,
+      "loss": 0.6408,
+      "step": 11949
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2673980935914833e-05,
+      "loss": 0.662,
+      "step": 11950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2672931705052238e-05,
+      "loss": 0.6209,
+      "step": 11951
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2671882442498654e-05,
+      "loss": 0.6286,
+      "step": 11952
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.267083314826652e-05,
+      "loss": 0.6548,
+      "step": 11953
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2669783822368285e-05,
+      "loss": 0.6473,
+      "step": 11954
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2668734464816383e-05,
+      "loss": 0.648,
+      "step": 11955
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2667685075623259e-05,
+      "loss": 0.6055,
+      "step": 11956
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2666635654801357e-05,
+      "loss": 0.5884,
+      "step": 11957
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.266558620236311e-05,
+      "loss": 0.6116,
+      "step": 11958
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2664536718320967e-05,
+      "loss": 0.62,
+      "step": 11959
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2663487202687374e-05,
+      "loss": 0.6028,
+      "step": 11960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2662437655474771e-05,
+      "loss": 0.6314,
+      "step": 11961
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.26613880766956e-05,
+      "loss": 0.6582,
+      "step": 11962
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2660338466362307e-05,
+      "loss": 0.6002,
+      "step": 11963
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2659288824487334e-05,
+      "loss": 0.5731,
+      "step": 11964
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.265823915108313e-05,
+      "loss": 0.5678,
+      "step": 11965
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.265718944616214e-05,
+      "loss": 0.5807,
+      "step": 11966
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2656139709736808e-05,
+      "loss": 0.6444,
+      "step": 11967
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2655089941819575e-05,
+      "loss": 0.6808,
+      "step": 11968
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2654040142422897e-05,
+      "loss": 0.6703,
+      "step": 11969
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2652990311559213e-05,
+      "loss": 0.6675,
+      "step": 11970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2651940449240972e-05,
+      "loss": 0.63,
+      "step": 11971
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2650890555480624e-05,
+      "loss": 0.6159,
+      "step": 11972
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2649840630290616e-05,
+      "loss": 0.6411,
+      "step": 11973
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2648790673683392e-05,
+      "loss": 0.643,
+      "step": 11974
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2647740685671406e-05,
+      "loss": 0.6262,
+      "step": 11975
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2646690666267105e-05,
+      "loss": 0.5891,
+      "step": 11976
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.2645640615482934e-05,
+      "loss": 0.6016,
+      "step": 11977
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2644590533331353e-05,
+      "loss": 0.6542,
+      "step": 11978
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2643540419824805e-05,
+      "loss": 0.5919,
+      "step": 11979
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2642490274975737e-05,
+      "loss": 0.6353,
+      "step": 11980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2641440098796607e-05,
+      "loss": 0.5982,
+      "step": 11981
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2640389891299863e-05,
+      "loss": 0.6446,
+      "step": 11982
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2639339652497954e-05,
+      "loss": 0.6476,
+      "step": 11983
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2638289382403336e-05,
+      "loss": 0.635,
+      "step": 11984
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2637239081028466e-05,
+      "loss": 0.5707,
+      "step": 11985
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.263618874838578e-05,
+      "loss": 0.6599,
+      "step": 11986
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2635138384487748e-05,
+      "loss": 0.6502,
+      "step": 11987
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2634087989346818e-05,
+      "loss": 0.6034,
+      "step": 11988
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.263303756297544e-05,
+      "loss": 0.6271,
+      "step": 11989
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2631987105386072e-05,
+      "loss": 0.6609,
+      "step": 11990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2630936616591167e-05,
+      "loss": 0.5422,
+      "step": 11991
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2629886096603178e-05,
+      "loss": 0.5951,
+      "step": 11992
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2628835545434565e-05,
+      "loss": 0.664,
+      "step": 11993
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2627784963097782e-05,
+      "loss": 0.5848,
+      "step": 11994
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.262673434960528e-05,
+      "loss": 0.6124,
+      "step": 11995
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2625683704969525e-05,
+      "loss": 0.6053,
+      "step": 11996
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2624633029202967e-05,
+      "loss": 0.6927,
+      "step": 11997
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2623582322318062e-05,
+      "loss": 0.6251,
+      "step": 11998
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2622531584327269e-05,
+      "loss": 0.6119,
+      "step": 11999
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.262148081524305e-05,
+      "loss": 0.6875,
+      "step": 12000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2620430015077858e-05,
+      "loss": 0.6582,
+      "step": 12001
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2619379183844152e-05,
+      "loss": 0.6218,
+      "step": 12002
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2618328321554394e-05,
+      "loss": 0.6661,
+      "step": 12003
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.261727742822104e-05,
+      "loss": 0.6616,
+      "step": 12004
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2616226503856552e-05,
+      "loss": 0.6252,
+      "step": 12005
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.261517554847339e-05,
+      "loss": 0.6658,
+      "step": 12006
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2614124562084014e-05,
+      "loss": 0.6617,
+      "step": 12007
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2613073544700882e-05,
+      "loss": 0.6358,
+      "step": 12008
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2612022496336461e-05,
+      "loss": 0.5801,
+      "step": 12009
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2610971417003206e-05,
+      "loss": 0.6359,
+      "step": 12010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.260992030671358e-05,
+      "loss": 0.7121,
+      "step": 12011
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2608869165480054e-05,
+      "loss": 0.6491,
+      "step": 12012
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.260781799331508e-05,
+      "loss": 0.6104,
+      "step": 12013
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2606766790231122e-05,
+      "loss": 0.6526,
+      "step": 12014
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2605715556240653e-05,
+      "loss": 0.661,
+      "step": 12015
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2604664291356123e-05,
+      "loss": 0.6072,
+      "step": 12016
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2603612995590004e-05,
+      "loss": 0.6028,
+      "step": 12017
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2602561668954762e-05,
+      "loss": 0.6294,
+      "step": 12018
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.260151031146286e-05,
+      "loss": 0.6005,
+      "step": 12019
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.260045892312676e-05,
+      "loss": 0.6817,
+      "step": 12020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2599407503958931e-05,
+      "loss": 0.5866,
+      "step": 12021
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2598356053971836e-05,
+      "loss": 0.6354,
+      "step": 12022
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2597304573177944e-05,
+      "loss": 0.5778,
+      "step": 12023
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2596253061589722e-05,
+      "loss": 0.6374,
+      "step": 12024
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2595201519219633e-05,
+      "loss": 0.5986,
+      "step": 12025
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2594149946080148e-05,
+      "loss": 0.6128,
+      "step": 12026
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2593098342183736e-05,
+      "loss": 0.624,
+      "step": 12027
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.259204670754286e-05,
+      "loss": 0.6458,
+      "step": 12028
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2590995042169996e-05,
+      "loss": 0.6434,
+      "step": 12029
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2589943346077602e-05,
+      "loss": 0.678,
+      "step": 12030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2588891619278156e-05,
+      "loss": 0.6706,
+      "step": 12031
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2587839861784126e-05,
+      "loss": 0.637,
+      "step": 12032
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2586788073607981e-05,
+      "loss": 0.6214,
+      "step": 12033
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2585736254762188e-05,
+      "loss": 0.667,
+      "step": 12034
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2584684405259223e-05,
+      "loss": 0.6364,
+      "step": 12035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2583632525111555e-05,
+      "loss": 0.6142,
+      "step": 12036
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2582580614331654e-05,
+      "loss": 0.6136,
+      "step": 12037
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2581528672931996e-05,
+      "loss": 0.6176,
+      "step": 12038
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2580476700925045e-05,
+      "loss": 0.6357,
+      "step": 12039
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.257942469832328e-05,
+      "loss": 0.6274,
+      "step": 12040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2578372665139174e-05,
+      "loss": 0.6293,
+      "step": 12041
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2577320601385197e-05,
+      "loss": 0.6523,
+      "step": 12042
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2576268507073822e-05,
+      "loss": 0.6445,
+      "step": 12043
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.257521638221753e-05,
+      "loss": 0.6114,
+      "step": 12044
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2574164226828785e-05,
+      "loss": 0.6372,
+      "step": 12045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2573112040920068e-05,
+      "loss": 0.6347,
+      "step": 12046
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2572059824503855e-05,
+      "loss": 0.6091,
+      "step": 12047
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2571007577592618e-05,
+      "loss": 0.6587,
+      "step": 12048
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2569955300198833e-05,
+      "loss": 0.609,
+      "step": 12049
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.256890299233498e-05,
+      "loss": 0.6242,
+      "step": 12050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2567850654013527e-05,
+      "loss": 0.5633,
+      "step": 12051
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2566798285246957e-05,
+      "loss": 0.5836,
+      "step": 12052
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2565745886047751e-05,
+      "loss": 0.6156,
+      "step": 12053
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2564693456428379e-05,
+      "loss": 0.6241,
+      "step": 12054
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2563640996401321e-05,
+      "loss": 0.6531,
+      "step": 12055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2562588505979059e-05,
+      "loss": 0.674,
+      "step": 12056
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2561535985174066e-05,
+      "loss": 0.5882,
+      "step": 12057
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2560483433998823e-05,
+      "loss": 0.671,
+      "step": 12058
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.255943085246581e-05,
+      "loss": 0.6189,
+      "step": 12059
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2558378240587508e-05,
+      "loss": 0.5531,
+      "step": 12060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2557325598376393e-05,
+      "loss": 0.6633,
+      "step": 12061
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2556272925844952e-05,
+      "loss": 0.6434,
+      "step": 12062
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.255522022300566e-05,
+      "loss": 0.6112,
+      "step": 12063
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2554167489870999e-05,
+      "loss": 0.5965,
+      "step": 12064
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.255311472645345e-05,
+      "loss": 0.6308,
+      "step": 12065
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2552061932765498e-05,
+      "loss": 0.6215,
+      "step": 12066
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2551009108819626e-05,
+      "loss": 0.5968,
+      "step": 12067
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.254995625462831e-05,
+      "loss": 0.5998,
+      "step": 12068
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2548903370204038e-05,
+      "loss": 0.6188,
+      "step": 12069
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2547850455559292e-05,
+      "loss": 0.6615,
+      "step": 12070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2546797510706559e-05,
+      "loss": 0.6478,
+      "step": 12071
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2545744535658316e-05,
+      "loss": 0.6617,
+      "step": 12072
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2544691530427052e-05,
+      "loss": 0.6658,
+      "step": 12073
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2543638495025252e-05,
+      "loss": 0.6031,
+      "step": 12074
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2542585429465398e-05,
+      "loss": 0.6657,
+      "step": 12075
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2541532333759978e-05,
+      "loss": 0.6226,
+      "step": 12076
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2540479207921478e-05,
+      "loss": 0.6292,
+      "step": 12077
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2539426051962383e-05,
+      "loss": 0.6425,
+      "step": 12078
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2538372865895179e-05,
+      "loss": 0.6334,
+      "step": 12079
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2537319649732356e-05,
+      "loss": 0.6184,
+      "step": 12080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2536266403486395e-05,
+      "loss": 0.6615,
+      "step": 12081
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.253521312716979e-05,
+      "loss": 0.6809,
+      "step": 12082
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2534159820795027e-05,
+      "loss": 0.6615,
+      "step": 12083
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2533106484374592e-05,
+      "loss": 0.6223,
+      "step": 12084
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2532053117920975e-05,
+      "loss": 0.6146,
+      "step": 12085
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2530999721446667e-05,
+      "loss": 0.6612,
+      "step": 12086
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2529946294964153e-05,
+      "loss": 0.6337,
+      "step": 12087
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2528892838485925e-05,
+      "loss": 0.6302,
+      "step": 12088
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2527839352024476e-05,
+      "loss": 0.6754,
+      "step": 12089
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2526785835592292e-05,
+      "loss": 0.6016,
+      "step": 12090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2525732289201868e-05,
+      "loss": 0.6102,
+      "step": 12091
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.252467871286569e-05,
+      "loss": 0.6833,
+      "step": 12092
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2523625106596254e-05,
+      "loss": 0.6276,
+      "step": 12093
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.252257147040605e-05,
+      "loss": 0.6095,
+      "step": 12094
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2521517804307569e-05,
+      "loss": 0.6116,
+      "step": 12095
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2520464108313301e-05,
+      "loss": 0.6423,
+      "step": 12096
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.251941038243575e-05,
+      "loss": 0.6469,
+      "step": 12097
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2518356626687397e-05,
+      "loss": 0.6447,
+      "step": 12098
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.251730284108074e-05,
+      "loss": 0.6662,
+      "step": 12099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2516249025628277e-05,
+      "loss": 0.624,
+      "step": 12100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2515195180342496e-05,
+      "loss": 0.6376,
+      "step": 12101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2514141305235893e-05,
+      "loss": 0.618,
+      "step": 12102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.251308740032097e-05,
+      "loss": 0.6191,
+      "step": 12103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2512033465610211e-05,
+      "loss": 0.6774,
+      "step": 12104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.251097950111612e-05,
+      "loss": 0.6078,
+      "step": 12105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2509925506851192e-05,
+      "loss": 0.6407,
+      "step": 12106
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.250887148282792e-05,
+      "loss": 0.678,
+      "step": 12107
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2507817429058804e-05,
+      "loss": 0.6598,
+      "step": 12108
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2506763345556341e-05,
+      "loss": 0.6309,
+      "step": 12109
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2505709232333025e-05,
+      "loss": 0.6805,
+      "step": 12110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2504655089401355e-05,
+      "loss": 0.6521,
+      "step": 12111
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2503600916773835e-05,
+      "loss": 0.567,
+      "step": 12112
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2502546714462955e-05,
+      "loss": 0.6558,
+      "step": 12113
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.250149248248122e-05,
+      "loss": 0.6203,
+      "step": 12114
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2500438220841129e-05,
+      "loss": 0.6358,
+      "step": 12115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2499383929555177e-05,
+      "loss": 0.6728,
+      "step": 12116
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2498329608635866e-05,
+      "loss": 0.6371,
+      "step": 12117
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.24972752580957e-05,
+      "loss": 0.5915,
+      "step": 12118
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2496220877947174e-05,
+      "loss": 0.7062,
+      "step": 12119
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2495166468202793e-05,
+      "loss": 0.6686,
+      "step": 12120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2494112028875059e-05,
+      "loss": 0.6482,
+      "step": 12121
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2493057559976471e-05,
+      "loss": 0.6178,
+      "step": 12122
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.249200306151953e-05,
+      "loss": 0.686,
+      "step": 12123
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2490948533516742e-05,
+      "loss": 0.6352,
+      "step": 12124
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2489893975980608e-05,
+      "loss": 0.6035,
+      "step": 12125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2488839388923631e-05,
+      "loss": 0.6175,
+      "step": 12126
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2487784772358313e-05,
+      "loss": 0.644,
+      "step": 12127
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2486730126297162e-05,
+      "loss": 0.5991,
+      "step": 12128
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2485675450752677e-05,
+      "loss": 0.598,
+      "step": 12129
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2484620745737365e-05,
+      "loss": 0.6522,
+      "step": 12130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2483566011263735e-05,
+      "loss": 0.6943,
+      "step": 12131
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2482511247344284e-05,
+      "loss": 0.6212,
+      "step": 12132
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2481456453991523e-05,
+      "loss": 0.561,
+      "step": 12133
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2480401631217957e-05,
+      "loss": 0.6157,
+      "step": 12134
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2479346779036092e-05,
+      "loss": 0.606,
+      "step": 12135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2478291897458434e-05,
+      "loss": 0.5476,
+      "step": 12136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2477236986497494e-05,
+      "loss": 0.6337,
+      "step": 12137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2476182046165772e-05,
+      "loss": 0.5823,
+      "step": 12138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.247512707647578e-05,
+      "loss": 0.656,
+      "step": 12139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2474072077440024e-05,
+      "loss": 0.6325,
+      "step": 12140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2473017049071016e-05,
+      "loss": 0.6555,
+      "step": 12141
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2471961991381263e-05,
+      "loss": 0.649,
+      "step": 12142
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2470906904383274e-05,
+      "loss": 0.6277,
+      "step": 12143
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2469851788089556e-05,
+      "loss": 0.591,
+      "step": 12144
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.246879664251262e-05,
+      "loss": 0.6563,
+      "step": 12145
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2467741467664979e-05,
+      "loss": 0.5781,
+      "step": 12146
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.246668626355914e-05,
+      "loss": 0.611,
+      "step": 12147
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2465631030207615e-05,
+      "loss": 0.6461,
+      "step": 12148
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2464575767622915e-05,
+      "loss": 0.612,
+      "step": 12149
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2463520475817553e-05,
+      "loss": 0.7178,
+      "step": 12150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2462465154804035e-05,
+      "loss": 0.6439,
+      "step": 12151
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2461409804594884e-05,
+      "loss": 0.6987,
+      "step": 12152
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.24603544252026e-05,
+      "loss": 0.5058,
+      "step": 12153
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2459299016639705e-05,
+      "loss": 0.6741,
+      "step": 12154
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.245824357891871e-05,
+      "loss": 0.6432,
+      "step": 12155
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2457188112052126e-05,
+      "loss": 0.6283,
+      "step": 12156
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.245613261605247e-05,
+      "loss": 0.5668,
+      "step": 12157
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2455077090932256e-05,
+      "loss": 0.635,
+      "step": 12158
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2454021536703993e-05,
+      "loss": 0.6168,
+      "step": 12159
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.24529659533802e-05,
+      "loss": 0.6612,
+      "step": 12160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2451910340973398e-05,
+      "loss": 0.6312,
+      "step": 12161
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2450854699496093e-05,
+      "loss": 0.5652,
+      "step": 12162
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2449799028960806e-05,
+      "loss": 0.6759,
+      "step": 12163
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2448743329380055e-05,
+      "loss": 0.6692,
+      "step": 12164
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2447687600766352e-05,
+      "loss": 0.572,
+      "step": 12165
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2446631843132215e-05,
+      "loss": 0.6181,
+      "step": 12166
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2445576056490168e-05,
+      "loss": 0.6539,
+      "step": 12167
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2444520240852716e-05,
+      "loss": 0.5925,
+      "step": 12168
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2443464396232387e-05,
+      "loss": 0.5918,
+      "step": 12169
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2442408522641698e-05,
+      "loss": 0.6459,
+      "step": 12170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2441352620093166e-05,
+      "loss": 0.6119,
+      "step": 12171
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.244029668859931e-05,
+      "loss": 0.6828,
+      "step": 12172
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.243924072817265e-05,
+      "loss": 0.6543,
+      "step": 12173
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2438184738825703e-05,
+      "loss": 0.6157,
+      "step": 12174
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2437128720570992e-05,
+      "loss": 0.6783,
+      "step": 12175
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.243607267342104e-05,
+      "loss": 0.6331,
+      "step": 12176
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2435016597388362e-05,
+      "loss": 0.634,
+      "step": 12177
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2433960492485485e-05,
+      "loss": 0.6329,
+      "step": 12178
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2432904358724926e-05,
+      "loss": 0.545,
+      "step": 12179
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2431848196119208e-05,
+      "loss": 0.6074,
+      "step": 12180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2430792004680851e-05,
+      "loss": 0.6627,
+      "step": 12181
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2429735784422384e-05,
+      "loss": 0.6496,
+      "step": 12182
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2428679535356323e-05,
+      "loss": 0.6198,
+      "step": 12183
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2427623257495196e-05,
+      "loss": 0.5881,
+      "step": 12184
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2426566950851522e-05,
+      "loss": 0.6862,
+      "step": 12185
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2425510615437828e-05,
+      "loss": 0.6652,
+      "step": 12186
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2424454251266637e-05,
+      "loss": 0.6216,
+      "step": 12187
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2423397858350475e-05,
+      "loss": 0.6247,
+      "step": 12188
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2422341436701863e-05,
+      "loss": 0.6774,
+      "step": 12189
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2421284986333333e-05,
+      "loss": 0.6166,
+      "step": 12190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2420228507257403e-05,
+      "loss": 0.6231,
+      "step": 12191
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2419171999486603e-05,
+      "loss": 0.6409,
+      "step": 12192
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2418115463033462e-05,
+      "loss": 0.5985,
+      "step": 12193
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2417058897910498e-05,
+      "loss": 0.6276,
+      "step": 12194
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2416002304130248e-05,
+      "loss": 0.6413,
+      "step": 12195
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2414945681705233e-05,
+      "loss": 0.5814,
+      "step": 12196
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2413889030647981e-05,
+      "loss": 0.6048,
+      "step": 12197
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.241283235097102e-05,
+      "loss": 0.6404,
+      "step": 12198
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2411775642686883e-05,
+      "loss": 0.6282,
+      "step": 12199
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2410718905808093e-05,
+      "loss": 0.6084,
+      "step": 12200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2409662140347178e-05,
+      "loss": 0.6341,
+      "step": 12201
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2408605346316673e-05,
+      "loss": 0.6786,
+      "step": 12202
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2407548523729105e-05,
+      "loss": 0.6336,
+      "step": 12203
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2406491672597e-05,
+      "loss": 0.6341,
+      "step": 12204
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2405434792932898e-05,
+      "loss": 0.617,
+      "step": 12205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2404377884749318e-05,
+      "loss": 0.6502,
+      "step": 12206
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2403320948058799e-05,
+      "loss": 0.6488,
+      "step": 12207
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2402263982873873e-05,
+      "loss": 0.6438,
+      "step": 12208
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2401206989207064e-05,
+      "loss": 0.6379,
+      "step": 12209
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.240014996707091e-05,
+      "loss": 0.6239,
+      "step": 12210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2399092916477944e-05,
+      "loss": 0.6578,
+      "step": 12211
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2398035837440693e-05,
+      "loss": 0.6387,
+      "step": 12212
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2396978729971696e-05,
+      "loss": 0.6515,
+      "step": 12213
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2395921594083485e-05,
+      "loss": 0.6241,
+      "step": 12214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.239486442978859e-05,
+      "loss": 0.6594,
+      "step": 12215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2393807237099547e-05,
+      "loss": 0.5963,
+      "step": 12216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2392750016028893e-05,
+      "loss": 0.6324,
+      "step": 12217
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2391692766589163e-05,
+      "loss": 0.6508,
+      "step": 12218
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2390635488792883e-05,
+      "loss": 0.6476,
+      "step": 12219
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2389578182652601e-05,
+      "loss": 0.6893,
+      "step": 12220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2388520848180848e-05,
+      "loss": 0.5969,
+      "step": 12221
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2387463485390152e-05,
+      "loss": 0.6893,
+      "step": 12222
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2386406094293062e-05,
+      "loss": 0.6315,
+      "step": 12223
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.238534867490211e-05,
+      "loss": 0.6589,
+      "step": 12224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2384291227229825e-05,
+      "loss": 0.6472,
+      "step": 12225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2383233751288757e-05,
+      "loss": 0.5981,
+      "step": 12226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2382176247091438e-05,
+      "loss": 0.5929,
+      "step": 12227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2381118714650402e-05,
+      "loss": 0.5563,
+      "step": 12228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2380061153978193e-05,
+      "loss": 0.598,
+      "step": 12229
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2379003565087352e-05,
+      "loss": 0.5917,
+      "step": 12230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2377945947990408e-05,
+      "loss": 0.6445,
+      "step": 12231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2376888302699913e-05,
+      "loss": 0.6779,
+      "step": 12232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2375830629228398e-05,
+      "loss": 0.624,
+      "step": 12233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2374772927588404e-05,
+      "loss": 0.6621,
+      "step": 12234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2373715197792475e-05,
+      "loss": 0.6125,
+      "step": 12235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.237265743985315e-05,
+      "loss": 0.6283,
+      "step": 12236
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2371599653782964e-05,
+      "loss": 0.6729,
+      "step": 12237
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.237054183959447e-05,
+      "loss": 0.6291,
+      "step": 12238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2369483997300205e-05,
+      "loss": 0.6862,
+      "step": 12239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2368426126912703e-05,
+      "loss": 0.6315,
+      "step": 12240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2367368228444519e-05,
+      "loss": 0.647,
+      "step": 12241
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2366310301908188e-05,
+      "loss": 0.6044,
+      "step": 12242
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2365252347316252e-05,
+      "loss": 0.6245,
+      "step": 12243
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2364194364681261e-05,
+      "loss": 0.66,
+      "step": 12244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2363136354015757e-05,
+      "loss": 0.6179,
+      "step": 12245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2362078315332278e-05,
+      "loss": 0.6039,
+      "step": 12246
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2361020248643375e-05,
+      "loss": 0.6148,
+      "step": 12247
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2359962153961589e-05,
+      "loss": 0.6658,
+      "step": 12248
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2358904031299465e-05,
+      "loss": 0.6039,
+      "step": 12249
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.235784588066955e-05,
+      "loss": 0.6582,
+      "step": 12250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2356787702084391e-05,
+      "loss": 0.6251,
+      "step": 12251
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2355729495556533e-05,
+      "loss": 0.6502,
+      "step": 12252
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.235467126109852e-05,
+      "loss": 0.6279,
+      "step": 12253
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2353612998722903e-05,
+      "loss": 0.6411,
+      "step": 12254
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2352554708442223e-05,
+      "loss": 0.6332,
+      "step": 12255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2351496390269034e-05,
+      "loss": 0.6682,
+      "step": 12256
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.235043804421588e-05,
+      "loss": 0.6139,
+      "step": 12257
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2349379670295308e-05,
+      "loss": 0.6324,
+      "step": 12258
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2348321268519869e-05,
+      "loss": 0.5844,
+      "step": 12259
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2347262838902113e-05,
+      "loss": 0.6301,
+      "step": 12260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2346204381454582e-05,
+      "loss": 0.6645,
+      "step": 12261
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2345145896189831e-05,
+      "loss": 0.5927,
+      "step": 12262
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2344087383120414e-05,
+      "loss": 0.6365,
+      "step": 12263
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.234302884225887e-05,
+      "loss": 0.6792,
+      "step": 12264
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2341970273617759e-05,
+      "loss": 0.6505,
+      "step": 12265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2340911677209625e-05,
+      "loss": 0.6822,
+      "step": 12266
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2339853053047024e-05,
+      "loss": 0.6174,
+      "step": 12267
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2338794401142501e-05,
+      "loss": 0.6704,
+      "step": 12268
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2337735721508619e-05,
+      "loss": 0.6253,
+      "step": 12269
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2336677014157916e-05,
+      "loss": 0.6455,
+      "step": 12270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2335618279102955e-05,
+      "loss": 0.644,
+      "step": 12271
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2334559516356284e-05,
+      "loss": 0.6063,
+      "step": 12272
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2333500725930456e-05,
+      "loss": 0.6483,
+      "step": 12273
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2332441907838023e-05,
+      "loss": 0.6238,
+      "step": 12274
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2331383062091544e-05,
+      "loss": 0.625,
+      "step": 12275
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2330324188703569e-05,
+      "loss": 0.5965,
+      "step": 12276
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2329265287686653e-05,
+      "loss": 0.634,
+      "step": 12277
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.232820635905335e-05,
+      "loss": 0.6173,
+      "step": 12278
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2327147402816213e-05,
+      "loss": 0.6554,
+      "step": 12279
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2326088418987804e-05,
+      "loss": 0.6247,
+      "step": 12280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2325029407580671e-05,
+      "loss": 0.6381,
+      "step": 12281
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2323970368607374e-05,
+      "loss": 0.6408,
+      "step": 12282
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.232291130208047e-05,
+      "loss": 0.6209,
+      "step": 12283
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2321852208012513e-05,
+      "loss": 0.5895,
+      "step": 12284
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2320793086416058e-05,
+      "loss": 0.6582,
+      "step": 12285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2319733937303671e-05,
+      "loss": 0.5961,
+      "step": 12286
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.23186747606879e-05,
+      "loss": 0.6508,
+      "step": 12287
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2317615556581306e-05,
+      "loss": 0.6607,
+      "step": 12288
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.231655632499645e-05,
+      "loss": 0.6647,
+      "step": 12289
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2315497065945887e-05,
+      "loss": 0.6718,
+      "step": 12290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2314437779442176e-05,
+      "loss": 0.5943,
+      "step": 12291
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.231337846549788e-05,
+      "loss": 0.6463,
+      "step": 12292
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2312319124125554e-05,
+      "loss": 0.5895,
+      "step": 12293
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.231125975533776e-05,
+      "loss": 0.6473,
+      "step": 12294
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.231020035914706e-05,
+      "loss": 0.6423,
+      "step": 12295
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2309140935566008e-05,
+      "loss": 0.6768,
+      "step": 12296
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2308081484607171e-05,
+      "loss": 0.6304,
+      "step": 12297
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2307022006283109e-05,
+      "loss": 0.6424,
+      "step": 12298
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2305962500606384e-05,
+      "loss": 0.6223,
+      "step": 12299
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2304902967589554e-05,
+      "loss": 0.6195,
+      "step": 12300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2303843407245186e-05,
+      "loss": 0.6531,
+      "step": 12301
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2302783819585838e-05,
+      "loss": 0.5924,
+      "step": 12302
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2301724204624075e-05,
+      "loss": 0.6594,
+      "step": 12303
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2300664562372462e-05,
+      "loss": 0.6832,
+      "step": 12304
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2299604892843555e-05,
+      "loss": 0.6227,
+      "step": 12305
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2298545196049927e-05,
+      "loss": 0.6847,
+      "step": 12306
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2297485472004139e-05,
+      "loss": 0.6335,
+      "step": 12307
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2296425720718752e-05,
+      "loss": 0.6472,
+      "step": 12308
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2295365942206335e-05,
+      "loss": 0.6287,
+      "step": 12309
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.229430613647945e-05,
+      "loss": 0.61,
+      "step": 12310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2293246303550664e-05,
+      "loss": 0.6117,
+      "step": 12311
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2292186443432538e-05,
+      "loss": 0.6681,
+      "step": 12312
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.229112655613765e-05,
+      "loss": 0.6618,
+      "step": 12313
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2290066641678554e-05,
+      "loss": 0.6469,
+      "step": 12314
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2289006700067818e-05,
+      "loss": 0.635,
+      "step": 12315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2287946731318017e-05,
+      "loss": 0.6534,
+      "step": 12316
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2286886735441712e-05,
+      "loss": 0.5996,
+      "step": 12317
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.228582671245147e-05,
+      "loss": 0.5855,
+      "step": 12318
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2284766662359861e-05,
+      "loss": 0.652,
+      "step": 12319
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2283706585179458e-05,
+      "loss": 0.6206,
+      "step": 12320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2282646480922814e-05,
+      "loss": 0.7107,
+      "step": 12321
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.228158634960252e-05,
+      "loss": 0.5949,
+      "step": 12322
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2280526191231128e-05,
+      "loss": 0.6002,
+      "step": 12323
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2279466005821212e-05,
+      "loss": 0.6046,
+      "step": 12324
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2278405793385346e-05,
+      "loss": 0.6437,
+      "step": 12325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2277345553936095e-05,
+      "loss": 0.6517,
+      "step": 12326
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.227628528748603e-05,
+      "loss": 0.6402,
+      "step": 12327
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2275224994047727e-05,
+      "loss": 0.6364,
+      "step": 12328
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2274164673633755e-05,
+      "loss": 0.5786,
+      "step": 12329
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2273104326256678e-05,
+      "loss": 0.6569,
+      "step": 12330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2272043951929079e-05,
+      "loss": 0.7115,
+      "step": 12331
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2270983550663522e-05,
+      "loss": 0.6018,
+      "step": 12332
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.226992312247258e-05,
+      "loss": 0.5841,
+      "step": 12333
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2268862667368833e-05,
+      "loss": 0.7047,
+      "step": 12334
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2267802185364848e-05,
+      "loss": 0.6713,
+      "step": 12335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2266741676473195e-05,
+      "loss": 0.6621,
+      "step": 12336
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2265681140706455e-05,
+      "loss": 0.6275,
+      "step": 12337
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2264620578077202e-05,
+      "loss": 0.6528,
+      "step": 12338
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2263559988598004e-05,
+      "loss": 0.6086,
+      "step": 12339
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2262499372281438e-05,
+      "loss": 0.6547,
+      "step": 12340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2261438729140083e-05,
+      "loss": 0.642,
+      "step": 12341
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2260378059186507e-05,
+      "loss": 0.6277,
+      "step": 12342
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2259317362433292e-05,
+      "loss": 0.6898,
+      "step": 12343
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2258256638893012e-05,
+      "loss": 0.6253,
+      "step": 12344
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2257195888578241e-05,
+      "loss": 0.6226,
+      "step": 12345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2256135111501559e-05,
+      "loss": 0.6332,
+      "step": 12346
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2255074307675543e-05,
+      "loss": 0.6378,
+      "step": 12347
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2254013477112765e-05,
+      "loss": 0.6684,
+      "step": 12348
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2252952619825807e-05,
+      "loss": 0.6281,
+      "step": 12349
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2251891735827246e-05,
+      "loss": 0.6346,
+      "step": 12350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.225083082512966e-05,
+      "loss": 0.6409,
+      "step": 12351
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2249769887745626e-05,
+      "loss": 0.6108,
+      "step": 12352
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2248708923687727e-05,
+      "loss": 0.6384,
+      "step": 12353
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2247647932968536e-05,
+      "loss": 0.6524,
+      "step": 12354
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2246586915600636e-05,
+      "loss": 0.5898,
+      "step": 12355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2245525871596606e-05,
+      "loss": 0.6734,
+      "step": 12356
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2244464800969026e-05,
+      "loss": 0.6042,
+      "step": 12357
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2243403703730479e-05,
+      "loss": 0.5873,
+      "step": 12358
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2242342579893543e-05,
+      "loss": 0.6386,
+      "step": 12359
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2241281429470796e-05,
+      "loss": 0.6482,
+      "step": 12360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2240220252474824e-05,
+      "loss": 0.6335,
+      "step": 12361
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2239159048918208e-05,
+      "loss": 0.6069,
+      "step": 12362
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2238097818813529e-05,
+      "loss": 0.633,
+      "step": 12363
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2237036562173365e-05,
+      "loss": 0.6432,
+      "step": 12364
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2235975279010308e-05,
+      "loss": 0.5738,
+      "step": 12365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2234913969336932e-05,
+      "loss": 0.681,
+      "step": 12366
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2233852633165826e-05,
+      "loss": 0.6669,
+      "step": 12367
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2232791270509572e-05,
+      "loss": 0.6509,
+      "step": 12368
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.223172988138075e-05,
+      "loss": 0.6331,
+      "step": 12369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2230668465791948e-05,
+      "loss": 0.7068,
+      "step": 12370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.222960702375575e-05,
+      "loss": 0.6557,
+      "step": 12371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2228545555284739e-05,
+      "loss": 0.644,
+      "step": 12372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2227484060391505e-05,
+      "loss": 0.5907,
+      "step": 12373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2226422539088628e-05,
+      "loss": 0.5993,
+      "step": 12374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2225360991388694e-05,
+      "loss": 0.616,
+      "step": 12375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2224299417304291e-05,
+      "loss": 0.5881,
+      "step": 12376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2223237816848003e-05,
+      "loss": 0.6058,
+      "step": 12377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.222217619003242e-05,
+      "loss": 0.583,
+      "step": 12378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2221114536870128e-05,
+      "loss": 0.6362,
+      "step": 12379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2220052857373712e-05,
+      "loss": 0.599,
+      "step": 12380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2218991151555761e-05,
+      "loss": 0.6883,
+      "step": 12381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2217929419428865e-05,
+      "loss": 0.6221,
+      "step": 12382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2216867661005608e-05,
+      "loss": 0.6056,
+      "step": 12383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2215805876298579e-05,
+      "loss": 0.5876,
+      "step": 12384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2214744065320373e-05,
+      "loss": 0.6774,
+      "step": 12385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2213682228083569e-05,
+      "loss": 0.5628,
+      "step": 12386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2212620364600765e-05,
+      "loss": 0.6386,
+      "step": 12387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2211558474884548e-05,
+      "loss": 0.6475,
+      "step": 12388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2210496558947506e-05,
+      "loss": 0.5916,
+      "step": 12389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2209434616802231e-05,
+      "loss": 0.5946,
+      "step": 12390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2208372648461318e-05,
+      "loss": 0.6063,
+      "step": 12391
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2207310653937348e-05,
+      "loss": 0.6688,
+      "step": 12392
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2206248633242919e-05,
+      "loss": 0.6337,
+      "step": 12393
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2205186586390623e-05,
+      "loss": 0.6551,
+      "step": 12394
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.220412451339305e-05,
+      "loss": 0.6222,
+      "step": 12395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2203062414262792e-05,
+      "loss": 0.6431,
+      "step": 12396
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2202000289012444e-05,
+      "loss": 0.6325,
+      "step": 12397
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2200938137654596e-05,
+      "loss": 0.6826,
+      "step": 12398
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2199875960201841e-05,
+      "loss": 0.6641,
+      "step": 12399
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2198813756666776e-05,
+      "loss": 0.5924,
+      "step": 12400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2197751527061992e-05,
+      "loss": 0.5875,
+      "step": 12401
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.219668927140008e-05,
+      "loss": 0.627,
+      "step": 12402
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2195626989693643e-05,
+      "loss": 0.6674,
+      "step": 12403
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2194564681955265e-05,
+      "loss": 0.5876,
+      "step": 12404
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.219350234819755e-05,
+      "loss": 0.6092,
+      "step": 12405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.219243998843309e-05,
+      "loss": 0.663,
+      "step": 12406
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2191377602674481e-05,
+      "loss": 0.6325,
+      "step": 12407
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2190315190934317e-05,
+      "loss": 0.616,
+      "step": 12408
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2189252753225196e-05,
+      "loss": 0.6257,
+      "step": 12409
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2188190289559713e-05,
+      "loss": 0.5655,
+      "step": 12410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2187127799950466e-05,
+      "loss": 0.6458,
+      "step": 12411
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2186065284410056e-05,
+      "loss": 0.616,
+      "step": 12412
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2185002742951073e-05,
+      "loss": 0.6164,
+      "step": 12413
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.218394017558612e-05,
+      "loss": 0.6779,
+      "step": 12414
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2182877582327794e-05,
+      "loss": 0.6123,
+      "step": 12415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2181814963188692e-05,
+      "loss": 0.6524,
+      "step": 12416
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2180752318181412e-05,
+      "loss": 0.657,
+      "step": 12417
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2179689647318558e-05,
+      "loss": 0.6229,
+      "step": 12418
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2178626950612722e-05,
+      "loss": 0.6851,
+      "step": 12419
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2177564228076508e-05,
+      "loss": 0.6361,
+      "step": 12420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.217650147972252e-05,
+      "loss": 0.6537,
+      "step": 12421
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.217543870556335e-05,
+      "loss": 0.6626,
+      "step": 12422
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2174375905611601e-05,
+      "loss": 0.6793,
+      "step": 12423
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.217331307987988e-05,
+      "loss": 0.6006,
+      "step": 12424
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.217225022838078e-05,
+      "loss": 0.6369,
+      "step": 12425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2171187351126902e-05,
+      "loss": 0.6287,
+      "step": 12426
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2170124448130858e-05,
+      "loss": 0.6221,
+      "step": 12427
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.216906151940524e-05,
+      "loss": 0.631,
+      "step": 12428
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2167998564962653e-05,
+      "loss": 0.6445,
+      "step": 12429
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2166935584815705e-05,
+      "loss": 0.6152,
+      "step": 12430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2165872578976989e-05,
+      "loss": 0.6574,
+      "step": 12431
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2164809547459117e-05,
+      "loss": 0.6178,
+      "step": 12432
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2163746490274689e-05,
+      "loss": 0.6603,
+      "step": 12433
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2162683407436308e-05,
+      "loss": 0.6618,
+      "step": 12434
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2161620298956583e-05,
+      "loss": 0.64,
+      "step": 12435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.216055716484811e-05,
+      "loss": 0.6079,
+      "step": 12436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.21594940051235e-05,
+      "loss": 0.6686,
+      "step": 12437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2158430819795357e-05,
+      "loss": 0.5726,
+      "step": 12438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2157367608876288e-05,
+      "loss": 0.6712,
+      "step": 12439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2156304372378892e-05,
+      "loss": 0.5716,
+      "step": 12440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2155241110315784e-05,
+      "loss": 0.6307,
+      "step": 12441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2154177822699566e-05,
+      "loss": 0.6387,
+      "step": 12442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2153114509542845e-05,
+      "loss": 0.6067,
+      "step": 12443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2152051170858227e-05,
+      "loss": 0.6397,
+      "step": 12444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2150987806658321e-05,
+      "loss": 0.581,
+      "step": 12445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2149924416955732e-05,
+      "loss": 0.6316,
+      "step": 12446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2148861001763072e-05,
+      "loss": 0.6368,
+      "step": 12447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2147797561092945e-05,
+      "loss": 0.5999,
+      "step": 12448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.214673409495796e-05,
+      "loss": 0.6573,
+      "step": 12449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2145670603370727e-05,
+      "loss": 0.6066,
+      "step": 12450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2144607086343853e-05,
+      "loss": 0.6622,
+      "step": 12451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2143543543889953e-05,
+      "loss": 0.6017,
+      "step": 12452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.214247997602163e-05,
+      "loss": 0.6889,
+      "step": 12453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.21414163827515e-05,
+      "loss": 0.5907,
+      "step": 12454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2140352764092164e-05,
+      "loss": 0.6209,
+      "step": 12455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2139289120056243e-05,
+      "loss": 0.6526,
+      "step": 12456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2138225450656343e-05,
+      "loss": 0.6629,
+      "step": 12457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2137161755905073e-05,
+      "loss": 0.6228,
+      "step": 12458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2136098035815048e-05,
+      "loss": 0.6258,
+      "step": 12459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.213503429039888e-05,
+      "loss": 0.6289,
+      "step": 12460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2133970519669176e-05,
+      "loss": 0.6202,
+      "step": 12461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2132906723638558e-05,
+      "loss": 0.5917,
+      "step": 12462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2131842902319631e-05,
+      "loss": 0.6072,
+      "step": 12463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2130779055725003e-05,
+      "loss": 0.658,
+      "step": 12464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2129715183867301e-05,
+      "loss": 0.6292,
+      "step": 12465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.212865128675913e-05,
+      "loss": 0.6454,
+      "step": 12466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.21275873644131e-05,
+      "loss": 0.6504,
+      "step": 12467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2126523416841838e-05,
+      "loss": 0.6619,
+      "step": 12468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2125459444057948e-05,
+      "loss": 0.6357,
+      "step": 12469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2124395446074044e-05,
+      "loss": 0.6244,
+      "step": 12470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2123331422902749e-05,
+      "loss": 0.5957,
+      "step": 12471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2122267374556676e-05,
+      "loss": 0.612,
+      "step": 12472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.212120330104843e-05,
+      "loss": 0.6665,
+      "step": 12473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2120139202390641e-05,
+      "loss": 0.5995,
+      "step": 12474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2119075078595923e-05,
+      "loss": 0.612,
+      "step": 12475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2118010929676882e-05,
+      "loss": 0.6767,
+      "step": 12476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.211694675564615e-05,
+      "loss": 0.669,
+      "step": 12477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2115882556516331e-05,
+      "loss": 0.6447,
+      "step": 12478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2114818332300047e-05,
+      "loss": 0.6006,
+      "step": 12479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2113754083009918e-05,
+      "loss": 0.6479,
+      "step": 12480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2112689808658562e-05,
+      "loss": 0.6225,
+      "step": 12481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2111625509258593e-05,
+      "loss": 0.5769,
+      "step": 12482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2110561184822635e-05,
+      "loss": 0.6168,
+      "step": 12483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2109496835363304e-05,
+      "loss": 0.5834,
+      "step": 12484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2108432460893215e-05,
+      "loss": 0.5912,
+      "step": 12485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2107368061424997e-05,
+      "loss": 0.6648,
+      "step": 12486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2106303636971266e-05,
+      "loss": 0.6185,
+      "step": 12487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2105239187544637e-05,
+      "loss": 0.6463,
+      "step": 12488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2104174713157735e-05,
+      "loss": 0.5882,
+      "step": 12489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2103110213823181e-05,
+      "loss": 0.6367,
+      "step": 12490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2102045689553595e-05,
+      "loss": 0.6445,
+      "step": 12491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2100981140361595e-05,
+      "loss": 0.6644,
+      "step": 12492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2099916566259812e-05,
+      "loss": 0.6485,
+      "step": 12493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2098851967260856e-05,
+      "loss": 0.6162,
+      "step": 12494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2097787343377359e-05,
+      "loss": 0.6561,
+      "step": 12495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2096722694621939e-05,
+      "loss": 0.6449,
+      "step": 12496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2095658021007215e-05,
+      "loss": 0.6697,
+      "step": 12497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2094593322545815e-05,
+      "loss": 0.6856,
+      "step": 12498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2093528599250366e-05,
+      "loss": 0.6774,
+      "step": 12499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2092463851133483e-05,
+      "loss": 0.5839,
+      "step": 12500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2091399078207795e-05,
+      "loss": 0.6057,
+      "step": 12501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2090334280485927e-05,
+      "loss": 0.6208,
+      "step": 12502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.20892694579805e-05,
+      "loss": 0.6477,
+      "step": 12503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208820461070414e-05,
+      "loss": 0.6095,
+      "step": 12504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2087139738669476e-05,
+      "loss": 0.6419,
+      "step": 12505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2086074841889126e-05,
+      "loss": 0.626,
+      "step": 12506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208500992037572e-05,
+      "loss": 0.5913,
+      "step": 12507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2083944974141888e-05,
+      "loss": 0.6286,
+      "step": 12508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208288000320025e-05,
+      "loss": 0.6541,
+      "step": 12509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2081815007563433e-05,
+      "loss": 0.616,
+      "step": 12510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.208074998724407e-05,
+      "loss": 0.6218,
+      "step": 12511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2079684942254779e-05,
+      "loss": 0.6568,
+      "step": 12512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2078619872608194e-05,
+      "loss": 0.6392,
+      "step": 12513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2077554778316942e-05,
+      "loss": 0.6055,
+      "step": 12514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2076489659393649e-05,
+      "loss": 0.5993,
+      "step": 12515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2075424515850944e-05,
+      "loss": 0.6349,
+      "step": 12516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2074359347701457e-05,
+      "loss": 0.6177,
+      "step": 12517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2073294154957815e-05,
+      "loss": 0.6057,
+      "step": 12518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2072228937632646e-05,
+      "loss": 0.6585,
+      "step": 12519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2071163695738586e-05,
+      "loss": 0.5935,
+      "step": 12520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2070098429288257e-05,
+      "loss": 0.6437,
+      "step": 12521
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2069033138294293e-05,
+      "loss": 0.6464,
+      "step": 12522
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2067967822769326e-05,
+      "loss": 0.6202,
+      "step": 12523
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2066902482725983e-05,
+      "loss": 0.6316,
+      "step": 12524
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2065837118176895e-05,
+      "loss": 0.6336,
+      "step": 12525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2064771729134696e-05,
+      "loss": 0.6096,
+      "step": 12526
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2063706315612016e-05,
+      "loss": 0.6139,
+      "step": 12527
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2062640877621487e-05,
+      "loss": 0.6533,
+      "step": 12528
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2061575415175742e-05,
+      "loss": 0.6362,
+      "step": 12529
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2060509928287412e-05,
+      "loss": 0.6271,
+      "step": 12530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.205944441696913e-05,
+      "loss": 0.618,
+      "step": 12531
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2058378881233528e-05,
+      "loss": 0.641,
+      "step": 12532
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2057313321093242e-05,
+      "loss": 0.5907,
+      "step": 12533
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2056247736560905e-05,
+      "loss": 0.634,
+      "step": 12534
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2055182127649148e-05,
+      "loss": 0.6343,
+      "step": 12535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2054116494370608e-05,
+      "loss": 0.682,
+      "step": 12536
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2053050836737919e-05,
+      "loss": 0.595,
+      "step": 12537
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2051985154763712e-05,
+      "loss": 0.7031,
+      "step": 12538
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2050919448460627e-05,
+      "loss": 0.6338,
+      "step": 12539
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.20498537178413e-05,
+      "loss": 0.649,
+      "step": 12540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2048787962918362e-05,
+      "loss": 0.6066,
+      "step": 12541
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2047722183704451e-05,
+      "loss": 0.6351,
+      "step": 12542
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2046656380212205e-05,
+      "loss": 0.5854,
+      "step": 12543
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2045590552454256e-05,
+      "loss": 0.6418,
+      "step": 12544
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2044524700443244e-05,
+      "loss": 0.638,
+      "step": 12545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2043458824191807e-05,
+      "loss": 0.5957,
+      "step": 12546
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2042392923712578e-05,
+      "loss": 0.65,
+      "step": 12547
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2041326999018197e-05,
+      "loss": 0.6112,
+      "step": 12548
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2040261050121307e-05,
+      "loss": 0.6233,
+      "step": 12549
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2039195077034536e-05,
+      "loss": 0.6537,
+      "step": 12550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2038129079770529e-05,
+      "loss": 0.6329,
+      "step": 12551
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2037063058341926e-05,
+      "loss": 0.67,
+      "step": 12552
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2035997012761361e-05,
+      "loss": 0.5886,
+      "step": 12553
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2034930943041477e-05,
+      "loss": 0.6083,
+      "step": 12554
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2033864849194912e-05,
+      "loss": 0.6487,
+      "step": 12555
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2032798731234308e-05,
+      "loss": 0.6312,
+      "step": 12556
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.20317325891723e-05,
+      "loss": 0.6452,
+      "step": 12557
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2030666423021536e-05,
+      "loss": 0.6501,
+      "step": 12558
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2029600232794652e-05,
+      "loss": 0.6429,
+      "step": 12559
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2028534018504285e-05,
+      "loss": 0.6527,
+      "step": 12560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2027467780163087e-05,
+      "loss": 0.6628,
+      "step": 12561
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2026401517783692e-05,
+      "loss": 0.6326,
+      "step": 12562
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2025335231378741e-05,
+      "loss": 0.6601,
+      "step": 12563
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2024268920960882e-05,
+      "loss": 0.5929,
+      "step": 12564
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2023202586542753e-05,
+      "loss": 0.6225,
+      "step": 12565
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2022136228136996e-05,
+      "loss": 0.6663,
+      "step": 12566
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.202106984575626e-05,
+      "loss": 0.6412,
+      "step": 12567
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2020003439413181e-05,
+      "loss": 0.6557,
+      "step": 12568
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2018937009120406e-05,
+      "loss": 0.669,
+      "step": 12569
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2017870554890581e-05,
+      "loss": 0.6428,
+      "step": 12570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2016804076736348e-05,
+      "loss": 0.6298,
+      "step": 12571
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2015737574670347e-05,
+      "loss": 0.6124,
+      "step": 12572
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2014671048705232e-05,
+      "loss": 0.6354,
+      "step": 12573
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2013604498853645e-05,
+      "loss": 0.5723,
+      "step": 12574
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2012537925128223e-05,
+      "loss": 0.6029,
+      "step": 12575
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2011471327541622e-05,
+      "loss": 0.6367,
+      "step": 12576
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2010404706106484e-05,
+      "loss": 0.6363,
+      "step": 12577
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2009338060835452e-05,
+      "loss": 0.6331,
+      "step": 12578
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2008271391741178e-05,
+      "loss": 0.651,
+      "step": 12579
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2007204698836305e-05,
+      "loss": 0.6465,
+      "step": 12580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.200613798213348e-05,
+      "loss": 0.6459,
+      "step": 12581
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2005071241645354e-05,
+      "loss": 0.6751,
+      "step": 12582
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2004004477384574e-05,
+      "loss": 0.5774,
+      "step": 12583
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2002937689363781e-05,
+      "loss": 0.6259,
+      "step": 12584
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.200187087759563e-05,
+      "loss": 0.6164,
+      "step": 12585
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2000804042092768e-05,
+      "loss": 0.6496,
+      "step": 12586
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1999737182867843e-05,
+      "loss": 0.6226,
+      "step": 12587
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1998670299933503e-05,
+      "loss": 0.5615,
+      "step": 12588
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.19976033933024e-05,
+      "loss": 0.665,
+      "step": 12589
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1996536462987179e-05,
+      "loss": 0.6256,
+      "step": 12590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1995469509000493e-05,
+      "loss": 0.6356,
+      "step": 12591
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1994402531354993e-05,
+      "loss": 0.6204,
+      "step": 12592
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1993335530063328e-05,
+      "loss": 0.5809,
+      "step": 12593
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.199226850513815e-05,
+      "loss": 0.6236,
+      "step": 12594
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1991201456592108e-05,
+      "loss": 0.6363,
+      "step": 12595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1990134384437853e-05,
+      "loss": 0.6434,
+      "step": 12596
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1989067288688037e-05,
+      "loss": 0.6244,
+      "step": 12597
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1988000169355311e-05,
+      "loss": 0.6103,
+      "step": 12598
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.198693302645233e-05,
+      "loss": 0.6208,
+      "step": 12599
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1985865859991742e-05,
+      "loss": 0.6299,
+      "step": 12600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1984798669986209e-05,
+      "loss": 0.6651,
+      "step": 12601
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.198373145644837e-05,
+      "loss": 0.6806,
+      "step": 12602
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1982664219390887e-05,
+      "loss": 0.6291,
+      "step": 12603
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1981596958826414e-05,
+      "loss": 0.6457,
+      "step": 12604
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1980529674767599e-05,
+      "loss": 0.6588,
+      "step": 12605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.19794623672271e-05,
+      "loss": 0.6166,
+      "step": 12606
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1978395036217572e-05,
+      "loss": 0.6155,
+      "step": 12607
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1977327681751668e-05,
+      "loss": 0.663,
+      "step": 12608
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.197626030384204e-05,
+      "loss": 0.6535,
+      "step": 12609
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1975192902501349e-05,
+      "loss": 0.6021,
+      "step": 12610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1974125477742247e-05,
+      "loss": 0.5813,
+      "step": 12611
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1973058029577388e-05,
+      "loss": 0.6916,
+      "step": 12612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1971990558019435e-05,
+      "loss": 0.6338,
+      "step": 12613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1970923063081038e-05,
+      "loss": 0.61,
+      "step": 12614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1969855544774851e-05,
+      "loss": 0.6133,
+      "step": 12615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1968788003113539e-05,
+      "loss": 0.6184,
+      "step": 12616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.196772043810975e-05,
+      "loss": 0.6692,
+      "step": 12617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1966652849776152e-05,
+      "loss": 0.6262,
+      "step": 12618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1965585238125392e-05,
+      "loss": 0.5753,
+      "step": 12619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1964517603170134e-05,
+      "loss": 0.6067,
+      "step": 12620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1963449944923036e-05,
+      "loss": 0.6148,
+      "step": 12621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1962382263396752e-05,
+      "loss": 0.5804,
+      "step": 12622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1961314558603946e-05,
+      "loss": 0.6628,
+      "step": 12623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1960246830557276e-05,
+      "loss": 0.6711,
+      "step": 12624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1959179079269399e-05,
+      "loss": 0.6151,
+      "step": 12625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1958111304752973e-05,
+      "loss": 0.656,
+      "step": 12626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1957043507020665e-05,
+      "loss": 0.6393,
+      "step": 12627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1955975686085129e-05,
+      "loss": 0.6423,
+      "step": 12628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1954907841959026e-05,
+      "loss": 0.6265,
+      "step": 12629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1953839974655018e-05,
+      "loss": 0.5778,
+      "step": 12630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1952772084185765e-05,
+      "loss": 0.6244,
+      "step": 12631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1951704170563927e-05,
+      "loss": 0.6063,
+      "step": 12632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1950636233802172e-05,
+      "loss": 0.5795,
+      "step": 12633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1949568273913153e-05,
+      "loss": 0.6208,
+      "step": 12634
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1948500290909535e-05,
+      "loss": 0.6288,
+      "step": 12635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1947432284803984e-05,
+      "loss": 0.6394,
+      "step": 12636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1946364255609157e-05,
+      "loss": 0.6421,
+      "step": 12637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1945296203337721e-05,
+      "loss": 0.6537,
+      "step": 12638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.194422812800234e-05,
+      "loss": 0.6441,
+      "step": 12639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1943160029615672e-05,
+      "loss": 0.5801,
+      "step": 12640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1942091908190385e-05,
+      "loss": 0.6717,
+      "step": 12641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1941023763739143e-05,
+      "loss": 0.551,
+      "step": 12642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1939955596274606e-05,
+      "loss": 0.6687,
+      "step": 12643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1938887405809443e-05,
+      "loss": 0.638,
+      "step": 12644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1937819192356317e-05,
+      "loss": 0.671,
+      "step": 12645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1936750955927891e-05,
+      "loss": 0.644,
+      "step": 12646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1935682696536835e-05,
+      "loss": 0.6247,
+      "step": 12647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1934614414195814e-05,
+      "loss": 0.5936,
+      "step": 12648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.193354610891749e-05,
+      "loss": 0.6279,
+      "step": 12649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1932477780714526e-05,
+      "loss": 0.628,
+      "step": 12650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.19314094295996e-05,
+      "loss": 0.6129,
+      "step": 12651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1930341055585369e-05,
+      "loss": 0.6277,
+      "step": 12652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1929272658684502e-05,
+      "loss": 0.6551,
+      "step": 12653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1928204238909671e-05,
+      "loss": 0.6629,
+      "step": 12654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1927135796273534e-05,
+      "loss": 0.6592,
+      "step": 12655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1926067330788769e-05,
+      "loss": 0.6123,
+      "step": 12656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1924998842468039e-05,
+      "loss": 0.6082,
+      "step": 12657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.192393033132401e-05,
+      "loss": 0.6154,
+      "step": 12658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1922861797369351e-05,
+      "loss": 0.6284,
+      "step": 12659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.192179324061674e-05,
+      "loss": 0.6903,
+      "step": 12660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1920724661078832e-05,
+      "loss": 0.5583,
+      "step": 12661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1919656058768308e-05,
+      "loss": 0.666,
+      "step": 12662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1918587433697832e-05,
+      "loss": 0.6793,
+      "step": 12663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1917518785880075e-05,
+      "loss": 0.5583,
+      "step": 12664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1916450115327707e-05,
+      "loss": 0.615,
+      "step": 12665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1915381422053398e-05,
+      "loss": 0.6252,
+      "step": 12666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1914312706069819e-05,
+      "loss": 0.6427,
+      "step": 12667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1913243967389643e-05,
+      "loss": 0.6572,
+      "step": 12668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1912175206025537e-05,
+      "loss": 0.5851,
+      "step": 12669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1911106421990178e-05,
+      "loss": 0.6841,
+      "step": 12670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1910037615296233e-05,
+      "loss": 0.6328,
+      "step": 12671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1908968785956376e-05,
+      "loss": 0.6006,
+      "step": 12672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1907899933983279e-05,
+      "loss": 0.6728,
+      "step": 12673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1906831059389614e-05,
+      "loss": 0.6277,
+      "step": 12674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1905762162188058e-05,
+      "loss": 0.6296,
+      "step": 12675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1904693242391277e-05,
+      "loss": 0.6412,
+      "step": 12676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.190362430001195e-05,
+      "loss": 0.6124,
+      "step": 12677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.190255533506275e-05,
+      "loss": 0.5927,
+      "step": 12678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1901486347556346e-05,
+      "loss": 0.663,
+      "step": 12679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.190041733750542e-05,
+      "loss": 0.6271,
+      "step": 12680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.189934830492264e-05,
+      "loss": 0.6393,
+      "step": 12681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1898279249820684e-05,
+      "loss": 0.6577,
+      "step": 12682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1897210172212226e-05,
+      "loss": 0.6092,
+      "step": 12683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.189614107210994e-05,
+      "loss": 0.6671,
+      "step": 12684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1895071949526507e-05,
+      "loss": 0.6303,
+      "step": 12685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1894002804474594e-05,
+      "loss": 0.6412,
+      "step": 12686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1892933636966884e-05,
+      "loss": 0.6027,
+      "step": 12687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1891864447016053e-05,
+      "loss": 0.6395,
+      "step": 12688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1890795234634772e-05,
+      "loss": 0.6234,
+      "step": 12689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1889725999835721e-05,
+      "loss": 0.6093,
+      "step": 12690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1888656742631583e-05,
+      "loss": 0.6951,
+      "step": 12691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1887587463035027e-05,
+      "loss": 0.6803,
+      "step": 12692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1886518161058734e-05,
+      "loss": 0.5904,
+      "step": 12693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.188544883671538e-05,
+      "loss": 0.6457,
+      "step": 12694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1884379490017647e-05,
+      "loss": 0.5845,
+      "step": 12695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1883310120978212e-05,
+      "loss": 0.6141,
+      "step": 12696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1882240729609753e-05,
+      "loss": 0.6106,
+      "step": 12697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1881171315924949e-05,
+      "loss": 0.6241,
+      "step": 12698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1880101879936477e-05,
+      "loss": 0.6347,
+      "step": 12699
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1879032421657021e-05,
+      "loss": 0.6591,
+      "step": 12700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.187796294109926e-05,
+      "loss": 0.5801,
+      "step": 12701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.187689343827587e-05,
+      "loss": 0.6208,
+      "step": 12702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1875823913199536e-05,
+      "loss": 0.653,
+      "step": 12703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1874754365882935e-05,
+      "loss": 0.6249,
+      "step": 12704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1873684796338746e-05,
+      "loss": 0.6585,
+      "step": 12705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1872615204579661e-05,
+      "loss": 0.6751,
+      "step": 12706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.187154559061835e-05,
+      "loss": 0.6403,
+      "step": 12707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1870475954467497e-05,
+      "loss": 0.6297,
+      "step": 12708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.186940629613979e-05,
+      "loss": 0.6718,
+      "step": 12709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1868336615647901e-05,
+      "loss": 0.66,
+      "step": 12710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1867266913004523e-05,
+      "loss": 0.6264,
+      "step": 12711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1866197188222332e-05,
+      "loss": 0.5979,
+      "step": 12712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1865127441314011e-05,
+      "loss": 0.666,
+      "step": 12713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1864057672292247e-05,
+      "loss": 0.5936,
+      "step": 12714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.186298788116972e-05,
+      "loss": 0.6099,
+      "step": 12715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1861918067959113e-05,
+      "loss": 0.6355,
+      "step": 12716
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1860848232673116e-05,
+      "loss": 0.617,
+      "step": 12717
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1859778375324406e-05,
+      "loss": 0.6489,
+      "step": 12718
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1858708495925672e-05,
+      "loss": 0.6423,
+      "step": 12719
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1857638594489598e-05,
+      "loss": 0.6393,
+      "step": 12720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1856568671028867e-05,
+      "loss": 0.5983,
+      "step": 12721
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1855498725556168e-05,
+      "loss": 0.654,
+      "step": 12722
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1854428758084184e-05,
+      "loss": 0.6328,
+      "step": 12723
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1853358768625598e-05,
+      "loss": 0.6081,
+      "step": 12724
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1852288757193103e-05,
+      "loss": 0.6682,
+      "step": 12725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1851218723799381e-05,
+      "loss": 0.6353,
+      "step": 12726
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1850148668457118e-05,
+      "loss": 0.6684,
+      "step": 12727
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1849078591179001e-05,
+      "loss": 0.653,
+      "step": 12728
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1848008491977723e-05,
+      "loss": 0.6375,
+      "step": 12729
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1846938370865962e-05,
+      "loss": 0.5974,
+      "step": 12730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.184586822785641e-05,
+      "loss": 0.647,
+      "step": 12731
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1844798062961758e-05,
+      "loss": 0.6253,
+      "step": 12732
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1843727876194691e-05,
+      "loss": 0.6054,
+      "step": 12733
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1842657667567896e-05,
+      "loss": 0.661,
+      "step": 12734
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1841587437094064e-05,
+      "loss": 0.5955,
+      "step": 12735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1840517184785883e-05,
+      "loss": 0.6415,
+      "step": 12736
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1839446910656043e-05,
+      "loss": 0.6603,
+      "step": 12737
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1838376614717234e-05,
+      "loss": 0.6977,
+      "step": 12738
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1837306296982142e-05,
+      "loss": 0.6252,
+      "step": 12739
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.183623595746346e-05,
+      "loss": 0.6178,
+      "step": 12740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1835165596173881e-05,
+      "loss": 0.6512,
+      "step": 12741
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.183409521312609e-05,
+      "loss": 0.561,
+      "step": 12742
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.183302480833278e-05,
+      "loss": 0.696,
+      "step": 12743
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1831954381806646e-05,
+      "loss": 0.7045,
+      "step": 12744
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1830883933560369e-05,
+      "loss": 0.5989,
+      "step": 12745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.182981346360665e-05,
+      "loss": 0.5945,
+      "step": 12746
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1828742971958178e-05,
+      "loss": 0.6287,
+      "step": 12747
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1827672458627643e-05,
+      "loss": 0.6408,
+      "step": 12748
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.182660192362774e-05,
+      "loss": 0.6803,
+      "step": 12749
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1825531366971163e-05,
+      "loss": 0.5672,
+      "step": 12750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1824460788670599e-05,
+      "loss": 0.6248,
+      "step": 12751
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1823390188738746e-05,
+      "loss": 0.5567,
+      "step": 12752
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1822319567188297e-05,
+      "loss": 0.6272,
+      "step": 12753
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1821248924031941e-05,
+      "loss": 0.6639,
+      "step": 12754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1820178259282376e-05,
+      "loss": 0.6313,
+      "step": 12755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1819107572952297e-05,
+      "loss": 0.6659,
+      "step": 12756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1818036865054394e-05,
+      "loss": 0.5762,
+      "step": 12757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1816966135601366e-05,
+      "loss": 0.6055,
+      "step": 12758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1815895384605907e-05,
+      "loss": 0.6459,
+      "step": 12759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1814824612080709e-05,
+      "loss": 0.5751,
+      "step": 12760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1813753818038469e-05,
+      "loss": 0.6589,
+      "step": 12761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1812683002491884e-05,
+      "loss": 0.6052,
+      "step": 12762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.181161216545365e-05,
+      "loss": 0.6245,
+      "step": 12763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.181054130693646e-05,
+      "loss": 0.582,
+      "step": 12764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1809470426953016e-05,
+      "loss": 0.6584,
+      "step": 12765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1808399525516008e-05,
+      "loss": 0.6141,
+      "step": 12766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1807328602638136e-05,
+      "loss": 0.6582,
+      "step": 12767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1806257658332099e-05,
+      "loss": 0.6801,
+      "step": 12768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1805186692610591e-05,
+      "loss": 0.66,
+      "step": 12769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.180411570548631e-05,
+      "loss": 0.6439,
+      "step": 12770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1803044696971957e-05,
+      "loss": 0.6388,
+      "step": 12771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.180197366708023e-05,
+      "loss": 0.5949,
+      "step": 12772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1800902615823823e-05,
+      "loss": 0.5963,
+      "step": 12773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1799831543215436e-05,
+      "loss": 0.6743,
+      "step": 12774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1798760449267771e-05,
+      "loss": 0.6653,
+      "step": 12775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1797689333993528e-05,
+      "loss": 0.6467,
+      "step": 12776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.17966181974054e-05,
+      "loss": 0.5774,
+      "step": 12777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.179554703951609e-05,
+      "loss": 0.6411,
+      "step": 12778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1794475860338304e-05,
+      "loss": 0.6009,
+      "step": 12779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1793404659884732e-05,
+      "loss": 0.659,
+      "step": 12780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.179233343816808e-05,
+      "loss": 0.6113,
+      "step": 12781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.179126219520105e-05,
+      "loss": 0.6872,
+      "step": 12782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1790190930996337e-05,
+      "loss": 0.6328,
+      "step": 12783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1789119645566648e-05,
+      "loss": 0.567,
+      "step": 12784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1788048338924683e-05,
+      "loss": 0.5578,
+      "step": 12785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1786977011083141e-05,
+      "loss": 0.6145,
+      "step": 12786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1785905662054726e-05,
+      "loss": 0.6104,
+      "step": 12787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1784834291852143e-05,
+      "loss": 0.6008,
+      "step": 12788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1783762900488087e-05,
+      "loss": 0.6401,
+      "step": 12789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1782691487975268e-05,
+      "loss": 0.6346,
+      "step": 12790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1781620054326388e-05,
+      "loss": 0.6142,
+      "step": 12791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1780548599554147e-05,
+      "loss": 0.6132,
+      "step": 12792
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.177947712367125e-05,
+      "loss": 0.575,
+      "step": 12793
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1778405626690402e-05,
+      "loss": 0.6869,
+      "step": 12794
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1777334108624303e-05,
+      "loss": 0.6506,
+      "step": 12795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1776262569485658e-05,
+      "loss": 0.5887,
+      "step": 12796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1775191009287178e-05,
+      "loss": 0.6164,
+      "step": 12797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1774119428041562e-05,
+      "loss": 0.6667,
+      "step": 12798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1773047825761515e-05,
+      "loss": 0.6498,
+      "step": 12799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1771976202459742e-05,
+      "loss": 0.6027,
+      "step": 12800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1770904558148952e-05,
+      "loss": 0.5909,
+      "step": 12801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1769832892841846e-05,
+      "loss": 0.6085,
+      "step": 12802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1768761206551136e-05,
+      "loss": 0.6451,
+      "step": 12803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1767689499289521e-05,
+      "loss": 0.6753,
+      "step": 12804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1766617771069711e-05,
+      "loss": 0.6221,
+      "step": 12805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1765546021904413e-05,
+      "loss": 0.6081,
+      "step": 12806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1764474251806335e-05,
+      "loss": 0.6608,
+      "step": 12807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1763402460788178e-05,
+      "loss": 0.6415,
+      "step": 12808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1762330648862658e-05,
+      "loss": 0.5866,
+      "step": 12809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.176125881604248e-05,
+      "loss": 0.6573,
+      "step": 12810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1760186962340345e-05,
+      "loss": 0.6003,
+      "step": 12811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1759115087768973e-05,
+      "loss": 0.628,
+      "step": 12812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1758043192341066e-05,
+      "loss": 0.6342,
+      "step": 12813
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1756971276069329e-05,
+      "loss": 0.6288,
+      "step": 12814
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1755899338966476e-05,
+      "loss": 0.6403,
+      "step": 12815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.175482738104522e-05,
+      "loss": 0.5832,
+      "step": 12816
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1753755402318259e-05,
+      "loss": 0.622,
+      "step": 12817
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1752683402798312e-05,
+      "loss": 0.6286,
+      "step": 12818
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1751611382498088e-05,
+      "loss": 0.631,
+      "step": 12819
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.175053934143029e-05,
+      "loss": 0.6502,
+      "step": 12820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1749467279607639e-05,
+      "loss": 0.6853,
+      "step": 12821
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1748395197042839e-05,
+      "loss": 0.5904,
+      "step": 12822
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1747323093748597e-05,
+      "loss": 0.6268,
+      "step": 12823
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1746250969737636e-05,
+      "loss": 0.6401,
+      "step": 12824
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1745178825022661e-05,
+      "loss": 0.6781,
+      "step": 12825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1744106659616377e-05,
+      "loss": 0.6067,
+      "step": 12826
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1743034473531506e-05,
+      "loss": 0.596,
+      "step": 12827
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.174196226678076e-05,
+      "loss": 0.6341,
+      "step": 12828
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.174089003937684e-05,
+      "loss": 0.6678,
+      "step": 12829
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1739817791332473e-05,
+      "loss": 0.6333,
+      "step": 12830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1738745522660364e-05,
+      "loss": 0.6204,
+      "step": 12831
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1737673233373224e-05,
+      "loss": 0.653,
+      "step": 12832
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1736600923483771e-05,
+      "loss": 0.6683,
+      "step": 12833
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1735528593004722e-05,
+      "loss": 0.6676,
+      "step": 12834
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.173445624194878e-05,
+      "loss": 0.6093,
+      "step": 12835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1733383870328668e-05,
+      "loss": 0.6074,
+      "step": 12836
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1732311478157096e-05,
+      "loss": 0.5753,
+      "step": 12837
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.173123906544678e-05,
+      "loss": 0.6033,
+      "step": 12838
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1730166632210435e-05,
+      "loss": 0.6675,
+      "step": 12839
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1729094178460778e-05,
+      "loss": 0.6254,
+      "step": 12840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.172802170421052e-05,
+      "loss": 0.5923,
+      "step": 12841
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1726949209472378e-05,
+      "loss": 0.6103,
+      "step": 12842
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.1725876694259073e-05,
+      "loss": 0.6022,
+      "step": 12843
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1724804158583312e-05,
+      "loss": 0.6015,
+      "step": 12844
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1723731602457817e-05,
+      "loss": 0.5921,
+      "step": 12845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1722659025895305e-05,
+      "loss": 0.6109,
+      "step": 12846
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1721586428908489e-05,
+      "loss": 0.6589,
+      "step": 12847
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1720513811510088e-05,
+      "loss": 0.6083,
+      "step": 12848
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.171944117371282e-05,
+      "loss": 0.6356,
+      "step": 12849
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1718368515529402e-05,
+      "loss": 0.6761,
+      "step": 12850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.171729583697255e-05,
+      "loss": 0.5828,
+      "step": 12851
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1716223138054987e-05,
+      "loss": 0.6625,
+      "step": 12852
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1715150418789424e-05,
+      "loss": 0.5551,
+      "step": 12853
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1714077679188585e-05,
+      "loss": 0.6082,
+      "step": 12854
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1713004919265187e-05,
+      "loss": 0.6137,
+      "step": 12855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1711932139031947e-05,
+      "loss": 0.6121,
+      "step": 12856
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1710859338501585e-05,
+      "loss": 0.6852,
+      "step": 12857
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1709786517686825e-05,
+      "loss": 0.6034,
+      "step": 12858
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.170871367660038e-05,
+      "loss": 0.5904,
+      "step": 12859
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1707640815254975e-05,
+      "loss": 0.5917,
+      "step": 12860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1706567933663329e-05,
+      "loss": 0.6627,
+      "step": 12861
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1705495031838159e-05,
+      "loss": 0.6481,
+      "step": 12862
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1704422109792186e-05,
+      "loss": 0.6013,
+      "step": 12863
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1703349167538136e-05,
+      "loss": 0.6952,
+      "step": 12864
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1702276205088726e-05,
+      "loss": 0.588,
+      "step": 12865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1701203222456679e-05,
+      "loss": 0.6356,
+      "step": 12866
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1700130219654715e-05,
+      "loss": 0.6685,
+      "step": 12867
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1699057196695555e-05,
+      "loss": 0.6197,
+      "step": 12868
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1697984153591925e-05,
+      "loss": 0.6586,
+      "step": 12869
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1696911090356545e-05,
+      "loss": 0.5921,
+      "step": 12870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1695838007002133e-05,
+      "loss": 0.666,
+      "step": 12871
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1694764903541422e-05,
+      "loss": 0.6341,
+      "step": 12872
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1693691779987127e-05,
+      "loss": 0.6945,
+      "step": 12873
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1692618636351972e-05,
+      "loss": 0.6396,
+      "step": 12874
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1691545472648684e-05,
+      "loss": 0.5988,
+      "step": 12875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1690472288889983e-05,
+      "loss": 0.6288,
+      "step": 12876
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1689399085088596e-05,
+      "loss": 0.639,
+      "step": 12877
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1688325861257246e-05,
+      "loss": 0.6437,
+      "step": 12878
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1687252617408654e-05,
+      "loss": 0.6187,
+      "step": 12879
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.168617935355555e-05,
+      "loss": 0.6187,
+      "step": 12880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1685106069710656e-05,
+      "loss": 0.6498,
+      "step": 12881
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1684032765886699e-05,
+      "loss": 0.6549,
+      "step": 12882
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1682959442096401e-05,
+      "loss": 0.5568,
+      "step": 12883
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1681886098352495e-05,
+      "loss": 0.6136,
+      "step": 12884
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1680812734667697e-05,
+      "loss": 0.5994,
+      "step": 12885
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1679739351054739e-05,
+      "loss": 0.6493,
+      "step": 12886
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1678665947526347e-05,
+      "loss": 0.6695,
+      "step": 12887
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1677592524095244e-05,
+      "loss": 0.662,
+      "step": 12888
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1676519080774162e-05,
+      "loss": 0.6363,
+      "step": 12889
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1675445617575826e-05,
+      "loss": 0.6173,
+      "step": 12890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1674372134512962e-05,
+      "loss": 0.6291,
+      "step": 12891
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1673298631598296e-05,
+      "loss": 0.6235,
+      "step": 12892
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1672225108844562e-05,
+      "loss": 0.6264,
+      "step": 12893
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.167115156626448e-05,
+      "loss": 0.622,
+      "step": 12894
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1670078003870783e-05,
+      "loss": 0.6081,
+      "step": 12895
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1669004421676202e-05,
+      "loss": 0.6006,
+      "step": 12896
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1667930819693459e-05,
+      "loss": 0.6155,
+      "step": 12897
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1666857197935287e-05,
+      "loss": 0.6629,
+      "step": 12898
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1665783556414416e-05,
+      "loss": 0.5829,
+      "step": 12899
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1664709895143573e-05,
+      "loss": 0.729,
+      "step": 12900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1663636214135485e-05,
+      "loss": 0.6168,
+      "step": 12901
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1662562513402888e-05,
+      "loss": 0.6493,
+      "step": 12902
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.166148879295851e-05,
+      "loss": 0.6308,
+      "step": 12903
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1660415052815078e-05,
+      "loss": 0.6472,
+      "step": 12904
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1659341292985326e-05,
+      "loss": 0.6644,
+      "step": 12905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1658267513481987e-05,
+      "loss": 0.6264,
+      "step": 12906
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1657193714317782e-05,
+      "loss": 0.6778,
+      "step": 12907
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1656119895505455e-05,
+      "loss": 0.6332,
+      "step": 12908
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1655046057057734e-05,
+      "loss": 0.613,
+      "step": 12909
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.165397219898734e-05,
+      "loss": 0.599,
+      "step": 12910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.165289832130702e-05,
+      "loss": 0.6266,
+      "step": 12911
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.16518244240295e-05,
+      "loss": 0.6392,
+      "step": 12912
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1650750507167505e-05,
+      "loss": 0.6259,
+      "step": 12913
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1649676570733779e-05,
+      "loss": 0.5986,
+      "step": 12914
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.164860261474105e-05,
+      "loss": 0.6175,
+      "step": 12915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.164752863920205e-05,
+      "loss": 0.6397,
+      "step": 12916
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1646454644129519e-05,
+      "loss": 0.6328,
+      "step": 12917
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1645380629536183e-05,
+      "loss": 0.6566,
+      "step": 12918
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1644306595434773e-05,
+      "loss": 0.6488,
+      "step": 12919
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1643232541838035e-05,
+      "loss": 0.6387,
+      "step": 12920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1642158468758696e-05,
+      "loss": 0.6428,
+      "step": 12921
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1641084376209484e-05,
+      "loss": 0.6656,
+      "step": 12922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1640010264203147e-05,
+      "loss": 0.604,
+      "step": 12923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1638936132752414e-05,
+      "loss": 0.6107,
+      "step": 12924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1637861981870016e-05,
+      "loss": 0.6045,
+      "step": 12925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1636787811568695e-05,
+      "loss": 0.6154,
+      "step": 12926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1635713621861185e-05,
+      "loss": 0.688,
+      "step": 12927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1634639412760216e-05,
+      "loss": 0.6213,
+      "step": 12928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1633565184278534e-05,
+      "loss": 0.5819,
+      "step": 12929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1632490936428869e-05,
+      "loss": 0.6162,
+      "step": 12930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1631416669223958e-05,
+      "loss": 0.6074,
+      "step": 12931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1630342382676539e-05,
+      "loss": 0.5821,
+      "step": 12932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1629268076799349e-05,
+      "loss": 0.6511,
+      "step": 12933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1628193751605124e-05,
+      "loss": 0.6594,
+      "step": 12934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1627119407106602e-05,
+      "loss": 0.6412,
+      "step": 12935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1626045043316523e-05,
+      "loss": 0.6453,
+      "step": 12936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1624970660247621e-05,
+      "loss": 0.6491,
+      "step": 12937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1623896257912637e-05,
+      "loss": 0.6291,
+      "step": 12938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.162282183632431e-05,
+      "loss": 0.5734,
+      "step": 12939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1621747395495375e-05,
+      "loss": 0.6188,
+      "step": 12940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1620672935438574e-05,
+      "loss": 0.6703,
+      "step": 12941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1619598456166647e-05,
+      "loss": 0.6109,
+      "step": 12942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1618523957692329e-05,
+      "loss": 0.5653,
+      "step": 12943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1617449440028363e-05,
+      "loss": 0.6472,
+      "step": 12944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1616374903187488e-05,
+      "loss": 0.642,
+      "step": 12945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1615300347182443e-05,
+      "loss": 0.6332,
+      "step": 12946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1614225772025969e-05,
+      "loss": 0.6386,
+      "step": 12947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1613151177730808e-05,
+      "loss": 0.6173,
+      "step": 12948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1612076564309698e-05,
+      "loss": 0.6881,
+      "step": 12949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.161100193177538e-05,
+      "loss": 0.6512,
+      "step": 12950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.16099272801406e-05,
+      "loss": 0.6447,
+      "step": 12951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1608852609418092e-05,
+      "loss": 0.6362,
+      "step": 12952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1607777919620603e-05,
+      "loss": 0.6278,
+      "step": 12953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1606703210760873e-05,
+      "loss": 0.6183,
+      "step": 12954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1605628482851641e-05,
+      "loss": 0.5996,
+      "step": 12955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1604553735905652e-05,
+      "loss": 0.6555,
+      "step": 12956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1603478969935654e-05,
+      "loss": 0.6738,
+      "step": 12957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1602404184954379e-05,
+      "loss": 0.6897,
+      "step": 12958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1601329380974577e-05,
+      "loss": 0.6417,
+      "step": 12959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1600254558008992e-05,
+      "loss": 0.6541,
+      "step": 12960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1599179716070364e-05,
+      "loss": 0.5951,
+      "step": 12961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1598104855171437e-05,
+      "loss": 0.6871,
+      "step": 12962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1597029975324954e-05,
+      "loss": 0.6106,
+      "step": 12963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.159595507654366e-05,
+      "loss": 0.6503,
+      "step": 12964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1594880158840302e-05,
+      "loss": 0.5702,
+      "step": 12965
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1593805222227621e-05,
+      "loss": 0.6176,
+      "step": 12966
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1592730266718361e-05,
+      "loss": 0.6347,
+      "step": 12967
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1591655292325273e-05,
+      "loss": 0.647,
+      "step": 12968
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1590580299061093e-05,
+      "loss": 0.6381,
+      "step": 12969
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1589505286938576e-05,
+      "loss": 0.6294,
+      "step": 12970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1588430255970463e-05,
+      "loss": 0.5892,
+      "step": 12971
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1587355206169497e-05,
+      "loss": 0.6004,
+      "step": 12972
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1586280137548426e-05,
+      "loss": 0.6176,
+      "step": 12973
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.158520505012e-05,
+      "loss": 0.6785,
+      "step": 12974
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1584129943896964e-05,
+      "loss": 0.675,
+      "step": 12975
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1583054818892062e-05,
+      "loss": 0.6408,
+      "step": 12976
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1581979675118042e-05,
+      "loss": 0.6678,
+      "step": 12977
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1580904512587652e-05,
+      "loss": 0.6258,
+      "step": 12978
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1579829331313639e-05,
+      "loss": 0.6432,
+      "step": 12979
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1578754131308754e-05,
+      "loss": 0.5698,
+      "step": 12980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1577678912585738e-05,
+      "loss": 0.6507,
+      "step": 12981
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1576603675157343e-05,
+      "loss": 0.6556,
+      "step": 12982
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1575528419036318e-05,
+      "loss": 0.6211,
+      "step": 12983
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1574453144235408e-05,
+      "loss": 0.6693,
+      "step": 12984
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1573377850767367e-05,
+      "loss": 0.6788,
+      "step": 12985
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1572302538644941e-05,
+      "loss": 0.5973,
+      "step": 12986
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.157122720788088e-05,
+      "loss": 0.6511,
+      "step": 12987
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.157015185848793e-05,
+      "loss": 0.6675,
+      "step": 12988
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1569076490478848e-05,
+      "loss": 0.6209,
+      "step": 12989
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1568001103866376e-05,
+      "loss": 0.6125,
+      "step": 12990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1566925698663267e-05,
+      "loss": 0.6608,
+      "step": 12991
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1565850274882275e-05,
+      "loss": 0.7055,
+      "step": 12992
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1564774832536143e-05,
+      "loss": 0.6081,
+      "step": 12993
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1563699371637628e-05,
+      "loss": 0.7101,
+      "step": 12994
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1562623892199481e-05,
+      "loss": 0.6328,
+      "step": 12995
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1561548394234447e-05,
+      "loss": 0.6602,
+      "step": 12996
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1560472877755282e-05,
+      "loss": 0.6755,
+      "step": 12997
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.155939734277474e-05,
+      "loss": 0.6179,
+      "step": 12998
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1558321789305568e-05,
+      "loss": 0.6628,
+      "step": 12999
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1557246217360519e-05,
+      "loss": 0.6616,
+      "step": 13000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1556170626952348e-05,
+      "loss": 0.6493,
+      "step": 13001
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1555095018093805e-05,
+      "loss": 0.6138,
+      "step": 13002
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.155401939079764e-05,
+      "loss": 0.6162,
+      "step": 13003
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1552943745076616e-05,
+      "loss": 0.6046,
+      "step": 13004
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1551868080943476e-05,
+      "loss": 0.5994,
+      "step": 13005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1550792398410975e-05,
+      "loss": 0.6387,
+      "step": 13006
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.154971669749187e-05,
+      "loss": 0.6179,
+      "step": 13007
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1548640978198912e-05,
+      "loss": 0.6032,
+      "step": 13008
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1547565240544858e-05,
+      "loss": 0.6375,
+      "step": 13009
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1546489484542462e-05,
+      "loss": 0.5921,
+      "step": 13010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1545413710204473e-05,
+      "loss": 0.6358,
+      "step": 13011
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1544337917543651e-05,
+      "loss": 0.598,
+      "step": 13012
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.154326210657275e-05,
+      "loss": 0.6083,
+      "step": 13013
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1542186277304524e-05,
+      "loss": 0.5891,
+      "step": 13014
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1541110429751728e-05,
+      "loss": 0.6261,
+      "step": 13015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.154003456392712e-05,
+      "loss": 0.6869,
+      "step": 13016
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1538958679843452e-05,
+      "loss": 0.6418,
+      "step": 13017
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1537882777513483e-05,
+      "loss": 0.6378,
+      "step": 13018
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.153680685694997e-05,
+      "loss": 0.7034,
+      "step": 13019
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1535730918165667e-05,
+      "loss": 0.6104,
+      "step": 13020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1534654961173328e-05,
+      "loss": 0.6699,
+      "step": 13021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1533578985985716e-05,
+      "loss": 0.6209,
+      "step": 13022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1532502992615584e-05,
+      "loss": 0.5928,
+      "step": 13023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1531426981075693e-05,
+      "loss": 0.5855,
+      "step": 13024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1530350951378794e-05,
+      "loss": 0.6426,
+      "step": 13025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.152927490353765e-05,
+      "loss": 0.6921,
+      "step": 13026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1528198837565018e-05,
+      "loss": 0.6438,
+      "step": 13027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1527122753473656e-05,
+      "loss": 0.6427,
+      "step": 13028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1526046651276322e-05,
+      "loss": 0.6499,
+      "step": 13029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1524970530985775e-05,
+      "loss": 0.6142,
+      "step": 13030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1523894392614772e-05,
+      "loss": 0.6444,
+      "step": 13031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1522818236176072e-05,
+      "loss": 0.6271,
+      "step": 13032
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1521742061682437e-05,
+      "loss": 0.6067,
+      "step": 13033
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1520665869146625e-05,
+      "loss": 0.6128,
+      "step": 13034
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1519589658581393e-05,
+      "loss": 0.6637,
+      "step": 13035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.151851342999951e-05,
+      "loss": 0.6564,
+      "step": 13036
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1517437183413723e-05,
+      "loss": 0.6257,
+      "step": 13037
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.15163609188368e-05,
+      "loss": 0.6536,
+      "step": 13038
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1515284636281502e-05,
+      "loss": 0.6317,
+      "step": 13039
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1514208335760583e-05,
+      "loss": 0.6449,
+      "step": 13040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1513132017286812e-05,
+      "loss": 0.6126,
+      "step": 13041
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1512055680872948e-05,
+      "loss": 0.6538,
+      "step": 13042
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.151097932653175e-05,
+      "loss": 0.6403,
+      "step": 13043
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.150990295427598e-05,
+      "loss": 0.6502,
+      "step": 13044
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1508826564118403e-05,
+      "loss": 0.5929,
+      "step": 13045
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1507750156071775e-05,
+      "loss": 0.6156,
+      "step": 13046
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1506673730148862e-05,
+      "loss": 0.6015,
+      "step": 13047
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.150559728636243e-05,
+      "loss": 0.6062,
+      "step": 13048
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1504520824725235e-05,
+      "loss": 0.6341,
+      "step": 13049
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.150344434525004e-05,
+      "loss": 0.625,
+      "step": 13050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1502367847949613e-05,
+      "loss": 0.7137,
+      "step": 13051
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1501291332836714e-05,
+      "loss": 0.5527,
+      "step": 13052
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.150021479992411e-05,
+      "loss": 0.6766,
+      "step": 13053
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1499138249224558e-05,
+      "loss": 0.5826,
+      "step": 13054
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1498061680750828e-05,
+      "loss": 0.637,
+      "step": 13055
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1496985094515678e-05,
+      "loss": 0.6409,
+      "step": 13056
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1495908490531877e-05,
+      "loss": 0.6557,
+      "step": 13057
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1494831868812192e-05,
+      "loss": 0.5936,
+      "step": 13058
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1493755229369384e-05,
+      "loss": 0.6825,
+      "step": 13059
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1492678572216213e-05,
+      "loss": 0.6786,
+      "step": 13060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1491601897365454e-05,
+      "loss": 0.6247,
+      "step": 13061
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1490525204829866e-05,
+      "loss": 0.6072,
+      "step": 13062
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1489448494622215e-05,
+      "loss": 0.6115,
+      "step": 13063
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1488371766755272e-05,
+      "loss": 0.6102,
+      "step": 13064
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1487295021241796e-05,
+      "loss": 0.644,
+      "step": 13065
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1486218258094555e-05,
+      "loss": 0.5936,
+      "step": 13066
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1485141477326319e-05,
+      "loss": 0.6576,
+      "step": 13067
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1484064678949849e-05,
+      "loss": 0.6626,
+      "step": 13068
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1482987862977917e-05,
+      "loss": 0.6341,
+      "step": 13069
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1481911029423288e-05,
+      "loss": 0.6949,
+      "step": 13070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1480834178298728e-05,
+      "loss": 0.5923,
+      "step": 13071
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1479757309617006e-05,
+      "loss": 0.6019,
+      "step": 13072
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1478680423390892e-05,
+      "loss": 0.6622,
+      "step": 13073
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1477603519633148e-05,
+      "loss": 0.6176,
+      "step": 13074
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1476526598356543e-05,
+      "loss": 0.6665,
+      "step": 13075
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1475449659573852e-05,
+      "loss": 0.618,
+      "step": 13076
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1474372703297836e-05,
+      "loss": 0.6221,
+      "step": 13077
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1473295729541265e-05,
+      "loss": 0.6467,
+      "step": 13078
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1472218738316913e-05,
+      "loss": 0.6755,
+      "step": 13079
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1471141729637545e-05,
+      "loss": 0.6494,
+      "step": 13080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1470064703515929e-05,
+      "loss": 0.5903,
+      "step": 13081
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.146898765996484e-05,
+      "loss": 0.5929,
+      "step": 13082
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1467910598997038e-05,
+      "loss": 0.6594,
+      "step": 13083
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1466833520625303e-05,
+      "loss": 0.6237,
+      "step": 13084
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1465756424862401e-05,
+      "loss": 0.5918,
+      "step": 13085
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1464679311721101e-05,
+      "loss": 0.6577,
+      "step": 13086
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1463602181214175e-05,
+      "loss": 0.596,
+      "step": 13087
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1462525033354395e-05,
+      "loss": 0.5822,
+      "step": 13088
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1461447868154529e-05,
+      "loss": 0.6419,
+      "step": 13089
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.146037068562735e-05,
+      "loss": 0.6466,
+      "step": 13090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1459293485785631e-05,
+      "loss": 0.5851,
+      "step": 13091
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1458216268642142e-05,
+      "loss": 0.6455,
+      "step": 13092
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.145713903420965e-05,
+      "loss": 0.655,
+      "step": 13093
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1456061782500936e-05,
+      "loss": 0.6087,
+      "step": 13094
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1454984513528767e-05,
+      "loss": 0.6339,
+      "step": 13095
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1453907227305914e-05,
+      "loss": 0.6711,
+      "step": 13096
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1452829923845155e-05,
+      "loss": 0.6196,
+      "step": 13097
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1451752603159254e-05,
+      "loss": 0.6838,
+      "step": 13098
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1450675265260994e-05,
+      "loss": 0.6299,
+      "step": 13099
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1449597910163141e-05,
+      "loss": 0.6129,
+      "step": 13100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1448520537878473e-05,
+      "loss": 0.6491,
+      "step": 13101
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1447443148419758e-05,
+      "loss": 0.6566,
+      "step": 13102
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1446365741799778e-05,
+      "loss": 0.6578,
+      "step": 13103
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1445288318031303e-05,
+      "loss": 0.5964,
+      "step": 13104
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1444210877127102e-05,
+      "loss": 0.6652,
+      "step": 13105
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1443133419099958e-05,
+      "loss": 0.6903,
+      "step": 13106
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.144205594396264e-05,
+      "loss": 0.6278,
+      "step": 13107
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1440978451727925e-05,
+      "loss": 0.6549,
+      "step": 13108
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1439900942408588e-05,
+      "loss": 0.6469,
+      "step": 13109
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1438823416017402e-05,
+      "loss": 0.6562,
+      "step": 13110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1437745872567145e-05,
+      "loss": 0.6548,
+      "step": 13111
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1436668312070593e-05,
+      "loss": 0.6375,
+      "step": 13112
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.143559073454052e-05,
+      "loss": 0.6003,
+      "step": 13113
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1434513139989702e-05,
+      "loss": 0.6561,
+      "step": 13114
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.143343552843092e-05,
+      "loss": 0.6298,
+      "step": 13115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1432357899876942e-05,
+      "loss": 0.6948,
+      "step": 13116
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1431280254340554e-05,
+      "loss": 0.612,
+      "step": 13117
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1430202591834522e-05,
+      "loss": 0.6371,
+      "step": 13118
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1429124912371632e-05,
+      "loss": 0.6164,
+      "step": 13119
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.142804721596466e-05,
+      "loss": 0.6021,
+      "step": 13120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1426969502626379e-05,
+      "loss": 0.67,
+      "step": 13121
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.142589177236957e-05,
+      "loss": 0.6808,
+      "step": 13122
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1424814025207012e-05,
+      "loss": 0.6564,
+      "step": 13123
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.142373626115148e-05,
+      "loss": 0.6448,
+      "step": 13124
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1422658480215751e-05,
+      "loss": 0.5703,
+      "step": 13125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1421580682412612e-05,
+      "loss": 0.6348,
+      "step": 13126
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.142050286775483e-05,
+      "loss": 0.6857,
+      "step": 13127
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1419425036255192e-05,
+      "loss": 0.6425,
+      "step": 13128
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1418347187926476e-05,
+      "loss": 0.5895,
+      "step": 13129
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1417269322781458e-05,
+      "loss": 0.6195,
+      "step": 13130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1416191440832918e-05,
+      "loss": 0.662,
+      "step": 13131
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1415113542093642e-05,
+      "loss": 0.623,
+      "step": 13132
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1414035626576399e-05,
+      "loss": 0.6554,
+      "step": 13133
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1412957694293975e-05,
+      "loss": 0.6277,
+      "step": 13134
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1411879745259156e-05,
+      "loss": 0.6253,
+      "step": 13135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1410801779484713e-05,
+      "loss": 0.6904,
+      "step": 13136
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.140972379698343e-05,
+      "loss": 0.6533,
+      "step": 13137
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1408645797768087e-05,
+      "loss": 0.5695,
+      "step": 13138
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1407567781851467e-05,
+      "loss": 0.5957,
+      "step": 13139
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1406489749246353e-05,
+      "loss": 0.6271,
+      "step": 13140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1405411699965521e-05,
+      "loss": 0.6013,
+      "step": 13141
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1404333634021756e-05,
+      "loss": 0.5898,
+      "step": 13142
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1403255551427839e-05,
+      "loss": 0.6373,
+      "step": 13143
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1402177452196555e-05,
+      "loss": 0.6224,
+      "step": 13144
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.140109933634068e-05,
+      "loss": 0.6604,
+      "step": 13145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1400021203872999e-05,
+      "loss": 0.5739,
+      "step": 13146
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1398943054806299e-05,
+      "loss": 0.6708,
+      "step": 13147
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1397864889153361e-05,
+      "loss": 0.6634,
+      "step": 13148
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1396786706926961e-05,
+      "loss": 0.6084,
+      "step": 13149
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1395708508139892e-05,
+      "loss": 0.6234,
+      "step": 13150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.139463029280493e-05,
+      "loss": 0.6006,
+      "step": 13151
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1393552060934863e-05,
+      "loss": 0.6156,
+      "step": 13152
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1392473812542475e-05,
+      "loss": 0.5976,
+      "step": 13153
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1391395547640547e-05,
+      "loss": 0.6317,
+      "step": 13154
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.139031726624186e-05,
+      "loss": 0.5952,
+      "step": 13155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.138923896835921e-05,
+      "loss": 0.6457,
+      "step": 13156
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1388160654005376e-05,
+      "loss": 0.5793,
+      "step": 13157
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1387082323193135e-05,
+      "loss": 0.6277,
+      "step": 13158
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1386003975935282e-05,
+      "loss": 0.6624,
+      "step": 13159
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1384925612244598e-05,
+      "loss": 0.65,
+      "step": 13160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1383847232133866e-05,
+      "loss": 0.5923,
+      "step": 13161
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1382768835615878e-05,
+      "loss": 0.6291,
+      "step": 13162
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1381690422703418e-05,
+      "loss": 0.5727,
+      "step": 13163
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1380611993409266e-05,
+      "loss": 0.6193,
+      "step": 13164
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1379533547746214e-05,
+      "loss": 0.6492,
+      "step": 13165
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1378455085727047e-05,
+      "loss": 0.5734,
+      "step": 13166
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1377376607364548e-05,
+      "loss": 0.6696,
+      "step": 13167
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.137629811267151e-05,
+      "loss": 0.5564,
+      "step": 13168
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1375219601660719e-05,
+      "loss": 0.5858,
+      "step": 13169
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1374141074344955e-05,
+      "loss": 0.5866,
+      "step": 13170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1373062530737013e-05,
+      "loss": 0.6675,
+      "step": 13171
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.137198397084968e-05,
+      "loss": 0.6616,
+      "step": 13172
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1370905394695736e-05,
+      "loss": 0.6376,
+      "step": 13173
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1369826802287978e-05,
+      "loss": 0.6633,
+      "step": 13174
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1368748193639192e-05,
+      "loss": 0.6265,
+      "step": 13175
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1367669568762158e-05,
+      "loss": 0.6522,
+      "step": 13176
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1366590927669679e-05,
+      "loss": 0.6426,
+      "step": 13177
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1365512270374532e-05,
+      "loss": 0.6338,
+      "step": 13178
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1364433596889511e-05,
+      "loss": 0.5951,
+      "step": 13179
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1363354907227402e-05,
+      "loss": 0.6111,
+      "step": 13180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1362276201400996e-05,
+      "loss": 0.599,
+      "step": 13181
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1361197479423083e-05,
+      "loss": 0.6808,
+      "step": 13182
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1360118741306451e-05,
+      "loss": 0.6512,
+      "step": 13183
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1359039987063893e-05,
+      "loss": 0.6187,
+      "step": 13184
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1357961216708196e-05,
+      "loss": 0.6537,
+      "step": 13185
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.135688243025215e-05,
+      "loss": 0.6476,
+      "step": 13186
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1355803627708548e-05,
+      "loss": 0.6289,
+      "step": 13187
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1354724809090176e-05,
+      "loss": 0.65,
+      "step": 13188
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1353645974409829e-05,
+      "loss": 0.6298,
+      "step": 13189
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1352567123680299e-05,
+      "loss": 0.6241,
+      "step": 13190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1351488256914371e-05,
+      "loss": 0.6574,
+      "step": 13191
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.135040937412484e-05,
+      "loss": 0.5514,
+      "step": 13192
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1349330475324502e-05,
+      "loss": 0.671,
+      "step": 13193
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.134825156052614e-05,
+      "loss": 0.6575,
+      "step": 13194
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.134717262974255e-05,
+      "loss": 0.6427,
+      "step": 13195
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1346093682986527e-05,
+      "loss": 0.6085,
+      "step": 13196
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1345014720270862e-05,
+      "loss": 0.6315,
+      "step": 13197
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1343935741608342e-05,
+      "loss": 0.6312,
+      "step": 13198
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1342856747011765e-05,
+      "loss": 0.5741,
+      "step": 13199
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1341777736493924e-05,
+      "loss": 0.6088,
+      "step": 13200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1340698710067609e-05,
+      "loss": 0.6708,
+      "step": 13201
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1339619667745616e-05,
+      "loss": 0.6282,
+      "step": 13202
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1338540609540736e-05,
+      "loss": 0.6778,
+      "step": 13203
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1337461535465763e-05,
+      "loss": 0.6098,
+      "step": 13204
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1336382445533495e-05,
+      "loss": 0.6112,
+      "step": 13205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.133530333975672e-05,
+      "loss": 0.6805,
+      "step": 13206
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1334224218148233e-05,
+      "loss": 0.6583,
+      "step": 13207
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1333145080720835e-05,
+      "loss": 0.6162,
+      "step": 13208
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1332065927487312e-05,
+      "loss": 0.5984,
+      "step": 13209
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1330986758460464e-05,
+      "loss": 0.6759,
+      "step": 13210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1329907573653083e-05,
+      "loss": 0.7285,
+      "step": 13211
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1328828373077964e-05,
+      "loss": 0.6233,
+      "step": 13212
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.132774915674791e-05,
+      "loss": 0.6319,
+      "step": 13213
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1326669924675704e-05,
+      "loss": 0.6233,
+      "step": 13214
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.132559067687415e-05,
+      "loss": 0.6435,
+      "step": 13215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.132451141335604e-05,
+      "loss": 0.6544,
+      "step": 13216
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1323432134134174e-05,
+      "loss": 0.5866,
+      "step": 13217
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1322352839221344e-05,
+      "loss": 0.6521,
+      "step": 13218
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.132127352863035e-05,
+      "loss": 0.616,
+      "step": 13219
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1320194202373985e-05,
+      "loss": 0.6287,
+      "step": 13220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1319114860465048e-05,
+      "loss": 0.6454,
+      "step": 13221
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1318035502916337e-05,
+      "loss": 0.6453,
+      "step": 13222
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1316956129740646e-05,
+      "loss": 0.6483,
+      "step": 13223
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1315876740950775e-05,
+      "loss": 0.6902,
+      "step": 13224
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1314797336559521e-05,
+      "loss": 0.6569,
+      "step": 13225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1313717916579683e-05,
+      "loss": 0.6156,
+      "step": 13226
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1312638481024054e-05,
+      "loss": 0.6566,
+      "step": 13227
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1311559029905438e-05,
+      "loss": 0.6681,
+      "step": 13228
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1310479563236629e-05,
+      "loss": 0.6003,
+      "step": 13229
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1309400081030427e-05,
+      "loss": 0.6174,
+      "step": 13230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1308320583299632e-05,
+      "loss": 0.6384,
+      "step": 13231
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.130724107005704e-05,
+      "loss": 0.6581,
+      "step": 13232
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.130616154131545e-05,
+      "loss": 0.6031,
+      "step": 13233
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1305081997087666e-05,
+      "loss": 0.6315,
+      "step": 13234
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1304002437386483e-05,
+      "loss": 0.6541,
+      "step": 13235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1302922862224698e-05,
+      "loss": 0.6425,
+      "step": 13236
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1301843271615121e-05,
+      "loss": 0.6593,
+      "step": 13237
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.130076366557054e-05,
+      "loss": 0.6484,
+      "step": 13238
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.129968404410376e-05,
+      "loss": 0.6111,
+      "step": 13239
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1298604407227588e-05,
+      "loss": 0.5867,
+      "step": 13240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1297524754954812e-05,
+      "loss": 0.6487,
+      "step": 13241
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1296445087298238e-05,
+      "loss": 0.636,
+      "step": 13242
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.129536540427067e-05,
+      "loss": 0.6219,
+      "step": 13243
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1294285705884907e-05,
+      "loss": 0.5666,
+      "step": 13244
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1293205992153746e-05,
+      "loss": 0.6162,
+      "step": 13245
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1292126263089995e-05,
+      "loss": 0.6349,
+      "step": 13246
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1291046518706455e-05,
+      "loss": 0.631,
+      "step": 13247
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.128996675901592e-05,
+      "loss": 0.6319,
+      "step": 13248
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.12888869840312e-05,
+      "loss": 0.6505,
+      "step": 13249
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1287807193765097e-05,
+      "loss": 0.5727,
+      "step": 13250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1286727388230405e-05,
+      "loss": 0.611,
+      "step": 13251
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1285647567439934e-05,
+      "loss": 0.6258,
+      "step": 13252
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1284567731406489e-05,
+      "loss": 0.6552,
+      "step": 13253
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1283487880142863e-05,
+      "loss": 0.6148,
+      "step": 13254
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1282408013661868e-05,
+      "loss": 0.6329,
+      "step": 13255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1281328131976303e-05,
+      "loss": 0.6737,
+      "step": 13256
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1280248235098969e-05,
+      "loss": 0.613,
+      "step": 13257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1279168323042677e-05,
+      "loss": 0.6191,
+      "step": 13258
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1278088395820226e-05,
+      "loss": 0.6693,
+      "step": 13259
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1277008453444415e-05,
+      "loss": 0.6527,
+      "step": 13260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.127592849592806e-05,
+      "loss": 0.6802,
+      "step": 13261
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1274848523283957e-05,
+      "loss": 0.6435,
+      "step": 13262
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1273768535524907e-05,
+      "loss": 0.6619,
+      "step": 13263
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1272688532663725e-05,
+      "loss": 0.6368,
+      "step": 13264
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.127160851471321e-05,
+      "loss": 0.5697,
+      "step": 13265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1270528481686165e-05,
+      "loss": 0.6688,
+      "step": 13266
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.12694484335954e-05,
+      "loss": 0.6685,
+      "step": 13267
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.126836837045372e-05,
+      "loss": 0.5905,
+      "step": 13268
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1267288292273921e-05,
+      "loss": 0.6665,
+      "step": 13269
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1266208199068822e-05,
+      "loss": 0.5939,
+      "step": 13270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1265128090851226e-05,
+      "loss": 0.6612,
+      "step": 13271
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1264047967633928e-05,
+      "loss": 0.656,
+      "step": 13272
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1262967829429748e-05,
+      "loss": 0.6426,
+      "step": 13273
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1261887676251487e-05,
+      "loss": 0.6045,
+      "step": 13274
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1260807508111946e-05,
+      "loss": 0.6366,
+      "step": 13275
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1259727325023942e-05,
+      "loss": 0.6422,
+      "step": 13276
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1258647127000276e-05,
+      "loss": 0.6398,
+      "step": 13277
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1257566914053754e-05,
+      "loss": 0.6694,
+      "step": 13278
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1256486686197186e-05,
+      "loss": 0.646,
+      "step": 13279
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1255406443443382e-05,
+      "loss": 0.6429,
+      "step": 13280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1254326185805142e-05,
+      "loss": 0.6641,
+      "step": 13281
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1253245913295282e-05,
+      "loss": 0.7042,
+      "step": 13282
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1252165625926606e-05,
+      "loss": 0.6653,
+      "step": 13283
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1251085323711918e-05,
+      "loss": 0.6324,
+      "step": 13284
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1250005006664036e-05,
+      "loss": 0.6568,
+      "step": 13285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1248924674795762e-05,
+      "loss": 0.64,
+      "step": 13286
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1247844328119906e-05,
+      "loss": 0.6848,
+      "step": 13287
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1246763966649274e-05,
+      "loss": 0.6165,
+      "step": 13288
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1245683590396684e-05,
+      "loss": 0.662,
+      "step": 13289
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1244603199374934e-05,
+      "loss": 0.6206,
+      "step": 13290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1243522793596839e-05,
+      "loss": 0.6686,
+      "step": 13291
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1242442373075214e-05,
+      "loss": 0.6635,
+      "step": 13292
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1241361937822857e-05,
+      "loss": 0.6173,
+      "step": 13293
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1240281487852584e-05,
+      "loss": 0.5893,
+      "step": 13294
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1239201023177209e-05,
+      "loss": 0.6682,
+      "step": 13295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1238120543809534e-05,
+      "loss": 0.596,
+      "step": 13296
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1237040049762376e-05,
+      "loss": 0.6511,
+      "step": 13297
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1235959541048544e-05,
+      "loss": 0.5979,
+      "step": 13298
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1234879017680846e-05,
+      "loss": 0.6006,
+      "step": 13299
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1233798479672095e-05,
+      "loss": 0.5825,
+      "step": 13300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1232717927035106e-05,
+      "loss": 0.6306,
+      "step": 13301
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1231637359782683e-05,
+      "loss": 0.6777,
+      "step": 13302
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1230556777927644e-05,
+      "loss": 0.6191,
+      "step": 13303
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1229476181482795e-05,
+      "loss": 0.6585,
+      "step": 13304
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.122839557046095e-05,
+      "loss": 0.5725,
+      "step": 13305
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1227314944874926e-05,
+      "loss": 0.61,
+      "step": 13306
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1226234304737527e-05,
+      "loss": 0.6451,
+      "step": 13307
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1225153650061568e-05,
+      "loss": 0.6504,
+      "step": 13308
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1224072980859865e-05,
+      "loss": 0.5982,
+      "step": 13309
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.122299229714523e-05,
+      "loss": 0.6157,
+      "step": 13310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.122191159893047e-05,
+      "loss": 0.6165,
+      "step": 13311
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1220830886228407e-05,
+      "loss": 0.6552,
+      "step": 13312
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1219750159051846e-05,
+      "loss": 0.6172,
+      "step": 13313
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1218669417413603e-05,
+      "loss": 0.6216,
+      "step": 13314
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1217588661326496e-05,
+      "loss": 0.6406,
+      "step": 13315
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1216507890803332e-05,
+      "loss": 0.6238,
+      "step": 13316
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.121542710585693e-05,
+      "loss": 0.6233,
+      "step": 13317
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1214346306500103e-05,
+      "loss": 0.6325,
+      "step": 13318
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1213265492745661e-05,
+      "loss": 0.6657,
+      "step": 13319
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1212184664606425e-05,
+      "loss": 0.5888,
+      "step": 13320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1211103822095206e-05,
+      "loss": 0.6301,
+      "step": 13321
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1210022965224816e-05,
+      "loss": 0.6319,
+      "step": 13322
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1208942094008076e-05,
+      "loss": 0.6164,
+      "step": 13323
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1207861208457798e-05,
+      "loss": 0.6642,
+      "step": 13324
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1206780308586796e-05,
+      "loss": 0.6516,
+      "step": 13325
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1205699394407887e-05,
+      "loss": 0.5835,
+      "step": 13326
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.120461846593389e-05,
+      "loss": 0.6635,
+      "step": 13327
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1203537523177613e-05,
+      "loss": 0.6369,
+      "step": 13328
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1202456566151878e-05,
+      "loss": 0.5877,
+      "step": 13329
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.12013755948695e-05,
+      "loss": 0.6441,
+      "step": 13330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1200294609343292e-05,
+      "loss": 0.5891,
+      "step": 13331
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1199213609586075e-05,
+      "loss": 0.585,
+      "step": 13332
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1198132595610663e-05,
+      "loss": 0.6114,
+      "step": 13333
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1197051567429873e-05,
+      "loss": 0.652,
+      "step": 13334
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1195970525056524e-05,
+      "loss": 0.6484,
+      "step": 13335
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1194889468503432e-05,
+      "loss": 0.6657,
+      "step": 13336
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.119380839778341e-05,
+      "loss": 0.6688,
+      "step": 13337
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1192727312909281e-05,
+      "loss": 0.7251,
+      "step": 13338
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1191646213893863e-05,
+      "loss": 0.6908,
+      "step": 13339
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.119056510074997e-05,
+      "loss": 0.6306,
+      "step": 13340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.118948397349042e-05,
+      "loss": 0.6725,
+      "step": 13341
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1188402832128035e-05,
+      "loss": 0.6115,
+      "step": 13342
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.118732167667563e-05,
+      "loss": 0.6239,
+      "step": 13343
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1186240507146026e-05,
+      "loss": 0.6119,
+      "step": 13344
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1185159323552041e-05,
+      "loss": 0.6771,
+      "step": 13345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.118407812590649e-05,
+      "loss": 0.6511,
+      "step": 13346
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1182996914222198e-05,
+      "loss": 0.6454,
+      "step": 13347
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.118191568851198e-05,
+      "loss": 0.5955,
+      "step": 13348
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1180834448788655e-05,
+      "loss": 0.6283,
+      "step": 13349
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1179753195065044e-05,
+      "loss": 0.5808,
+      "step": 13350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1178671927353967e-05,
+      "loss": 0.6028,
+      "step": 13351
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1177590645668244e-05,
+      "loss": 0.6145,
+      "step": 13352
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1176509350020692e-05,
+      "loss": 0.67,
+      "step": 13353
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1175428040424137e-05,
+      "loss": 0.6748,
+      "step": 13354
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1174346716891393e-05,
+      "loss": 0.6493,
+      "step": 13355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1173265379435283e-05,
+      "loss": 0.5574,
+      "step": 13356
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1172184028068631e-05,
+      "loss": 0.6672,
+      "step": 13357
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1171102662804251e-05,
+      "loss": 0.665,
+      "step": 13358
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1170021283654968e-05,
+      "loss": 0.6815,
+      "step": 13359
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1168939890633603e-05,
+      "loss": 0.6366,
+      "step": 13360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1167858483752978e-05,
+      "loss": 0.5672,
+      "step": 13361
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.116677706302591e-05,
+      "loss": 0.6351,
+      "step": 13362
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.116569562846523e-05,
+      "loss": 0.6258,
+      "step": 13363
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1164614180083748e-05,
+      "loss": 0.6354,
+      "step": 13364
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1163532717894297e-05,
+      "loss": 0.699,
+      "step": 13365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1162451241909689e-05,
+      "loss": 0.7008,
+      "step": 13366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.116136975214275e-05,
+      "loss": 0.6794,
+      "step": 13367
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1160288248606307e-05,
+      "loss": 0.6721,
+      "step": 13368
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.115920673131318e-05,
+      "loss": 0.6116,
+      "step": 13369
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1158125200276186e-05,
+      "loss": 0.6026,
+      "step": 13370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1157043655508157e-05,
+      "loss": 0.6497,
+      "step": 13371
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.115596209702191e-05,
+      "loss": 0.6383,
+      "step": 13372
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.115488052483027e-05,
+      "loss": 0.6307,
+      "step": 13373
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.115379893894606e-05,
+      "loss": 0.6296,
+      "step": 13374
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1152717339382106e-05,
+      "loss": 0.6631,
+      "step": 13375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1151635726151228e-05,
+      "loss": 0.5982,
+      "step": 13376
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1150554099266253e-05,
+      "loss": 0.5995,
+      "step": 13377
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1149472458740002e-05,
+      "loss": 0.6356,
+      "step": 13378
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1148390804585301e-05,
+      "loss": 0.6093,
+      "step": 13379
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1147309136814976e-05,
+      "loss": 0.6579,
+      "step": 13380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1146227455441849e-05,
+      "loss": 0.6217,
+      "step": 13381
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1145145760478744e-05,
+      "loss": 0.6709,
+      "step": 13382
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.114406405193849e-05,
+      "loss": 0.5822,
+      "step": 13383
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1142982329833909e-05,
+      "loss": 0.6246,
+      "step": 13384
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1141900594177823e-05,
+      "loss": 0.6284,
+      "step": 13385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1140818844983066e-05,
+      "loss": 0.6262,
+      "step": 13386
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1139737082262456e-05,
+      "loss": 0.6381,
+      "step": 13387
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.113865530602882e-05,
+      "loss": 0.5982,
+      "step": 13388
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1137573516294988e-05,
+      "loss": 0.6574,
+      "step": 13389
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.113649171307378e-05,
+      "loss": 0.6212,
+      "step": 13390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1135409896378027e-05,
+      "loss": 0.6171,
+      "step": 13391
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1134328066220554e-05,
+      "loss": 0.5767,
+      "step": 13392
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1133246222614185e-05,
+      "loss": 0.6559,
+      "step": 13393
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1132164365571747e-05,
+      "loss": 0.6503,
+      "step": 13394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1131082495106072e-05,
+      "loss": 0.6388,
+      "step": 13395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1130000611229983e-05,
+      "loss": 0.6009,
+      "step": 13396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1128918713956303e-05,
+      "loss": 0.6135,
+      "step": 13397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1127836803297867e-05,
+      "loss": 0.6378,
+      "step": 13398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.11267548792675e-05,
+      "loss": 0.6526,
+      "step": 13399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1125672941878023e-05,
+      "loss": 0.6657,
+      "step": 13400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1124590991142274e-05,
+      "loss": 0.628,
+      "step": 13401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1123509027073078e-05,
+      "loss": 0.6265,
+      "step": 13402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1122427049683256e-05,
+      "loss": 0.6982,
+      "step": 13403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1121345058985644e-05,
+      "loss": 0.654,
+      "step": 13404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.112026305499307e-05,
+      "loss": 0.5654,
+      "step": 13405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1119181037718353e-05,
+      "loss": 0.666,
+      "step": 13406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1118099007174337e-05,
+      "loss": 0.6496,
+      "step": 13407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1117016963373842e-05,
+      "loss": 0.578,
+      "step": 13408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1115934906329693e-05,
+      "loss": 0.6948,
+      "step": 13409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1114852836054727e-05,
+      "loss": 0.6593,
+      "step": 13410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1113770752561772e-05,
+      "loss": 0.578,
+      "step": 13411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1112688655863652e-05,
+      "loss": 0.6307,
+      "step": 13412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1111606545973204e-05,
+      "loss": 0.6356,
+      "step": 13413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1110524422903254e-05,
+      "loss": 0.655,
+      "step": 13414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.110944228666663e-05,
+      "loss": 0.605,
+      "step": 13415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1108360137276164e-05,
+      "loss": 0.6202,
+      "step": 13416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.110727797474469e-05,
+      "loss": 0.6215,
+      "step": 13417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1106195799085031e-05,
+      "loss": 0.6105,
+      "step": 13418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1105113610310024e-05,
+      "loss": 0.614,
+      "step": 13419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1104031408432496e-05,
+      "loss": 0.5971,
+      "step": 13420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1102949193465277e-05,
+      "loss": 0.6273,
+      "step": 13421
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1101866965421202e-05,
+      "loss": 0.6172,
+      "step": 13422
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1100784724313103e-05,
+      "loss": 0.5756,
+      "step": 13423
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1099702470153803e-05,
+      "loss": 0.6384,
+      "step": 13424
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1098620202956144e-05,
+      "loss": 0.6153,
+      "step": 13425
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.109753792273295e-05,
+      "loss": 0.6046,
+      "step": 13426
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1096455629497056e-05,
+      "loss": 0.6075,
+      "step": 13427
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1095373323261293e-05,
+      "loss": 0.6233,
+      "step": 13428
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1094291004038497e-05,
+      "loss": 0.6435,
+      "step": 13429
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1093208671841493e-05,
+      "loss": 0.6423,
+      "step": 13430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.109212632668312e-05,
+      "loss": 0.6118,
+      "step": 13431
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1091043968576206e-05,
+      "loss": 0.6157,
+      "step": 13432
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1089961597533584e-05,
+      "loss": 0.5838,
+      "step": 13433
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1088879213568091e-05,
+      "loss": 0.665,
+      "step": 13434
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1087796816692557e-05,
+      "loss": 0.627,
+      "step": 13435
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1086714406919817e-05,
+      "loss": 0.6029,
+      "step": 13436
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1085631984262699e-05,
+      "loss": 0.6158,
+      "step": 13437
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1084549548734045e-05,
+      "loss": 0.6376,
+      "step": 13438
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1083467100346683e-05,
+      "loss": 0.6452,
+      "step": 13439
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1082384639113446e-05,
+      "loss": 0.6334,
+      "step": 13440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1081302165047172e-05,
+      "loss": 0.6732,
+      "step": 13441
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1080219678160691e-05,
+      "loss": 0.6803,
+      "step": 13442
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1079137178466838e-05,
+      "loss": 0.6201,
+      "step": 13443
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1078054665978454e-05,
+      "loss": 0.5896,
+      "step": 13444
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1076972140708364e-05,
+      "loss": 0.5944,
+      "step": 13445
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1075889602669407e-05,
+      "loss": 0.6492,
+      "step": 13446
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.107480705187442e-05,
+      "loss": 0.621,
+      "step": 13447
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1073724488336235e-05,
+      "loss": 0.6463,
+      "step": 13448
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1072641912067688e-05,
+      "loss": 0.6578,
+      "step": 13449
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1071559323081612e-05,
+      "loss": 0.621,
+      "step": 13450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1070476721390849e-05,
+      "loss": 0.658,
+      "step": 13451
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1069394107008225e-05,
+      "loss": 0.6043,
+      "step": 13452
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1068311479946585e-05,
+      "loss": 0.6731,
+      "step": 13453
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.106722884021876e-05,
+      "loss": 0.6485,
+      "step": 13454
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1066146187837586e-05,
+      "loss": 0.6677,
+      "step": 13455
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1065063522815902e-05,
+      "loss": 0.6203,
+      "step": 13456
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1063980845166542e-05,
+      "loss": 0.702,
+      "step": 13457
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1062898154902344e-05,
+      "loss": 0.6448,
+      "step": 13458
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1061815452036141e-05,
+      "loss": 0.6103,
+      "step": 13459
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1060732736580772e-05,
+      "loss": 0.6517,
+      "step": 13460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1059650008549078e-05,
+      "loss": 0.6433,
+      "step": 13461
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1058567267953893e-05,
+      "loss": 0.6374,
+      "step": 13462
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.105748451480805e-05,
+      "loss": 0.6511,
+      "step": 13463
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1056401749124394e-05,
+      "loss": 0.6729,
+      "step": 13464
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1055318970915758e-05,
+      "loss": 0.6558,
+      "step": 13465
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1054236180194979e-05,
+      "loss": 0.6079,
+      "step": 13466
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.10531533769749e-05,
+      "loss": 0.6028,
+      "step": 13467
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1052070561268353e-05,
+      "loss": 0.6026,
+      "step": 13468
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1050987733088177e-05,
+      "loss": 0.6434,
+      "step": 13469
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1049904892447217e-05,
+      "loss": 0.6521,
+      "step": 13470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1048822039358301e-05,
+      "loss": 0.6863,
+      "step": 13471
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1047739173834276e-05,
+      "loss": 0.6523,
+      "step": 13472
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1046656295887977e-05,
+      "loss": 0.6159,
+      "step": 13473
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1045573405532242e-05,
+      "loss": 0.6457,
+      "step": 13474
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1044490502779913e-05,
+      "loss": 0.6399,
+      "step": 13475
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1043407587643828e-05,
+      "loss": 0.6124,
+      "step": 13476
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1042324660136825e-05,
+      "loss": 0.6302,
+      "step": 13477
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1041241720271745e-05,
+      "loss": 0.6324,
+      "step": 13478
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1040158768061431e-05,
+      "loss": 0.687,
+      "step": 13479
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1039075803518716e-05,
+      "loss": 0.6366,
+      "step": 13480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1037992826656441e-05,
+      "loss": 0.6483,
+      "step": 13481
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.103690983748745e-05,
+      "loss": 0.6692,
+      "step": 13482
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.103582683602458e-05,
+      "loss": 0.6248,
+      "step": 13483
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1034743822280673e-05,
+      "loss": 0.6048,
+      "step": 13484
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.103366079626857e-05,
+      "loss": 0.6387,
+      "step": 13485
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1032577758001111e-05,
+      "loss": 0.6153,
+      "step": 13486
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1031494707491133e-05,
+      "loss": 0.6356,
+      "step": 13487
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1030411644751485e-05,
+      "loss": 0.6511,
+      "step": 13488
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1029328569795004e-05,
+      "loss": 0.6372,
+      "step": 13489
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1028245482634524e-05,
+      "loss": 0.6225,
+      "step": 13490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.10271623832829e-05,
+      "loss": 0.6485,
+      "step": 13491
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1026079271752962e-05,
+      "loss": 0.613,
+      "step": 13492
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1024996148057557e-05,
+      "loss": 0.5794,
+      "step": 13493
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1023913012209528e-05,
+      "loss": 0.6058,
+      "step": 13494
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1022829864221716e-05,
+      "loss": 0.6873,
+      "step": 13495
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1021746704106956e-05,
+      "loss": 0.6295,
+      "step": 13496
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1020663531878101e-05,
+      "loss": 0.5953,
+      "step": 13497
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1019580347547992e-05,
+      "loss": 0.6116,
+      "step": 13498
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.101849715112946e-05,
+      "loss": 0.6132,
+      "step": 13499
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1017413942635362e-05,
+      "loss": 0.6396,
+      "step": 13500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1016330722078534e-05,
+      "loss": 0.6318,
+      "step": 13501
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1015247489471817e-05,
+      "loss": 0.5908,
+      "step": 13502
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1014164244828058e-05,
+      "loss": 0.6126,
+      "step": 13503
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1013080988160102e-05,
+      "loss": 0.6527,
+      "step": 13504
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1011997719480782e-05,
+      "loss": 0.6413,
+      "step": 13505
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1010914438802955e-05,
+      "loss": 0.5822,
+      "step": 13506
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1009831146139457e-05,
+      "loss": 0.5845,
+      "step": 13507
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1008747841503132e-05,
+      "loss": 0.5952,
+      "step": 13508
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1007664524906827e-05,
+      "loss": 0.6347,
+      "step": 13509
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1006581196363385e-05,
+      "loss": 0.6655,
+      "step": 13510
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1005497855885644e-05,
+      "loss": 0.6096,
+      "step": 13511
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1004414503486461e-05,
+      "loss": 0.6523,
+      "step": 13512
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1003331139178672e-05,
+      "loss": 0.6417,
+      "step": 13513
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.100224776297512e-05,
+      "loss": 0.6017,
+      "step": 13514
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1001164374888655e-05,
+      "loss": 0.6331,
+      "step": 13515
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.100008097493212e-05,
+      "loss": 0.6117,
+      "step": 13516
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0998997563118358e-05,
+      "loss": 0.6577,
+      "step": 13517
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.099791413946022e-05,
+      "loss": 0.6294,
+      "step": 13518
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0996830703970546e-05,
+      "loss": 0.6329,
+      "step": 13519
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0995747256662181e-05,
+      "loss": 0.6664,
+      "step": 13520
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0994663797547973e-05,
+      "loss": 0.645,
+      "step": 13521
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0993580326640771e-05,
+      "loss": 0.6457,
+      "step": 13522
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0992496843953413e-05,
+      "loss": 0.6066,
+      "step": 13523
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.099141334949875e-05,
+      "loss": 0.5728,
+      "step": 13524
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.099032984328963e-05,
+      "loss": 0.6577,
+      "step": 13525
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0989246325338896e-05,
+      "loss": 0.6301,
+      "step": 13526
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0988162795659395e-05,
+      "loss": 0.6092,
+      "step": 13527
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0987079254263974e-05,
+      "loss": 0.7135,
+      "step": 13528
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0985995701165479e-05,
+      "loss": 0.648,
+      "step": 13529
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0984912136376758e-05,
+      "loss": 0.6105,
+      "step": 13530
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0983828559910658e-05,
+      "loss": 0.6361,
+      "step": 13531
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0982744971780026e-05,
+      "loss": 0.5803,
+      "step": 13532
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0981661371997709e-05,
+      "loss": 0.6186,
+      "step": 13533
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0980577760576557e-05,
+      "loss": 0.6317,
+      "step": 13534
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0979494137529412e-05,
+      "loss": 0.6621,
+      "step": 13535
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0978410502869126e-05,
+      "loss": 0.6183,
+      "step": 13536
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0977326856608545e-05,
+      "loss": 0.5791,
+      "step": 13537
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0976243198760519e-05,
+      "loss": 0.5823,
+      "step": 13538
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0975159529337893e-05,
+      "loss": 0.6013,
+      "step": 13539
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.097407584835352e-05,
+      "loss": 0.6534,
+      "step": 13540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0972992155820245e-05,
+      "loss": 0.6136,
+      "step": 13541
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0971908451750915e-05,
+      "loss": 0.6328,
+      "step": 13542
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0970824736158385e-05,
+      "loss": 0.6365,
+      "step": 13543
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0969741009055495e-05,
+      "loss": 0.6466,
+      "step": 13544
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.09686572704551e-05,
+      "loss": 0.5768,
+      "step": 13545
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0967573520370049e-05,
+      "loss": 0.6433,
+      "step": 13546
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0966489758813191e-05,
+      "loss": 0.6272,
+      "step": 13547
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.096540598579737e-05,
+      "loss": 0.6393,
+      "step": 13548
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0964322201335445e-05,
+      "loss": 0.6556,
+      "step": 13549
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0963238405440254e-05,
+      "loss": 0.6388,
+      "step": 13550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.096215459812466e-05,
+      "loss": 0.6536,
+      "step": 13551
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0961070779401502e-05,
+      "loss": 0.6474,
+      "step": 13552
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0959986949283634e-05,
+      "loss": 0.6377,
+      "step": 13553
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0958903107783908e-05,
+      "loss": 0.6266,
+      "step": 13554
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0957819254915174e-05,
+      "loss": 0.6152,
+      "step": 13555
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0956735390690278e-05,
+      "loss": 0.6471,
+      "step": 13556
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0955651515122077e-05,
+      "loss": 0.6229,
+      "step": 13557
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0954567628223413e-05,
+      "loss": 0.6415,
+      "step": 13558
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0953483730007146e-05,
+      "loss": 0.6609,
+      "step": 13559
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0952399820486124e-05,
+      "loss": 0.6318,
+      "step": 13560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0951315899673196e-05,
+      "loss": 0.6379,
+      "step": 13561
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0950231967581213e-05,
+      "loss": 0.6786,
+      "step": 13562
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0949148024223031e-05,
+      "loss": 0.6356,
+      "step": 13563
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0948064069611496e-05,
+      "loss": 0.6631,
+      "step": 13564
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0946980103759463e-05,
+      "loss": 0.5814,
+      "step": 13565
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0945896126679785e-05,
+      "loss": 0.6885,
+      "step": 13566
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.094481213838531e-05,
+      "loss": 0.6504,
+      "step": 13567
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.094372813888889e-05,
+      "loss": 0.6231,
+      "step": 13568
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0942644128203383e-05,
+      "loss": 0.6024,
+      "step": 13569
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0941560106341635e-05,
+      "loss": 0.6192,
+      "step": 13570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.09404760733165e-05,
+      "loss": 0.6582,
+      "step": 13571
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0939392029140834e-05,
+      "loss": 0.6151,
+      "step": 13572
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0938307973827485e-05,
+      "loss": 0.603,
+      "step": 13573
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0937223907389309e-05,
+      "loss": 0.6809,
+      "step": 13574
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0936139829839158e-05,
+      "loss": 0.6087,
+      "step": 13575
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0935055741189884e-05,
+      "loss": 0.6146,
+      "step": 13576
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0933971641454342e-05,
+      "loss": 0.6162,
+      "step": 13577
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0932887530645387e-05,
+      "loss": 0.6149,
+      "step": 13578
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0931803408775867e-05,
+      "loss": 0.6643,
+      "step": 13579
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0930719275858638e-05,
+      "loss": 0.6261,
+      "step": 13580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0929635131906559e-05,
+      "loss": 0.6894,
+      "step": 13581
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0928550976932477e-05,
+      "loss": 0.6291,
+      "step": 13582
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0927466810949247e-05,
+      "loss": 0.6386,
+      "step": 13583
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.092638263396973e-05,
+      "loss": 0.6213,
+      "step": 13584
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0925298446006769e-05,
+      "loss": 0.6935,
+      "step": 13585
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0924214247073229e-05,
+      "loss": 0.607,
+      "step": 13586
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0923130037181957e-05,
+      "loss": 0.6277,
+      "step": 13587
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0922045816345812e-05,
+      "loss": 0.6176,
+      "step": 13588
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0920961584577645e-05,
+      "loss": 0.649,
+      "step": 13589
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0919877341890316e-05,
+      "loss": 0.5653,
+      "step": 13590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0918793088296678e-05,
+      "loss": 0.6616,
+      "step": 13591
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0917708823809583e-05,
+      "loss": 0.6312,
+      "step": 13592
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0916624548441892e-05,
+      "loss": 0.639,
+      "step": 13593
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0915540262206454e-05,
+      "loss": 0.5957,
+      "step": 13594
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.091445596511613e-05,
+      "loss": 0.6272,
+      "step": 13595
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0913371657183775e-05,
+      "loss": 0.6703,
+      "step": 13596
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0912287338422243e-05,
+      "loss": 0.6591,
+      "step": 13597
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0911203008844389e-05,
+      "loss": 0.6365,
+      "step": 13598
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0910118668463072e-05,
+      "loss": 0.6201,
+      "step": 13599
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0909034317291145e-05,
+      "loss": 0.6201,
+      "step": 13600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0907949955341468e-05,
+      "loss": 0.6035,
+      "step": 13601
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0906865582626895e-05,
+      "loss": 0.6452,
+      "step": 13602
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0905781199160283e-05,
+      "loss": 0.6198,
+      "step": 13603
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0904696804954488e-05,
+      "loss": 0.6601,
+      "step": 13604
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0903612400022371e-05,
+      "loss": 0.6323,
+      "step": 13605
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0902527984376785e-05,
+      "loss": 0.6398,
+      "step": 13606
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0901443558030587e-05,
+      "loss": 0.6019,
+      "step": 13607
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0900359120996638e-05,
+      "loss": 0.5718,
+      "step": 13608
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0899274673287792e-05,
+      "loss": 0.6008,
+      "step": 13609
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0898190214916902e-05,
+      "loss": 0.5941,
+      "step": 13610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0897105745896837e-05,
+      "loss": 0.6257,
+      "step": 13611
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0896021266240446e-05,
+      "loss": 0.687,
+      "step": 13612
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.089493677596059e-05,
+      "loss": 0.6308,
+      "step": 13613
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0893852275070124e-05,
+      "loss": 0.5766,
+      "step": 13614
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0892767763581913e-05,
+      "loss": 0.6333,
+      "step": 13615
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0891683241508807e-05,
+      "loss": 0.6281,
+      "step": 13616
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0890598708863669e-05,
+      "loss": 0.6674,
+      "step": 13617
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0889514165659358e-05,
+      "loss": 0.6336,
+      "step": 13618
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0888429611908729e-05,
+      "loss": 0.6164,
+      "step": 13619
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0887345047624644e-05,
+      "loss": 0.6734,
+      "step": 13620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0886260472819963e-05,
+      "loss": 0.6222,
+      "step": 13621
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.088517588750754e-05,
+      "loss": 0.6875,
+      "step": 13622
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0884091291700239e-05,
+      "loss": 0.6059,
+      "step": 13623
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0883006685410918e-05,
+      "loss": 0.6396,
+      "step": 13624
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0881922068652432e-05,
+      "loss": 0.6449,
+      "step": 13625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0880837441437648e-05,
+      "loss": 0.5635,
+      "step": 13626
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0879752803779422e-05,
+      "loss": 0.609,
+      "step": 13627
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0878668155690611e-05,
+      "loss": 0.6364,
+      "step": 13628
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0877583497184078e-05,
+      "loss": 0.6692,
+      "step": 13629
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0876498828272687e-05,
+      "loss": 0.5993,
+      "step": 13630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0875414148969288e-05,
+      "loss": 0.6342,
+      "step": 13631
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0874329459286747e-05,
+      "loss": 0.6792,
+      "step": 13632
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0873244759237928e-05,
+      "loss": 0.6223,
+      "step": 13633
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0872160048835684e-05,
+      "loss": 0.5986,
+      "step": 13634
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.087107532809288e-05,
+      "loss": 0.6698,
+      "step": 13635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.086999059702238e-05,
+      "loss": 0.6659,
+      "step": 13636
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0868905855637037e-05,
+      "loss": 0.6679,
+      "step": 13637
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0867821103949717e-05,
+      "loss": 0.6639,
+      "step": 13638
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0866736341973282e-05,
+      "loss": 0.624,
+      "step": 13639
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0865651569720587e-05,
+      "loss": 0.5748,
+      "step": 13640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0864566787204498e-05,
+      "loss": 0.5586,
+      "step": 13641
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0863481994437881e-05,
+      "loss": 0.6231,
+      "step": 13642
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.086239719143359e-05,
+      "loss": 0.608,
+      "step": 13643
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.086131237820449e-05,
+      "loss": 0.6599,
+      "step": 13644
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0860227554763438e-05,
+      "loss": 0.6618,
+      "step": 13645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0859142721123303e-05,
+      "loss": 0.6475,
+      "step": 13646
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0858057877296946e-05,
+      "loss": 0.6225,
+      "step": 13647
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0856973023297225e-05,
+      "loss": 0.5563,
+      "step": 13648
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0855888159137006e-05,
+      "loss": 0.5632,
+      "step": 13649
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.085480328482915e-05,
+      "loss": 0.6418,
+      "step": 13650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.085371840038652e-05,
+      "loss": 0.6205,
+      "step": 13651
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0852633505821976e-05,
+      "loss": 0.5486,
+      "step": 13652
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0851548601148387e-05,
+      "loss": 0.6044,
+      "step": 13653
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.085046368637861e-05,
+      "loss": 0.5538,
+      "step": 13654
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0849378761525508e-05,
+      "loss": 0.613,
+      "step": 13655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.084829382660195e-05,
+      "loss": 0.6527,
+      "step": 13656
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0847208881620794e-05,
+      "loss": 0.653,
+      "step": 13657
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0846123926594905e-05,
+      "loss": 0.6282,
+      "step": 13658
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0845038961537147e-05,
+      "loss": 0.6093,
+      "step": 13659
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0843953986460382e-05,
+      "loss": 0.6504,
+      "step": 13660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0842869001377477e-05,
+      "loss": 0.6177,
+      "step": 13661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0841784006301293e-05,
+      "loss": 0.6766,
+      "step": 13662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0840699001244693e-05,
+      "loss": 0.5841,
+      "step": 13663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0839613986220543e-05,
+      "loss": 0.7012,
+      "step": 13664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.083852896124171e-05,
+      "loss": 0.6078,
+      "step": 13665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0837443926321054e-05,
+      "loss": 0.676,
+      "step": 13666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0836358881471439e-05,
+      "loss": 0.6235,
+      "step": 13667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0835273826705734e-05,
+      "loss": 0.5992,
+      "step": 13668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0834188762036799e-05,
+      "loss": 0.6664,
+      "step": 13669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0833103687477501e-05,
+      "loss": 0.6269,
+      "step": 13670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0832018603040709e-05,
+      "loss": 0.6622,
+      "step": 13671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.083093350873928e-05,
+      "loss": 0.6403,
+      "step": 13672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0829848404586082e-05,
+      "loss": 0.6557,
+      "step": 13673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0828763290593984e-05,
+      "loss": 0.6163,
+      "step": 13674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0827678166775849e-05,
+      "loss": 0.6183,
+      "step": 13675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0826593033144541e-05,
+      "loss": 0.626,
+      "step": 13676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0825507889712929e-05,
+      "loss": 0.5759,
+      "step": 13677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0824422736493873e-05,
+      "loss": 0.6598,
+      "step": 13678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0823337573500245e-05,
+      "loss": 0.611,
+      "step": 13679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0822252400744911e-05,
+      "loss": 0.5606,
+      "step": 13680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.082116721824073e-05,
+      "loss": 0.6334,
+      "step": 13681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0820082026000573e-05,
+      "loss": 0.5907,
+      "step": 13682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.081899682403731e-05,
+      "loss": 0.6175,
+      "step": 13683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0817911612363801e-05,
+      "loss": 0.6544,
+      "step": 13684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0816826390992915e-05,
+      "loss": 0.6716,
+      "step": 13685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0815741159937521e-05,
+      "loss": 0.6086,
+      "step": 13686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0814655919210482e-05,
+      "loss": 0.6719,
+      "step": 13687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0813570668824665e-05,
+      "loss": 0.6006,
+      "step": 13688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0812485408792943e-05,
+      "loss": 0.6162,
+      "step": 13689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0811400139128175e-05,
+      "loss": 0.61,
+      "step": 13690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0810314859843232e-05,
+      "loss": 0.6058,
+      "step": 13691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0809229570950983e-05,
+      "loss": 0.6704,
+      "step": 13692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0808144272464291e-05,
+      "loss": 0.6011,
+      "step": 13693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0807058964396029e-05,
+      "loss": 0.6394,
+      "step": 13694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0805973646759063e-05,
+      "loss": 0.6172,
+      "step": 13695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0804888319566259e-05,
+      "loss": 0.6503,
+      "step": 13696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0803802982830483e-05,
+      "loss": 0.6216,
+      "step": 13697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0802717636564608e-05,
+      "loss": 0.6013,
+      "step": 13698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0801632280781499e-05,
+      "loss": 0.6931,
+      "step": 13699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0800546915494026e-05,
+      "loss": 0.6324,
+      "step": 13700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0799461540715058e-05,
+      "loss": 0.6826,
+      "step": 13701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0798376156457459e-05,
+      "loss": 0.5789,
+      "step": 13702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0797290762734102e-05,
+      "loss": 0.6286,
+      "step": 13703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0796205359557856e-05,
+      "loss": 0.6081,
+      "step": 13704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0795119946941586e-05,
+      "loss": 0.6074,
+      "step": 13705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0794034524898167e-05,
+      "loss": 0.6286,
+      "step": 13706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0792949093440459e-05,
+      "loss": 0.6333,
+      "step": 13707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.079186365258134e-05,
+      "loss": 0.6463,
+      "step": 13708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.0790778202333676e-05,
+      "loss": 0.6011,
+      "step": 13709
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0789692742710334e-05,
+      "loss": 0.6302,
+      "step": 13710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0788607273724185e-05,
+      "loss": 0.6638,
+      "step": 13711
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0787521795388102e-05,
+      "loss": 0.6042,
+      "step": 13712
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.078643630771495e-05,
+      "loss": 0.6976,
+      "step": 13713
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.07853508107176e-05,
+      "loss": 0.6206,
+      "step": 13714
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0784265304408925e-05,
+      "loss": 0.5983,
+      "step": 13715
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.078317978880179e-05,
+      "loss": 0.6225,
+      "step": 13716
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.078209426390907e-05,
+      "loss": 0.639,
+      "step": 13717
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0781008729743634e-05,
+      "loss": 0.6492,
+      "step": 13718
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0779923186318349e-05,
+      "loss": 0.6418,
+      "step": 13719
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0778837633646088e-05,
+      "loss": 0.6445,
+      "step": 13720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0777752071739724e-05,
+      "loss": 0.6005,
+      "step": 13721
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0776666500612123e-05,
+      "loss": 0.6229,
+      "step": 13722
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0775580920276158e-05,
+      "loss": 0.6234,
+      "step": 13723
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0774495330744703e-05,
+      "loss": 0.6195,
+      "step": 13724
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0773409732030624e-05,
+      "loss": 0.6139,
+      "step": 13725
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0772324124146794e-05,
+      "loss": 0.6141,
+      "step": 13726
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0771238507106086e-05,
+      "loss": 0.6703,
+      "step": 13727
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0770152880921368e-05,
+      "loss": 0.5913,
+      "step": 13728
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0769067245605514e-05,
+      "loss": 0.619,
+      "step": 13729
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0767981601171398e-05,
+      "loss": 0.6056,
+      "step": 13730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0766895947631884e-05,
+      "loss": 0.6577,
+      "step": 13731
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.076581028499985e-05,
+      "loss": 0.6094,
+      "step": 13732
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0764724613288167e-05,
+      "loss": 0.623,
+      "step": 13733
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0763638932509705e-05,
+      "loss": 0.6423,
+      "step": 13734
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0762553242677339e-05,
+      "loss": 0.6375,
+      "step": 13735
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.076146754380394e-05,
+      "loss": 0.6379,
+      "step": 13736
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.076038183590238e-05,
+      "loss": 0.588,
+      "step": 13737
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0759296118985526e-05,
+      "loss": 0.6493,
+      "step": 13738
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0758210393066265e-05,
+      "loss": 0.5952,
+      "step": 13739
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0757124658157456e-05,
+      "loss": 0.6306,
+      "step": 13740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0756038914271973e-05,
+      "loss": 0.6152,
+      "step": 13741
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0754953161422698e-05,
+      "loss": 0.6593,
+      "step": 13742
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0753867399622498e-05,
+      "loss": 0.607,
+      "step": 13743
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0752781628884241e-05,
+      "loss": 0.6708,
+      "step": 13744
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0751695849220812e-05,
+      "loss": 0.6384,
+      "step": 13745
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0750610060645075e-05,
+      "loss": 0.6276,
+      "step": 13746
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0749524263169901e-05,
+      "loss": 0.6732,
+      "step": 13747
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0748438456808176e-05,
+      "loss": 0.6404,
+      "step": 13748
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0747352641572766e-05,
+      "loss": 0.6047,
+      "step": 13749
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.074626681747654e-05,
+      "loss": 0.5643,
+      "step": 13750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0745180984532382e-05,
+      "loss": 0.6933,
+      "step": 13751
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0744095142753159e-05,
+      "loss": 0.6465,
+      "step": 13752
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0743009292151743e-05,
+      "loss": 0.6201,
+      "step": 13753
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0741923432741017e-05,
+      "loss": 0.6157,
+      "step": 13754
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0740837564533851e-05,
+      "loss": 0.6369,
+      "step": 13755
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0739751687543112e-05,
+      "loss": 0.6125,
+      "step": 13756
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0738665801781686e-05,
+      "loss": 0.6647,
+      "step": 13757
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0737579907262443e-05,
+      "loss": 0.616,
+      "step": 13758
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0736494003998253e-05,
+      "loss": 0.5979,
+      "step": 13759
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0735408092002e-05,
+      "loss": 0.6407,
+      "step": 13760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0734322171286554e-05,
+      "loss": 0.6271,
+      "step": 13761
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0733236241864782e-05,
+      "loss": 0.6332,
+      "step": 13762
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0732150303749573e-05,
+      "loss": 0.576,
+      "step": 13763
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0731064356953797e-05,
+      "loss": 0.6187,
+      "step": 13764
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0729978401490322e-05,
+      "loss": 0.6511,
+      "step": 13765
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0728892437372037e-05,
+      "loss": 0.6247,
+      "step": 13766
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0727806464611807e-05,
+      "loss": 0.6172,
+      "step": 13767
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.072672048322251e-05,
+      "loss": 0.7048,
+      "step": 13768
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0725634493217023e-05,
+      "loss": 0.617,
+      "step": 13769
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0724548494608222e-05,
+      "loss": 0.6033,
+      "step": 13770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.072346248740898e-05,
+      "loss": 0.6395,
+      "step": 13771
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0722376471632174e-05,
+      "loss": 0.6665,
+      "step": 13772
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0721290447290687e-05,
+      "loss": 0.6857,
+      "step": 13773
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0720204414397383e-05,
+      "loss": 0.6243,
+      "step": 13774
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0719118372965147e-05,
+      "loss": 0.6708,
+      "step": 13775
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0718032323006855e-05,
+      "loss": 0.5741,
+      "step": 13776
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0716946264535377e-05,
+      "loss": 0.6192,
+      "step": 13777
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0715860197563597e-05,
+      "loss": 0.5971,
+      "step": 13778
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.071477412210439e-05,
+      "loss": 0.6243,
+      "step": 13779
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.071368803817063e-05,
+      "loss": 0.6443,
+      "step": 13780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0712601945775194e-05,
+      "loss": 0.6257,
+      "step": 13781
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0711515844930964e-05,
+      "loss": 0.7077,
+      "step": 13782
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.071042973565081e-05,
+      "loss": 0.6194,
+      "step": 13783
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0709343617947613e-05,
+      "loss": 0.6184,
+      "step": 13784
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0708257491834253e-05,
+      "loss": 0.6046,
+      "step": 13785
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0707171357323602e-05,
+      "loss": 0.6557,
+      "step": 13786
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0706085214428542e-05,
+      "loss": 0.6072,
+      "step": 13787
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0704999063161946e-05,
+      "loss": 0.6108,
+      "step": 13788
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0703912903536698e-05,
+      "loss": 0.6211,
+      "step": 13789
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0702826735565668e-05,
+      "loss": 0.5851,
+      "step": 13790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0701740559261742e-05,
+      "loss": 0.624,
+      "step": 13791
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0700654374637792e-05,
+      "loss": 0.6727,
+      "step": 13792
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0699568181706696e-05,
+      "loss": 0.6211,
+      "step": 13793
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0698481980481337e-05,
+      "loss": 0.5634,
+      "step": 13794
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.069739577097459e-05,
+      "loss": 0.6225,
+      "step": 13795
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0696309553199331e-05,
+      "loss": 0.6687,
+      "step": 13796
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0695223327168446e-05,
+      "loss": 0.6384,
+      "step": 13797
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0694137092894807e-05,
+      "loss": 0.6717,
+      "step": 13798
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0693050850391295e-05,
+      "loss": 0.6998,
+      "step": 13799
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0691964599670787e-05,
+      "loss": 0.6244,
+      "step": 13800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0690878340746165e-05,
+      "loss": 0.6552,
+      "step": 13801
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0689792073630308e-05,
+      "loss": 0.5736,
+      "step": 13802
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.068870579833609e-05,
+      "loss": 0.6635,
+      "step": 13803
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0687619514876393e-05,
+      "loss": 0.675,
+      "step": 13804
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0686533223264099e-05,
+      "loss": 0.6444,
+      "step": 13805
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0685446923512084e-05,
+      "loss": 0.6044,
+      "step": 13806
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0684360615633229e-05,
+      "loss": 0.6029,
+      "step": 13807
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0683274299640416e-05,
+      "loss": 0.6321,
+      "step": 13808
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.068218797554652e-05,
+      "loss": 0.5886,
+      "step": 13809
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.068110164336442e-05,
+      "loss": 0.6916,
+      "step": 13810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0680015303107e-05,
+      "loss": 0.5769,
+      "step": 13811
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0678928954787138e-05,
+      "loss": 0.6613,
+      "step": 13812
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0677842598417713e-05,
+      "loss": 0.6298,
+      "step": 13813
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0676756234011612e-05,
+      "loss": 0.6517,
+      "step": 13814
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0675669861581706e-05,
+      "loss": 0.6257,
+      "step": 13815
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0674583481140878e-05,
+      "loss": 0.6033,
+      "step": 13816
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0673497092702012e-05,
+      "loss": 0.599,
+      "step": 13817
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0672410696277985e-05,
+      "loss": 0.6549,
+      "step": 13818
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0671324291881677e-05,
+      "loss": 0.6364,
+      "step": 13819
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0670237879525973e-05,
+      "loss": 0.6866,
+      "step": 13820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.066915145922375e-05,
+      "loss": 0.6314,
+      "step": 13821
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0668065030987889e-05,
+      "loss": 0.6491,
+      "step": 13822
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0666978594831274e-05,
+      "loss": 0.6658,
+      "step": 13823
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0665892150766781e-05,
+      "loss": 0.6675,
+      "step": 13824
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0664805698807294e-05,
+      "loss": 0.6646,
+      "step": 13825
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.06637192389657e-05,
+      "loss": 0.6205,
+      "step": 13826
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.066263277125487e-05,
+      "loss": 0.6333,
+      "step": 13827
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0661546295687691e-05,
+      "loss": 0.627,
+      "step": 13828
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0660459812277044e-05,
+      "loss": 0.6256,
+      "step": 13829
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0659373321035813e-05,
+      "loss": 0.6283,
+      "step": 13830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0658286821976872e-05,
+      "loss": 0.6493,
+      "step": 13831
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0657200315113113e-05,
+      "loss": 0.6936,
+      "step": 13832
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0656113800457413e-05,
+      "loss": 0.614,
+      "step": 13833
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0655027278022647e-05,
+      "loss": 0.6534,
+      "step": 13834
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0653940747821712e-05,
+      "loss": 0.6572,
+      "step": 13835
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0652854209867478e-05,
+      "loss": 0.5895,
+      "step": 13836
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.065176766417283e-05,
+      "loss": 0.6788,
+      "step": 13837
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0650681110750657e-05,
+      "loss": 0.6705,
+      "step": 13838
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0649594549613834e-05,
+      "loss": 0.5674,
+      "step": 13839
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0648507980775242e-05,
+      "loss": 0.6304,
+      "step": 13840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0647421404247773e-05,
+      "loss": 0.5926,
+      "step": 13841
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0646334820044304e-05,
+      "loss": 0.6278,
+      "step": 13842
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0645248228177712e-05,
+      "loss": 0.6666,
+      "step": 13843
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0644161628660891e-05,
+      "loss": 0.6459,
+      "step": 13844
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0643075021506722e-05,
+      "loss": 0.6491,
+      "step": 13845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0641988406728077e-05,
+      "loss": 0.6616,
+      "step": 13846
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0640901784337853e-05,
+      "loss": 0.7036,
+      "step": 13847
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.063981515434893e-05,
+      "loss": 0.6163,
+      "step": 13848
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0638728516774181e-05,
+      "loss": 0.6349,
+      "step": 13849
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0637641871626503e-05,
+      "loss": 0.6101,
+      "step": 13850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0636555218918773e-05,
+      "loss": 0.5951,
+      "step": 13851
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0635468558663873e-05,
+      "loss": 0.6266,
+      "step": 13852
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0634381890874694e-05,
+      "loss": 0.582,
+      "step": 13853
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0633295215564115e-05,
+      "loss": 0.6255,
+      "step": 13854
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0632208532745014e-05,
+      "loss": 0.6454,
+      "step": 13855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0631121842430287e-05,
+      "loss": 0.6879,
+      "step": 13856
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0630035144632811e-05,
+      "loss": 0.6097,
+      "step": 13857
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0628948439365467e-05,
+      "loss": 0.6104,
+      "step": 13858
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0627861726641151e-05,
+      "loss": 0.5824,
+      "step": 13859
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0626775006472736e-05,
+      "loss": 0.6894,
+      "step": 13860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.062568827887311e-05,
+      "loss": 0.6125,
+      "step": 13861
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0624601543855157e-05,
+      "loss": 0.6032,
+      "step": 13862
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0623514801431767e-05,
+      "loss": 0.6361,
+      "step": 13863
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0622428051615816e-05,
+      "loss": 0.6877,
+      "step": 13864
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0621341294420196e-05,
+      "loss": 0.5843,
+      "step": 13865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0620254529857787e-05,
+      "loss": 0.6027,
+      "step": 13866
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0619167757941476e-05,
+      "loss": 0.6263,
+      "step": 13867
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0618080978684149e-05,
+      "loss": 0.6729,
+      "step": 13868
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.061699419209869e-05,
+      "loss": 0.6673,
+      "step": 13869
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0615907398197981e-05,
+      "loss": 0.6506,
+      "step": 13870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0614820596994915e-05,
+      "loss": 0.5879,
+      "step": 13871
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0613733788502374e-05,
+      "loss": 0.6295,
+      "step": 13872
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0612646972733238e-05,
+      "loss": 0.6023,
+      "step": 13873
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.06115601497004e-05,
+      "loss": 0.6267,
+      "step": 13874
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0610473319416743e-05,
+      "loss": 0.6084,
+      "step": 13875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.060938648189515e-05,
+      "loss": 0.6343,
+      "step": 13876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0608299637148512e-05,
+      "loss": 0.6458,
+      "step": 13877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0607212785189712e-05,
+      "loss": 0.6455,
+      "step": 13878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0606125926031634e-05,
+      "loss": 0.6393,
+      "step": 13879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0605039059687167e-05,
+      "loss": 0.633,
+      "step": 13880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.06039521861692e-05,
+      "loss": 0.5964,
+      "step": 13881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0602865305490612e-05,
+      "loss": 0.5854,
+      "step": 13882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0601778417664295e-05,
+      "loss": 0.5932,
+      "step": 13883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0600691522703134e-05,
+      "loss": 0.6716,
+      "step": 13884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0599604620620014e-05,
+      "loss": 0.6719,
+      "step": 13885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0598517711427823e-05,
+      "loss": 0.6288,
+      "step": 13886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0597430795139447e-05,
+      "loss": 0.6817,
+      "step": 13887
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0596343871767773e-05,
+      "loss": 0.6289,
+      "step": 13888
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.059525694132569e-05,
+      "loss": 0.6298,
+      "step": 13889
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.059417000382608e-05,
+      "loss": 0.6272,
+      "step": 13890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0593083059281837e-05,
+      "loss": 0.616,
+      "step": 13891
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0591996107705842e-05,
+      "loss": 0.6127,
+      "step": 13892
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0590909149110984e-05,
+      "loss": 0.621,
+      "step": 13893
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0589822183510147e-05,
+      "loss": 0.635,
+      "step": 13894
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0588735210916228e-05,
+      "loss": 0.6053,
+      "step": 13895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0587648231342104e-05,
+      "loss": 0.6302,
+      "step": 13896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0586561244800668e-05,
+      "loss": 0.6513,
+      "step": 13897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.058547425130481e-05,
+      "loss": 0.6041,
+      "step": 13898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0584387250867409e-05,
+      "loss": 0.6887,
+      "step": 13899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.058330024350136e-05,
+      "loss": 0.6141,
+      "step": 13900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.058221322921955e-05,
+      "loss": 0.6403,
+      "step": 13901
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0581126208034863e-05,
+      "loss": 0.6178,
+      "step": 13902
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0580039179960192e-05,
+      "loss": 0.5864,
+      "step": 13903
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0578952145008424e-05,
+      "loss": 0.6189,
+      "step": 13904
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0577865103192442e-05,
+      "loss": 0.6888,
+      "step": 13905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0576778054525139e-05,
+      "loss": 0.6514,
+      "step": 13906
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0575690999019405e-05,
+      "loss": 0.6053,
+      "step": 13907
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0574603936688124e-05,
+      "loss": 0.6343,
+      "step": 13908
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0573516867544184e-05,
+      "loss": 0.5559,
+      "step": 13909
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0572429791600482e-05,
+      "loss": 0.6391,
+      "step": 13910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0571342708869896e-05,
+      "loss": 0.6493,
+      "step": 13911
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0570255619365322e-05,
+      "loss": 0.636,
+      "step": 13912
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0569168523099646e-05,
+      "loss": 0.6449,
+      "step": 13913
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0568081420085756e-05,
+      "loss": 0.5966,
+      "step": 13914
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.056699431033654e-05,
+      "loss": 0.6349,
+      "step": 13915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0565907193864895e-05,
+      "loss": 0.6282,
+      "step": 13916
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0564820070683699e-05,
+      "loss": 0.5425,
+      "step": 13917
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0563732940805848e-05,
+      "loss": 0.6126,
+      "step": 13918
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0562645804244234e-05,
+      "loss": 0.5584,
+      "step": 13919
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0561558661011738e-05,
+      "loss": 0.6429,
+      "step": 13920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0560471511121253e-05,
+      "loss": 0.6413,
+      "step": 13921
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0559384354585671e-05,
+      "loss": 0.6106,
+      "step": 13922
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.055829719141788e-05,
+      "loss": 0.6315,
+      "step": 13923
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0557210021630769e-05,
+      "loss": 0.6053,
+      "step": 13924
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.055612284523723e-05,
+      "loss": 0.5879,
+      "step": 13925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0555035662250149e-05,
+      "loss": 0.6246,
+      "step": 13926
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0553948472682417e-05,
+      "loss": 0.6332,
+      "step": 13927
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0552861276546928e-05,
+      "loss": 0.5793,
+      "step": 13928
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0551774073856567e-05,
+      "loss": 0.6267,
+      "step": 13929
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0550686864624226e-05,
+      "loss": 0.6593,
+      "step": 13930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0549599648862798e-05,
+      "loss": 0.6902,
+      "step": 13931
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0548512426585171e-05,
+      "loss": 0.6531,
+      "step": 13932
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0547425197804233e-05,
+      "loss": 0.639,
+      "step": 13933
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.054633796253288e-05,
+      "loss": 0.6337,
+      "step": 13934
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0545250720783995e-05,
+      "loss": 0.6622,
+      "step": 13935
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0544163472570474e-05,
+      "loss": 0.626,
+      "step": 13936
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0543076217905211e-05,
+      "loss": 0.6425,
+      "step": 13937
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0541988956801088e-05,
+      "loss": 0.6417,
+      "step": 13938
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0540901689271001e-05,
+      "loss": 0.5841,
+      "step": 13939
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0539814415327842e-05,
+      "loss": 0.7055,
+      "step": 13940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0538727134984499e-05,
+      "loss": 0.6202,
+      "step": 13941
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0537639848253865e-05,
+      "loss": 0.6659,
+      "step": 13942
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0536552555148831e-05,
+      "loss": 0.6353,
+      "step": 13943
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0535465255682284e-05,
+      "loss": 0.5997,
+      "step": 13944
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0534377949867124e-05,
+      "loss": 0.577,
+      "step": 13945
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0533290637716234e-05,
+      "loss": 0.6096,
+      "step": 13946
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0532203319242509e-05,
+      "loss": 0.6309,
+      "step": 13947
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0531115994458843e-05,
+      "loss": 0.6551,
+      "step": 13948
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0530028663378124e-05,
+      "loss": 0.6861,
+      "step": 13949
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0528941326013243e-05,
+      "loss": 0.6093,
+      "step": 13950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0527853982377094e-05,
+      "loss": 0.6115,
+      "step": 13951
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.052676663248257e-05,
+      "loss": 0.6123,
+      "step": 13952
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.052567927634256e-05,
+      "loss": 0.6617,
+      "step": 13953
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0524591913969954e-05,
+      "loss": 0.5918,
+      "step": 13954
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0523504545377655e-05,
+      "loss": 0.6221,
+      "step": 13955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.052241717057854e-05,
+      "loss": 0.6424,
+      "step": 13956
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0521329789585516e-05,
+      "loss": 0.6211,
+      "step": 13957
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.052024240241146e-05,
+      "loss": 0.6422,
+      "step": 13958
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0519155009069274e-05,
+      "loss": 0.6488,
+      "step": 13959
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0518067609571853e-05,
+      "loss": 0.6533,
+      "step": 13960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.051698020393208e-05,
+      "loss": 0.5822,
+      "step": 13961
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0515892792162856e-05,
+      "loss": 0.6646,
+      "step": 13962
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0514805374277071e-05,
+      "loss": 0.68,
+      "step": 13963
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0513717950287614e-05,
+      "loss": 0.6392,
+      "step": 13964
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.051263052020738e-05,
+      "loss": 0.6402,
+      "step": 13965
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0511543084049267e-05,
+      "loss": 0.6755,
+      "step": 13966
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0510455641826162e-05,
+      "loss": 0.6227,
+      "step": 13967
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.050936819355096e-05,
+      "loss": 0.6829,
+      "step": 13968
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0508280739236554e-05,
+      "loss": 0.6318,
+      "step": 13969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0507193278895833e-05,
+      "loss": 0.6704,
+      "step": 13970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0506105812541696e-05,
+      "loss": 0.6498,
+      "step": 13971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0505018340187039e-05,
+      "loss": 0.5643,
+      "step": 13972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0503930861844745e-05,
+      "loss": 0.643,
+      "step": 13973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0502843377527716e-05,
+      "loss": 0.617,
+      "step": 13974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0501755887248842e-05,
+      "loss": 0.6641,
+      "step": 13975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0500668391021017e-05,
+      "loss": 0.6636,
+      "step": 13976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0499580888857134e-05,
+      "loss": 0.607,
+      "step": 13977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0498493380770089e-05,
+      "loss": 0.6238,
+      "step": 13978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0497405866772775e-05,
+      "loss": 0.6386,
+      "step": 13979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0496318346878084e-05,
+      "loss": 0.6604,
+      "step": 13980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.049523082109891e-05,
+      "loss": 0.6662,
+      "step": 13981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0494143289448149e-05,
+      "loss": 0.6236,
+      "step": 13982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0493055751938696e-05,
+      "loss": 0.7105,
+      "step": 13983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.049196820858344e-05,
+      "loss": 0.6131,
+      "step": 13984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.049088065939528e-05,
+      "loss": 0.6403,
+      "step": 13985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0489793104387112e-05,
+      "loss": 0.5793,
+      "step": 13986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0488705543571822e-05,
+      "loss": 0.6318,
+      "step": 13987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0487617976962315e-05,
+      "loss": 0.6126,
+      "step": 13988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0486530404571476e-05,
+      "loss": 0.5748,
+      "step": 13989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0485442826412204e-05,
+      "loss": 0.6016,
+      "step": 13990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0484355242497396e-05,
+      "loss": 0.6169,
+      "step": 13991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0483267652839939e-05,
+      "loss": 0.6131,
+      "step": 13992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0482180057452735e-05,
+      "loss": 0.6196,
+      "step": 13993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0481092456348679e-05,
+      "loss": 0.5654,
+      "step": 13994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.048000484954066e-05,
+      "loss": 0.6098,
+      "step": 13995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0478917237041578e-05,
+      "loss": 0.5889,
+      "step": 13996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0477829618864325e-05,
+      "loss": 0.6379,
+      "step": 13997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.04767419950218e-05,
+      "loss": 0.6097,
+      "step": 13998
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0475654365526893e-05,
+      "loss": 0.6435,
+      "step": 13999
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0474566730392504e-05,
+      "loss": 0.6362,
+      "step": 14000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0473479089631522e-05,
+      "loss": 0.6563,
+      "step": 14001
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0472391443256848e-05,
+      "loss": 0.6142,
+      "step": 14002
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.047130379128138e-05,
+      "loss": 0.6879,
+      "step": 14003
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0470216133718005e-05,
+      "loss": 0.6213,
+      "step": 14004
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0469128470579624e-05,
+      "loss": 0.6758,
+      "step": 14005
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0468040801879133e-05,
+      "loss": 0.5881,
+      "step": 14006
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0466953127629425e-05,
+      "loss": 0.6658,
+      "step": 14007
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0465865447843396e-05,
+      "loss": 0.6333,
+      "step": 14008
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0464777762533944e-05,
+      "loss": 0.6361,
+      "step": 14009
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0463690071713964e-05,
+      "loss": 0.6493,
+      "step": 14010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.046260237539635e-05,
+      "loss": 0.6919,
+      "step": 14011
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0461514673594003e-05,
+      "loss": 0.6091,
+      "step": 14012
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.046042696631981e-05,
+      "loss": 0.6685,
+      "step": 14013
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0459339253586675e-05,
+      "loss": 0.5884,
+      "step": 14014
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0458251535407495e-05,
+      "loss": 0.6356,
+      "step": 14015
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0457163811795161e-05,
+      "loss": 0.5936,
+      "step": 14016
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0456076082762572e-05,
+      "loss": 0.6481,
+      "step": 14017
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0454988348322628e-05,
+      "loss": 0.6016,
+      "step": 14018
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0453900608488217e-05,
+      "loss": 0.606,
+      "step": 14019
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.045281286327224e-05,
+      "loss": 0.6241,
+      "step": 14020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0451725112687597e-05,
+      "loss": 0.6739,
+      "step": 14021
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0450637356747178e-05,
+      "loss": 0.625,
+      "step": 14022
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0449549595463884e-05,
+      "loss": 0.6411,
+      "step": 14023
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.044846182885061e-05,
+      "loss": 0.6123,
+      "step": 14024
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0447374056920256e-05,
+      "loss": 0.6183,
+      "step": 14025
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0446286279685714e-05,
+      "loss": 0.6426,
+      "step": 14026
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0445198497159887e-05,
+      "loss": 0.6252,
+      "step": 14027
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0444110709355666e-05,
+      "loss": 0.5421,
+      "step": 14028
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.044302291628595e-05,
+      "loss": 0.693,
+      "step": 14029
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0441935117963639e-05,
+      "loss": 0.6667,
+      "step": 14030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0440847314401627e-05,
+      "loss": 0.6173,
+      "step": 14031
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0439759505612813e-05,
+      "loss": 0.6596,
+      "step": 14032
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0438671691610095e-05,
+      "loss": 0.5867,
+      "step": 14033
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0437583872406366e-05,
+      "loss": 0.6569,
+      "step": 14034
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0436496048014528e-05,
+      "loss": 0.5929,
+      "step": 14035
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.043540821844748e-05,
+      "loss": 0.6524,
+      "step": 14036
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0434320383718113e-05,
+      "loss": 0.6087,
+      "step": 14037
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0433232543839331e-05,
+      "loss": 0.678,
+      "step": 14038
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.043214469882403e-05,
+      "loss": 0.6358,
+      "step": 14039
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0431056848685102e-05,
+      "loss": 0.6403,
+      "step": 14040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0429968993435453e-05,
+      "loss": 0.6263,
+      "step": 14041
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0428881133087977e-05,
+      "loss": 0.6511,
+      "step": 14042
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0427793267655575e-05,
+      "loss": 0.6118,
+      "step": 14043
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0426705397151138e-05,
+      "loss": 0.6433,
+      "step": 14044
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0425617521587575e-05,
+      "loss": 0.6182,
+      "step": 14045
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0424529640977772e-05,
+      "loss": 0.5992,
+      "step": 14046
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0423441755334636e-05,
+      "loss": 0.6416,
+      "step": 14047
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0422353864671062e-05,
+      "loss": 0.6547,
+      "step": 14048
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0421265968999948e-05,
+      "loss": 0.5908,
+      "step": 14049
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0420178068334195e-05,
+      "loss": 0.6374,
+      "step": 14050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0419090162686695e-05,
+      "loss": 0.6463,
+      "step": 14051
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0418002252070354e-05,
+      "loss": 0.6529,
+      "step": 14052
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.041691433649807e-05,
+      "loss": 0.6237,
+      "step": 14053
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0415826415982734e-05,
+      "loss": 0.6287,
+      "step": 14054
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0414738490537254e-05,
+      "loss": 0.5682,
+      "step": 14055
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0413650560174521e-05,
+      "loss": 0.5872,
+      "step": 14056
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.041256262490744e-05,
+      "loss": 0.6776,
+      "step": 14057
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0411474684748906e-05,
+      "loss": 0.65,
+      "step": 14058
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0410386739711823e-05,
+      "loss": 0.6386,
+      "step": 14059
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0409298789809078e-05,
+      "loss": 0.6705,
+      "step": 14060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0408210835053583e-05,
+      "loss": 0.6189,
+      "step": 14061
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0407122875458234e-05,
+      "loss": 0.6253,
+      "step": 14062
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0406034911035925e-05,
+      "loss": 0.6037,
+      "step": 14063
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.040494694179956e-05,
+      "loss": 0.6142,
+      "step": 14064
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0403858967762038e-05,
+      "loss": 0.618,
+      "step": 14065
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0402770988936253e-05,
+      "loss": 0.6268,
+      "step": 14066
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.040168300533511e-05,
+      "loss": 0.6632,
+      "step": 14067
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.040059501697151e-05,
+      "loss": 0.6256,
+      "step": 14068
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0399507023858347e-05,
+      "loss": 0.6857,
+      "step": 14069
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.039841902600852e-05,
+      "loss": 0.6355,
+      "step": 14070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0397331023434936e-05,
+      "loss": 0.6196,
+      "step": 14071
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0396243016150487e-05,
+      "loss": 0.655,
+      "step": 14072
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0395155004168078e-05,
+      "loss": 0.6175,
+      "step": 14073
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0394066987500607e-05,
+      "loss": 0.6075,
+      "step": 14074
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0392978966160972e-05,
+      "loss": 0.669,
+      "step": 14075
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0391890940162072e-05,
+      "loss": 0.6226,
+      "step": 14076
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0390802909516812e-05,
+      "loss": 0.6474,
+      "step": 14077
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0389714874238088e-05,
+      "loss": 0.5774,
+      "step": 14078
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0388626834338802e-05,
+      "loss": 0.5968,
+      "step": 14079
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0387538789831854e-05,
+      "loss": 0.6932,
+      "step": 14080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0386450740730142e-05,
+      "loss": 0.6165,
+      "step": 14081
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0385362687046567e-05,
+      "loss": 0.6284,
+      "step": 14082
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0384274628794033e-05,
+      "loss": 0.6323,
+      "step": 14083
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0383186565985438e-05,
+      "loss": 0.5797,
+      "step": 14084
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0382098498633672e-05,
+      "loss": 0.6452,
+      "step": 14085
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0381010426751655e-05,
+      "loss": 0.6507,
+      "step": 14086
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0379922350352274e-05,
+      "loss": 0.6613,
+      "step": 14087
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.037883426944843e-05,
+      "loss": 0.6444,
+      "step": 14088
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0377746184053032e-05,
+      "loss": 0.6317,
+      "step": 14089
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0376658094178973e-05,
+      "loss": 0.5835,
+      "step": 14090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0375569999839153e-05,
+      "loss": 0.6118,
+      "step": 14091
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.037448190104648e-05,
+      "loss": 0.5663,
+      "step": 14092
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0373393797813849e-05,
+      "loss": 0.6112,
+      "step": 14093
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0372305690154159e-05,
+      "loss": 0.6616,
+      "step": 14094
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0371217578080316e-05,
+      "loss": 0.6521,
+      "step": 14095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0370129461605223e-05,
+      "loss": 0.6408,
+      "step": 14096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.036904134074177e-05,
+      "loss": 0.7004,
+      "step": 14097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.036795321550287e-05,
+      "loss": 0.6609,
+      "step": 14098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.036686508590142e-05,
+      "loss": 0.6985,
+      "step": 14099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0365776951950315e-05,
+      "loss": 0.6153,
+      "step": 14100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0364688813662464e-05,
+      "loss": 0.6448,
+      "step": 14101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0363600671050768e-05,
+      "loss": 0.6472,
+      "step": 14102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0362512524128123e-05,
+      "loss": 0.6471,
+      "step": 14103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0361424372907435e-05,
+      "loss": 0.567,
+      "step": 14104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0360336217401606e-05,
+      "loss": 0.6206,
+      "step": 14105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.035924805762353e-05,
+      "loss": 0.5929,
+      "step": 14106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.035815989358612e-05,
+      "loss": 0.6364,
+      "step": 14107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0357071725302268e-05,
+      "loss": 0.6043,
+      "step": 14108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0355983552784876e-05,
+      "loss": 0.6118,
+      "step": 14109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0354895376046854e-05,
+      "loss": 0.6908,
+      "step": 14110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0353807195101098e-05,
+      "loss": 0.5775,
+      "step": 14111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0352719009960508e-05,
+      "loss": 0.6707,
+      "step": 14112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0351630820637987e-05,
+      "loss": 0.643,
+      "step": 14113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.035054262714644e-05,
+      "loss": 0.6315,
+      "step": 14114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0349454429498766e-05,
+      "loss": 0.6064,
+      "step": 14115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0348366227707867e-05,
+      "loss": 0.6262,
+      "step": 14116
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0347278021786646e-05,
+      "loss": 0.6346,
+      "step": 14117
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0346189811748007e-05,
+      "loss": 0.6543,
+      "step": 14118
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0345101597604846e-05,
+      "loss": 0.6004,
+      "step": 14119
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0344013379370072e-05,
+      "loss": 0.6166,
+      "step": 14120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0342925157056582e-05,
+      "loss": 0.6356,
+      "step": 14121
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.034183693067728e-05,
+      "loss": 0.6298,
+      "step": 14122
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0340748700245074e-05,
+      "loss": 0.685,
+      "step": 14123
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0339660465772856e-05,
+      "loss": 0.6713,
+      "step": 14124
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0338572227273534e-05,
+      "loss": 0.6424,
+      "step": 14125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0337483984760012e-05,
+      "loss": 0.662,
+      "step": 14126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0336395738245188e-05,
+      "loss": 0.6023,
+      "step": 14127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0335307487741968e-05,
+      "loss": 0.6508,
+      "step": 14128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0334219233263257e-05,
+      "loss": 0.608,
+      "step": 14129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0333130974821948e-05,
+      "loss": 0.6381,
+      "step": 14130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0332042712430952e-05,
+      "loss": 0.6287,
+      "step": 14131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0330954446103172e-05,
+      "loss": 0.6146,
+      "step": 14132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0329866175851507e-05,
+      "loss": 0.7095,
+      "step": 14133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.032877790168886e-05,
+      "loss": 0.5877,
+      "step": 14134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0327689623628139e-05,
+      "loss": 0.6407,
+      "step": 14135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0326601341682238e-05,
+      "loss": 0.5947,
+      "step": 14136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0325513055864068e-05,
+      "loss": 0.6704,
+      "step": 14137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0324424766186527e-05,
+      "loss": 0.5956,
+      "step": 14138
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0323336472662521e-05,
+      "loss": 0.674,
+      "step": 14139
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0322248175304951e-05,
+      "loss": 0.6373,
+      "step": 14140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0321159874126726e-05,
+      "loss": 0.6505,
+      "step": 14141
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.032007156914074e-05,
+      "loss": 0.6687,
+      "step": 14142
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0318983260359902e-05,
+      "loss": 0.6339,
+      "step": 14143
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0317894947797113e-05,
+      "loss": 0.6528,
+      "step": 14144
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0316806631465277e-05,
+      "loss": 0.594,
+      "step": 14145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0315718311377298e-05,
+      "loss": 0.6113,
+      "step": 14146
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.031462998754608e-05,
+      "loss": 0.6321,
+      "step": 14147
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0313541659984523e-05,
+      "loss": 0.6829,
+      "step": 14148
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0312453328705536e-05,
+      "loss": 0.5906,
+      "step": 14149
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0311364993722019e-05,
+      "loss": 0.6104,
+      "step": 14150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0310276655046873e-05,
+      "loss": 0.6161,
+      "step": 14151
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.030918831269301e-05,
+      "loss": 0.6145,
+      "step": 14152
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0308099966673323e-05,
+      "loss": 0.6475,
+      "step": 14153
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0307011617000723e-05,
+      "loss": 0.684,
+      "step": 14154
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0305923263688116e-05,
+      "loss": 0.6269,
+      "step": 14155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0304834906748397e-05,
+      "loss": 0.6442,
+      "step": 14156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0303746546194475e-05,
+      "loss": 0.6646,
+      "step": 14157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0302658182039256e-05,
+      "loss": 0.6239,
+      "step": 14158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.030156981429564e-05,
+      "loss": 0.5904,
+      "step": 14159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0300481442976531e-05,
+      "loss": 0.6186,
+      "step": 14160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0299393068094836e-05,
+      "loss": 0.6213,
+      "step": 14161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0298304689663457e-05,
+      "loss": 0.5326,
+      "step": 14162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0297216307695297e-05,
+      "loss": 0.6033,
+      "step": 14163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0296127922203266e-05,
+      "loss": 0.6596,
+      "step": 14164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.029503953320026e-05,
+      "loss": 0.6443,
+      "step": 14165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0293951140699187e-05,
+      "loss": 0.617,
+      "step": 14166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0292862744712955e-05,
+      "loss": 0.6583,
+      "step": 14167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0291774345254462e-05,
+      "loss": 0.5442,
+      "step": 14168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0290685942336614e-05,
+      "loss": 0.6071,
+      "step": 14169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0289597535972317e-05,
+      "loss": 0.6174,
+      "step": 14170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0288509126174478e-05,
+      "loss": 0.6008,
+      "step": 14171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0287420712955994e-05,
+      "loss": 0.6568,
+      "step": 14172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0286332296329777e-05,
+      "loss": 0.5977,
+      "step": 14173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0285243876308729e-05,
+      "loss": 0.6735,
+      "step": 14174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0284155452905748e-05,
+      "loss": 0.6427,
+      "step": 14175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0283067026133751e-05,
+      "loss": 0.6732,
+      "step": 14176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0281978596005635e-05,
+      "loss": 0.6386,
+      "step": 14177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0280890162534302e-05,
+      "loss": 0.6206,
+      "step": 14178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0279801725732668e-05,
+      "loss": 0.6374,
+      "step": 14179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0278713285613626e-05,
+      "loss": 0.6468,
+      "step": 14180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0277624842190084e-05,
+      "loss": 0.6149,
+      "step": 14181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0276536395474952e-05,
+      "loss": 0.617,
+      "step": 14182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0275447945481132e-05,
+      "loss": 0.6007,
+      "step": 14183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0274359492221522e-05,
+      "loss": 0.588,
+      "step": 14184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0273271035709038e-05,
+      "loss": 0.5834,
+      "step": 14185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.027218257595658e-05,
+      "loss": 0.6129,
+      "step": 14186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0271094112977049e-05,
+      "loss": 0.5998,
+      "step": 14187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0270005646783361e-05,
+      "loss": 0.6228,
+      "step": 14188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.026891717738841e-05,
+      "loss": 0.6286,
+      "step": 14189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0267828704805106e-05,
+      "loss": 0.6805,
+      "step": 14190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0266740229046354e-05,
+      "loss": 0.6399,
+      "step": 14191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0265651750125063e-05,
+      "loss": 0.5576,
+      "step": 14192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0264563268054129e-05,
+      "loss": 0.6188,
+      "step": 14193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0263474782846466e-05,
+      "loss": 0.6553,
+      "step": 14194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0262386294514977e-05,
+      "loss": 0.6794,
+      "step": 14195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.026129780307256e-05,
+      "loss": 0.6024,
+      "step": 14196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0260209308532134e-05,
+      "loss": 0.6641,
+      "step": 14197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0259120810906597e-05,
+      "loss": 0.6601,
+      "step": 14198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0258032310208852e-05,
+      "loss": 0.6242,
+      "step": 14199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.025694380645181e-05,
+      "loss": 0.6437,
+      "step": 14200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0255855299648374e-05,
+      "loss": 0.6112,
+      "step": 14201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0254766789811446e-05,
+      "loss": 0.6684,
+      "step": 14202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.025367827695394e-05,
+      "loss": 0.6396,
+      "step": 14203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0252589761088757e-05,
+      "loss": 0.6494,
+      "step": 14204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0251501242228801e-05,
+      "loss": 0.6465,
+      "step": 14205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0250412720386977e-05,
+      "loss": 0.6368,
+      "step": 14206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0249324195576198e-05,
+      "loss": 0.6628,
+      "step": 14207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0248235667809363e-05,
+      "loss": 0.6429,
+      "step": 14208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.024714713709938e-05,
+      "loss": 0.5724,
+      "step": 14209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0246058603459157e-05,
+      "loss": 0.6376,
+      "step": 14210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0244970066901597e-05,
+      "loss": 0.6494,
+      "step": 14211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0243881527439605e-05,
+      "loss": 0.6358,
+      "step": 14212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0242792985086091e-05,
+      "loss": 0.6258,
+      "step": 14213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0241704439853959e-05,
+      "loss": 0.6759,
+      "step": 14214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0240615891756113e-05,
+      "loss": 0.6443,
+      "step": 14215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0239527340805466e-05,
+      "loss": 0.6298,
+      "step": 14216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0238438787014916e-05,
+      "loss": 0.6055,
+      "step": 14217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.023735023039737e-05,
+      "loss": 0.6967,
+      "step": 14218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0236261670965742e-05,
+      "loss": 0.6275,
+      "step": 14219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.023517310873293e-05,
+      "loss": 0.614,
+      "step": 14220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0234084543711843e-05,
+      "loss": 0.654,
+      "step": 14221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.023299597591539e-05,
+      "loss": 0.6679,
+      "step": 14222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0231907405356474e-05,
+      "loss": 0.6493,
+      "step": 14223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0230818832047999e-05,
+      "loss": 0.6303,
+      "step": 14224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0229730256002881e-05,
+      "loss": 0.6201,
+      "step": 14225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0228641677234015e-05,
+      "loss": 0.6403,
+      "step": 14226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0227553095754315e-05,
+      "loss": 0.5952,
+      "step": 14227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0226464511576686e-05,
+      "loss": 0.6579,
+      "step": 14228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0225375924714031e-05,
+      "loss": 0.6448,
+      "step": 14229
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0224287335179261e-05,
+      "loss": 0.6036,
+      "step": 14230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.022319874298528e-05,
+      "loss": 0.6498,
+      "step": 14231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0222110148144998e-05,
+      "loss": 0.6534,
+      "step": 14232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0221021550671315e-05,
+      "loss": 0.6513,
+      "step": 14233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0219932950577148e-05,
+      "loss": 0.643,
+      "step": 14234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0218844347875391e-05,
+      "loss": 0.6684,
+      "step": 14235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0217755742578963e-05,
+      "loss": 0.6158,
+      "step": 14236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0216667134700763e-05,
+      "loss": 0.605,
+      "step": 14237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0215578524253698e-05,
+      "loss": 0.6328,
+      "step": 14238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.021448991125068e-05,
+      "loss": 0.6793,
+      "step": 14239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.021340129570461e-05,
+      "loss": 0.6477,
+      "step": 14240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0212312677628399e-05,
+      "loss": 0.6496,
+      "step": 14241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0211224057034954e-05,
+      "loss": 0.6061,
+      "step": 14242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0210135433937178e-05,
+      "loss": 0.6116,
+      "step": 14243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0209046808347983e-05,
+      "loss": 0.6327,
+      "step": 14244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0207958180280273e-05,
+      "loss": 0.629,
+      "step": 14245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0206869549746953e-05,
+      "loss": 0.6181,
+      "step": 14246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0205780916760937e-05,
+      "loss": 0.6604,
+      "step": 14247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0204692281335126e-05,
+      "loss": 0.6095,
+      "step": 14248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0203603643482429e-05,
+      "loss": 0.606,
+      "step": 14249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.020251500321575e-05,
+      "loss": 0.5874,
+      "step": 14250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0201426360548005e-05,
+      "loss": 0.6236,
+      "step": 14251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0200337715492094e-05,
+      "loss": 0.6173,
+      "step": 14252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0199249068060923e-05,
+      "loss": 0.6142,
+      "step": 14253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.019816041826741e-05,
+      "loss": 0.6341,
+      "step": 14254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0197071766124447e-05,
+      "loss": 0.6715,
+      "step": 14255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.019598311164495e-05,
+      "loss": 0.6577,
+      "step": 14256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0194894454841829e-05,
+      "loss": 0.6367,
+      "step": 14257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0193805795727983e-05,
+      "loss": 0.6122,
+      "step": 14258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0192717134316325e-05,
+      "loss": 0.6568,
+      "step": 14259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0191628470619765e-05,
+      "loss": 0.6435,
+      "step": 14260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0190539804651203e-05,
+      "loss": 0.671,
+      "step": 14261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0189451136423552e-05,
+      "loss": 0.6483,
+      "step": 14262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0188362465949719e-05,
+      "loss": 0.6661,
+      "step": 14263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.018727379324261e-05,
+      "loss": 0.5965,
+      "step": 14264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0186185118315132e-05,
+      "loss": 0.6336,
+      "step": 14265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0185096441180198e-05,
+      "loss": 0.6746,
+      "step": 14266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.018400776185071e-05,
+      "loss": 0.633,
+      "step": 14267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0182919080339575e-05,
+      "loss": 0.6058,
+      "step": 14268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0181830396659705e-05,
+      "loss": 0.6331,
+      "step": 14269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0180741710824004e-05,
+      "loss": 0.5502,
+      "step": 14270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0179653022845381e-05,
+      "loss": 0.6438,
+      "step": 14271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0178564332736747e-05,
+      "loss": 0.5867,
+      "step": 14272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0177475640511006e-05,
+      "loss": 0.6337,
+      "step": 14273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0176386946181062e-05,
+      "loss": 0.6395,
+      "step": 14274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0175298249759835e-05,
+      "loss": 0.649,
+      "step": 14275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.017420955126022e-05,
+      "loss": 0.616,
+      "step": 14276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0173120850695132e-05,
+      "loss": 0.594,
+      "step": 14277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.017203214807748e-05,
+      "loss": 0.6655,
+      "step": 14278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0170943443420168e-05,
+      "loss": 0.5961,
+      "step": 14279
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0169854736736103e-05,
+      "loss": 0.5546,
+      "step": 14280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0168766028038197e-05,
+      "loss": 0.6158,
+      "step": 14281
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0167677317339355e-05,
+      "loss": 0.6491,
+      "step": 14282
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0166588604652488e-05,
+      "loss": 0.631,
+      "step": 14283
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0165499889990503e-05,
+      "loss": 0.5973,
+      "step": 14284
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0164411173366305e-05,
+      "loss": 0.6228,
+      "step": 14285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0163322454792803e-05,
+      "loss": 0.6392,
+      "step": 14286
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0162233734282912e-05,
+      "loss": 0.6721,
+      "step": 14287
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.016114501184953e-05,
+      "loss": 0.6023,
+      "step": 14288
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.016005628750557e-05,
+      "loss": 0.6532,
+      "step": 14289
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0158967561263946e-05,
+      "loss": 0.6428,
+      "step": 14290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0157878833137556e-05,
+      "loss": 0.5904,
+      "step": 14291
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.015679010313931e-05,
+      "loss": 0.6743,
+      "step": 14292
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0155701371282125e-05,
+      "loss": 0.594,
+      "step": 14293
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.01546126375789e-05,
+      "loss": 0.6725,
+      "step": 14294
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0153523902042548e-05,
+      "loss": 0.6545,
+      "step": 14295
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0152435164685977e-05,
+      "loss": 0.6466,
+      "step": 14296
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.015134642552209e-05,
+      "loss": 0.6234,
+      "step": 14297
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0150257684563804e-05,
+      "loss": 0.5935,
+      "step": 14298
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0149168941824022e-05,
+      "loss": 0.6225,
+      "step": 14299
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0148080197315651e-05,
+      "loss": 0.6116,
+      "step": 14300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0146991451051604e-05,
+      "loss": 0.6247,
+      "step": 14301
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0145902703044788e-05,
+      "loss": 0.6081,
+      "step": 14302
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0144813953308108e-05,
+      "loss": 0.6317,
+      "step": 14303
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.014372520185448e-05,
+      "loss": 0.6771,
+      "step": 14304
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0142636448696803e-05,
+      "loss": 0.6132,
+      "step": 14305
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0141547693847992e-05,
+      "loss": 0.6341,
+      "step": 14306
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0140458937320958e-05,
+      "loss": 0.6659,
+      "step": 14307
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0139370179128601e-05,
+      "loss": 0.6558,
+      "step": 14308
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0138281419283835e-05,
+      "loss": 0.5942,
+      "step": 14309
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0137192657799571e-05,
+      "loss": 0.6049,
+      "step": 14310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0136103894688709e-05,
+      "loss": 0.6065,
+      "step": 14311
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0135015129964168e-05,
+      "loss": 0.6078,
+      "step": 14312
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0133926363638851e-05,
+      "loss": 0.6473,
+      "step": 14313
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0132837595725665e-05,
+      "loss": 0.6344,
+      "step": 14314
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0131748826237522e-05,
+      "loss": 0.6871,
+      "step": 14315
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.013066005518733e-05,
+      "loss": 0.6406,
+      "step": 14316
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0129571282588e-05,
+      "loss": 0.6617,
+      "step": 14317
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0128482508452435e-05,
+      "loss": 0.6405,
+      "step": 14318
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.012739373279355e-05,
+      "loss": 0.6599,
+      "step": 14319
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.012630495562425e-05,
+      "loss": 0.6301,
+      "step": 14320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0125216176957446e-05,
+      "loss": 0.5889,
+      "step": 14321
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0124127396806048e-05,
+      "loss": 0.6571,
+      "step": 14322
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0123038615182959e-05,
+      "loss": 0.6243,
+      "step": 14323
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0121949832101092e-05,
+      "loss": 0.6351,
+      "step": 14324
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0120861047573358e-05,
+      "loss": 0.704,
+      "step": 14325
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0119772261612659e-05,
+      "loss": 0.6067,
+      "step": 14326
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0118683474231911e-05,
+      "loss": 0.6304,
+      "step": 14327
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0117594685444021e-05,
+      "loss": 0.6418,
+      "step": 14328
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.01165058952619e-05,
+      "loss": 0.6265,
+      "step": 14329
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0115417103698447e-05,
+      "loss": 0.6492,
+      "step": 14330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0114328310766582e-05,
+      "loss": 0.5989,
+      "step": 14331
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0113239516479212e-05,
+      "loss": 0.5833,
+      "step": 14332
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0112150720849239e-05,
+      "loss": 0.6569,
+      "step": 14333
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0111061923889582e-05,
+      "loss": 0.6336,
+      "step": 14334
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0109973125613146e-05,
+      "loss": 0.6327,
+      "step": 14335
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0108884326032834e-05,
+      "loss": 0.617,
+      "step": 14336
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0107795525161565e-05,
+      "loss": 0.6322,
+      "step": 14337
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0106706723012246e-05,
+      "loss": 0.6144,
+      "step": 14338
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0105617919597777e-05,
+      "loss": 0.6724,
+      "step": 14339
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0104529114931079e-05,
+      "loss": 0.6715,
+      "step": 14340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0103440309025056e-05,
+      "loss": 0.6476,
+      "step": 14341
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0102351501892616e-05,
+      "loss": 0.694,
+      "step": 14342
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0101262693546668e-05,
+      "loss": 0.5982,
+      "step": 14343
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0100173884000126e-05,
+      "loss": 0.6505,
+      "step": 14344
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0099085073265894e-05,
+      "loss": 0.6528,
+      "step": 14345
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.009799626135688e-05,
+      "loss": 0.6333,
+      "step": 14346
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0096907448286001e-05,
+      "loss": 0.6289,
+      "step": 14347
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.009581863406616e-05,
+      "loss": 0.5271,
+      "step": 14348
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0094729818710267e-05,
+      "loss": 0.6871,
+      "step": 14349
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0093641002231235e-05,
+      "loss": 0.6226,
+      "step": 14350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0092552184641969e-05,
+      "loss": 0.6057,
+      "step": 14351
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0091463365955378e-05,
+      "loss": 0.6424,
+      "step": 14352
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0090374546184376e-05,
+      "loss": 0.6524,
+      "step": 14353
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0089285725341866e-05,
+      "loss": 0.6352,
+      "step": 14354
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0088196903440763e-05,
+      "loss": 0.5802,
+      "step": 14355
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0087108080493975e-05,
+      "loss": 0.6602,
+      "step": 14356
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0086019256514407e-05,
+      "loss": 0.6205,
+      "step": 14357
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0084930431514975e-05,
+      "loss": 0.6405,
+      "step": 14358
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0083841605508586e-05,
+      "loss": 0.6411,
+      "step": 14359
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0082752778508148e-05,
+      "loss": 0.6281,
+      "step": 14360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0081663950526568e-05,
+      "loss": 0.6148,
+      "step": 14361
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0080575121576761e-05,
+      "loss": 0.617,
+      "step": 14362
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0079486291671634e-05,
+      "loss": 0.6497,
+      "step": 14363
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0078397460824096e-05,
+      "loss": 0.6476,
+      "step": 14364
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0077308629047058e-05,
+      "loss": 0.6297,
+      "step": 14365
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0076219796353428e-05,
+      "loss": 0.6908,
+      "step": 14366
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0075130962756115e-05,
+      "loss": 0.5614,
+      "step": 14367
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.007404212826803e-05,
+      "loss": 0.6123,
+      "step": 14368
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0072953292902082e-05,
+      "loss": 0.6676,
+      "step": 14369
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0071864456671179e-05,
+      "loss": 0.5812,
+      "step": 14370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0070775619588236e-05,
+      "loss": 0.6192,
+      "step": 14371
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0069686781666156e-05,
+      "loss": 0.5759,
+      "step": 14372
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0068597942917848e-05,
+      "loss": 0.6694,
+      "step": 14373
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.006750910335623e-05,
+      "loss": 0.6671,
+      "step": 14374
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0066420262994203e-05,
+      "loss": 0.6339,
+      "step": 14375
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.006533142184468e-05,
+      "loss": 0.5978,
+      "step": 14376
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0064242579920571e-05,
+      "loss": 0.6054,
+      "step": 14377
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0063153737234784e-05,
+      "loss": 0.6203,
+      "step": 14378
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.006206489380023e-05,
+      "loss": 0.6096,
+      "step": 14379
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.006097604962982e-05,
+      "loss": 0.583,
+      "step": 14380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0059887204736458e-05,
+      "loss": 0.6662,
+      "step": 14381
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.005879835913306e-05,
+      "loss": 0.6413,
+      "step": 14382
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0057709512832533e-05,
+      "loss": 0.6244,
+      "step": 14383
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0056620665847788e-05,
+      "loss": 0.7072,
+      "step": 14384
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.005553181819173e-05,
+      "loss": 0.6486,
+      "step": 14385
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0054442969877276e-05,
+      "loss": 0.6265,
+      "step": 14386
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0053354120917326e-05,
+      "loss": 0.6269,
+      "step": 14387
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.00522652713248e-05,
+      "loss": 0.5966,
+      "step": 14388
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0051176421112603e-05,
+      "loss": 0.6834,
+      "step": 14389
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0050087570293643e-05,
+      "loss": 0.6106,
+      "step": 14390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0048998718880834e-05,
+      "loss": 0.677,
+      "step": 14391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0047909866887081e-05,
+      "loss": 0.5819,
+      "step": 14392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0046821014325297e-05,
+      "loss": 0.6472,
+      "step": 14393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.004573216120839e-05,
+      "loss": 0.5778,
+      "step": 14394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.004464330754927e-05,
+      "loss": 0.5917,
+      "step": 14395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0043554453360847e-05,
+      "loss": 0.68,
+      "step": 14396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0042465598656033e-05,
+      "loss": 0.6033,
+      "step": 14397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0041376743447732e-05,
+      "loss": 0.6473,
+      "step": 14398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0040287887748861e-05,
+      "loss": 0.668,
+      "step": 14399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0039199031572326e-05,
+      "loss": 0.6483,
+      "step": 14400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0038110174931035e-05,
+      "loss": 0.5752,
+      "step": 14401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0037021317837898e-05,
+      "loss": 0.5742,
+      "step": 14402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0035932460305831e-05,
+      "loss": 0.6958,
+      "step": 14403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0034843602347735e-05,
+      "loss": 0.5937,
+      "step": 14404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0033754743976525e-05,
+      "loss": 0.6463,
+      "step": 14405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0032665885205113e-05,
+      "loss": 0.5964,
+      "step": 14406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0031577026046401e-05,
+      "loss": 0.6406,
+      "step": 14407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0030488166513307e-05,
+      "loss": 0.6369,
+      "step": 14408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0029399306618737e-05,
+      "loss": 0.6658,
+      "step": 14409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0028310446375598e-05,
+      "loss": 0.6818,
+      "step": 14410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0027221585796804e-05,
+      "loss": 0.6566,
+      "step": 14411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0026132724895265e-05,
+      "loss": 0.6225,
+      "step": 14412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.002504386368389e-05,
+      "loss": 0.6283,
+      "step": 14413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0023955002175587e-05,
+      "loss": 0.6394,
+      "step": 14414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0022866140383267e-05,
+      "loss": 0.6216,
+      "step": 14415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0021777278319839e-05,
+      "loss": 0.7218,
+      "step": 14416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0020688415998217e-05,
+      "loss": 0.6324,
+      "step": 14417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0019599553431308e-05,
+      "loss": 0.6472,
+      "step": 14418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0018510690632017e-05,
+      "loss": 0.6632,
+      "step": 14419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.001742182761326e-05,
+      "loss": 0.6385,
+      "step": 14420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0016332964387948e-05,
+      "loss": 0.6636,
+      "step": 14421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0015244100968986e-05,
+      "loss": 0.6804,
+      "step": 14422
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0014155237369283e-05,
+      "loss": 0.6259,
+      "step": 14423
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0013066373601757e-05,
+      "loss": 0.6352,
+      "step": 14424
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0011977509679312e-05,
+      "loss": 0.5801,
+      "step": 14425
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0010888645614853e-05,
+      "loss": 0.6359,
+      "step": 14426
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0009799781421302e-05,
+      "loss": 0.6629,
+      "step": 14427
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.000871091711156e-05,
+      "loss": 0.5948,
+      "step": 14428
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0007622052698533e-05,
+      "loss": 0.673,
+      "step": 14429
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0006533188195144e-05,
+      "loss": 0.6555,
+      "step": 14430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0005444323614295e-05,
+      "loss": 0.6262,
+      "step": 14431
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0004355458968892e-05,
+      "loss": 0.5915,
+      "step": 14432
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0003266594271856e-05,
+      "loss": 0.5959,
+      "step": 14433
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0002177729536089e-05,
+      "loss": 0.6094,
+      "step": 14434
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0001088864774498e-05,
+      "loss": 0.6435,
+      "step": 14435
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1e-05,
+      "loss": 0.5964,
+      "step": 14436
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.998911135225503e-06,
+      "loss": 0.6196,
+      "step": 14437
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.997822270463916e-06,
+      "loss": 0.6596,
+      "step": 14438
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.996733405728145e-06,
+      "loss": 0.6125,
+      "step": 14439
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.995644541031106e-06,
+      "loss": 0.6454,
+      "step": 14440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.994555676385708e-06,
+      "loss": 0.5861,
+      "step": 14441
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.993466811804858e-06,
+      "loss": 0.6523,
+      "step": 14442
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.992377947301467e-06,
+      "loss": 0.6652,
+      "step": 14443
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.991289082888447e-06,
+      "loss": 0.6316,
+      "step": 14444
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.990200218578703e-06,
+      "loss": 0.6482,
+      "step": 14445
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.989111354385149e-06,
+      "loss": 0.6018,
+      "step": 14446
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.988022490320695e-06,
+      "loss": 0.6747,
+      "step": 14447
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.986933626398247e-06,
+      "loss": 0.6602,
+      "step": 14448
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.985844762630717e-06,
+      "loss": 0.5925,
+      "step": 14449
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.984755899031019e-06,
+      "loss": 0.5598,
+      "step": 14450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.983667035612056e-06,
+      "loss": 0.6174,
+      "step": 14451
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.98257817238674e-06,
+      "loss": 0.674,
+      "step": 14452
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.981489309367987e-06,
+      "loss": 0.6325,
+      "step": 14453
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.980400446568696e-06,
+      "loss": 0.6354,
+      "step": 14454
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.979311584001785e-06,
+      "loss": 0.6064,
+      "step": 14455
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.978222721680163e-06,
+      "loss": 0.6084,
+      "step": 14456
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.977133859616735e-06,
+      "loss": 0.6178,
+      "step": 14457
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.976044997824415e-06,
+      "loss": 0.5926,
+      "step": 14458
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.974956136316112e-06,
+      "loss": 0.6833,
+      "step": 14459
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.973867275104737e-06,
+      "loss": 0.5931,
+      "step": 14460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.972778414203196e-06,
+      "loss": 0.6011,
+      "step": 14461
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.971689553624403e-06,
+      "loss": 0.6079,
+      "step": 14462
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.970600693381266e-06,
+      "loss": 0.6535,
+      "step": 14463
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.969511833486695e-06,
+      "loss": 0.663,
+      "step": 14464
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.9684229739536e-06,
+      "loss": 0.6005,
+      "step": 14465
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.96733411479489e-06,
+      "loss": 0.668,
+      "step": 14466
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.966245256023475e-06,
+      "loss": 0.6567,
+      "step": 14467
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.965156397652267e-06,
+      "loss": 0.6447,
+      "step": 14468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.964067539694172e-06,
+      "loss": 0.6386,
+      "step": 14469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.962978682162105e-06,
+      "loss": 0.663,
+      "step": 14470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.961889825068968e-06,
+      "loss": 0.6387,
+      "step": 14471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.960800968427678e-06,
+      "loss": 0.6515,
+      "step": 14472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.959712112251142e-06,
+      "loss": 0.6732,
+      "step": 14473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.95862325655227e-06,
+      "loss": 0.6081,
+      "step": 14474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.95753440134397e-06,
+      "loss": 0.6725,
+      "step": 14475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.956445546639155e-06,
+      "loss": 0.6721,
+      "step": 14476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.955356692450733e-06,
+      "loss": 0.599,
+      "step": 14477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.954267838791612e-06,
+      "loss": 0.6345,
+      "step": 14478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.953178985674707e-06,
+      "loss": 0.6217,
+      "step": 14479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.952090133112922e-06,
+      "loss": 0.6304,
+      "step": 14480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.951001281119167e-06,
+      "loss": 0.6613,
+      "step": 14481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.949912429706359e-06,
+      "loss": 0.6128,
+      "step": 14482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.948823578887398e-06,
+      "loss": 0.5912,
+      "step": 14483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.9477347286752e-06,
+      "loss": 0.6151,
+      "step": 14484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.946645879082675e-06,
+      "loss": 0.64,
+      "step": 14485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.945557030122729e-06,
+      "loss": 0.6723,
+      "step": 14486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.944468181808272e-06,
+      "loss": 0.625,
+      "step": 14487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.943379334152217e-06,
+      "loss": 0.5778,
+      "step": 14488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.942290487167468e-06,
+      "loss": 0.6405,
+      "step": 14489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.94120164086694e-06,
+      "loss": 0.6057,
+      "step": 14490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.940112795263543e-06,
+      "loss": 0.6302,
+      "step": 14491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.939023950370183e-06,
+      "loss": 0.5888,
+      "step": 14492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.93793510619977e-06,
+      "loss": 0.6217,
+      "step": 14493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.936846262765218e-06,
+      "loss": 0.6609,
+      "step": 14494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.935757420079432e-06,
+      "loss": 0.6587,
+      "step": 14495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.934668578155322e-06,
+      "loss": 0.6141,
+      "step": 14496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.9335797370058e-06,
+      "loss": 0.5736,
+      "step": 14497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.932490896643773e-06,
+      "loss": 0.6504,
+      "step": 14498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.931402057082152e-06,
+      "loss": 0.6797,
+      "step": 14499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.930313218333848e-06,
+      "loss": 0.635,
+      "step": 14500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.929224380411767e-06,
+      "loss": 0.6465,
+      "step": 14501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.928135543328821e-06,
+      "loss": 0.5917,
+      "step": 14502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.92704670709792e-06,
+      "loss": 0.6649,
+      "step": 14503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.92595787173197e-06,
+      "loss": 0.648,
+      "step": 14504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.924869037243887e-06,
+      "loss": 0.6347,
+      "step": 14505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.923780203646575e-06,
+      "loss": 0.597,
+      "step": 14506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.922691370952944e-06,
+      "loss": 0.6365,
+      "step": 14507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.921602539175906e-06,
+      "loss": 0.5844,
+      "step": 14508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.920513708328368e-06,
+      "loss": 0.6469,
+      "step": 14509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.91942487842324e-06,
+      "loss": 0.6118,
+      "step": 14510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.918336049473432e-06,
+      "loss": 0.5715,
+      "step": 14511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.917247221491857e-06,
+      "loss": 0.6116,
+      "step": 14512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.916158394491418e-06,
+      "loss": 0.5763,
+      "step": 14513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.915069568485027e-06,
+      "loss": 0.6082,
+      "step": 14514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.913980743485594e-06,
+      "loss": 0.602,
+      "step": 14515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.912891919506028e-06,
+      "loss": 0.5904,
+      "step": 14516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.911803096559239e-06,
+      "loss": 0.6552,
+      "step": 14517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.910714274658136e-06,
+      "loss": 0.6089,
+      "step": 14518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.909625453815627e-06,
+      "loss": 0.6539,
+      "step": 14519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.908536634044624e-06,
+      "loss": 0.5691,
+      "step": 14520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.907447815358035e-06,
+      "loss": 0.6198,
+      "step": 14521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.906358997768768e-06,
+      "loss": 0.6596,
+      "step": 14522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.905270181289733e-06,
+      "loss": 0.6031,
+      "step": 14523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.904181365933843e-06,
+      "loss": 0.6013,
+      "step": 14524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.903092551714e-06,
+      "loss": 0.629,
+      "step": 14525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.90200373864312e-06,
+      "loss": 0.6538,
+      "step": 14526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.90091492673411e-06,
+      "loss": 0.6503,
+      "step": 14527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.899826115999877e-06,
+      "loss": 0.6162,
+      "step": 14528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.898737306453334e-06,
+      "loss": 0.6299,
+      "step": 14529
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.897648498107387e-06,
+      "loss": 0.6852,
+      "step": 14530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.896559690974947e-06,
+      "loss": 0.6092,
+      "step": 14531
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.895470885068926e-06,
+      "loss": 0.5865,
+      "step": 14532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.894382080402226e-06,
+      "loss": 0.6395,
+      "step": 14533
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.893293276987758e-06,
+      "loss": 0.5891,
+      "step": 14534
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.89220447483844e-06,
+      "loss": 0.6695,
+      "step": 14535
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.891115673967169e-06,
+      "loss": 0.6624,
+      "step": 14536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.890026874386856e-06,
+      "loss": 0.577,
+      "step": 14537
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.888938076110423e-06,
+      "loss": 0.661,
+      "step": 14538
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.887849279150764e-06,
+      "loss": 0.5913,
+      "step": 14539
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.886760483520791e-06,
+      "loss": 0.6459,
+      "step": 14540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.885671689233423e-06,
+      "loss": 0.6933,
+      "step": 14541
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.884582896301556e-06,
+      "loss": 0.6602,
+      "step": 14542
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.883494104738105e-06,
+      "loss": 0.6041,
+      "step": 14543
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.882405314555984e-06,
+      "loss": 0.6398,
+      "step": 14544
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.881316525768092e-06,
+      "loss": 0.6133,
+      "step": 14545
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.880227738387341e-06,
+      "loss": 0.6455,
+      "step": 14546
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.879138952426649e-06,
+      "loss": 0.6153,
+      "step": 14547
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.878050167898911e-06,
+      "loss": 0.6437,
+      "step": 14548
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.876961384817041e-06,
+      "loss": 0.6763,
+      "step": 14549
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.875872603193957e-06,
+      "loss": 0.6521,
+      "step": 14550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.874783823042556e-06,
+      "loss": 0.565,
+      "step": 14551
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.873695044375749e-06,
+      "loss": 0.6458,
+      "step": 14552
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.872606267206453e-06,
+      "loss": 0.6755,
+      "step": 14553
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.871517491547566e-06,
+      "loss": 0.6366,
+      "step": 14554
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.870428717412001e-06,
+      "loss": 0.5852,
+      "step": 14555
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.869339944812673e-06,
+      "loss": 0.6154,
+      "step": 14556
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.868251173762481e-06,
+      "loss": 0.61,
+      "step": 14557
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.867162404274337e-06,
+      "loss": 0.6795,
+      "step": 14558
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.866073636361155e-06,
+      "loss": 0.6026,
+      "step": 14559
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.864984870035837e-06,
+      "loss": 0.6724,
+      "step": 14560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.86389610531129e-06,
+      "loss": 0.6225,
+      "step": 14561
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.862807342200436e-06,
+      "loss": 0.6769,
+      "step": 14562
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.861718580716168e-06,
+      "loss": 0.611,
+      "step": 14563
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.860629820871399e-06,
+      "loss": 0.6158,
+      "step": 14564
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.859541062679047e-06,
+      "loss": 0.6637,
+      "step": 14565
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.85845230615201e-06,
+      "loss": 0.6854,
+      "step": 14566
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.857363551303197e-06,
+      "loss": 0.6237,
+      "step": 14567
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.856274798145526e-06,
+      "loss": 0.6432,
+      "step": 14568
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.855186046691895e-06,
+      "loss": 0.67,
+      "step": 14569
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.854097296955212e-06,
+      "loss": 0.6384,
+      "step": 14570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.8530085489484e-06,
+      "loss": 0.6221,
+      "step": 14571
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.851919802684352e-06,
+      "loss": 0.6629,
+      "step": 14572
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.85083105817598e-06,
+      "loss": 0.6443,
+      "step": 14573
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.849742315436201e-06,
+      "loss": 0.6317,
+      "step": 14574
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.848653574477912e-06,
+      "loss": 0.6129,
+      "step": 14575
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.847564835314025e-06,
+      "loss": 0.6149,
+      "step": 14576
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.846476097957455e-06,
+      "loss": 0.6387,
+      "step": 14577
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.845387362421102e-06,
+      "loss": 0.6517,
+      "step": 14578
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.844298628717875e-06,
+      "loss": 0.6128,
+      "step": 14579
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.843209896860691e-06,
+      "loss": 0.6061,
+      "step": 14580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.842121166862449e-06,
+      "loss": 0.685,
+      "step": 14581
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.841032438736055e-06,
+      "loss": 0.585,
+      "step": 14582
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.839943712494432e-06,
+      "loss": 0.6597,
+      "step": 14583
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.838854988150472e-06,
+      "loss": 0.6667,
+      "step": 14584
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.83776626571709e-06,
+      "loss": 0.6645,
+      "step": 14585
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.836677545207199e-06,
+      "loss": 0.6191,
+      "step": 14586
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.8355888266337e-06,
+      "loss": 0.622,
+      "step": 14587
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.834500110009499e-06,
+      "loss": 0.6157,
+      "step": 14588
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.833411395347517e-06,
+      "loss": 0.659,
+      "step": 14589
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.832322682660648e-06,
+      "loss": 0.6266,
+      "step": 14590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.831233971961803e-06,
+      "loss": 0.6417,
+      "step": 14591
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.8301452632639e-06,
+      "loss": 0.6175,
+      "step": 14592
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.829056556579837e-06,
+      "loss": 0.6736,
+      "step": 14593
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.827967851922526e-06,
+      "loss": 0.639,
+      "step": 14594
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.82687914930487e-06,
+      "loss": 0.5941,
+      "step": 14595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.825790448739781e-06,
+      "loss": 0.6258,
+      "step": 14596
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.82470175024017e-06,
+      "loss": 0.6018,
+      "step": 14597
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.82361305381894e-06,
+      "loss": 0.5657,
+      "step": 14598
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.822524359489e-06,
+      "loss": 0.6397,
+      "step": 14599
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.821435667263258e-06,
+      "loss": 0.6593,
+      "step": 14600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.820346977154622e-06,
+      "loss": 0.6024,
+      "step": 14601
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.819258289175999e-06,
+      "loss": 0.6571,
+      "step": 14602
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.8181696033403e-06,
+      "loss": 0.6387,
+      "step": 14603
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.817080919660429e-06,
+      "loss": 0.6102,
+      "step": 14604
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.815992238149295e-06,
+      "loss": 0.6266,
+      "step": 14605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.814903558819805e-06,
+      "loss": 0.649,
+      "step": 14606
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.81381488168487e-06,
+      "loss": 0.6409,
+      "step": 14607
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.812726206757393e-06,
+      "loss": 0.6208,
+      "step": 14608
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.811637534050284e-06,
+      "loss": 0.6322,
+      "step": 14609
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.81054886357645e-06,
+      "loss": 0.6205,
+      "step": 14610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.809460195348798e-06,
+      "loss": 0.6741,
+      "step": 14611
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.80837152938024e-06,
+      "loss": 0.6011,
+      "step": 14612
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.807282865683677e-06,
+      "loss": 0.6691,
+      "step": 14613
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.80619420427202e-06,
+      "loss": 0.6331,
+      "step": 14614
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.805105545158178e-06,
+      "loss": 0.6311,
+      "step": 14615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.804016888355053e-06,
+      "loss": 0.6648,
+      "step": 14616
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.802928233875557e-06,
+      "loss": 0.6106,
+      "step": 14617
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.801839581732597e-06,
+      "loss": 0.6252,
+      "step": 14618
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.800750931939079e-06,
+      "loss": 0.6165,
+      "step": 14619
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.79966228450791e-06,
+      "loss": 0.6671,
+      "step": 14620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.798573639452e-06,
+      "loss": 0.6481,
+      "step": 14621
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.797484996784251e-06,
+      "loss": 0.6656,
+      "step": 14622
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.796396356517574e-06,
+      "loss": 0.5814,
+      "step": 14623
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.795307718664879e-06,
+      "loss": 0.6701,
+      "step": 14624
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.794219083239067e-06,
+      "loss": 0.6142,
+      "step": 14625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.793130450253048e-06,
+      "loss": 0.6327,
+      "step": 14626
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.792041819719732e-06,
+      "loss": 0.6063,
+      "step": 14627
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.79095319165202e-06,
+      "loss": 0.6405,
+      "step": 14628
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.789864566062824e-06,
+      "loss": 0.6929,
+      "step": 14629
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.788775942965051e-06,
+      "loss": 0.6551,
+      "step": 14630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.787687322371604e-06,
+      "loss": 0.6485,
+      "step": 14631
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.786598704295392e-06,
+      "loss": 0.5886,
+      "step": 14632
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.785510088749324e-06,
+      "loss": 0.6315,
+      "step": 14633
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.784421475746305e-06,
+      "loss": 0.6886,
+      "step": 14634
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.78333286529924e-06,
+      "loss": 0.636,
+      "step": 14635
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.782244257421042e-06,
+      "loss": 0.6388,
+      "step": 14636
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.78115565212461e-06,
+      "loss": 0.5824,
+      "step": 14637
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.780067049422856e-06,
+      "loss": 0.603,
+      "step": 14638
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.778978449328687e-06,
+      "loss": 0.6087,
+      "step": 14639
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.777889851855006e-06,
+      "loss": 0.5591,
+      "step": 14640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.776801257014721e-06,
+      "loss": 0.6743,
+      "step": 14641
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.775712664820742e-06,
+      "loss": 0.6898,
+      "step": 14642
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.77462407528597e-06,
+      "loss": 0.6795,
+      "step": 14643
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.773535488423317e-06,
+      "loss": 0.6879,
+      "step": 14644
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.772446904245688e-06,
+      "loss": 0.6423,
+      "step": 14645
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.771358322765987e-06,
+      "loss": 0.6055,
+      "step": 14646
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.770269743997122e-06,
+      "loss": 0.6475,
+      "step": 14647
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.769181167952003e-06,
+      "loss": 0.6273,
+      "step": 14648
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.76809259464353e-06,
+      "loss": 0.6295,
+      "step": 14649
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.767004024084612e-06,
+      "loss": 0.659,
+      "step": 14650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.76591545628816e-06,
+      "loss": 0.6461,
+      "step": 14651
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.764826891267072e-06,
+      "loss": 0.6297,
+      "step": 14652
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.76373832903426e-06,
+      "loss": 0.6196,
+      "step": 14653
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.76264976960263e-06,
+      "loss": 0.6212,
+      "step": 14654
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.761561212985088e-06,
+      "loss": 0.7113,
+      "step": 14655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.76047265919454e-06,
+      "loss": 0.6363,
+      "step": 14656
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.759384108243888e-06,
+      "loss": 0.7063,
+      "step": 14657
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.758295560146043e-06,
+      "loss": 0.5932,
+      "step": 14658
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.757207014913914e-06,
+      "loss": 0.6946,
+      "step": 14659
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.756118472560397e-06,
+      "loss": 0.6272,
+      "step": 14660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.755029933098406e-06,
+      "loss": 0.6346,
+      "step": 14661
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.753941396540848e-06,
+      "loss": 0.6384,
+      "step": 14662
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.752852862900622e-06,
+      "loss": 0.6941,
+      "step": 14663
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.75176433219064e-06,
+      "loss": 0.6138,
+      "step": 14664
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.750675804423807e-06,
+      "loss": 0.6552,
+      "step": 14665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.749587279613025e-06,
+      "loss": 0.6726,
+      "step": 14666
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.748498757771204e-06,
+      "loss": 0.5734,
+      "step": 14667
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.747410238911248e-06,
+      "loss": 0.6662,
+      "step": 14668
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.746321723046065e-06,
+      "loss": 0.619,
+      "step": 14669
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.745233210188556e-06,
+      "loss": 0.6619,
+      "step": 14670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.744144700351631e-06,
+      "loss": 0.6019,
+      "step": 14671
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.743056193548194e-06,
+      "loss": 0.6445,
+      "step": 14672
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.74196768979115e-06,
+      "loss": 0.6211,
+      "step": 14673
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.740879189093407e-06,
+      "loss": 0.6269,
+      "step": 14674
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.739790691467868e-06,
+      "loss": 0.6688,
+      "step": 14675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.738702196927438e-06,
+      "loss": 0.6497,
+      "step": 14676
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.737613705485028e-06,
+      "loss": 0.6703,
+      "step": 14677
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.736525217153538e-06,
+      "loss": 0.702,
+      "step": 14678
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.735436731945873e-06,
+      "loss": 0.5889,
+      "step": 14679
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.734348249874942e-06,
+      "loss": 0.6538,
+      "step": 14680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.733259770953647e-06,
+      "loss": 0.656,
+      "step": 14681
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.732171295194896e-06,
+      "loss": 0.6302,
+      "step": 14682
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.731082822611595e-06,
+      "loss": 0.6782,
+      "step": 14683
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.729994353216642e-06,
+      "loss": 0.6403,
+      "step": 14684
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.728905887022951e-06,
+      "loss": 0.6689,
+      "step": 14685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.727817424043425e-06,
+      "loss": 0.5857,
+      "step": 14686
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.726728964290965e-06,
+      "loss": 0.6475,
+      "step": 14687
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.72564050777848e-06,
+      "loss": 0.6117,
+      "step": 14688
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.724552054518875e-06,
+      "loss": 0.5994,
+      "step": 14689
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.723463604525051e-06,
+      "loss": 0.6295,
+      "step": 14690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.722375157809918e-06,
+      "loss": 0.5951,
+      "step": 14691
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.72128671438638e-06,
+      "loss": 0.65,
+      "step": 14692
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.720198274267335e-06,
+      "loss": 0.622,
+      "step": 14693
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.719109837465698e-06,
+      "loss": 0.6435,
+      "step": 14694
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.718021403994368e-06,
+      "loss": 0.6224,
+      "step": 14695
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.71693297386625e-06,
+      "loss": 0.6418,
+      "step": 14696
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.715844547094252e-06,
+      "loss": 0.6183,
+      "step": 14697
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.714756123691276e-06,
+      "loss": 0.6598,
+      "step": 14698
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.713667703670225e-06,
+      "loss": 0.6544,
+      "step": 14699
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.712579287044008e-06,
+      "loss": 0.5403,
+      "step": 14700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.711490873825527e-06,
+      "loss": 0.6384,
+      "step": 14701
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.710402464027684e-06,
+      "loss": 0.6541,
+      "step": 14702
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.709314057663388e-06,
+      "loss": 0.663,
+      "step": 14703
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.708225654745543e-06,
+      "loss": 0.6576,
+      "step": 14704
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.707137255287048e-06,
+      "loss": 0.6134,
+      "step": 14705
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.706048859300813e-06,
+      "loss": 0.6992,
+      "step": 14706
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.704960466799744e-06,
+      "loss": 0.6695,
+      "step": 14707
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.703872077796738e-06,
+      "loss": 0.6668,
+      "step": 14708
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.702783692304702e-06,
+      "loss": 0.6516,
+      "step": 14709
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.701695310336545e-06,
+      "loss": 0.6339,
+      "step": 14710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.700606931905165e-06,
+      "loss": 0.6413,
+      "step": 14711
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.699518557023469e-06,
+      "loss": 0.66,
+      "step": 14712
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.698430185704365e-06,
+      "loss": 0.6512,
+      "step": 14713
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.697341817960745e-06,
+      "loss": 0.6733,
+      "step": 14714
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.696253453805525e-06,
+      "loss": 0.6065,
+      "step": 14715
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.695165093251605e-06,
+      "loss": 0.6262,
+      "step": 14716
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.694076736311887e-06,
+      "loss": 0.6666,
+      "step": 14717
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.692988382999278e-06,
+      "loss": 0.645,
+      "step": 14718
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.691900033326679e-06,
+      "loss": 0.645,
+      "step": 14719
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.690811687306993e-06,
+      "loss": 0.618,
+      "step": 14720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.689723344953129e-06,
+      "loss": 0.6263,
+      "step": 14721
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.688635006277984e-06,
+      "loss": 0.611,
+      "step": 14722
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.687546671294465e-06,
+      "loss": 0.5928,
+      "step": 14723
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.686458340015479e-06,
+      "loss": 0.6181,
+      "step": 14724
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.685370012453922e-06,
+      "loss": 0.6287,
+      "step": 14725
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.684281688622705e-06,
+      "loss": 0.6137,
+      "step": 14726
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.68319336853473e-06,
+      "loss": 0.6363,
+      "step": 14727
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.68210505220289e-06,
+      "loss": 0.5927,
+      "step": 14728
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.681016739640101e-06,
+      "loss": 0.6603,
+      "step": 14729
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.679928430859267e-06,
+      "loss": 0.63,
+      "step": 14730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.678840125873278e-06,
+      "loss": 0.6509,
+      "step": 14731
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.677751824695049e-06,
+      "loss": 0.6565,
+      "step": 14732
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.676663527337484e-06,
+      "loss": 0.6327,
+      "step": 14733
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.675575233813475e-06,
+      "loss": 0.6069,
+      "step": 14734
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.674486944135934e-06,
+      "loss": 0.616,
+      "step": 14735
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.673398658317767e-06,
+      "loss": 0.679,
+      "step": 14736
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.672310376371864e-06,
+      "loss": 0.6205,
+      "step": 14737
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.67122209831114e-06,
+      "loss": 0.6213,
+      "step": 14738
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.670133824148498e-06,
+      "loss": 0.6688,
+      "step": 14739
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.66904555389683e-06,
+      "loss": 0.6048,
+      "step": 14740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.667957287569048e-06,
+      "loss": 0.6012,
+      "step": 14741
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.666869025178057e-06,
+      "loss": 0.6662,
+      "step": 14742
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.665780766736748e-06,
+      "loss": 0.6301,
+      "step": 14743
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.664692512258032e-06,
+      "loss": 0.6211,
+      "step": 14744
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.663604261754817e-06,
+      "loss": 0.6781,
+      "step": 14745
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.662516015239991e-06,
+      "loss": 0.6432,
+      "step": 14746
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.661427772726468e-06,
+      "loss": 0.6354,
+      "step": 14747
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.66033953422715e-06,
+      "loss": 0.6279,
+      "step": 14748
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.65925129975493e-06,
+      "loss": 0.7499,
+      "step": 14749
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.658163069322719e-06,
+      "loss": 0.621,
+      "step": 14750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.657074842943423e-06,
+      "loss": 0.6323,
+      "step": 14751
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.655986620629931e-06,
+      "loss": 0.5504,
+      "step": 14752
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.654898402395156e-06,
+      "loss": 0.6585,
+      "step": 14753
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.653810188252e-06,
+      "loss": 0.6097,
+      "step": 14754
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.652721978213356e-06,
+      "loss": 0.6675,
+      "step": 14755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.651633772292133e-06,
+      "loss": 0.654,
+      "step": 14756
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.65054557050124e-06,
+      "loss": 0.6111,
+      "step": 14757
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.649457372853562e-06,
+      "loss": 0.6188,
+      "step": 14758
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.648369179362014e-06,
+      "loss": 0.6271,
+      "step": 14759
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.647280990039499e-06,
+      "loss": 0.6378,
+      "step": 14760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.646192804898905e-06,
+      "loss": 0.6642,
+      "step": 14761
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.645104623953147e-06,
+      "loss": 0.6246,
+      "step": 14762
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.644016447215127e-06,
+      "loss": 0.6539,
+      "step": 14763
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.642928274697735e-06,
+      "loss": 0.6239,
+      "step": 14764
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.641840106413882e-06,
+      "loss": 0.6315,
+      "step": 14765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.640751942376474e-06,
+      "loss": 0.5964,
+      "step": 14766
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.639663782598398e-06,
+      "loss": 0.6681,
+      "step": 14767
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.638575627092565e-06,
+      "loss": 0.6615,
+      "step": 14768
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.63748747587188e-06,
+      "loss": 0.6569,
+      "step": 14769
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.636399328949235e-06,
+      "loss": 0.6536,
+      "step": 14770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.635311186337535e-06,
+      "loss": 0.6164,
+      "step": 14771
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.634223048049688e-06,
+      "loss": 0.6199,
+      "step": 14772
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.633134914098584e-06,
+      "loss": 0.6165,
+      "step": 14773
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.632046784497131e-06,
+      "loss": 0.5698,
+      "step": 14774
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.630958659258233e-06,
+      "loss": 0.647,
+      "step": 14775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.62987053839478e-06,
+      "loss": 0.6432,
+      "step": 14776
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.628782421919683e-06,
+      "loss": 0.6597,
+      "step": 14777
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.627694309845844e-06,
+      "loss": 0.6257,
+      "step": 14778
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.626606202186155e-06,
+      "loss": 0.6414,
+      "step": 14779
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.625518098953525e-06,
+      "loss": 0.6715,
+      "step": 14780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.62443000016085e-06,
+      "loss": 0.6182,
+      "step": 14781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.62334190582103e-06,
+      "loss": 0.6305,
+      "step": 14782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.622253815946974e-06,
+      "loss": 0.6659,
+      "step": 14783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.621165730551571e-06,
+      "loss": 0.6019,
+      "step": 14784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.620077649647728e-06,
+      "loss": 0.6715,
+      "step": 14785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.61898957324835e-06,
+      "loss": 0.6827,
+      "step": 14786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.617901501366329e-06,
+      "loss": 0.6267,
+      "step": 14787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.616813434014567e-06,
+      "loss": 0.6514,
+      "step": 14788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.615725371205972e-06,
+      "loss": 0.6904,
+      "step": 14789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.614637312953435e-06,
+      "loss": 0.6222,
+      "step": 14790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.61354925926986e-06,
+      "loss": 0.6448,
+      "step": 14791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.61246121016815e-06,
+      "loss": 0.5713,
+      "step": 14792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.611373165661201e-06,
+      "loss": 0.5277,
+      "step": 14793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.610285125761911e-06,
+      "loss": 0.6271,
+      "step": 14794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.609197090483193e-06,
+      "loss": 0.6609,
+      "step": 14795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.608109059837931e-06,
+      "loss": 0.6346,
+      "step": 14796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.60702103383903e-06,
+      "loss": 0.6351,
+      "step": 14797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.605933012499398e-06,
+      "loss": 0.6518,
+      "step": 14798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.604844995831925e-06,
+      "loss": 0.6238,
+      "step": 14799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.603756983849513e-06,
+      "loss": 0.6296,
+      "step": 14800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.602668976565069e-06,
+      "loss": 0.683,
+      "step": 14801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.601580973991482e-06,
+      "loss": 0.6815,
+      "step": 14802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.600492976141655e-06,
+      "loss": 0.6441,
+      "step": 14803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.599404983028496e-06,
+      "loss": 0.6384,
+      "step": 14804
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.598316994664893e-06,
+      "loss": 0.6328,
+      "step": 14805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.597229011063747e-06,
+      "loss": 0.6581,
+      "step": 14806
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.596141032237969e-06,
+      "loss": 0.6856,
+      "step": 14807
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.595053058200444e-06,
+      "loss": 0.6334,
+      "step": 14808
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.593965088964076e-06,
+      "loss": 0.5781,
+      "step": 14809
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.592877124541771e-06,
+      "loss": 0.6416,
+      "step": 14810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.591789164946419e-06,
+      "loss": 0.6827,
+      "step": 14811
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.590701210190922e-06,
+      "loss": 0.6743,
+      "step": 14812
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.589613260288184e-06,
+      "loss": 0.6292,
+      "step": 14813
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.588525315251097e-06,
+      "loss": 0.6116,
+      "step": 14814
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.58743737509256e-06,
+      "loss": 0.6558,
+      "step": 14815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.586349439825482e-06,
+      "loss": 0.6751,
+      "step": 14816
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.58526150946275e-06,
+      "loss": 0.6484,
+      "step": 14817
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.584173584017265e-06,
+      "loss": 0.5941,
+      "step": 14818
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.583085663501935e-06,
+      "loss": 0.5996,
+      "step": 14819
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.58199774792965e-06,
+      "loss": 0.5826,
+      "step": 14820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.580909837313305e-06,
+      "loss": 0.594,
+      "step": 14821
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.57982193166581e-06,
+      "loss": 0.6444,
+      "step": 14822
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.578734031000056e-06,
+      "loss": 0.6181,
+      "step": 14823
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.577646135328938e-06,
+      "loss": 0.5792,
+      "step": 14824
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.576558244665368e-06,
+      "loss": 0.6334,
+      "step": 14825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.575470359022231e-06,
+      "loss": 0.6446,
+      "step": 14826
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.574382478412427e-06,
+      "loss": 0.5876,
+      "step": 14827
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.573294602848863e-06,
+      "loss": 0.6486,
+      "step": 14828
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.57220673234443e-06,
+      "loss": 0.7033,
+      "step": 14829
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.571118866912023e-06,
+      "loss": 0.6477,
+      "step": 14830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.570031006564549e-06,
+      "loss": 0.6406,
+      "step": 14831
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.5689431513149e-06,
+      "loss": 0.6861,
+      "step": 14832
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.567855301175972e-06,
+      "loss": 0.6367,
+      "step": 14833
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.566767456160672e-06,
+      "loss": 0.6114,
+      "step": 14834
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.565679616281889e-06,
+      "loss": 0.6637,
+      "step": 14835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.56459178155252e-06,
+      "loss": 0.6698,
+      "step": 14836
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.563503951985473e-06,
+      "loss": 0.6353,
+      "step": 14837
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.562416127593635e-06,
+      "loss": 0.6059,
+      "step": 14838
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.561328308389906e-06,
+      "loss": 0.6374,
+      "step": 14839
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.56024049438719e-06,
+      "loss": 0.5979,
+      "step": 14840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.559152685598375e-06,
+      "loss": 0.6532,
+      "step": 14841
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.558064882036366e-06,
+      "loss": 0.6308,
+      "step": 14842
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.556977083714052e-06,
+      "loss": 0.6521,
+      "step": 14843
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.555889290644337e-06,
+      "loss": 0.6192,
+      "step": 14844
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.554801502840118e-06,
+      "loss": 0.6273,
+      "step": 14845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.553713720314288e-06,
+      "loss": 0.5594,
+      "step": 14846
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.552625943079747e-06,
+      "loss": 0.6173,
+      "step": 14847
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.551538171149394e-06,
+      "loss": 0.6583,
+      "step": 14848
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.55045040453612e-06,
+      "loss": 0.6784,
+      "step": 14849
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.549362643252825e-06,
+      "loss": 0.5996,
+      "step": 14850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.54827488731241e-06,
+      "loss": 0.5794,
+      "step": 14851
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.547187136727764e-06,
+      "loss": 0.6288,
+      "step": 14852
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.546099391511787e-06,
+      "loss": 0.5801,
+      "step": 14853
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.545011651677379e-06,
+      "loss": 0.6564,
+      "step": 14854
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.54392391723743e-06,
+      "loss": 0.6251,
+      "step": 14855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.54283618820484e-06,
+      "loss": 0.6109,
+      "step": 14856
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.541748464592508e-06,
+      "loss": 0.6153,
+      "step": 14857
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.540660746413326e-06,
+      "loss": 0.6239,
+      "step": 14858
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.539573033680192e-06,
+      "loss": 0.6303,
+      "step": 14859
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.538485326406004e-06,
+      "loss": 0.6767,
+      "step": 14860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.537397624603653e-06,
+      "loss": 0.659,
+      "step": 14861
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.53630992828604e-06,
+      "loss": 0.6791,
+      "step": 14862
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.535222237466061e-06,
+      "loss": 0.6056,
+      "step": 14863
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.534134552156608e-06,
+      "loss": 0.6752,
+      "step": 14864
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.53304687237058e-06,
+      "loss": 0.5812,
+      "step": 14865
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.531959198120873e-06,
+      "loss": 0.6399,
+      "step": 14866
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.530871529420379e-06,
+      "loss": 0.6038,
+      "step": 14867
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.529783866281998e-06,
+      "loss": 0.6259,
+      "step": 14868
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.528696208718627e-06,
+      "loss": 0.6321,
+      "step": 14869
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.527608556743154e-06,
+      "loss": 0.6649,
+      "step": 14870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.52652091036848e-06,
+      "loss": 0.6529,
+      "step": 14871
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.525433269607503e-06,
+      "loss": 0.6187,
+      "step": 14872
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.524345634473112e-06,
+      "loss": 0.6715,
+      "step": 14873
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.523258004978205e-06,
+      "loss": 0.6667,
+      "step": 14874
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.52217038113568e-06,
+      "loss": 0.6181,
+      "step": 14875
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.521082762958426e-06,
+      "loss": 0.6275,
+      "step": 14876
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.519995150459342e-06,
+      "loss": 0.6031,
+      "step": 14877
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.518907543651326e-06,
+      "loss": 0.6156,
+      "step": 14878
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.517819942547267e-06,
+      "loss": 0.5958,
+      "step": 14879
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.516732347160063e-06,
+      "loss": 0.6437,
+      "step": 14880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.51564475750261e-06,
+      "loss": 0.5937,
+      "step": 14881
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.5145571735878e-06,
+      "loss": 0.6385,
+      "step": 14882
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.513469595428527e-06,
+      "loss": 0.6138,
+      "step": 14883
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.51238202303769e-06,
+      "loss": 0.6594,
+      "step": 14884
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.51129445642818e-06,
+      "loss": 0.6227,
+      "step": 14885
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.510206895612891e-06,
+      "loss": 0.6561,
+      "step": 14886
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.509119340604722e-06,
+      "loss": 0.635,
+      "step": 14887
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.508031791416562e-06,
+      "loss": 0.6101,
+      "step": 14888
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.506944248061307e-06,
+      "loss": 0.6492,
+      "step": 14889
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.505856710551854e-06,
+      "loss": 0.6411,
+      "step": 14890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.504769178901093e-06,
+      "loss": 0.6841,
+      "step": 14891
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.503681653121918e-06,
+      "loss": 0.5888,
+      "step": 14892
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.50259413322723e-06,
+      "loss": 0.6564,
+      "step": 14893
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.501506619229914e-06,
+      "loss": 0.6485,
+      "step": 14894
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.500419111142867e-06,
+      "loss": 0.6504,
+      "step": 14895
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.499331608978986e-06,
+      "loss": 0.624,
+      "step": 14896
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.498244112751161e-06,
+      "loss": 0.6401,
+      "step": 14897
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.497156622472286e-06,
+      "loss": 0.6248,
+      "step": 14898
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.496069138155259e-06,
+      "loss": 0.6508,
+      "step": 14899
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.494981659812965e-06,
+      "loss": 0.5966,
+      "step": 14900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.493894187458304e-06,
+      "loss": 0.6268,
+      "step": 14901
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.492806721104169e-06,
+      "loss": 0.7078,
+      "step": 14902
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.49171926076345e-06,
+      "loss": 0.6654,
+      "step": 14903
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.490631806449044e-06,
+      "loss": 0.5936,
+      "step": 14904
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.489544358173841e-06,
+      "loss": 0.6486,
+      "step": 14905
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.488456915950735e-06,
+      "loss": 0.5852,
+      "step": 14906
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.487369479792621e-06,
+      "loss": 0.6001,
+      "step": 14907
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.486282049712389e-06,
+      "loss": 0.6642,
+      "step": 14908
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.485194625722932e-06,
+      "loss": 0.5982,
+      "step": 14909
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.484107207837147e-06,
+      "loss": 0.6498,
+      "step": 14910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.483019796067921e-06,
+      "loss": 0.6405,
+      "step": 14911
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.48193239042815e-06,
+      "loss": 0.6288,
+      "step": 14912
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.480844990930727e-06,
+      "loss": 0.6431,
+      "step": 14913
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.479757597588542e-06,
+      "loss": 0.6262,
+      "step": 14914
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.47867021041449e-06,
+      "loss": 0.6226,
+      "step": 14915
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.477582829421462e-06,
+      "loss": 0.6372,
+      "step": 14916
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.476495454622349e-06,
+      "loss": 0.6028,
+      "step": 14917
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.475408086030044e-06,
+      "loss": 0.6114,
+      "step": 14918
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.474320723657445e-06,
+      "loss": 0.6049,
+      "step": 14919
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.473233367517433e-06,
+      "loss": 0.6994,
+      "step": 14920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.472146017622908e-06,
+      "loss": 0.6398,
+      "step": 14921
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.47105867398676e-06,
+      "loss": 0.5931,
+      "step": 14922
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.469971336621879e-06,
+      "loss": 0.5744,
+      "step": 14923
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.468884005541159e-06,
+      "loss": 0.6483,
+      "step": 14924
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.467796680757493e-06,
+      "loss": 0.6245,
+      "step": 14925
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.466709362283767e-06,
+      "loss": 0.5654,
+      "step": 14926
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.465622050132878e-06,
+      "loss": 0.649,
+      "step": 14927
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.464534744317717e-06,
+      "loss": 0.6065,
+      "step": 14928
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.463447444851172e-06,
+      "loss": 0.6108,
+      "step": 14929
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.462360151746137e-06,
+      "loss": 0.6411,
+      "step": 14930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.461272865015505e-06,
+      "loss": 0.6565,
+      "step": 14931
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.46018558467216e-06,
+      "loss": 0.6474,
+      "step": 14932
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.459098310728999e-06,
+      "loss": 0.6706,
+      "step": 14933
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.458011043198914e-06,
+      "loss": 0.6259,
+      "step": 14934
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.456923782094792e-06,
+      "loss": 0.6659,
+      "step": 14935
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.455836527429526e-06,
+      "loss": 0.5808,
+      "step": 14936
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.454749279216007e-06,
+      "loss": 0.6076,
+      "step": 14937
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.453662037467123e-06,
+      "loss": 0.6439,
+      "step": 14938
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.452574802195767e-06,
+      "loss": 0.602,
+      "step": 14939
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.451487573414832e-06,
+      "loss": 0.5997,
+      "step": 14940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.450400351137204e-06,
+      "loss": 0.6262,
+      "step": 14941
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.449313135375774e-06,
+      "loss": 0.662,
+      "step": 14942
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.448225926143435e-06,
+      "loss": 0.573,
+      "step": 14943
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.447138723453075e-06,
+      "loss": 0.6086,
+      "step": 14944
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.446051527317583e-06,
+      "loss": 0.6579,
+      "step": 14945
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.444964337749856e-06,
+      "loss": 0.6191,
+      "step": 14946
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.443877154762774e-06,
+      "loss": 0.6044,
+      "step": 14947
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.442789978369233e-06,
+      "loss": 0.606,
+      "step": 14948
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.441702808582122e-06,
+      "loss": 0.5633,
+      "step": 14949
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.44061564541433e-06,
+      "loss": 0.5971,
+      "step": 14950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.439528488878747e-06,
+      "loss": 0.6317,
+      "step": 14951
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.438441338988265e-06,
+      "loss": 0.6512,
+      "step": 14952
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.43735419575577e-06,
+      "loss": 0.6329,
+      "step": 14953
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.436267059194152e-06,
+      "loss": 0.5846,
+      "step": 14954
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.435179929316303e-06,
+      "loss": 0.6539,
+      "step": 14955
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.434092806135108e-06,
+      "loss": 0.6338,
+      "step": 14956
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.43300568966346e-06,
+      "loss": 0.6563,
+      "step": 14957
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.431918579914248e-06,
+      "loss": 0.5685,
+      "step": 14958
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.430831476900357e-06,
+      "loss": 0.6547,
+      "step": 14959
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.42974438063468e-06,
+      "loss": 0.6413,
+      "step": 14960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.428657291130107e-06,
+      "loss": 0.6138,
+      "step": 14961
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.427570208399521e-06,
+      "loss": 0.6521,
+      "step": 14962
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.426483132455814e-06,
+      "loss": 0.6622,
+      "step": 14963
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.42539606331188e-06,
+      "loss": 0.6101,
+      "step": 14964
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.424309000980598e-06,
+      "loss": 0.5879,
+      "step": 14965
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.423221945474866e-06,
+      "loss": 0.612,
+      "step": 14966
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.422134896807562e-06,
+      "loss": 0.6374,
+      "step": 14967
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.421047854991581e-06,
+      "loss": 0.6765,
+      "step": 14968
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.419960820039813e-06,
+      "loss": 0.5899,
+      "step": 14969
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.418873791965138e-06,
+      "loss": 0.6482,
+      "step": 14970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.417786770780453e-06,
+      "loss": 0.6297,
+      "step": 14971
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.416699756498645e-06,
+      "loss": 0.6217,
+      "step": 14972
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.415612749132593e-06,
+      "loss": 0.6808,
+      "step": 14973
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.414525748695193e-06,
+      "loss": 0.6329,
+      "step": 14974
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.413438755199335e-06,
+      "loss": 0.5937,
+      "step": 14975
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.412351768657898e-06,
+      "loss": 0.6437,
+      "step": 14976
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.411264789083773e-06,
+      "loss": 0.6308,
+      "step": 14977
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.410177816489856e-06,
+      "loss": 0.6052,
+      "step": 14978
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.40909085088902e-06,
+      "loss": 0.6237,
+      "step": 14979
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.408003892294162e-06,
+      "loss": 0.6225,
+      "step": 14980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.40691694071817e-06,
+      "loss": 0.6195,
+      "step": 14981
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.405829996173921e-06,
+      "loss": 0.6443,
+      "step": 14982
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.404743058674312e-06,
+      "loss": 0.5896,
+      "step": 14983
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.40365612823223e-06,
+      "loss": 0.64,
+      "step": 14984
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.402569204860555e-06,
+      "loss": 0.6309,
+      "step": 14985
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.401482288572179e-06,
+      "loss": 0.599,
+      "step": 14986
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.400395379379993e-06,
+      "loss": 0.6727,
+      "step": 14987
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.399308477296869e-06,
+      "loss": 0.607,
+      "step": 14988
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.398221582335705e-06,
+      "loss": 0.5919,
+      "step": 14989
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.397134694509393e-06,
+      "loss": 0.6538,
+      "step": 14990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.396047813830804e-06,
+      "loss": 0.6296,
+      "step": 14991
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.394960940312833e-06,
+      "loss": 0.6214,
+      "step": 14992
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.39387407396837e-06,
+      "loss": 0.6238,
+      "step": 14993
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.392787214810292e-06,
+      "loss": 0.5973,
+      "step": 14994
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.39170036285149e-06,
+      "loss": 0.6029,
+      "step": 14995
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.390613518104855e-06,
+      "loss": 0.6498,
+      "step": 14996
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.38952668058326e-06,
+      "loss": 0.6421,
+      "step": 14997
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.388439850299602e-06,
+      "loss": 0.6153,
+      "step": 14998
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.387353027266767e-06,
+      "loss": 0.5939,
+      "step": 14999
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.38626621149763e-06,
+      "loss": 0.6051,
+      "step": 15000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.385179403005085e-06,
+      "loss": 0.6587,
+      "step": 15001
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.384092601802022e-06,
+      "loss": 0.6316,
+      "step": 15002
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.383005807901313e-06,
+      "loss": 0.6879,
+      "step": 15003
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.381919021315853e-06,
+      "loss": 0.6424,
+      "step": 15004
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.380832242058529e-06,
+      "loss": 0.6424,
+      "step": 15005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.379745470142214e-06,
+      "loss": 0.624,
+      "step": 15006
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.378658705579806e-06,
+      "loss": 0.7022,
+      "step": 15007
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.377571948384189e-06,
+      "loss": 0.6527,
+      "step": 15008
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.376485198568236e-06,
+      "loss": 0.6471,
+      "step": 15009
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.375398456144841e-06,
+      "loss": 0.6621,
+      "step": 15010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.374311721126895e-06,
+      "loss": 0.6266,
+      "step": 15011
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.373224993527266e-06,
+      "loss": 0.6688,
+      "step": 15012
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.37213827335885e-06,
+      "loss": 0.6491,
+      "step": 15013
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.371051560634535e-06,
+      "loss": 0.6735,
+      "step": 15014
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.369964855367192e-06,
+      "loss": 0.6487,
+      "step": 15015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.368878157569715e-06,
+      "loss": 0.6352,
+      "step": 15016
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.367791467254989e-06,
+      "loss": 0.624,
+      "step": 15017
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.36670478443589e-06,
+      "loss": 0.5881,
+      "step": 15018
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.365618109125308e-06,
+      "loss": 0.582,
+      "step": 15019
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.364531441336129e-06,
+      "loss": 0.6688,
+      "step": 15020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.363444781081228e-06,
+      "loss": 0.6131,
+      "step": 15021
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.362358128373498e-06,
+      "loss": 0.6148,
+      "step": 15022
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.361271483225822e-06,
+      "loss": 0.6608,
+      "step": 15023
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.360184845651074e-06,
+      "loss": 0.5763,
+      "step": 15024
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.359098215662147e-06,
+      "loss": 0.6597,
+      "step": 15025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.358011593271925e-06,
+      "loss": 0.5868,
+      "step": 15026
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.356924978493283e-06,
+      "loss": 0.6589,
+      "step": 15027
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.355838371339114e-06,
+      "loss": 0.7261,
+      "step": 15028
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.35475177182229e-06,
+      "loss": 0.6176,
+      "step": 15029
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.3536651799557e-06,
+      "loss": 0.614,
+      "step": 15030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.352578595752233e-06,
+      "loss": 0.6413,
+      "step": 15031
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.351492019224761e-06,
+      "loss": 0.649,
+      "step": 15032
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.350405450386168e-06,
+      "loss": 0.5996,
+      "step": 15033
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.349318889249349e-06,
+      "loss": 0.65,
+      "step": 15034
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.348232335827173e-06,
+      "loss": 0.6628,
+      "step": 15035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.347145790132524e-06,
+      "loss": 0.6519,
+      "step": 15036
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.346059252178295e-06,
+      "loss": 0.6246,
+      "step": 15037
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.344972721977355e-06,
+      "loss": 0.6101,
+      "step": 15038
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.34388619954259e-06,
+      "loss": 0.5932,
+      "step": 15039
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.342799684886892e-06,
+      "loss": 0.6617,
+      "step": 15040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.34171317802313e-06,
+      "loss": 0.6454,
+      "step": 15041
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.34062667896419e-06,
+      "loss": 0.6358,
+      "step": 15042
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.339540187722959e-06,
+      "loss": 0.6418,
+      "step": 15043
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.338453704312312e-06,
+      "loss": 0.6438,
+      "step": 15044
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.337367228745131e-06,
+      "loss": 0.6495,
+      "step": 15045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.336280761034306e-06,
+      "loss": 0.597,
+      "step": 15046
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.335194301192707e-06,
+      "loss": 0.622,
+      "step": 15047
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.334107849233219e-06,
+      "loss": 0.6069,
+      "step": 15048
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.333021405168731e-06,
+      "loss": 0.5994,
+      "step": 15049
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.331934969012115e-06,
+      "loss": 0.6712,
+      "step": 15050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.330848540776252e-06,
+      "loss": 0.5952,
+      "step": 15051
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.329762120474032e-06,
+      "loss": 0.6624,
+      "step": 15052
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.328675708118327e-06,
+      "loss": 0.6075,
+      "step": 15053
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.327589303722017e-06,
+      "loss": 0.6166,
+      "step": 15054
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.326502907297993e-06,
+      "loss": 0.6167,
+      "step": 15055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.325416518859125e-06,
+      "loss": 0.7058,
+      "step": 15056
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.324330138418294e-06,
+      "loss": 0.6052,
+      "step": 15057
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.323243765988393e-06,
+      "loss": 0.609,
+      "step": 15058
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.322157401582289e-06,
+      "loss": 0.6096,
+      "step": 15059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.321071045212862e-06,
+      "loss": 0.6551,
+      "step": 15060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.319984696893005e-06,
+      "loss": 0.6294,
+      "step": 15061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.318898356635583e-06,
+      "loss": 0.6294,
+      "step": 15062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.317812024453482e-06,
+      "loss": 0.6051,
+      "step": 15063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.31672570035959e-06,
+      "loss": 0.6294,
+      "step": 15064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.315639384366773e-06,
+      "loss": 0.6413,
+      "step": 15065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.314553076487916e-06,
+      "loss": 0.6043,
+      "step": 15066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.313466776735906e-06,
+      "loss": 0.6656,
+      "step": 15067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.31238048512361e-06,
+      "loss": 0.6296,
+      "step": 15068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.311294201663912e-06,
+      "loss": 0.6349,
+      "step": 15069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.310207926369699e-06,
+      "loss": 0.602,
+      "step": 15070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.309121659253838e-06,
+      "loss": 0.6289,
+      "step": 15071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.308035400329213e-06,
+      "loss": 0.6521,
+      "step": 15072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.30694914960871e-06,
+      "loss": 0.5987,
+      "step": 15073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.305862907105196e-06,
+      "loss": 0.6208,
+      "step": 15074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.304776672831557e-06,
+      "loss": 0.6328,
+      "step": 15075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.30369044680067e-06,
+      "loss": 0.6236,
+      "step": 15076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.302604229025414e-06,
+      "loss": 0.604,
+      "step": 15077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.301518019518666e-06,
+      "loss": 0.6154,
+      "step": 15078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.300431818293307e-06,
+      "loss": 0.6412,
+      "step": 15079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.299345625362213e-06,
+      "loss": 0.629,
+      "step": 15080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.298259440738263e-06,
+      "loss": 0.6422,
+      "step": 15081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.297173264434334e-06,
+      "loss": 0.6138,
+      "step": 15082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.296087096463305e-06,
+      "loss": 0.6829,
+      "step": 15083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.295000936838055e-06,
+      "loss": 0.6444,
+      "step": 15084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.293914785571462e-06,
+      "loss": 0.586,
+      "step": 15085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.2928286426764e-06,
+      "loss": 0.6064,
+      "step": 15086
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.291742508165749e-06,
+      "loss": 0.6323,
+      "step": 15087
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.290656382052388e-06,
+      "loss": 0.6689,
+      "step": 15088
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.289570264349192e-06,
+      "loss": 0.6109,
+      "step": 15089
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.288484155069041e-06,
+      "loss": 0.6462,
+      "step": 15090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.28739805422481e-06,
+      "loss": 0.5916,
+      "step": 15091
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.286311961829374e-06,
+      "loss": 0.6417,
+      "step": 15092
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.285225877895615e-06,
+      "loss": 0.633,
+      "step": 15093
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.284139802436407e-06,
+      "loss": 0.6589,
+      "step": 15094
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.283053735464624e-06,
+      "loss": 0.644,
+      "step": 15095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.281967676993152e-06,
+      "loss": 0.5839,
+      "step": 15096
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.280881627034856e-06,
+      "loss": 0.6912,
+      "step": 15097
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.27979558560262e-06,
+      "loss": 0.6841,
+      "step": 15098
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.27870955270932e-06,
+      "loss": 0.604,
+      "step": 15099
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.277623528367828e-06,
+      "loss": 0.6198,
+      "step": 15100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.276537512591022e-06,
+      "loss": 0.5535,
+      "step": 15101
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.275451505391783e-06,
+      "loss": 0.6636,
+      "step": 15102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.27436550678298e-06,
+      "loss": 0.6461,
+      "step": 15103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.273279516777494e-06,
+      "loss": 0.6287,
+      "step": 15104
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.272193535388198e-06,
+      "loss": 0.6432,
+      "step": 15105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.271107562627967e-06,
+      "loss": 0.6356,
+      "step": 15106
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.270021598509677e-06,
+      "loss": 0.6085,
+      "step": 15107
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.268935643046208e-06,
+      "loss": 0.6442,
+      "step": 15108
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.267849696250429e-06,
+      "loss": 0.6168,
+      "step": 15109
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.26676375813522e-06,
+      "loss": 0.6542,
+      "step": 15110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.265677828713453e-06,
+      "loss": 0.603,
+      "step": 15111
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.264591907998004e-06,
+      "loss": 0.6499,
+      "step": 15112
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.263505996001747e-06,
+      "loss": 0.6191,
+      "step": 15113
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.262420092737562e-06,
+      "loss": 0.6771,
+      "step": 15114
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.261334198218315e-06,
+      "loss": 0.6284,
+      "step": 15115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.26024831245689e-06,
+      "loss": 0.6286,
+      "step": 15116
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.259162435466154e-06,
+      "loss": 0.6647,
+      "step": 15117
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.258076567258986e-06,
+      "loss": 0.6555,
+      "step": 15118
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.256990707848257e-06,
+      "loss": 0.6474,
+      "step": 15119
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.255904857246846e-06,
+      "loss": 0.6723,
+      "step": 15120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.254819015467621e-06,
+      "loss": 0.61,
+      "step": 15121
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.253733182523462e-06,
+      "loss": 0.6129,
+      "step": 15122
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.252647358427239e-06,
+      "loss": 0.6526,
+      "step": 15123
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.251561543191825e-06,
+      "loss": 0.5749,
+      "step": 15124
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.250475736830097e-06,
+      "loss": 0.6076,
+      "step": 15125
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.24938993935493e-06,
+      "loss": 0.6543,
+      "step": 15126
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.248304150779193e-06,
+      "loss": 0.611,
+      "step": 15127
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.24721837111576e-06,
+      "loss": 0.6483,
+      "step": 15128
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.246132600377507e-06,
+      "loss": 0.6158,
+      "step": 15129
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.245046838577305e-06,
+      "loss": 0.6318,
+      "step": 15130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.243961085728027e-06,
+      "loss": 0.6582,
+      "step": 15131
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.242875341842549e-06,
+      "loss": 0.5546,
+      "step": 15132
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.24178960693374e-06,
+      "loss": 0.626,
+      "step": 15133
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.240703881014474e-06,
+      "loss": 0.6366,
+      "step": 15134
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.239618164097624e-06,
+      "loss": 0.5975,
+      "step": 15135
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.238532456196062e-06,
+      "loss": 0.6332,
+      "step": 15136
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.237446757322663e-06,
+      "loss": 0.6079,
+      "step": 15137
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.236361067490296e-06,
+      "loss": 0.5868,
+      "step": 15138
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.235275386711834e-06,
+      "loss": 0.6525,
+      "step": 15139
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.234189715000152e-06,
+      "loss": 0.6697,
+      "step": 15140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.23310405236812e-06,
+      "loss": 0.6151,
+      "step": 15141
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.232018398828606e-06,
+      "loss": 0.6294,
+      "step": 15142
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.230932754394488e-06,
+      "loss": 0.6626,
+      "step": 15143
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.229847119078634e-06,
+      "loss": 0.6523,
+      "step": 15144
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.228761492893916e-06,
+      "loss": 0.5719,
+      "step": 15145
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.227675875853207e-06,
+      "loss": 0.6027,
+      "step": 15146
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.226590267969378e-06,
+      "loss": 0.5914,
+      "step": 15147
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.2255046692553e-06,
+      "loss": 0.6164,
+      "step": 15148
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.224419079723842e-06,
+      "loss": 0.5621,
+      "step": 15149
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.22333349938788e-06,
+      "loss": 0.6095,
+      "step": 15150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.22224792826028e-06,
+      "loss": 0.6436,
+      "step": 15151
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.221162366353915e-06,
+      "loss": 0.6611,
+      "step": 15152
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.220076813681655e-06,
+      "loss": 0.5903,
+      "step": 15153
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.21899127025637e-06,
+      "loss": 0.5882,
+      "step": 15154
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.217905736090932e-06,
+      "loss": 0.6175,
+      "step": 15155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.216820211198212e-06,
+      "loss": 0.6613,
+      "step": 15156
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.215734695591078e-06,
+      "loss": 0.5931,
+      "step": 15157
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.214649189282403e-06,
+      "loss": 0.643,
+      "step": 15158
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.213563692285054e-06,
+      "loss": 0.6606,
+      "step": 15159
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.212478204611901e-06,
+      "loss": 0.6722,
+      "step": 15160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.211392726275818e-06,
+      "loss": 0.6461,
+      "step": 15161
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.21030725728967e-06,
+      "loss": 0.6539,
+      "step": 15162
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.209221797666328e-06,
+      "loss": 0.5857,
+      "step": 15163
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.208136347418664e-06,
+      "loss": 0.6551,
+      "step": 15164
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.207050906559543e-06,
+      "loss": 0.6405,
+      "step": 15165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.205965475101836e-06,
+      "loss": 0.5674,
+      "step": 15166
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.204880053058416e-06,
+      "loss": 0.6813,
+      "step": 15167
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.203794640442147e-06,
+      "loss": 0.5786,
+      "step": 15168
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.2027092372659e-06,
+      "loss": 0.6674,
+      "step": 15169
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.201623843542543e-06,
+      "loss": 0.6438,
+      "step": 15170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.200538459284945e-06,
+      "loss": 0.6084,
+      "step": 15171
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.199453084505975e-06,
+      "loss": 0.5994,
+      "step": 15172
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.198367719218504e-06,
+      "loss": 0.673,
+      "step": 15173
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.197282363435395e-06,
+      "loss": 0.6551,
+      "step": 15174
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.196197017169519e-06,
+      "loss": 0.6167,
+      "step": 15175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.195111680433746e-06,
+      "loss": 0.6244,
+      "step": 15176
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.19402635324094e-06,
+      "loss": 0.6489,
+      "step": 15177
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.192941035603971e-06,
+      "loss": 0.6932,
+      "step": 15178
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.191855727535712e-06,
+      "loss": 0.6511,
+      "step": 15179
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.19077042904902e-06,
+      "loss": 0.6462,
+      "step": 15180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.189685140156768e-06,
+      "loss": 0.6056,
+      "step": 15181
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.188599860871828e-06,
+      "loss": 0.5765,
+      "step": 15182
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.18751459120706e-06,
+      "loss": 0.6698,
+      "step": 15183
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.186429331175335e-06,
+      "loss": 0.5973,
+      "step": 15184
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.185344080789522e-06,
+      "loss": 0.6599,
+      "step": 15185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.184258840062482e-06,
+      "loss": 0.651,
+      "step": 15186
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.183173609007085e-06,
+      "loss": 0.5947,
+      "step": 15187
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.182088387636202e-06,
+      "loss": 0.6601,
+      "step": 15188
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.181003175962693e-06,
+      "loss": 0.6376,
+      "step": 15189
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.179917973999427e-06,
+      "loss": 0.6686,
+      "step": 15190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.178832781759272e-06,
+      "loss": 0.659,
+      "step": 15191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.177747599255094e-06,
+      "loss": 0.6268,
+      "step": 15192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.176662426499757e-06,
+      "loss": 0.6315,
+      "step": 15193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.175577263506129e-06,
+      "loss": 0.6701,
+      "step": 15194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.174492110287074e-06,
+      "loss": 0.615,
+      "step": 15195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.17340696685546e-06,
+      "loss": 0.6114,
+      "step": 15196
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.172321833224153e-06,
+      "loss": 0.6404,
+      "step": 15197
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.171236709406017e-06,
+      "loss": 0.5947,
+      "step": 15198
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.17015159541392e-06,
+      "loss": 0.5986,
+      "step": 15199
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.169066491260724e-06,
+      "loss": 0.6469,
+      "step": 15200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.167981396959294e-06,
+      "loss": 0.6466,
+      "step": 15201
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.166896312522499e-06,
+      "loss": 0.6043,
+      "step": 15202
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.165811237963203e-06,
+      "loss": 0.6457,
+      "step": 15203
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.164726173294269e-06,
+      "loss": 0.6134,
+      "step": 15204
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.163641118528563e-06,
+      "loss": 0.609,
+      "step": 15205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.16255607367895e-06,
+      "loss": 0.6316,
+      "step": 15206
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.161471038758292e-06,
+      "loss": 0.612,
+      "step": 15207
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.160386013779457e-06,
+      "loss": 0.6672,
+      "step": 15208
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.159300998755309e-06,
+      "loss": 0.619,
+      "step": 15209
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.15821599369871e-06,
+      "loss": 0.7012,
+      "step": 15210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.157130998622525e-06,
+      "loss": 0.6108,
+      "step": 15211
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.15604601353962e-06,
+      "loss": 0.6274,
+      "step": 15212
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.154961038462854e-06,
+      "loss": 0.6791,
+      "step": 15213
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.1538760734051e-06,
+      "loss": 0.6391,
+      "step": 15214
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.152791118379209e-06,
+      "loss": 0.6766,
+      "step": 15215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.151706173398053e-06,
+      "loss": 0.6807,
+      "step": 15216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.150621238474497e-06,
+      "loss": 0.5966,
+      "step": 15217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.149536313621394e-06,
+      "loss": 0.6922,
+      "step": 15218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.148451398851616e-06,
+      "loss": 0.6279,
+      "step": 15219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.147366494178029e-06,
+      "loss": 0.6662,
+      "step": 15220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.146281599613484e-06,
+      "loss": 0.5877,
+      "step": 15221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.145196715170853e-06,
+      "loss": 0.6046,
+      "step": 15222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.144111840863e-06,
+      "loss": 0.6208,
+      "step": 15223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.143026976702778e-06,
+      "loss": 0.6714,
+      "step": 15224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.141942122703057e-06,
+      "loss": 0.643,
+      "step": 15225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.140857278876702e-06,
+      "loss": 0.6048,
+      "step": 15226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.139772445236564e-06,
+      "loss": 0.6686,
+      "step": 15227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.138687621795515e-06,
+      "loss": 0.6344,
+      "step": 15228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.137602808566417e-06,
+      "loss": 0.6198,
+      "step": 15229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.136518005562122e-06,
+      "loss": 0.6149,
+      "step": 15230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.1354332127955e-06,
+      "loss": 0.6538,
+      "step": 15231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.134348430279418e-06,
+      "loss": 0.6279,
+      "step": 15232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.133263658026722e-06,
+      "loss": 0.6588,
+      "step": 15233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.132178896050285e-06,
+      "loss": 0.6424,
+      "step": 15234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.13109414436297e-06,
+      "loss": 0.5943,
+      "step": 15235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.130009402977622e-06,
+      "loss": 0.6001,
+      "step": 15236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.12892467190712e-06,
+      "loss": 0.5578,
+      "step": 15237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.127839951164321e-06,
+      "loss": 0.6642,
+      "step": 15238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.126755240762075e-06,
+      "loss": 0.6447,
+      "step": 15239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.125670540713253e-06,
+      "loss": 0.5916,
+      "step": 15240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.124585851030719e-06,
+      "loss": 0.6483,
+      "step": 15241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.123501171727318e-06,
+      "loss": 0.6345,
+      "step": 15242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.122416502815923e-06,
+      "loss": 0.6138,
+      "step": 15243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.121331844309394e-06,
+      "loss": 0.6374,
+      "step": 15244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.120247196220581e-06,
+      "loss": 0.683,
+      "step": 15245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.119162558562353e-06,
+      "loss": 0.6589,
+      "step": 15246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.118077931347571e-06,
+      "loss": 0.5562,
+      "step": 15247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.116993314589085e-06,
+      "loss": 0.6954,
+      "step": 15248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.115908708299762e-06,
+      "loss": 0.5741,
+      "step": 15249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.114824112492464e-06,
+      "loss": 0.6855,
+      "step": 15250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.11373952718004e-06,
+      "loss": 0.6381,
+      "step": 15251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.112654952375358e-06,
+      "loss": 0.6546,
+      "step": 15252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.111570388091276e-06,
+      "loss": 0.6518,
+      "step": 15253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.110485834340646e-06,
+      "loss": 0.5756,
+      "step": 15254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.109401291136333e-06,
+      "loss": 0.6336,
+      "step": 15255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.108316758491198e-06,
+      "loss": 0.643,
+      "step": 15256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.10723223641809e-06,
+      "loss": 0.6306,
+      "step": 15257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.106147724929878e-06,
+      "loss": 0.6248,
+      "step": 15258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.105063224039416e-06,
+      "loss": 0.6698,
+      "step": 15259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.103978733759556e-06,
+      "loss": 0.6328,
+      "step": 15260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.102894254103165e-06,
+      "loss": 0.6287,
+      "step": 15261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.1018097850831e-06,
+      "loss": 0.6377,
+      "step": 15262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.100725326712211e-06,
+      "loss": 0.6712,
+      "step": 15263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.099640879003363e-06,
+      "loss": 0.6812,
+      "step": 15264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.098556441969414e-06,
+      "loss": 0.6583,
+      "step": 15265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.097472015623218e-06,
+      "loss": 0.6226,
+      "step": 15266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.096387599977627e-06,
+      "loss": 0.5932,
+      "step": 15267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.095303195045513e-06,
+      "loss": 0.6344,
+      "step": 15268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.094218800839719e-06,
+      "loss": 0.6336,
+      "step": 15269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.093134417373105e-06,
+      "loss": 0.6219,
+      "step": 15270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.092050044658535e-06,
+      "loss": 0.658,
+      "step": 15271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.090965682708858e-06,
+      "loss": 0.5875,
+      "step": 15272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.08988133153693e-06,
+      "loss": 0.6308,
+      "step": 15273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.088796991155614e-06,
+      "loss": 0.6476,
+      "step": 15274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.08771266157776e-06,
+      "loss": 0.6019,
+      "step": 15275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.08662834281623e-06,
+      "loss": 0.6411,
+      "step": 15276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.085544034883873e-06,
+      "loss": 0.6228,
+      "step": 15277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.084459737793548e-06,
+      "loss": 0.67,
+      "step": 15278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.083375451558113e-06,
+      "loss": 0.6171,
+      "step": 15279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.08229117619042e-06,
+      "loss": 0.5955,
+      "step": 15280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.081206911703325e-06,
+      "loss": 0.6443,
+      "step": 15281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.080122658109689e-06,
+      "loss": 0.6671,
+      "step": 15282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.079038415422357e-06,
+      "loss": 0.6563,
+      "step": 15283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.077954183654191e-06,
+      "loss": 0.6091,
+      "step": 15284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.076869962818048e-06,
+      "loss": 0.6616,
+      "step": 15285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.075785752926776e-06,
+      "loss": 0.6327,
+      "step": 15286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.074701553993233e-06,
+      "loss": 0.6694,
+      "step": 15287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.073617366030276e-06,
+      "loss": 0.6034,
+      "step": 15288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.072533189050754e-06,
+      "loss": 0.6384,
+      "step": 15289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.071449023067526e-06,
+      "loss": 0.6448,
+      "step": 15290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.070364868093446e-06,
+      "loss": 0.606,
+      "step": 15291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.069280724141364e-06,
+      "loss": 0.6146,
+      "step": 15292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.068196591224135e-06,
+      "loss": 0.6686,
+      "step": 15293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.067112469354618e-06,
+      "loss": 0.6855,
+      "step": 15294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.06602835854566e-06,
+      "loss": 0.6301,
+      "step": 15295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.064944258810118e-06,
+      "loss": 0.6358,
+      "step": 15296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.063860170160847e-06,
+      "loss": 0.5933,
+      "step": 15297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.062776092610695e-06,
+      "loss": 0.6437,
+      "step": 15298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.061692026172517e-06,
+      "loss": 0.6004,
+      "step": 15299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.060607970859171e-06,
+      "loss": 0.6553,
+      "step": 15300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.059523926683504e-06,
+      "loss": 0.6246,
+      "step": 15301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.058439893658368e-06,
+      "loss": 0.6347,
+      "step": 15302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.057355871796622e-06,
+      "loss": 0.5601,
+      "step": 15303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.056271861111112e-06,
+      "loss": 0.6398,
+      "step": 15304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.055187861614694e-06,
+      "loss": 0.6053,
+      "step": 15305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.05410387332022e-06,
+      "loss": 0.6093,
+      "step": 15306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.05301989624054e-06,
+      "loss": 0.6038,
+      "step": 15307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.051935930388505e-06,
+      "loss": 0.6216,
+      "step": 15308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.050851975776974e-06,
+      "loss": 0.6264,
+      "step": 15309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.04976803241879e-06,
+      "loss": 0.6111,
+      "step": 15310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.048684100326807e-06,
+      "loss": 0.6115,
+      "step": 15311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.04760017951388e-06,
+      "loss": 0.5825,
+      "step": 15312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.046516269992856e-06,
+      "loss": 0.6246,
+      "step": 15313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.045432371776588e-06,
+      "loss": 0.6647,
+      "step": 15314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.04434848487793e-06,
+      "loss": 0.6266,
+      "step": 15315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.043264609309725e-06,
+      "loss": 0.6508,
+      "step": 15316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.042180745084829e-06,
+      "loss": 0.6313,
+      "step": 15317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.041096892216095e-06,
+      "loss": 0.6503,
+      "step": 15318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.04001305071637e-06,
+      "loss": 0.6051,
+      "step": 15319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.0389292205985e-06,
+      "loss": 0.6323,
+      "step": 15320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.037845401875344e-06,
+      "loss": 0.618,
+      "step": 15321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.036761594559747e-06,
+      "loss": 0.6328,
+      "step": 15322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.03567779866456e-06,
+      "loss": 0.6471,
+      "step": 15323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.034594014202634e-06,
+      "loss": 0.6947,
+      "step": 15324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.033510241186814e-06,
+      "loss": 0.6369,
+      "step": 15325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.032426479629953e-06,
+      "loss": 0.6208,
+      "step": 15326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.031342729544903e-06,
+      "loss": 0.612,
+      "step": 15327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.030258990944506e-06,
+      "loss": 0.6023,
+      "step": 15328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.029175263841618e-06,
+      "loss": 0.6194,
+      "step": 15329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.028091548249086e-06,
+      "loss": 0.5634,
+      "step": 15330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.027007844179758e-06,
+      "loss": 0.6433,
+      "step": 15331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.025924151646482e-06,
+      "loss": 0.5847,
+      "step": 15332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.024840470662109e-06,
+      "loss": 0.6398,
+      "step": 15333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.023756801239484e-06,
+      "loss": 0.6121,
+      "step": 15334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.022673143391457e-06,
+      "loss": 0.6553,
+      "step": 15335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.021589497130879e-06,
+      "loss": 0.6194,
+      "step": 15336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.020505862470592e-06,
+      "loss": 0.5737,
+      "step": 15337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.019422239423448e-06,
+      "loss": 0.641,
+      "step": 15338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.018338628002293e-06,
+      "loss": 0.6431,
+      "step": 15339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.017255028219976e-06,
+      "loss": 0.6718,
+      "step": 15340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.016171440089345e-06,
+      "loss": 0.6694,
+      "step": 15341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.015087863623245e-06,
+      "loss": 0.5901,
+      "step": 15342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.014004298834523e-06,
+      "loss": 0.6643,
+      "step": 15343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.012920745736031e-06,
+      "loss": 0.6222,
+      "step": 15344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.011837204340609e-06,
+      "loss": 0.6434,
+      "step": 15345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.010753674661107e-06,
+      "loss": 0.5609,
+      "step": 15346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.009670156710374e-06,
+      "loss": 0.6292,
+      "step": 15347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.008586650501251e-06,
+      "loss": 0.6305,
+      "step": 15348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.007503156046589e-06,
+      "loss": 0.6056,
+      "step": 15349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.006419673359234e-06,
+      "loss": 0.6867,
+      "step": 15350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.00533620245203e-06,
+      "loss": 0.6548,
+      "step": 15351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.004252743337822e-06,
+      "loss": 0.6314,
+      "step": 15352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.003169296029459e-06,
+      "loss": 0.6701,
+      "step": 15353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.002085860539782e-06,
+      "loss": 0.6231,
+      "step": 15354
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.001002436881643e-06,
+      "loss": 0.6338,
+      "step": 15355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.999919025067883e-06,
+      "loss": 0.6378,
+      "step": 15356
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.998835625111347e-06,
+      "loss": 0.6105,
+      "step": 15357
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.997752237024882e-06,
+      "loss": 0.5909,
+      "step": 15358
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.996668860821333e-06,
+      "loss": 0.6494,
+      "step": 15359
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.99558549651354e-06,
+      "loss": 0.6103,
+      "step": 15360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.994502144114356e-06,
+      "loss": 0.6183,
+      "step": 15361
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.99341880363662e-06,
+      "loss": 0.6801,
+      "step": 15362
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.992335475093176e-06,
+      "loss": 0.5923,
+      "step": 15363
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.99125215849687e-06,
+      "loss": 0.6284,
+      "step": 15364
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.990168853860548e-06,
+      "loss": 0.6361,
+      "step": 15365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.989085561197048e-06,
+      "loss": 0.5774,
+      "step": 15366
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.98800228051922e-06,
+      "loss": 0.5968,
+      "step": 15367
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.986919011839903e-06,
+      "loss": 0.6179,
+      "step": 15368
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.985835755171945e-06,
+      "loss": 0.6429,
+      "step": 15369
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.984752510528185e-06,
+      "loss": 0.6472,
+      "step": 15370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.983669277921471e-06,
+      "loss": 0.6247,
+      "step": 15371
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.982586057364642e-06,
+      "loss": 0.6443,
+      "step": 15372
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.98150284887054e-06,
+      "loss": 0.6061,
+      "step": 15373
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.980419652452013e-06,
+      "loss": 0.6396,
+      "step": 15374
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.9793364681219e-06,
+      "loss": 0.6241,
+      "step": 15375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.978253295893045e-06,
+      "loss": 0.5908,
+      "step": 15376
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.97717013577829e-06,
+      "loss": 0.6134,
+      "step": 15377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.976086987790474e-06,
+      "loss": 0.6224,
+      "step": 15378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.975003851942445e-06,
+      "loss": 0.6466,
+      "step": 15379
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.973920728247041e-06,
+      "loss": 0.6273,
+      "step": 15380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.972837616717106e-06,
+      "loss": 0.6261,
+      "step": 15381
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.971754517365476e-06,
+      "loss": 0.6788,
+      "step": 15382
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.970671430205003e-06,
+      "loss": 0.5891,
+      "step": 15383
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.969588355248517e-06,
+      "loss": 0.6602,
+      "step": 15384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.968505292508867e-06,
+      "loss": 0.633,
+      "step": 15385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.967422241998894e-06,
+      "loss": 0.6791,
+      "step": 15386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.966339203731432e-06,
+      "loss": 0.6922,
+      "step": 15387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.965256177719328e-06,
+      "loss": 0.6028,
+      "step": 15388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.964173163975422e-06,
+      "loss": 0.612,
+      "step": 15389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.963090162512553e-06,
+      "loss": 0.6519,
+      "step": 15390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.96200717334356e-06,
+      "loss": 0.6538,
+      "step": 15391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.96092419648129e-06,
+      "loss": 0.6388,
+      "step": 15392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.959841231938574e-06,
+      "loss": 0.6433,
+      "step": 15393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.958758279728255e-06,
+      "loss": 0.6312,
+      "step": 15394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.957675339863177e-06,
+      "loss": 0.602,
+      "step": 15395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.956592412356174e-06,
+      "loss": 0.6522,
+      "step": 15396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.955509497220089e-06,
+      "loss": 0.6247,
+      "step": 15397
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.95442659446776e-06,
+      "loss": 0.6027,
+      "step": 15398
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.953343704112026e-06,
+      "loss": 0.619,
+      "step": 15399
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.952260826165728e-06,
+      "loss": 0.6302,
+      "step": 15400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.9511779606417e-06,
+      "loss": 0.6249,
+      "step": 15401
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.950095107552788e-06,
+      "loss": 0.6345,
+      "step": 15402
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.949012266911825e-06,
+      "loss": 0.5572,
+      "step": 15403
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.94792943873165e-06,
+      "loss": 0.6535,
+      "step": 15404
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.946846623025103e-06,
+      "loss": 0.6541,
+      "step": 15405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.945763819805023e-06,
+      "loss": 0.5769,
+      "step": 15406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.944681029084244e-06,
+      "loss": 0.6007,
+      "step": 15407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.943598250875607e-06,
+      "loss": 0.6088,
+      "step": 15408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.942515485191952e-06,
+      "loss": 0.6371,
+      "step": 15409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.94143273204611e-06,
+      "loss": 0.6156,
+      "step": 15410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.940349991450923e-06,
+      "loss": 0.6685,
+      "step": 15411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.93926726341923e-06,
+      "loss": 0.6311,
+      "step": 15412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.938184547963862e-06,
+      "loss": 0.6995,
+      "step": 15413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.93710184509766e-06,
+      "loss": 0.5842,
+      "step": 15414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.936019154833461e-06,
+      "loss": 0.672,
+      "step": 15415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.9349364771841e-06,
+      "loss": 0.6925,
+      "step": 15416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.933853812162416e-06,
+      "loss": 0.6428,
+      "step": 15417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.932771159781243e-06,
+      "loss": 0.6419,
+      "step": 15418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.931688520053418e-06,
+      "loss": 0.6367,
+      "step": 15419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.930605892991775e-06,
+      "loss": 0.6153,
+      "step": 15420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.929523278609156e-06,
+      "loss": 0.7123,
+      "step": 15421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.92844067691839e-06,
+      "loss": 0.6983,
+      "step": 15422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.927358087932313e-06,
+      "loss": 0.6168,
+      "step": 15423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.926275511663768e-06,
+      "loss": 0.655,
+      "step": 15424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.925192948125581e-06,
+      "loss": 0.5992,
+      "step": 15425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.924110397330591e-06,
+      "loss": 0.6032,
+      "step": 15426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.923027859291638e-06,
+      "loss": 0.6102,
+      "step": 15427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.921945334021549e-06,
+      "loss": 0.6638,
+      "step": 15428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.92086282153316e-06,
+      "loss": 0.6656,
+      "step": 15429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.91978032183931e-06,
+      "loss": 0.647,
+      "step": 15430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.918697834952831e-06,
+      "loss": 0.6058,
+      "step": 15431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.917615360886554e-06,
+      "loss": 0.6469,
+      "step": 15432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.91653289965332e-06,
+      "loss": 0.6501,
+      "step": 15433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.915450451265957e-06,
+      "loss": 0.6073,
+      "step": 15434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.914368015737301e-06,
+      "loss": 0.582,
+      "step": 15435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.913285593080186e-06,
+      "loss": 0.6556,
+      "step": 15436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.912203183307444e-06,
+      "loss": 0.6219,
+      "step": 15437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.911120786431909e-06,
+      "loss": 0.6437,
+      "step": 15438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.910038402466417e-06,
+      "loss": 0.6135,
+      "step": 15439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.908956031423796e-06,
+      "loss": 0.6879,
+      "step": 15440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.907873673316882e-06,
+      "loss": 0.6935,
+      "step": 15441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.90679132815851e-06,
+      "loss": 0.6241,
+      "step": 15442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.905708995961506e-06,
+      "loss": 0.6311,
+      "step": 15443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.904626676738707e-06,
+      "loss": 0.6006,
+      "step": 15444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.903544370502945e-06,
+      "loss": 0.5973,
+      "step": 15445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.902462077267051e-06,
+      "loss": 0.695,
+      "step": 15446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.901379797043858e-06,
+      "loss": 0.6492,
+      "step": 15447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.900297529846199e-06,
+      "loss": 0.6438,
+      "step": 15448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.8992152756869e-06,
+      "loss": 0.6312,
+      "step": 15449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.8981330345788e-06,
+      "loss": 0.6297,
+      "step": 15450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.897050806534725e-06,
+      "loss": 0.6497,
+      "step": 15451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.895968591567507e-06,
+      "loss": 0.628,
+      "step": 15452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.894886389689978e-06,
+      "loss": 0.6482,
+      "step": 15453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.893804200914972e-06,
+      "loss": 0.6254,
+      "step": 15454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.892722025255315e-06,
+      "loss": 0.6059,
+      "step": 15455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.891639862723836e-06,
+      "loss": 0.6199,
+      "step": 15456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.890557713333374e-06,
+      "loss": 0.6587,
+      "step": 15457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.889475577096749e-06,
+      "loss": 0.6111,
+      "step": 15458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.888393454026797e-06,
+      "loss": 0.6671,
+      "step": 15459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.887311344136351e-06,
+      "loss": 0.5976,
+      "step": 15460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.88622924743823e-06,
+      "loss": 0.6454,
+      "step": 15461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.885147163945278e-06,
+      "loss": 0.6168,
+      "step": 15462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.88406509367031e-06,
+      "loss": 0.6503,
+      "step": 15463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.88298303662616e-06,
+      "loss": 0.6239,
+      "step": 15464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.881900992825666e-06,
+      "loss": 0.6188,
+      "step": 15465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.880818962281648e-06,
+      "loss": 0.661,
+      "step": 15466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.879736945006933e-06,
+      "loss": 0.6735,
+      "step": 15467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.87865494101436e-06,
+      "loss": 0.6144,
+      "step": 15468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.877572950316748e-06,
+      "loss": 0.6628,
+      "step": 15469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.876490972926925e-06,
+      "loss": 0.6396,
+      "step": 15470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.87540900885773e-06,
+      "loss": 0.6278,
+      "step": 15471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.874327058121979e-06,
+      "loss": 0.6256,
+      "step": 15472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.873245120732503e-06,
+      "loss": 0.6062,
+      "step": 15473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.872163196702138e-06,
+      "loss": 0.6361,
+      "step": 15474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.8710812860437e-06,
+      "loss": 0.6851,
+      "step": 15475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.86999938877002e-06,
+      "loss": 0.609,
+      "step": 15476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.868917504893933e-06,
+      "loss": 0.5786,
+      "step": 15477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.867835634428256e-06,
+      "loss": 0.6109,
+      "step": 15478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.866753777385817e-06,
+      "loss": 0.639,
+      "step": 15479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.865671933779451e-06,
+      "loss": 0.6635,
+      "step": 15480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.864590103621978e-06,
+      "loss": 0.6577,
+      "step": 15481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.86350828692622e-06,
+      "loss": 0.6773,
+      "step": 15482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.862426483705017e-06,
+      "loss": 0.6384,
+      "step": 15483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.861344693971183e-06,
+      "loss": 0.6307,
+      "step": 15484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.860262917737544e-06,
+      "loss": 0.6849,
+      "step": 15485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.859181155016937e-06,
+      "loss": 0.6583,
+      "step": 15486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.858099405822179e-06,
+      "loss": 0.6323,
+      "step": 15487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.857017670166093e-06,
+      "loss": 0.6567,
+      "step": 15488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.855935948061513e-06,
+      "loss": 0.7013,
+      "step": 15489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.854854239521258e-06,
+      "loss": 0.6186,
+      "step": 15490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.853772544558153e-06,
+      "loss": 0.6125,
+      "step": 15491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.852690863185029e-06,
+      "loss": 0.6037,
+      "step": 15492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.851609195414702e-06,
+      "loss": 0.5744,
+      "step": 15493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.85052754126e-06,
+      "loss": 0.6525,
+      "step": 15494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.849445900733752e-06,
+      "loss": 0.63,
+      "step": 15495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.848364273848775e-06,
+      "loss": 0.6385,
+      "step": 15496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.847282660617895e-06,
+      "loss": 0.6516,
+      "step": 15497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.846201061053943e-06,
+      "loss": 0.599,
+      "step": 15498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.845119475169733e-06,
+      "loss": 0.6232,
+      "step": 15499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.844037902978091e-06,
+      "loss": 0.6556,
+      "step": 15500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.842956344491848e-06,
+      "loss": 0.6273,
+      "step": 15501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.841874799723816e-06,
+      "loss": 0.622,
+      "step": 15502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.840793268686822e-06,
+      "loss": 0.6295,
+      "step": 15503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.839711751393698e-06,
+      "loss": 0.603,
+      "step": 15504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.838630247857252e-06,
+      "loss": 0.597,
+      "step": 15505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.837548758090311e-06,
+      "loss": 0.6222,
+      "step": 15506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.83646728210571e-06,
+      "loss": 0.6117,
+      "step": 15507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.835385819916254e-06,
+      "loss": 0.623,
+      "step": 15508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.834304371534772e-06,
+      "loss": 0.6333,
+      "step": 15509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.833222936974091e-06,
+      "loss": 0.6381,
+      "step": 15510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.832141516247025e-06,
+      "loss": 0.5541,
+      "step": 15511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.831060109366397e-06,
+      "loss": 0.6512,
+      "step": 15512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.829978716345035e-06,
+      "loss": 0.633,
+      "step": 15513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.828897337195752e-06,
+      "loss": 0.6432,
+      "step": 15514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.82781597193137e-06,
+      "loss": 0.6365,
+      "step": 15515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.826734620564719e-06,
+      "loss": 0.6715,
+      "step": 15516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.82565328310861e-06,
+      "loss": 0.6399,
+      "step": 15517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.824571959575863e-06,
+      "loss": 0.608,
+      "step": 15518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.82349064997931e-06,
+      "loss": 0.6137,
+      "step": 15519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.82240935433176e-06,
+      "loss": 0.6201,
+      "step": 15520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.821328072646033e-06,
+      "loss": 0.6376,
+      "step": 15521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.820246804934959e-06,
+      "loss": 0.635,
+      "step": 15522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.819165551211349e-06,
+      "loss": 0.6351,
+      "step": 15523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.81808431148802e-06,
+      "loss": 0.6581,
+      "step": 15524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.817003085777807e-06,
+      "loss": 0.6331,
+      "step": 15525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.81592187409351e-06,
+      "loss": 0.6412,
+      "step": 15526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.814840676447964e-06,
+      "loss": 0.5772,
+      "step": 15527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.813759492853977e-06,
+      "loss": 0.6578,
+      "step": 15528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.812678323324371e-06,
+      "loss": 0.6269,
+      "step": 15529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.811597167871968e-06,
+      "loss": 0.6469,
+      "step": 15530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.810516026509582e-06,
+      "loss": 0.6241,
+      "step": 15531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.809434899250033e-06,
+      "loss": 0.6124,
+      "step": 15532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.808353786106142e-06,
+      "loss": 0.605,
+      "step": 15533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.807272687090722e-06,
+      "loss": 0.6254,
+      "step": 15534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.806191602216593e-06,
+      "loss": 0.602,
+      "step": 15535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.805110531496575e-06,
+      "loss": 0.6102,
+      "step": 15536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.804029474943481e-06,
+      "loss": 0.6566,
+      "step": 15537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.802948432570129e-06,
+      "loss": 0.6274,
+      "step": 15538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.80186740438934e-06,
+      "loss": 0.6492,
+      "step": 15539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.800786390413929e-06,
+      "loss": 0.6156,
+      "step": 15540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.799705390656711e-06,
+      "loss": 0.6948,
+      "step": 15541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.798624405130505e-06,
+      "loss": 0.6209,
+      "step": 15542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.797543433848125e-06,
+      "loss": 0.6697,
+      "step": 15543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.796462476822389e-06,
+      "loss": 0.5758,
+      "step": 15544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.795381534066117e-06,
+      "loss": 0.7098,
+      "step": 15545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.794300605592116e-06,
+      "loss": 0.6235,
+      "step": 15546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.793219691413206e-06,
+      "loss": 0.5821,
+      "step": 15547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.792138791542207e-06,
+      "loss": 0.6348,
+      "step": 15548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.791057905991928e-06,
+      "loss": 0.67,
+      "step": 15549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.789977034775186e-06,
+      "loss": 0.6272,
+      "step": 15550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.7888961779048e-06,
+      "loss": 0.6442,
+      "step": 15551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.787815335393578e-06,
+      "loss": 0.6569,
+      "step": 15552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.78673450725434e-06,
+      "loss": 0.6676,
+      "step": 15553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.785653693499902e-06,
+      "loss": 0.6672,
+      "step": 15554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.784572894143072e-06,
+      "loss": 0.5729,
+      "step": 15555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.78349210919667e-06,
+      "loss": 0.6646,
+      "step": 15556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.78241133867351e-06,
+      "loss": 0.6473,
+      "step": 15557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.781330582586399e-06,
+      "loss": 0.6444,
+      "step": 15558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.780249840948157e-06,
+      "loss": 0.6693,
+      "step": 15559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.7791691137716e-06,
+      "loss": 0.6201,
+      "step": 15560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.778088401069532e-06,
+      "loss": 0.6117,
+      "step": 15561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.777007702854773e-06,
+      "loss": 0.6222,
+      "step": 15562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.775927019140138e-06,
+      "loss": 0.6674,
+      "step": 15563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.774846349938433e-06,
+      "loss": 0.5967,
+      "step": 15564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.773765695262476e-06,
+      "loss": 0.6043,
+      "step": 15565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.77268505512508e-06,
+      "loss": 0.6439,
+      "step": 15566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.771604429539052e-06,
+      "loss": 0.611,
+      "step": 15567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.770523818517208e-06,
+      "loss": 0.5959,
+      "step": 15568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.769443222072361e-06,
+      "loss": 0.6297,
+      "step": 15569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.768362640217319e-06,
+      "loss": 0.6458,
+      "step": 15570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.767282072964896e-06,
+      "loss": 0.6917,
+      "step": 15571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.766201520327906e-06,
+      "loss": 0.6541,
+      "step": 15572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.765120982319155e-06,
+      "loss": 0.6698,
+      "step": 15573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.764040458951459e-06,
+      "loss": 0.6149,
+      "step": 15574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.762959950237627e-06,
+      "loss": 0.6609,
+      "step": 15575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.761879456190469e-06,
+      "loss": 0.6502,
+      "step": 15576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.760798976822795e-06,
+      "loss": 0.6498,
+      "step": 15577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.759718512147419e-06,
+      "loss": 0.6356,
+      "step": 15578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.758638062177146e-06,
+      "loss": 0.5795,
+      "step": 15579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.757557626924791e-06,
+      "loss": 0.6373,
+      "step": 15580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.756477206403163e-06,
+      "loss": 0.5713,
+      "step": 15581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.755396800625069e-06,
+      "loss": 0.5928,
+      "step": 15582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.75431640960332e-06,
+      "loss": 0.6188,
+      "step": 15583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.753236033350728e-06,
+      "loss": 0.5659,
+      "step": 15584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.752155671880097e-06,
+      "loss": 0.668,
+      "step": 15585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.751075325204241e-06,
+      "loss": 0.537,
+      "step": 15586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.749994993335969e-06,
+      "loss": 0.6796,
+      "step": 15587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.748914676288083e-06,
+      "loss": 0.6361,
+      "step": 15588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.747834374073401e-06,
+      "loss": 0.6252,
+      "step": 15589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.746754086704723e-06,
+      "loss": 0.6413,
+      "step": 15590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.74567381419486e-06,
+      "loss": 0.6617,
+      "step": 15591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.744593556556623e-06,
+      "loss": 0.6401,
+      "step": 15592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.743513313802818e-06,
+      "loss": 0.6155,
+      "step": 15593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.74243308594625e-06,
+      "loss": 0.6601,
+      "step": 15594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.74135287299973e-06,
+      "loss": 0.6114,
+      "step": 15595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.740272674976063e-06,
+      "loss": 0.6516,
+      "step": 15596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.739192491888056e-06,
+      "loss": 0.6603,
+      "step": 15597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.73811232374852e-06,
+      "loss": 0.639,
+      "step": 15598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.737032170570257e-06,
+      "loss": 0.6427,
+      "step": 15599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.735952032366074e-06,
+      "loss": 0.6691,
+      "step": 15600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.734871909148781e-06,
+      "loss": 0.6238,
+      "step": 15601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.73379180093118e-06,
+      "loss": 0.6004,
+      "step": 15602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.732711707726079e-06,
+      "loss": 0.6707,
+      "step": 15603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.731631629546287e-06,
+      "loss": 0.6348,
+      "step": 15604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.730551566404603e-06,
+      "loss": 0.6652,
+      "step": 15605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.729471518313835e-06,
+      "loss": 0.6212,
+      "step": 15606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.728391485286793e-06,
+      "loss": 0.6028,
+      "step": 15607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.727311467336277e-06,
+      "loss": 0.6347,
+      "step": 15608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.726231464475093e-06,
+      "loss": 0.6248,
+      "step": 15609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.725151476716048e-06,
+      "loss": 0.6426,
+      "step": 15610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.724071504071944e-06,
+      "loss": 0.645,
+      "step": 15611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.722991546555586e-06,
+      "loss": 0.6255,
+      "step": 15612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.72191160417978e-06,
+      "loss": 0.6277,
+      "step": 15613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.720831676957327e-06,
+      "loss": 0.6209,
+      "step": 15614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.719751764901033e-06,
+      "loss": 0.6597,
+      "step": 15615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.718671868023702e-06,
+      "loss": 0.6283,
+      "step": 15616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.717591986338137e-06,
+      "loss": 0.6005,
+      "step": 15617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.716512119857139e-06,
+      "loss": 0.6283,
+      "step": 15618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.715432268593516e-06,
+      "loss": 0.6166,
+      "step": 15619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.714352432560068e-06,
+      "loss": 0.6054,
+      "step": 15620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.713272611769597e-06,
+      "loss": 0.6574,
+      "step": 15621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.71219280623491e-06,
+      "loss": 0.6767,
+      "step": 15622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.711113015968801e-06,
+      "loss": 0.644,
+      "step": 15623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.710033240984082e-06,
+      "loss": 0.5871,
+      "step": 15624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.70895348129355e-06,
+      "loss": 0.6122,
+      "step": 15625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.707873736910007e-06,
+      "loss": 0.6601,
+      "step": 15626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.706794007846254e-06,
+      "loss": 0.6234,
+      "step": 15627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.705714294115098e-06,
+      "loss": 0.641,
+      "step": 15628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.704634595729333e-06,
+      "loss": 0.641,
+      "step": 15629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.703554912701763e-06,
+      "loss": 0.6203,
+      "step": 15630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.702475245045192e-06,
+      "loss": 0.6256,
+      "step": 15631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.701395592772417e-06,
+      "loss": 0.6474,
+      "step": 15632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.700315955896239e-06,
+      "loss": 0.6457,
+      "step": 15633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.699236334429463e-06,
+      "loss": 0.6453,
+      "step": 15634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.698156728384882e-06,
+      "loss": 0.6573,
+      "step": 15635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.6970771377753e-06,
+      "loss": 0.6261,
+      "step": 15636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.69599756261352e-06,
+      "loss": 0.5837,
+      "step": 15637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.694918002912336e-06,
+      "loss": 0.6068,
+      "step": 15638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.69383845868455e-06,
+      "loss": 0.6792,
+      "step": 15639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.692758929942964e-06,
+      "loss": 0.602,
+      "step": 15640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.691679416700372e-06,
+      "loss": 0.6159,
+      "step": 15641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.690599918969575e-06,
+      "loss": 0.642,
+      "step": 15642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.689520436763374e-06,
+      "loss": 0.6154,
+      "step": 15643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.688440970094565e-06,
+      "loss": 0.5988,
+      "step": 15644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.687361518975946e-06,
+      "loss": 0.6663,
+      "step": 15645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.68628208342032e-06,
+      "loss": 0.6037,
+      "step": 15646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.68520266344048e-06,
+      "loss": 0.6091,
+      "step": 15647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.684123259049225e-06,
+      "loss": 0.6516,
+      "step": 15648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.683043870259355e-06,
+      "loss": 0.6017,
+      "step": 15649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.681964497083664e-06,
+      "loss": 0.6163,
+      "step": 15650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.680885139534955e-06,
+      "loss": 0.6484,
+      "step": 15651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.679805797626017e-06,
+      "loss": 0.6376,
+      "step": 15652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.678726471369652e-06,
+      "loss": 0.6579,
+      "step": 15653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.677647160778661e-06,
+      "loss": 0.6515,
+      "step": 15654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.67656786586583e-06,
+      "loss": 0.6063,
+      "step": 15655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.675488586643961e-06,
+      "loss": 0.6476,
+      "step": 15656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.674409323125857e-06,
+      "loss": 0.6503,
+      "step": 15657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.6733300753243e-06,
+      "loss": 0.6663,
+      "step": 15658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.672250843252095e-06,
+      "loss": 0.6344,
+      "step": 15659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.671171626922039e-06,
+      "loss": 0.6377,
+      "step": 15660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.670092426346919e-06,
+      "loss": 0.6213,
+      "step": 15661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.669013241539538e-06,
+      "loss": 0.5963,
+      "step": 15662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.667934072512693e-06,
+      "loss": 0.6286,
+      "step": 15663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.666854919279168e-06,
+      "loss": 0.5959,
+      "step": 15664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.665775781851766e-06,
+      "loss": 0.5732,
+      "step": 15665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.664696660243286e-06,
+      "loss": 0.6336,
+      "step": 15666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.663617554466508e-06,
+      "loss": 0.638,
+      "step": 15667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.662538464534237e-06,
+      "loss": 0.6072,
+      "step": 15668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.66145939045927e-06,
+      "loss": 0.6638,
+      "step": 15669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.660380332254387e-06,
+      "loss": 0.5753,
+      "step": 15670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.659301289932393e-06,
+      "loss": 0.6247,
+      "step": 15671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.658222263506081e-06,
+      "loss": 0.6027,
+      "step": 15672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.657143252988236e-06,
+      "loss": 0.6444,
+      "step": 15673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.65606425839166e-06,
+      "loss": 0.5666,
+      "step": 15674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.654985279729144e-06,
+      "loss": 0.6273,
+      "step": 15675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.653906317013475e-06,
+      "loss": 0.6246,
+      "step": 15676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.65282737025745e-06,
+      "loss": 0.6605,
+      "step": 15677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.651748439473865e-06,
+      "loss": 0.6256,
+      "step": 15678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.650669524675501e-06,
+      "loss": 0.6191,
+      "step": 15679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.649590625875159e-06,
+      "loss": 0.5964,
+      "step": 15680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.648511743085635e-06,
+      "loss": 0.634,
+      "step": 15681
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.647432876319705e-06,
+      "loss": 0.6548,
+      "step": 15682
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.646354025590173e-06,
+      "loss": 0.5876,
+      "step": 15683
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.645275190909829e-06,
+      "loss": 0.6172,
+      "step": 15684
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.644196372291455e-06,
+      "loss": 0.6657,
+      "step": 15685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.643117569747852e-06,
+      "loss": 0.6644,
+      "step": 15686
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.642038783291809e-06,
+      "loss": 0.6204,
+      "step": 15687
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.640960012936109e-06,
+      "loss": 0.673,
+      "step": 15688
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.639881258693549e-06,
+      "loss": 0.6547,
+      "step": 15689
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.638802520576922e-06,
+      "loss": 0.6377,
+      "step": 15690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.637723798599005e-06,
+      "loss": 0.6608,
+      "step": 15691
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.6366450927726e-06,
+      "loss": 0.5818,
+      "step": 15692
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.635566403110494e-06,
+      "loss": 0.6213,
+      "step": 15693
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.634487729625471e-06,
+      "loss": 0.6766,
+      "step": 15694
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.633409072330323e-06,
+      "loss": 0.6245,
+      "step": 15695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.632330431237844e-06,
+      "loss": 0.6588,
+      "step": 15696
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.631251806360811e-06,
+      "loss": 0.6711,
+      "step": 15697
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.630173197712022e-06,
+      "loss": 0.6036,
+      "step": 15698
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.629094605304266e-06,
+      "loss": 0.6432,
+      "step": 15699
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.628016029150324e-06,
+      "loss": 0.6396,
+      "step": 15700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.626937469262987e-06,
+      "loss": 0.6006,
+      "step": 15701
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.625858925655049e-06,
+      "loss": 0.6151,
+      "step": 15702
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.624780398339284e-06,
+      "loss": 0.6058,
+      "step": 15703
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.62370188732849e-06,
+      "loss": 0.6721,
+      "step": 15704
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.622623392635456e-06,
+      "loss": 0.6449,
+      "step": 15705
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.621544914272956e-06,
+      "loss": 0.7221,
+      "step": 15706
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.620466452253788e-06,
+      "loss": 0.6216,
+      "step": 15707
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.61938800659074e-06,
+      "loss": 0.625,
+      "step": 15708
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.618309577296586e-06,
+      "loss": 0.6085,
+      "step": 15709
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.617231164384122e-06,
+      "loss": 0.6431,
+      "step": 15710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.616152767866137e-06,
+      "loss": 0.5791,
+      "step": 15711
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.615074387755404e-06,
+      "loss": 0.6237,
+      "step": 15712
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.613996024064724e-06,
+      "loss": 0.6466,
+      "step": 15713
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.612917676806869e-06,
+      "loss": 0.5954,
+      "step": 15714
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.611839345994629e-06,
+      "loss": 0.6751,
+      "step": 15715
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.610761031640794e-06,
+      "loss": 0.6463,
+      "step": 15716
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.609682733758141e-06,
+      "loss": 0.6346,
+      "step": 15717
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.608604452359455e-06,
+      "loss": 0.6062,
+      "step": 15718
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.60752618745753e-06,
+      "loss": 0.6283,
+      "step": 15719
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.606447939065142e-06,
+      "loss": 0.5451,
+      "step": 15720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.605369707195071e-06,
+      "loss": 0.6871,
+      "step": 15721
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.604291491860113e-06,
+      "loss": 0.643,
+      "step": 15722
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.603213293073042e-06,
+      "loss": 0.6363,
+      "step": 15723
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.602135110846642e-06,
+      "loss": 0.6239,
+      "step": 15724
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.601056945193706e-06,
+      "loss": 0.6266,
+      "step": 15725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.599978796127003e-06,
+      "loss": 0.664,
+      "step": 15726
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.598900663659321e-06,
+      "loss": 0.6396,
+      "step": 15727
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.597822547803452e-06,
+      "loss": 0.6182,
+      "step": 15728
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.596744448572164e-06,
+      "loss": 0.6661,
+      "step": 15729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.595666365978244e-06,
+      "loss": 0.6473,
+      "step": 15730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.594588300034484e-06,
+      "loss": 0.6596,
+      "step": 15731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.593510250753652e-06,
+      "loss": 0.5922,
+      "step": 15732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.592432218148531e-06,
+      "loss": 0.5765,
+      "step": 15733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.591354202231916e-06,
+      "loss": 0.6367,
+      "step": 15734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.590276203016575e-06,
+      "loss": 0.6733,
+      "step": 15735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.589198220515289e-06,
+      "loss": 0.6588,
+      "step": 15736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.588120254740849e-06,
+      "loss": 0.6834,
+      "step": 15737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.587042305706026e-06,
+      "loss": 0.625,
+      "step": 15738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.585964373423601e-06,
+      "loss": 0.6367,
+      "step": 15739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.584886457906365e-06,
+      "loss": 0.6979,
+      "step": 15740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.583808559167085e-06,
+      "loss": 0.5661,
+      "step": 15741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.582730677218544e-06,
+      "loss": 0.6691,
+      "step": 15742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.581652812073529e-06,
+      "loss": 0.6268,
+      "step": 15743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.580574963744812e-06,
+      "loss": 0.6453,
+      "step": 15744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.57949713224517e-06,
+      "loss": 0.5716,
+      "step": 15745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.578419317587395e-06,
+      "loss": 0.6082,
+      "step": 15746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.577341519784252e-06,
+      "loss": 0.6472,
+      "step": 15747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.576263738848522e-06,
+      "loss": 0.6141,
+      "step": 15748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.575185974792993e-06,
+      "loss": 0.6638,
+      "step": 15749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.574108227630433e-06,
+      "loss": 0.5777,
+      "step": 15750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.573030497373623e-06,
+      "loss": 0.6617,
+      "step": 15751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.571952784035345e-06,
+      "loss": 0.6612,
+      "step": 15752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.57087508762837e-06,
+      "loss": 0.6321,
+      "step": 15753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.569797408165478e-06,
+      "loss": 0.6374,
+      "step": 15754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.568719745659451e-06,
+      "loss": 0.6129,
+      "step": 15755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.567642100123061e-06,
+      "loss": 0.561,
+      "step": 15756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.566564471569081e-06,
+      "loss": 0.6191,
+      "step": 15757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.5654868600103e-06,
+      "loss": 0.6621,
+      "step": 15758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.564409265459482e-06,
+      "loss": 0.6083,
+      "step": 15759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.563331687929405e-06,
+      "loss": 0.6476,
+      "step": 15760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.562254127432858e-06,
+      "loss": 0.6526,
+      "step": 15761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.5611765839826e-06,
+      "loss": 0.6219,
+      "step": 15762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.560099057591414e-06,
+      "loss": 0.6449,
+      "step": 15763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.55902154827208e-06,
+      "loss": 0.6518,
+      "step": 15764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.557944056037362e-06,
+      "loss": 0.577,
+      "step": 15765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.556866580900043e-06,
+      "loss": 0.6697,
+      "step": 15766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.5557891228729e-06,
+      "loss": 0.6264,
+      "step": 15767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.5547116819687e-06,
+      "loss": 0.6008,
+      "step": 15768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.55363425820022e-06,
+      "loss": 0.6045,
+      "step": 15769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.552556851580243e-06,
+      "loss": 0.6219,
+      "step": 15770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.55147946212153e-06,
+      "loss": 0.6259,
+      "step": 15771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.550402089836857e-06,
+      "loss": 0.6619,
+      "step": 15772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.549324734739011e-06,
+      "loss": 0.653,
+      "step": 15773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.548247396840747e-06,
+      "loss": 0.6315,
+      "step": 15774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.547170076154852e-06,
+      "loss": 0.6329,
+      "step": 15775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.54609277269409e-06,
+      "loss": 0.6899,
+      "step": 15776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.545015486471238e-06,
+      "loss": 0.6467,
+      "step": 15777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.543938217499069e-06,
+      "loss": 0.6104,
+      "step": 15778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.542860965790351e-06,
+      "loss": 0.671,
+      "step": 15779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.541783731357863e-06,
+      "loss": 0.6216,
+      "step": 15780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.540706514214374e-06,
+      "loss": 0.627,
+      "step": 15781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.539629314372653e-06,
+      "loss": 0.622,
+      "step": 15782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.538552131845473e-06,
+      "loss": 0.6355,
+      "step": 15783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.537474966645609e-06,
+      "loss": 0.6662,
+      "step": 15784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.536397818785829e-06,
+      "loss": 0.6795,
+      "step": 15785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.535320688278902e-06,
+      "loss": 0.6578,
+      "step": 15786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.534243575137604e-06,
+      "loss": 0.6532,
+      "step": 15787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.5331664793747e-06,
+      "loss": 0.6729,
+      "step": 15788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.532089401002964e-06,
+      "loss": 0.6336,
+      "step": 15789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.531012340035166e-06,
+      "loss": 0.6393,
+      "step": 15790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.529935296484075e-06,
+      "loss": 0.5994,
+      "step": 15791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.528858270362458e-06,
+      "loss": 0.6631,
+      "step": 15792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.527781261683092e-06,
+      "loss": 0.6126,
+      "step": 15793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.526704270458737e-06,
+      "loss": 0.6838,
+      "step": 15794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.525627296702168e-06,
+      "loss": 0.6731,
+      "step": 15795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.524550340426154e-06,
+      "loss": 0.6304,
+      "step": 15796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.523473401643459e-06,
+      "loss": 0.5694,
+      "step": 15797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.522396480366857e-06,
+      "loss": 0.6113,
+      "step": 15798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.521319576609115e-06,
+      "loss": 0.6927,
+      "step": 15799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.520242690382998e-06,
+      "loss": 0.6473,
+      "step": 15800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.519165821701274e-06,
+      "loss": 0.6611,
+      "step": 15801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.518088970576717e-06,
+      "loss": 0.6794,
+      "step": 15802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.517012137022087e-06,
+      "loss": 0.6254,
+      "step": 15803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.515935321050154e-06,
+      "loss": 0.6101,
+      "step": 15804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.514858522673686e-06,
+      "loss": 0.612,
+      "step": 15805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.51378174190545e-06,
+      "loss": 0.6594,
+      "step": 15806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.512704978758209e-06,
+      "loss": 0.5907,
+      "step": 15807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.511628233244733e-06,
+      "loss": 0.6827,
+      "step": 15808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.510551505377787e-06,
+      "loss": 0.6486,
+      "step": 15809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.509474795170136e-06,
+      "loss": 0.5388,
+      "step": 15810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.508398102634551e-06,
+      "loss": 0.6365,
+      "step": 15811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.50732142778379e-06,
+      "loss": 0.6526,
+      "step": 15812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.50624477063062e-06,
+      "loss": 0.63,
+      "step": 15813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.505168131187813e-06,
+      "loss": 0.6436,
+      "step": 15814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.504091509468125e-06,
+      "loss": 0.6645,
+      "step": 15815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.503014905484324e-06,
+      "loss": 0.5861,
+      "step": 15816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.501938319249177e-06,
+      "loss": 0.6387,
+      "step": 15817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.500861750775446e-06,
+      "loss": 0.6538,
+      "step": 15818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.499785200075894e-06,
+      "loss": 0.5885,
+      "step": 15819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.498708667163289e-06,
+      "loss": 0.6387,
+      "step": 15820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.497632152050389e-06,
+      "loss": 0.6498,
+      "step": 15821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.49655565474996e-06,
+      "loss": 0.6868,
+      "step": 15822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.495479175274769e-06,
+      "loss": 0.64,
+      "step": 15823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.494402713637572e-06,
+      "loss": 0.6602,
+      "step": 15824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.493326269851138e-06,
+      "loss": 0.6231,
+      "step": 15825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.492249843928228e-06,
+      "loss": 0.7122,
+      "step": 15826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.4911734358816e-06,
+      "loss": 0.6656,
+      "step": 15827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.490097045724021e-06,
+      "loss": 0.6266,
+      "step": 15828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.489020673468252e-06,
+      "loss": 0.587,
+      "step": 15829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.487944319127053e-06,
+      "loss": 0.6208,
+      "step": 15830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.486867982713187e-06,
+      "loss": 0.591,
+      "step": 15831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.485791664239418e-06,
+      "loss": 0.6869,
+      "step": 15832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.484715363718502e-06,
+      "loss": 0.6482,
+      "step": 15833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.483639081163201e-06,
+      "loss": 0.6031,
+      "step": 15834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.48256281658628e-06,
+      "loss": 0.677,
+      "step": 15835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.481486570000495e-06,
+      "loss": 0.6722,
+      "step": 15836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.480410341418608e-06,
+      "loss": 0.6004,
+      "step": 15837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.479334130853378e-06,
+      "loss": 0.6116,
+      "step": 15838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.478257938317565e-06,
+      "loss": 0.6488,
+      "step": 15839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.477181763823931e-06,
+      "loss": 0.6708,
+      "step": 15840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.476105607385233e-06,
+      "loss": 0.5877,
+      "step": 15841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.475029469014229e-06,
+      "loss": 0.6267,
+      "step": 15842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.473953348723682e-06,
+      "loss": 0.639,
+      "step": 15843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.472877246526346e-06,
+      "loss": 0.6632,
+      "step": 15844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.471801162434983e-06,
+      "loss": 0.6795,
+      "step": 15845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.470725096462352e-06,
+      "loss": 0.6667,
+      "step": 15846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.469649048621207e-06,
+      "loss": 0.6775,
+      "step": 15847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.46857301892431e-06,
+      "loss": 0.6389,
+      "step": 15848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.467497007384418e-06,
+      "loss": 0.6527,
+      "step": 15849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.466421014014285e-06,
+      "loss": 0.6187,
+      "step": 15850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.465345038826674e-06,
+      "loss": 0.6733,
+      "step": 15851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.464269081834338e-06,
+      "loss": 0.616,
+      "step": 15852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.463193143050032e-06,
+      "loss": 0.6146,
+      "step": 15853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.462117222486519e-06,
+      "loss": 0.6763,
+      "step": 15854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.46104132015655e-06,
+      "loss": 0.6907,
+      "step": 15855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.459965436072883e-06,
+      "loss": 0.6599,
+      "step": 15856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.458889570248273e-06,
+      "loss": 0.624,
+      "step": 15857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.45781372269548e-06,
+      "loss": 0.6506,
+      "step": 15858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.456737893427252e-06,
+      "loss": 0.6315,
+      "step": 15859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.45566208245635e-06,
+      "loss": 0.6149,
+      "step": 15860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.45458628979553e-06,
+      "loss": 0.6394,
+      "step": 15861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.453510515457542e-06,
+      "loss": 0.621,
+      "step": 15862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.452434759455143e-06,
+      "loss": 0.6496,
+      "step": 15863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.45135902180109e-06,
+      "loss": 0.5886,
+      "step": 15864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.450283302508132e-06,
+      "loss": 0.6263,
+      "step": 15865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.449207601589026e-06,
+      "loss": 0.6391,
+      "step": 15866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.448131919056527e-06,
+      "loss": 0.6383,
+      "step": 15867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.447056254923387e-06,
+      "loss": 0.6051,
+      "step": 15868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.445980609202359e-06,
+      "loss": 0.6034,
+      "step": 15869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.444904981906199e-06,
+      "loss": 0.5929,
+      "step": 15870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.443829373047653e-06,
+      "loss": 0.6378,
+      "step": 15871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.442753782639481e-06,
+      "loss": 0.6822,
+      "step": 15872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.441678210694435e-06,
+      "loss": 0.6443,
+      "step": 15873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.440602657225263e-06,
+      "loss": 0.611,
+      "step": 15874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.439527122244716e-06,
+      "loss": 0.6417,
+      "step": 15875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.438451605765555e-06,
+      "loss": 0.5969,
+      "step": 15876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.437376107800522e-06,
+      "loss": 0.621,
+      "step": 15877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.436300628362374e-06,
+      "loss": 0.6406,
+      "step": 15878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.435225167463859e-06,
+      "loss": 0.6461,
+      "step": 15879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.434149725117728e-06,
+      "loss": 0.6212,
+      "step": 15880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.433074301336732e-06,
+      "loss": 0.6502,
+      "step": 15881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.431998896133626e-06,
+      "loss": 0.6007,
+      "step": 15882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.430923509521155e-06,
+      "loss": 0.6513,
+      "step": 15883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.42984814151207e-06,
+      "loss": 0.5828,
+      "step": 15884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.428772792119123e-06,
+      "loss": 0.6651,
+      "step": 15885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.42769746135506e-06,
+      "loss": 0.6231,
+      "step": 15886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.426622149232633e-06,
+      "loss": 0.628,
+      "step": 15887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.425546855764593e-06,
+      "loss": 0.6443,
+      "step": 15888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.424471580963684e-06,
+      "loss": 0.5774,
+      "step": 15889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.423396324842658e-06,
+      "loss": 0.6123,
+      "step": 15890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.422321087414266e-06,
+      "loss": 0.6148,
+      "step": 15891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.42124586869125e-06,
+      "loss": 0.6827,
+      "step": 15892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.420170668686361e-06,
+      "loss": 0.6238,
+      "step": 15893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.41909548741235e-06,
+      "loss": 0.6401,
+      "step": 15894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.41802032488196e-06,
+      "loss": 0.6241,
+      "step": 15895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.41694518110794e-06,
+      "loss": 0.645,
+      "step": 15896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.415870056103039e-06,
+      "loss": 0.5522,
+      "step": 15897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.41479494988e-06,
+      "loss": 0.5968,
+      "step": 15898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.413719862451579e-06,
+      "loss": 0.6363,
+      "step": 15899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.412644793830506e-06,
+      "loss": 0.6316,
+      "step": 15900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.41156974402954e-06,
+      "loss": 0.6144,
+      "step": 15901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.41049471306143e-06,
+      "loss": 0.6119,
+      "step": 15902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.409419700938908e-06,
+      "loss": 0.6391,
+      "step": 15903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.40834470767473e-06,
+      "loss": 0.6575,
+      "step": 15904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.407269733281644e-06,
+      "loss": 0.6259,
+      "step": 15905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.406194777772382e-06,
+      "loss": 0.6205,
+      "step": 15906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.405119841159701e-06,
+      "loss": 0.596,
+      "step": 15907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.404044923456346e-06,
+      "loss": 0.6377,
+      "step": 15908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.402970024675051e-06,
+      "loss": 0.6009,
+      "step": 15909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.401895144828566e-06,
+      "loss": 0.6416,
+      "step": 15910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.400820283929642e-06,
+      "loss": 0.6816,
+      "step": 15911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.399745441991011e-06,
+      "loss": 0.6945,
+      "step": 15912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.398670619025423e-06,
+      "loss": 0.7075,
+      "step": 15913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.397595815045624e-06,
+      "loss": 0.6292,
+      "step": 15914
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.39652103006435e-06,
+      "loss": 0.6116,
+      "step": 15915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.395446264094346e-06,
+      "loss": 0.6145,
+      "step": 15916
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.394371517148364e-06,
+      "loss": 0.6521,
+      "step": 15917
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.39329678923913e-06,
+      "loss": 0.6437,
+      "step": 15918
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.392222080379399e-06,
+      "loss": 0.6439,
+      "step": 15919
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.391147390581914e-06,
+      "loss": 0.6211,
+      "step": 15920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.390072719859403e-06,
+      "loss": 0.6467,
+      "step": 15921
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.38899806822462e-06,
+      "loss": 0.6532,
+      "step": 15922
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.387923435690307e-06,
+      "loss": 0.6378,
+      "step": 15923
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.386848822269194e-06,
+      "loss": 0.6793,
+      "step": 15924
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.385774227974031e-06,
+      "loss": 0.6021,
+      "step": 15925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.38469965281756e-06,
+      "loss": 0.6557,
+      "step": 15926
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.383625096812514e-06,
+      "loss": 0.612,
+      "step": 15927
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.382550559971638e-06,
+      "loss": 0.5951,
+      "step": 15928
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.381476042307677e-06,
+      "loss": 0.6479,
+      "step": 15929
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.380401543833355e-06,
+      "loss": 0.6312,
+      "step": 15930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.379327064561427e-06,
+      "loss": 0.5972,
+      "step": 15931
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.378252604504629e-06,
+      "loss": 0.558,
+      "step": 15932
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.377178163675691e-06,
+      "loss": 0.6214,
+      "step": 15933
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.376103742087363e-06,
+      "loss": 0.5856,
+      "step": 15934
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.375029339752382e-06,
+      "loss": 0.6308,
+      "step": 15935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.373954956683479e-06,
+      "loss": 0.6608,
+      "step": 15936
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.3728805928934e-06,
+      "loss": 0.5926,
+      "step": 15937
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.371806248394882e-06,
+      "loss": 0.6344,
+      "step": 15938
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.370731923200653e-06,
+      "loss": 0.5892,
+      "step": 15939
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.369657617323463e-06,
+      "loss": 0.6765,
+      "step": 15940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.368583330776047e-06,
+      "loss": 0.6044,
+      "step": 15941
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.367509063571133e-06,
+      "loss": 0.6431,
+      "step": 15942
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.366434815721468e-06,
+      "loss": 0.6296,
+      "step": 15943
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.365360587239787e-06,
+      "loss": 0.6366,
+      "step": 15944
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.364286378138819e-06,
+      "loss": 0.6565,
+      "step": 15945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.363212188431306e-06,
+      "loss": 0.6141,
+      "step": 15946
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.362138018129988e-06,
+      "loss": 0.6519,
+      "step": 15947
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.36106386724759e-06,
+      "loss": 0.6446,
+      "step": 15948
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.359989735796855e-06,
+      "loss": 0.6267,
+      "step": 15949
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.358915623790519e-06,
+      "loss": 0.6187,
+      "step": 15950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.35784153124131e-06,
+      "loss": 0.6135,
+      "step": 15951
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.356767458161967e-06,
+      "loss": 0.6262,
+      "step": 15952
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.355693404565229e-06,
+      "loss": 0.6099,
+      "step": 15953
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.354619370463822e-06,
+      "loss": 0.6406,
+      "step": 15954
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.353545355870483e-06,
+      "loss": 0.5869,
+      "step": 15955
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.352471360797952e-06,
+      "loss": 0.602,
+      "step": 15956
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.351397385258951e-06,
+      "loss": 0.6008,
+      "step": 15957
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.350323429266221e-06,
+      "loss": 0.6297,
+      "step": 15958
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.349249492832499e-06,
+      "loss": 0.5861,
+      "step": 15959
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.348175575970504e-06,
+      "loss": 0.6269,
+      "step": 15960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.347101678692985e-06,
+      "loss": 0.6461,
+      "step": 15961
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.346027801012661e-06,
+      "loss": 0.6479,
+      "step": 15962
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.34495394294227e-06,
+      "loss": 0.613,
+      "step": 15963
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.343880104494548e-06,
+      "loss": 0.6238,
+      "step": 15964
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.34280628568222e-06,
+      "loss": 0.6095,
+      "step": 15965
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.341732486518016e-06,
+      "loss": 0.6475,
+      "step": 15966
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.340658707014677e-06,
+      "loss": 0.6012,
+      "step": 15967
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.339584947184925e-06,
+      "loss": 0.616,
+      "step": 15968
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.338511207041491e-06,
+      "loss": 0.6296,
+      "step": 15969
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.337437486597117e-06,
+      "loss": 0.584,
+      "step": 15970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.336363785864518e-06,
+      "loss": 0.6468,
+      "step": 15971
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.33529010485643e-06,
+      "loss": 0.639,
+      "step": 15972
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.33421644358559e-06,
+      "loss": 0.62,
+      "step": 15973
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.333142802064717e-06,
+      "loss": 0.6378,
+      "step": 15974
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.332069180306541e-06,
+      "loss": 0.6866,
+      "step": 15975
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.330995578323803e-06,
+      "loss": 0.666,
+      "step": 15976
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.329921996129219e-06,
+      "loss": 0.6148,
+      "step": 15977
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.32884843373552e-06,
+      "loss": 0.641,
+      "step": 15978
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.327774891155443e-06,
+      "loss": 0.6619,
+      "step": 15979
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.326701368401706e-06,
+      "loss": 0.6351,
+      "step": 15980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.32562786548704e-06,
+      "loss": 0.6452,
+      "step": 15981
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.324554382424179e-06,
+      "loss": 0.6608,
+      "step": 15982
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.323480919225841e-06,
+      "loss": 0.6763,
+      "step": 15983
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.322407475904754e-06,
+      "loss": 0.6136,
+      "step": 15984
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.321334052473658e-06,
+      "loss": 0.6196,
+      "step": 15985
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.320260648945264e-06,
+      "loss": 0.59,
+      "step": 15986
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.319187265332304e-06,
+      "loss": 0.6217,
+      "step": 15987
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.31811390164751e-06,
+      "loss": 0.6002,
+      "step": 15988
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.3170405579036e-06,
+      "loss": 0.5791,
+      "step": 15989
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.3159672341133e-06,
+      "loss": 0.5553,
+      "step": 15990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.314893930289347e-06,
+      "loss": 0.6886,
+      "step": 15991
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.313820646444454e-06,
+      "loss": 0.6047,
+      "step": 15992
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.312747382591346e-06,
+      "loss": 0.6964,
+      "step": 15993
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.31167413874276e-06,
+      "loss": 0.6396,
+      "step": 15994
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.31060091491141e-06,
+      "loss": 0.6077,
+      "step": 15995
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.309527711110017e-06,
+      "loss": 0.6124,
+      "step": 15996
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.308454527351321e-06,
+      "loss": 0.6438,
+      "step": 15997
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.30738136364803e-06,
+      "loss": 0.6348,
+      "step": 15998
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.306308220012877e-06,
+      "loss": 0.656,
+      "step": 15999
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.305235096458583e-06,
+      "loss": 0.6023,
+      "step": 16000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.304161992997869e-06,
+      "loss": 0.606,
+      "step": 16001
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.303088909643458e-06,
+      "loss": 0.6243,
+      "step": 16002
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.30201584640808e-06,
+      "loss": 0.6446,
+      "step": 16003
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.300942803304448e-06,
+      "loss": 0.6267,
+      "step": 16004
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.29986978034529e-06,
+      "loss": 0.6406,
+      "step": 16005
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.298796777543325e-06,
+      "loss": 0.6834,
+      "step": 16006
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.297723794911276e-06,
+      "loss": 0.6936,
+      "step": 16007
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.296650832461866e-06,
+      "loss": 0.6098,
+      "step": 16008
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.295577890207817e-06,
+      "loss": 0.6341,
+      "step": 16009
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.294504968161844e-06,
+      "loss": 0.631,
+      "step": 16010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.293432066336675e-06,
+      "loss": 0.6008,
+      "step": 16011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.292359184745028e-06,
+      "loss": 0.6289,
+      "step": 16012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.291286323399622e-06,
+      "loss": 0.6357,
+      "step": 16013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.290213482313177e-06,
+      "loss": 0.6316,
+      "step": 16014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.289140661498416e-06,
+      "loss": 0.5915,
+      "step": 16015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.288067860968055e-06,
+      "loss": 0.5847,
+      "step": 16016
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.286995080734817e-06,
+      "loss": 0.7003,
+      "step": 16017
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.28592232081142e-06,
+      "loss": 0.6289,
+      "step": 16018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.28484958121058e-06,
+      "loss": 0.667,
+      "step": 16019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.283776861945016e-06,
+      "loss": 0.6516,
+      "step": 16020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.282704163027453e-06,
+      "loss": 0.6496,
+      "step": 16021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.2816314844706e-06,
+      "loss": 0.6157,
+      "step": 16022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.280558826287184e-06,
+      "loss": 0.6098,
+      "step": 16023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.279486188489916e-06,
+      "loss": 0.6017,
+      "step": 16024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.278413571091515e-06,
+      "loss": 0.622,
+      "step": 16025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.2773409741047e-06,
+      "loss": 0.593,
+      "step": 16026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.276268397542186e-06,
+      "loss": 0.6061,
+      "step": 16027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.275195841416692e-06,
+      "loss": 0.5843,
+      "step": 16028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.274123305740934e-06,
+      "loss": 0.5989,
+      "step": 16029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.273050790527624e-06,
+      "loss": 0.6095,
+      "step": 16030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.271978295789482e-06,
+      "loss": 0.6377,
+      "step": 16031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.270905821539227e-06,
+      "loss": 0.6519,
+      "step": 16032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.269833367789566e-06,
+      "loss": 0.66,
+      "step": 16033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.268760934553221e-06,
+      "loss": 0.5921,
+      "step": 16034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.267688521842909e-06,
+      "loss": 0.6414,
+      "step": 16035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.266616129671337e-06,
+      "loss": 0.6318,
+      "step": 16036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.265543758051221e-06,
+      "loss": 0.6414,
+      "step": 16037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.264471406995285e-06,
+      "loss": 0.6676,
+      "step": 16038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.26339907651623e-06,
+      "loss": 0.6371,
+      "step": 16039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.262326766626778e-06,
+      "loss": 0.6064,
+      "step": 16040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.261254477339642e-06,
+      "loss": 0.6559,
+      "step": 16041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.26018220866753e-06,
+      "loss": 0.6175,
+      "step": 16042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.259109960623159e-06,
+      "loss": 0.648,
+      "step": 16043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.258037733219245e-06,
+      "loss": 0.6477,
+      "step": 16044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.256965526468495e-06,
+      "loss": 0.5611,
+      "step": 16045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.255893340383623e-06,
+      "loss": 0.5739,
+      "step": 16046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.254821174977344e-06,
+      "loss": 0.6173,
+      "step": 16047
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.253749030262366e-06,
+      "loss": 0.6672,
+      "step": 16048
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.252676906251402e-06,
+      "loss": 0.6431,
+      "step": 16049
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.251604802957166e-06,
+      "loss": 0.6154,
+      "step": 16050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.250532720392364e-06,
+      "loss": 0.5979,
+      "step": 16051
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.24946065856971e-06,
+      "loss": 0.7005,
+      "step": 16052
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.248388617501917e-06,
+      "loss": 0.6006,
+      "step": 16053
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.247316597201691e-06,
+      "loss": 0.6,
+      "step": 16054
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.246244597681743e-06,
+      "loss": 0.6235,
+      "step": 16055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.245172618954786e-06,
+      "loss": 0.6438,
+      "step": 16056
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.244100661033526e-06,
+      "loss": 0.6333,
+      "step": 16057
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.243028723930673e-06,
+      "loss": 0.635,
+      "step": 16058
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.241956807658939e-06,
+      "loss": 0.6223,
+      "step": 16059
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.24088491223103e-06,
+      "loss": 0.5888,
+      "step": 16060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.239813037659655e-06,
+      "loss": 0.6142,
+      "step": 16061
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.238741183957525e-06,
+      "loss": 0.6352,
+      "step": 16062
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.237669351137343e-06,
+      "loss": 0.6082,
+      "step": 16063
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.236597539211823e-06,
+      "loss": 0.648,
+      "step": 16064
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.23552574819367e-06,
+      "loss": 0.599,
+      "step": 16065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.23445397809559e-06,
+      "loss": 0.661,
+      "step": 16066
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.23338222893029e-06,
+      "loss": 0.5972,
+      "step": 16067
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.232310500710482e-06,
+      "loss": 0.5544,
+      "step": 16068
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.231238793448868e-06,
+      "loss": 0.6713,
+      "step": 16069
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.230167107158154e-06,
+      "loss": 0.6246,
+      "step": 16070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.22909544185105e-06,
+      "loss": 0.6547,
+      "step": 16071
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.22802379754026e-06,
+      "loss": 0.6382,
+      "step": 16072
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.226952174238487e-06,
+      "loss": 0.5706,
+      "step": 16073
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.225880571958441e-06,
+      "loss": 0.6091,
+      "step": 16074
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.224808990712823e-06,
+      "loss": 0.6164,
+      "step": 16075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.22373743051434e-06,
+      "loss": 0.6524,
+      "step": 16076
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.2226658913757e-06,
+      "loss": 0.6246,
+      "step": 16077
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.221594373309602e-06,
+      "loss": 0.616,
+      "step": 16078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.220522876328751e-06,
+      "loss": 0.5527,
+      "step": 16079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.219451400445857e-06,
+      "loss": 0.6284,
+      "step": 16080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.218379945673613e-06,
+      "loss": 0.6334,
+      "step": 16081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.217308512024732e-06,
+      "loss": 0.631,
+      "step": 16082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.216237099511915e-06,
+      "loss": 0.6122,
+      "step": 16083
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.21516570814786e-06,
+      "loss": 0.6206,
+      "step": 16084
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.214094337945277e-06,
+      "loss": 0.6795,
+      "step": 16085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.213022988916862e-06,
+      "loss": 0.6296,
+      "step": 16086
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.211951661075322e-06,
+      "loss": 0.6439,
+      "step": 16087
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.210880354433356e-06,
+      "loss": 0.6588,
+      "step": 16088
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.209809069003666e-06,
+      "loss": 0.592,
+      "step": 16089
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.208737804798954e-06,
+      "loss": 0.6018,
+      "step": 16090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.207666561831924e-06,
+      "loss": 0.6516,
+      "step": 16091
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.20659534011527e-06,
+      "loss": 0.658,
+      "step": 16092
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.2055241396617e-06,
+      "loss": 0.6664,
+      "step": 16093
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.204452960483911e-06,
+      "loss": 0.6111,
+      "step": 16094
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.203381802594602e-06,
+      "loss": 0.6254,
+      "step": 16095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.202310666006475e-06,
+      "loss": 0.6335,
+      "step": 16096
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.201239550732232e-06,
+      "loss": 0.5991,
+      "step": 16097
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.200168456784566e-06,
+      "loss": 0.5647,
+      "step": 16098
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.19909738417618e-06,
+      "loss": 0.653,
+      "step": 16099
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.198026332919776e-06,
+      "loss": 0.6257,
+      "step": 16100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.196955303028044e-06,
+      "loss": 0.6135,
+      "step": 16101
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.195884294513691e-06,
+      "loss": 0.6582,
+      "step": 16102
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.194813307389412e-06,
+      "loss": 0.6441,
+      "step": 16103
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.193742341667906e-06,
+      "loss": 0.6418,
+      "step": 16104
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.192671397361866e-06,
+      "loss": 0.628,
+      "step": 16105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.191600474483996e-06,
+      "loss": 0.6289,
+      "step": 16106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.190529573046988e-06,
+      "loss": 0.6397,
+      "step": 16107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.189458693063542e-06,
+      "loss": 0.5768,
+      "step": 16108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.188387834546352e-06,
+      "loss": 0.6278,
+      "step": 16109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.187316997508117e-06,
+      "loss": 0.6272,
+      "step": 16110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.186246181961533e-06,
+      "loss": 0.5775,
+      "step": 16111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.185175387919295e-06,
+      "loss": 0.6359,
+      "step": 16112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.184104615394096e-06,
+      "loss": 0.6186,
+      "step": 16113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.183033864398635e-06,
+      "loss": 0.6593,
+      "step": 16114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.181963134945609e-06,
+      "loss": 0.6414,
+      "step": 16115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.180892427047706e-06,
+      "loss": 0.6862,
+      "step": 16116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.179821740717625e-06,
+      "loss": 0.571,
+      "step": 16117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.178751075968062e-06,
+      "loss": 0.5926,
+      "step": 16118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.177680432811706e-06,
+      "loss": 0.6377,
+      "step": 16119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.176609811261255e-06,
+      "loss": 0.6679,
+      "step": 16120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.175539211329403e-06,
+      "loss": 0.6573,
+      "step": 16121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.17446863302884e-06,
+      "loss": 0.6244,
+      "step": 16122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.17339807637226e-06,
+      "loss": 0.6355,
+      "step": 16123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.172327541372359e-06,
+      "loss": 0.6316,
+      "step": 16124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.171257028041824e-06,
+      "loss": 0.6,
+      "step": 16125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.170186536393352e-06,
+      "loss": 0.632,
+      "step": 16126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.169116066439633e-06,
+      "loss": 0.675,
+      "step": 16127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.16804561819336e-06,
+      "loss": 0.6468,
+      "step": 16128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.16697519166722e-06,
+      "loss": 0.6696,
+      "step": 16129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.165904786873913e-06,
+      "loss": 0.6415,
+      "step": 16130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.16483440382612e-06,
+      "loss": 0.5946,
+      "step": 16131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.163764042536539e-06,
+      "loss": 0.6012,
+      "step": 16132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.16269370301786e-06,
+      "loss": 0.6395,
+      "step": 16133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.161623385282768e-06,
+      "loss": 0.6408,
+      "step": 16134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.160553089343959e-06,
+      "loss": 0.6101,
+      "step": 16135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.159482815214118e-06,
+      "loss": 0.6463,
+      "step": 16136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.158412562905939e-06,
+      "loss": 0.6316,
+      "step": 16137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157342332432106e-06,
+      "loss": 0.6105,
+      "step": 16138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.156272123805314e-06,
+      "loss": 0.5959,
+      "step": 16139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155201937038244e-06,
+      "loss": 0.5968,
+      "step": 16140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15413177214359e-06,
+      "loss": 0.6655,
+      "step": 16141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153061629134042e-06,
+      "loss": 0.5845,
+      "step": 16142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15199150802228e-06,
+      "loss": 0.6192,
+      "step": 16143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.150921408820998e-06,
+      "loss": 0.631,
+      "step": 16144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.149851331542886e-06,
+      "loss": 0.6561,
+      "step": 16145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.14878127620062e-06,
+      "loss": 0.6287,
+      "step": 16146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.147711242806902e-06,
+      "loss": 0.6151,
+      "step": 16147
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.146641231374403e-06,
+      "loss": 0.5951,
+      "step": 16148
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.14557124191582e-06,
+      "loss": 0.6946,
+      "step": 16149
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.144501274443838e-06,
+      "loss": 0.6447,
+      "step": 16150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.143431328971136e-06,
+      "loss": 0.6163,
+      "step": 16151
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.142361405510403e-06,
+      "loss": 0.5746,
+      "step": 16152
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.141291504074333e-06,
+      "loss": 0.6929,
+      "step": 16153
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.140221624675595e-06,
+      "loss": 0.5798,
+      "step": 16154
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.139151767326887e-06,
+      "loss": 0.6721,
+      "step": 16155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.138081932040892e-06,
+      "loss": 0.5891,
+      "step": 16156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.137012118830282e-06,
+      "loss": 0.5941,
+      "step": 16157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.135942327707755e-06,
+      "loss": 0.6498,
+      "step": 16158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.134872558685994e-06,
+      "loss": 0.6388,
+      "step": 16159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.133802811777671e-06,
+      "loss": 0.6422,
+      "step": 16160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.132733086995479e-06,
+      "loss": 0.5906,
+      "step": 16161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.131663384352102e-06,
+      "loss": 0.5965,
+      "step": 16162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.130593703860214e-06,
+      "loss": 0.6307,
+      "step": 16163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.129524045532503e-06,
+      "loss": 0.6424,
+      "step": 16164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.128454409381655e-06,
+      "loss": 0.6388,
+      "step": 16165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.127384795420342e-06,
+      "loss": 0.667,
+      "step": 16166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.126315203661252e-06,
+      "loss": 0.6603,
+      "step": 16167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.125245634117072e-06,
+      "loss": 0.6196,
+      "step": 16168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.124176086800469e-06,
+      "loss": 0.6552,
+      "step": 16169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.123106561724132e-06,
+      "loss": 0.6184,
+      "step": 16170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.122037058900747e-06,
+      "loss": 0.6389,
+      "step": 16171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.12096757834298e-06,
+      "loss": 0.6286,
+      "step": 16172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.119898120063525e-06,
+      "loss": 0.5879,
+      "step": 16173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.118828684075058e-06,
+      "loss": 0.6226,
+      "step": 16174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.11775927039025e-06,
+      "loss": 0.6648,
+      "step": 16175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.11668987902179e-06,
+      "loss": 0.6537,
+      "step": 16176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.115620509982356e-06,
+      "loss": 0.6503,
+      "step": 16177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.114551163284621e-06,
+      "loss": 0.6175,
+      "step": 16178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.113481838941268e-06,
+      "loss": 0.6474,
+      "step": 16179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.112412536964978e-06,
+      "loss": 0.5978,
+      "step": 16180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.11134325736842e-06,
+      "loss": 0.669,
+      "step": 16181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.110274000164278e-06,
+      "loss": 0.6331,
+      "step": 16182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.109204765365233e-06,
+      "loss": 0.6207,
+      "step": 16183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.10813555298395e-06,
+      "loss": 0.6633,
+      "step": 16184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.107066363033118e-06,
+      "loss": 0.618,
+      "step": 16185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.105997195525411e-06,
+      "loss": 0.6218,
+      "step": 16186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.104928050473497e-06,
+      "loss": 0.615,
+      "step": 16187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.10385892789006e-06,
+      "loss": 0.6746,
+      "step": 16188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.102789827787779e-06,
+      "loss": 0.5992,
+      "step": 16189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.101720750179318e-06,
+      "loss": 0.6116,
+      "step": 16190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.100651695077362e-06,
+      "loss": 0.5912,
+      "step": 16191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.099582662494585e-06,
+      "loss": 0.5646,
+      "step": 16192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.098513652443658e-06,
+      "loss": 0.6445,
+      "step": 16193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.097444664937252e-06,
+      "loss": 0.6469,
+      "step": 16194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.096375699988056e-06,
+      "loss": 0.6156,
+      "step": 16195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.095306757608726e-06,
+      "loss": 0.6128,
+      "step": 16196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.094237837811944e-06,
+      "loss": 0.6448,
+      "step": 16197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.09316894061039e-06,
+      "loss": 0.6132,
+      "step": 16198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.092100066016723e-06,
+      "loss": 0.644,
+      "step": 16199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.091031214043624e-06,
+      "loss": 0.6296,
+      "step": 16200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.08996238470377e-06,
+      "loss": 0.6581,
+      "step": 16201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.088893578009826e-06,
+      "loss": 0.6227,
+      "step": 16202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.087824793974463e-06,
+      "loss": 0.6604,
+      "step": 16203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.086756032610362e-06,
+      "loss": 0.6616,
+      "step": 16204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.085687293930183e-06,
+      "loss": 0.6263,
+      "step": 16205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.084618577946602e-06,
+      "loss": 0.6647,
+      "step": 16206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.083549884672297e-06,
+      "loss": 0.6087,
+      "step": 16207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.082481214119928e-06,
+      "loss": 0.6031,
+      "step": 16208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.081412566302173e-06,
+      "loss": 0.689,
+      "step": 16209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.080343941231695e-06,
+      "loss": 0.6248,
+      "step": 16210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.07927533892117e-06,
+      "loss": 0.6581,
+      "step": 16211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.078206759383265e-06,
+      "loss": 0.6246,
+      "step": 16212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.07713820263065e-06,
+      "loss": 0.6441,
+      "step": 16213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.076069668675995e-06,
+      "loss": 0.6661,
+      "step": 16214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.075001157531968e-06,
+      "loss": 0.6619,
+      "step": 16215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.073932669211234e-06,
+      "loss": 0.6361,
+      "step": 16216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.072864203726467e-06,
+      "loss": 0.6101,
+      "step": 16217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.071795761090336e-06,
+      "loss": 0.618,
+      "step": 16218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.0707273413155e-06,
+      "loss": 0.6136,
+      "step": 16219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.069658944414635e-06,
+      "loss": 0.6268,
+      "step": 16220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.068590570400405e-06,
+      "loss": 0.6074,
+      "step": 16221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.067522219285475e-06,
+      "loss": 0.6554,
+      "step": 16222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.066453891082516e-06,
+      "loss": 0.6587,
+      "step": 16223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.065385585804191e-06,
+      "loss": 0.6381,
+      "step": 16224
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.064317303463167e-06,
+      "loss": 0.5806,
+      "step": 16225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.06324904407211e-06,
+      "loss": 0.6356,
+      "step": 16226
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.062180807643687e-06,
+      "loss": 0.6665,
+      "step": 16227
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.06111259419056e-06,
+      "loss": 0.6393,
+      "step": 16228
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.060044403725397e-06,
+      "loss": 0.7185,
+      "step": 16229
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.058976236260863e-06,
+      "loss": 0.6395,
+      "step": 16230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.057908091809618e-06,
+      "loss": 0.6693,
+      "step": 16231
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.05683997038433e-06,
+      "loss": 0.6204,
+      "step": 16232
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.055771871997665e-06,
+      "loss": 0.6652,
+      "step": 16233
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.05470379666228e-06,
+      "loss": 0.6441,
+      "step": 16234
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.053635744390844e-06,
+      "loss": 0.6546,
+      "step": 16235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.052567715196021e-06,
+      "loss": 0.6519,
+      "step": 16236
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.051499709090469e-06,
+      "loss": 0.6498,
+      "step": 16237
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.05043172608685e-06,
+      "loss": 0.6259,
+      "step": 16238
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.049363766197835e-06,
+      "loss": 0.5938,
+      "step": 16239
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.048295829436075e-06,
+      "loss": 0.6327,
+      "step": 16240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.04722791581424e-06,
+      "loss": 0.5804,
+      "step": 16241
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.046160025344987e-06,
+      "loss": 0.5878,
+      "step": 16242
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.04509215804098e-06,
+      "loss": 0.6253,
+      "step": 16243
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.044024313914876e-06,
+      "loss": 0.6135,
+      "step": 16244
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.04295649297934e-06,
+      "loss": 0.6665,
+      "step": 16245
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.041888695247029e-06,
+      "loss": 0.592,
+      "step": 16246
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.040820920730605e-06,
+      "loss": 0.6274,
+      "step": 16247
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.03975316944273e-06,
+      "loss": 0.6246,
+      "step": 16248
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.038685441396057e-06,
+      "loss": 0.6679,
+      "step": 16249
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.03761773660325e-06,
+      "loss": 0.6178,
+      "step": 16250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.036550055076967e-06,
+      "loss": 0.628,
+      "step": 16251
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.035482396829867e-06,
+      "loss": 0.6327,
+      "step": 16252
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.03441476187461e-06,
+      "loss": 0.595,
+      "step": 16253
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.033347150223853e-06,
+      "loss": 0.6323,
+      "step": 16254
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.032279561890251e-06,
+      "loss": 0.6294,
+      "step": 16255
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.031211996886464e-06,
+      "loss": 0.662,
+      "step": 16256
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.030144455225152e-06,
+      "loss": 0.6098,
+      "step": 16257
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.029076936918967e-06,
+      "loss": 0.6379,
+      "step": 16258
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.028009441980568e-06,
+      "loss": 0.6093,
+      "step": 16259
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.026941970422613e-06,
+      "loss": 0.6249,
+      "step": 16260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.025874522257755e-06,
+      "loss": 0.6168,
+      "step": 16261
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.024807097498653e-06,
+      "loss": 0.6193,
+      "step": 16262
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.023739696157962e-06,
+      "loss": 0.5716,
+      "step": 16263
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.022672318248337e-06,
+      "loss": 0.6638,
+      "step": 16264
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.021604963782431e-06,
+      "loss": 0.5949,
+      "step": 16265
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.020537632772904e-06,
+      "loss": 0.597,
+      "step": 16266
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.019470325232405e-06,
+      "loss": 0.6387,
+      "step": 16267
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.01840304117359e-06,
+      "loss": 0.635,
+      "step": 16268
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.017335780609118e-06,
+      "loss": 0.7025,
+      "step": 16269
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.016268543551632e-06,
+      "loss": 0.6876,
+      "step": 16270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.015201330013798e-06,
+      "loss": 0.6964,
+      "step": 16271
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.01413414000826e-06,
+      "loss": 0.6509,
+      "step": 16272
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.013066973547674e-06,
+      "loss": 0.6799,
+      "step": 16273
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.011999830644692e-06,
+      "loss": 0.612,
+      "step": 16274
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.010932711311968e-06,
+      "loss": 0.6375,
+      "step": 16275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.009865615562152e-06,
+      "loss": 0.6487,
+      "step": 16276
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.008798543407897e-06,
+      "loss": 0.5973,
+      "step": 16277
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.007731494861855e-06,
+      "loss": 0.643,
+      "step": 16278
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.006664469936673e-06,
+      "loss": 0.5965,
+      "step": 16279
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.00559746864501e-06,
+      "loss": 0.6642,
+      "step": 16280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.004530490999509e-06,
+      "loss": 0.6122,
+      "step": 16281
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.003463537012823e-06,
+      "loss": 0.6403,
+      "step": 16282
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.002396606697605e-06,
+      "loss": 0.6835,
+      "step": 16283
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.0013297000665e-06,
+      "loss": 0.6607,
+      "step": 16284
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.00026281713216e-06,
+      "loss": 0.6264,
+      "step": 16285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.999195957907235e-06,
+      "loss": 0.6489,
+      "step": 16286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.998129122404372e-06,
+      "loss": 0.6315,
+      "step": 16287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.99706231063622e-06,
+      "loss": 0.6301,
+      "step": 16288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.995995522615431e-06,
+      "loss": 0.6187,
+      "step": 16289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.994928758354647e-06,
+      "loss": 0.5762,
+      "step": 16290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.993862017866521e-06,
+      "loss": 0.6689,
+      "step": 16291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.992795301163698e-06,
+      "loss": 0.6029,
+      "step": 16292
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.991728608258825e-06,
+      "loss": 0.6551,
+      "step": 16293
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.99066193916455e-06,
+      "loss": 0.6366,
+      "step": 16294
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.989595293893521e-06,
+      "loss": 0.5968,
+      "step": 16295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.988528672458383e-06,
+      "loss": 0.6386,
+      "step": 16296
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.987462074871779e-06,
+      "loss": 0.6902,
+      "step": 16297
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.98639550114636e-06,
+      "loss": 0.6391,
+      "step": 16298
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.985328951294771e-06,
+      "loss": 0.6091,
+      "step": 16299
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.984262425329653e-06,
+      "loss": 0.6059,
+      "step": 16300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.983195923263657e-06,
+      "loss": 0.5954,
+      "step": 16301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.98212944510942e-06,
+      "loss": 0.6154,
+      "step": 16302
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.981062990879596e-06,
+      "loss": 0.6301,
+      "step": 16303
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.979996560586822e-06,
+      "loss": 0.5853,
+      "step": 16304
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.978930154243744e-06,
+      "loss": 0.6106,
+      "step": 16305
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.977863771863006e-06,
+      "loss": 0.6112,
+      "step": 16306
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.97679741345725e-06,
+      "loss": 0.5976,
+      "step": 16307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.975731079039121e-06,
+      "loss": 0.629,
+      "step": 16308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.97466476862126e-06,
+      "loss": 0.5699,
+      "step": 16309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.973598482216313e-06,
+      "loss": 0.6753,
+      "step": 16310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.972532219836916e-06,
+      "loss": 0.6337,
+      "step": 16311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.971465981495717e-06,
+      "loss": 0.6245,
+      "step": 16312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.970399767205353e-06,
+      "loss": 0.6299,
+      "step": 16313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.969333576978468e-06,
+      "loss": 0.6682,
+      "step": 16314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.968267410827701e-06,
+      "loss": 0.6589,
+      "step": 16315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.967201268765697e-06,
+      "loss": 0.6341,
+      "step": 16316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.96613515080509e-06,
+      "loss": 0.6446,
+      "step": 16317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.965069056958525e-06,
+      "loss": 0.6667,
+      "step": 16318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.96400298723864e-06,
+      "loss": 0.674,
+      "step": 16319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.962936941658077e-06,
+      "loss": 0.6585,
+      "step": 16320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.961870920229471e-06,
+      "loss": 0.6354,
+      "step": 16321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.960804922965466e-06,
+      "loss": 0.632,
+      "step": 16322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.959738949878697e-06,
+      "loss": 0.6481,
+      "step": 16323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.958673000981803e-06,
+      "loss": 0.6172,
+      "step": 16324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.957607076287425e-06,
+      "loss": 0.5516,
+      "step": 16325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.956541175808196e-06,
+      "loss": 0.6273,
+      "step": 16326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.955475299556756e-06,
+      "loss": 0.6327,
+      "step": 16327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.954409447545747e-06,
+      "loss": 0.651,
+      "step": 16328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.953343619787797e-06,
+      "loss": 0.6474,
+      "step": 16329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.95227781629555e-06,
+      "loss": 0.6552,
+      "step": 16330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.951212037081641e-06,
+      "loss": 0.5994,
+      "step": 16331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.950146282158702e-06,
+      "loss": 0.6353,
+      "step": 16332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.949080551539375e-06,
+      "loss": 0.6027,
+      "step": 16333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.94801484523629e-06,
+      "loss": 0.6061,
+      "step": 16334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.946949163262085e-06,
+      "loss": 0.5612,
+      "step": 16335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.945883505629396e-06,
+      "loss": 0.5853,
+      "step": 16336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.944817872350856e-06,
+      "loss": 0.6532,
+      "step": 16337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.943752263439097e-06,
+      "loss": 0.5998,
+      "step": 16338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.942686678906761e-06,
+      "loss": 0.6492,
+      "step": 16339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.941621118766473e-06,
+      "loss": 0.6043,
+      "step": 16340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.940555583030872e-06,
+      "loss": 0.6436,
+      "step": 16341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.939490071712593e-06,
+      "loss": 0.6496,
+      "step": 16342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.938424584824262e-06,
+      "loss": 0.6707,
+      "step": 16343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.937359122378515e-06,
+      "loss": 0.6348,
+      "step": 16344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.936293684387986e-06,
+      "loss": 0.6165,
+      "step": 16345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.935228270865307e-06,
+      "loss": 0.6152,
+      "step": 16346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.934162881823107e-06,
+      "loss": 0.6021,
+      "step": 16347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.93309751727402e-06,
+      "loss": 0.6434,
+      "step": 16348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.932032177230677e-06,
+      "loss": 0.6506,
+      "step": 16349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.930966861705707e-06,
+      "loss": 0.6628,
+      "step": 16350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.929901570711746e-06,
+      "loss": 0.5559,
+      "step": 16351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.928836304261417e-06,
+      "loss": 0.6131,
+      "step": 16352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.927771062367354e-06,
+      "loss": 0.6201,
+      "step": 16353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.926705845042188e-06,
+      "loss": 0.5791,
+      "step": 16354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.925640652298547e-06,
+      "loss": 0.6426,
+      "step": 16355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.924575484149057e-06,
+      "loss": 0.5816,
+      "step": 16356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.923510340606355e-06,
+      "loss": 0.6238,
+      "step": 16357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.922445221683061e-06,
+      "loss": 0.6514,
+      "step": 16358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.921380127391807e-06,
+      "loss": 0.6679,
+      "step": 16359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.920315057745223e-06,
+      "loss": 0.6301,
+      "step": 16360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.919250012755934e-06,
+      "loss": 0.633,
+      "step": 16361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.918184992436565e-06,
+      "loss": 0.6756,
+      "step": 16362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.917119996799753e-06,
+      "loss": 0.5748,
+      "step": 16363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.916055025858114e-06,
+      "loss": 0.6991,
+      "step": 16364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.914990079624279e-06,
+      "loss": 0.582,
+      "step": 16365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.913925158110875e-06,
+      "loss": 0.6681,
+      "step": 16366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.912860261330527e-06,
+      "loss": 0.6346,
+      "step": 16367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.91179538929586e-06,
+      "loss": 0.6188,
+      "step": 16368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.910730542019502e-06,
+      "loss": 0.6915,
+      "step": 16369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.909665719514076e-06,
+      "loss": 0.6366,
+      "step": 16370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.908600921792205e-06,
+      "loss": 0.6166,
+      "step": 16371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.907536148866518e-06,
+      "loss": 0.6423,
+      "step": 16372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.906471400749636e-06,
+      "loss": 0.5878,
+      "step": 16373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.905406677454183e-06,
+      "loss": 0.6019,
+      "step": 16374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.904341978992788e-06,
+      "loss": 0.5727,
+      "step": 16375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.903277305378066e-06,
+      "loss": 0.6576,
+      "step": 16376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.902212656622643e-06,
+      "loss": 0.6231,
+      "step": 16377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.901148032739146e-06,
+      "loss": 0.6393,
+      "step": 16378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.900083433740191e-06,
+      "loss": 0.5992,
+      "step": 16379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.899018859638405e-06,
+      "loss": 0.565,
+      "step": 16380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.897954310446408e-06,
+      "loss": 0.6272,
+      "step": 16381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.89688978617682e-06,
+      "loss": 0.6056,
+      "step": 16382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.895825286842266e-06,
+      "loss": 0.6331,
+      "step": 16383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.894760812455365e-06,
+      "loss": 0.6375,
+      "step": 16384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.893696363028737e-06,
+      "loss": 0.652,
+      "step": 16385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.892631938575002e-06,
+      "loss": 0.6436,
+      "step": 16386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.891567539106786e-06,
+      "loss": 0.6408,
+      "step": 16387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.890503164636698e-06,
+      "loss": 0.6222,
+      "step": 16388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.889438815177366e-06,
+      "loss": 0.5626,
+      "step": 16389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.88837449074141e-06,
+      "loss": 0.6628,
+      "step": 16390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.88731019134144e-06,
+      "loss": 0.5883,
+      "step": 16391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.886245916990082e-06,
+      "loss": 0.6062,
+      "step": 16392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.885181667699957e-06,
+      "loss": 0.6703,
+      "step": 16393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.884117443483672e-06,
+      "loss": 0.6293,
+      "step": 16394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.883053244353858e-06,
+      "loss": 0.6479,
+      "step": 16395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.88198907032312e-06,
+      "loss": 0.6524,
+      "step": 16396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.88092492140408e-06,
+      "loss": 0.6394,
+      "step": 16397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.879860797609362e-06,
+      "loss": 0.6513,
+      "step": 16398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.878796698951572e-06,
+      "loss": 0.6204,
+      "step": 16399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.877732625443328e-06,
+      "loss": 0.656,
+      "step": 16400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.876668577097255e-06,
+      "loss": 0.6588,
+      "step": 16401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.875604553925958e-06,
+      "loss": 0.635,
+      "step": 16402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.874540555942055e-06,
+      "loss": 0.6512,
+      "step": 16403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.873476583158169e-06,
+      "loss": 0.6402,
+      "step": 16404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.872412635586902e-06,
+      "loss": 0.6628,
+      "step": 16405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.871348713240873e-06,
+      "loss": 0.5821,
+      "step": 16406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.870284816132705e-06,
+      "loss": 0.637,
+      "step": 16407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.869220944275e-06,
+      "loss": 0.6194,
+      "step": 16408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.868157097680374e-06,
+      "loss": 0.6324,
+      "step": 16409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.867093276361449e-06,
+      "loss": 0.6594,
+      "step": 16410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.866029480330826e-06,
+      "loss": 0.5651,
+      "step": 16411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.86496570960112e-06,
+      "loss": 0.6136,
+      "step": 16412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.863901964184955e-06,
+      "loss": 0.6592,
+      "step": 16413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.86283824409493e-06,
+      "loss": 0.6086,
+      "step": 16414
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.861774549343659e-06,
+      "loss": 0.6537,
+      "step": 16415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.860710879943762e-06,
+      "loss": 0.6308,
+      "step": 16416
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.859647235907838e-06,
+      "loss": 0.6277,
+      "step": 16417
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.858583617248502e-06,
+      "loss": 0.6375,
+      "step": 16418
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.857520023978373e-06,
+      "loss": 0.5829,
+      "step": 16419
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.85645645611005e-06,
+      "loss": 0.6266,
+      "step": 16420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.855392913656145e-06,
+      "loss": 0.6366,
+      "step": 16421
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.854329396629278e-06,
+      "loss": 0.6062,
+      "step": 16422
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.853265905042045e-06,
+      "loss": 0.6579,
+      "step": 16423
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.852202438907057e-06,
+      "loss": 0.6823,
+      "step": 16424
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.851138998236933e-06,
+      "loss": 0.5903,
+      "step": 16425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.85007558304427e-06,
+      "loss": 0.6119,
+      "step": 16426
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.849012193341679e-06,
+      "loss": 0.6475,
+      "step": 16427
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.847948829141776e-06,
+      "loss": 0.6235,
+      "step": 16428
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.846885490457157e-06,
+      "loss": 0.6354,
+      "step": 16429
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.845822177300432e-06,
+      "loss": 0.6739,
+      "step": 16430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.844758889684219e-06,
+      "loss": 0.6748,
+      "step": 16431
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.84369562762111e-06,
+      "loss": 0.6285,
+      "step": 16432
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.842632391123714e-06,
+      "loss": 0.6066,
+      "step": 16433
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.841569180204647e-06,
+      "loss": 0.6681,
+      "step": 16434
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.840505994876505e-06,
+      "loss": 0.6628,
+      "step": 16435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.839442835151892e-06,
+      "loss": 0.5883,
+      "step": 16436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.838379701043424e-06,
+      "loss": 0.6511,
+      "step": 16437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.837316592563695e-06,
+      "loss": 0.6263,
+      "step": 16438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.836253509725311e-06,
+      "loss": 0.6298,
+      "step": 16439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.835190452540886e-06,
+      "loss": 0.6431,
+      "step": 16440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.834127421023013e-06,
+      "loss": 0.6519,
+      "step": 16441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.833064415184296e-06,
+      "loss": 0.623,
+      "step": 16442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.832001435037348e-06,
+      "loss": 0.6796,
+      "step": 16443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.830938480594763e-06,
+      "loss": 0.6511,
+      "step": 16444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.829875551869142e-06,
+      "loss": 0.6263,
+      "step": 16445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.8288126488731e-06,
+      "loss": 0.5954,
+      "step": 16446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.827749771619224e-06,
+      "loss": 0.6322,
+      "step": 16447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.826686920120121e-06,
+      "loss": 0.6214,
+      "step": 16448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.8256240943884e-06,
+      "loss": 0.5828,
+      "step": 16449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.824561294436651e-06,
+      "loss": 0.5921,
+      "step": 16450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.82349852027748e-06,
+      "loss": 0.6005,
+      "step": 16451
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.822435771923494e-06,
+      "loss": 0.6073,
+      "step": 16452
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.82137304938728e-06,
+      "loss": 0.6672,
+      "step": 16453
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.820310352681444e-06,
+      "loss": 0.6306,
+      "step": 16454
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.819247681818591e-06,
+      "loss": 0.6915,
+      "step": 16455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.818185036811311e-06,
+      "loss": 0.5854,
+      "step": 16456
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.81712241767221e-06,
+      "loss": 0.6861,
+      "step": 16457
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.816059824413884e-06,
+      "loss": 0.6716,
+      "step": 16458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.81499725704893e-06,
+      "loss": 0.6627,
+      "step": 16459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.813934715589949e-06,
+      "loss": 0.604,
+      "step": 16460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.812872200049535e-06,
+      "loss": 0.6451,
+      "step": 16461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.811809710440289e-06,
+      "loss": 0.6223,
+      "step": 16462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.810747246774807e-06,
+      "loss": 0.6672,
+      "step": 16463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.809684809065687e-06,
+      "loss": 0.6069,
+      "step": 16464
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.808622397325522e-06,
+      "loss": 0.6714,
+      "step": 16465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.807560011566914e-06,
+      "loss": 0.6941,
+      "step": 16466
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.806497651802453e-06,
+      "loss": 0.6293,
+      "step": 16467
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.805435318044736e-06,
+      "loss": 0.6363,
+      "step": 16468
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.804373010306364e-06,
+      "loss": 0.6332,
+      "step": 16469
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.803310728599922e-06,
+      "loss": 0.6021,
+      "step": 16470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.802248472938013e-06,
+      "loss": 0.5834,
+      "step": 16471
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.801186243333229e-06,
+      "loss": 0.5908,
+      "step": 16472
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.800124039798162e-06,
+      "loss": 0.6319,
+      "step": 16473
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.799061862345407e-06,
+      "loss": 0.6253,
+      "step": 16474
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.79799971098756e-06,
+      "loss": 0.6312,
+      "step": 16475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.796937585737211e-06,
+      "loss": 0.6182,
+      "step": 16476
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.795875486606951e-06,
+      "loss": 0.5957,
+      "step": 16477
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.79481341360938e-06,
+      "loss": 0.6314,
+      "step": 16478
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.793751366757085e-06,
+      "loss": 0.6549,
+      "step": 16479
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.792689346062655e-06,
+      "loss": 0.5995,
+      "step": 16480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.791627351538689e-06,
+      "loss": 0.6981,
+      "step": 16481
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.790565383197772e-06,
+      "loss": 0.6036,
+      "step": 16482
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.789503441052496e-06,
+      "loss": 0.5908,
+      "step": 16483
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.788441525115457e-06,
+      "loss": 0.6364,
+      "step": 16484
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.787379635399238e-06,
+      "loss": 0.6655,
+      "step": 16485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.786317771916433e-06,
+      "loss": 0.6255,
+      "step": 16486
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.785255934679632e-06,
+      "loss": 0.6307,
+      "step": 16487
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.784194123701425e-06,
+      "loss": 0.6398,
+      "step": 16488
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.783132338994396e-06,
+      "loss": 0.6425,
+      "step": 16489
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.78207058057114e-06,
+      "loss": 0.6354,
+      "step": 16490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.781008848444242e-06,
+      "loss": 0.6596,
+      "step": 16491
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.779947142626291e-06,
+      "loss": 0.6286,
+      "step": 16492
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.778885463129877e-06,
+      "loss": 0.666,
+      "step": 16493
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.777823809967583e-06,
+      "loss": 0.6295,
+      "step": 16494
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.776762183151999e-06,
+      "loss": 0.647,
+      "step": 16495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.775700582695714e-06,
+      "loss": 0.6821,
+      "step": 16496
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.774639008611311e-06,
+      "loss": 0.6772,
+      "step": 16497
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.773577460911377e-06,
+      "loss": 0.6211,
+      "step": 16498
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.7725159396085e-06,
+      "loss": 0.5781,
+      "step": 16499
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.771454444715263e-06,
+      "loss": 0.5484,
+      "step": 16500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.770392976244251e-06,
+      "loss": 0.622,
+      "step": 16501
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.769331534208056e-06,
+      "loss": 0.6469,
+      "step": 16502
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.768270118619254e-06,
+      "loss": 0.6126,
+      "step": 16503
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.767208729490432e-06,
+      "loss": 0.6687,
+      "step": 16504
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.766147366834178e-06,
+      "loss": 0.6574,
+      "step": 16505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.76508603066307e-06,
+      "loss": 0.599,
+      "step": 16506
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.764024720989694e-06,
+      "loss": 0.6047,
+      "step": 16507
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.762963437826637e-06,
+      "loss": 0.6348,
+      "step": 16508
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.761902181186476e-06,
+      "loss": 0.6387,
+      "step": 16509
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.760840951081795e-06,
+      "loss": 0.585,
+      "step": 16510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.759779747525181e-06,
+      "loss": 0.6243,
+      "step": 16511
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.758718570529206e-06,
+      "loss": 0.6388,
+      "step": 16512
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.75765742010646e-06,
+      "loss": 0.6334,
+      "step": 16513
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.756596296269525e-06,
+      "loss": 0.6227,
+      "step": 16514
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.755535199030975e-06,
+      "loss": 0.6495,
+      "step": 16515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.754474128403395e-06,
+      "loss": 0.5827,
+      "step": 16516
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.753413084399369e-06,
+      "loss": 0.6437,
+      "step": 16517
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.752352067031468e-06,
+      "loss": 0.6462,
+      "step": 16518
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.751291076312277e-06,
+      "loss": 0.6437,
+      "step": 16519
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.750230112254377e-06,
+      "loss": 0.6501,
+      "step": 16520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.749169174870344e-06,
+      "loss": 0.5534,
+      "step": 16521
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.748108264172759e-06,
+      "loss": 0.6195,
+      "step": 16522
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.747047380174197e-06,
+      "loss": 0.6699,
+      "step": 16523
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.745986522887238e-06,
+      "loss": 0.5926,
+      "step": 16524
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.744925692324463e-06,
+      "loss": 0.6356,
+      "step": 16525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.743864888498444e-06,
+      "loss": 0.6318,
+      "step": 16526
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.74280411142176e-06,
+      "loss": 0.6042,
+      "step": 16527
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.741743361106993e-06,
+      "loss": 0.6716,
+      "step": 16528
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.740682637566711e-06,
+      "loss": 0.6609,
+      "step": 16529
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.739621940813495e-06,
+      "loss": 0.6368,
+      "step": 16530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.738561270859924e-06,
+      "loss": 0.6382,
+      "step": 16531
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.737500627718567e-06,
+      "loss": 0.6251,
+      "step": 16532
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.736440011402e-06,
+      "loss": 0.5992,
+      "step": 16533
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.735379421922805e-06,
+      "loss": 0.6075,
+      "step": 16534
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.734318859293547e-06,
+      "loss": 0.6419,
+      "step": 16535
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.733258323526807e-06,
+      "loss": 0.6475,
+      "step": 16536
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.732197814635157e-06,
+      "loss": 0.6029,
+      "step": 16537
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.73113733263117e-06,
+      "loss": 0.6021,
+      "step": 16538
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.730076877527421e-06,
+      "loss": 0.6657,
+      "step": 16539
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.729016449336483e-06,
+      "loss": 0.6528,
+      "step": 16540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.727956048070926e-06,
+      "loss": 0.6374,
+      "step": 16541
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.726895673743323e-06,
+      "loss": 0.6582,
+      "step": 16542
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.72583532636625e-06,
+      "loss": 0.6518,
+      "step": 16543
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.724775005952276e-06,
+      "loss": 0.6781,
+      "step": 16544
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.72371471251397e-06,
+      "loss": 0.6796,
+      "step": 16545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.722654446063909e-06,
+      "loss": 0.6439,
+      "step": 16546
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.721594206614657e-06,
+      "loss": 0.6034,
+      "step": 16547
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.72053399417879e-06,
+      "loss": 0.5832,
+      "step": 16548
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.719473808768877e-06,
+      "loss": 0.6628,
+      "step": 16549
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.718413650397484e-06,
+      "loss": 0.6433,
+      "step": 16550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.717353519077184e-06,
+      "loss": 0.6375,
+      "step": 16551
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.716293414820549e-06,
+      "loss": 0.6087,
+      "step": 16552
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.71523333764014e-06,
+      "loss": 0.6436,
+      "step": 16553
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.714173287548532e-06,
+      "loss": 0.6159,
+      "step": 16554
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.713113264558291e-06,
+      "loss": 0.6117,
+      "step": 16555
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.712053268681985e-06,
+      "loss": 0.6721,
+      "step": 16556
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.710993299932182e-06,
+      "loss": 0.6502,
+      "step": 16557
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.709933358321451e-06,
+      "loss": 0.6014,
+      "step": 16558
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.708873443862354e-06,
+      "loss": 0.5773,
+      "step": 16559
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.70781355656746e-06,
+      "loss": 0.6242,
+      "step": 16560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.70675369644934e-06,
+      "loss": 0.6918,
+      "step": 16561
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.705693863520553e-06,
+      "loss": 0.6409,
+      "step": 16562
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.704634057793667e-06,
+      "loss": 0.6553,
+      "step": 16563
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.703574279281251e-06,
+      "loss": 0.6784,
+      "step": 16564
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.702514527995863e-06,
+      "loss": 0.623,
+      "step": 16565
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.701454803950073e-06,
+      "loss": 0.7079,
+      "step": 16566
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.700395107156447e-06,
+      "loss": 0.6257,
+      "step": 16567
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.699335437627543e-06,
+      "loss": 0.6623,
+      "step": 16568
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.698275795375927e-06,
+      "loss": 0.6324,
+      "step": 16569
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.697216180414166e-06,
+      "loss": 0.6176,
+      "step": 16570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.696156592754818e-06,
+      "loss": 0.6059,
+      "step": 16571
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.695097032410447e-06,
+      "loss": 0.5999,
+      "step": 16572
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.69403749939362e-06,
+      "loss": 0.6007,
+      "step": 16573
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.692977993716893e-06,
+      "loss": 0.6183,
+      "step": 16574
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.691918515392829e-06,
+      "loss": 0.6434,
+      "step": 16575
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.690859064433994e-06,
+      "loss": 0.6969,
+      "step": 16576
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.689799640852944e-06,
+      "loss": 0.6713,
+      "step": 16577
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.688740244662242e-06,
+      "loss": 0.6068,
+      "step": 16578
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.68768087587445e-06,
+      "loss": 0.6422,
+      "step": 16579
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.686621534502123e-06,
+      "loss": 0.58,
+      "step": 16580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.685562220557826e-06,
+      "loss": 0.608,
+      "step": 16581
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.684502934054117e-06,
+      "loss": 0.605,
+      "step": 16582
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.683443675003552e-06,
+      "loss": 0.6217,
+      "step": 16583
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.6823844434187e-06,
+      "loss": 0.631,
+      "step": 16584
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.681325239312102e-06,
+      "loss": 0.6846,
+      "step": 16585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.680266062696332e-06,
+      "loss": 0.5832,
+      "step": 16586
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.679206913583945e-06,
+      "loss": 0.6305,
+      "step": 16587
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.67814779198749e-06,
+      "loss": 0.6514,
+      "step": 16588
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.677088697919533e-06,
+      "loss": 0.6143,
+      "step": 16589
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.676029631392631e-06,
+      "loss": 0.6219,
+      "step": 16590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.67497059241933e-06,
+      "loss": 0.6631,
+      "step": 16591
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.673911581012199e-06,
+      "loss": 0.62,
+      "step": 16592
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.67285259718379e-06,
+      "loss": 0.6714,
+      "step": 16593
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.671793640946654e-06,
+      "loss": 0.6439,
+      "step": 16594
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.670734712313349e-06,
+      "loss": 0.6802,
+      "step": 16595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.669675811296436e-06,
+      "loss": 0.6347,
+      "step": 16596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.668616937908458e-06,
+      "loss": 0.616,
+      "step": 16597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.667558092161978e-06,
+      "loss": 0.6659,
+      "step": 16598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.66649927406955e-06,
+      "loss": 0.6359,
+      "step": 16599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.66544048364372e-06,
+      "loss": 0.6619,
+      "step": 16600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.664381720897046e-06,
+      "loss": 0.6438,
+      "step": 16601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.663322985842088e-06,
+      "loss": 0.6461,
+      "step": 16602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.662264278491385e-06,
+      "loss": 0.6417,
+      "step": 16603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.661205598857497e-06,
+      "loss": 0.6051,
+      "step": 16604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.660146946952983e-06,
+      "loss": 0.6461,
+      "step": 16605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.659088322790376e-06,
+      "loss": 0.6082,
+      "step": 16606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.658029726382243e-06,
+      "loss": 0.6083,
+      "step": 16607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.656971157741134e-06,
+      "loss": 0.6646,
+      "step": 16608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.65591261687959e-06,
+      "loss": 0.6136,
+      "step": 16609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.654854103810167e-06,
+      "loss": 0.6288,
+      "step": 16610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.653795618545423e-06,
+      "loss": 0.6355,
+      "step": 16611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.65273716109789e-06,
+      "loss": 0.6241,
+      "step": 16612
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.651678731480133e-06,
+      "loss": 0.6377,
+      "step": 16613
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.650620329704697e-06,
+      "loss": 0.5676,
+      "step": 16614
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.649561955784123e-06,
+      "loss": 0.5978,
+      "step": 16615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.648503609730968e-06,
+      "loss": 0.6266,
+      "step": 16616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.64744529155778e-06,
+      "loss": 0.6689,
+      "step": 16617
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.6463870012771e-06,
+      "loss": 0.6316,
+      "step": 16618
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.645328738901481e-06,
+      "loss": 0.6407,
+      "step": 16619
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.644270504443472e-06,
+      "loss": 0.6395,
+      "step": 16620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.64321229791561e-06,
+      "loss": 0.6163,
+      "step": 16621
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.642154119330449e-06,
+      "loss": 0.626,
+      "step": 16622
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.64109596870054e-06,
+      "loss": 0.6072,
+      "step": 16623
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.640037846038415e-06,
+      "loss": 0.6237,
+      "step": 16624
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.638979751356627e-06,
+      "loss": 0.6877,
+      "step": 16625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.637921684667727e-06,
+      "loss": 0.6504,
+      "step": 16626
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.636863645984248e-06,
+      "loss": 0.6038,
+      "step": 16627
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.635805635318739e-06,
+      "loss": 0.6293,
+      "step": 16628
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.63474765268375e-06,
+      "loss": 0.5988,
+      "step": 16629
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.633689698091815e-06,
+      "loss": 0.6225,
+      "step": 16630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.632631771555483e-06,
+      "loss": 0.6363,
+      "step": 16631
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.6315738730873e-06,
+      "loss": 0.6723,
+      "step": 16632
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.630516002699799e-06,
+      "loss": 0.6383,
+      "step": 16633
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.62945816040553e-06,
+      "loss": 0.6274,
+      "step": 16634
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.628400346217038e-06,
+      "loss": 0.6458,
+      "step": 16635
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.6273425601468535e-06,
+      "loss": 0.6049,
+      "step": 16636
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.626284802207526e-06,
+      "loss": 0.6462,
+      "step": 16637
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.6252270724116e-06,
+      "loss": 0.66,
+      "step": 16638
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.624169370771603e-06,
+      "loss": 0.6806,
+      "step": 16639
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.623111697300087e-06,
+      "loss": 0.6631,
+      "step": 16640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.6220540520095935e-06,
+      "loss": 0.6163,
+      "step": 16641
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.620996434912651e-06,
+      "loss": 0.622,
+      "step": 16642
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.619938846021806e-06,
+      "loss": 0.6957,
+      "step": 16643
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.618881285349602e-06,
+      "loss": 0.6042,
+      "step": 16644
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.6178237529085656e-06,
+      "loss": 0.6015,
+      "step": 16645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.6167662487112485e-06,
+      "loss": 0.7127,
+      "step": 16646
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.615708772770178e-06,
+      "loss": 0.6317,
+      "step": 16647
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.614651325097895e-06,
+      "loss": 0.6222,
+      "step": 16648
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.613593905706943e-06,
+      "loss": 0.6592,
+      "step": 16649
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.612536514609851e-06,
+      "loss": 0.6389,
+      "step": 16650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.6114791518191556e-06,
+      "loss": 0.6507,
+      "step": 16651
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.610421817347403e-06,
+      "loss": 0.5809,
+      "step": 16652
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.609364511207118e-06,
+      "loss": 0.636,
+      "step": 16653
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.60830723341084e-06,
+      "loss": 0.6505,
+      "step": 16654
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.607249983971111e-06,
+      "loss": 0.6319,
+      "step": 16655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.606192762900456e-06,
+      "loss": 0.6449,
+      "step": 16656
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.605135570211411e-06,
+      "loss": 0.6384,
+      "step": 16657
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.604078405916521e-06,
+      "loss": 0.6755,
+      "step": 16658
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.603021270028308e-06,
+      "loss": 0.6456,
+      "step": 16659
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.601964162559307e-06,
+      "loss": 0.615,
+      "step": 16660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.600907083522062e-06,
+      "loss": 0.6033,
+      "step": 16661
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5998500329290925e-06,
+      "loss": 0.6004,
+      "step": 16662
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.598793010792936e-06,
+      "loss": 0.5714,
+      "step": 16663
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.597736017126132e-06,
+      "loss": 0.6072,
+      "step": 16664
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.596679051941203e-06,
+      "loss": 0.5976,
+      "step": 16665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.595622115250681e-06,
+      "loss": 0.6679,
+      "step": 16666
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.594565207067107e-06,
+      "loss": 0.6319,
+      "step": 16667
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.593508327403001e-06,
+      "loss": 0.6424,
+      "step": 16668
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.592451476270897e-06,
+      "loss": 0.57,
+      "step": 16669
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.59139465368333e-06,
+      "loss": 0.6582,
+      "step": 16670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.590337859652825e-06,
+      "loss": 0.622,
+      "step": 16671
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5892810941919094e-06,
+      "loss": 0.627,
+      "step": 16672
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.588224357313122e-06,
+      "loss": 0.6154,
+      "step": 16673
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.587167649028981e-06,
+      "loss": 0.6289,
+      "step": 16674
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.58611096935202e-06,
+      "loss": 0.6161,
+      "step": 16675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.585054318294771e-06,
+      "loss": 0.5702,
+      "step": 16676
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.583997695869756e-06,
+      "loss": 0.6753,
+      "step": 16677
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5829411020895e-06,
+      "loss": 0.6157,
+      "step": 16678
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5818845369665425e-06,
+      "loss": 0.6274,
+      "step": 16679
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.580828000513398e-06,
+      "loss": 0.6218,
+      "step": 16680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.579771492742597e-06,
+      "loss": 0.6182,
+      "step": 16681
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.578715013666672e-06,
+      "loss": 0.6482,
+      "step": 16682
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5776585632981395e-06,
+      "loss": 0.6266,
+      "step": 16683
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5766021416495265e-06,
+      "loss": 0.6446,
+      "step": 16684
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.575545748733367e-06,
+      "loss": 0.6733,
+      "step": 16685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5744893845621754e-06,
+      "loss": 0.6322,
+      "step": 16686
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.573433049148479e-06,
+      "loss": 0.6498,
+      "step": 16687
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.572376742504809e-06,
+      "loss": 0.6482,
+      "step": 16688
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.57132046464368e-06,
+      "loss": 0.6391,
+      "step": 16689
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.570264215577616e-06,
+      "loss": 0.6531,
+      "step": 16690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.569207995319151e-06,
+      "loss": 0.635,
+      "step": 16691
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.568151803880795e-06,
+      "loss": 0.6368,
+      "step": 16692
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.567095641275075e-06,
+      "loss": 0.6575,
+      "step": 16693
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.566039507514519e-06,
+      "loss": 0.6432,
+      "step": 16694
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.56498340261164e-06,
+      "loss": 0.6395,
+      "step": 16695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.563927326578959e-06,
+      "loss": 0.6004,
+      "step": 16696
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.562871279429009e-06,
+      "loss": 0.5941,
+      "step": 16697
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5618152611743e-06,
+      "loss": 0.6693,
+      "step": 16698
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.560759271827351e-06,
+      "loss": 0.6199,
+      "step": 16699
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5597033114006945e-06,
+      "loss": 0.6307,
+      "step": 16700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.558647379906837e-06,
+      "loss": 0.6246,
+      "step": 16701
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5575914773583015e-06,
+      "loss": 0.6172,
+      "step": 16702
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.556535603767615e-06,
+      "loss": 0.6586,
+      "step": 16703
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.555479759147286e-06,
+      "loss": 0.6447,
+      "step": 16704
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.554423943509834e-06,
+      "loss": 0.6286,
+      "step": 16705
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.553368156867787e-06,
+      "loss": 0.6091,
+      "step": 16706
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.55231239923365e-06,
+      "loss": 0.632,
+      "step": 16707
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.55125667061995e-06,
+      "loss": 0.6203,
+      "step": 16708
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.550200971039195e-06,
+      "loss": 0.6688,
+      "step": 16709
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.549145300503908e-06,
+      "loss": 0.6429,
+      "step": 16710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5480896590266075e-06,
+      "loss": 0.6791,
+      "step": 16711
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.547034046619801e-06,
+      "loss": 0.6851,
+      "step": 16712
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.54597846329601e-06,
+      "loss": 0.5796,
+      "step": 16713
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.544922909067751e-06,
+      "loss": 0.6243,
+      "step": 16714
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.543867383947534e-06,
+      "loss": 0.5869,
+      "step": 16715
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.542811887947876e-06,
+      "loss": 0.591,
+      "step": 16716
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5417564210812946e-06,
+      "loss": 0.6421,
+      "step": 16717
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.540700983360298e-06,
+      "loss": 0.6261,
+      "step": 16718
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.539645574797402e-06,
+      "loss": 0.6227,
+      "step": 16719
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5385901954051225e-06,
+      "loss": 0.6485,
+      "step": 16720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5375348451959665e-06,
+      "loss": 0.6386,
+      "step": 16721
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.536479524182452e-06,
+      "loss": 0.6041,
+      "step": 16722
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5354242323770894e-06,
+      "loss": 0.6005,
+      "step": 16723
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.53436896979239e-06,
+      "loss": 0.6163,
+      "step": 16724
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.533313736440864e-06,
+      "loss": 0.6036,
+      "step": 16725
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.532258532335026e-06,
+      "loss": 0.6347,
+      "step": 16726
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.531203357487383e-06,
+      "loss": 0.6584,
+      "step": 16727
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5301482119104475e-06,
+      "loss": 0.6299,
+      "step": 16728
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.529093095616732e-06,
+      "loss": 0.6332,
+      "step": 16729
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.528038008618741e-06,
+      "loss": 0.6158,
+      "step": 16730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.526982950928986e-06,
+      "loss": 0.6597,
+      "step": 16731
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.525927922559979e-06,
+      "loss": 0.6103,
+      "step": 16732
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.524872923524224e-06,
+      "loss": 0.5963,
+      "step": 16733
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.523817953834232e-06,
+      "loss": 0.6305,
+      "step": 16734
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.522763013502512e-06,
+      "loss": 0.5799,
+      "step": 16735
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.521708102541569e-06,
+      "loss": 0.6496,
+      "step": 16736
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.520653220963909e-06,
+      "loss": 0.7029,
+      "step": 16737
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.519598368782046e-06,
+      "loss": 0.6218,
+      "step": 16738
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.51854354600848e-06,
+      "loss": 0.6285,
+      "step": 16739
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.517488752655718e-06,
+      "loss": 0.6589,
+      "step": 16740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5164339887362705e-06,
+      "loss": 0.6358,
+      "step": 16741
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5153792542626365e-06,
+      "loss": 0.6942,
+      "step": 16742
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.514324549247326e-06,
+      "loss": 0.539,
+      "step": 16743
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.513269873702843e-06,
+      "loss": 0.6304,
+      "step": 16744
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.51221522764169e-06,
+      "loss": 0.628,
+      "step": 16745
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5111606110763724e-06,
+      "loss": 0.6404,
+      "step": 16746
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.510106024019397e-06,
+      "loss": 0.6228,
+      "step": 16747
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.509051466483262e-06,
+      "loss": 0.6762,
+      "step": 16748
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5079969384804715e-06,
+      "loss": 0.6216,
+      "step": 16749
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.506942440023534e-06,
+      "loss": 0.7054,
+      "step": 16750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.505887971124943e-06,
+      "loss": 0.6458,
+      "step": 16751
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5048335317972066e-06,
+      "loss": 0.6157,
+      "step": 16752
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.503779122052828e-06,
+      "loss": 0.624,
+      "step": 16753
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.5027247419043026e-06,
+      "loss": 0.5743,
+      "step": 16754
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.501670391364134e-06,
+      "loss": 0.5938,
+      "step": 16755
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.500616070444826e-06,
+      "loss": 0.6067,
+      "step": 16756
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.499561779158875e-06,
+      "loss": 0.6231,
+      "step": 16757
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.49850751751878e-06,
+      "loss": 0.6162,
+      "step": 16758
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4974532855370465e-06,
+      "loss": 0.5941,
+      "step": 16759
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.496399083226168e-06,
+      "loss": 0.555,
+      "step": 16760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.495344910598645e-06,
+      "loss": 0.6572,
+      "step": 16761
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.494290767666979e-06,
+      "loss": 0.6235,
+      "step": 16762
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.493236654443663e-06,
+      "loss": 0.6754,
+      "step": 16763
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.492182570941198e-06,
+      "loss": 0.6837,
+      "step": 16764
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.491128517172082e-06,
+      "loss": 0.6087,
+      "step": 16765
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.49007449314881e-06,
+      "loss": 0.6411,
+      "step": 16766
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.48902049888388e-06,
+      "loss": 0.6117,
+      "step": 16767
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.48796653438979e-06,
+      "loss": 0.6515,
+      "step": 16768
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.486912599679032e-06,
+      "loss": 0.6156,
+      "step": 16769
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.485858694764108e-06,
+      "loss": 0.6565,
+      "step": 16770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4848048196575065e-06,
+      "loss": 0.6706,
+      "step": 16771
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.483750974371726e-06,
+      "loss": 0.6483,
+      "step": 16772
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4826971589192624e-06,
+      "loss": 0.6373,
+      "step": 16773
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.481643373312606e-06,
+      "loss": 0.6568,
+      "step": 16774
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4805896175642535e-06,
+      "loss": 0.6296,
+      "step": 16775
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4795358916867e-06,
+      "loss": 0.6305,
+      "step": 16776
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.478482195692435e-06,
+      "loss": 0.6645,
+      "step": 16777
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.477428529593953e-06,
+      "loss": 0.6354,
+      "step": 16778
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4763748934037505e-06,
+      "loss": 0.6212,
+      "step": 16779
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4753212871343115e-06,
+      "loss": 0.6476,
+      "step": 16780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.474267710798135e-06,
+      "loss": 0.6302,
+      "step": 16781
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.473214164407709e-06,
+      "loss": 0.6161,
+      "step": 16782
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.472160647975526e-06,
+      "loss": 0.6514,
+      "step": 16783
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.471107161514075e-06,
+      "loss": 0.6198,
+      "step": 16784
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4700537050358515e-06,
+      "loss": 0.6262,
+      "step": 16785
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.469000278553338e-06,
+      "loss": 0.5988,
+      "step": 16786
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.467946882079027e-06,
+      "loss": 0.6324,
+      "step": 16787
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.466893515625412e-06,
+      "loss": 0.6099,
+      "step": 16788
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.465840179204978e-06,
+      "loss": 0.5932,
+      "step": 16789
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.464786872830212e-06,
+      "loss": 0.6281,
+      "step": 16790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4637335965136074e-06,
+      "loss": 0.6572,
+      "step": 16791
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.462680350267648e-06,
+      "loss": 0.6264,
+      "step": 16792
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.461627134104823e-06,
+      "loss": 0.6113,
+      "step": 16793
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.46057394803762e-06,
+      "loss": 0.6708,
+      "step": 16794
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.459520792078525e-06,
+      "loss": 0.6225,
+      "step": 16795
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.458467666240024e-06,
+      "loss": 0.7009,
+      "step": 16796
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.457414570534605e-06,
+      "loss": 0.6247,
+      "step": 16797
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4563615049747515e-06,
+      "loss": 0.5782,
+      "step": 16798
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.455308469572951e-06,
+      "loss": 0.6442,
+      "step": 16799
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.454255464341687e-06,
+      "loss": 0.6455,
+      "step": 16800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.453202489293445e-06,
+      "loss": 0.6534,
+      "step": 16801
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.452149544440707e-06,
+      "loss": 0.6626,
+      "step": 16802
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4510966297959645e-06,
+      "loss": 0.6471,
+      "step": 16803
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.450043745371692e-06,
+      "loss": 0.6061,
+      "step": 16804
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.448990891180377e-06,
+      "loss": 0.6105,
+      "step": 16805
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.447938067234504e-06,
+      "loss": 0.6549,
+      "step": 16806
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4468852735465515e-06,
+      "loss": 0.5996,
+      "step": 16807
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.445832510129003e-06,
+      "loss": 0.6601,
+      "step": 16808
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4447797769943444e-06,
+      "loss": 0.7043,
+      "step": 16809
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.443727074155051e-06,
+      "loss": 0.6465,
+      "step": 16810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.442674401623607e-06,
+      "loss": 0.6098,
+      "step": 16811
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4416217594124965e-06,
+      "loss": 0.6114,
+      "step": 16812
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.440569147534193e-06,
+      "loss": 0.6211,
+      "step": 16813
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.439516566001179e-06,
+      "loss": 0.6895,
+      "step": 16814
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.438464014825939e-06,
+      "loss": 0.6096,
+      "step": 16815
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.437411494020945e-06,
+      "loss": 0.6051,
+      "step": 16816
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.436359003598681e-06,
+      "loss": 0.6457,
+      "step": 16817
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.435306543571624e-06,
+      "loss": 0.5923,
+      "step": 16818
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.434254113952251e-06,
+      "loss": 0.6119,
+      "step": 16819
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.433201714753042e-06,
+      "loss": 0.6478,
+      "step": 16820
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.432149345986476e-06,
+      "loss": 0.6463,
+      "step": 16821
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.431097007665025e-06,
+      "loss": 0.6554,
+      "step": 16822
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.430044699801169e-06,
+      "loss": 0.5864,
+      "step": 16823
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.428992422407386e-06,
+      "loss": 0.6253,
+      "step": 16824
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.427940175496146e-06,
+      "loss": 0.6222,
+      "step": 16825
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.426887959079933e-06,
+      "loss": 0.6131,
+      "step": 16826
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.425835773171217e-06,
+      "loss": 0.6142,
+      "step": 16827
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.424783617782474e-06,
+      "loss": 0.626,
+      "step": 16828
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.423731492926178e-06,
+      "loss": 0.6085,
+      "step": 16829
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.422679398614807e-06,
+      "loss": 0.581,
+      "step": 16830
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.421627334860829e-06,
+      "loss": 0.6278,
+      "step": 16831
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.420575301676726e-06,
+      "loss": 0.6273,
+      "step": 16832
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.419523299074958e-06,
+      "loss": 0.6932,
+      "step": 16833
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.418471327068008e-06,
+      "loss": 0.6745,
+      "step": 16834
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.417419385668352e-06,
+      "loss": 0.6282,
+      "step": 16835
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.416367474888449e-06,
+      "loss": 0.6108,
+      "step": 16836
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.415315594740778e-06,
+      "loss": 0.6249,
+      "step": 16837
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.414263745237818e-06,
+      "loss": 0.6168,
+      "step": 16838
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.413211926392024e-06,
+      "loss": 0.622,
+      "step": 16839
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.412160138215876e-06,
+      "loss": 0.5797,
+      "step": 16840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.411108380721849e-06,
+      "loss": 0.6383,
+      "step": 16841
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.410056653922399e-06,
+      "loss": 0.6231,
+      "step": 16842
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.409004957830008e-06,
+      "loss": 0.6967,
+      "step": 16843
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.407953292457144e-06,
+      "loss": 0.6584,
+      "step": 16844
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.406901657816266e-06,
+      "loss": 0.6105,
+      "step": 16845
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.405850053919852e-06,
+      "loss": 0.6745,
+      "step": 16846
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.404798480780371e-06,
+      "loss": 0.6835,
+      "step": 16847
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.403746938410281e-06,
+      "loss": 0.6105,
+      "step": 16848
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.402695426822058e-06,
+      "loss": 0.6181,
+      "step": 16849
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4016439460281685e-06,
+      "loss": 0.651,
+      "step": 16850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.4005924960410724e-06,
+      "loss": 0.6827,
+      "step": 16851
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.399541076873242e-06,
+      "loss": 0.6268,
+      "step": 16852
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.398489688537146e-06,
+      "loss": 0.6309,
+      "step": 16853
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.397438331045239e-06,
+      "loss": 0.7056,
+      "step": 16854
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.396387004409995e-06,
+      "loss": 0.6223,
+      "step": 16855
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.395335708643882e-06,
+      "loss": 0.6462,
+      "step": 16856
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.3942844437593516e-06,
+      "loss": 0.634,
+      "step": 16857
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.3932332097688775e-06,
+      "loss": 0.6185,
+      "step": 16858
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.392182006684927e-06,
+      "loss": 0.6337,
+      "step": 16859
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.39113083451995e-06,
+      "loss": 0.6533,
+      "step": 16860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.390079693286418e-06,
+      "loss": 0.6223,
+      "step": 16861
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.3890285829968e-06,
+      "loss": 0.6101,
+      "step": 16862
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.387977503663544e-06,
+      "loss": 0.6834,
+      "step": 16863
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.386926455299119e-06,
+      "loss": 0.5818,
+      "step": 16864
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.385875437915993e-06,
+      "loss": 0.6276,
+      "step": 16865
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.384824451526612e-06,
+      "loss": 0.6083,
+      "step": 16866
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.383773496143449e-06,
+      "loss": 0.6021,
+      "step": 16867
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.382722571778964e-06,
+      "loss": 0.655,
+      "step": 16868
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.381671678445609e-06,
+      "loss": 0.6445,
+      "step": 16869
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.380620816155849e-06,
+      "loss": 0.6254,
+      "step": 16870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.379569984922148e-06,
+      "loss": 0.6333,
+      "step": 16871
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.3785191847569526e-06,
+      "loss": 0.624,
+      "step": 16872
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.377468415672731e-06,
+      "loss": 0.66,
+      "step": 16873
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.376417677681943e-06,
+      "loss": 0.6173,
+      "step": 16874
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.375366970797035e-06,
+      "loss": 0.5939,
+      "step": 16875
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.374316295030476e-06,
+      "loss": 0.6419,
+      "step": 16876
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.373265650394722e-06,
+      "loss": 0.6181,
+      "step": 16877
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.372215036902219e-06,
+      "loss": 0.6637,
+      "step": 16878
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.371164454565435e-06,
+      "loss": 0.6313,
+      "step": 16879
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.370113903396825e-06,
+      "loss": 0.5598,
+      "step": 16880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.369063383408836e-06,
+      "loss": 0.6078,
+      "step": 16881
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.3680128946139295e-06,
+      "loss": 0.6392,
+      "step": 16882
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.366962437024565e-06,
+      "loss": 0.6357,
+      "step": 16883
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.365912010653185e-06,
+      "loss": 0.591,
+      "step": 16884
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3648616155122506e-06,
+      "loss": 0.6119,
+      "step": 16885
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3638112516142225e-06,
+      "loss": 0.6401,
+      "step": 16886
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.362760918971539e-06,
+      "loss": 0.6224,
+      "step": 16887
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.361710617596663e-06,
+      "loss": 0.6451,
+      "step": 16888
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.360660347502049e-06,
+      "loss": 0.629,
+      "step": 16889
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.359610108700139e-06,
+      "loss": 0.6023,
+      "step": 16890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.358559901203395e-06,
+      "loss": 0.619,
+      "step": 16891
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.357509725024267e-06,
+      "loss": 0.6251,
+      "step": 16892
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3564595801751996e-06,
+      "loss": 0.631,
+      "step": 16893
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.355409466668653e-06,
+      "loss": 0.6091,
+      "step": 16894
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3543593845170665e-06,
+      "loss": 0.6303,
+      "step": 16895
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.353309333732896e-06,
+      "loss": 0.6646,
+      "step": 16896
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.352259314328599e-06,
+      "loss": 0.6254,
+      "step": 16897
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.351209326316613e-06,
+      "loss": 0.5657,
+      "step": 16898
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3501593697093865e-06,
+      "loss": 0.5906,
+      "step": 16899
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.349109444519381e-06,
+      "loss": 0.6737,
+      "step": 16900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.348059550759032e-06,
+      "loss": 0.6234,
+      "step": 16901
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.34700968844079e-06,
+      "loss": 0.6379,
+      "step": 16902
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.34595985757711e-06,
+      "loss": 0.6373,
+      "step": 16903
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.344910058180428e-06,
+      "loss": 0.6344,
+      "step": 16904
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.343860290263195e-06,
+      "loss": 0.6407,
+      "step": 16905
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3428105538378645e-06,
+      "loss": 0.6106,
+      "step": 16906
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3417608489168725e-06,
+      "loss": 0.6344,
+      "step": 16907
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3407111755126644e-06,
+      "loss": 0.6698,
+      "step": 16908
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.339661533637698e-06,
+      "loss": 0.6364,
+      "step": 16909
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.338611923304404e-06,
+      "loss": 0.6044,
+      "step": 16910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.337562344525231e-06,
+      "loss": 0.5599,
+      "step": 16911
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.336512797312629e-06,
+      "loss": 0.6746,
+      "step": 16912
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.335463281679035e-06,
+      "loss": 0.6469,
+      "step": 16913
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.334413797636891e-06,
+      "loss": 0.6794,
+      "step": 16914
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.33336434519865e-06,
+      "loss": 0.6147,
+      "step": 16915
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.332314924376745e-06,
+      "loss": 0.649,
+      "step": 16916
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.331265535183617e-06,
+      "loss": 0.6791,
+      "step": 16917
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.330216177631719e-06,
+      "loss": 0.5869,
+      "step": 16918
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3291668517334825e-06,
+      "loss": 0.6027,
+      "step": 16919
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.328117557501349e-06,
+      "loss": 0.6506,
+      "step": 16920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.327068294947768e-06,
+      "loss": 0.6591,
+      "step": 16921
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.326019064085168e-06,
+      "loss": 0.6932,
+      "step": 16922
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.324969864925995e-06,
+      "loss": 0.6263,
+      "step": 16923
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.323920697482693e-06,
+      "loss": 0.6386,
+      "step": 16924
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3228715617676925e-06,
+      "loss": 0.6878,
+      "step": 16925
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.321822457793435e-06,
+      "loss": 0.6369,
+      "step": 16926
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.320773385572364e-06,
+      "loss": 0.6816,
+      "step": 16927
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.319724345116911e-06,
+      "loss": 0.6238,
+      "step": 16928
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3186753364395155e-06,
+      "loss": 0.6375,
+      "step": 16929
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.31762635955262e-06,
+      "loss": 0.6014,
+      "step": 16930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.316577414468654e-06,
+      "loss": 0.6732,
+      "step": 16931
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.315528501200058e-06,
+      "loss": 0.6281,
+      "step": 16932
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.314479619759268e-06,
+      "loss": 0.629,
+      "step": 16933
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3134307701587185e-06,
+      "loss": 0.6285,
+      "step": 16934
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.312381952410845e-06,
+      "loss": 0.6702,
+      "step": 16935
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.311333166528085e-06,
+      "loss": 0.6052,
+      "step": 16936
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.310284412522871e-06,
+      "loss": 0.6789,
+      "step": 16937
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.309235690407637e-06,
+      "loss": 0.6112,
+      "step": 16938
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.308187000194819e-06,
+      "loss": 0.5996,
+      "step": 16939
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.307138341896849e-06,
+      "loss": 0.6128,
+      "step": 16940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.306089715526159e-06,
+      "loss": 0.5784,
+      "step": 16941
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.305041121095187e-06,
+      "loss": 0.6456,
+      "step": 16942
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.303992558616357e-06,
+      "loss": 0.6176,
+      "step": 16943
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.302944028102107e-06,
+      "loss": 0.6503,
+      "step": 16944
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.30189552956487e-06,
+      "loss": 0.6651,
+      "step": 16945
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.300847063017072e-06,
+      "loss": 0.654,
+      "step": 16946
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2997986284711465e-06,
+      "loss": 0.6297,
+      "step": 16947
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.298750225939526e-06,
+      "loss": 0.6254,
+      "step": 16948
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2977018554346356e-06,
+      "loss": 0.6382,
+      "step": 16949
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.29665351696891e-06,
+      "loss": 0.5727,
+      "step": 16950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.295605210554777e-06,
+      "loss": 0.6204,
+      "step": 16951
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.294556936204664e-06,
+      "loss": 0.6512,
+      "step": 16952
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.293508693931002e-06,
+      "loss": 0.6151,
+      "step": 16953
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.29246048374622e-06,
+      "loss": 0.6123,
+      "step": 16954
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.291412305662741e-06,
+      "loss": 0.6372,
+      "step": 16955
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.290364159692997e-06,
+      "loss": 0.6619,
+      "step": 16956
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.289316045849413e-06,
+      "loss": 0.6563,
+      "step": 16957
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.288267964144415e-06,
+      "loss": 0.683,
+      "step": 16958
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.287219914590434e-06,
+      "loss": 0.604,
+      "step": 16959
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2861718971998894e-06,
+      "loss": 0.6273,
+      "step": 16960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.285123911985211e-06,
+      "loss": 0.6935,
+      "step": 16961
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.284075958958825e-06,
+      "loss": 0.6557,
+      "step": 16962
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.283028038133151e-06,
+      "loss": 0.6347,
+      "step": 16963
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.281980149520618e-06,
+      "loss": 0.6645,
+      "step": 16964
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.280932293133649e-06,
+      "loss": 0.6468,
+      "step": 16965
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.279884468984667e-06,
+      "loss": 0.6216,
+      "step": 16966
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.278836677086094e-06,
+      "loss": 0.6182,
+      "step": 16967
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.277788917450358e-06,
+      "loss": 0.5817,
+      "step": 16968
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.276741190089875e-06,
+      "loss": 0.6421,
+      "step": 16969
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.275693495017071e-06,
+      "loss": 0.645,
+      "step": 16970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.274645832244368e-06,
+      "loss": 0.6557,
+      "step": 16971
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.273598201784185e-06,
+      "loss": 0.6341,
+      "step": 16972
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2725506036489425e-06,
+      "loss": 0.6427,
+      "step": 16973
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.271503037851066e-06,
+      "loss": 0.653,
+      "step": 16974
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.270455504402971e-06,
+      "loss": 0.6342,
+      "step": 16975
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.269408003317079e-06,
+      "loss": 0.6369,
+      "step": 16976
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2683605346058114e-06,
+      "loss": 0.6373,
+      "step": 16977
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.267313098281582e-06,
+      "loss": 0.6544,
+      "step": 16978
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.266265694356813e-06,
+      "loss": 0.6373,
+      "step": 16979
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.265218322843926e-06,
+      "loss": 0.5853,
+      "step": 16980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.264170983755332e-06,
+      "loss": 0.6599,
+      "step": 16981
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.263123677103452e-06,
+      "loss": 0.6192,
+      "step": 16982
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.262076402900704e-06,
+      "loss": 0.6411,
+      "step": 16983
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.261029161159503e-06,
+      "loss": 0.5934,
+      "step": 16984
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.259981951892266e-06,
+      "loss": 0.6259,
+      "step": 16985
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.25893477511141e-06,
+      "loss": 0.5919,
+      "step": 16986
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.257887630829349e-06,
+      "loss": 0.5626,
+      "step": 16987
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.256840519058499e-06,
+      "loss": 0.5241,
+      "step": 16988
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.255793439811276e-06,
+      "loss": 0.6775,
+      "step": 16989
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2547463931000916e-06,
+      "loss": 0.6178,
+      "step": 16990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.25369937893736e-06,
+      "loss": 0.6495,
+      "step": 16991
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2526523973355005e-06,
+      "loss": 0.6155,
+      "step": 16992
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2516054483069194e-06,
+      "loss": 0.6063,
+      "step": 16993
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.250558531864031e-06,
+      "loss": 0.6316,
+      "step": 16994
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.249511648019254e-06,
+      "loss": 0.6102,
+      "step": 16995
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.248464796784993e-06,
+      "loss": 0.6309,
+      "step": 16996
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.247417978173661e-06,
+      "loss": 0.5975,
+      "step": 16997
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.246371192197674e-06,
+      "loss": 0.6437,
+      "step": 16998
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2453244388694385e-06,
+      "loss": 0.5876,
+      "step": 16999
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.244277718201364e-06,
+      "loss": 0.6177,
+      "step": 17000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2432310302058685e-06,
+      "loss": 0.5962,
+      "step": 17001
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.242184374895354e-06,
+      "loss": 0.616,
+      "step": 17002
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.24113775228223e-06,
+      "loss": 0.6472,
+      "step": 17003
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.24009116237891e-06,
+      "loss": 0.6221,
+      "step": 17004
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2390446051978e-06,
+      "loss": 0.6257,
+      "step": 17005
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2379980807513096e-06,
+      "loss": 0.6324,
+      "step": 17006
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.236951589051847e-06,
+      "loss": 0.669,
+      "step": 17007
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.235905130111816e-06,
+      "loss": 0.6292,
+      "step": 17008
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.234858703943628e-06,
+      "loss": 0.6084,
+      "step": 17009
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.233812310559688e-06,
+      "loss": 0.5943,
+      "step": 17010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.232765949972401e-06,
+      "loss": 0.6301,
+      "step": 17011
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.231719622194175e-06,
+      "loss": 0.6138,
+      "step": 17012
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.230673327237416e-06,
+      "loss": 0.6331,
+      "step": 17013
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.229627065114526e-06,
+      "loss": 0.6329,
+      "step": 17014
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.228580835837911e-06,
+      "loss": 0.6013,
+      "step": 17015
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.22753463941998e-06,
+      "loss": 0.6875,
+      "step": 17016
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.226488475873129e-06,
+      "loss": 0.6435,
+      "step": 17017
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.225442345209768e-06,
+      "loss": 0.6069,
+      "step": 17018
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2243962474422955e-06,
+      "loss": 0.6209,
+      "step": 17019
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.223350182583118e-06,
+      "loss": 0.5896,
+      "step": 17020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.222304150644639e-06,
+      "loss": 0.6002,
+      "step": 17021
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.221258151639254e-06,
+      "loss": 0.674,
+      "step": 17022
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.220212185579369e-06,
+      "loss": 0.6539,
+      "step": 17023
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.21916625247739e-06,
+      "loss": 0.6568,
+      "step": 17024
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.218120352345708e-06,
+      "loss": 0.6647,
+      "step": 17025
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.217074485196729e-06,
+      "loss": 0.6297,
+      "step": 17026
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.216028651042855e-06,
+      "loss": 0.617,
+      "step": 17027
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.214982849896479e-06,
+      "loss": 0.6368,
+      "step": 17028
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.213937081770006e-06,
+      "loss": 0.6239,
+      "step": 17029
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2128913466758355e-06,
+      "loss": 0.6353,
+      "step": 17030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.211845644626361e-06,
+      "loss": 0.6097,
+      "step": 17031
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.210799975633984e-06,
+      "loss": 0.5661,
+      "step": 17032
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.209754339711103e-06,
+      "loss": 0.6078,
+      "step": 17033
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.208708736870113e-06,
+      "loss": 0.6533,
+      "step": 17034
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.207663167123411e-06,
+      "loss": 0.6576,
+      "step": 17035
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.206617630483397e-06,
+      "loss": 0.6497,
+      "step": 17036
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.205572126962462e-06,
+      "loss": 0.662,
+      "step": 17037
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2045266565730054e-06,
+      "loss": 0.5739,
+      "step": 17038
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.203481219327424e-06,
+      "loss": 0.5715,
+      "step": 17039
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.202435815238107e-06,
+      "loss": 0.6349,
+      "step": 17040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.201390444317452e-06,
+      "loss": 0.6516,
+      "step": 17041
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.200345106577857e-06,
+      "loss": 0.6712,
+      "step": 17042
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.199299802031709e-06,
+      "loss": 0.5989,
+      "step": 17043
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.198254530691405e-06,
+      "loss": 0.6193,
+      "step": 17044
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1972092925693405e-06,
+      "loss": 0.6232,
+      "step": 17045
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.196164087677904e-06,
+      "loss": 0.5917,
+      "step": 17046
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.195118916029487e-06,
+      "loss": 0.6527,
+      "step": 17047
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.194073777636488e-06,
+      "loss": 0.6665,
+      "step": 17048
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.19302867251129e-06,
+      "loss": 0.6361,
+      "step": 17049
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.191983600666288e-06,
+      "loss": 0.6514,
+      "step": 17050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.190938562113875e-06,
+      "loss": 0.5436,
+      "step": 17051
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.189893556866436e-06,
+      "loss": 0.6262,
+      "step": 17052
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.188848584936365e-06,
+      "loss": 0.6409,
+      "step": 17053
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.187803646336051e-06,
+      "loss": 0.6126,
+      "step": 17054
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.186758741077882e-06,
+      "loss": 0.6209,
+      "step": 17055
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1857138691742456e-06,
+      "loss": 0.6438,
+      "step": 17056
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.184669030637534e-06,
+      "loss": 0.582,
+      "step": 17057
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.18362422548013e-06,
+      "loss": 0.6622,
+      "step": 17058
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.182579453714422e-06,
+      "loss": 0.6702,
+      "step": 17059
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.181534715352802e-06,
+      "loss": 0.6122,
+      "step": 17060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1804900104076525e-06,
+      "loss": 0.5699,
+      "step": 17061
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.17944533889136e-06,
+      "loss": 0.6446,
+      "step": 17062
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.178400700816313e-06,
+      "loss": 0.6861,
+      "step": 17063
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.177356096194892e-06,
+      "loss": 0.648,
+      "step": 17064
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.176311525039486e-06,
+      "loss": 0.6653,
+      "step": 17065
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.175266987362481e-06,
+      "loss": 0.6795,
+      "step": 17066
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.174222483176257e-06,
+      "loss": 0.6564,
+      "step": 17067
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.173178012493199e-06,
+      "loss": 0.6493,
+      "step": 17068
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.172133575325695e-06,
+      "loss": 0.6497,
+      "step": 17069
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.171089171686123e-06,
+      "loss": 0.6016,
+      "step": 17070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.170044801586866e-06,
+      "loss": 0.6143,
+      "step": 17071
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.16900046504031e-06,
+      "loss": 0.6587,
+      "step": 17072
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.167956162058833e-06,
+      "loss": 0.6633,
+      "step": 17073
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1669118926548195e-06,
+      "loss": 0.6591,
+      "step": 17074
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.16586765684065e-06,
+      "loss": 0.6887,
+      "step": 17075
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.164823454628704e-06,
+      "loss": 0.6506,
+      "step": 17076
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.16377928603136e-06,
+      "loss": 0.6307,
+      "step": 17077
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.162735151061004e-06,
+      "loss": 0.6493,
+      "step": 17078
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.161691049730011e-06,
+      "loss": 0.585,
+      "step": 17079
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.160646982050763e-06,
+      "loss": 0.6511,
+      "step": 17080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.159602948035632e-06,
+      "loss": 0.5743,
+      "step": 17081
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.158558947697003e-06,
+      "loss": 0.6342,
+      "step": 17082
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.157514981047256e-06,
+      "loss": 0.624,
+      "step": 17083
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.156471048098759e-06,
+      "loss": 0.6007,
+      "step": 17084
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.155427148863896e-06,
+      "loss": 0.6484,
+      "step": 17085
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1543832833550475e-06,
+      "loss": 0.6546,
+      "step": 17086
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.153339451584578e-06,
+      "loss": 0.648,
+      "step": 17087
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.152295653564874e-06,
+      "loss": 0.6133,
+      "step": 17088
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.15125188930831e-06,
+      "loss": 0.6623,
+      "step": 17089
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.150208158827252e-06,
+      "loss": 0.6435,
+      "step": 17090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.149164462134084e-06,
+      "loss": 0.6288,
+      "step": 17091
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.148120799241182e-06,
+      "loss": 0.6664,
+      "step": 17092
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.147077170160909e-06,
+      "loss": 0.6192,
+      "step": 17093
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.146033574905647e-06,
+      "loss": 0.645,
+      "step": 17094
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.144990013487772e-06,
+      "loss": 0.6552,
+      "step": 17095
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.143946485919645e-06,
+      "loss": 0.6339,
+      "step": 17096
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.142902992213648e-06,
+      "loss": 0.669,
+      "step": 17097
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.141859532382154e-06,
+      "loss": 0.6252,
+      "step": 17098
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.140816106437525e-06,
+      "loss": 0.5985,
+      "step": 17099
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.13977271439214e-06,
+      "loss": 0.5871,
+      "step": 17100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1387293562583714e-06,
+      "loss": 0.5765,
+      "step": 17101
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.137686032048581e-06,
+      "loss": 0.6366,
+      "step": 17102
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.136642741775144e-06,
+      "loss": 0.6523,
+      "step": 17103
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.135599485450434e-06,
+      "loss": 0.6152,
+      "step": 17104
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1345562630868096e-06,
+      "loss": 0.6182,
+      "step": 17105
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1335130746966475e-06,
+      "loss": 0.6111,
+      "step": 17106
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.132469920292318e-06,
+      "loss": 0.5838,
+      "step": 17107
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.131426799886179e-06,
+      "loss": 0.6281,
+      "step": 17108
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.130383713490607e-06,
+      "loss": 0.6659,
+      "step": 17109
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1293406611179685e-06,
+      "loss": 0.6288,
+      "step": 17110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.128297642780622e-06,
+      "loss": 0.6738,
+      "step": 17111
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.127254658490943e-06,
+      "loss": 0.6178,
+      "step": 17112
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.126211708261297e-06,
+      "loss": 0.6593,
+      "step": 17113
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.125168792104041e-06,
+      "loss": 0.676,
+      "step": 17114
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.124125910031547e-06,
+      "loss": 0.6834,
+      "step": 17115
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.123083062056184e-06,
+      "loss": 0.6319,
+      "step": 17116
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.122040248190305e-06,
+      "loss": 0.6275,
+      "step": 17117
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.120997468446278e-06,
+      "loss": 0.627,
+      "step": 17118
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.119954722836474e-06,
+      "loss": 0.6148,
+      "step": 17119
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.118912011373247e-06,
+      "loss": 0.6421,
+      "step": 17120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.117869334068959e-06,
+      "loss": 0.5998,
+      "step": 17121
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.116826690935983e-06,
+      "loss": 0.5802,
+      "step": 17122
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.11578408198667e-06,
+      "loss": 0.6102,
+      "step": 17123
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.114741507233383e-06,
+      "loss": 0.5927,
+      "step": 17124
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1136989666884904e-06,
+      "loss": 0.6644,
+      "step": 17125
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.112656460364344e-06,
+      "loss": 0.5892,
+      "step": 17126
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.111613988273306e-06,
+      "loss": 0.5768,
+      "step": 17127
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.110571550427745e-06,
+      "loss": 0.6252,
+      "step": 17128
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1095291468400084e-06,
+      "loss": 0.6567,
+      "step": 17129
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.108486777522459e-06,
+      "loss": 0.6312,
+      "step": 17130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1074444424874615e-06,
+      "loss": 0.6211,
+      "step": 17131
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.106402141747366e-06,
+      "loss": 0.6178,
+      "step": 17132
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1053598753145304e-06,
+      "loss": 0.6414,
+      "step": 17133
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.104317643201322e-06,
+      "loss": 0.6708,
+      "step": 17134
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.103275445420088e-06,
+      "loss": 0.6191,
+      "step": 17135
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1022332819831845e-06,
+      "loss": 0.6219,
+      "step": 17136
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.101191152902978e-06,
+      "loss": 0.576,
+      "step": 17137
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.100149058191813e-06,
+      "loss": 0.6445,
+      "step": 17138
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.099106997862048e-06,
+      "loss": 0.6548,
+      "step": 17139
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.098064971926044e-06,
+      "loss": 0.6602,
+      "step": 17140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.097022980396148e-06,
+      "loss": 0.6225,
+      "step": 17141
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.095981023284717e-06,
+      "loss": 0.5852,
+      "step": 17142
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.094939100604104e-06,
+      "loss": 0.638,
+      "step": 17143
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.093897212366662e-06,
+      "loss": 0.6842,
+      "step": 17144
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.0928553585847496e-06,
+      "loss": 0.6031,
+      "step": 17145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.091813539270709e-06,
+      "loss": 0.6531,
+      "step": 17146
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.090771754436899e-06,
+      "loss": 0.6572,
+      "step": 17147
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.089730004095672e-06,
+      "loss": 0.6212,
+      "step": 17148
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.088688288259375e-06,
+      "loss": 0.648,
+      "step": 17149
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.087646606940361e-06,
+      "loss": 0.5691,
+      "step": 17150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.086604960150985e-06,
+      "loss": 0.6707,
+      "step": 17151
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.085563347903588e-06,
+      "loss": 0.6356,
+      "step": 17152
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.0845217702105255e-06,
+      "loss": 0.6131,
+      "step": 17153
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.083480227084147e-06,
+      "loss": 0.6169,
+      "step": 17154
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.082438718536798e-06,
+      "loss": 0.6576,
+      "step": 17155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.0813972445808285e-06,
+      "loss": 0.6906,
+      "step": 17156
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.080355805228589e-06,
+      "loss": 0.6384,
+      "step": 17157
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.079314400492423e-06,
+      "loss": 0.6916,
+      "step": 17158
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.0782730303846794e-06,
+      "loss": 0.6564,
+      "step": 17159
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.077231694917707e-06,
+      "loss": 0.651,
+      "step": 17160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.076190394103848e-06,
+      "loss": 0.6077,
+      "step": 17161
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.07514912795545e-06,
+      "loss": 0.6881,
+      "step": 17162
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.074107896484861e-06,
+      "loss": 0.6655,
+      "step": 17163
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.073066699704423e-06,
+      "loss": 0.6216,
+      "step": 17164
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.072025537626482e-06,
+      "loss": 0.6369,
+      "step": 17165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.070984410263384e-06,
+      "loss": 0.6524,
+      "step": 17166
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.069943317627468e-06,
+      "loss": 0.6267,
+      "step": 17167
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.068902259731082e-06,
+      "loss": 0.6333,
+      "step": 17168
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.067861236586569e-06,
+      "loss": 0.659,
+      "step": 17169
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.066820248206267e-06,
+      "loss": 0.672,
+      "step": 17170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.065779294602524e-06,
+      "loss": 0.613,
+      "step": 17171
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.064738375787678e-06,
+      "loss": 0.6424,
+      "step": 17172
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.063697491774073e-06,
+      "loss": 0.6335,
+      "step": 17173
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.062656642574046e-06,
+      "loss": 0.6138,
+      "step": 17174
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.0616158281999435e-06,
+      "loss": 0.6752,
+      "step": 17175
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.060575048664101e-06,
+      "loss": 0.6305,
+      "step": 17176
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.059534303978859e-06,
+      "loss": 0.6168,
+      "step": 17177
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.058493594156559e-06,
+      "loss": 0.6592,
+      "step": 17178
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.057452919209538e-06,
+      "loss": 0.6007,
+      "step": 17179
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.056412279150133e-06,
+      "loss": 0.5947,
+      "step": 17180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.0553716739906875e-06,
+      "loss": 0.6876,
+      "step": 17181
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.054331103743533e-06,
+      "loss": 0.6322,
+      "step": 17182
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.0532905684210105e-06,
+      "loss": 0.6088,
+      "step": 17183
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.052250068035458e-06,
+      "loss": 0.6112,
+      "step": 17184
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.0512096025992074e-06,
+      "loss": 0.6347,
+      "step": 17185
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.050169172124597e-06,
+      "loss": 0.6078,
+      "step": 17186
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.049128776623964e-06,
+      "loss": 0.6314,
+      "step": 17187
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.048088416109641e-06,
+      "loss": 0.6074,
+      "step": 17188
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.047048090593964e-06,
+      "loss": 0.6331,
+      "step": 17189
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.04600780008927e-06,
+      "loss": 0.6293,
+      "step": 17190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.044967544607885e-06,
+      "loss": 0.7025,
+      "step": 17191
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.043927324162152e-06,
+      "loss": 0.6068,
+      "step": 17192
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.0428871387644e-06,
+      "loss": 0.6318,
+      "step": 17193
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.041846988426961e-06,
+      "loss": 0.6723,
+      "step": 17194
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.040806873162167e-06,
+      "loss": 0.6528,
+      "step": 17195
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.039766792982355e-06,
+      "loss": 0.6677,
+      "step": 17196
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.038726747899847e-06,
+      "loss": 0.6836,
+      "step": 17197
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.037686737926981e-06,
+      "loss": 0.6596,
+      "step": 17198
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.036646763076088e-06,
+      "loss": 0.6303,
+      "step": 17199
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.035606823359495e-06,
+      "loss": 0.6526,
+      "step": 17200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.034566918789531e-06,
+      "loss": 0.6783,
+      "step": 17201
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.033527049378532e-06,
+      "loss": 0.6338,
+      "step": 17202
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.032487215138819e-06,
+      "loss": 0.6658,
+      "step": 17203
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.031447416082727e-06,
+      "loss": 0.6316,
+      "step": 17204
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.030407652222578e-06,
+      "loss": 0.6385,
+      "step": 17205
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.029367923570704e-06,
+      "loss": 0.6539,
+      "step": 17206
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.028328230139432e-06,
+      "loss": 0.6896,
+      "step": 17207
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.027288571941086e-06,
+      "loss": 0.6489,
+      "step": 17208
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.026248948987996e-06,
+      "loss": 0.6526,
+      "step": 17209
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.025209361292487e-06,
+      "loss": 0.6694,
+      "step": 17210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.024169808866883e-06,
+      "loss": 0.6159,
+      "step": 17211
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.02313029172351e-06,
+      "loss": 0.5899,
+      "step": 17212
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.0220908098746954e-06,
+      "loss": 0.6615,
+      "step": 17213
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.021051363332758e-06,
+      "loss": 0.6166,
+      "step": 17214
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.020011952110026e-06,
+      "loss": 0.6104,
+      "step": 17215
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.0189725762188256e-06,
+      "loss": 0.6314,
+      "step": 17216
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.017933235671472e-06,
+      "loss": 0.5998,
+      "step": 17217
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.016893930480293e-06,
+      "loss": 0.6651,
+      "step": 17218
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.015854660657613e-06,
+      "loss": 0.625,
+      "step": 17219
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.014815426215749e-06,
+      "loss": 0.5964,
+      "step": 17220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.013776227167022e-06,
+      "loss": 0.6501,
+      "step": 17221
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.01273706352376e-06,
+      "loss": 0.6155,
+      "step": 17222
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.011697935298276e-06,
+      "loss": 0.5903,
+      "step": 17223
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.010658842502893e-06,
+      "loss": 0.6482,
+      "step": 17224
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.009619785149932e-06,
+      "loss": 0.6404,
+      "step": 17225
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.008580763251711e-06,
+      "loss": 0.6418,
+      "step": 17226
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.00754177682055e-06,
+      "loss": 0.6114,
+      "step": 17227
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.006502825868766e-06,
+      "loss": 0.6081,
+      "step": 17228
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.005463910408679e-06,
+      "loss": 0.5921,
+      "step": 17229
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.004425030452602e-06,
+      "loss": 0.6708,
+      "step": 17230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.003386186012859e-06,
+      "loss": 0.621,
+      "step": 17231
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.0023473771017635e-06,
+      "loss": 0.6855,
+      "step": 17232
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.001308603731629e-06,
+      "loss": 0.6292,
+      "step": 17233
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.000269865914779e-06,
+      "loss": 0.6106,
+      "step": 17234
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.99923116366352e-06,
+      "loss": 0.6212,
+      "step": 17235
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.998192496990172e-06,
+      "loss": 0.6022,
+      "step": 17236
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.997153865907052e-06,
+      "loss": 0.6598,
+      "step": 17237
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9961152704264704e-06,
+      "loss": 0.6018,
+      "step": 17238
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.995076710560741e-06,
+      "loss": 0.5991,
+      "step": 17239
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9940381863221805e-06,
+      "loss": 0.6972,
+      "step": 17240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9929996977230975e-06,
+      "loss": 0.6948,
+      "step": 17241
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.991961244775808e-06,
+      "loss": 0.6309,
+      "step": 17242
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.990922827492624e-06,
+      "loss": 0.6478,
+      "step": 17243
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.989884445885855e-06,
+      "loss": 0.6248,
+      "step": 17244
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.988846099967814e-06,
+      "loss": 0.6119,
+      "step": 17245
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.987807789750814e-06,
+      "loss": 0.63,
+      "step": 17246
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.98676951524716e-06,
+      "loss": 0.608,
+      "step": 17247
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.985731276469165e-06,
+      "loss": 0.6182,
+      "step": 17248
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.984693073429141e-06,
+      "loss": 0.6462,
+      "step": 17249
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.983654906139394e-06,
+      "loss": 0.6508,
+      "step": 17250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.982616774612233e-06,
+      "loss": 0.6153,
+      "step": 17251
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.981578678859969e-06,
+      "loss": 0.6264,
+      "step": 17252
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9805406188949065e-06,
+      "loss": 0.6635,
+      "step": 17253
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.979502594729355e-06,
+      "loss": 0.5959,
+      "step": 17254
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.978464606375623e-06,
+      "loss": 0.6475,
+      "step": 17255
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.977426653846011e-06,
+      "loss": 0.5892,
+      "step": 17256
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.976388737152833e-06,
+      "loss": 0.5915,
+      "step": 17257
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.975350856308392e-06,
+      "loss": 0.6169,
+      "step": 17258
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.97431301132499e-06,
+      "loss": 0.5932,
+      "step": 17259
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.973275202214935e-06,
+      "loss": 0.5905,
+      "step": 17260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9722374289905345e-06,
+      "loss": 0.6494,
+      "step": 17261
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.971199691664087e-06,
+      "loss": 0.6792,
+      "step": 17262
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.970161990247899e-06,
+      "loss": 0.6621,
+      "step": 17263
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.969124324754274e-06,
+      "loss": 0.661,
+      "step": 17264
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.968086695195513e-06,
+      "loss": 0.6315,
+      "step": 17265
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.967049101583921e-06,
+      "loss": 0.6702,
+      "step": 17266
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.966011543931798e-06,
+      "loss": 0.6218,
+      "step": 17267
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.964974022251445e-06,
+      "loss": 0.6007,
+      "step": 17268
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.963936536555167e-06,
+      "loss": 0.6464,
+      "step": 17269
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9628990868552595e-06,
+      "loss": 0.6684,
+      "step": 17270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.961861673164026e-06,
+      "loss": 0.5905,
+      "step": 17271
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.960824295493767e-06,
+      "loss": 0.6386,
+      "step": 17272
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9597869538567775e-06,
+      "loss": 0.6503,
+      "step": 17273
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9587496482653615e-06,
+      "loss": 0.6054,
+      "step": 17274
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.957712378731815e-06,
+      "loss": 0.5945,
+      "step": 17275
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.956675145268438e-06,
+      "loss": 0.6992,
+      "step": 17276
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.955637947887526e-06,
+      "loss": 0.6224,
+      "step": 17277
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.954600786601378e-06,
+      "loss": 0.5706,
+      "step": 17278
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.953563661422289e-06,
+      "loss": 0.6653,
+      "step": 17279
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9525265723625565e-06,
+      "loss": 0.6452,
+      "step": 17280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.951489519434478e-06,
+      "loss": 0.5959,
+      "step": 17281
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9504525026503465e-06,
+      "loss": 0.6994,
+      "step": 17282
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.949415522022458e-06,
+      "loss": 0.6225,
+      "step": 17283
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.948378577563108e-06,
+      "loss": 0.6815,
+      "step": 17284
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.947341669284591e-06,
+      "loss": 0.6674,
+      "step": 17285
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9463047971991995e-06,
+      "loss": 0.5994,
+      "step": 17286
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9452679613192285e-06,
+      "loss": 0.65,
+      "step": 17287
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.944231161656968e-06,
+      "loss": 0.6278,
+      "step": 17288
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.943194398224713e-06,
+      "loss": 0.5739,
+      "step": 17289
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.942157671034757e-06,
+      "loss": 0.6031,
+      "step": 17290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.94112098009939e-06,
+      "loss": 0.6196,
+      "step": 17291
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9400843254309005e-06,
+      "loss": 0.5708,
+      "step": 17292
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.939047707041586e-06,
+      "loss": 0.6373,
+      "step": 17293
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.938011124943731e-06,
+      "loss": 0.6204,
+      "step": 17294
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.936974579149627e-06,
+      "loss": 0.6595,
+      "step": 17295
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.935938069671565e-06,
+      "loss": 0.6513,
+      "step": 17296
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.934901596521832e-06,
+      "loss": 0.6394,
+      "step": 17297
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.933865159712718e-06,
+      "loss": 0.6416,
+      "step": 17298
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9328287592565135e-06,
+      "loss": 0.5924,
+      "step": 17299
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.931792395165502e-06,
+      "loss": 0.6929,
+      "step": 17300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.930756067451974e-06,
+      "loss": 0.6585,
+      "step": 17301
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.929719776128217e-06,
+      "loss": 0.6617,
+      "step": 17302
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.928683521206512e-06,
+      "loss": 0.6293,
+      "step": 17303
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9276473026991505e-06,
+      "loss": 0.5743,
+      "step": 17304
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.926611120618419e-06,
+      "loss": 0.6305,
+      "step": 17305
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.925574974976599e-06,
+      "loss": 0.6715,
+      "step": 17306
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.924538865785976e-06,
+      "loss": 0.646,
+      "step": 17307
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.923502793058836e-06,
+      "loss": 0.6078,
+      "step": 17308
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9224667568074624e-06,
+      "loss": 0.6378,
+      "step": 17309
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.921430757044137e-06,
+      "loss": 0.6543,
+      "step": 17310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.92039479378115e-06,
+      "loss": 0.6207,
+      "step": 17311
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.919358867030771e-06,
+      "loss": 0.6427,
+      "step": 17312
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.918322976805292e-06,
+      "loss": 0.6299,
+      "step": 17313
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.917287123116997e-06,
+      "loss": 0.6497,
+      "step": 17314
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.916251305978156e-06,
+      "loss": 0.5954,
+      "step": 17315
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.915215525401059e-06,
+      "loss": 0.6281,
+      "step": 17316
+    },
+    {
+      "epoch": 0.6,
+      "eval_loss": 0.5753005146980286,
+      "eval_runtime": 4931.9747,
+      "eval_samples_per_second": 111.85,
+      "eval_steps_per_second": 37.283,
+      "step": 17316
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.914179781397989e-06,
+      "loss": 0.6366,
+      "step": 17317
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.913144073981214e-06,
+      "loss": 0.6408,
+      "step": 17318
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.912108403163024e-06,
+      "loss": 0.6058,
+      "step": 17319
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.911072768955698e-06,
+      "loss": 0.5771,
+      "step": 17320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.910037171371506e-06,
+      "loss": 0.5992,
+      "step": 17321
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.909001610422735e-06,
+      "loss": 0.6282,
+      "step": 17322
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.907966086121663e-06,
+      "loss": 0.6064,
+      "step": 17323
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.906930598480559e-06,
+      "loss": 0.6187,
+      "step": 17324
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.9058951475117075e-06,
+      "loss": 0.6173,
+      "step": 17325
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.904859733227388e-06,
+      "loss": 0.6615,
+      "step": 17326
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.903824355639865e-06,
+      "loss": 0.6239,
+      "step": 17327
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.902789014761426e-06,
+      "loss": 0.6735,
+      "step": 17328
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.90175371060434e-06,
+      "loss": 0.6182,
+      "step": 17329
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.900718443180878e-06,
+      "loss": 0.694,
+      "step": 17330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.899683212503327e-06,
+      "loss": 0.6443,
+      "step": 17331
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8986480185839486e-06,
+      "loss": 0.6039,
+      "step": 17332
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.897612861435019e-06,
+      "loss": 0.6139,
+      "step": 17333
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.896577741068822e-06,
+      "loss": 0.6422,
+      "step": 17334
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.895542657497614e-06,
+      "loss": 0.6709,
+      "step": 17335
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.894507610733676e-06,
+      "loss": 0.64,
+      "step": 17336
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.893472600789284e-06,
+      "loss": 0.6255,
+      "step": 17337
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.892437627676698e-06,
+      "loss": 0.6559,
+      "step": 17338
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.891402691408196e-06,
+      "loss": 0.6344,
+      "step": 17339
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.890367791996051e-06,
+      "loss": 0.5857,
+      "step": 17340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.889332929452527e-06,
+      "loss": 0.5676,
+      "step": 17341
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.888298103789894e-06,
+      "loss": 0.6266,
+      "step": 17342
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.88726331502043e-06,
+      "loss": 0.6919,
+      "step": 17343
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.886228563156391e-06,
+      "loss": 0.6636,
+      "step": 17344
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.88519384821005e-06,
+      "loss": 0.6084,
+      "step": 17345
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8841591701936826e-06,
+      "loss": 0.599,
+      "step": 17346
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8831245291195466e-06,
+      "loss": 0.6275,
+      "step": 17347
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.882089924999909e-06,
+      "loss": 0.6216,
+      "step": 17348
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.881055357847046e-06,
+      "loss": 0.6447,
+      "step": 17349
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.880020827673213e-06,
+      "loss": 0.594,
+      "step": 17350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.878986334490678e-06,
+      "loss": 0.632,
+      "step": 17351
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.877951878311715e-06,
+      "loss": 0.6086,
+      "step": 17352
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.876917459148576e-06,
+      "loss": 0.6538,
+      "step": 17353
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.87588307701353e-06,
+      "loss": 0.6549,
+      "step": 17354
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.874848731918848e-06,
+      "loss": 0.6416,
+      "step": 17355
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8738144238767855e-06,
+      "loss": 0.6793,
+      "step": 17356
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.872780152899603e-06,
+      "loss": 0.6852,
+      "step": 17357
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.871745918999575e-06,
+      "loss": 0.6034,
+      "step": 17358
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.870711722188953e-06,
+      "loss": 0.6415,
+      "step": 17359
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.869677562479999e-06,
+      "loss": 0.6733,
+      "step": 17360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.868643439884983e-06,
+      "loss": 0.6851,
+      "step": 17361
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.867609354416158e-06,
+      "loss": 0.6481,
+      "step": 17362
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.866575306085783e-06,
+      "loss": 0.6428,
+      "step": 17363
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8655412949061286e-06,
+      "loss": 0.6216,
+      "step": 17364
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8645073208894434e-06,
+      "loss": 0.6659,
+      "step": 17365
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.863473384047987e-06,
+      "loss": 0.6617,
+      "step": 17366
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.862439484394028e-06,
+      "loss": 0.6306,
+      "step": 17367
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8614056219398136e-06,
+      "loss": 0.6407,
+      "step": 17368
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.860371796697605e-06,
+      "loss": 0.6566,
+      "step": 17369
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8593380086796655e-06,
+      "loss": 0.6577,
+      "step": 17370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.858304257898243e-06,
+      "loss": 0.6003,
+      "step": 17371
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.857270544365597e-06,
+      "loss": 0.5939,
+      "step": 17372
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8562368680939885e-06,
+      "loss": 0.5741,
+      "step": 17373
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.855203229095666e-06,
+      "loss": 0.6332,
+      "step": 17374
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.854169627382884e-06,
+      "loss": 0.6276,
+      "step": 17375
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.853136062967908e-06,
+      "loss": 0.6045,
+      "step": 17376
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.85210253586298e-06,
+      "loss": 0.604,
+      "step": 17377
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.851069046080356e-06,
+      "loss": 0.611,
+      "step": 17378
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.850035593632299e-06,
+      "loss": 0.6683,
+      "step": 17379
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.84900217853105e-06,
+      "loss": 0.6329,
+      "step": 17380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8479688007888645e-06,
+      "loss": 0.6667,
+      "step": 17381
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.846935460418003e-06,
+      "loss": 0.627,
+      "step": 17382
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.845902157430706e-06,
+      "loss": 0.6366,
+      "step": 17383
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.844868891839226e-06,
+      "loss": 0.5724,
+      "step": 17384
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.843835663655824e-06,
+      "loss": 0.5909,
+      "step": 17385
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.842802472892737e-06,
+      "loss": 0.645,
+      "step": 17386
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8417693195622195e-06,
+      "loss": 0.6295,
+      "step": 17387
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.840736203676528e-06,
+      "loss": 0.5996,
+      "step": 17388
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8397031252479004e-06,
+      "loss": 0.6485,
+      "step": 17389
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.838670084288592e-06,
+      "loss": 0.6567,
+      "step": 17390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.837637080810848e-06,
+      "loss": 0.5989,
+      "step": 17391
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.836604114826917e-06,
+      "loss": 0.6137,
+      "step": 17392
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.835571186349047e-06,
+      "loss": 0.6729,
+      "step": 17393
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.834538295389481e-06,
+      "loss": 0.5682,
+      "step": 17394
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.83350544196047e-06,
+      "loss": 0.6401,
+      "step": 17395
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.832472626074258e-06,
+      "loss": 0.643,
+      "step": 17396
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.831439847743089e-06,
+      "loss": 0.6723,
+      "step": 17397
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.830407106979207e-06,
+      "loss": 0.6446,
+      "step": 17398
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8293744037948614e-06,
+      "loss": 0.6194,
+      "step": 17399
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.828341738202291e-06,
+      "loss": 0.6328,
+      "step": 17400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.827309110213742e-06,
+      "loss": 0.6083,
+      "step": 17401
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.82627651984146e-06,
+      "loss": 0.5512,
+      "step": 17402
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.82524396709768e-06,
+      "loss": 0.6647,
+      "step": 17403
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.82421145199465e-06,
+      "loss": 0.5857,
+      "step": 17404
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.823178974544613e-06,
+      "loss": 0.5913,
+      "step": 17405
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.822146534759806e-06,
+      "loss": 0.6046,
+      "step": 17406
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.821114132652471e-06,
+      "loss": 0.6593,
+      "step": 17407
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.820081768234852e-06,
+      "loss": 0.6317,
+      "step": 17408
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.819049441519184e-06,
+      "loss": 0.6545,
+      "step": 17409
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.818017152517708e-06,
+      "loss": 0.6305,
+      "step": 17410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.816984901242667e-06,
+      "loss": 0.5646,
+      "step": 17411
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8159526877062955e-06,
+      "loss": 0.6531,
+      "step": 17412
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.814920511920831e-06,
+      "loss": 0.6498,
+      "step": 17413
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.813888373898517e-06,
+      "loss": 0.6435,
+      "step": 17414
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.812856273651582e-06,
+      "loss": 0.6479,
+      "step": 17415
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.8118242111922684e-06,
+      "loss": 0.6321,
+      "step": 17416
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.810792186532815e-06,
+      "loss": 0.5938,
+      "step": 17417
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.809760199685451e-06,
+      "loss": 0.6284,
+      "step": 17418
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.808728250662417e-06,
+      "loss": 0.6504,
+      "step": 17419
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.807696339475949e-06,
+      "loss": 0.6288,
+      "step": 17420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.806664466138276e-06,
+      "loss": 0.6487,
+      "step": 17421
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.805632630661634e-06,
+      "loss": 0.6733,
+      "step": 17422
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.804600833058262e-06,
+      "loss": 0.6077,
+      "step": 17423
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.803569073340387e-06,
+      "loss": 0.6316,
+      "step": 17424
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.802537351520244e-06,
+      "loss": 0.6459,
+      "step": 17425
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.801505667610067e-06,
+      "loss": 0.5881,
+      "step": 17426
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.800474021622086e-06,
+      "loss": 0.6438,
+      "step": 17427
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.799442413568531e-06,
+      "loss": 0.5765,
+      "step": 17428
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.798410843461638e-06,
+      "loss": 0.6457,
+      "step": 17429
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.797379311313632e-06,
+      "loss": 0.6167,
+      "step": 17430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.796347817136746e-06,
+      "loss": 0.6505,
+      "step": 17431
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.795316360943212e-06,
+      "loss": 0.6464,
+      "step": 17432
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.794284942745253e-06,
+      "loss": 0.6334,
+      "step": 17433
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.793253562555102e-06,
+      "loss": 0.6035,
+      "step": 17434
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.792222220384989e-06,
+      "loss": 0.6119,
+      "step": 17435
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.791190916247137e-06,
+      "loss": 0.6153,
+      "step": 17436
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.790159650153776e-06,
+      "loss": 0.5794,
+      "step": 17437
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.7891284221171336e-06,
+      "loss": 0.5851,
+      "step": 17438
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.788097232149436e-06,
+      "loss": 0.6214,
+      "step": 17439
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.787066080262905e-06,
+      "loss": 0.6875,
+      "step": 17440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.786034966469775e-06,
+      "loss": 0.6156,
+      "step": 17441
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.7850038907822635e-06,
+      "loss": 0.6196,
+      "step": 17442
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.783972853212598e-06,
+      "loss": 0.6214,
+      "step": 17443
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.782941853773004e-06,
+      "loss": 0.6379,
+      "step": 17444
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.7819108924757025e-06,
+      "loss": 0.6493,
+      "step": 17445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.780879969332919e-06,
+      "loss": 0.6068,
+      "step": 17446
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.779849084356877e-06,
+      "loss": 0.6096,
+      "step": 17447
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.778818237559794e-06,
+      "loss": 0.6243,
+      "step": 17448
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.777787428953897e-06,
+      "loss": 0.6653,
+      "step": 17449
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.7767566585514085e-06,
+      "loss": 0.6407,
+      "step": 17450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.775725926364545e-06,
+      "loss": 0.5867,
+      "step": 17451
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.774695232405532e-06,
+      "loss": 0.6794,
+      "step": 17452
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.773664576686584e-06,
+      "loss": 0.6282,
+      "step": 17453
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.772633959219924e-06,
+      "loss": 0.6085,
+      "step": 17454
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.771603380017773e-06,
+      "loss": 0.6651,
+      "step": 17455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.770572839092346e-06,
+      "loss": 0.6759,
+      "step": 17456
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.769542336455863e-06,
+      "loss": 0.637,
+      "step": 17457
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.768511872120544e-06,
+      "loss": 0.6422,
+      "step": 17458
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.767481446098604e-06,
+      "loss": 0.5852,
+      "step": 17459
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.766451058402259e-06,
+      "loss": 0.6128,
+      "step": 17460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.765420709043731e-06,
+      "loss": 0.6354,
+      "step": 17461
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.764390398035229e-06,
+      "loss": 0.6459,
+      "step": 17462
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7633601253889715e-06,
+      "loss": 0.6205,
+      "step": 17463
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7623298911171766e-06,
+      "loss": 0.6271,
+      "step": 17464
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.761299695232054e-06,
+      "loss": 0.6428,
+      "step": 17465
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.760269537745821e-06,
+      "loss": 0.6482,
+      "step": 17466
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7592394186706935e-06,
+      "loss": 0.6259,
+      "step": 17467
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.758209338018879e-06,
+      "loss": 0.6069,
+      "step": 17468
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.757179295802594e-06,
+      "loss": 0.5661,
+      "step": 17469
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.756149292034053e-06,
+      "loss": 0.6214,
+      "step": 17470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.755119326725464e-06,
+      "loss": 0.6246,
+      "step": 17471
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.75408939988904e-06,
+      "loss": 0.6616,
+      "step": 17472
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.753059511536994e-06,
+      "loss": 0.6642,
+      "step": 17473
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.752029661681533e-06,
+      "loss": 0.6424,
+      "step": 17474
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.75099985033487e-06,
+      "loss": 0.5834,
+      "step": 17475
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.749970077509214e-06,
+      "loss": 0.6235,
+      "step": 17476
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.748940343216772e-06,
+      "loss": 0.6032,
+      "step": 17477
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.747910647469756e-06,
+      "loss": 0.6163,
+      "step": 17478
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.746880990280376e-06,
+      "loss": 0.6845,
+      "step": 17479
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.745851371660833e-06,
+      "loss": 0.6096,
+      "step": 17480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.744821791623339e-06,
+      "loss": 0.6021,
+      "step": 17481
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.743792250180103e-06,
+      "loss": 0.6343,
+      "step": 17482
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7427627473433276e-06,
+      "loss": 0.652,
+      "step": 17483
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.741733283125219e-06,
+      "loss": 0.5946,
+      "step": 17484
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.740703857537987e-06,
+      "loss": 0.5993,
+      "step": 17485
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.739674470593831e-06,
+      "loss": 0.659,
+      "step": 17486
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.738645122304959e-06,
+      "loss": 0.634,
+      "step": 17487
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7376158126835775e-06,
+      "loss": 0.6155,
+      "step": 17488
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.736586541741884e-06,
+      "loss": 0.6274,
+      "step": 17489
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.735557309492087e-06,
+      "loss": 0.6061,
+      "step": 17490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.734528115946388e-06,
+      "loss": 0.6894,
+      "step": 17491
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.733498961116986e-06,
+      "loss": 0.6645,
+      "step": 17492
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.732469845016089e-06,
+      "loss": 0.698,
+      "step": 17493
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7314407676558955e-06,
+      "loss": 0.6315,
+      "step": 17494
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.730411729048605e-06,
+      "loss": 0.6778,
+      "step": 17495
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.729382729206419e-06,
+      "loss": 0.6722,
+      "step": 17496
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7283537681415416e-06,
+      "loss": 0.6518,
+      "step": 17497
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.727324845866166e-06,
+      "loss": 0.605,
+      "step": 17498
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.726295962392493e-06,
+      "loss": 0.6284,
+      "step": 17499
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.725267117732726e-06,
+      "loss": 0.645,
+      "step": 17500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.724238311899058e-06,
+      "loss": 0.6187,
+      "step": 17501
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.723209544903689e-06,
+      "loss": 0.6381,
+      "step": 17502
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.722180816758817e-06,
+      "loss": 0.6317,
+      "step": 17503
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.721152127476635e-06,
+      "loss": 0.6458,
+      "step": 17504
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.720123477069345e-06,
+      "loss": 0.6263,
+      "step": 17505
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.719094865549141e-06,
+      "loss": 0.6118,
+      "step": 17506
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.718066292928214e-06,
+      "loss": 0.6431,
+      "step": 17507
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.717037759218765e-06,
+      "loss": 0.5734,
+      "step": 17508
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7160092644329875e-06,
+      "loss": 0.5988,
+      "step": 17509
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.714980808583071e-06,
+      "loss": 0.6262,
+      "step": 17510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7139523916812154e-06,
+      "loss": 0.6057,
+      "step": 17511
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7129240137396125e-06,
+      "loss": 0.6198,
+      "step": 17512
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.71189567477045e-06,
+      "loss": 0.6532,
+      "step": 17513
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.710867374785929e-06,
+      "loss": 0.6365,
+      "step": 17514
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.709839113798229e-06,
+      "loss": 0.6877,
+      "step": 17515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.708810891819551e-06,
+      "loss": 0.6188,
+      "step": 17516
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.707782708862087e-06,
+      "loss": 0.5743,
+      "step": 17517
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.7067545649380185e-06,
+      "loss": 0.6271,
+      "step": 17518
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.705726460059543e-06,
+      "loss": 0.6757,
+      "step": 17519
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.70469839423885e-06,
+      "loss": 0.6265,
+      "step": 17520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.70367036748812e-06,
+      "loss": 0.6115,
+      "step": 17521
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.702642379819551e-06,
+      "loss": 0.6588,
+      "step": 17522
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.701614431245331e-06,
+      "loss": 0.6608,
+      "step": 17523
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.700586521777639e-06,
+      "loss": 0.6151,
+      "step": 17524
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.699558651428668e-06,
+      "loss": 0.5909,
+      "step": 17525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.69853082021061e-06,
+      "loss": 0.6779,
+      "step": 17526
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.697503028135639e-06,
+      "loss": 0.6585,
+      "step": 17527
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.69647527521595e-06,
+      "loss": 0.6502,
+      "step": 17528
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.695447561463728e-06,
+      "loss": 0.612,
+      "step": 17529
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.69441988689115e-06,
+      "loss": 0.6242,
+      "step": 17530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.693392251510409e-06,
+      "loss": 0.6951,
+      "step": 17531
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6923646553336885e-06,
+      "loss": 0.6559,
+      "step": 17532
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.691337098373163e-06,
+      "loss": 0.6274,
+      "step": 17533
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.690309580641025e-06,
+      "loss": 0.605,
+      "step": 17534
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6892821021494566e-06,
+      "loss": 0.5955,
+      "step": 17535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.688254662910632e-06,
+      "loss": 0.6372,
+      "step": 17536
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.687227262936739e-06,
+      "loss": 0.5825,
+      "step": 17537
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6861999022399605e-06,
+      "loss": 0.6076,
+      "step": 17538
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.685172580832468e-06,
+      "loss": 0.6765,
+      "step": 17539
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.684145298726452e-06,
+      "loss": 0.6394,
+      "step": 17540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.683118055934091e-06,
+      "loss": 0.6417,
+      "step": 17541
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.682090852467556e-06,
+      "loss": 0.639,
+      "step": 17542
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.681063688339032e-06,
+      "loss": 0.5924,
+      "step": 17543
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.680036563560701e-06,
+      "loss": 0.6012,
+      "step": 17544
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.67900947814473e-06,
+      "loss": 0.5833,
+      "step": 17545
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.677982432103305e-06,
+      "loss": 0.6109,
+      "step": 17546
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.676955425448605e-06,
+      "loss": 0.5765,
+      "step": 17547
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.675928458192795e-06,
+      "loss": 0.601,
+      "step": 17548
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.674901530348062e-06,
+      "loss": 0.6092,
+      "step": 17549
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.67387464192658e-06,
+      "loss": 0.6449,
+      "step": 17550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.672847792940515e-06,
+      "loss": 0.6457,
+      "step": 17551
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.671820983402052e-06,
+      "loss": 0.6622,
+      "step": 17552
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.670794213323365e-06,
+      "loss": 0.6035,
+      "step": 17553
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.669767482716618e-06,
+      "loss": 0.6371,
+      "step": 17554
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.668740791593992e-06,
+      "loss": 0.6042,
+      "step": 17555
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6677141399676626e-06,
+      "loss": 0.5917,
+      "step": 17556
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.666687527849791e-06,
+      "loss": 0.6698,
+      "step": 17557
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.665660955252557e-06,
+      "loss": 0.6151,
+      "step": 17558
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.664634422188134e-06,
+      "loss": 0.644,
+      "step": 17559
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.663607928668685e-06,
+      "loss": 0.6264,
+      "step": 17560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.662581474706385e-06,
+      "loss": 0.6872,
+      "step": 17561
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.661555060313407e-06,
+      "loss": 0.62,
+      "step": 17562
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6605286855019125e-06,
+      "loss": 0.6473,
+      "step": 17563
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.659502350284076e-06,
+      "loss": 0.6489,
+      "step": 17564
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6584760546720694e-06,
+      "loss": 0.6034,
+      "step": 17565
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6574497986780495e-06,
+      "loss": 0.6373,
+      "step": 17566
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.656423582314193e-06,
+      "loss": 0.6377,
+      "step": 17567
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6553974055926694e-06,
+      "loss": 0.6393,
+      "step": 17568
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.654371268525632e-06,
+      "loss": 0.6006,
+      "step": 17569
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.653345171125261e-06,
+      "loss": 0.6113,
+      "step": 17570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.652319113403718e-06,
+      "loss": 0.6262,
+      "step": 17571
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6512930953731614e-06,
+      "loss": 0.6687,
+      "step": 17572
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.650267117045763e-06,
+      "loss": 0.6622,
+      "step": 17573
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.64924117843369e-06,
+      "loss": 0.6826,
+      "step": 17574
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.648215279549095e-06,
+      "loss": 0.5894,
+      "step": 17575
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.64718942040415e-06,
+      "loss": 0.6357,
+      "step": 17576
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6461636010110195e-06,
+      "loss": 0.6043,
+      "step": 17577
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.645137821381858e-06,
+      "loss": 0.6767,
+      "step": 17578
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6441120815288354e-06,
+      "loss": 0.6269,
+      "step": 17579
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.643086381464106e-06,
+      "loss": 0.6174,
+      "step": 17580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6420607211998325e-06,
+      "loss": 0.633,
+      "step": 17581
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.641035100748182e-06,
+      "loss": 0.5701,
+      "step": 17582
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.640009520121307e-06,
+      "loss": 0.6379,
+      "step": 17583
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.638983979331366e-06,
+      "loss": 0.6118,
+      "step": 17584
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.637958478390527e-06,
+      "loss": 0.6438,
+      "step": 17585
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6369330173109405e-06,
+      "loss": 0.6707,
+      "step": 17586
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.635907596104766e-06,
+      "loss": 0.6208,
+      "step": 17587
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6348822147841686e-06,
+      "loss": 0.6147,
+      "step": 17588
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.633856873361293e-06,
+      "loss": 0.6357,
+      "step": 17589
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.632831571848302e-06,
+      "loss": 0.6676,
+      "step": 17590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.631806310257358e-06,
+      "loss": 0.5974,
+      "step": 17591
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.630781088600608e-06,
+      "loss": 0.6069,
+      "step": 17592
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6297559068902074e-06,
+      "loss": 0.6333,
+      "step": 17593
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.628730765138321e-06,
+      "loss": 0.6293,
+      "step": 17594
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6277056633570904e-06,
+      "loss": 0.684,
+      "step": 17595
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.626680601558674e-06,
+      "loss": 0.6147,
+      "step": 17596
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.625655579755232e-06,
+      "loss": 0.6885,
+      "step": 17597
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.624630597958907e-06,
+      "loss": 0.7076,
+      "step": 17598
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.623605656181855e-06,
+      "loss": 0.684,
+      "step": 17599
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6225807544362366e-06,
+      "loss": 0.5813,
+      "step": 17600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6215558927341895e-06,
+      "loss": 0.6182,
+      "step": 17601
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6205310710878705e-06,
+      "loss": 0.6254,
+      "step": 17602
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.619506289509435e-06,
+      "loss": 0.6521,
+      "step": 17603
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6184815480110265e-06,
+      "loss": 0.6675,
+      "step": 17604
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.617456846604793e-06,
+      "loss": 0.6017,
+      "step": 17605
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.616432185302895e-06,
+      "loss": 0.6236,
+      "step": 17606
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.61540756411747e-06,
+      "loss": 0.6182,
+      "step": 17607
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.614382983060667e-06,
+      "loss": 0.6288,
+      "step": 17608
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.613358442144643e-06,
+      "loss": 0.666,
+      "step": 17609
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.6123339413815345e-06,
+      "loss": 0.6447,
+      "step": 17610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.61130948078349e-06,
+      "loss": 0.6459,
+      "step": 17611
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.610285060362665e-06,
+      "loss": 0.6545,
+      "step": 17612
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.609260680131195e-06,
+      "loss": 0.6215,
+      "step": 17613
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.608236340101226e-06,
+      "loss": 0.6501,
+      "step": 17614
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.607212040284913e-06,
+      "loss": 0.6556,
+      "step": 17615
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.606187780694389e-06,
+      "loss": 0.6412,
+      "step": 17616
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.605163561341798e-06,
+      "loss": 0.6339,
+      "step": 17617
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.604139382239296e-06,
+      "loss": 0.6383,
+      "step": 17618
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.603115243399012e-06,
+      "loss": 0.6344,
+      "step": 17619
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.602091144833092e-06,
+      "loss": 0.5936,
+      "step": 17620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.601067086553687e-06,
+      "loss": 0.605,
+      "step": 17621
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.600043068572926e-06,
+      "loss": 0.633,
+      "step": 17622
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5990190909029535e-06,
+      "loss": 0.6525,
+      "step": 17623
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.597995153555919e-06,
+      "loss": 0.662,
+      "step": 17624
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5969712565439515e-06,
+      "loss": 0.6407,
+      "step": 17625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.595947399879191e-06,
+      "loss": 0.6498,
+      "step": 17626
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.594923583573787e-06,
+      "loss": 0.6659,
+      "step": 17627
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.593899807639868e-06,
+      "loss": 0.6472,
+      "step": 17628
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.592876072089573e-06,
+      "loss": 0.6446,
+      "step": 17629
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5918523769350485e-06,
+      "loss": 0.6653,
+      "step": 17630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.59082872218842e-06,
+      "loss": 0.6583,
+      "step": 17631
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.58980510786183e-06,
+      "loss": 0.6212,
+      "step": 17632
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5887815339674185e-06,
+      "loss": 0.6651,
+      "step": 17633
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.587758000517314e-06,
+      "loss": 0.6514,
+      "step": 17634
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5867345075236534e-06,
+      "loss": 0.6734,
+      "step": 17635
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.585711054998578e-06,
+      "loss": 0.6312,
+      "step": 17636
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.584687642954213e-06,
+      "loss": 0.6094,
+      "step": 17637
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5836642714026946e-06,
+      "loss": 0.6127,
+      "step": 17638
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.582640940356164e-06,
+      "loss": 0.6393,
+      "step": 17639
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.581617649826745e-06,
+      "loss": 0.6672,
+      "step": 17640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.580594399826574e-06,
+      "loss": 0.6234,
+      "step": 17641
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5795711903677796e-06,
+      "loss": 0.6508,
+      "step": 17642
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.578548021462496e-06,
+      "loss": 0.6432,
+      "step": 17643
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.577524893122858e-06,
+      "loss": 0.63,
+      "step": 17644
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.576501805360986e-06,
+      "loss": 0.7267,
+      "step": 17645
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.575478758189018e-06,
+      "loss": 0.6641,
+      "step": 17646
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.574455751619083e-06,
+      "loss": 0.6049,
+      "step": 17647
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5734327856633075e-06,
+      "loss": 0.5648,
+      "step": 17648
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.57240986033382e-06,
+      "loss": 0.6157,
+      "step": 17649
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.571386975642751e-06,
+      "loss": 0.6743,
+      "step": 17650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5703641316022246e-06,
+      "loss": 0.655,
+      "step": 17651
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5693413282243705e-06,
+      "loss": 0.6255,
+      "step": 17652
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.568318565521316e-06,
+      "loss": 0.6258,
+      "step": 17653
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.567295843505185e-06,
+      "loss": 0.6265,
+      "step": 17654
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.566273162188104e-06,
+      "loss": 0.62,
+      "step": 17655
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5652505215822005e-06,
+      "loss": 0.6515,
+      "step": 17656
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.564227921699594e-06,
+      "loss": 0.5967,
+      "step": 17657
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.563205362552413e-06,
+      "loss": 0.6054,
+      "step": 17658
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5621828441527815e-06,
+      "loss": 0.6272,
+      "step": 17659
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.561160366512819e-06,
+      "loss": 0.6804,
+      "step": 17660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.56013792964465e-06,
+      "loss": 0.6934,
+      "step": 17661
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5591155335604e-06,
+      "loss": 0.6376,
+      "step": 17662
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.558093178272187e-06,
+      "loss": 0.6384,
+      "step": 17663
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.557070863792132e-06,
+      "loss": 0.6584,
+      "step": 17664
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.55604859013236e-06,
+      "loss": 0.5966,
+      "step": 17665
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.555026357304986e-06,
+      "loss": 0.6708,
+      "step": 17666
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5540041653221334e-06,
+      "loss": 0.6222,
+      "step": 17667
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.552982014195921e-06,
+      "loss": 0.6003,
+      "step": 17668
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.551959903938468e-06,
+      "loss": 0.6305,
+      "step": 17669
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.550937834561889e-06,
+      "loss": 0.6479,
+      "step": 17670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.549915806078309e-06,
+      "loss": 0.6365,
+      "step": 17671
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.548893818499839e-06,
+      "loss": 0.6118,
+      "step": 17672
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.547871871838598e-06,
+      "loss": 0.6171,
+      "step": 17673
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.546849966106705e-06,
+      "loss": 0.5966,
+      "step": 17674
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.545828101316273e-06,
+      "loss": 0.7067,
+      "step": 17675
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5448062774794175e-06,
+      "loss": 0.61,
+      "step": 17676
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.543784494608256e-06,
+      "loss": 0.6273,
+      "step": 17677
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.542762752714901e-06,
+      "loss": 0.6237,
+      "step": 17678
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5417410518114655e-06,
+      "loss": 0.5987,
+      "step": 17679
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.540719391910068e-06,
+      "loss": 0.6584,
+      "step": 17680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.539697773022815e-06,
+      "loss": 0.604,
+      "step": 17681
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.538676195161821e-06,
+      "loss": 0.6601,
+      "step": 17682
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.537654658339204e-06,
+      "loss": 0.5864,
+      "step": 17683
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.536633162567067e-06,
+      "loss": 0.626,
+      "step": 17684
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.535611707857527e-06,
+      "loss": 0.6325,
+      "step": 17685
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.534590294222693e-06,
+      "loss": 0.6233,
+      "step": 17686
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.533568921674673e-06,
+      "loss": 0.6599,
+      "step": 17687
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.532547590225578e-06,
+      "loss": 0.6194,
+      "step": 17688
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.531526299887521e-06,
+      "loss": 0.6645,
+      "step": 17689
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.530505050672604e-06,
+      "loss": 0.6437,
+      "step": 17690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.52948384259294e-06,
+      "loss": 0.6236,
+      "step": 17691
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.528462675660636e-06,
+      "loss": 0.6278,
+      "step": 17692
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.5274415498877975e-06,
+      "loss": 0.67,
+      "step": 17693
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.526420465286532e-06,
+      "loss": 0.6417,
+      "step": 17694
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.525399421868948e-06,
+      "loss": 0.6168,
+      "step": 17695
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.524378419647146e-06,
+      "loss": 0.6358,
+      "step": 17696
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.523357458633237e-06,
+      "loss": 0.6456,
+      "step": 17697
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.522336538839324e-06,
+      "loss": 0.6466,
+      "step": 17698
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.521315660277508e-06,
+      "loss": 0.5973,
+      "step": 17699
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.520294822959896e-06,
+      "loss": 0.6708,
+      "step": 17700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.519274026898592e-06,
+      "loss": 0.6263,
+      "step": 17701
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.518253272105697e-06,
+      "loss": 0.6553,
+      "step": 17702
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.517232558593314e-06,
+      "loss": 0.6244,
+      "step": 17703
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.516211886373545e-06,
+      "loss": 0.687,
+      "step": 17704
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.515191255458489e-06,
+      "loss": 0.6059,
+      "step": 17705
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.514170665860251e-06,
+      "loss": 0.6115,
+      "step": 17706
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.513150117590929e-06,
+      "loss": 0.6498,
+      "step": 17707
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.512129610662623e-06,
+      "loss": 0.645,
+      "step": 17708
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.511109145087433e-06,
+      "loss": 0.6419,
+      "step": 17709
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.510088720877457e-06,
+      "loss": 0.6881,
+      "step": 17710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.509068338044794e-06,
+      "loss": 0.6866,
+      "step": 17711
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.508047996601543e-06,
+      "loss": 0.5954,
+      "step": 17712
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.507027696559799e-06,
+      "loss": 0.6581,
+      "step": 17713
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.506007437931661e-06,
+      "loss": 0.6275,
+      "step": 17714
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.504987220729225e-06,
+      "loss": 0.6208,
+      "step": 17715
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.503967044964585e-06,
+      "loss": 0.6264,
+      "step": 17716
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.502946910649839e-06,
+      "loss": 0.6423,
+      "step": 17717
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.501926817797084e-06,
+      "loss": 0.6142,
+      "step": 17718
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.500906766418408e-06,
+      "loss": 0.6744,
+      "step": 17719
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.499886756525909e-06,
+      "loss": 0.7066,
+      "step": 17720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.498866788131684e-06,
+      "loss": 0.6359,
+      "step": 17721
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.497846861247818e-06,
+      "loss": 0.6071,
+      "step": 17722
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.496826975886408e-06,
+      "loss": 0.5949,
+      "step": 17723
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.495807132059549e-06,
+      "loss": 0.6795,
+      "step": 17724
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.4947873297793255e-06,
+      "loss": 0.6223,
+      "step": 17725
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.493767569057833e-06,
+      "loss": 0.5776,
+      "step": 17726
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.492747849907164e-06,
+      "loss": 0.6408,
+      "step": 17727
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.491728172339402e-06,
+      "loss": 0.583,
+      "step": 17728
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.490708536366643e-06,
+      "loss": 0.6322,
+      "step": 17729
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.489688942000975e-06,
+      "loss": 0.6473,
+      "step": 17730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.488669389254483e-06,
+      "loss": 0.608,
+      "step": 17731
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.487649878139256e-06,
+      "loss": 0.6452,
+      "step": 17732
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.486630408667386e-06,
+      "loss": 0.6567,
+      "step": 17733
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.4856109808509545e-06,
+      "loss": 0.6912,
+      "step": 17734
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.484591594702051e-06,
+      "loss": 0.669,
+      "step": 17735
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.483572250232764e-06,
+      "loss": 0.6363,
+      "step": 17736
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.482552947455172e-06,
+      "loss": 0.6089,
+      "step": 17737
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.481533686381366e-06,
+      "loss": 0.6559,
+      "step": 17738
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.480514467023431e-06,
+      "loss": 0.5954,
+      "step": 17739
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.479495289393447e-06,
+      "loss": 0.602,
+      "step": 17740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.478476153503499e-06,
+      "loss": 0.625,
+      "step": 17741
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.477457059365674e-06,
+      "loss": 0.5956,
+      "step": 17742
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.47643800699205e-06,
+      "loss": 0.6198,
+      "step": 17743
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.475418996394709e-06,
+      "loss": 0.6095,
+      "step": 17744
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.474400027585736e-06,
+      "loss": 0.5854,
+      "step": 17745
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.473381100577212e-06,
+      "loss": 0.6284,
+      "step": 17746
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.472362215381214e-06,
+      "loss": 0.6503,
+      "step": 17747
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.471343372009826e-06,
+      "loss": 0.6056,
+      "step": 17748
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.470324570475124e-06,
+      "loss": 0.6414,
+      "step": 17749
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.4693058107891905e-06,
+      "loss": 0.6234,
+      "step": 17750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.468287092964105e-06,
+      "loss": 0.5816,
+      "step": 17751
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.467268417011941e-06,
+      "loss": 0.6531,
+      "step": 17752
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4662497829447776e-06,
+      "loss": 0.6344,
+      "step": 17753
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.465231190774698e-06,
+      "loss": 0.6364,
+      "step": 17754
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.46421264051377e-06,
+      "loss": 0.6527,
+      "step": 17755
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.463194132174075e-06,
+      "loss": 0.6515,
+      "step": 17756
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4621756657676895e-06,
+      "loss": 0.637,
+      "step": 17757
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.461157241306685e-06,
+      "loss": 0.6179,
+      "step": 17758
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4601388588031375e-06,
+      "loss": 0.6542,
+      "step": 17759
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.459120518269124e-06,
+      "loss": 0.5916,
+      "step": 17760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.458102219716713e-06,
+      "loss": 0.6721,
+      "step": 17761
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.457083963157982e-06,
+      "loss": 0.6035,
+      "step": 17762
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.456065748605005e-06,
+      "loss": 0.6379,
+      "step": 17763
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.455047576069849e-06,
+      "loss": 0.6374,
+      "step": 17764
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.454029445564591e-06,
+      "loss": 0.5721,
+      "step": 17765
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.453011357101295e-06,
+      "loss": 0.6202,
+      "step": 17766
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4519933106920385e-06,
+      "loss": 0.6332,
+      "step": 17767
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4509753063488934e-06,
+      "loss": 0.6507,
+      "step": 17768
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.449957344083919e-06,
+      "loss": 0.5545,
+      "step": 17769
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.448939423909193e-06,
+      "loss": 0.6396,
+      "step": 17770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.447921545836788e-06,
+      "loss": 0.6553,
+      "step": 17771
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4469037098787576e-06,
+      "loss": 0.6258,
+      "step": 17772
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.445885916047183e-06,
+      "loss": 0.6077,
+      "step": 17773
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.44486816435413e-06,
+      "loss": 0.6488,
+      "step": 17774
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.443850454811655e-06,
+      "loss": 0.6115,
+      "step": 17775
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.442832787431835e-06,
+      "loss": 0.6814,
+      "step": 17776
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.441815162226737e-06,
+      "loss": 0.5696,
+      "step": 17777
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.440797579208413e-06,
+      "loss": 0.6594,
+      "step": 17778
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4397800383889384e-06,
+      "loss": 0.6622,
+      "step": 17779
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.438762539780381e-06,
+      "loss": 0.63,
+      "step": 17780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.437745083394791e-06,
+      "loss": 0.6695,
+      "step": 17781
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.436727669244241e-06,
+      "loss": 0.6653,
+      "step": 17782
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.435710297340798e-06,
+      "loss": 0.6671,
+      "step": 17783
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.43469296769651e-06,
+      "loss": 0.6332,
+      "step": 17784
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.43367568032345e-06,
+      "loss": 0.5894,
+      "step": 17785
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.432658435233681e-06,
+      "loss": 0.6141,
+      "step": 17786
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.431641232439252e-06,
+      "loss": 0.5973,
+      "step": 17787
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4306240719522315e-06,
+      "loss": 0.6262,
+      "step": 17788
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.429606953784684e-06,
+      "loss": 0.6737,
+      "step": 17789
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.428589877948656e-06,
+      "loss": 0.6762,
+      "step": 17790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.427572844456214e-06,
+      "loss": 0.6054,
+      "step": 17791
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.42655585331942e-06,
+      "loss": 0.562,
+      "step": 17792
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.425538904550319e-06,
+      "loss": 0.659,
+      "step": 17793
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.42452199816098e-06,
+      "loss": 0.6595,
+      "step": 17794
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.423505134163459e-06,
+      "loss": 0.6599,
+      "step": 17795
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.422488312569803e-06,
+      "loss": 0.6463,
+      "step": 17796
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.421471533392075e-06,
+      "loss": 0.5992,
+      "step": 17797
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4204547966423336e-06,
+      "loss": 0.5774,
+      "step": 17798
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4194381023326225e-06,
+      "loss": 0.6253,
+      "step": 17799
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.418421450475004e-06,
+      "loss": 0.6,
+      "step": 17800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4174048410815335e-06,
+      "loss": 0.627,
+      "step": 17801
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4163882741642555e-06,
+      "loss": 0.6128,
+      "step": 17802
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.41537174973523e-06,
+      "loss": 0.604,
+      "step": 17803
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4143552678065105e-06,
+      "loss": 0.638,
+      "step": 17804
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.413338828390139e-06,
+      "loss": 0.6172,
+      "step": 17805
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.412322431498175e-06,
+      "loss": 0.6139,
+      "step": 17806
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4113060771426715e-06,
+      "loss": 0.5956,
+      "step": 17807
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.410289765335667e-06,
+      "loss": 0.6071,
+      "step": 17808
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.409273496089221e-06,
+      "loss": 0.6889,
+      "step": 17809
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.408257269415383e-06,
+      "loss": 0.6666,
+      "step": 17810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.407241085326194e-06,
+      "loss": 0.6175,
+      "step": 17811
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.406224943833708e-06,
+      "loss": 0.6797,
+      "step": 17812
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4052088449499775e-06,
+      "loss": 0.5861,
+      "step": 17813
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.404192788687036e-06,
+      "loss": 0.6242,
+      "step": 17814
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.4031767750569385e-06,
+      "loss": 0.6271,
+      "step": 17815
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.402160804071735e-06,
+      "loss": 0.6414,
+      "step": 17816
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.401144875743461e-06,
+      "loss": 0.7015,
+      "step": 17817
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.400128990084169e-06,
+      "loss": 0.6642,
+      "step": 17818
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.399113147105907e-06,
+      "loss": 0.6282,
+      "step": 17819
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.398097346820706e-06,
+      "loss": 0.5933,
+      "step": 17820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.397081589240619e-06,
+      "loss": 0.6421,
+      "step": 17821
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.396065874377693e-06,
+      "loss": 0.601,
+      "step": 17822
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.395050202243959e-06,
+      "loss": 0.6699,
+      "step": 17823
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.394034572851466e-06,
+      "loss": 0.6109,
+      "step": 17824
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.393018986212259e-06,
+      "loss": 0.6555,
+      "step": 17825
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.392003442338369e-06,
+      "loss": 0.6266,
+      "step": 17826
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.390987941241849e-06,
+      "loss": 0.6542,
+      "step": 17827
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.389972482934728e-06,
+      "loss": 0.599,
+      "step": 17828
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.388957067429047e-06,
+      "loss": 0.6618,
+      "step": 17829
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.387941694736854e-06,
+      "loss": 0.6309,
+      "step": 17830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.386926364870178e-06,
+      "loss": 0.6269,
+      "step": 17831
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.385911077841059e-06,
+      "loss": 0.6502,
+      "step": 17832
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.384895833661542e-06,
+      "loss": 0.5487,
+      "step": 17833
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.383880632343653e-06,
+      "loss": 0.6177,
+      "step": 17834
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.382865473899434e-06,
+      "loss": 0.6528,
+      "step": 17835
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.381850358340924e-06,
+      "loss": 0.6709,
+      "step": 17836
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.380835285680151e-06,
+      "loss": 0.674,
+      "step": 17837
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.3798202559291526e-06,
+      "loss": 0.5911,
+      "step": 17838
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.378805269099971e-06,
+      "loss": 0.6359,
+      "step": 17839
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.377790325204629e-06,
+      "loss": 0.6054,
+      "step": 17840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.376775424255164e-06,
+      "loss": 0.6234,
+      "step": 17841
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.375760566263615e-06,
+      "loss": 0.6442,
+      "step": 17842
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.374745751242004e-06,
+      "loss": 0.6487,
+      "step": 17843
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.373730979202368e-06,
+      "loss": 0.5905,
+      "step": 17844
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.372716250156743e-06,
+      "loss": 0.6326,
+      "step": 17845
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.371701564117152e-06,
+      "loss": 0.6497,
+      "step": 17846
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.370686921095628e-06,
+      "loss": 0.6278,
+      "step": 17847
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.3696723211042054e-06,
+      "loss": 0.632,
+      "step": 17848
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.368657764154908e-06,
+      "loss": 0.6359,
+      "step": 17849
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.367643250259765e-06,
+      "loss": 0.6147,
+      "step": 17850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.366628779430809e-06,
+      "loss": 0.6725,
+      "step": 17851
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.365614351680063e-06,
+      "loss": 0.6396,
+      "step": 17852
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.364599967019556e-06,
+      "loss": 0.5996,
+      "step": 17853
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.363585625461317e-06,
+      "loss": 0.6297,
+      "step": 17854
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.362571327017368e-06,
+      "loss": 0.6834,
+      "step": 17855
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.361557071699739e-06,
+      "loss": 0.645,
+      "step": 17856
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.360542859520455e-06,
+      "loss": 0.6271,
+      "step": 17857
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.359528690491536e-06,
+      "loss": 0.6628,
+      "step": 17858
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.358514564625011e-06,
+      "loss": 0.6592,
+      "step": 17859
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.357500481932904e-06,
+      "loss": 0.6691,
+      "step": 17860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.356486442427235e-06,
+      "loss": 0.6546,
+      "step": 17861
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.355472446120028e-06,
+      "loss": 0.6415,
+      "step": 17862
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.354458493023307e-06,
+      "loss": 0.6094,
+      "step": 17863
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.353444583149091e-06,
+      "loss": 0.5785,
+      "step": 17864
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.352430716509401e-06,
+      "loss": 0.6565,
+      "step": 17865
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.351416893116262e-06,
+      "loss": 0.6251,
+      "step": 17866
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.350403112981689e-06,
+      "loss": 0.6573,
+      "step": 17867
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.349389376117705e-06,
+      "loss": 0.6033,
+      "step": 17868
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.348375682536328e-06,
+      "loss": 0.63,
+      "step": 17869
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.347362032249576e-06,
+      "loss": 0.6451,
+      "step": 17870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.346348425269468e-06,
+      "loss": 0.614,
+      "step": 17871
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.345334861608022e-06,
+      "loss": 0.6384,
+      "step": 17872
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.344321341277253e-06,
+      "loss": 0.6093,
+      "step": 17873
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.343307864289178e-06,
+      "loss": 0.5925,
+      "step": 17874
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.3422944306558175e-06,
+      "loss": 0.637,
+      "step": 17875
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.341281040389182e-06,
+      "loss": 0.6255,
+      "step": 17876
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.340267693501287e-06,
+      "loss": 0.7493,
+      "step": 17877
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.33925439000415e-06,
+      "loss": 0.618,
+      "step": 17878
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.338241129909781e-06,
+      "loss": 0.6267,
+      "step": 17879
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.3372279132301965e-06,
+      "loss": 0.6391,
+      "step": 17880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.336214739977409e-06,
+      "loss": 0.6535,
+      "step": 17881
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.3352016101634295e-06,
+      "loss": 0.6384,
+      "step": 17882
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.334188523800271e-06,
+      "loss": 0.625,
+      "step": 17883
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.333175480899947e-06,
+      "loss": 0.6549,
+      "step": 17884
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.332162481474463e-06,
+      "loss": 0.581,
+      "step": 17885
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.331149525535833e-06,
+      "loss": 0.6407,
+      "step": 17886
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.330136613096069e-06,
+      "loss": 0.6189,
+      "step": 17887
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.329123744167176e-06,
+      "loss": 0.674,
+      "step": 17888
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.328110918761166e-06,
+      "loss": 0.6662,
+      "step": 17889
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.327098136890045e-06,
+      "loss": 0.5915,
+      "step": 17890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.326085398565822e-06,
+      "loss": 0.5704,
+      "step": 17891
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.325072703800506e-06,
+      "loss": 0.673,
+      "step": 17892
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.324060052606098e-06,
+      "loss": 0.6333,
+      "step": 17893
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.32304744499461e-06,
+      "loss": 0.6579,
+      "step": 17894
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.322034880978047e-06,
+      "loss": 0.6574,
+      "step": 17895
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.3210223605684115e-06,
+      "loss": 0.672,
+      "step": 17896
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.32000988377771e-06,
+      "loss": 0.6264,
+      "step": 17897
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.3189974506179485e-06,
+      "loss": 0.6344,
+      "step": 17898
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.317985061101126e-06,
+      "loss": 0.6462,
+      "step": 17899
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.316972715239249e-06,
+      "loss": 0.6013,
+      "step": 17900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.315960413044322e-06,
+      "loss": 0.652,
+      "step": 17901
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.314948154528342e-06,
+      "loss": 0.6363,
+      "step": 17902
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.313935939703314e-06,
+      "loss": 0.6764,
+      "step": 17903
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.312923768581239e-06,
+      "loss": 0.6359,
+      "step": 17904
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.311911641174115e-06,
+      "loss": 0.6737,
+      "step": 17905
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.3108995574939456e-06,
+      "loss": 0.6028,
+      "step": 17906
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.30988751755273e-06,
+      "loss": 0.6362,
+      "step": 17907
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.308875521362463e-06,
+      "loss": 0.6595,
+      "step": 17908
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.307863568935147e-06,
+      "loss": 0.6665,
+      "step": 17909
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.30685166028278e-06,
+      "loss": 0.6124,
+      "step": 17910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.305839795417358e-06,
+      "loss": 0.6746,
+      "step": 17911
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.304827974350877e-06,
+      "loss": 0.6305,
+      "step": 17912
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.303816197095337e-06,
+      "loss": 0.6634,
+      "step": 17913
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.302804463662729e-06,
+      "loss": 0.6668,
+      "step": 17914
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.301792774065052e-06,
+      "loss": 0.6454,
+      "step": 17915
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.300781128314302e-06,
+      "loss": 0.6378,
+      "step": 17916
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.29976952642247e-06,
+      "loss": 0.6705,
+      "step": 17917
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.29875796840155e-06,
+      "loss": 0.6186,
+      "step": 17918
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.297746454263538e-06,
+      "loss": 0.6445,
+      "step": 17919
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.296734984020423e-06,
+      "loss": 0.5933,
+      "step": 17920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.295723557684201e-06,
+      "loss": 0.6272,
+      "step": 17921
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.294712175266863e-06,
+      "loss": 0.6214,
+      "step": 17922
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.293700836780397e-06,
+      "loss": 0.7059,
+      "step": 17923
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.292689542236796e-06,
+      "loss": 0.6113,
+      "step": 17924
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2916782916480534e-06,
+      "loss": 0.6026,
+      "step": 17925
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.290667085026154e-06,
+      "loss": 0.6733,
+      "step": 17926
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.289655922383088e-06,
+      "loss": 0.6509,
+      "step": 17927
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.288644803730846e-06,
+      "loss": 0.6586,
+      "step": 17928
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.287633729081413e-06,
+      "loss": 0.6652,
+      "step": 17929
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.286622698446779e-06,
+      "loss": 0.6557,
+      "step": 17930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.285611711838933e-06,
+      "loss": 0.6372,
+      "step": 17931
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.284600769269855e-06,
+      "loss": 0.59,
+      "step": 17932
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.283589870751535e-06,
+      "loss": 0.6253,
+      "step": 17933
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2825790162959625e-06,
+      "loss": 0.5909,
+      "step": 17934
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.281568205915114e-06,
+      "loss": 0.5736,
+      "step": 17935
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.280557439620979e-06,
+      "loss": 0.6336,
+      "step": 17936
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.279546717425542e-06,
+      "loss": 0.529,
+      "step": 17937
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.278536039340784e-06,
+      "loss": 0.6202,
+      "step": 17938
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.277525405378689e-06,
+      "loss": 0.6133,
+      "step": 17939
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2765148155512414e-06,
+      "loss": 0.6054,
+      "step": 17940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.275504269870418e-06,
+      "loss": 0.6264,
+      "step": 17941
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.274493768348204e-06,
+      "loss": 0.6574,
+      "step": 17942
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.273483310996581e-06,
+      "loss": 0.6116,
+      "step": 17943
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.272472897827524e-06,
+      "loss": 0.577,
+      "step": 17944
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2714625288530185e-06,
+      "loss": 0.624,
+      "step": 17945
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.270452204085042e-06,
+      "loss": 0.6414,
+      "step": 17946
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.269441923535571e-06,
+      "loss": 0.615,
+      "step": 17947
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.268431687216584e-06,
+      "loss": 0.6544,
+      "step": 17948
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.267421495140063e-06,
+      "loss": 0.6038,
+      "step": 17949
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.26641134731798e-06,
+      "loss": 0.6387,
+      "step": 17950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.265401243762315e-06,
+      "loss": 0.6532,
+      "step": 17951
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.264391184485039e-06,
+      "loss": 0.687,
+      "step": 17952
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.263381169498133e-06,
+      "loss": 0.5828,
+      "step": 17953
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.262371198813573e-06,
+      "loss": 0.6184,
+      "step": 17954
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.261361272443327e-06,
+      "loss": 0.6643,
+      "step": 17955
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.260351390399373e-06,
+      "loss": 0.6505,
+      "step": 17956
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.259341552693685e-06,
+      "loss": 0.6111,
+      "step": 17957
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.258331759338234e-06,
+      "loss": 0.6412,
+      "step": 17958
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2573220103449925e-06,
+      "loss": 0.6912,
+      "step": 17959
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.256312305725936e-06,
+      "loss": 0.6243,
+      "step": 17960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.25530264549303e-06,
+      "loss": 0.6231,
+      "step": 17961
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.254293029658248e-06,
+      "loss": 0.5785,
+      "step": 17962
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.253283458233563e-06,
+      "loss": 0.6496,
+      "step": 17963
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.252273931230941e-06,
+      "loss": 0.6622,
+      "step": 17964
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2512644486623505e-06,
+      "loss": 0.6321,
+      "step": 17965
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.250255010539766e-06,
+      "loss": 0.6664,
+      "step": 17966
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.249245616875149e-06,
+      "loss": 0.6455,
+      "step": 17967
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.24823626768047e-06,
+      "loss": 0.5916,
+      "step": 17968
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.247226962967698e-06,
+      "loss": 0.6227,
+      "step": 17969
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.246217702748795e-06,
+      "loss": 0.6125,
+      "step": 17970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.245208487035732e-06,
+      "loss": 0.6613,
+      "step": 17971
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.244199315840472e-06,
+      "loss": 0.6407,
+      "step": 17972
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.243190189174979e-06,
+      "loss": 0.577,
+      "step": 17973
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2421811070512196e-06,
+      "loss": 0.6448,
+      "step": 17974
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.241172069481156e-06,
+      "loss": 0.5768,
+      "step": 17975
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.240163076476753e-06,
+      "loss": 0.6799,
+      "step": 17976
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.239154128049972e-06,
+      "loss": 0.6763,
+      "step": 17977
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.238145224212779e-06,
+      "loss": 0.6446,
+      "step": 17978
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.23713636497713e-06,
+      "loss": 0.636,
+      "step": 17979
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.236127550354991e-06,
+      "loss": 0.6582,
+      "step": 17980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.235118780358323e-06,
+      "loss": 0.6533,
+      "step": 17981
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.234110054999081e-06,
+      "loss": 0.6372,
+      "step": 17982
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2331013742892285e-06,
+      "loss": 0.6689,
+      "step": 17983
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.232092738240727e-06,
+      "loss": 0.6555,
+      "step": 17984
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.231084146865531e-06,
+      "loss": 0.6305,
+      "step": 17985
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.230075600175599e-06,
+      "loss": 0.5996,
+      "step": 17986
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.229067098182892e-06,
+      "loss": 0.6622,
+      "step": 17987
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.228058640899364e-06,
+      "loss": 0.6141,
+      "step": 17988
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.22705022833697e-06,
+      "loss": 0.6645,
+      "step": 17989
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.226041860507673e-06,
+      "loss": 0.6248,
+      "step": 17990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.22503353742342e-06,
+      "loss": 0.6586,
+      "step": 17991
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.22402525909617e-06,
+      "loss": 0.6733,
+      "step": 17992
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.223017025537879e-06,
+      "loss": 0.656,
+      "step": 17993
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.222008836760498e-06,
+      "loss": 0.637,
+      "step": 17994
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.22100069277598e-06,
+      "loss": 0.6647,
+      "step": 17995
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2199925935962835e-06,
+      "loss": 0.6516,
+      "step": 17996
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2189845392333525e-06,
+      "loss": 0.6815,
+      "step": 17997
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.217976529699144e-06,
+      "loss": 0.5854,
+      "step": 17998
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.21696856500561e-06,
+      "loss": 0.6183,
+      "step": 17999
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.215960645164696e-06,
+      "loss": 0.603,
+      "step": 18000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.214952770188356e-06,
+      "loss": 0.6306,
+      "step": 18001
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.213944940088542e-06,
+      "loss": 0.5969,
+      "step": 18002
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.212937154877197e-06,
+      "loss": 0.6443,
+      "step": 18003
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.211929414566274e-06,
+      "loss": 0.6358,
+      "step": 18004
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.21092171916772e-06,
+      "loss": 0.7036,
+      "step": 18005
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.209914068693481e-06,
+      "loss": 0.6406,
+      "step": 18006
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.208906463155505e-06,
+      "loss": 0.6385,
+      "step": 18007
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.207898902565741e-06,
+      "loss": 0.5996,
+      "step": 18008
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.20689138693613e-06,
+      "loss": 0.6447,
+      "step": 18009
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.205883916278621e-06,
+      "loss": 0.5805,
+      "step": 18010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.204876490605158e-06,
+      "loss": 0.6088,
+      "step": 18011
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2038691099276845e-06,
+      "loss": 0.6437,
+      "step": 18012
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.2028617742581485e-06,
+      "loss": 0.6246,
+      "step": 18013
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.201854483608484e-06,
+      "loss": 0.6326,
+      "step": 18014
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.200847237990641e-06,
+      "loss": 0.6693,
+      "step": 18015
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.199840037416564e-06,
+      "loss": 0.6339,
+      "step": 18016
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.1988328818981836e-06,
+      "loss": 0.6834,
+      "step": 18017
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.19782577144745e-06,
+      "loss": 0.6489,
+      "step": 18018
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.196818706076306e-06,
+      "loss": 0.6384,
+      "step": 18019
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.195811685796682e-06,
+      "loss": 0.6686,
+      "step": 18020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.194804710620524e-06,
+      "loss": 0.647,
+      "step": 18021
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.193797780559775e-06,
+      "loss": 0.6296,
+      "step": 18022
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.192790895626361e-06,
+      "loss": 0.6705,
+      "step": 18023
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.1917840558322285e-06,
+      "loss": 0.6598,
+      "step": 18024
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.190777261189319e-06,
+      "loss": 0.6242,
+      "step": 18025
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.189770511709556e-06,
+      "loss": 0.613,
+      "step": 18026
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.188763807404887e-06,
+      "loss": 0.6254,
+      "step": 18027
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.187757148287248e-06,
+      "loss": 0.6813,
+      "step": 18028
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.1867505343685645e-06,
+      "loss": 0.6395,
+      "step": 18029
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.18574396566078e-06,
+      "loss": 0.6311,
+      "step": 18030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.1847374421758274e-06,
+      "loss": 0.6158,
+      "step": 18031
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.183730963925636e-06,
+      "loss": 0.6245,
+      "step": 18032
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.182724530922142e-06,
+      "loss": 0.6044,
+      "step": 18033
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.181718143177281e-06,
+      "loss": 0.6772,
+      "step": 18034
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.180711800702977e-06,
+      "loss": 0.6521,
+      "step": 18035
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.179705503511166e-06,
+      "loss": 0.6794,
+      "step": 18036
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.178699251613784e-06,
+      "loss": 0.614,
+      "step": 18037
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.17769304502275e-06,
+      "loss": 0.6678,
+      "step": 18038
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.176686883750002e-06,
+      "loss": 0.6217,
+      "step": 18039
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1756807678074725e-06,
+      "loss": 0.6745,
+      "step": 18040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.17467469720708e-06,
+      "loss": 0.599,
+      "step": 18041
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1736686719607575e-06,
+      "loss": 0.6503,
+      "step": 18042
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.17266269208044e-06,
+      "loss": 0.6165,
+      "step": 18043
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.171656757578042e-06,
+      "loss": 0.6387,
+      "step": 18044
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.170650868465494e-06,
+      "loss": 0.6063,
+      "step": 18045
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.169645024754731e-06,
+      "loss": 0.6307,
+      "step": 18046
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.168639226457667e-06,
+      "loss": 0.6102,
+      "step": 18047
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.16763347358623e-06,
+      "loss": 0.6656,
+      "step": 18048
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.166627766152352e-06,
+      "loss": 0.6442,
+      "step": 18049
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.165622104167948e-06,
+      "loss": 0.5778,
+      "step": 18050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.164616487644941e-06,
+      "loss": 0.6218,
+      "step": 18051
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.163610916595264e-06,
+      "loss": 0.6363,
+      "step": 18052
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.162605391030829e-06,
+      "loss": 0.6506,
+      "step": 18053
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.161599910963558e-06,
+      "loss": 0.6188,
+      "step": 18054
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.160594476405384e-06,
+      "loss": 0.6676,
+      "step": 18055
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.159589087368214e-06,
+      "loss": 0.68,
+      "step": 18056
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.158583743863971e-06,
+      "loss": 0.6255,
+      "step": 18057
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.157578445904584e-06,
+      "loss": 0.6454,
+      "step": 18058
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.15657319350196e-06,
+      "loss": 0.6324,
+      "step": 18059
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.155567986668023e-06,
+      "loss": 0.5761,
+      "step": 18060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.154562825414695e-06,
+      "loss": 0.6104,
+      "step": 18061
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.153557709753886e-06,
+      "loss": 0.5507,
+      "step": 18062
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.152552639697513e-06,
+      "loss": 0.6177,
+      "step": 18063
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.151547615257503e-06,
+      "loss": 0.637,
+      "step": 18064
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.150542636445759e-06,
+      "loss": 0.6696,
+      "step": 18065
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1495377032742e-06,
+      "loss": 0.6415,
+      "step": 18066
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1485328157547485e-06,
+      "loss": 0.5698,
+      "step": 18067
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.147527973899309e-06,
+      "loss": 0.5807,
+      "step": 18068
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.146523177719797e-06,
+      "loss": 0.652,
+      "step": 18069
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.145518427228132e-06,
+      "loss": 0.6685,
+      "step": 18070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.144513722436218e-06,
+      "loss": 0.609,
+      "step": 18071
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.14350906335597e-06,
+      "loss": 0.6277,
+      "step": 18072
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.142504449999307e-06,
+      "loss": 0.6466,
+      "step": 18073
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.141499882378129e-06,
+      "loss": 0.5837,
+      "step": 18074
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.140495360504353e-06,
+      "loss": 0.6675,
+      "step": 18075
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.139490884389885e-06,
+      "loss": 0.6592,
+      "step": 18076
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.138486454046635e-06,
+      "loss": 0.6735,
+      "step": 18077
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1374820694865156e-06,
+      "loss": 0.6221,
+      "step": 18078
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.136477730721431e-06,
+      "loss": 0.6259,
+      "step": 18079
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.135473437763291e-06,
+      "loss": 0.6189,
+      "step": 18080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.134469190624002e-06,
+      "loss": 0.6539,
+      "step": 18081
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1334649893154696e-06,
+      "loss": 0.6294,
+      "step": 18082
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1324608338496e-06,
+      "loss": 0.6591,
+      "step": 18083
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.131456724238302e-06,
+      "loss": 0.606,
+      "step": 18084
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.130452660493478e-06,
+      "loss": 0.6044,
+      "step": 18085
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1294486426270316e-06,
+      "loss": 0.6143,
+      "step": 18086
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1284446706508695e-06,
+      "loss": 0.6528,
+      "step": 18087
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.127440744576892e-06,
+      "loss": 0.6051,
+      "step": 18088
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.126436864417002e-06,
+      "loss": 0.5845,
+      "step": 18089
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.125433030183107e-06,
+      "loss": 0.6575,
+      "step": 18090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.124429241887101e-06,
+      "loss": 0.5894,
+      "step": 18091
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.123425499540889e-06,
+      "loss": 0.6479,
+      "step": 18092
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1224218031563754e-06,
+      "loss": 0.6402,
+      "step": 18093
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.121418152745453e-06,
+      "loss": 0.6002,
+      "step": 18094
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.120414548320025e-06,
+      "loss": 0.6433,
+      "step": 18095
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.119410989891992e-06,
+      "loss": 0.6165,
+      "step": 18096
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.11840747747325e-06,
+      "loss": 0.5903,
+      "step": 18097
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.117404011075696e-06,
+      "loss": 0.6395,
+      "step": 18098
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.116400590711233e-06,
+      "loss": 0.6388,
+      "step": 18099
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.115397216391749e-06,
+      "loss": 0.6669,
+      "step": 18100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.114393888129146e-06,
+      "loss": 0.6123,
+      "step": 18101
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.113390605935322e-06,
+      "loss": 0.6921,
+      "step": 18102
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.112387369822166e-06,
+      "loss": 0.6836,
+      "step": 18103
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.111384179801576e-06,
+      "loss": 0.5878,
+      "step": 18104
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1103810358854475e-06,
+      "loss": 0.5786,
+      "step": 18105
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.10937793808567e-06,
+      "loss": 0.6354,
+      "step": 18106
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.108374886414139e-06,
+      "loss": 0.6053,
+      "step": 18107
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.107371880882749e-06,
+      "loss": 0.6448,
+      "step": 18108
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.106368921503387e-06,
+      "loss": 0.6614,
+      "step": 18109
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.1053660082879475e-06,
+      "loss": 0.6245,
+      "step": 18110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.104363141248322e-06,
+      "loss": 0.6614,
+      "step": 18111
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.103360320396396e-06,
+      "loss": 0.6434,
+      "step": 18112
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.102357545744066e-06,
+      "loss": 0.6441,
+      "step": 18113
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.101354817303218e-06,
+      "loss": 0.6419,
+      "step": 18114
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.100352135085739e-06,
+      "loss": 0.609,
+      "step": 18115
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.099349499103517e-06,
+      "loss": 0.614,
+      "step": 18116
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0983469093684444e-06,
+      "loss": 0.6056,
+      "step": 18117
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0973443658924025e-06,
+      "loss": 0.6263,
+      "step": 18118
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.096341868687281e-06,
+      "loss": 0.6534,
+      "step": 18119
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.095339417764965e-06,
+      "loss": 0.6897,
+      "step": 18120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.094337013137339e-06,
+      "loss": 0.651,
+      "step": 18121
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.093334654816288e-06,
+      "loss": 0.6012,
+      "step": 18122
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.092332342813698e-06,
+      "loss": 0.6359,
+      "step": 18123
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0913300771414504e-06,
+      "loss": 0.6822,
+      "step": 18124
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.090327857811429e-06,
+      "loss": 0.6156,
+      "step": 18125
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.089325684835517e-06,
+      "loss": 0.6342,
+      "step": 18126
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.088323558225596e-06,
+      "loss": 0.6487,
+      "step": 18127
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.087321477993546e-06,
+      "loss": 0.6425,
+      "step": 18128
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.086319444151253e-06,
+      "loss": 0.6104,
+      "step": 18129
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.085317456710591e-06,
+      "loss": 0.6176,
+      "step": 18130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.084315515683443e-06,
+      "loss": 0.6861,
+      "step": 18131
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.08331362108169e-06,
+      "loss": 0.627,
+      "step": 18132
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.082311772917206e-06,
+      "loss": 0.6312,
+      "step": 18133
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.081309971201872e-06,
+      "loss": 0.635,
+      "step": 18134
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.080308215947568e-06,
+      "loss": 0.6295,
+      "step": 18135
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.079306507166166e-06,
+      "loss": 0.6589,
+      "step": 18136
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.078304844869547e-06,
+      "loss": 0.61,
+      "step": 18137
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.077303229069585e-06,
+      "loss": 0.6811,
+      "step": 18138
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.076301659778152e-06,
+      "loss": 0.6076,
+      "step": 18139
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.07530013700713e-06,
+      "loss": 0.6468,
+      "step": 18140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.074298660768387e-06,
+      "loss": 0.6246,
+      "step": 18141
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.073297231073802e-06,
+      "loss": 0.6055,
+      "step": 18142
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.072295847935246e-06,
+      "loss": 0.6478,
+      "step": 18143
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.071294511364588e-06,
+      "loss": 0.6629,
+      "step": 18144
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.070293221373704e-06,
+      "loss": 0.6556,
+      "step": 18145
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.069291977974468e-06,
+      "loss": 0.6003,
+      "step": 18146
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0682907811787455e-06,
+      "loss": 0.6237,
+      "step": 18147
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.06728963099841e-06,
+      "loss": 0.6454,
+      "step": 18148
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.066288527445332e-06,
+      "loss": 0.5898,
+      "step": 18149
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.065287470531377e-06,
+      "loss": 0.6638,
+      "step": 18150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.064286460268418e-06,
+      "loss": 0.6608,
+      "step": 18151
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.063285496668323e-06,
+      "loss": 0.6669,
+      "step": 18152
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.062284579742957e-06,
+      "loss": 0.655,
+      "step": 18153
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.061283709504186e-06,
+      "loss": 0.5978,
+      "step": 18154
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.060282885963884e-06,
+      "loss": 0.6726,
+      "step": 18155
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.059282109133908e-06,
+      "loss": 0.5884,
+      "step": 18156
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.05828137902613e-06,
+      "loss": 0.6042,
+      "step": 18157
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.057280695652413e-06,
+      "loss": 0.5931,
+      "step": 18158
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0562800590246196e-06,
+      "loss": 0.629,
+      "step": 18159
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.055279469154614e-06,
+      "loss": 0.6273,
+      "step": 18160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.054278926054264e-06,
+      "loss": 0.6688,
+      "step": 18161
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.053278429735425e-06,
+      "loss": 0.6449,
+      "step": 18162
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.052277980209964e-06,
+      "loss": 0.5741,
+      "step": 18163
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.051277577489743e-06,
+      "loss": 0.6348,
+      "step": 18164
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.05027722158662e-06,
+      "loss": 0.6258,
+      "step": 18165
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0492769125124566e-06,
+      "loss": 0.6693,
+      "step": 18166
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.048276650279114e-06,
+      "loss": 0.5892,
+      "step": 18167
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.04727643489845e-06,
+      "loss": 0.6178,
+      "step": 18168
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.046276266382324e-06,
+      "loss": 0.6226,
+      "step": 18169
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.045276144742596e-06,
+      "loss": 0.6292,
+      "step": 18170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.044276069991119e-06,
+      "loss": 0.6258,
+      "step": 18171
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0432760421397566e-06,
+      "loss": 0.6924,
+      "step": 18172
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.042276061200361e-06,
+      "loss": 0.6175,
+      "step": 18173
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.041276127184787e-06,
+      "loss": 0.6054,
+      "step": 18174
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.040276240104894e-06,
+      "loss": 0.6571,
+      "step": 18175
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0392763999725365e-06,
+      "loss": 0.6526,
+      "step": 18176
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0382766067995656e-06,
+      "loss": 0.6498,
+      "step": 18177
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.037276860597837e-06,
+      "loss": 0.6756,
+      "step": 18178
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.036277161379207e-06,
+      "loss": 0.6287,
+      "step": 18179
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.035277509155521e-06,
+      "loss": 0.6342,
+      "step": 18180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.034277903938639e-06,
+      "loss": 0.5902,
+      "step": 18181
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0332783457404085e-06,
+      "loss": 0.6792,
+      "step": 18182
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.03227883457268e-06,
+      "loss": 0.6685,
+      "step": 18183
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.031279370447305e-06,
+      "loss": 0.5822,
+      "step": 18184
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.030279953376136e-06,
+      "loss": 0.6476,
+      "step": 18185
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.029280583371018e-06,
+      "loss": 0.6241,
+      "step": 18186
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0282812604438e-06,
+      "loss": 0.6184,
+      "step": 18187
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0272819846063356e-06,
+      "loss": 0.6224,
+      "step": 18188
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0262827558704666e-06,
+      "loss": 0.6439,
+      "step": 18189
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.025283574248042e-06,
+      "loss": 0.6301,
+      "step": 18190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.02428443975091e-06,
+      "loss": 0.6372,
+      "step": 18191
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.023285352390914e-06,
+      "loss": 0.6242,
+      "step": 18192
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0222863121799016e-06,
+      "loss": 0.6297,
+      "step": 18193
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.021287319129717e-06,
+      "loss": 0.6423,
+      "step": 18194
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.020288373252203e-06,
+      "loss": 0.5896,
+      "step": 18195
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.019289474559204e-06,
+      "loss": 0.7093,
+      "step": 18196
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.018290623062568e-06,
+      "loss": 0.6622,
+      "step": 18197
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.017291818774129e-06,
+      "loss": 0.6436,
+      "step": 18198
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.016293061705738e-06,
+      "loss": 0.6276,
+      "step": 18199
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.01529435186923e-06,
+      "loss": 0.5899,
+      "step": 18200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.014295689276446e-06,
+      "loss": 0.6606,
+      "step": 18201
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.013297073939233e-06,
+      "loss": 0.6471,
+      "step": 18202
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0122985058694226e-06,
+      "loss": 0.6415,
+      "step": 18203
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.011299985078858e-06,
+      "loss": 0.6312,
+      "step": 18204
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.010301511579381e-06,
+      "loss": 0.6584,
+      "step": 18205
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.009303085382823e-06,
+      "loss": 0.6691,
+      "step": 18206
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.008304706501027e-06,
+      "loss": 0.6144,
+      "step": 18207
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.007306374945828e-06,
+      "loss": 0.6536,
+      "step": 18208
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.006308090729062e-06,
+      "loss": 0.6308,
+      "step": 18209
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.005309853862567e-06,
+      "loss": 0.6185,
+      "step": 18210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0043116643581766e-06,
+      "loss": 0.6129,
+      "step": 18211
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.003313522227725e-06,
+      "loss": 0.6309,
+      "step": 18212
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.002315427483048e-06,
+      "loss": 0.6286,
+      "step": 18213
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.0013173801359805e-06,
+      "loss": 0.6094,
+      "step": 18214
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.000319380198351e-06,
+      "loss": 0.6329,
+      "step": 18215
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.999321427681995e-06,
+      "loss": 0.6219,
+      "step": 18216
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.998323522598746e-06,
+      "loss": 0.5722,
+      "step": 18217
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.997325664960433e-06,
+      "loss": 0.668,
+      "step": 18218
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9963278547788876e-06,
+      "loss": 0.6038,
+      "step": 18219
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.995330092065941e-06,
+      "loss": 0.5847,
+      "step": 18220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.99433237683342e-06,
+      "loss": 0.6037,
+      "step": 18221
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.993334709093157e-06,
+      "loss": 0.6617,
+      "step": 18222
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9923370888569806e-06,
+      "loss": 0.6272,
+      "step": 18223
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.991339516136717e-06,
+      "loss": 0.6504,
+      "step": 18224
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.990341990944193e-06,
+      "loss": 0.6622,
+      "step": 18225
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.98934451329124e-06,
+      "loss": 0.6406,
+      "step": 18226
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.988347083189678e-06,
+      "loss": 0.6013,
+      "step": 18227
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.987349700651338e-06,
+      "loss": 0.6548,
+      "step": 18228
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.986352365688044e-06,
+      "loss": 0.6912,
+      "step": 18229
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9853550783116186e-06,
+      "loss": 0.64,
+      "step": 18230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.984357838533888e-06,
+      "loss": 0.6673,
+      "step": 18231
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9833606463666746e-06,
+      "loss": 0.6724,
+      "step": 18232
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.982363501821802e-06,
+      "loss": 0.5916,
+      "step": 18233
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.981366404911091e-06,
+      "loss": 0.6193,
+      "step": 18234
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.980369355646367e-06,
+      "loss": 0.6355,
+      "step": 18235
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.979372354039449e-06,
+      "loss": 0.6055,
+      "step": 18236
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.978375400102156e-06,
+      "loss": 0.6479,
+      "step": 18237
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.977378493846315e-06,
+      "loss": 0.5804,
+      "step": 18238
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9763816352837345e-06,
+      "loss": 0.6029,
+      "step": 18239
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.97538482442624e-06,
+      "loss": 0.6583,
+      "step": 18240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.974388061285656e-06,
+      "loss": 0.6467,
+      "step": 18241
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.973391345873787e-06,
+      "loss": 0.618,
+      "step": 18242
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.972394678202458e-06,
+      "loss": 0.663,
+      "step": 18243
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9713980582834906e-06,
+      "loss": 0.6129,
+      "step": 18244
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.97040148612869e-06,
+      "loss": 0.6375,
+      "step": 18245
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.969404961749877e-06,
+      "loss": 0.6063,
+      "step": 18246
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.968408485158872e-06,
+      "loss": 0.6026,
+      "step": 18247
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.967412056367479e-06,
+      "loss": 0.6062,
+      "step": 18248
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.966415675387518e-06,
+      "loss": 0.6742,
+      "step": 18249
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.965419342230807e-06,
+      "loss": 0.7021,
+      "step": 18250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.964423056909147e-06,
+      "loss": 0.6768,
+      "step": 18251
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.963426819434359e-06,
+      "loss": 0.6258,
+      "step": 18252
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.962430629818256e-06,
+      "loss": 0.6729,
+      "step": 18253
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.961434488072639e-06,
+      "loss": 0.591,
+      "step": 18254
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.960438394209327e-06,
+      "loss": 0.6835,
+      "step": 18255
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.959442348240132e-06,
+      "loss": 0.654,
+      "step": 18256
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.958446350176854e-06,
+      "loss": 0.6501,
+      "step": 18257
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9574504000313095e-06,
+      "loss": 0.6574,
+      "step": 18258
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.956454497815307e-06,
+      "loss": 0.6122,
+      "step": 18259
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.955458643540646e-06,
+      "loss": 0.679,
+      "step": 18260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.954462837219146e-06,
+      "loss": 0.589,
+      "step": 18261
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9534670788626015e-06,
+      "loss": 0.5774,
+      "step": 18262
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.952471368482822e-06,
+      "loss": 0.6061,
+      "step": 18263
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9514757060916205e-06,
+      "loss": 0.6288,
+      "step": 18264
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.950480091700792e-06,
+      "loss": 0.6403,
+      "step": 18265
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.949484525322144e-06,
+      "loss": 0.5892,
+      "step": 18266
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.948489006967485e-06,
+      "loss": 0.6486,
+      "step": 18267
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.947493536648611e-06,
+      "loss": 0.6122,
+      "step": 18268
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.946498114377325e-06,
+      "loss": 0.6412,
+      "step": 18269
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9455027401654385e-06,
+      "loss": 0.5987,
+      "step": 18270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9445074140247406e-06,
+      "loss": 0.6529,
+      "step": 18271
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.943512135967036e-06,
+      "loss": 0.6732,
+      "step": 18272
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.942516906004133e-06,
+      "loss": 0.6601,
+      "step": 18273
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.941521724147821e-06,
+      "loss": 0.6049,
+      "step": 18274
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.940526590409901e-06,
+      "loss": 0.6322,
+      "step": 18275
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.939531504802179e-06,
+      "loss": 0.5867,
+      "step": 18276
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.938536467336443e-06,
+      "loss": 0.6333,
+      "step": 18277
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.937541478024493e-06,
+      "loss": 0.6306,
+      "step": 18278
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.936546536878135e-06,
+      "loss": 0.6743,
+      "step": 18279
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.935551643909153e-06,
+      "loss": 0.6725,
+      "step": 18280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9345567991293474e-06,
+      "loss": 0.6483,
+      "step": 18281
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.933562002550518e-06,
+      "loss": 0.6324,
+      "step": 18282
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.932567254184451e-06,
+      "loss": 0.6156,
+      "step": 18283
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.931572554042944e-06,
+      "loss": 0.6523,
+      "step": 18284
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.930577902137796e-06,
+      "loss": 0.6354,
+      "step": 18285
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.929583298480791e-06,
+      "loss": 0.687,
+      "step": 18286
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.928588743083723e-06,
+      "loss": 0.6783,
+      "step": 18287
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9275942359583925e-06,
+      "loss": 0.6054,
+      "step": 18288
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.926599777116579e-06,
+      "loss": 0.6442,
+      "step": 18289
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.925605366570078e-06,
+      "loss": 0.6184,
+      "step": 18290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.924611004330682e-06,
+      "loss": 0.5859,
+      "step": 18291
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.923616690410176e-06,
+      "loss": 0.5711,
+      "step": 18292
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9226224248203504e-06,
+      "loss": 0.66,
+      "step": 18293
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.921628207572998e-06,
+      "loss": 0.6717,
+      "step": 18294
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.920634038679898e-06,
+      "loss": 0.6488,
+      "step": 18295
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.91963991815284e-06,
+      "loss": 0.6368,
+      "step": 18296
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9186458460036186e-06,
+      "loss": 0.6677,
+      "step": 18297
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.917651822244011e-06,
+      "loss": 0.6027,
+      "step": 18298
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.916657846885802e-06,
+      "loss": 0.6566,
+      "step": 18299
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.915663919940786e-06,
+      "loss": 0.6642,
+      "step": 18300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.914670041420736e-06,
+      "loss": 0.658,
+      "step": 18301
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.91367621133744e-06,
+      "loss": 0.6295,
+      "step": 18302
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.912682429702687e-06,
+      "loss": 0.5991,
+      "step": 18303
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9116886965282506e-06,
+      "loss": 0.5927,
+      "step": 18304
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.910695011825913e-06,
+      "loss": 0.6295,
+      "step": 18305
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.909701375607467e-06,
+      "loss": 0.612,
+      "step": 18306
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.908707787884681e-06,
+      "loss": 0.6506,
+      "step": 18307
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9077142486693375e-06,
+      "loss": 0.6134,
+      "step": 18308
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.906720757973224e-06,
+      "loss": 0.6715,
+      "step": 18309
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.905727315808111e-06,
+      "loss": 0.658,
+      "step": 18310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.904733922185777e-06,
+      "loss": 0.651,
+      "step": 18311
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.903740577118009e-06,
+      "loss": 0.6209,
+      "step": 18312
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.902747280616575e-06,
+      "loss": 0.7044,
+      "step": 18313
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.901754032693252e-06,
+      "loss": 0.6521,
+      "step": 18314
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.9007608333598255e-06,
+      "loss": 0.5915,
+      "step": 18315
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.8997676826280615e-06,
+      "loss": 0.6724,
+      "step": 18316
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.898774580509735e-06,
+      "loss": 0.6321,
+      "step": 18317
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.897781527016631e-06,
+      "loss": 0.6337,
+      "step": 18318
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.896788522160512e-06,
+      "loss": 0.6,
+      "step": 18319
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.895795565953154e-06,
+      "loss": 0.6386,
+      "step": 18320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.894802658406338e-06,
+      "loss": 0.6549,
+      "step": 18321
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.893809799531825e-06,
+      "loss": 0.5791,
+      "step": 18322
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.892816989341393e-06,
+      "loss": 0.6455,
+      "step": 18323
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.8918242278468085e-06,
+      "loss": 0.6658,
+      "step": 18324
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.890831515059845e-06,
+      "loss": 0.634,
+      "step": 18325
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.889838850992275e-06,
+      "loss": 0.6009,
+      "step": 18326
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 5.888846235655863e-06,
+      "loss": 0.71,
+      "step": 18327
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.887853669062378e-06,
+      "loss": 0.5983,
+      "step": 18328
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.886861151223593e-06,
+      "loss": 0.625,
+      "step": 18329
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.885868682151269e-06,
+      "loss": 0.6372,
+      "step": 18330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8848762618571765e-06,
+      "loss": 0.702,
+      "step": 18331
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8838838903530835e-06,
+      "loss": 0.6527,
+      "step": 18332
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.882891567650752e-06,
+      "loss": 0.6041,
+      "step": 18333
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8818992937619504e-06,
+      "loss": 0.6749,
+      "step": 18334
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.880907068698443e-06,
+      "loss": 0.6714,
+      "step": 18335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.879914892471991e-06,
+      "loss": 0.6645,
+      "step": 18336
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8789227650943595e-06,
+      "loss": 0.6411,
+      "step": 18337
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.877930686577315e-06,
+      "loss": 0.6347,
+      "step": 18338
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.876938656932614e-06,
+      "loss": 0.6272,
+      "step": 18339
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.87594667617202e-06,
+      "loss": 0.6508,
+      "step": 18340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8749547443072974e-06,
+      "loss": 0.6343,
+      "step": 18341
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.873962861350204e-06,
+      "loss": 0.6393,
+      "step": 18342
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8729710273125e-06,
+      "loss": 0.6578,
+      "step": 18343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.871979242205945e-06,
+      "loss": 0.6485,
+      "step": 18344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.870987506042299e-06,
+      "loss": 0.6127,
+      "step": 18345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.869995818833317e-06,
+      "loss": 0.5931,
+      "step": 18346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.869004180590763e-06,
+      "loss": 0.6519,
+      "step": 18347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.868012591326386e-06,
+      "loss": 0.649,
+      "step": 18348
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.867021051051949e-06,
+      "loss": 0.5766,
+      "step": 18349
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.866029559779206e-06,
+      "loss": 0.6484,
+      "step": 18350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.86503811751991e-06,
+      "loss": 0.6725,
+      "step": 18351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.864046724285819e-06,
+      "loss": 0.5812,
+      "step": 18352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.863055380088687e-06,
+      "loss": 0.6391,
+      "step": 18353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.862064084940267e-06,
+      "loss": 0.643,
+      "step": 18354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.861072838852309e-06,
+      "loss": 0.5896,
+      "step": 18355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.860081641836571e-06,
+      "loss": 0.6033,
+      "step": 18356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.859090493904801e-06,
+      "loss": 0.6143,
+      "step": 18357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8580993950687514e-06,
+      "loss": 0.6106,
+      "step": 18358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.857108345340175e-06,
+      "loss": 0.6879,
+      "step": 18359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8561173447308174e-06,
+      "loss": 0.633,
+      "step": 18360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.855126393252431e-06,
+      "loss": 0.591,
+      "step": 18361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.854135490916766e-06,
+      "loss": 0.664,
+      "step": 18362
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8531446377355684e-06,
+      "loss": 0.6566,
+      "step": 18363
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.852153833720586e-06,
+      "loss": 0.6321,
+      "step": 18364
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.851163078883568e-06,
+      "loss": 0.6353,
+      "step": 18365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.85017237323626e-06,
+      "loss": 0.6291,
+      "step": 18366
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.849181716790406e-06,
+      "loss": 0.6562,
+      "step": 18367
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8481911095577585e-06,
+      "loss": 0.7064,
+      "step": 18368
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.847200551550054e-06,
+      "loss": 0.5965,
+      "step": 18369
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.846210042779039e-06,
+      "loss": 0.6466,
+      "step": 18370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.845219583256462e-06,
+      "loss": 0.6814,
+      "step": 18371
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.84422917299406e-06,
+      "loss": 0.6128,
+      "step": 18372
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.843238812003579e-06,
+      "loss": 0.614,
+      "step": 18373
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.842248500296761e-06,
+      "loss": 0.6695,
+      "step": 18374
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.841258237885345e-06,
+      "loss": 0.6227,
+      "step": 18375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.840268024781075e-06,
+      "loss": 0.7205,
+      "step": 18376
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.839277860995689e-06,
+      "loss": 0.6554,
+      "step": 18377
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.838287746540927e-06,
+      "loss": 0.5762,
+      "step": 18378
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.837297681428529e-06,
+      "loss": 0.5861,
+      "step": 18379
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.836307665670236e-06,
+      "loss": 0.6945,
+      "step": 18380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.835317699277774e-06,
+      "loss": 0.5967,
+      "step": 18381
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.834327782262894e-06,
+      "loss": 0.6365,
+      "step": 18382
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.833337914637331e-06,
+      "loss": 0.6216,
+      "step": 18383
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.832348096412814e-06,
+      "loss": 0.5834,
+      "step": 18384
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.831358327601087e-06,
+      "loss": 0.6221,
+      "step": 18385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.830368608213876e-06,
+      "loss": 0.6294,
+      "step": 18386
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.829378938262915e-06,
+      "loss": 0.6371,
+      "step": 18387
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.828389317759953e-06,
+      "loss": 0.6047,
+      "step": 18388
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.827399746716707e-06,
+      "loss": 0.6135,
+      "step": 18389
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8264102251449154e-06,
+      "loss": 0.6323,
+      "step": 18390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8254207530563145e-06,
+      "loss": 0.6413,
+      "step": 18391
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.824431330462626e-06,
+      "loss": 0.6534,
+      "step": 18392
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.823441957375584e-06,
+      "loss": 0.6142,
+      "step": 18393
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.822452633806928e-06,
+      "loss": 0.637,
+      "step": 18394
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.821463359768378e-06,
+      "loss": 0.6188,
+      "step": 18395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.820474135271664e-06,
+      "loss": 0.5949,
+      "step": 18396
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8194849603285205e-06,
+      "loss": 0.6531,
+      "step": 18397
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8184958349506685e-06,
+      "loss": 0.618,
+      "step": 18398
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.817506759149834e-06,
+      "loss": 0.627,
+      "step": 18399
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.816517732937754e-06,
+      "loss": 0.63,
+      "step": 18400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.815528756326145e-06,
+      "loss": 0.6509,
+      "step": 18401
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.814539829326737e-06,
+      "loss": 0.6335,
+      "step": 18402
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.813550951951257e-06,
+      "loss": 0.5991,
+      "step": 18403
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.812562124211423e-06,
+      "loss": 0.6797,
+      "step": 18404
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.811573346118957e-06,
+      "loss": 0.6927,
+      "step": 18405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.810584617685597e-06,
+      "loss": 0.6544,
+      "step": 18406
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.809595938923051e-06,
+      "loss": 0.5932,
+      "step": 18407
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.808607309843046e-06,
+      "loss": 0.6231,
+      "step": 18408
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.807618730457307e-06,
+      "loss": 0.6401,
+      "step": 18409
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.80663020077755e-06,
+      "loss": 0.6642,
+      "step": 18410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.805641720815489e-06,
+      "loss": 0.6358,
+      "step": 18411
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.804653290582862e-06,
+      "loss": 0.6258,
+      "step": 18412
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.803664910091372e-06,
+      "loss": 0.5932,
+      "step": 18413
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.802676579352744e-06,
+      "loss": 0.6202,
+      "step": 18414
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.801688298378697e-06,
+      "loss": 0.601,
+      "step": 18415
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.800700067180943e-06,
+      "loss": 0.6008,
+      "step": 18416
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.799711885771198e-06,
+      "loss": 0.6468,
+      "step": 18417
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.798723754161191e-06,
+      "loss": 0.6057,
+      "step": 18418
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7977356723626235e-06,
+      "loss": 0.6014,
+      "step": 18419
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.796747640387215e-06,
+      "loss": 0.601,
+      "step": 18420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.795759658246684e-06,
+      "loss": 0.618,
+      "step": 18421
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.794771725952738e-06,
+      "loss": 0.631,
+      "step": 18422
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.793783843517089e-06,
+      "loss": 0.6554,
+      "step": 18423
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.792796010951461e-06,
+      "loss": 0.6432,
+      "step": 18424
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.791808228267554e-06,
+      "loss": 0.6487,
+      "step": 18425
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.790820495477083e-06,
+      "loss": 0.6173,
+      "step": 18426
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.789832812591764e-06,
+      "loss": 0.6434,
+      "step": 18427
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.788845179623301e-06,
+      "loss": 0.5973,
+      "step": 18428
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.787857596583399e-06,
+      "loss": 0.5873,
+      "step": 18429
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.786870063483784e-06,
+      "loss": 0.6481,
+      "step": 18430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.785882580336148e-06,
+      "loss": 0.5765,
+      "step": 18431
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.784895147152206e-06,
+      "loss": 0.6154,
+      "step": 18432
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.783907763943668e-06,
+      "loss": 0.6633,
+      "step": 18433
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.782920430722232e-06,
+      "loss": 0.6305,
+      "step": 18434
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.781933147499605e-06,
+      "loss": 0.6345,
+      "step": 18435
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.780945914287505e-06,
+      "loss": 0.6399,
+      "step": 18436
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.779958731097626e-06,
+      "loss": 0.5914,
+      "step": 18437
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.778971597941671e-06,
+      "loss": 0.695,
+      "step": 18438
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.777984514831354e-06,
+      "loss": 0.6668,
+      "step": 18439
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.776997481778367e-06,
+      "loss": 0.6328,
+      "step": 18440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.776010498794414e-06,
+      "loss": 0.6685,
+      "step": 18441
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7750235658912045e-06,
+      "loss": 0.6609,
+      "step": 18442
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.774036683080435e-06,
+      "loss": 0.6175,
+      "step": 18443
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.773049850373804e-06,
+      "loss": 0.655,
+      "step": 18444
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7720630677830194e-06,
+      "loss": 0.6347,
+      "step": 18445
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.771076335319771e-06,
+      "loss": 0.6485,
+      "step": 18446
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.770089652995763e-06,
+      "loss": 0.6577,
+      "step": 18447
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.769103020822692e-06,
+      "loss": 0.6742,
+      "step": 18448
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.768116438812257e-06,
+      "loss": 0.6585,
+      "step": 18449
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.767129906976159e-06,
+      "loss": 0.6695,
+      "step": 18450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.766143425326087e-06,
+      "loss": 0.6016,
+      "step": 18451
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7651569938737395e-06,
+      "loss": 0.5678,
+      "step": 18452
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.764170612630814e-06,
+      "loss": 0.6535,
+      "step": 18453
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.763184281609004e-06,
+      "loss": 0.6216,
+      "step": 18454
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.762198000820004e-06,
+      "loss": 0.6383,
+      "step": 18455
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.761211770275512e-06,
+      "loss": 0.6483,
+      "step": 18456
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.760225589987212e-06,
+      "loss": 0.6113,
+      "step": 18457
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.759239459966801e-06,
+      "loss": 0.6571,
+      "step": 18458
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.75825338022597e-06,
+      "loss": 0.5742,
+      "step": 18459
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.757267350776412e-06,
+      "loss": 0.6476,
+      "step": 18460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.756281371629817e-06,
+      "loss": 0.6199,
+      "step": 18461
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.755295442797877e-06,
+      "loss": 0.6304,
+      "step": 18462
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.754309564292276e-06,
+      "loss": 0.6455,
+      "step": 18463
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.753323736124707e-06,
+      "loss": 0.6818,
+      "step": 18464
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.752337958306856e-06,
+      "loss": 0.672,
+      "step": 18465
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.751352230850412e-06,
+      "loss": 0.6221,
+      "step": 18466
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.750366553767061e-06,
+      "loss": 0.6119,
+      "step": 18467
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7493809270684965e-06,
+      "loss": 0.6296,
+      "step": 18468
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.748395350766394e-06,
+      "loss": 0.6199,
+      "step": 18469
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.747409824872441e-06,
+      "loss": 0.6244,
+      "step": 18470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7464243493983256e-06,
+      "loss": 0.6685,
+      "step": 18471
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.745438924355729e-06,
+      "loss": 0.6264,
+      "step": 18472
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.744453549756337e-06,
+      "loss": 0.6076,
+      "step": 18473
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.743468225611834e-06,
+      "loss": 0.6249,
+      "step": 18474
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.742482951933899e-06,
+      "loss": 0.6141,
+      "step": 18475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.741497728734211e-06,
+      "loss": 0.6314,
+      "step": 18476
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.740512556024455e-06,
+      "loss": 0.625,
+      "step": 18477
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.739527433816313e-06,
+      "loss": 0.5982,
+      "step": 18478
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.738542362121462e-06,
+      "loss": 0.6448,
+      "step": 18479
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.737557340951585e-06,
+      "loss": 0.6305,
+      "step": 18480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.736572370318354e-06,
+      "loss": 0.5644,
+      "step": 18481
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7355874502334505e-06,
+      "loss": 0.649,
+      "step": 18482
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.734602580708554e-06,
+      "loss": 0.656,
+      "step": 18483
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.733617761755339e-06,
+      "loss": 0.6611,
+      "step": 18484
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.732632993385481e-06,
+      "loss": 0.655,
+      "step": 18485
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.731648275610663e-06,
+      "loss": 0.6325,
+      "step": 18486
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.730663608442549e-06,
+      "loss": 0.628,
+      "step": 18487
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.729678991892819e-06,
+      "loss": 0.6326,
+      "step": 18488
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.728694425973147e-06,
+      "loss": 0.6195,
+      "step": 18489
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.727709910695205e-06,
+      "loss": 0.6486,
+      "step": 18490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.726725446070667e-06,
+      "loss": 0.6689,
+      "step": 18491
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.725741032111208e-06,
+      "loss": 0.646,
+      "step": 18492
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.724756668828493e-06,
+      "loss": 0.6355,
+      "step": 18493
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.723772356234194e-06,
+      "loss": 0.6673,
+      "step": 18494
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.722788094339985e-06,
+      "loss": 0.6282,
+      "step": 18495
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.721803883157533e-06,
+      "loss": 0.6325,
+      "step": 18496
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.720819722698509e-06,
+      "loss": 0.6026,
+      "step": 18497
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.719835612974583e-06,
+      "loss": 0.5965,
+      "step": 18498
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.718851553997418e-06,
+      "loss": 0.656,
+      "step": 18499
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.717867545778681e-06,
+      "loss": 0.5985,
+      "step": 18500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.716883588330043e-06,
+      "loss": 0.6137,
+      "step": 18501
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.715899681663168e-06,
+      "loss": 0.6414,
+      "step": 18502
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.714915825789722e-06,
+      "loss": 0.6264,
+      "step": 18503
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.713932020721372e-06,
+      "loss": 0.6285,
+      "step": 18504
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7129482664697775e-06,
+      "loss": 0.6455,
+      "step": 18505
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.711964563046604e-06,
+      "loss": 0.6454,
+      "step": 18506
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.710980910463514e-06,
+      "loss": 0.663,
+      "step": 18507
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.709997308732171e-06,
+      "loss": 0.5771,
+      "step": 18508
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.709013757864243e-06,
+      "loss": 0.6843,
+      "step": 18509
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7080302578713786e-06,
+      "loss": 0.6139,
+      "step": 18510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.707046808765246e-06,
+      "loss": 0.6547,
+      "step": 18511
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.706063410557503e-06,
+      "loss": 0.6701,
+      "step": 18512
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.705080063259811e-06,
+      "loss": 0.6222,
+      "step": 18513
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7040967668838286e-06,
+      "loss": 0.6315,
+      "step": 18514
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.703113521441217e-06,
+      "loss": 0.6334,
+      "step": 18515
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.702130326943625e-06,
+      "loss": 0.5964,
+      "step": 18516
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.701147183402715e-06,
+      "loss": 0.6433,
+      "step": 18517
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.700164090830145e-06,
+      "loss": 0.6044,
+      "step": 18518
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.699181049237568e-06,
+      "loss": 0.6298,
+      "step": 18519
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.69819805863664e-06,
+      "loss": 0.6542,
+      "step": 18520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.69721511903902e-06,
+      "loss": 0.6473,
+      "step": 18521
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.696232230456353e-06,
+      "loss": 0.6288,
+      "step": 18522
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.695249392900299e-06,
+      "loss": 0.629,
+      "step": 18523
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.694266606382508e-06,
+      "loss": 0.6146,
+      "step": 18524
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.693283870914633e-06,
+      "loss": 0.6777,
+      "step": 18525
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.692301186508327e-06,
+      "loss": 0.606,
+      "step": 18526
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6913185531752425e-06,
+      "loss": 0.6198,
+      "step": 18527
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.690335970927022e-06,
+      "loss": 0.6426,
+      "step": 18528
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.689353439775323e-06,
+      "loss": 0.6636,
+      "step": 18529
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.68837095973179e-06,
+      "loss": 0.6337,
+      "step": 18530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.687388530808073e-06,
+      "loss": 0.6394,
+      "step": 18531
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.686406153015822e-06,
+      "loss": 0.6539,
+      "step": 18532
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.685423826366686e-06,
+      "loss": 0.6678,
+      "step": 18533
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.684441550872304e-06,
+      "loss": 0.6449,
+      "step": 18534
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.683459326544327e-06,
+      "loss": 0.6104,
+      "step": 18535
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.682477153394401e-06,
+      "loss": 0.6503,
+      "step": 18536
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.68149503143417e-06,
+      "loss": 0.6039,
+      "step": 18537
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.680512960675277e-06,
+      "loss": 0.6505,
+      "step": 18538
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.679530941129372e-06,
+      "loss": 0.6557,
+      "step": 18539
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.678548972808089e-06,
+      "loss": 0.6932,
+      "step": 18540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.677567055723074e-06,
+      "loss": 0.6243,
+      "step": 18541
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.676585189885971e-06,
+      "loss": 0.6445,
+      "step": 18542
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.675603375308419e-06,
+      "loss": 0.6841,
+      "step": 18543
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.674621612002059e-06,
+      "loss": 0.6448,
+      "step": 18544
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.673639899978536e-06,
+      "loss": 0.6119,
+      "step": 18545
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.67265823924948e-06,
+      "loss": 0.6367,
+      "step": 18546
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.671676629826535e-06,
+      "loss": 0.6428,
+      "step": 18547
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.670695071721339e-06,
+      "loss": 0.6887,
+      "step": 18548
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.669713564945529e-06,
+      "loss": 0.7034,
+      "step": 18549
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.668732109510744e-06,
+      "loss": 0.6404,
+      "step": 18550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.667750705428622e-06,
+      "loss": 0.6096,
+      "step": 18551
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.666769352710791e-06,
+      "loss": 0.6785,
+      "step": 18552
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.665788051368892e-06,
+      "loss": 0.6129,
+      "step": 18553
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.664806801414557e-06,
+      "loss": 0.6169,
+      "step": 18554
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.663825602859423e-06,
+      "loss": 0.6397,
+      "step": 18555
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.662844455715122e-06,
+      "loss": 0.607,
+      "step": 18556
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.661863359993289e-06,
+      "loss": 0.622,
+      "step": 18557
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.660882315705551e-06,
+      "loss": 0.6601,
+      "step": 18558
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.659901322863541e-06,
+      "loss": 0.627,
+      "step": 18559
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6589203814788915e-06,
+      "loss": 0.5909,
+      "step": 18560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6579394915632336e-06,
+      "loss": 0.5948,
+      "step": 18561
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.656958653128194e-06,
+      "loss": 0.6977,
+      "step": 18562
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.655977866185408e-06,
+      "loss": 0.6197,
+      "step": 18563
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.654997130746496e-06,
+      "loss": 0.6291,
+      "step": 18564
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.654016446823089e-06,
+      "loss": 0.6803,
+      "step": 18565
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.653035814426815e-06,
+      "loss": 0.6495,
+      "step": 18566
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6520552335693e-06,
+      "loss": 0.6599,
+      "step": 18567
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.651074704262172e-06,
+      "loss": 0.6138,
+      "step": 18568
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.650094226517056e-06,
+      "loss": 0.6902,
+      "step": 18569
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.649113800345573e-06,
+      "loss": 0.6242,
+      "step": 18570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.64813342575935e-06,
+      "loss": 0.6066,
+      "step": 18571
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.647153102770011e-06,
+      "loss": 0.6436,
+      "step": 18572
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.646172831389178e-06,
+      "loss": 0.5881,
+      "step": 18573
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6451926116284774e-06,
+      "loss": 0.6671,
+      "step": 18574
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.644212443499523e-06,
+      "loss": 0.6635,
+      "step": 18575
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.643232327013941e-06,
+      "loss": 0.6364,
+      "step": 18576
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.642252262183354e-06,
+      "loss": 0.6471,
+      "step": 18577
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.641272249019372e-06,
+      "loss": 0.615,
+      "step": 18578
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6402922875336245e-06,
+      "loss": 0.6188,
+      "step": 18579
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.639312377737731e-06,
+      "loss": 0.6152,
+      "step": 18580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.638332519643302e-06,
+      "loss": 0.6198,
+      "step": 18581
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.637352713261959e-06,
+      "loss": 0.6897,
+      "step": 18582
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6363729586053206e-06,
+      "loss": 0.6419,
+      "step": 18583
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.635393255684995e-06,
+      "loss": 0.6437,
+      "step": 18584
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.634413604512605e-06,
+      "loss": 0.7086,
+      "step": 18585
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.63343400509977e-06,
+      "loss": 0.6161,
+      "step": 18586
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.632454457458094e-06,
+      "loss": 0.6126,
+      "step": 18587
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.631474961599197e-06,
+      "loss": 0.6599,
+      "step": 18588
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.630495517534693e-06,
+      "loss": 0.5881,
+      "step": 18589
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.629516125276184e-06,
+      "loss": 0.6963,
+      "step": 18590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.628536784835295e-06,
+      "loss": 0.6453,
+      "step": 18591
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6275574962236354e-06,
+      "loss": 0.6465,
+      "step": 18592
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.626578259452808e-06,
+      "loss": 0.6325,
+      "step": 18593
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.62559907453443e-06,
+      "loss": 0.6277,
+      "step": 18594
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6246199414801115e-06,
+      "loss": 0.6301,
+      "step": 18595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.623640860301452e-06,
+      "loss": 0.6481,
+      "step": 18596
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6226618310100675e-06,
+      "loss": 0.6101,
+      "step": 18597
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6216828536175694e-06,
+      "loss": 0.6409,
+      "step": 18598
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.620703928135556e-06,
+      "loss": 0.6581,
+      "step": 18599
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.619725054575638e-06,
+      "loss": 0.6462,
+      "step": 18600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.618746232949423e-06,
+      "loss": 0.6213,
+      "step": 18601
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.617767463268506e-06,
+      "loss": 0.6092,
+      "step": 18602
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6167887455445044e-06,
+      "loss": 0.6093,
+      "step": 18603
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.615810079789022e-06,
+      "loss": 0.6661,
+      "step": 18604
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.61483146601365e-06,
+      "loss": 0.6163,
+      "step": 18605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.613852904229999e-06,
+      "loss": 0.6147,
+      "step": 18606
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.612874394449676e-06,
+      "loss": 0.6331,
+      "step": 18607
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.611895936684269e-06,
+      "loss": 0.6213,
+      "step": 18608
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.610917530945388e-06,
+      "loss": 0.6206,
+      "step": 18609
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.609939177244639e-06,
+      "loss": 0.6367,
+      "step": 18610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.608960875593608e-06,
+      "loss": 0.6473,
+      "step": 18611
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.607982626003901e-06,
+      "loss": 0.6446,
+      "step": 18612
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.60700442848712e-06,
+      "loss": 0.5724,
+      "step": 18613
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.606026283054852e-06,
+      "loss": 0.6074,
+      "step": 18614
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.605048189718704e-06,
+      "loss": 0.6849,
+      "step": 18615
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.604070148490273e-06,
+      "loss": 0.6634,
+      "step": 18616
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.603092159381147e-06,
+      "loss": 0.5905,
+      "step": 18617
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.602114222402925e-06,
+      "loss": 0.5544,
+      "step": 18618
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.601136337567209e-06,
+      "loss": 0.6187,
+      "step": 18619
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.600158504885577e-06,
+      "loss": 0.6261,
+      "step": 18620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.599180724369635e-06,
+      "loss": 0.6219,
+      "step": 18621
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.598202996030977e-06,
+      "loss": 0.628,
+      "step": 18622
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.597225319881187e-06,
+      "loss": 0.6022,
+      "step": 18623
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5962476959318605e-06,
+      "loss": 0.6413,
+      "step": 18624
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5952701241945915e-06,
+      "loss": 0.5863,
+      "step": 18625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.59429260468096e-06,
+      "loss": 0.6635,
+      "step": 18626
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5933151374025664e-06,
+      "loss": 0.6542,
+      "step": 18627
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.592337722371e-06,
+      "loss": 0.6838,
+      "step": 18628
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.591360359597842e-06,
+      "loss": 0.675,
+      "step": 18629
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.590383049094683e-06,
+      "loss": 0.6635,
+      "step": 18630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.589405790873117e-06,
+      "loss": 0.6485,
+      "step": 18631
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.588428584944716e-06,
+      "loss": 0.5744,
+      "step": 18632
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.587451431321078e-06,
+      "loss": 0.6118,
+      "step": 18633
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.586474330013791e-06,
+      "loss": 0.6431,
+      "step": 18634
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.585497281034428e-06,
+      "loss": 0.6574,
+      "step": 18635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.584520284394584e-06,
+      "loss": 0.606,
+      "step": 18636
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.583543340105835e-06,
+      "loss": 0.6273,
+      "step": 18637
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.582566448179761e-06,
+      "loss": 0.5939,
+      "step": 18638
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.581589608627958e-06,
+      "loss": 0.5921,
+      "step": 18639
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.580612821461995e-06,
+      "loss": 0.6303,
+      "step": 18640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.579636086693459e-06,
+      "loss": 0.6122,
+      "step": 18641
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.57865940433393e-06,
+      "loss": 0.663,
+      "step": 18642
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.577682774394985e-06,
+      "loss": 0.6854,
+      "step": 18643
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.576706196888202e-06,
+      "loss": 0.6024,
+      "step": 18644
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.575729671825169e-06,
+      "loss": 0.6712,
+      "step": 18645
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.574753199217453e-06,
+      "loss": 0.6662,
+      "step": 18646
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5737767790766375e-06,
+      "loss": 0.6502,
+      "step": 18647
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.572800411414301e-06,
+      "loss": 0.6865,
+      "step": 18648
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.571824096242011e-06,
+      "loss": 0.5867,
+      "step": 18649
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.570847833571344e-06,
+      "loss": 0.6256,
+      "step": 18650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.569871623413887e-06,
+      "loss": 0.6101,
+      "step": 18651
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.568895465781203e-06,
+      "loss": 0.6574,
+      "step": 18652
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.567919360684868e-06,
+      "loss": 0.6543,
+      "step": 18653
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.566943308136461e-06,
+      "loss": 0.6294,
+      "step": 18654
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.565967308147543e-06,
+      "loss": 0.5899,
+      "step": 18655
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.564991360729688e-06,
+      "loss": 0.6456,
+      "step": 18656
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.564015465894479e-06,
+      "loss": 0.6163,
+      "step": 18657
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.563039623653474e-06,
+      "loss": 0.6577,
+      "step": 18658
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.562063834018247e-06,
+      "loss": 0.6126,
+      "step": 18659
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.561088097000373e-06,
+      "loss": 0.6038,
+      "step": 18660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.56011241261141e-06,
+      "loss": 0.65,
+      "step": 18661
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.559136780862926e-06,
+      "loss": 0.6834,
+      "step": 18662
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.558161201766501e-06,
+      "loss": 0.6208,
+      "step": 18663
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.557185675333689e-06,
+      "loss": 0.6537,
+      "step": 18664
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.556210201576063e-06,
+      "loss": 0.6154,
+      "step": 18665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.55523478050519e-06,
+      "loss": 0.6564,
+      "step": 18666
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.554259412132628e-06,
+      "loss": 0.6231,
+      "step": 18667
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5532840964699396e-06,
+      "loss": 0.6143,
+      "step": 18668
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.552308833528701e-06,
+      "loss": 0.6424,
+      "step": 18669
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.551333623320464e-06,
+      "loss": 0.6034,
+      "step": 18670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.550358465856796e-06,
+      "loss": 0.5745,
+      "step": 18671
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.549383361149259e-06,
+      "loss": 0.6572,
+      "step": 18672
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.548408309209409e-06,
+      "loss": 0.6309,
+      "step": 18673
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5474333100488056e-06,
+      "loss": 0.6482,
+      "step": 18674
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5464583636790214e-06,
+      "loss": 0.6908,
+      "step": 18675
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.545483470111603e-06,
+      "loss": 0.6263,
+      "step": 18676
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.544508629358112e-06,
+      "loss": 0.6475,
+      "step": 18677
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.543533841430112e-06,
+      "loss": 0.6703,
+      "step": 18678
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.542559106339152e-06,
+      "loss": 0.6379,
+      "step": 18679
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.541584424096793e-06,
+      "loss": 0.5784,
+      "step": 18680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.54060979471459e-06,
+      "loss": 0.6447,
+      "step": 18681
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.539635218204099e-06,
+      "loss": 0.6204,
+      "step": 18682
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.538660694576875e-06,
+      "loss": 0.6124,
+      "step": 18683
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.537686223844476e-06,
+      "loss": 0.5753,
+      "step": 18684
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.536711806018448e-06,
+      "loss": 0.7029,
+      "step": 18685
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.535737441110347e-06,
+      "loss": 0.613,
+      "step": 18686
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.534763129131726e-06,
+      "loss": 0.6263,
+      "step": 18687
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.533788870094137e-06,
+      "loss": 0.6055,
+      "step": 18688
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5328146640091315e-06,
+      "loss": 0.6021,
+      "step": 18689
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.531840510888262e-06,
+      "loss": 0.6555,
+      "step": 18690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.530866410743072e-06,
+      "loss": 0.6272,
+      "step": 18691
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.529892363585114e-06,
+      "loss": 0.64,
+      "step": 18692
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.528918369425937e-06,
+      "loss": 0.5969,
+      "step": 18693
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.527944428277088e-06,
+      "loss": 0.6023,
+      "step": 18694
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.526970540150116e-06,
+      "loss": 0.6231,
+      "step": 18695
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.52599670505657e-06,
+      "loss": 0.6478,
+      "step": 18696
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.525022923007989e-06,
+      "loss": 0.6373,
+      "step": 18697
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.524049194015921e-06,
+      "loss": 0.584,
+      "step": 18698
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.523075518091914e-06,
+      "loss": 0.5929,
+      "step": 18699
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.522101895247509e-06,
+      "loss": 0.6489,
+      "step": 18700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.521128325494254e-06,
+      "loss": 0.7073,
+      "step": 18701
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.520154808843685e-06,
+      "loss": 0.62,
+      "step": 18702
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.519181345307348e-06,
+      "loss": 0.6222,
+      "step": 18703
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.518207934896784e-06,
+      "loss": 0.6505,
+      "step": 18704
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.517234577623535e-06,
+      "loss": 0.6894,
+      "step": 18705
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.516261273499139e-06,
+      "loss": 0.5831,
+      "step": 18706
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.515288022535144e-06,
+      "loss": 0.6026,
+      "step": 18707
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.514314824743078e-06,
+      "loss": 0.624,
+      "step": 18708
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.513341680134483e-06,
+      "loss": 0.6011,
+      "step": 18709
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5123685887209e-06,
+      "loss": 0.6429,
+      "step": 18710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.511395550513863e-06,
+      "loss": 0.6626,
+      "step": 18711
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.51042256552491e-06,
+      "loss": 0.6476,
+      "step": 18712
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.50944963376558e-06,
+      "loss": 0.6487,
+      "step": 18713
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.508476755247403e-06,
+      "loss": 0.6274,
+      "step": 18714
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.507503929981914e-06,
+      "loss": 0.6728,
+      "step": 18715
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.50653115798065e-06,
+      "loss": 0.6294,
+      "step": 18716
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.505558439255144e-06,
+      "loss": 0.6568,
+      "step": 18717
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5045857738169274e-06,
+      "loss": 0.6457,
+      "step": 18718
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5036131616775365e-06,
+      "loss": 0.612,
+      "step": 18719
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.502640602848497e-06,
+      "loss": 0.6166,
+      "step": 18720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.5016680973413416e-06,
+      "loss": 0.5848,
+      "step": 18721
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.500695645167602e-06,
+      "loss": 0.6077,
+      "step": 18722
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.499723246338806e-06,
+      "loss": 0.6225,
+      "step": 18723
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.498750900866484e-06,
+      "loss": 0.552,
+      "step": 18724
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.497778608762168e-06,
+      "loss": 0.6493,
+      "step": 18725
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.49680637003738e-06,
+      "loss": 0.6355,
+      "step": 18726
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.495834184703647e-06,
+      "loss": 0.6216,
+      "step": 18727
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.494862052772498e-06,
+      "loss": 0.6353,
+      "step": 18728
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.493889974255458e-06,
+      "loss": 0.6077,
+      "step": 18729
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4929179491640535e-06,
+      "loss": 0.627,
+      "step": 18730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4919459775098115e-06,
+      "loss": 0.6438,
+      "step": 18731
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.490974059304248e-06,
+      "loss": 0.6157,
+      "step": 18732
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.490002194558892e-06,
+      "loss": 0.6238,
+      "step": 18733
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4890303832852655e-06,
+      "loss": 0.625,
+      "step": 18734
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.48805862549489e-06,
+      "loss": 0.5963,
+      "step": 18735
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.487086921199287e-06,
+      "loss": 0.6191,
+      "step": 18736
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4861152704099805e-06,
+      "loss": 0.6446,
+      "step": 18737
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4851436731384845e-06,
+      "loss": 0.6068,
+      "step": 18738
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.484172129396322e-06,
+      "loss": 0.6133,
+      "step": 18739
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.48320063919501e-06,
+      "loss": 0.6453,
+      "step": 18740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.48222920254607e-06,
+      "loss": 0.6771,
+      "step": 18741
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.481257819461016e-06,
+      "loss": 0.6446,
+      "step": 18742
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.480286489951372e-06,
+      "loss": 0.6516,
+      "step": 18743
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.479315214028645e-06,
+      "loss": 0.6111,
+      "step": 18744
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.478343991704354e-06,
+      "loss": 0.58,
+      "step": 18745
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.477372822990016e-06,
+      "loss": 0.6264,
+      "step": 18746
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4764017078971436e-06,
+      "loss": 0.6105,
+      "step": 18747
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.475430646437252e-06,
+      "loss": 0.6469,
+      "step": 18748
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.474459638621857e-06,
+      "loss": 0.5758,
+      "step": 18749
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.473488684462465e-06,
+      "loss": 0.6332,
+      "step": 18750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.47251778397059e-06,
+      "loss": 0.6013,
+      "step": 18751
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.471546937157743e-06,
+      "loss": 0.6139,
+      "step": 18752
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.470576144035437e-06,
+      "loss": 0.6841,
+      "step": 18753
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4696054046151795e-06,
+      "loss": 0.6233,
+      "step": 18754
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.468634718908486e-06,
+      "loss": 0.6672,
+      "step": 18755
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4676640869268545e-06,
+      "loss": 0.6378,
+      "step": 18756
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4666935086818e-06,
+      "loss": 0.5949,
+      "step": 18757
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.465722984184828e-06,
+      "loss": 0.6301,
+      "step": 18758
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.464752513447447e-06,
+      "loss": 0.6329,
+      "step": 18759
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.463782096481165e-06,
+      "loss": 0.6536,
+      "step": 18760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.46281173329748e-06,
+      "loss": 0.5711,
+      "step": 18761
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.461841423907903e-06,
+      "loss": 0.6513,
+      "step": 18762
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.460871168323936e-06,
+      "loss": 0.614,
+      "step": 18763
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.459900966557084e-06,
+      "loss": 0.6025,
+      "step": 18764
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.458930818618848e-06,
+      "loss": 0.6519,
+      "step": 18765
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.457960724520737e-06,
+      "loss": 0.6151,
+      "step": 18766
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.456990684274241e-06,
+      "loss": 0.6653,
+      "step": 18767
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.456020697890869e-06,
+      "loss": 0.6357,
+      "step": 18768
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.455050765382121e-06,
+      "loss": 0.6588,
+      "step": 18769
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.454080886759494e-06,
+      "loss": 0.6052,
+      "step": 18770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.453111062034489e-06,
+      "loss": 0.6235,
+      "step": 18771
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.452141291218609e-06,
+      "loss": 0.6141,
+      "step": 18772
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.451171574323342e-06,
+      "loss": 0.6605,
+      "step": 18773
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.45020191136019e-06,
+      "loss": 0.5592,
+      "step": 18774
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.44923230234065e-06,
+      "loss": 0.5957,
+      "step": 18775
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.448262747276219e-06,
+      "loss": 0.6443,
+      "step": 18776
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.44729324617839e-06,
+      "loss": 0.6179,
+      "step": 18777
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.446323799058664e-06,
+      "loss": 0.6301,
+      "step": 18778
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.445354405928524e-06,
+      "loss": 0.6541,
+      "step": 18779
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.444385066799471e-06,
+      "loss": 0.63,
+      "step": 18780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.443415781682995e-06,
+      "loss": 0.5873,
+      "step": 18781
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.442446550590589e-06,
+      "loss": 0.6209,
+      "step": 18782
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.441477373533744e-06,
+      "loss": 0.6232,
+      "step": 18783
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.440508250523956e-06,
+      "loss": 0.5795,
+      "step": 18784
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.439539181572706e-06,
+      "loss": 0.6263,
+      "step": 18785
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.438570166691488e-06,
+      "loss": 0.6504,
+      "step": 18786
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.437601205891791e-06,
+      "loss": 0.6331,
+      "step": 18787
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.436632299185105e-06,
+      "loss": 0.5978,
+      "step": 18788
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.435663446582913e-06,
+      "loss": 0.5912,
+      "step": 18789
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.43469464809671e-06,
+      "loss": 0.6343,
+      "step": 18790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.433725903737974e-06,
+      "loss": 0.6708,
+      "step": 18791
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4327572135181915e-06,
+      "loss": 0.6066,
+      "step": 18792
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.431788577448852e-06,
+      "loss": 0.661,
+      "step": 18793
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4308199955414365e-06,
+      "loss": 0.6266,
+      "step": 18794
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.42985146780743e-06,
+      "loss": 0.6041,
+      "step": 18795
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.42888299425832e-06,
+      "loss": 0.634,
+      "step": 18796
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4279145749055814e-06,
+      "loss": 0.6433,
+      "step": 18797
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.426946209760698e-06,
+      "loss": 0.641,
+      "step": 18798
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.425977898835153e-06,
+      "loss": 0.6884,
+      "step": 18799
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.425009642140426e-06,
+      "loss": 0.6571,
+      "step": 18800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.424041439687997e-06,
+      "loss": 0.6164,
+      "step": 18801
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.423073291489349e-06,
+      "loss": 0.6044,
+      "step": 18802
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.422105197555954e-06,
+      "loss": 0.5461,
+      "step": 18803
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.421137157899291e-06,
+      "loss": 0.6546,
+      "step": 18804
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.42016917253084e-06,
+      "loss": 0.7119,
+      "step": 18805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.419201241462078e-06,
+      "loss": 0.6135,
+      "step": 18806
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.418233364704479e-06,
+      "loss": 0.5968,
+      "step": 18807
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4172655422695246e-06,
+      "loss": 0.6308,
+      "step": 18808
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.41629777416868e-06,
+      "loss": 0.6069,
+      "step": 18809
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.415330060413423e-06,
+      "loss": 0.6512,
+      "step": 18810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.414362401015228e-06,
+      "loss": 0.6018,
+      "step": 18811
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.413394795985567e-06,
+      "loss": 0.6111,
+      "step": 18812
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.412427245335914e-06,
+      "loss": 0.6455,
+      "step": 18813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.411459749077743e-06,
+      "loss": 0.6587,
+      "step": 18814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.410492307222517e-06,
+      "loss": 0.6637,
+      "step": 18815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4095249197817105e-06,
+      "loss": 0.6533,
+      "step": 18816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.408557586766794e-06,
+      "loss": 0.6098,
+      "step": 18817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4075903081892345e-06,
+      "loss": 0.6572,
+      "step": 18818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4066230840605005e-06,
+      "loss": 0.6085,
+      "step": 18819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.405655914392065e-06,
+      "loss": 0.6952,
+      "step": 18820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4046887991953874e-06,
+      "loss": 0.6558,
+      "step": 18821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.403721738481941e-06,
+      "loss": 0.6022,
+      "step": 18822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.402754732263179e-06,
+      "loss": 0.6361,
+      "step": 18823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.4017877805505806e-06,
+      "loss": 0.6465,
+      "step": 18824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.400820883355606e-06,
+      "loss": 0.6199,
+      "step": 18825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.399854040689716e-06,
+      "loss": 0.6287,
+      "step": 18826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3988872525643755e-06,
+      "loss": 0.6634,
+      "step": 18827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.397920518991051e-06,
+      "loss": 0.681,
+      "step": 18828
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.396953839981191e-06,
+      "loss": 0.6088,
+      "step": 18829
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.395987215546272e-06,
+      "loss": 0.6391,
+      "step": 18830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.395020645697753e-06,
+      "loss": 0.6673,
+      "step": 18831
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.394054130447084e-06,
+      "loss": 0.6078,
+      "step": 18832
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.39308766980573e-06,
+      "loss": 0.6446,
+      "step": 18833
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.392121263785154e-06,
+      "loss": 0.6705,
+      "step": 18834
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.391154912396801e-06,
+      "loss": 0.5928,
+      "step": 18835
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.390188615652141e-06,
+      "loss": 0.6418,
+      "step": 18836
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.38922237356263e-06,
+      "loss": 0.6615,
+      "step": 18837
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.388256186139718e-06,
+      "loss": 0.6757,
+      "step": 18838
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.38729005339486e-06,
+      "loss": 0.6844,
+      "step": 18839
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.38632397533952e-06,
+      "loss": 0.6342,
+      "step": 18840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.385357951985137e-06,
+      "loss": 0.6112,
+      "step": 18841
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.384391983343175e-06,
+      "loss": 0.6574,
+      "step": 18842
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3834260694250915e-06,
+      "loss": 0.6427,
+      "step": 18843
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.382460210242328e-06,
+      "loss": 0.5852,
+      "step": 18844
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.381494405806338e-06,
+      "loss": 0.5983,
+      "step": 18845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.38052865612858e-06,
+      "loss": 0.6795,
+      "step": 18846
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.379562961220489e-06,
+      "loss": 0.6414,
+      "step": 18847
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.378597321093528e-06,
+      "loss": 0.611,
+      "step": 18848
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.377631735759148e-06,
+      "loss": 0.6227,
+      "step": 18849
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.376666205228784e-06,
+      "loss": 0.6264,
+      "step": 18850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.375700729513893e-06,
+      "loss": 0.5987,
+      "step": 18851
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.374735308625923e-06,
+      "loss": 0.6112,
+      "step": 18852
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.373769942576308e-06,
+      "loss": 0.6333,
+      "step": 18853
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.372804631376508e-06,
+      "loss": 0.6481,
+      "step": 18854
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.371839375037966e-06,
+      "loss": 0.5923,
+      "step": 18855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.370874173572118e-06,
+      "loss": 0.6754,
+      "step": 18856
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.369909026990414e-06,
+      "loss": 0.6217,
+      "step": 18857
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3689439353042985e-06,
+      "loss": 0.6056,
+      "step": 18858
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.367978898525203e-06,
+      "loss": 0.6598,
+      "step": 18859
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.367013916664582e-06,
+      "loss": 0.639,
+      "step": 18860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.366048989733875e-06,
+      "loss": 0.6622,
+      "step": 18861
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.365084117744515e-06,
+      "loss": 0.655,
+      "step": 18862
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.364119300707947e-06,
+      "loss": 0.6028,
+      "step": 18863
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.363154538635613e-06,
+      "loss": 0.6266,
+      "step": 18864
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.362189831538938e-06,
+      "loss": 0.6886,
+      "step": 18865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.361225179429375e-06,
+      "loss": 0.6604,
+      "step": 18866
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3602605823183596e-06,
+      "loss": 0.6141,
+      "step": 18867
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.359296040217319e-06,
+      "loss": 0.6597,
+      "step": 18868
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.358331553137696e-06,
+      "loss": 0.6426,
+      "step": 18869
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3573671210909275e-06,
+      "loss": 0.6203,
+      "step": 18870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.356402744088436e-06,
+      "loss": 0.6152,
+      "step": 18871
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.355438422141669e-06,
+      "loss": 0.5862,
+      "step": 18872
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.354474155262059e-06,
+      "loss": 0.6558,
+      "step": 18873
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3535099434610305e-06,
+      "loss": 0.6316,
+      "step": 18874
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.352545786750019e-06,
+      "loss": 0.5731,
+      "step": 18875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.35158168514046e-06,
+      "loss": 0.6332,
+      "step": 18876
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.350617638643777e-06,
+      "loss": 0.6569,
+      "step": 18877
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3496536472713986e-06,
+      "loss": 0.6348,
+      "step": 18878
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.348689711034768e-06,
+      "loss": 0.6197,
+      "step": 18879
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.347725829945301e-06,
+      "loss": 0.6331,
+      "step": 18880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3467620040144275e-06,
+      "loss": 0.6237,
+      "step": 18881
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3457982332535826e-06,
+      "loss": 0.6315,
+      "step": 18882
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.344834517674182e-06,
+      "loss": 0.6083,
+      "step": 18883
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3438708572876586e-06,
+      "loss": 0.5961,
+      "step": 18884
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.342907252105436e-06,
+      "loss": 0.6464,
+      "step": 18885
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.341943702138939e-06,
+      "loss": 0.6502,
+      "step": 18886
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.340980207399594e-06,
+      "loss": 0.5785,
+      "step": 18887
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.34001676789882e-06,
+      "loss": 0.5916,
+      "step": 18888
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3390533836480414e-06,
+      "loss": 0.6119,
+      "step": 18889
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.338090054658682e-06,
+      "loss": 0.6071,
+      "step": 18890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.337126780942161e-06,
+      "loss": 0.6453,
+      "step": 18891
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3361635625099016e-06,
+      "loss": 0.6277,
+      "step": 18892
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3352003993733285e-06,
+      "loss": 0.6931,
+      "step": 18893
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.334237291543849e-06,
+      "loss": 0.6514,
+      "step": 18894
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.33327423903289e-06,
+      "loss": 0.6294,
+      "step": 18895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.332311241851869e-06,
+      "loss": 0.5796,
+      "step": 18896
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3313483000122e-06,
+      "loss": 0.6742,
+      "step": 18897
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3303854135253054e-06,
+      "loss": 0.5695,
+      "step": 18898
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.329422582402602e-06,
+      "loss": 0.5951,
+      "step": 18899
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.3284598066554985e-06,
+      "loss": 0.6495,
+      "step": 18900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.327497086295412e-06,
+      "loss": 0.6013,
+      "step": 18901
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.32653442133376e-06,
+      "loss": 0.5737,
+      "step": 18902
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.325571811781954e-06,
+      "loss": 0.5779,
+      "step": 18903
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.324609257651407e-06,
+      "loss": 0.6271,
+      "step": 18904
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.323646758953537e-06,
+      "loss": 0.5923,
+      "step": 18905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.322684315699744e-06,
+      "loss": 0.6211,
+      "step": 18906
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.321721927901448e-06,
+      "loss": 0.6183,
+      "step": 18907
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.320759595570055e-06,
+      "loss": 0.6006,
+      "step": 18908
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.319797318716976e-06,
+      "loss": 0.6158,
+      "step": 18909
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3188350973536205e-06,
+      "loss": 0.6429,
+      "step": 18910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.317872931491402e-06,
+      "loss": 0.6683,
+      "step": 18911
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.316910821141716e-06,
+      "loss": 0.5719,
+      "step": 18912
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.315948766315978e-06,
+      "loss": 0.643,
+      "step": 18913
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.314986767025593e-06,
+      "loss": 0.6013,
+      "step": 18914
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3140248232819665e-06,
+      "loss": 0.6021,
+      "step": 18915
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.313062935096502e-06,
+      "loss": 0.5982,
+      "step": 18916
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.312101102480611e-06,
+      "loss": 0.6555,
+      "step": 18917
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.311139325445687e-06,
+      "loss": 0.6555,
+      "step": 18918
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.310177604003138e-06,
+      "loss": 0.6999,
+      "step": 18919
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.309215938164367e-06,
+      "loss": 0.6087,
+      "step": 18920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.308254327940774e-06,
+      "loss": 0.6263,
+      "step": 18921
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.307292773343761e-06,
+      "loss": 0.6671,
+      "step": 18922
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.306331274384733e-06,
+      "loss": 0.6593,
+      "step": 18923
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.30536983107508e-06,
+      "loss": 0.6298,
+      "step": 18924
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.304408443426208e-06,
+      "loss": 0.633,
+      "step": 18925
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.303447111449512e-06,
+      "loss": 0.5582,
+      "step": 18926
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.302485835156392e-06,
+      "loss": 0.6083,
+      "step": 18927
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.301524614558247e-06,
+      "loss": 0.616,
+      "step": 18928
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.300563449666474e-06,
+      "loss": 0.6465,
+      "step": 18929
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.299602340492461e-06,
+      "loss": 0.5625,
+      "step": 18930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.29864128704761e-06,
+      "loss": 0.6688,
+      "step": 18931
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.297680289343312e-06,
+      "loss": 0.6615,
+      "step": 18932
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.296719347390964e-06,
+      "loss": 0.6228,
+      "step": 18933
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.295758461201957e-06,
+      "loss": 0.6261,
+      "step": 18934
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.294797630787689e-06,
+      "loss": 0.6084,
+      "step": 18935
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.293836856159544e-06,
+      "loss": 0.6616,
+      "step": 18936
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.292876137328915e-06,
+      "loss": 0.6259,
+      "step": 18937
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.291915474307194e-06,
+      "loss": 0.7038,
+      "step": 18938
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.290954867105771e-06,
+      "loss": 0.618,
+      "step": 18939
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.289994315736035e-06,
+      "loss": 0.7054,
+      "step": 18940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.28903382020938e-06,
+      "loss": 0.6369,
+      "step": 18941
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.288073380537182e-06,
+      "loss": 0.6876,
+      "step": 18942
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.287112996730837e-06,
+      "loss": 0.6247,
+      "step": 18943
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.286152668801727e-06,
+      "loss": 0.6239,
+      "step": 18944
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.285192396761242e-06,
+      "loss": 0.6843,
+      "step": 18945
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.284232180620769e-06,
+      "loss": 0.6264,
+      "step": 18946
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.283272020391685e-06,
+      "loss": 0.6569,
+      "step": 18947
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2823119160853765e-06,
+      "loss": 0.5611,
+      "step": 18948
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.281351867713229e-06,
+      "loss": 0.6034,
+      "step": 18949
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2803918752866255e-06,
+      "loss": 0.6552,
+      "step": 18950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.279431938816946e-06,
+      "loss": 0.5946,
+      "step": 18951
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.278472058315576e-06,
+      "loss": 0.6188,
+      "step": 18952
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.277512233793888e-06,
+      "loss": 0.6244,
+      "step": 18953
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2765524652632674e-06,
+      "loss": 0.6397,
+      "step": 18954
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2755927527350915e-06,
+      "loss": 0.6675,
+      "step": 18955
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.27463309622074e-06,
+      "loss": 0.6575,
+      "step": 18956
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.273673495731592e-06,
+      "loss": 0.6899,
+      "step": 18957
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.272713951279027e-06,
+      "loss": 0.6638,
+      "step": 18958
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2717544628744136e-06,
+      "loss": 0.649,
+      "step": 18959
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2707950305291326e-06,
+      "loss": 0.6651,
+      "step": 18960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2698356542545595e-06,
+      "loss": 0.6067,
+      "step": 18961
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.268876334062068e-06,
+      "loss": 0.6348,
+      "step": 18962
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.267917069963032e-06,
+      "loss": 0.6343,
+      "step": 18963
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.266957861968829e-06,
+      "loss": 0.6142,
+      "step": 18964
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.265998710090825e-06,
+      "loss": 0.5991,
+      "step": 18965
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.265039614340395e-06,
+      "loss": 0.5973,
+      "step": 18966
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.264080574728909e-06,
+      "loss": 0.6488,
+      "step": 18967
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.263121591267739e-06,
+      "loss": 0.6177,
+      "step": 18968
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.262162663968255e-06,
+      "loss": 0.6192,
+      "step": 18969
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.26120379284183e-06,
+      "loss": 0.6106,
+      "step": 18970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.260244977899824e-06,
+      "loss": 0.6634,
+      "step": 18971
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.259286219153612e-06,
+      "loss": 0.64,
+      "step": 18972
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.258327516614557e-06,
+      "loss": 0.633,
+      "step": 18973
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.257368870294027e-06,
+      "loss": 0.6076,
+      "step": 18974
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.25641028020339e-06,
+      "loss": 0.5662,
+      "step": 18975
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.255451746354012e-06,
+      "loss": 0.6377,
+      "step": 18976
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2544932687572525e-06,
+      "loss": 0.6791,
+      "step": 18977
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.253534847424479e-06,
+      "loss": 0.5663,
+      "step": 18978
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.252576482367054e-06,
+      "loss": 0.6217,
+      "step": 18979
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.25161817359634e-06,
+      "loss": 0.6252,
+      "step": 18980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.250659921123698e-06,
+      "loss": 0.6355,
+      "step": 18981
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.249701724960496e-06,
+      "loss": 0.6298,
+      "step": 18982
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.248743585118086e-06,
+      "loss": 0.6755,
+      "step": 18983
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.24778550160783e-06,
+      "loss": 0.6335,
+      "step": 18984
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.246827474441087e-06,
+      "loss": 0.6033,
+      "step": 18985
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.245869503629219e-06,
+      "loss": 0.6952,
+      "step": 18986
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2449115891835825e-06,
+      "loss": 0.6496,
+      "step": 18987
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.243953731115536e-06,
+      "loss": 0.6879,
+      "step": 18988
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.242995929436433e-06,
+      "loss": 0.6359,
+      "step": 18989
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.24203818415763e-06,
+      "loss": 0.6451,
+      "step": 18990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.241080495290483e-06,
+      "loss": 0.5954,
+      "step": 18991
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.240122862846346e-06,
+      "loss": 0.6397,
+      "step": 18992
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.239165286836576e-06,
+      "loss": 0.632,
+      "step": 18993
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.238207767272526e-06,
+      "loss": 0.6314,
+      "step": 18994
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.237250304165543e-06,
+      "loss": 0.5481,
+      "step": 18995
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.236292897526983e-06,
+      "loss": 0.6524,
+      "step": 18996
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.235335547368197e-06,
+      "loss": 0.5878,
+      "step": 18997
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.234378253700535e-06,
+      "loss": 0.6727,
+      "step": 18998
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2334210165353474e-06,
+      "loss": 0.5854,
+      "step": 18999
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.232463835883988e-06,
+      "loss": 0.6339,
+      "step": 19000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.231506711757795e-06,
+      "loss": 0.6173,
+      "step": 19001
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2305496441681236e-06,
+      "loss": 0.5807,
+      "step": 19002
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2295926331263194e-06,
+      "loss": 0.6281,
+      "step": 19003
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.228635678643728e-06,
+      "loss": 0.6259,
+      "step": 19004
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.227678780731697e-06,
+      "loss": 0.6547,
+      "step": 19005
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.226721939401575e-06,
+      "loss": 0.6203,
+      "step": 19006
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2257651546647e-06,
+      "loss": 0.6589,
+      "step": 19007
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.224808426532417e-06,
+      "loss": 0.6365,
+      "step": 19008
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.223851755016071e-06,
+      "loss": 0.6354,
+      "step": 19009
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.222895140127007e-06,
+      "loss": 0.5961,
+      "step": 19010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.221938581876565e-06,
+      "loss": 0.5876,
+      "step": 19011
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.220982080276083e-06,
+      "loss": 0.6239,
+      "step": 19012
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.220025635336903e-06,
+      "loss": 0.6839,
+      "step": 19013
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.219069247070365e-06,
+      "loss": 0.6519,
+      "step": 19014
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.218112915487809e-06,
+      "loss": 0.6076,
+      "step": 19015
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.217156640600576e-06,
+      "loss": 0.6254,
+      "step": 19016
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.216200422420003e-06,
+      "loss": 0.5866,
+      "step": 19017
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.215244260957423e-06,
+      "loss": 0.6334,
+      "step": 19018
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.214288156224174e-06,
+      "loss": 0.6251,
+      "step": 19019
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.213332108231594e-06,
+      "loss": 0.6264,
+      "step": 19020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2123761169910155e-06,
+      "loss": 0.641,
+      "step": 19021
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.211420182513776e-06,
+      "loss": 0.6543,
+      "step": 19022
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2104643048112115e-06,
+      "loss": 0.6171,
+      "step": 19023
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.209508483894647e-06,
+      "loss": 0.6282,
+      "step": 19024
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.208552719775419e-06,
+      "loss": 0.6074,
+      "step": 19025
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.20759701246486e-06,
+      "loss": 0.6662,
+      "step": 19026
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.2066413619743e-06,
+      "loss": 0.6302,
+      "step": 19027
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.205685768315072e-06,
+      "loss": 0.6172,
+      "step": 19028
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.204730231498507e-06,
+      "loss": 0.6472,
+      "step": 19029
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.203774751535928e-06,
+      "loss": 0.6288,
+      "step": 19030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.202819328438667e-06,
+      "loss": 0.5946,
+      "step": 19031
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.20186396221805e-06,
+      "loss": 0.672,
+      "step": 19032
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.200908652885406e-06,
+      "loss": 0.6405,
+      "step": 19033
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.19995340045206e-06,
+      "loss": 0.6297,
+      "step": 19034
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.198998204929343e-06,
+      "loss": 0.6118,
+      "step": 19035
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.198043066328572e-06,
+      "loss": 0.6573,
+      "step": 19036
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.197087984661075e-06,
+      "loss": 0.6171,
+      "step": 19037
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.196132959938175e-06,
+      "loss": 0.6302,
+      "step": 19038
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.195177992171197e-06,
+      "loss": 0.6445,
+      "step": 19039
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.19422308137146e-06,
+      "loss": 0.6094,
+      "step": 19040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.193268227550292e-06,
+      "loss": 0.6063,
+      "step": 19041
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.192313430719007e-06,
+      "loss": 0.6079,
+      "step": 19042
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.191358690888928e-06,
+      "loss": 0.6543,
+      "step": 19043
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.190404008071375e-06,
+      "loss": 0.5577,
+      "step": 19044
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.189449382277665e-06,
+      "loss": 0.6546,
+      "step": 19045
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.188494813519118e-06,
+      "loss": 0.6216,
+      "step": 19046
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.187540301807057e-06,
+      "loss": 0.6464,
+      "step": 19047
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.18658584715279e-06,
+      "loss": 0.6533,
+      "step": 19048
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.185631449567634e-06,
+      "loss": 0.613,
+      "step": 19049
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.18467710906291e-06,
+      "loss": 0.6274,
+      "step": 19050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1837228256499285e-06,
+      "loss": 0.6461,
+      "step": 19051
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1827685993400065e-06,
+      "loss": 0.6888,
+      "step": 19052
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.181814430144461e-06,
+      "loss": 0.6296,
+      "step": 19053
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1808603180745966e-06,
+      "loss": 0.6505,
+      "step": 19054
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.179906263141728e-06,
+      "loss": 0.5986,
+      "step": 19055
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.178952265357169e-06,
+      "loss": 0.6359,
+      "step": 19056
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.17799832473223e-06,
+      "loss": 0.6801,
+      "step": 19057
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1770444412782185e-06,
+      "loss": 0.6945,
+      "step": 19058
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1760906150064525e-06,
+      "loss": 0.6224,
+      "step": 19059
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.17513684592823e-06,
+      "loss": 0.6724,
+      "step": 19060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1741831340548636e-06,
+      "loss": 0.546,
+      "step": 19061
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.173229479397661e-06,
+      "loss": 0.6167,
+      "step": 19062
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.172275881967928e-06,
+      "loss": 0.6503,
+      "step": 19063
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.171322341776972e-06,
+      "loss": 0.6213,
+      "step": 19064
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.170368858836101e-06,
+      "loss": 0.6283,
+      "step": 19065
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.169415433156615e-06,
+      "loss": 0.6057,
+      "step": 19066
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1684620647498175e-06,
+      "loss": 0.6094,
+      "step": 19067
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1675087536270195e-06,
+      "loss": 0.6624,
+      "step": 19068
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.16655549979951e-06,
+      "loss": 0.6058,
+      "step": 19069
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1656023032786075e-06,
+      "loss": 0.6245,
+      "step": 19070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.164649164075602e-06,
+      "loss": 0.6774,
+      "step": 19071
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.163696082201796e-06,
+      "loss": 0.608,
+      "step": 19072
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.162743057668496e-06,
+      "loss": 0.6156,
+      "step": 19073
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.161790090486991e-06,
+      "loss": 0.6478,
+      "step": 19074
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1608371806685804e-06,
+      "loss": 0.6416,
+      "step": 19075
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.159884328224576e-06,
+      "loss": 0.6423,
+      "step": 19076
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.158931533166262e-06,
+      "loss": 0.6726,
+      "step": 19077
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.157978795504939e-06,
+      "loss": 0.6501,
+      "step": 19078
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.157026115251904e-06,
+      "loss": 0.6029,
+      "step": 19079
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.156073492418448e-06,
+      "loss": 0.6264,
+      "step": 19080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.155120927015867e-06,
+      "loss": 0.615,
+      "step": 19081
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.154168419055461e-06,
+      "loss": 0.6187,
+      "step": 19082
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.153215968548515e-06,
+      "loss": 0.6377,
+      "step": 19083
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1522635755063244e-06,
+      "loss": 0.5765,
+      "step": 19084
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.151311239940186e-06,
+      "loss": 0.6029,
+      "step": 19085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.150358961861384e-06,
+      "loss": 0.6211,
+      "step": 19086
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1494067412812064e-06,
+      "loss": 0.6621,
+      "step": 19087
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.148454578210955e-06,
+      "loss": 0.6454,
+      "step": 19088
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.14750247266191e-06,
+      "loss": 0.637,
+      "step": 19089
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.146550424645359e-06,
+      "loss": 0.5911,
+      "step": 19090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.145598434172597e-06,
+      "loss": 0.5839,
+      "step": 19091
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.144646501254904e-06,
+      "loss": 0.6837,
+      "step": 19092
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.143694625903564e-06,
+      "loss": 0.6114,
+      "step": 19093
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.142742808129875e-06,
+      "loss": 0.613,
+      "step": 19094
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.141791047945111e-06,
+      "loss": 0.6116,
+      "step": 19095
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.140839345360557e-06,
+      "loss": 0.6257,
+      "step": 19096
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.139887700387507e-06,
+      "loss": 0.5938,
+      "step": 19097
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.138936113037231e-06,
+      "loss": 0.6904,
+      "step": 19098
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.137984583321014e-06,
+      "loss": 0.6233,
+      "step": 19099
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.137033111250147e-06,
+      "loss": 0.6037,
+      "step": 19100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1360816968359e-06,
+      "loss": 0.6224,
+      "step": 19101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.135130340089558e-06,
+      "loss": 0.6169,
+      "step": 19102
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.134179041022403e-06,
+      "loss": 0.6205,
+      "step": 19103
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.133227799645707e-06,
+      "loss": 0.6213,
+      "step": 19104
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.132276615970748e-06,
+      "loss": 0.5931,
+      "step": 19105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.131325490008816e-06,
+      "loss": 0.621,
+      "step": 19106
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.130374421771175e-06,
+      "loss": 0.6089,
+      "step": 19107
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.129423411269106e-06,
+      "loss": 0.711,
+      "step": 19108
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.128472458513886e-06,
+      "loss": 0.6542,
+      "step": 19109
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.127521563516784e-06,
+      "loss": 0.6422,
+      "step": 19110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.1265707262890755e-06,
+      "loss": 0.6146,
+      "step": 19111
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.125619946842042e-06,
+      "loss": 0.6424,
+      "step": 19112
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.124669225186947e-06,
+      "loss": 0.6761,
+      "step": 19113
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.123718561335065e-06,
+      "loss": 0.5805,
+      "step": 19114
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.122767955297671e-06,
+      "loss": 0.61,
+      "step": 19115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.121817407086031e-06,
+      "loss": 0.6069,
+      "step": 19116
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.120866916711412e-06,
+      "loss": 0.64,
+      "step": 19117
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.119916484185095e-06,
+      "loss": 0.6148,
+      "step": 19118
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.118966109518339e-06,
+      "loss": 0.6542,
+      "step": 19119
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.118015792722413e-06,
+      "loss": 0.6703,
+      "step": 19120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.117065533808589e-06,
+      "loss": 0.6556,
+      "step": 19121
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.116115332788127e-06,
+      "loss": 0.5817,
+      "step": 19122
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.115165189672291e-06,
+      "loss": 0.5933,
+      "step": 19123
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.11421510447236e-06,
+      "loss": 0.6283,
+      "step": 19124
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.113265077199585e-06,
+      "loss": 0.6525,
+      "step": 19125
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.112315107865235e-06,
+      "loss": 0.6165,
+      "step": 19126
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.111365196480575e-06,
+      "loss": 0.6092,
+      "step": 19127
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.110415343056862e-06,
+      "loss": 0.5979,
+      "step": 19128
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.109465547605355e-06,
+      "loss": 0.6677,
+      "step": 19129
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.10851581013733e-06,
+      "loss": 0.6472,
+      "step": 19130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.107566130664034e-06,
+      "loss": 0.6982,
+      "step": 19131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.106616509196734e-06,
+      "loss": 0.6356,
+      "step": 19132
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.10566694574668e-06,
+      "loss": 0.6501,
+      "step": 19133
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.104717440325138e-06,
+      "loss": 0.5985,
+      "step": 19134
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.10376799294336e-06,
+      "loss": 0.6496,
+      "step": 19135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.102818603612609e-06,
+      "loss": 0.6243,
+      "step": 19136
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.101869272344137e-06,
+      "loss": 0.6067,
+      "step": 19137
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.100919999149207e-06,
+      "loss": 0.6925,
+      "step": 19138
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.099970784039064e-06,
+      "loss": 0.6578,
+      "step": 19139
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0990216270249625e-06,
+      "loss": 0.6039,
+      "step": 19140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.098072528118162e-06,
+      "loss": 0.6455,
+      "step": 19141
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.097123487329912e-06,
+      "loss": 0.6737,
+      "step": 19142
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.096174504671465e-06,
+      "loss": 0.5761,
+      "step": 19143
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.095225580154077e-06,
+      "loss": 0.5588,
+      "step": 19144
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.09427671378899e-06,
+      "loss": 0.6346,
+      "step": 19145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.093327905587459e-06,
+      "loss": 0.6066,
+      "step": 19146
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.092379155560731e-06,
+      "loss": 0.6498,
+      "step": 19147
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.091430463720059e-06,
+      "loss": 0.6168,
+      "step": 19148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0904818300766854e-06,
+      "loss": 0.6242,
+      "step": 19149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.089533254641865e-06,
+      "loss": 0.6036,
+      "step": 19150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.088584737426837e-06,
+      "loss": 0.6843,
+      "step": 19151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.087636278442849e-06,
+      "loss": 0.6361,
+      "step": 19152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0866878777011475e-06,
+      "loss": 0.6024,
+      "step": 19153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0857395352129765e-06,
+      "loss": 0.6642,
+      "step": 19154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0847912509895795e-06,
+      "loss": 0.6216,
+      "step": 19155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0838430250422035e-06,
+      "loss": 0.5942,
+      "step": 19156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.082894857382085e-06,
+      "loss": 0.6165,
+      "step": 19157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.081946748020466e-06,
+      "loss": 0.6865,
+      "step": 19158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.080998696968591e-06,
+      "loss": 0.6314,
+      "step": 19159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0800507042377e-06,
+      "loss": 0.659,
+      "step": 19160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.079102769839031e-06,
+      "loss": 0.6437,
+      "step": 19161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.078154893783826e-06,
+      "loss": 0.6234,
+      "step": 19162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0772070760833185e-06,
+      "loss": 0.5831,
+      "step": 19163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0762593167487486e-06,
+      "loss": 0.6059,
+      "step": 19164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.075311615791352e-06,
+      "loss": 0.6552,
+      "step": 19165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.074363973222366e-06,
+      "loss": 0.6213,
+      "step": 19166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.073416389053027e-06,
+      "loss": 0.6703,
+      "step": 19167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.072468863294572e-06,
+      "loss": 0.6752,
+      "step": 19168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0715213959582275e-06,
+      "loss": 0.6402,
+      "step": 19169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.070573987055234e-06,
+      "loss": 0.6414,
+      "step": 19170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.069626636596818e-06,
+      "loss": 0.639,
+      "step": 19171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.068679344594218e-06,
+      "loss": 0.6375,
+      "step": 19172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.067732111058662e-06,
+      "loss": 0.5936,
+      "step": 19173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.066784936001384e-06,
+      "loss": 0.6423,
+      "step": 19174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.065837819433608e-06,
+      "loss": 0.6663,
+      "step": 19175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.064890761366567e-06,
+      "loss": 0.6053,
+      "step": 19176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.063943761811489e-06,
+      "loss": 0.583,
+      "step": 19177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0629968207796e-06,
+      "loss": 0.6396,
+      "step": 19178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.062049938282129e-06,
+      "loss": 0.6438,
+      "step": 19179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.061103114330309e-06,
+      "loss": 0.6191,
+      "step": 19180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.060156348935355e-06,
+      "loss": 0.6302,
+      "step": 19181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0592096421084946e-06,
+      "loss": 0.671,
+      "step": 19182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.058262993860955e-06,
+      "loss": 0.6244,
+      "step": 19183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.057316404203958e-06,
+      "loss": 0.6807,
+      "step": 19184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.0563698731487295e-06,
+      "loss": 0.6167,
+      "step": 19185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.055423400706492e-06,
+      "loss": 0.6523,
+      "step": 19186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.054476986888461e-06,
+      "loss": 0.5998,
+      "step": 19187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.053530631705863e-06,
+      "loss": 0.6089,
+      "step": 19188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.052584335169917e-06,
+      "loss": 0.6687,
+      "step": 19189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.05163809729184e-06,
+      "loss": 0.6538,
+      "step": 19190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.050691918082855e-06,
+      "loss": 0.6367,
+      "step": 19191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.049745797554181e-06,
+      "loss": 0.6324,
+      "step": 19192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.048799735717029e-06,
+      "loss": 0.6676,
+      "step": 19193
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.04785373258262e-06,
+      "loss": 0.6514,
+      "step": 19194
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.046907788162168e-06,
+      "loss": 0.6549,
+      "step": 19195
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.045961902466892e-06,
+      "loss": 0.6482,
+      "step": 19196
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.045016075508007e-06,
+      "loss": 0.6352,
+      "step": 19197
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.04407030729672e-06,
+      "loss": 0.642,
+      "step": 19198
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.04312459784425e-06,
+      "loss": 0.6127,
+      "step": 19199
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.042178947161808e-06,
+      "loss": 0.6205,
+      "step": 19200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.041233355260607e-06,
+      "loss": 0.6248,
+      "step": 19201
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.040287822151856e-06,
+      "loss": 0.6189,
+      "step": 19202
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.039342347846772e-06,
+      "loss": 0.6298,
+      "step": 19203
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.038396932356555e-06,
+      "loss": 0.6411,
+      "step": 19204
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.037451575692419e-06,
+      "loss": 0.6134,
+      "step": 19205
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0365062778655726e-06,
+      "loss": 0.6618,
+      "step": 19206
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.035561038887223e-06,
+      "loss": 0.6323,
+      "step": 19207
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.034615858768578e-06,
+      "loss": 0.5962,
+      "step": 19208
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.033670737520846e-06,
+      "loss": 0.5988,
+      "step": 19209
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.032725675155227e-06,
+      "loss": 0.6799,
+      "step": 19210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.031780671682928e-06,
+      "loss": 0.6712,
+      "step": 19211
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.030835727115153e-06,
+      "loss": 0.6234,
+      "step": 19212
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0298908414631085e-06,
+      "loss": 0.691,
+      "step": 19213
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.028946014737994e-06,
+      "loss": 0.6253,
+      "step": 19214
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.028001246951018e-06,
+      "loss": 0.6192,
+      "step": 19215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.027056538113372e-06,
+      "loss": 0.6296,
+      "step": 19216
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.026111888236261e-06,
+      "loss": 0.6989,
+      "step": 19217
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.025167297330887e-06,
+      "loss": 0.6401,
+      "step": 19218
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0242227654084465e-06,
+      "loss": 0.6598,
+      "step": 19219
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.02327829248014e-06,
+      "loss": 0.6414,
+      "step": 19220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.022333878557169e-06,
+      "loss": 0.6226,
+      "step": 19221
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.021389523650724e-06,
+      "loss": 0.6184,
+      "step": 19222
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.020445227772002e-06,
+      "loss": 0.6273,
+      "step": 19223
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0195009909322025e-06,
+      "loss": 0.63,
+      "step": 19224
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.018556813142519e-06,
+      "loss": 0.5806,
+      "step": 19225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.017612694414145e-06,
+      "loss": 0.6209,
+      "step": 19226
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.016668634758279e-06,
+      "loss": 0.677,
+      "step": 19227
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.015724634186108e-06,
+      "loss": 0.5977,
+      "step": 19228
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.014780692708825e-06,
+      "loss": 0.6397,
+      "step": 19229
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0138368103376245e-06,
+      "loss": 0.625,
+      "step": 19230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.012892987083695e-06,
+      "loss": 0.5872,
+      "step": 19231
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.011949222958229e-06,
+      "loss": 0.6559,
+      "step": 19232
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.011005517972418e-06,
+      "loss": 0.6464,
+      "step": 19233
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.010061872137444e-06,
+      "loss": 0.58,
+      "step": 19234
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.009118285464498e-06,
+      "loss": 0.6166,
+      "step": 19235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.00817475796477e-06,
+      "loss": 0.6067,
+      "step": 19236
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0072312896494434e-06,
+      "loss": 0.68,
+      "step": 19237
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.006287880529708e-06,
+      "loss": 0.5958,
+      "step": 19238
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.005344530616748e-06,
+      "loss": 0.6749,
+      "step": 19239
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.004401239921744e-06,
+      "loss": 0.5367,
+      "step": 19240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.003458008455884e-06,
+      "loss": 0.6413,
+      "step": 19241
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.002514836230348e-06,
+      "loss": 0.6252,
+      "step": 19242
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.001571723256321e-06,
+      "loss": 0.6464,
+      "step": 19243
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.000628669544985e-06,
+      "loss": 0.6425,
+      "step": 19244
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.999685675107523e-06,
+      "loss": 0.6155,
+      "step": 19245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.99874273995511e-06,
+      "loss": 0.5693,
+      "step": 19246
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.997799864098928e-06,
+      "loss": 0.6719,
+      "step": 19247
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9968570475501565e-06,
+      "loss": 0.6817,
+      "step": 19248
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.995914290319974e-06,
+      "loss": 0.6332,
+      "step": 19249
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.994971592419557e-06,
+      "loss": 0.5901,
+      "step": 19250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.994028953860087e-06,
+      "loss": 0.6277,
+      "step": 19251
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.993086374652733e-06,
+      "loss": 0.6204,
+      "step": 19252
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.992143854808674e-06,
+      "loss": 0.6126,
+      "step": 19253
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.991201394339085e-06,
+      "loss": 0.5891,
+      "step": 19254
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.990258993255139e-06,
+      "loss": 0.6389,
+      "step": 19255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.989316651568014e-06,
+      "loss": 0.6813,
+      "step": 19256
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.988374369288874e-06,
+      "loss": 0.6609,
+      "step": 19257
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.987432146428896e-06,
+      "loss": 0.6399,
+      "step": 19258
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.986489982999255e-06,
+      "loss": 0.5839,
+      "step": 19259
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.98554787901111e-06,
+      "loss": 0.6154,
+      "step": 19260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.984605834475642e-06,
+      "loss": 0.6745,
+      "step": 19261
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.98366384940402e-06,
+      "loss": 0.6782,
+      "step": 19262
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.982721923807404e-06,
+      "loss": 0.6637,
+      "step": 19263
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.981780057696968e-06,
+      "loss": 0.6315,
+      "step": 19264
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.980838251083881e-06,
+      "loss": 0.6353,
+      "step": 19265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.979896503979297e-06,
+      "loss": 0.6373,
+      "step": 19266
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.978954816394394e-06,
+      "loss": 0.6287,
+      "step": 19267
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.978013188340339e-06,
+      "loss": 0.6501,
+      "step": 19268
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.977071619828286e-06,
+      "loss": 0.641,
+      "step": 19269
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.976130110869403e-06,
+      "loss": 0.6395,
+      "step": 19270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.975188661474856e-06,
+      "loss": 0.5658,
+      "step": 19271
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.974247271655796e-06,
+      "loss": 0.6616,
+      "step": 19272
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.973305941423398e-06,
+      "loss": 0.6532,
+      "step": 19273
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.972364670788818e-06,
+      "loss": 0.6821,
+      "step": 19274
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.971423459763213e-06,
+      "loss": 0.6254,
+      "step": 19275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.970482308357743e-06,
+      "loss": 0.6189,
+      "step": 19276
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.969541216583572e-06,
+      "loss": 0.591,
+      "step": 19277
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.968600184451845e-06,
+      "loss": 0.6011,
+      "step": 19278
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.967659211973732e-06,
+      "loss": 0.5782,
+      "step": 19279
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.966718299160389e-06,
+      "loss": 0.6399,
+      "step": 19280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.965777446022964e-06,
+      "loss": 0.6275,
+      "step": 19281
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.964836652572615e-06,
+      "loss": 0.6391,
+      "step": 19282
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.963895918820502e-06,
+      "loss": 0.6591,
+      "step": 19283
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.962955244777764e-06,
+      "loss": 0.587,
+      "step": 19284
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.962014630455568e-06,
+      "loss": 0.5803,
+      "step": 19285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.961074075865067e-06,
+      "loss": 0.6391,
+      "step": 19286
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.960133581017401e-06,
+      "loss": 0.6147,
+      "step": 19287
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.959193145923728e-06,
+      "loss": 0.6534,
+      "step": 19288
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.958252770595201e-06,
+      "loss": 0.6339,
+      "step": 19289
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.957312455042957e-06,
+      "loss": 0.6479,
+      "step": 19290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.956372199278157e-06,
+      "loss": 0.6513,
+      "step": 19291
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.95543200331195e-06,
+      "loss": 0.64,
+      "step": 19292
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.954491867155472e-06,
+      "loss": 0.6182,
+      "step": 19293
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.953551790819876e-06,
+      "loss": 0.635,
+      "step": 19294
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.952611774316311e-06,
+      "loss": 0.6117,
+      "step": 19295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.951671817655912e-06,
+      "loss": 0.6152,
+      "step": 19296
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.950731920849833e-06,
+      "loss": 0.6099,
+      "step": 19297
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9497920839092185e-06,
+      "loss": 0.6175,
+      "step": 19298
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.948852306845202e-06,
+      "loss": 0.5533,
+      "step": 19299
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.947912589668933e-06,
+      "loss": 0.6294,
+      "step": 19300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9469729323915536e-06,
+      "loss": 0.6214,
+      "step": 19301
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.946033335024196e-06,
+      "loss": 0.6487,
+      "step": 19302
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.945093797578009e-06,
+      "loss": 0.6442,
+      "step": 19303
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.944154320064134e-06,
+      "loss": 0.6493,
+      "step": 19304
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.943214902493701e-06,
+      "loss": 0.62,
+      "step": 19305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.942275544877853e-06,
+      "loss": 0.6142,
+      "step": 19306
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.94133624722773e-06,
+      "loss": 0.6313,
+      "step": 19307
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.940397009554457e-06,
+      "loss": 0.655,
+      "step": 19308
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.939457831869182e-06,
+      "loss": 0.6531,
+      "step": 19309
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.938518714183039e-06,
+      "loss": 0.5914,
+      "step": 19310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9375796565071574e-06,
+      "loss": 0.6212,
+      "step": 19311
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.936640658852672e-06,
+      "loss": 0.6533,
+      "step": 19312
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.93570172123072e-06,
+      "loss": 0.6513,
+      "step": 19313
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9347628436524235e-06,
+      "loss": 0.5454,
+      "step": 19314
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.933824026128925e-06,
+      "loss": 0.6414,
+      "step": 19315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.932885268671354e-06,
+      "loss": 0.6441,
+      "step": 19316
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.931946571290835e-06,
+      "loss": 0.6649,
+      "step": 19317
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.931007933998505e-06,
+      "loss": 0.5917,
+      "step": 19318
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.930069356805485e-06,
+      "loss": 0.5911,
+      "step": 19319
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.929130839722901e-06,
+      "loss": 0.6994,
+      "step": 19320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.928192382761894e-06,
+      "loss": 0.6292,
+      "step": 19321
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.927253985933577e-06,
+      "loss": 0.6741,
+      "step": 19322
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.926315649249083e-06,
+      "loss": 0.6793,
+      "step": 19323
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9253773727195375e-06,
+      "loss": 0.5943,
+      "step": 19324
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.924439156356059e-06,
+      "loss": 0.6124,
+      "step": 19325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.923501000169772e-06,
+      "loss": 0.5871,
+      "step": 19326
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.922562904171809e-06,
+      "loss": 0.6482,
+      "step": 19327
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.921624868373284e-06,
+      "loss": 0.6606,
+      "step": 19328
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.920686892785318e-06,
+      "loss": 0.6444,
+      "step": 19329
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.919748977419039e-06,
+      "loss": 0.6351,
+      "step": 19330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.918811122285558e-06,
+      "loss": 0.5547,
+      "step": 19331
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.917873327395997e-06,
+      "loss": 0.6513,
+      "step": 19332
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9169355927614824e-06,
+      "loss": 0.5924,
+      "step": 19333
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.915997918393123e-06,
+      "loss": 0.6786,
+      "step": 19334
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.91506030430204e-06,
+      "loss": 0.5969,
+      "step": 19335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.914122750499353e-06,
+      "loss": 0.6334,
+      "step": 19336
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9131852569961714e-06,
+      "loss": 0.6848,
+      "step": 19337
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.912247823803609e-06,
+      "loss": 0.5924,
+      "step": 19338
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.911310450932791e-06,
+      "loss": 0.6107,
+      "step": 19339
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.910373138394823e-06,
+      "loss": 0.6173,
+      "step": 19340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9094358862008196e-06,
+      "loss": 0.6105,
+      "step": 19341
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.908498694361896e-06,
+      "loss": 0.6396,
+      "step": 19342
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.907561562889158e-06,
+      "loss": 0.6631,
+      "step": 19343
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.906624491793717e-06,
+      "loss": 0.6442,
+      "step": 19344
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.905687481086691e-06,
+      "loss": 0.6169,
+      "step": 19345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.904750530779183e-06,
+      "loss": 0.6811,
+      "step": 19346
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.903813640882301e-06,
+      "loss": 0.5898,
+      "step": 19347
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.90287681140716e-06,
+      "loss": 0.6129,
+      "step": 19348
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9019400423648575e-06,
+      "loss": 0.6023,
+      "step": 19349
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.901003333766502e-06,
+      "loss": 0.6175,
+      "step": 19350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.9000666856232105e-06,
+      "loss": 0.6379,
+      "step": 19351
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8991300979460744e-06,
+      "loss": 0.6515,
+      "step": 19352
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.898193570746205e-06,
+      "loss": 0.6148,
+      "step": 19353
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.897257104034707e-06,
+      "loss": 0.6397,
+      "step": 19354
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.896320697822678e-06,
+      "loss": 0.6171,
+      "step": 19355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8953843521212195e-06,
+      "loss": 0.6485,
+      "step": 19356
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8944480669414455e-06,
+      "loss": 0.6244,
+      "step": 19357
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.893511842294442e-06,
+      "loss": 0.6469,
+      "step": 19358
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.892575678191317e-06,
+      "loss": 0.6035,
+      "step": 19359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.891639574643172e-06,
+      "loss": 0.6654,
+      "step": 19360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.890703531661099e-06,
+      "loss": 0.6345,
+      "step": 19361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.889767549256194e-06,
+      "loss": 0.6297,
+      "step": 19362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.888831627439566e-06,
+      "loss": 0.5963,
+      "step": 19363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8878957662223024e-06,
+      "loss": 0.6114,
+      "step": 19364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8869599656155e-06,
+      "loss": 0.6116,
+      "step": 19365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.886024225630261e-06,
+      "loss": 0.6402,
+      "step": 19366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.885088546277667e-06,
+      "loss": 0.6106,
+      "step": 19367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.884152927568818e-06,
+      "loss": 0.5858,
+      "step": 19368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.883217369514815e-06,
+      "loss": 0.6748,
+      "step": 19369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8822818721267374e-06,
+      "loss": 0.6087,
+      "step": 19370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8813464354156824e-06,
+      "loss": 0.6264,
+      "step": 19371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.880411059392746e-06,
+      "loss": 0.6589,
+      "step": 19372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.879475744069007e-06,
+      "loss": 0.6075,
+      "step": 19373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.878540489455558e-06,
+      "loss": 0.6823,
+      "step": 19374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.877605295563497e-06,
+      "loss": 0.6425,
+      "step": 19375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.876670162403901e-06,
+      "loss": 0.6755,
+      "step": 19376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.875735089987863e-06,
+      "loss": 0.6052,
+      "step": 19377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.87480007832647e-06,
+      "loss": 0.6605,
+      "step": 19378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.873865127430802e-06,
+      "loss": 0.6251,
+      "step": 19379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.872930237311948e-06,
+      "loss": 0.6196,
+      "step": 19380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.871995407980991e-06,
+      "loss": 0.5773,
+      "step": 19381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8710606394490154e-06,
+      "loss": 0.6771,
+      "step": 19382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.870125931727109e-06,
+      "loss": 0.6854,
+      "step": 19383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.869191284826344e-06,
+      "loss": 0.6306,
+      "step": 19384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.868256698757808e-06,
+      "loss": 0.6052,
+      "step": 19385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.867322173532579e-06,
+      "loss": 0.6094,
+      "step": 19386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.866387709161741e-06,
+      "loss": 0.6063,
+      "step": 19387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.865453305656368e-06,
+      "loss": 0.6035,
+      "step": 19388
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.864518963027547e-06,
+      "loss": 0.656,
+      "step": 19389
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8635846812863455e-06,
+      "loss": 0.5892,
+      "step": 19390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.862650460443845e-06,
+      "loss": 0.6641,
+      "step": 19391
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.861716300511122e-06,
+      "loss": 0.673,
+      "step": 19392
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8607822014992525e-06,
+      "loss": 0.695,
+      "step": 19393
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.859848163419312e-06,
+      "loss": 0.6501,
+      "step": 19394
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.858914186282376e-06,
+      "loss": 0.6556,
+      "step": 19395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8579802700995125e-06,
+      "loss": 0.6273,
+      "step": 19396
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.857046414881799e-06,
+      "loss": 0.635,
+      "step": 19397
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.856112620640304e-06,
+      "loss": 0.6786,
+      "step": 19398
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.855178887386103e-06,
+      "loss": 0.6283,
+      "step": 19399
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.854245215130263e-06,
+      "loss": 0.6198,
+      "step": 19400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.85331160388386e-06,
+      "loss": 0.6392,
+      "step": 19401
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.852378053657954e-06,
+      "loss": 0.6317,
+      "step": 19402
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8514445644636185e-06,
+      "loss": 0.6212,
+      "step": 19403
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.850511136311921e-06,
+      "loss": 0.632,
+      "step": 19404
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.849577769213928e-06,
+      "loss": 0.6245,
+      "step": 19405
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.848644463180705e-06,
+      "loss": 0.5849,
+      "step": 19406
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.847711218223324e-06,
+      "loss": 0.5918,
+      "step": 19407
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.846778034352838e-06,
+      "loss": 0.6555,
+      "step": 19408
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.845844911580319e-06,
+      "loss": 0.6679,
+      "step": 19409
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.844911849916827e-06,
+      "loss": 0.6478,
+      "step": 19410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8439788493734276e-06,
+      "loss": 0.6816,
+      "step": 19411
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.84304590996118e-06,
+      "loss": 0.6188,
+      "step": 19412
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.842113031691153e-06,
+      "loss": 0.6356,
+      "step": 19413
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.841180214574395e-06,
+      "loss": 0.6516,
+      "step": 19414
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.840247458621972e-06,
+      "loss": 0.6137,
+      "step": 19415
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.839314763844942e-06,
+      "loss": 0.5905,
+      "step": 19416
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.838382130254364e-06,
+      "loss": 0.6305,
+      "step": 19417
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.837449557861296e-06,
+      "loss": 0.6482,
+      "step": 19418
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8365170466767975e-06,
+      "loss": 0.6598,
+      "step": 19419
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.835584596711917e-06,
+      "loss": 0.6363,
+      "step": 19420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8346522079777145e-06,
+      "loss": 0.6448,
+      "step": 19421
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.833719880485245e-06,
+      "loss": 0.6331,
+      "step": 19422
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.832787614245561e-06,
+      "loss": 0.6406,
+      "step": 19423
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.831855409269717e-06,
+      "loss": 0.6214,
+      "step": 19424
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.830923265568768e-06,
+      "loss": 0.6824,
+      "step": 19425
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.829991183153758e-06,
+      "loss": 0.6461,
+      "step": 19426
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.829059162035744e-06,
+      "loss": 0.6637,
+      "step": 19427
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.828127202225774e-06,
+      "loss": 0.6525,
+      "step": 19428
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8271953037349e-06,
+      "loss": 0.6152,
+      "step": 19429
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.826263466574167e-06,
+      "loss": 0.6346,
+      "step": 19430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.825331690754631e-06,
+      "loss": 0.6872,
+      "step": 19431
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8243999762873295e-06,
+      "loss": 0.6111,
+      "step": 19432
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.823468323183313e-06,
+      "loss": 0.6134,
+      "step": 19433
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.822536731453629e-06,
+      "loss": 0.6091,
+      "step": 19434
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.821605201109319e-06,
+      "loss": 0.6001,
+      "step": 19435
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.820673732161433e-06,
+      "loss": 0.5851,
+      "step": 19436
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.819742324621014e-06,
+      "loss": 0.6648,
+      "step": 19437
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8188109784991e-06,
+      "loss": 0.6632,
+      "step": 19438
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.817879693806735e-06,
+      "loss": 0.6638,
+      "step": 19439
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.816948470554963e-06,
+      "loss": 0.6385,
+      "step": 19440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.816017308754822e-06,
+      "loss": 0.7217,
+      "step": 19441
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8150862084173575e-06,
+      "loss": 0.6706,
+      "step": 19442
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.814155169553603e-06,
+      "loss": 0.6391,
+      "step": 19443
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.813224192174597e-06,
+      "loss": 0.6848,
+      "step": 19444
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.81229327629138e-06,
+      "loss": 0.624,
+      "step": 19445
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.811362421914989e-06,
+      "loss": 0.6535,
+      "step": 19446
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.81043162905646e-06,
+      "loss": 0.6122,
+      "step": 19447
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8095008977268335e-06,
+      "loss": 0.6486,
+      "step": 19448
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.808570227937135e-06,
+      "loss": 0.6337,
+      "step": 19449
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.807639619698404e-06,
+      "loss": 0.6367,
+      "step": 19450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.806709073021673e-06,
+      "loss": 0.6371,
+      "step": 19451
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.805778587917977e-06,
+      "loss": 0.6276,
+      "step": 19452
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.804848164398346e-06,
+      "loss": 0.6386,
+      "step": 19453
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.803917802473816e-06,
+      "loss": 0.6395,
+      "step": 19454
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.802987502155409e-06,
+      "loss": 0.6004,
+      "step": 19455
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.8020572634541586e-06,
+      "loss": 0.6796,
+      "step": 19456
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.801127086381097e-06,
+      "loss": 0.6424,
+      "step": 19457
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.800196970947249e-06,
+      "loss": 0.628,
+      "step": 19458
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.799266917163643e-06,
+      "loss": 0.6695,
+      "step": 19459
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.798336925041312e-06,
+      "loss": 0.6071,
+      "step": 19460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.797406994591272e-06,
+      "loss": 0.6021,
+      "step": 19461
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.796477125824553e-06,
+      "loss": 0.6304,
+      "step": 19462
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.795547318752181e-06,
+      "loss": 0.6068,
+      "step": 19463
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.79461757338518e-06,
+      "loss": 0.6089,
+      "step": 19464
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.7936878897345715e-06,
+      "loss": 0.6877,
+      "step": 19465
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.792758267811383e-06,
+      "loss": 0.6469,
+      "step": 19466
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.791828707626629e-06,
+      "loss": 0.6421,
+      "step": 19467
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.790899209191333e-06,
+      "loss": 0.6136,
+      "step": 19468
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.789969772516517e-06,
+      "loss": 0.6362,
+      "step": 19469
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.7890403976132e-06,
+      "loss": 0.6172,
+      "step": 19470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.788111084492401e-06,
+      "loss": 0.6018,
+      "step": 19471
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.787181833165143e-06,
+      "loss": 0.6383,
+      "step": 19472
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.786252643642434e-06,
+      "loss": 0.6389,
+      "step": 19473
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.785323515935296e-06,
+      "loss": 0.5998,
+      "step": 19474
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.784394450054743e-06,
+      "loss": 0.6142,
+      "step": 19475
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.783465446011793e-06,
+      "loss": 0.5801,
+      "step": 19476
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.7825365038174574e-06,
+      "loss": 0.6306,
+      "step": 19477
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.781607623482757e-06,
+      "loss": 0.5811,
+      "step": 19478
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.780678805018695e-06,
+      "loss": 0.6394,
+      "step": 19479
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.779750048436288e-06,
+      "loss": 0.6685,
+      "step": 19480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.778821353746549e-06,
+      "loss": 0.6404,
+      "step": 19481
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.7778927209604865e-06,
+      "loss": 0.6231,
+      "step": 19482
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.776964150089112e-06,
+      "loss": 0.6183,
+      "step": 19483
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.776035641143438e-06,
+      "loss": 0.6355,
+      "step": 19484
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775107194134467e-06,
+      "loss": 0.622,
+      "step": 19485
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7741788090732095e-06,
+      "loss": 0.6156,
+      "step": 19486
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.773250485970673e-06,
+      "loss": 0.6826,
+      "step": 19487
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772322224837864e-06,
+      "loss": 0.625,
+      "step": 19488
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.771394025685787e-06,
+      "loss": 0.6357,
+      "step": 19489
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.770465888525453e-06,
+      "loss": 0.6216,
+      "step": 19490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769537813367857e-06,
+      "loss": 0.6377,
+      "step": 19491
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.768609800224007e-06,
+      "loss": 0.5962,
+      "step": 19492
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.767681849104905e-06,
+      "loss": 0.5874,
+      "step": 19493
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.766753960021554e-06,
+      "loss": 0.6322,
+      "step": 19494
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.765826132984955e-06,
+      "loss": 0.6271,
+      "step": 19495
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.764898368006112e-06,
+      "loss": 0.6619,
+      "step": 19496
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.763970665096018e-06,
+      "loss": 0.6039,
+      "step": 19497
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7630430242656755e-06,
+      "loss": 0.66,
+      "step": 19498
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.762115445526082e-06,
+      "loss": 0.6471,
+      "step": 19499
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.761187928888236e-06,
+      "loss": 0.6473,
+      "step": 19500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.760260474363134e-06,
+      "loss": 0.6022,
+      "step": 19501
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.759333081961777e-06,
+      "loss": 0.5932,
+      "step": 19502
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.758405751695151e-06,
+      "loss": 0.6279,
+      "step": 19503
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.75747848357426e-06,
+      "loss": 0.6428,
+      "step": 19504
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.756551277610085e-06,
+      "loss": 0.6154,
+      "step": 19505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.755624133813634e-06,
+      "loss": 0.6223,
+      "step": 19506
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.754697052195894e-06,
+      "loss": 0.6144,
+      "step": 19507
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.753770032767853e-06,
+      "loss": 0.6472,
+      "step": 19508
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7528430755405046e-06,
+      "loss": 0.6316,
+      "step": 19509
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.751916180524843e-06,
+      "loss": 0.5937,
+      "step": 19510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.750989347731846e-06,
+      "loss": 0.6516,
+      "step": 19511
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.750062577172514e-06,
+      "loss": 0.6622,
+      "step": 19512
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.749135868857836e-06,
+      "loss": 0.6623,
+      "step": 19513
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7482092227987905e-06,
+      "loss": 0.6316,
+      "step": 19514
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7472826390063685e-06,
+      "loss": 0.6417,
+      "step": 19515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7463561174915586e-06,
+      "loss": 0.6427,
+      "step": 19516
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.745429658265336e-06,
+      "loss": 0.6765,
+      "step": 19517
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.744503261338695e-06,
+      "loss": 0.6019,
+      "step": 19518
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.743576926722621e-06,
+      "loss": 0.6152,
+      "step": 19519
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.742650654428087e-06,
+      "loss": 0.6028,
+      "step": 19520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.741724444466081e-06,
+      "loss": 0.6107,
+      "step": 19521
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.740798296847586e-06,
+      "loss": 0.6307,
+      "step": 19522
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.739872211583573e-06,
+      "loss": 0.5874,
+      "step": 19523
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.738946188685033e-06,
+      "loss": 0.6147,
+      "step": 19524
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.738020228162945e-06,
+      "loss": 0.6495,
+      "step": 19525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.73709433002828e-06,
+      "loss": 0.6137,
+      "step": 19526
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.73616849429202e-06,
+      "loss": 0.6134,
+      "step": 19527
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.735242720965144e-06,
+      "loss": 0.6552,
+      "step": 19528
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7343170100586174e-06,
+      "loss": 0.6548,
+      "step": 19529
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.733391361583429e-06,
+      "loss": 0.6438,
+      "step": 19530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.732465775550551e-06,
+      "loss": 0.5927,
+      "step": 19531
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.731540251970952e-06,
+      "loss": 0.7165,
+      "step": 19532
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.730614790855606e-06,
+      "loss": 0.6395,
+      "step": 19533
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.729689392215493e-06,
+      "loss": 0.6199,
+      "step": 19534
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.728764056061572e-06,
+      "loss": 0.6587,
+      "step": 19535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.727838782404824e-06,
+      "loss": 0.6677,
+      "step": 19536
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7269135712562205e-06,
+      "loss": 0.6736,
+      "step": 19537
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.725988422626724e-06,
+      "loss": 0.6387,
+      "step": 19538
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.725063336527306e-06,
+      "loss": 0.6184,
+      "step": 19539
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.724138312968937e-06,
+      "loss": 0.5866,
+      "step": 19540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7232133519625765e-06,
+      "loss": 0.6974,
+      "step": 19541
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7222884535192016e-06,
+      "loss": 0.5929,
+      "step": 19542
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.721363617649776e-06,
+      "loss": 0.624,
+      "step": 19543
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.720438844365258e-06,
+      "loss": 0.6369,
+      "step": 19544
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.719514133676617e-06,
+      "loss": 0.6119,
+      "step": 19545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.718589485594819e-06,
+      "loss": 0.6953,
+      "step": 19546
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7176649001308165e-06,
+      "loss": 0.6061,
+      "step": 19547
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.716740377295582e-06,
+      "loss": 0.5974,
+      "step": 19548
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7158159171000785e-06,
+      "loss": 0.6825,
+      "step": 19549
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.714891519555257e-06,
+      "loss": 0.5959,
+      "step": 19550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.713967184672084e-06,
+      "loss": 0.6327,
+      "step": 19551
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.713042912461519e-06,
+      "loss": 0.6465,
+      "step": 19552
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.712118702934511e-06,
+      "loss": 0.6757,
+      "step": 19553
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7111945561020286e-06,
+      "loss": 0.5604,
+      "step": 19554
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7102704719750295e-06,
+      "loss": 0.6149,
+      "step": 19555
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.709346450564462e-06,
+      "loss": 0.6541,
+      "step": 19556
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.708422491881284e-06,
+      "loss": 0.6501,
+      "step": 19557
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7074985959364546e-06,
+      "loss": 0.5633,
+      "step": 19558
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.706574762740919e-06,
+      "loss": 0.6042,
+      "step": 19559
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.705650992305637e-06,
+      "loss": 0.6907,
+      "step": 19560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.704727284641565e-06,
+      "loss": 0.6579,
+      "step": 19561
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.703803639759646e-06,
+      "loss": 0.6295,
+      "step": 19562
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.702880057670834e-06,
+      "loss": 0.6738,
+      "step": 19563
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.701956538386084e-06,
+      "loss": 0.6266,
+      "step": 19564
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.701033081916334e-06,
+      "loss": 0.6291,
+      "step": 19565
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.700109688272547e-06,
+      "loss": 0.6434,
+      "step": 19566
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.699186357465661e-06,
+      "loss": 0.6349,
+      "step": 19567
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.698263089506625e-06,
+      "loss": 0.6336,
+      "step": 19568
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.697339884406392e-06,
+      "loss": 0.6435,
+      "step": 19569
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.696416742175898e-06,
+      "loss": 0.6314,
+      "step": 19570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6954936628260895e-06,
+      "loss": 0.6825,
+      "step": 19571
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.694570646367921e-06,
+      "loss": 0.6288,
+      "step": 19572
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.693647692812324e-06,
+      "loss": 0.6296,
+      "step": 19573
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.692724802170248e-06,
+      "loss": 0.6872,
+      "step": 19574
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.691801974452635e-06,
+      "loss": 0.6296,
+      "step": 19575
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.690879209670423e-06,
+      "loss": 0.6004,
+      "step": 19576
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.689956507834548e-06,
+      "loss": 0.6537,
+      "step": 19577
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6890338689559656e-06,
+      "loss": 0.6243,
+      "step": 19578
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.688111293045601e-06,
+      "loss": 0.6239,
+      "step": 19579
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.687188780114395e-06,
+      "loss": 0.5997,
+      "step": 19580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.686266330173291e-06,
+      "loss": 0.6447,
+      "step": 19581
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.685343943233219e-06,
+      "loss": 0.6356,
+      "step": 19582
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6844216193051124e-06,
+      "loss": 0.6351,
+      "step": 19583
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.68349935839992e-06,
+      "loss": 0.6207,
+      "step": 19584
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.682577160528564e-06,
+      "loss": 0.6562,
+      "step": 19585
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.681655025701983e-06,
+      "loss": 0.6319,
+      "step": 19586
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.680732953931113e-06,
+      "loss": 0.6071,
+      "step": 19587
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.679810945226879e-06,
+      "loss": 0.6687,
+      "step": 19588
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.678888999600212e-06,
+      "loss": 0.6415,
+      "step": 19589
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.677967117062055e-06,
+      "loss": 0.6246,
+      "step": 19590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.677045297623327e-06,
+      "loss": 0.6536,
+      "step": 19591
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.67612354129496e-06,
+      "loss": 0.6288,
+      "step": 19592
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.675201848087887e-06,
+      "loss": 0.601,
+      "step": 19593
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.674280218013028e-06,
+      "loss": 0.6191,
+      "step": 19594
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.67335865108131e-06,
+      "loss": 0.6456,
+      "step": 19595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.672437147303672e-06,
+      "loss": 0.7093,
+      "step": 19596
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6715157066910265e-06,
+      "loss": 0.5736,
+      "step": 19597
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.670594329254303e-06,
+      "loss": 0.6074,
+      "step": 19598
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.669673015004429e-06,
+      "loss": 0.6311,
+      "step": 19599
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.668751763952321e-06,
+      "loss": 0.6388,
+      "step": 19600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.667830576108902e-06,
+      "loss": 0.644,
+      "step": 19601
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.666909451485102e-06,
+      "loss": 0.6249,
+      "step": 19602
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.665988390091835e-06,
+      "loss": 0.6606,
+      "step": 19603
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.665067391940022e-06,
+      "loss": 0.6468,
+      "step": 19604
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.664146457040588e-06,
+      "loss": 0.6827,
+      "step": 19605
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.663225585404445e-06,
+      "loss": 0.6938,
+      "step": 19606
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.662304777042513e-06,
+      "loss": 0.632,
+      "step": 19607
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.661384031965709e-06,
+      "loss": 0.6447,
+      "step": 19608
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.660463350184951e-06,
+      "loss": 0.6551,
+      "step": 19609
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6595427317111554e-06,
+      "loss": 0.6312,
+      "step": 19610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.65862217655524e-06,
+      "loss": 0.6336,
+      "step": 19611
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.657701684728112e-06,
+      "loss": 0.5983,
+      "step": 19612
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.656781256240688e-06,
+      "loss": 0.624,
+      "step": 19613
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.655860891103881e-06,
+      "loss": 0.6128,
+      "step": 19614
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.654940589328605e-06,
+      "loss": 0.595,
+      "step": 19615
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.654020350925769e-06,
+      "loss": 0.6713,
+      "step": 19616
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.653100175906288e-06,
+      "loss": 0.6028,
+      "step": 19617
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.652180064281065e-06,
+      "loss": 0.5821,
+      "step": 19618
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.651260016061011e-06,
+      "loss": 0.6308,
+      "step": 19619
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.650340031257037e-06,
+      "loss": 0.6181,
+      "step": 19620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.649420109880049e-06,
+      "loss": 0.6051,
+      "step": 19621
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.648500251940955e-06,
+      "loss": 0.5843,
+      "step": 19622
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.647580457450662e-06,
+      "loss": 0.6571,
+      "step": 19623
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6466607264200715e-06,
+      "loss": 0.5883,
+      "step": 19624
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.645741058860089e-06,
+      "loss": 0.6595,
+      "step": 19625
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.644821454781621e-06,
+      "loss": 0.6247,
+      "step": 19626
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.643901914195568e-06,
+      "loss": 0.68,
+      "step": 19627
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.642982437112833e-06,
+      "loss": 0.6109,
+      "step": 19628
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6420630235443205e-06,
+      "loss": 0.5708,
+      "step": 19629
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6411436735009275e-06,
+      "loss": 0.6079,
+      "step": 19630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.640224386993554e-06,
+      "loss": 0.6485,
+      "step": 19631
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6393051640331e-06,
+      "loss": 0.6567,
+      "step": 19632
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.638386004630465e-06,
+      "loss": 0.6091,
+      "step": 19633
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6374669087965505e-06,
+      "loss": 0.6065,
+      "step": 19634
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.636547876542246e-06,
+      "loss": 0.6507,
+      "step": 19635
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.63562890787845e-06,
+      "loss": 0.6314,
+      "step": 19636
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.63471000281606e-06,
+      "loss": 0.6383,
+      "step": 19637
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.63379116136597e-06,
+      "loss": 0.5873,
+      "step": 19638
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.632872383539074e-06,
+      "loss": 0.6142,
+      "step": 19639
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.63195366934627e-06,
+      "loss": 0.6494,
+      "step": 19640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.631035018798441e-06,
+      "loss": 0.6485,
+      "step": 19641
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.630116431906484e-06,
+      "loss": 0.579,
+      "step": 19642
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.629197908681289e-06,
+      "loss": 0.6378,
+      "step": 19643
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.628279449133747e-06,
+      "loss": 0.6416,
+      "step": 19644
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.627361053274748e-06,
+      "loss": 0.6579,
+      "step": 19645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.626442721115184e-06,
+      "loss": 0.6522,
+      "step": 19646
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.625524452665935e-06,
+      "loss": 0.6249,
+      "step": 19647
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6246062479378925e-06,
+      "loss": 0.6403,
+      "step": 19648
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.623688106941942e-06,
+      "loss": 0.6015,
+      "step": 19649
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.622770029688971e-06,
+      "loss": 0.6144,
+      "step": 19650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.621852016189864e-06,
+      "loss": 0.6497,
+      "step": 19651
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.620934066455508e-06,
+      "loss": 0.6614,
+      "step": 19652
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.620016180496779e-06,
+      "loss": 0.6747,
+      "step": 19653
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.619098358324564e-06,
+      "loss": 0.625,
+      "step": 19654
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6181805999497456e-06,
+      "loss": 0.6354,
+      "step": 19655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.617262905383203e-06,
+      "loss": 0.6271,
+      "step": 19656
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.616345274635818e-06,
+      "loss": 0.63,
+      "step": 19657
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.615427707718475e-06,
+      "loss": 0.6315,
+      "step": 19658
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.614510204642044e-06,
+      "loss": 0.6602,
+      "step": 19659
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.613592765417407e-06,
+      "loss": 0.6869,
+      "step": 19660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6126753900554415e-06,
+      "loss": 0.6451,
+      "step": 19661
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.611758078567024e-06,
+      "loss": 0.6753,
+      "step": 19662
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.610840830963032e-06,
+      "loss": 0.5915,
+      "step": 19663
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.609923647254341e-06,
+      "loss": 0.7493,
+      "step": 19664
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.60900652745182e-06,
+      "loss": 0.6328,
+      "step": 19665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.608089471566348e-06,
+      "loss": 0.6417,
+      "step": 19666
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.607172479608794e-06,
+      "loss": 0.6189,
+      "step": 19667
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.606255551590034e-06,
+      "loss": 0.5844,
+      "step": 19668
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.605338687520937e-06,
+      "loss": 0.6639,
+      "step": 19669
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.604421887412378e-06,
+      "loss": 0.5901,
+      "step": 19670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.6035051512752194e-06,
+      "loss": 0.6642,
+      "step": 19671
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.602588479120334e-06,
+      "loss": 0.6055,
+      "step": 19672
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.601671870958588e-06,
+      "loss": 0.602,
+      "step": 19673
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.600755326800853e-06,
+      "loss": 0.6557,
+      "step": 19674
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.599838846657994e-06,
+      "loss": 0.5882,
+      "step": 19675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.598922430540878e-06,
+      "loss": 0.6271,
+      "step": 19676
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.598006078460368e-06,
+      "loss": 0.5891,
+      "step": 19677
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.597089790427327e-06,
+      "loss": 0.5751,
+      "step": 19678
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.596173566452623e-06,
+      "loss": 0.6386,
+      "step": 19679
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.595257406547117e-06,
+      "loss": 0.6244,
+      "step": 19680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.594341310721671e-06,
+      "loss": 0.6189,
+      "step": 19681
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.59342527898715e-06,
+      "loss": 0.6293,
+      "step": 19682
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.59250931135441e-06,
+      "loss": 0.6317,
+      "step": 19683
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.59159340783431e-06,
+      "loss": 0.6311,
+      "step": 19684
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.590677568437714e-06,
+      "loss": 0.6653,
+      "step": 19685
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5897617931754776e-06,
+      "loss": 0.6377,
+      "step": 19686
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.588846082058459e-06,
+      "loss": 0.5736,
+      "step": 19687
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.58793043509752e-06,
+      "loss": 0.5991,
+      "step": 19688
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.587014852303507e-06,
+      "loss": 0.6633,
+      "step": 19689
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.586099333687281e-06,
+      "loss": 0.6367,
+      "step": 19690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.585183879259695e-06,
+      "loss": 0.6185,
+      "step": 19691
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.584268489031604e-06,
+      "loss": 0.6469,
+      "step": 19692
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.583353163013866e-06,
+      "loss": 0.6061,
+      "step": 19693
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.582437901217324e-06,
+      "loss": 0.6695,
+      "step": 19694
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.581522703652833e-06,
+      "loss": 0.589,
+      "step": 19695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.580607570331246e-06,
+      "loss": 0.6327,
+      "step": 19696
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.579692501263412e-06,
+      "loss": 0.6323,
+      "step": 19697
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.578777496460179e-06,
+      "loss": 0.5684,
+      "step": 19698
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.577862555932401e-06,
+      "loss": 0.648,
+      "step": 19699
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.576947679690917e-06,
+      "loss": 0.6761,
+      "step": 19700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.576032867746579e-06,
+      "loss": 0.5994,
+      "step": 19701
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.575118120110234e-06,
+      "loss": 0.5972,
+      "step": 19702
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.574203436792725e-06,
+      "loss": 0.607,
+      "step": 19703
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.573288817804898e-06,
+      "loss": 0.614,
+      "step": 19704
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5723742631576e-06,
+      "loss": 0.652,
+      "step": 19705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.571459772861668e-06,
+      "loss": 0.6275,
+      "step": 19706
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.570545346927947e-06,
+      "loss": 0.6117,
+      "step": 19707
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.569630985367281e-06,
+      "loss": 0.6497,
+      "step": 19708
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5687166881905066e-06,
+      "loss": 0.6438,
+      "step": 19709
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.567802455408468e-06,
+      "loss": 0.6383,
+      "step": 19710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.566888287032007e-06,
+      "loss": 0.5872,
+      "step": 19711
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.565974183071953e-06,
+      "loss": 0.6382,
+      "step": 19712
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5650601435391505e-06,
+      "loss": 0.5928,
+      "step": 19713
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.564146168444435e-06,
+      "loss": 0.6543,
+      "step": 19714
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.563232257798643e-06,
+      "loss": 0.597,
+      "step": 19715
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.562318411612609e-06,
+      "loss": 0.7204,
+      "step": 19716
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5614046298971746e-06,
+      "loss": 0.6557,
+      "step": 19717
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.560490912663164e-06,
+      "loss": 0.5873,
+      "step": 19718
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.559577259921414e-06,
+      "loss": 0.5665,
+      "step": 19719
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.558663671682757e-06,
+      "loss": 0.5563,
+      "step": 19720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.557750147958027e-06,
+      "loss": 0.6688,
+      "step": 19721
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.556836688758053e-06,
+      "loss": 0.6681,
+      "step": 19722
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.555923294093669e-06,
+      "loss": 0.6284,
+      "step": 19723
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.555009963975698e-06,
+      "loss": 0.5921,
+      "step": 19724
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5540966984149714e-06,
+      "loss": 0.6278,
+      "step": 19725
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5531834974223175e-06,
+      "loss": 0.6641,
+      "step": 19726
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.552270361008564e-06,
+      "loss": 0.612,
+      "step": 19727
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.551357289184537e-06,
+      "loss": 0.598,
+      "step": 19728
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.550444281961065e-06,
+      "loss": 0.6506,
+      "step": 19729
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.549531339348966e-06,
+      "loss": 0.6365,
+      "step": 19730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.548618461359068e-06,
+      "loss": 0.6369,
+      "step": 19731
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5477056480021945e-06,
+      "loss": 0.6181,
+      "step": 19732
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.546792899289168e-06,
+      "loss": 0.6388,
+      "step": 19733
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.545880215230809e-06,
+      "loss": 0.6362,
+      "step": 19734
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.544967595837943e-06,
+      "loss": 0.6313,
+      "step": 19735
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.544055041121385e-06,
+      "loss": 0.6584,
+      "step": 19736
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.543142551091954e-06,
+      "loss": 0.6009,
+      "step": 19737
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.542230125760473e-06,
+      "loss": 0.6242,
+      "step": 19738
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.541317765137756e-06,
+      "loss": 0.6317,
+      "step": 19739
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.540405469234623e-06,
+      "loss": 0.6022,
+      "step": 19740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.539493238061894e-06,
+      "loss": 0.5851,
+      "step": 19741
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.538581071630377e-06,
+      "loss": 0.6848,
+      "step": 19742
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5376689699508895e-06,
+      "loss": 0.6151,
+      "step": 19743
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.536756933034245e-06,
+      "loss": 0.6021,
+      "step": 19744
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.535844960891259e-06,
+      "loss": 0.6208,
+      "step": 19745
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.534933053532743e-06,
+      "loss": 0.6892,
+      "step": 19746
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.534021210969514e-06,
+      "loss": 0.6347,
+      "step": 19747
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.533109433212373e-06,
+      "loss": 0.6274,
+      "step": 19748
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.532197720272135e-06,
+      "loss": 0.6221,
+      "step": 19749
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.531286072159611e-06,
+      "loss": 0.6422,
+      "step": 19750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.530374488885607e-06,
+      "loss": 0.5949,
+      "step": 19751
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.529462970460934e-06,
+      "loss": 0.6406,
+      "step": 19752
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5285515168964014e-06,
+      "loss": 0.6548,
+      "step": 19753
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.527640128202808e-06,
+      "loss": 0.6333,
+      "step": 19754
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.526728804390966e-06,
+      "loss": 0.6162,
+      "step": 19755
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.525817545471672e-06,
+      "loss": 0.6548,
+      "step": 19756
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.524906351455739e-06,
+      "loss": 0.633,
+      "step": 19757
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.523995222353971e-06,
+      "loss": 0.6046,
+      "step": 19758
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.523084158177163e-06,
+      "loss": 0.6116,
+      "step": 19759
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.52217315893612e-06,
+      "loss": 0.631,
+      "step": 19760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.521262224641648e-06,
+      "loss": 0.6338,
+      "step": 19761
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.520351355304536e-06,
+      "loss": 0.6688,
+      "step": 19762
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.519440550935595e-06,
+      "loss": 0.7049,
+      "step": 19763
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.518529811545621e-06,
+      "loss": 0.6203,
+      "step": 19764
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.517619137145407e-06,
+      "loss": 0.6233,
+      "step": 19765
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.516708527745754e-06,
+      "loss": 0.6138,
+      "step": 19766
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.515797983357461e-06,
+      "loss": 0.6054,
+      "step": 19767
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.514887503991313e-06,
+      "loss": 0.6319,
+      "step": 19768
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.5139770896581155e-06,
+      "loss": 0.6627,
+      "step": 19769
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.513066740368664e-06,
+      "loss": 0.5835,
+      "step": 19770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.512156456133744e-06,
+      "loss": 0.6091,
+      "step": 19771
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.511246236964151e-06,
+      "loss": 0.6335,
+      "step": 19772
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.51033608287068e-06,
+      "loss": 0.6108,
+      "step": 19773
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.509425993864113e-06,
+      "loss": 0.6085,
+      "step": 19774
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.50851596995525e-06,
+      "loss": 0.617,
+      "step": 19775
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.507606011154879e-06,
+      "loss": 0.5889,
+      "step": 19776
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.506696117473784e-06,
+      "loss": 0.6321,
+      "step": 19777
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.505786288922756e-06,
+      "loss": 0.5589,
+      "step": 19778
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5048765255125855e-06,
+      "loss": 0.5887,
+      "step": 19779
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.503966827254047e-06,
+      "loss": 0.6026,
+      "step": 19780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.503057194157939e-06,
+      "loss": 0.636,
+      "step": 19781
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502147626235045e-06,
+      "loss": 0.6644,
+      "step": 19782
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5012381234961425e-06,
+      "loss": 0.5988,
+      "step": 19783
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5003286859520164e-06,
+      "loss": 0.6393,
+      "step": 19784
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.499419313613456e-06,
+      "loss": 0.6784,
+      "step": 19785
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4985100064912294e-06,
+      "loss": 0.6303,
+      "step": 19786
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4976007645961305e-06,
+      "loss": 0.6567,
+      "step": 19787
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.496691587938938e-06,
+      "loss": 0.609,
+      "step": 19788
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.495782476530426e-06,
+      "loss": 0.6013,
+      "step": 19789
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.494873430381376e-06,
+      "loss": 0.6579,
+      "step": 19790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.493964449502568e-06,
+      "loss": 0.6451,
+      "step": 19791
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.493055533904769e-06,
+      "loss": 0.627,
+      "step": 19792
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.492146683598767e-06,
+      "loss": 0.6733,
+      "step": 19793
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4912378985953375e-06,
+      "loss": 0.6866,
+      "step": 19794
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.490329178905248e-06,
+      "loss": 0.5703,
+      "step": 19795
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4894205245392755e-06,
+      "loss": 0.6543,
+      "step": 19796
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.488511935508196e-06,
+      "loss": 0.6767,
+      "step": 19797
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.487603411822777e-06,
+      "loss": 0.5935,
+      "step": 19798
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.48669495349379e-06,
+      "loss": 0.6695,
+      "step": 19799
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.485786560532015e-06,
+      "loss": 0.6105,
+      "step": 19800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.484878232948212e-06,
+      "loss": 0.6009,
+      "step": 19801
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4839699707531545e-06,
+      "loss": 0.6591,
+      "step": 19802
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.483061773957615e-06,
+      "loss": 0.6472,
+      "step": 19803
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.482153642572353e-06,
+      "loss": 0.6642,
+      "step": 19804
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.481245576608136e-06,
+      "loss": 0.5677,
+      "step": 19805
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.480337576075742e-06,
+      "loss": 0.6012,
+      "step": 19806
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.479429640985924e-06,
+      "loss": 0.6323,
+      "step": 19807
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.478521771349451e-06,
+      "loss": 0.6316,
+      "step": 19808
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.477613967177091e-06,
+      "loss": 0.6226,
+      "step": 19809
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.476706228479599e-06,
+      "loss": 0.5952,
+      "step": 19810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.47579855526774e-06,
+      "loss": 0.6228,
+      "step": 19811
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.474890947552283e-06,
+      "loss": 0.5687,
+      "step": 19812
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.473983405343979e-06,
+      "loss": 0.6593,
+      "step": 19813
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.473075928653593e-06,
+      "loss": 0.6253,
+      "step": 19814
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4721685174918875e-06,
+      "loss": 0.6578,
+      "step": 19815
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.471261171869612e-06,
+      "loss": 0.7089,
+      "step": 19816
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.47035389179753e-06,
+      "loss": 0.6711,
+      "step": 19817
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.469446677286398e-06,
+      "loss": 0.6306,
+      "step": 19818
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.468539528346971e-06,
+      "loss": 0.6534,
+      "step": 19819
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.46763244499001e-06,
+      "loss": 0.6499,
+      "step": 19820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.46672542722626e-06,
+      "loss": 0.6455,
+      "step": 19821
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.46581847506648e-06,
+      "loss": 0.6298,
+      "step": 19822
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.464911588521424e-06,
+      "loss": 0.6588,
+      "step": 19823
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4640047676018415e-06,
+      "loss": 0.5348,
+      "step": 19824
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.463098012318487e-06,
+      "loss": 0.6073,
+      "step": 19825
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.462191322682112e-06,
+      "loss": 0.606,
+      "step": 19826
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.461284698703461e-06,
+      "loss": 0.6137,
+      "step": 19827
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.460378140393286e-06,
+      "loss": 0.6496,
+      "step": 19828
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.459471647762337e-06,
+      "loss": 0.6328,
+      "step": 19829
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.458565220821359e-06,
+      "loss": 0.5947,
+      "step": 19830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.457658859581101e-06,
+      "loss": 0.6675,
+      "step": 19831
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4567525640523125e-06,
+      "loss": 0.6631,
+      "step": 19832
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4558463342457305e-06,
+      "loss": 0.5976,
+      "step": 19833
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.454940170172103e-06,
+      "loss": 0.5639,
+      "step": 19834
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.454034071842175e-06,
+      "loss": 0.6598,
+      "step": 19835
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.453128039266689e-06,
+      "loss": 0.6205,
+      "step": 19836
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4522220724563855e-06,
+      "loss": 0.6143,
+      "step": 19837
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.451316171422013e-06,
+      "loss": 0.612,
+      "step": 19838
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.450410336174302e-06,
+      "loss": 0.6068,
+      "step": 19839
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.449504566723996e-06,
+      "loss": 0.6202,
+      "step": 19840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.448598863081836e-06,
+      "loss": 0.6061,
+      "step": 19841
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.447693225258557e-06,
+      "loss": 0.6512,
+      "step": 19842
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4467876532649e-06,
+      "loss": 0.5988,
+      "step": 19843
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.445882147111604e-06,
+      "loss": 0.6862,
+      "step": 19844
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.444976706809397e-06,
+      "loss": 0.6448,
+      "step": 19845
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4440713323690185e-06,
+      "loss": 0.6874,
+      "step": 19846
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.443166023801203e-06,
+      "loss": 0.6736,
+      "step": 19847
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.442260781116683e-06,
+      "loss": 0.5913,
+      "step": 19848
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.441355604326192e-06,
+      "loss": 0.628,
+      "step": 19849
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4404504934404656e-06,
+      "loss": 0.609,
+      "step": 19850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.43954544847023e-06,
+      "loss": 0.6507,
+      "step": 19851
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.438640469426215e-06,
+      "loss": 0.6445,
+      "step": 19852
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4377355563191515e-06,
+      "loss": 0.6468,
+      "step": 19853
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.436830709159771e-06,
+      "loss": 0.5951,
+      "step": 19854
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4359259279588e-06,
+      "loss": 0.6363,
+      "step": 19855
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.435021212726968e-06,
+      "loss": 0.5999,
+      "step": 19856
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.434116563474996e-06,
+      "loss": 0.6603,
+      "step": 19857
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.433211980213614e-06,
+      "loss": 0.6142,
+      "step": 19858
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.432307462953546e-06,
+      "loss": 0.6086,
+      "step": 19859
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.431403011705515e-06,
+      "loss": 0.5975,
+      "step": 19860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.430498626480247e-06,
+      "loss": 0.5835,
+      "step": 19861
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.429594307288465e-06,
+      "loss": 0.6366,
+      "step": 19862
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.428690054140886e-06,
+      "loss": 0.6305,
+      "step": 19863
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.427785867048236e-06,
+      "loss": 0.6396,
+      "step": 19864
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.426881746021231e-06,
+      "loss": 0.722,
+      "step": 19865
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.425977691070594e-06,
+      "loss": 0.6404,
+      "step": 19866
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4250737022070435e-06,
+      "loss": 0.6514,
+      "step": 19867
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.424169779441299e-06,
+      "loss": 0.6119,
+      "step": 19868
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4232659227840725e-06,
+      "loss": 0.615,
+      "step": 19869
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.422362132246082e-06,
+      "loss": 0.657,
+      "step": 19870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4214584078380454e-06,
+      "loss": 0.6002,
+      "step": 19871
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.420554749570675e-06,
+      "loss": 0.5999,
+      "step": 19872
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.419651157454686e-06,
+      "loss": 0.602,
+      "step": 19873
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4187476315007975e-06,
+      "loss": 0.6553,
+      "step": 19874
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.417844171719711e-06,
+      "loss": 0.6359,
+      "step": 19875
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.416940778122143e-06,
+      "loss": 0.6143,
+      "step": 19876
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.416037450718804e-06,
+      "loss": 0.5992,
+      "step": 19877
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.415134189520407e-06,
+      "loss": 0.5569,
+      "step": 19878
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.414230994537661e-06,
+      "loss": 0.6477,
+      "step": 19879
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.41332786578127e-06,
+      "loss": 0.6825,
+      "step": 19880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.412424803261943e-06,
+      "loss": 0.6733,
+      "step": 19881
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.41152180699039e-06,
+      "loss": 0.6342,
+      "step": 19882
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.410618876977314e-06,
+      "loss": 0.6148,
+      "step": 19883
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.409716013233423e-06,
+      "loss": 0.581,
+      "step": 19884
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.408813215769423e-06,
+      "loss": 0.5995,
+      "step": 19885
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.407910484596012e-06,
+      "loss": 0.5907,
+      "step": 19886
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.407007819723896e-06,
+      "loss": 0.6696,
+      "step": 19887
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4061052211637775e-06,
+      "loss": 0.5544,
+      "step": 19888
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.405202688926358e-06,
+      "loss": 0.6297,
+      "step": 19889
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.404300223022339e-06,
+      "loss": 0.6429,
+      "step": 19890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.403397823462422e-06,
+      "loss": 0.6542,
+      "step": 19891
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.402495490257299e-06,
+      "loss": 0.6116,
+      "step": 19892
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.401593223417675e-06,
+      "loss": 0.6255,
+      "step": 19893
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.400691022954244e-06,
+      "loss": 0.6161,
+      "step": 19894
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.399788888877704e-06,
+      "loss": 0.6597,
+      "step": 19895
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.398886821198752e-06,
+      "loss": 0.6322,
+      "step": 19896
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.397984819928085e-06,
+      "loss": 0.5931,
+      "step": 19897
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.397082885076392e-06,
+      "loss": 0.6107,
+      "step": 19898
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.396181016654367e-06,
+      "loss": 0.6443,
+      "step": 19899
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.395279214672707e-06,
+      "loss": 0.6288,
+      "step": 19900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.394377479142102e-06,
+      "loss": 0.6382,
+      "step": 19901
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.393475810073241e-06,
+      "loss": 0.6256,
+      "step": 19902
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.392574207476822e-06,
+      "loss": 0.653,
+      "step": 19903
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.391672671363525e-06,
+      "loss": 0.6464,
+      "step": 19904
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.390771201744043e-06,
+      "loss": 0.6191,
+      "step": 19905
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.389869798629064e-06,
+      "loss": 0.6545,
+      "step": 19906
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.388968462029276e-06,
+      "loss": 0.6186,
+      "step": 19907
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.388067191955364e-06,
+      "loss": 0.6437,
+      "step": 19908
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.387165988418019e-06,
+      "loss": 0.6139,
+      "step": 19909
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.386264851427917e-06,
+      "loss": 0.6315,
+      "step": 19910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.385363780995747e-06,
+      "loss": 0.6331,
+      "step": 19911
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.384462777132192e-06,
+      "loss": 0.613,
+      "step": 19912
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.383561839847935e-06,
+      "loss": 0.6535,
+      "step": 19913
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.382660969153657e-06,
+      "loss": 0.6393,
+      "step": 19914
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.381760165060043e-06,
+      "loss": 0.6598,
+      "step": 19915
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.380859427577765e-06,
+      "loss": 0.6142,
+      "step": 19916
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3799587567175075e-06,
+      "loss": 0.6148,
+      "step": 19917
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.379058152489947e-06,
+      "loss": 0.6654,
+      "step": 19918
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.378157614905763e-06,
+      "loss": 0.6433,
+      "step": 19919
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3772571439756325e-06,
+      "loss": 0.6417,
+      "step": 19920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3763567397102355e-06,
+      "loss": 0.6371,
+      "step": 19921
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.37545640212024e-06,
+      "loss": 0.6661,
+      "step": 19922
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3745561312163235e-06,
+      "loss": 0.6725,
+      "step": 19923
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.373655927009159e-06,
+      "loss": 0.566,
+      "step": 19924
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.372755789509422e-06,
+      "loss": 0.5838,
+      "step": 19925
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3718557187277835e-06,
+      "loss": 0.6388,
+      "step": 19926
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3709557146749195e-06,
+      "loss": 0.6527,
+      "step": 19927
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3700557773614915e-06,
+      "loss": 0.6128,
+      "step": 19928
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.369155906798175e-06,
+      "loss": 0.6458,
+      "step": 19929
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.368256102995637e-06,
+      "loss": 0.6256,
+      "step": 19930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.367356365964548e-06,
+      "loss": 0.5933,
+      "step": 19931
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.366456695715574e-06,
+      "loss": 0.7225,
+      "step": 19932
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.365557092259387e-06,
+      "loss": 0.6562,
+      "step": 19933
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.364657555606644e-06,
+      "loss": 0.5998,
+      "step": 19934
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.363758085768015e-06,
+      "loss": 0.5975,
+      "step": 19935
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3628586827541634e-06,
+      "loss": 0.644,
+      "step": 19936
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3619593465757525e-06,
+      "loss": 0.6381,
+      "step": 19937
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.361060077243446e-06,
+      "loss": 0.6314,
+      "step": 19938
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.36016087476791e-06,
+      "loss": 0.6804,
+      "step": 19939
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.359261739159797e-06,
+      "loss": 0.6315,
+      "step": 19940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.358362670429772e-06,
+      "loss": 0.6614,
+      "step": 19941
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.357463668588494e-06,
+      "loss": 0.6031,
+      "step": 19942
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.356564733646622e-06,
+      "loss": 0.6581,
+      "step": 19943
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.355665865614818e-06,
+      "loss": 0.6784,
+      "step": 19944
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.354767064503731e-06,
+      "loss": 0.6254,
+      "step": 19945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.353868330324022e-06,
+      "loss": 0.6424,
+      "step": 19946
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.352969663086347e-06,
+      "loss": 0.6516,
+      "step": 19947
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.35207106280136e-06,
+      "loss": 0.6875,
+      "step": 19948
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.351172529479714e-06,
+      "loss": 0.6699,
+      "step": 19949
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.350274063132069e-06,
+      "loss": 0.6773,
+      "step": 19950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.349375663769067e-06,
+      "loss": 0.6611,
+      "step": 19951
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.348477331401364e-06,
+      "loss": 0.6618,
+      "step": 19952
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.347579066039612e-06,
+      "loss": 0.6159,
+      "step": 19953
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3466808676944615e-06,
+      "loss": 0.6265,
+      "step": 19954
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.34578273637656e-06,
+      "loss": 0.624,
+      "step": 19955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.34488467209656e-06,
+      "loss": 0.6501,
+      "step": 19956
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.343986674865103e-06,
+      "loss": 0.6198,
+      "step": 19957
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.34308874469284e-06,
+      "loss": 0.6423,
+      "step": 19958
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.342190881590413e-06,
+      "loss": 0.6676,
+      "step": 19959
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.341293085568472e-06,
+      "loss": 0.6572,
+      "step": 19960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.34039535663766e-06,
+      "loss": 0.6089,
+      "step": 19961
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.339497694808624e-06,
+      "loss": 0.6366,
+      "step": 19962
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.338600100092e-06,
+      "loss": 0.6367,
+      "step": 19963
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.337702572498432e-06,
+      "loss": 0.6324,
+      "step": 19964
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.336805112038564e-06,
+      "loss": 0.6507,
+      "step": 19965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3359077187230355e-06,
+      "loss": 0.6318,
+      "step": 19966
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.335010392562486e-06,
+      "loss": 0.6309,
+      "step": 19967
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.334113133567559e-06,
+      "loss": 0.6244,
+      "step": 19968
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3332159417488826e-06,
+      "loss": 0.6401,
+      "step": 19969
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3323188171171014e-06,
+      "loss": 0.6236,
+      "step": 19970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3314217596828504e-06,
+      "loss": 0.6527,
+      "step": 19971
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.330524769456764e-06,
+      "loss": 0.5808,
+      "step": 19972
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.329627846449479e-06,
+      "loss": 0.5816,
+      "step": 19973
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3287309906716325e-06,
+      "loss": 0.6395,
+      "step": 19974
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.32783420213385e-06,
+      "loss": 0.6072,
+      "step": 19975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.326937480846769e-06,
+      "loss": 0.6799,
+      "step": 19976
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.326040826821021e-06,
+      "loss": 0.5986,
+      "step": 19977
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3251442400672365e-06,
+      "loss": 0.5944,
+      "step": 19978
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3242477205960456e-06,
+      "loss": 0.6243,
+      "step": 19979
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.323351268418081e-06,
+      "loss": 0.6498,
+      "step": 19980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.322454883543965e-06,
+      "loss": 0.6105,
+      "step": 19981
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.321558565984328e-06,
+      "loss": 0.6157,
+      "step": 19982
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3206623157497976e-06,
+      "loss": 0.6146,
+      "step": 19983
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.319766132851e-06,
+      "loss": 0.6309,
+      "step": 19984
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.318870017298561e-06,
+      "loss": 0.6627,
+      "step": 19985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3179739691031075e-06,
+      "loss": 0.6312,
+      "step": 19986
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.317077988275257e-06,
+      "loss": 0.6214,
+      "step": 19987
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3161820748256355e-06,
+      "loss": 0.6124,
+      "step": 19988
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.315286228764865e-06,
+      "loss": 0.6117,
+      "step": 19989
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.31439045010357e-06,
+      "loss": 0.636,
+      "step": 19990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.313494738852366e-06,
+      "loss": 0.6605,
+      "step": 19991
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3125990950218795e-06,
+      "loss": 0.623,
+      "step": 19992
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.311703518622723e-06,
+      "loss": 0.6486,
+      "step": 19993
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3108080096655155e-06,
+      "loss": 0.71,
+      "step": 19994
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.309912568160881e-06,
+      "loss": 0.6082,
+      "step": 19995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.309017194119422e-06,
+      "loss": 0.6022,
+      "step": 19996
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.308121887551768e-06,
+      "loss": 0.6257,
+      "step": 19997
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.307226648468533e-06,
+      "loss": 0.6832,
+      "step": 19998
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.306331476880323e-06,
+      "loss": 0.6151,
+      "step": 19999
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.3054363727977565e-06,
+      "loss": 0.6243,
+      "step": 20000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.304541336231449e-06,
+      "loss": 0.5962,
+      "step": 20001
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.303646367192003e-06,
+      "loss": 0.6466,
+      "step": 20002
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.302751465690041e-06,
+      "loss": 0.6307,
+      "step": 20003
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.301856631736163e-06,
+      "loss": 0.6254,
+      "step": 20004
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.300961865340983e-06,
+      "loss": 0.6484,
+      "step": 20005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.300067166515114e-06,
+      "loss": 0.5931,
+      "step": 20006
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.299172535269156e-06,
+      "loss": 0.6189,
+      "step": 20007
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.298277971613715e-06,
+      "loss": 0.6662,
+      "step": 20008
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.297383475559408e-06,
+      "loss": 0.6255,
+      "step": 20009
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.2964890471168305e-06,
+      "loss": 0.6425,
+      "step": 20010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.29559468629659e-06,
+      "loss": 0.6592,
+      "step": 20011
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.294700393109294e-06,
+      "loss": 0.6631,
+      "step": 20012
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.29380616756554e-06,
+      "loss": 0.6211,
+      "step": 20013
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.292912009675927e-06,
+      "loss": 0.6324,
+      "step": 20014
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.2920179194510705e-06,
+      "loss": 0.6281,
+      "step": 20015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.291123896901557e-06,
+      "loss": 0.6117,
+      "step": 20016
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.290229942037992e-06,
+      "loss": 0.5733,
+      "step": 20017
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.289336054870978e-06,
+      "loss": 0.6476,
+      "step": 20018
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.288442235411105e-06,
+      "loss": 0.6186,
+      "step": 20019
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.28754848366897e-06,
+      "loss": 0.6595,
+      "step": 20020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.286654799655183e-06,
+      "loss": 0.6036,
+      "step": 20021
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.285761183380327e-06,
+      "loss": 0.6176,
+      "step": 20022
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.284867634855e-06,
+      "loss": 0.6143,
+      "step": 20023
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.283974154089801e-06,
+      "loss": 0.5879,
+      "step": 20024
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.283080741095315e-06,
+      "loss": 0.624,
+      "step": 20025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.282187395882134e-06,
+      "loss": 0.6304,
+      "step": 20026
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.281294118460864e-06,
+      "loss": 0.6717,
+      "step": 20027
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.280400908842082e-06,
+      "loss": 0.6587,
+      "step": 20028
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.279507767036382e-06,
+      "loss": 0.5914,
+      "step": 20029
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.278614693054358e-06,
+      "loss": 0.6571,
+      "step": 20030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.277721686906591e-06,
+      "loss": 0.6703,
+      "step": 20031
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.276828748603668e-06,
+      "loss": 0.6595,
+      "step": 20032
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.275935878156188e-06,
+      "loss": 0.6073,
+      "step": 20033
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.275043075574724e-06,
+      "loss": 0.6336,
+      "step": 20034
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.2741503408698685e-06,
+      "loss": 0.5774,
+      "step": 20035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.273257674052206e-06,
+      "loss": 0.5999,
+      "step": 20036
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.272365075132315e-06,
+      "loss": 0.5932,
+      "step": 20037
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.271472544120777e-06,
+      "loss": 0.6838,
+      "step": 20038
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.270580081028187e-06,
+      "loss": 0.6093,
+      "step": 20039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.269687685865115e-06,
+      "loss": 0.6371,
+      "step": 20040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.2687953586421425e-06,
+      "loss": 0.6005,
+      "step": 20041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.267903099369856e-06,
+      "loss": 0.6368,
+      "step": 20042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.267010908058825e-06,
+      "loss": 0.6252,
+      "step": 20043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.266118784719629e-06,
+      "loss": 0.6332,
+      "step": 20044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.265226729362855e-06,
+      "loss": 0.6105,
+      "step": 20045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.264334741999069e-06,
+      "loss": 0.6077,
+      "step": 20046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.263442822638851e-06,
+      "loss": 0.6287,
+      "step": 20047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.262550971292778e-06,
+      "loss": 0.5856,
+      "step": 20048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.261659187971418e-06,
+      "loss": 0.6493,
+      "step": 20049
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.260767472685343e-06,
+      "loss": 0.6606,
+      "step": 20050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.259875825445138e-06,
+      "loss": 0.636,
+      "step": 20051
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.258984246261361e-06,
+      "loss": 0.6003,
+      "step": 20052
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.25809273514459e-06,
+      "loss": 0.5945,
+      "step": 20053
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.257201292105396e-06,
+      "loss": 0.6104,
+      "step": 20054
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.256309917154342e-06,
+      "loss": 0.6139,
+      "step": 20055
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.255418610301996e-06,
+      "loss": 0.6555,
+      "step": 20056
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.2545273715589375e-06,
+      "loss": 0.6344,
+      "step": 20057
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.25363620093572e-06,
+      "loss": 0.6768,
+      "step": 20058
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.252745098442915e-06,
+      "loss": 0.6591,
+      "step": 20059
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.251854064091091e-06,
+      "loss": 0.6253,
+      "step": 20060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.250963097890806e-06,
+      "loss": 0.686,
+      "step": 20061
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.250072199852621e-06,
+      "loss": 0.6469,
+      "step": 20062
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.249181369987111e-06,
+      "loss": 0.6914,
+      "step": 20063
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.248290608304827e-06,
+      "loss": 0.6242,
+      "step": 20064
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247399914816338e-06,
+      "loss": 0.6166,
+      "step": 20065
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.246509289532196e-06,
+      "loss": 0.6141,
+      "step": 20066
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.245618732462965e-06,
+      "loss": 0.673,
+      "step": 20067
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2447282436192025e-06,
+      "loss": 0.6274,
+      "step": 20068
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2438378230114676e-06,
+      "loss": 0.6622,
+      "step": 20069
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.242947470650317e-06,
+      "loss": 0.6076,
+      "step": 20070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.24205718654631e-06,
+      "loss": 0.6206,
+      "step": 20071
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.241166970709995e-06,
+      "loss": 0.5725,
+      "step": 20072
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.240276823151932e-06,
+      "loss": 0.6524,
+      "step": 20073
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.239386743882673e-06,
+      "loss": 0.6423,
+      "step": 20074
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.23849673291277e-06,
+      "loss": 0.6657,
+      "step": 20075
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2376067902527785e-06,
+      "loss": 0.5551,
+      "step": 20076
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.236716915913251e-06,
+      "loss": 0.5939,
+      "step": 20077
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.23582710990473e-06,
+      "loss": 0.6663,
+      "step": 20078
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.234937372237772e-06,
+      "loss": 0.6352,
+      "step": 20079
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.234047702922923e-06,
+      "loss": 0.6231,
+      "step": 20080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2331581019707336e-06,
+      "loss": 0.6397,
+      "step": 20081
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2322685693917484e-06,
+      "loss": 0.605,
+      "step": 20082
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.231379105196521e-06,
+      "loss": 0.611,
+      "step": 20083
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.230489709395587e-06,
+      "loss": 0.6186,
+      "step": 20084
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229600381999496e-06,
+      "loss": 0.5823,
+      "step": 20085
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.228711123018793e-06,
+      "loss": 0.643,
+      "step": 20086
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.227821932464019e-06,
+      "loss": 0.6181,
+      "step": 20087
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2269328103457186e-06,
+      "loss": 0.6431,
+      "step": 20088
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.226043756674436e-06,
+      "loss": 0.655,
+      "step": 20089
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.225154771460706e-06,
+      "loss": 0.6418,
+      "step": 20090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.22426585471507e-06,
+      "loss": 0.687,
+      "step": 20091
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.223377006448069e-06,
+      "loss": 0.6291,
+      "step": 20092
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.222488226670242e-06,
+      "loss": 0.669,
+      "step": 20093
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.221599515392125e-06,
+      "loss": 0.6537,
+      "step": 20094
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.220710872624259e-06,
+      "loss": 0.6476,
+      "step": 20095
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.219822298377173e-06,
+      "loss": 0.6359,
+      "step": 20096
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.218933792661405e-06,
+      "loss": 0.6609,
+      "step": 20097
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.218045355487492e-06,
+      "loss": 0.6429,
+      "step": 20098
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.217156986865964e-06,
+      "loss": 0.5924,
+      "step": 20099
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.216268686807354e-06,
+      "loss": 0.7067,
+      "step": 20100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2153804553222e-06,
+      "loss": 0.629,
+      "step": 20101
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.214492292421025e-06,
+      "loss": 0.6186,
+      "step": 20102
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.213604198114363e-06,
+      "loss": 0.6062,
+      "step": 20103
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2127161724127415e-06,
+      "loss": 0.601,
+      "step": 20104
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.211828215326692e-06,
+      "loss": 0.6598,
+      "step": 20105
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.210940326866741e-06,
+      "loss": 0.6369,
+      "step": 20106
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.210052507043418e-06,
+      "loss": 0.5633,
+      "step": 20107
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.209164755867245e-06,
+      "loss": 0.6358,
+      "step": 20108
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.208277073348749e-06,
+      "loss": 0.6504,
+      "step": 20109
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.207389459498454e-06,
+      "loss": 0.6218,
+      "step": 20110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.206501914326885e-06,
+      "loss": 0.6255,
+      "step": 20111
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.205614437844564e-06,
+      "loss": 0.6534,
+      "step": 20112
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.204727030062018e-06,
+      "loss": 0.6683,
+      "step": 20113
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.203839690989761e-06,
+      "loss": 0.6529,
+      "step": 20114
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2029524206383164e-06,
+      "loss": 0.6288,
+      "step": 20115
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.202065219018203e-06,
+      "loss": 0.6436,
+      "step": 20116
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.201178086139942e-06,
+      "loss": 0.6635,
+      "step": 20117
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.200291022014051e-06,
+      "loss": 0.6514,
+      "step": 20118
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.199404026651048e-06,
+      "loss": 0.6283,
+      "step": 20119
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.198517100061447e-06,
+      "loss": 0.6363,
+      "step": 20120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.197630242255762e-06,
+      "loss": 0.6182,
+      "step": 20121
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.196743453244513e-06,
+      "loss": 0.6016,
+      "step": 20122
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.195856733038211e-06,
+      "loss": 0.6392,
+      "step": 20123
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1949700816473685e-06,
+      "loss": 0.5917,
+      "step": 20124
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.194083499082503e-06,
+      "loss": 0.6545,
+      "step": 20125
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1931969853541185e-06,
+      "loss": 0.6583,
+      "step": 20126
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.19231054047273e-06,
+      "loss": 0.6699,
+      "step": 20127
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.191424164448847e-06,
+      "loss": 0.5973,
+      "step": 20128
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.190537857292978e-06,
+      "loss": 0.6239,
+      "step": 20129
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.189651619015635e-06,
+      "loss": 0.7279,
+      "step": 20130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.188765449627319e-06,
+      "loss": 0.6509,
+      "step": 20131
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.18787934913854e-06,
+      "loss": 0.6468,
+      "step": 20132
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.186993317559803e-06,
+      "loss": 0.6178,
+      "step": 20133
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.186107354901615e-06,
+      "loss": 0.6071,
+      "step": 20134
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.185221461174478e-06,
+      "loss": 0.6226,
+      "step": 20135
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1843356363889e-06,
+      "loss": 0.5877,
+      "step": 20136
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.183449880555376e-06,
+      "loss": 0.6521,
+      "step": 20137
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1825641936844115e-06,
+      "loss": 0.6192,
+      "step": 20138
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.181678575786508e-06,
+      "loss": 0.6362,
+      "step": 20139
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.180793026872165e-06,
+      "loss": 0.6416,
+      "step": 20140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1799075469518825e-06,
+      "loss": 0.6107,
+      "step": 20141
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1790221360361615e-06,
+      "loss": 0.6642,
+      "step": 20142
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.178136794135494e-06,
+      "loss": 0.623,
+      "step": 20143
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.177251521260379e-06,
+      "loss": 0.5895,
+      "step": 20144
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.176366317421312e-06,
+      "loss": 0.6115,
+      "step": 20145
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.175481182628791e-06,
+      "loss": 0.6077,
+      "step": 20146
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.174596116893307e-06,
+      "loss": 0.6446,
+      "step": 20147
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.17371112022536e-06,
+      "loss": 0.6324,
+      "step": 20148
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.172826192635434e-06,
+      "loss": 0.6689,
+      "step": 20149
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.171941334134024e-06,
+      "loss": 0.6062,
+      "step": 20150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1710565447316235e-06,
+      "loss": 0.6205,
+      "step": 20151
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1701718244387205e-06,
+      "loss": 0.5958,
+      "step": 20152
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.169287173265806e-06,
+      "loss": 0.6046,
+      "step": 20153
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.168402591223371e-06,
+      "loss": 0.6161,
+      "step": 20154
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.167518078321896e-06,
+      "loss": 0.6363,
+      "step": 20155
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1666336345718725e-06,
+      "loss": 0.6277,
+      "step": 20156
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.165749259983788e-06,
+      "loss": 0.5815,
+      "step": 20157
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.164864954568126e-06,
+      "loss": 0.6483,
+      "step": 20158
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.16398071833537e-06,
+      "loss": 0.5912,
+      "step": 20159
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.16309655129601e-06,
+      "loss": 0.6428,
+      "step": 20160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.162212453460519e-06,
+      "loss": 0.5807,
+      "step": 20161
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.161328424839386e-06,
+      "loss": 0.675,
+      "step": 20162
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1604444654430896e-06,
+      "loss": 0.6486,
+      "step": 20163
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.159560575282111e-06,
+      "loss": 0.6223,
+      "step": 20164
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.15867675436693e-06,
+      "loss": 0.6412,
+      "step": 20165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.157793002708031e-06,
+      "loss": 0.5883,
+      "step": 20166
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1569093203158805e-06,
+      "loss": 0.6151,
+      "step": 20167
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.156025707200961e-06,
+      "loss": 0.6619,
+      "step": 20168
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.155142163373751e-06,
+      "loss": 0.5968,
+      "step": 20169
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.154258688844724e-06,
+      "loss": 0.6171,
+      "step": 20170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.153375283624356e-06,
+      "loss": 0.6379,
+      "step": 20171
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1524919477231215e-06,
+      "loss": 0.6492,
+      "step": 20172
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.15160868115149e-06,
+      "loss": 0.6881,
+      "step": 20173
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.150725483919936e-06,
+      "loss": 0.6225,
+      "step": 20174
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.14984235603893e-06,
+      "loss": 0.6301,
+      "step": 20175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1489592975189445e-06,
+      "loss": 0.6057,
+      "step": 20176
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.148076308370447e-06,
+      "loss": 0.7109,
+      "step": 20177
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.147193388603912e-06,
+      "loss": 0.6015,
+      "step": 20178
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.146310538229799e-06,
+      "loss": 0.7126,
+      "step": 20179
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.145427757258579e-06,
+      "loss": 0.6539,
+      "step": 20180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1445450457007196e-06,
+      "loss": 0.5999,
+      "step": 20181
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.143662403566686e-06,
+      "loss": 0.6043,
+      "step": 20182
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.142779830866943e-06,
+      "loss": 0.6052,
+      "step": 20183
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.141897327611958e-06,
+      "loss": 0.65,
+      "step": 20184
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.141014893812187e-06,
+      "loss": 0.6205,
+      "step": 20185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.140132529478097e-06,
+      "loss": 0.6265,
+      "step": 20186
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.139250234620147e-06,
+      "loss": 0.6049,
+      "step": 20187
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.138368009248801e-06,
+      "loss": 0.6517,
+      "step": 20188
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.13748585337452e-06,
+      "loss": 0.6718,
+      "step": 20189
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.136603767007756e-06,
+      "loss": 0.639,
+      "step": 20190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.135721750158973e-06,
+      "loss": 0.6072,
+      "step": 20191
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1348398028386295e-06,
+      "loss": 0.6301,
+      "step": 20192
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1339579250571725e-06,
+      "loss": 0.6376,
+      "step": 20193
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.133076116825069e-06,
+      "loss": 0.6871,
+      "step": 20194
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.132194378152774e-06,
+      "loss": 0.5868,
+      "step": 20195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.131312709050733e-06,
+      "loss": 0.6219,
+      "step": 20196
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.130431109529404e-06,
+      "loss": 0.613,
+      "step": 20197
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.129549579599242e-06,
+      "loss": 0.5883,
+      "step": 20198
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.12866811927069e-06,
+      "loss": 0.5762,
+      "step": 20199
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1277867285542075e-06,
+      "loss": 0.64,
+      "step": 20200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.126905407460246e-06,
+      "loss": 0.6337,
+      "step": 20201
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.126024155999246e-06,
+      "loss": 0.6423,
+      "step": 20202
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1251429741816605e-06,
+      "loss": 0.591,
+      "step": 20203
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1242618620179406e-06,
+      "loss": 0.6364,
+      "step": 20204
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.123380819518522e-06,
+      "loss": 0.6657,
+      "step": 20205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.122499846693862e-06,
+      "loss": 0.689,
+      "step": 20206
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.121618943554403e-06,
+      "loss": 0.7169,
+      "step": 20207
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.120738110110586e-06,
+      "loss": 0.6643,
+      "step": 20208
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.119857346372854e-06,
+      "loss": 0.5889,
+      "step": 20209
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.118976652351657e-06,
+      "loss": 0.6639,
+      "step": 20210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.118096028057424e-06,
+      "loss": 0.6073,
+      "step": 20211
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.117215473500606e-06,
+      "loss": 0.6284,
+      "step": 20212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.116334988691644e-06,
+      "loss": 0.5747,
+      "step": 20213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.11545457364097e-06,
+      "loss": 0.6309,
+      "step": 20214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.114574228359026e-06,
+      "loss": 0.5752,
+      "step": 20215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.113693952856254e-06,
+      "loss": 0.5602,
+      "step": 20216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.112813747143078e-06,
+      "loss": 0.5878,
+      "step": 20217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.111933611229948e-06,
+      "loss": 0.6677,
+      "step": 20218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1110535451272945e-06,
+      "loss": 0.6589,
+      "step": 20219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.11017354884555e-06,
+      "loss": 0.6793,
+      "step": 20220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1092936223951465e-06,
+      "loss": 0.6091,
+      "step": 20221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.108413765786523e-06,
+      "loss": 0.602,
+      "step": 20222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.107533979030101e-06,
+      "loss": 0.6076,
+      "step": 20223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.106654262136319e-06,
+      "loss": 0.6135,
+      "step": 20224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.105774615115612e-06,
+      "loss": 0.6192,
+      "step": 20225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.104895037978398e-06,
+      "loss": 0.633,
+      "step": 20226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.104015530735111e-06,
+      "loss": 0.639,
+      "step": 20227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.103136093396182e-06,
+      "loss": 0.653,
+      "step": 20228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.102256725972027e-06,
+      "loss": 0.627,
+      "step": 20229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.101377428473083e-06,
+      "loss": 0.65,
+      "step": 20230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.1004982009097746e-06,
+      "loss": 0.6657,
+      "step": 20231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.09961904329252e-06,
+      "loss": 0.5715,
+      "step": 20232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0987399556317456e-06,
+      "loss": 0.6392,
+      "step": 20233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.097860937937877e-06,
+      "loss": 0.6097,
+      "step": 20234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.096981990221326e-06,
+      "loss": 0.6198,
+      "step": 20235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.096103112492526e-06,
+      "loss": 0.6635,
+      "step": 20236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.095224304761894e-06,
+      "loss": 0.6375,
+      "step": 20237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.094345567039845e-06,
+      "loss": 0.6475,
+      "step": 20238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.093466899336799e-06,
+      "loss": 0.636,
+      "step": 20239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.092588301663179e-06,
+      "loss": 0.6566,
+      "step": 20240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.09170977402939e-06,
+      "loss": 0.6342,
+      "step": 20241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.090831316445859e-06,
+      "loss": 0.5917,
+      "step": 20242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.089952928923001e-06,
+      "loss": 0.6128,
+      "step": 20243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0890746114712255e-06,
+      "loss": 0.6124,
+      "step": 20244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.088196364100947e-06,
+      "loss": 0.6203,
+      "step": 20245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.087318186822582e-06,
+      "loss": 0.5698,
+      "step": 20246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.086440079646533e-06,
+      "loss": 0.651,
+      "step": 20247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.085562042583221e-06,
+      "loss": 0.6617,
+      "step": 20248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.084684075643056e-06,
+      "loss": 0.5998,
+      "step": 20249
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.083806178836442e-06,
+      "loss": 0.6197,
+      "step": 20250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0829283521737915e-06,
+      "loss": 0.5615,
+      "step": 20251
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.082050595665508e-06,
+      "loss": 0.6352,
+      "step": 20252
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.081172909321997e-06,
+      "loss": 0.6451,
+      "step": 20253
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.080295293153676e-06,
+      "loss": 0.6742,
+      "step": 20254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.079417747170938e-06,
+      "loss": 0.5807,
+      "step": 20255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.078540271384194e-06,
+      "loss": 0.6304,
+      "step": 20256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.077662865803847e-06,
+      "loss": 0.6371,
+      "step": 20257
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.076785530440297e-06,
+      "loss": 0.6007,
+      "step": 20258
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.075908265303944e-06,
+      "loss": 0.6617,
+      "step": 20259
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.075031070405199e-06,
+      "loss": 0.6031,
+      "step": 20260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.074153945754452e-06,
+      "loss": 0.6407,
+      "step": 20261
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.073276891362107e-06,
+      "loss": 0.6678,
+      "step": 20262
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.072399907238564e-06,
+      "loss": 0.6559,
+      "step": 20263
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.071522993394216e-06,
+      "loss": 0.6893,
+      "step": 20264
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.070646149839459e-06,
+      "loss": 0.6086,
+      "step": 20265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.069769376584699e-06,
+      "loss": 0.6401,
+      "step": 20266
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0688926736403215e-06,
+      "loss": 0.6915,
+      "step": 20267
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0680160410167246e-06,
+      "loss": 0.6093,
+      "step": 20268
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0671394787243045e-06,
+      "loss": 0.6157,
+      "step": 20269
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.066262986773448e-06,
+      "loss": 0.6917,
+      "step": 20270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.065386565174546e-06,
+      "loss": 0.6569,
+      "step": 20271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.064510213938001e-06,
+      "loss": 0.6127,
+      "step": 20272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.063633933074192e-06,
+      "loss": 0.6197,
+      "step": 20273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.062757722593512e-06,
+      "loss": 0.6497,
+      "step": 20274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.061881582506354e-06,
+      "loss": 0.6153,
+      "step": 20275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.061005512823099e-06,
+      "loss": 0.6525,
+      "step": 20276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.060129513554131e-06,
+      "loss": 0.5962,
+      "step": 20277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0592535847098505e-06,
+      "loss": 0.667,
+      "step": 20278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.058377726300629e-06,
+      "loss": 0.6769,
+      "step": 20279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.057501938336856e-06,
+      "loss": 0.6136,
+      "step": 20280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.056626220828918e-06,
+      "loss": 0.6123,
+      "step": 20281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.055750573787191e-06,
+      "loss": 0.5904,
+      "step": 20282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.054874997222058e-06,
+      "loss": 0.6507,
+      "step": 20283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.05399949114391e-06,
+      "loss": 0.6664,
+      "step": 20284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.053124055563115e-06,
+      "loss": 0.673,
+      "step": 20285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.052248690490057e-06,
+      "loss": 0.6037,
+      "step": 20286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0513733959351185e-06,
+      "loss": 0.6199,
+      "step": 20287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.05049817190867e-06,
+      "loss": 0.6944,
+      "step": 20288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0496230184210885e-06,
+      "loss": 0.6015,
+      "step": 20289
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.048747935482761e-06,
+      "loss": 0.6396,
+      "step": 20290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.04787292310405e-06,
+      "loss": 0.6152,
+      "step": 20291
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.046997981295336e-06,
+      "loss": 0.6066,
+      "step": 20292
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0461231100669955e-06,
+      "loss": 0.5795,
+      "step": 20293
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.045248309429393e-06,
+      "loss": 0.64,
+      "step": 20294
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.044373579392901e-06,
+      "loss": 0.6689,
+      "step": 20295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.043498919967901e-06,
+      "loss": 0.6429,
+      "step": 20296
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0426243311647525e-06,
+      "loss": 0.6719,
+      "step": 20297
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.04174981299383e-06,
+      "loss": 0.6174,
+      "step": 20298
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.040875365465504e-06,
+      "loss": 0.6399,
+      "step": 20299
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.040000988590135e-06,
+      "loss": 0.6189,
+      "step": 20300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.039126682378089e-06,
+      "loss": 0.6669,
+      "step": 20301
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0382524468397455e-06,
+      "loss": 0.6477,
+      "step": 20302
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.037378281985457e-06,
+      "loss": 0.6225,
+      "step": 20303
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.036504187825591e-06,
+      "loss": 0.6406,
+      "step": 20304
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.035630164370516e-06,
+      "loss": 0.6192,
+      "step": 20305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.034756211630586e-06,
+      "loss": 0.6633,
+      "step": 20306
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.033882329616164e-06,
+      "loss": 0.6058,
+      "step": 20307
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.033008518337622e-06,
+      "loss": 0.6195,
+      "step": 20308
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.032134777805307e-06,
+      "loss": 0.6307,
+      "step": 20309
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.031261108029586e-06,
+      "loss": 0.6001,
+      "step": 20310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0303875090208166e-06,
+      "loss": 0.6723,
+      "step": 20311
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.029513980789352e-06,
+      "loss": 0.6143,
+      "step": 20312
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.028640523345553e-06,
+      "loss": 0.6388,
+      "step": 20313
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.027767136699773e-06,
+      "loss": 0.6111,
+      "step": 20314
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.026893820862369e-06,
+      "loss": 0.6213,
+      "step": 20315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.026020575843698e-06,
+      "loss": 0.6424,
+      "step": 20316
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.025147401654107e-06,
+      "loss": 0.62,
+      "step": 20317
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.024274298303952e-06,
+      "loss": 0.6123,
+      "step": 20318
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.023401265803584e-06,
+      "loss": 0.6012,
+      "step": 20319
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.022528304163355e-06,
+      "loss": 0.634,
+      "step": 20320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0216554133936136e-06,
+      "loss": 0.6728,
+      "step": 20321
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.020782593504714e-06,
+      "loss": 0.5991,
+      "step": 20322
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.019909844506996e-06,
+      "loss": 0.628,
+      "step": 20323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.019037166410813e-06,
+      "loss": 0.6379,
+      "step": 20324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.018164559226509e-06,
+      "loss": 0.6009,
+      "step": 20325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.017292022964433e-06,
+      "loss": 0.6412,
+      "step": 20326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.016419557634927e-06,
+      "loss": 0.6276,
+      "step": 20327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.015547163248341e-06,
+      "loss": 0.6062,
+      "step": 20328
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.014674839815009e-06,
+      "loss": 0.5865,
+      "step": 20329
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.013802587345279e-06,
+      "loss": 0.6424,
+      "step": 20330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.012930405849491e-06,
+      "loss": 0.6567,
+      "step": 20331
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.012058295337988e-06,
+      "loss": 0.6389,
+      "step": 20332
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.011186255821108e-06,
+      "loss": 0.6296,
+      "step": 20333
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.010314287309195e-06,
+      "loss": 0.6241,
+      "step": 20334
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.009442389812579e-06,
+      "loss": 0.6863,
+      "step": 20335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.008570563341601e-06,
+      "loss": 0.6758,
+      "step": 20336
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.007698807906599e-06,
+      "loss": 0.6636,
+      "step": 20337
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.006827123517908e-06,
+      "loss": 0.5573,
+      "step": 20338
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.005955510185862e-06,
+      "loss": 0.6598,
+      "step": 20339
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0050839679208e-06,
+      "loss": 0.6342,
+      "step": 20340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.004212496733048e-06,
+      "loss": 0.6581,
+      "step": 20341
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.003341096632941e-06,
+      "loss": 0.6031,
+      "step": 20342
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.002469767630811e-06,
+      "loss": 0.6335,
+      "step": 20343
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.001598509736989e-06,
+      "loss": 0.6356,
+      "step": 20344
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0007273229618045e-06,
+      "loss": 0.5756,
+      "step": 20345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.9998562073155914e-06,
+      "loss": 0.6508,
+      "step": 20346
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.998985162808669e-06,
+      "loss": 0.6467,
+      "step": 20347
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.998114189451367e-06,
+      "loss": 0.6263,
+      "step": 20348
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.997243287254017e-06,
+      "loss": 0.6459,
+      "step": 20349
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.996372456226939e-06,
+      "loss": 0.6243,
+      "step": 20350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9955016963804604e-06,
+      "loss": 0.6246,
+      "step": 20351
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.99463100772491e-06,
+      "loss": 0.634,
+      "step": 20352
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9937603902706e-06,
+      "loss": 0.6437,
+      "step": 20353
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.992889844027861e-06,
+      "loss": 0.6102,
+      "step": 20354
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9920193690070095e-06,
+      "loss": 0.5704,
+      "step": 20355
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.99114896521837e-06,
+      "loss": 0.6527,
+      "step": 20356
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9902786326722605e-06,
+      "loss": 0.6461,
+      "step": 20357
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9894083713790035e-06,
+      "loss": 0.6228,
+      "step": 20358
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98853818134891e-06,
+      "loss": 0.6092,
+      "step": 20359
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.987668062592302e-06,
+      "loss": 0.6553,
+      "step": 20360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.986798015119493e-06,
+      "loss": 0.638,
+      "step": 20361
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.985928038940802e-06,
+      "loss": 0.6242,
+      "step": 20362
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9850581340665405e-06,
+      "loss": 0.6648,
+      "step": 20363
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984188300507028e-06,
+      "loss": 0.6687,
+      "step": 20364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.983318538272569e-06,
+      "loss": 0.6731,
+      "step": 20365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9824488473734814e-06,
+      "loss": 0.6895,
+      "step": 20366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.981579227820074e-06,
+      "loss": 0.57,
+      "step": 20367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980709679622659e-06,
+      "loss": 0.5959,
+      "step": 20368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.979840202791544e-06,
+      "loss": 0.5914,
+      "step": 20369
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.978970797337043e-06,
+      "loss": 0.6496,
+      "step": 20370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9781014632694556e-06,
+      "loss": 0.701,
+      "step": 20371
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.977232200599093e-06,
+      "loss": 0.6576,
+      "step": 20372
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9763630093362625e-06,
+      "loss": 0.6981,
+      "step": 20373
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.975493889491269e-06,
+      "loss": 0.6505,
+      "step": 20374
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.974624841074419e-06,
+      "loss": 0.6434,
+      "step": 20375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.97375586409601e-06,
+      "loss": 0.6584,
+      "step": 20376
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.972886958566349e-06,
+      "loss": 0.6432,
+      "step": 20377
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9720181244957366e-06,
+      "loss": 0.6022,
+      "step": 20378
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.971149361894476e-06,
+      "loss": 0.652,
+      "step": 20379
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970280670772866e-06,
+      "loss": 0.598,
+      "step": 20380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.969412051141209e-06,
+      "loss": 0.6291,
+      "step": 20381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.968543503009798e-06,
+      "loss": 0.5904,
+      "step": 20382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.967675026388933e-06,
+      "loss": 0.6206,
+      "step": 20383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.966806621288912e-06,
+      "loss": 0.6366,
+      "step": 20384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.965938287720031e-06,
+      "loss": 0.5908,
+      "step": 20385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.965070025692584e-06,
+      "loss": 0.729,
+      "step": 20386
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9642018352168685e-06,
+      "loss": 0.5753,
+      "step": 20387
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.963333716303174e-06,
+      "loss": 0.6851,
+      "step": 20388
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.962465668961793e-06,
+      "loss": 0.6365,
+      "step": 20389
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9615976932030206e-06,
+      "loss": 0.6341,
+      "step": 20390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.960729789037145e-06,
+      "loss": 0.6367,
+      "step": 20391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.959861956474459e-06,
+      "loss": 0.6559,
+      "step": 20392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.958994195525253e-06,
+      "loss": 0.6312,
+      "step": 20393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.958126506199809e-06,
+      "loss": 0.5851,
+      "step": 20394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.957258888508418e-06,
+      "loss": 0.5839,
+      "step": 20395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.956391342461368e-06,
+      "loss": 0.6292,
+      "step": 20396
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.955523868068942e-06,
+      "loss": 0.6164,
+      "step": 20397
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.954656465341428e-06,
+      "loss": 0.6116,
+      "step": 20398
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9537891342891134e-06,
+      "loss": 0.6103,
+      "step": 20399
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9529218749222734e-06,
+      "loss": 0.6518,
+      "step": 20400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.952054687251193e-06,
+      "loss": 0.6523,
+      "step": 20401
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.951187571286156e-06,
+      "loss": 0.6238,
+      "step": 20402
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.950320527037441e-06,
+      "loss": 0.5679,
+      "step": 20403
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.94945355451533e-06,
+      "loss": 0.5808,
+      "step": 20404
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.948586653730105e-06,
+      "loss": 0.6632,
+      "step": 20405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.947719824692035e-06,
+      "loss": 0.7228,
+      "step": 20406
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9468530674114034e-06,
+      "loss": 0.6272,
+      "step": 20407
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9459863818984864e-06,
+      "loss": 0.6481,
+      "step": 20408
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.945119768163559e-06,
+      "loss": 0.6804,
+      "step": 20409
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.944253226216896e-06,
+      "loss": 0.6106,
+      "step": 20410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.943386756068776e-06,
+      "loss": 0.5714,
+      "step": 20411
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.942520357729464e-06,
+      "loss": 0.6681,
+      "step": 20412
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.941654031209235e-06,
+      "loss": 0.6467,
+      "step": 20413
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9407877765183615e-06,
+      "loss": 0.6729,
+      "step": 20414
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.939921593667114e-06,
+      "loss": 0.6892,
+      "step": 20415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.939055482665762e-06,
+      "loss": 0.6244,
+      "step": 20416
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.938189443524578e-06,
+      "loss": 0.6796,
+      "step": 20417
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.937323476253822e-06,
+      "loss": 0.6434,
+      "step": 20418
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.936457580863767e-06,
+      "loss": 0.5862,
+      "step": 20419
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.935591757364677e-06,
+      "loss": 0.6358,
+      "step": 20420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.934726005766819e-06,
+      "loss": 0.623,
+      "step": 20421
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.933860326080456e-06,
+      "loss": 0.6507,
+      "step": 20422
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.932994718315857e-06,
+      "loss": 0.6345,
+      "step": 20423
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.932129182483275e-06,
+      "loss": 0.627,
+      "step": 20424
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.931263718592979e-06,
+      "loss": 0.6411,
+      "step": 20425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.930398326655228e-06,
+      "loss": 0.6984,
+      "step": 20426
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.929533006680284e-06,
+      "loss": 0.6464,
+      "step": 20427
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.928667758678404e-06,
+      "loss": 0.5991,
+      "step": 20428
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.927802582659852e-06,
+      "loss": 0.6358,
+      "step": 20429
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.926937478634879e-06,
+      "loss": 0.6444,
+      "step": 20430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.926072446613744e-06,
+      "loss": 0.6223,
+      "step": 20431
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.925207486606704e-06,
+      "loss": 0.6461,
+      "step": 20432
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.924342598624012e-06,
+      "loss": 0.6486,
+      "step": 20433
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.923477782675926e-06,
+      "loss": 0.637,
+      "step": 20434
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.922613038772701e-06,
+      "loss": 0.6618,
+      "step": 20435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9217483669245834e-06,
+      "loss": 0.6277,
+      "step": 20436
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.920883767141831e-06,
+      "loss": 0.6729,
+      "step": 20437
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.920019239434684e-06,
+      "loss": 0.6156,
+      "step": 20438
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9191547838134046e-06,
+      "loss": 0.6337,
+      "step": 20439
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.91829040028824e-06,
+      "loss": 0.64,
+      "step": 20440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.917426088869434e-06,
+      "loss": 0.6171,
+      "step": 20441
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.916561849567235e-06,
+      "loss": 0.6666,
+      "step": 20442
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.915697682391895e-06,
+      "loss": 0.5972,
+      "step": 20443
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.914833587353649e-06,
+      "loss": 0.5728,
+      "step": 20444
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.913969564462752e-06,
+      "loss": 0.6056,
+      "step": 20445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.913105613729448e-06,
+      "loss": 0.6709,
+      "step": 20446
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.912241735163973e-06,
+      "loss": 0.6832,
+      "step": 20447
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.911377928776573e-06,
+      "loss": 0.6663,
+      "step": 20448
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.910514194577494e-06,
+      "loss": 0.6459,
+      "step": 20449
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9096505325769655e-06,
+      "loss": 0.6665,
+      "step": 20450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.908786942785237e-06,
+      "loss": 0.6137,
+      "step": 20451
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9079234252125486e-06,
+      "loss": 0.6665,
+      "step": 20452
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.907059979869131e-06,
+      "loss": 0.6434,
+      "step": 20453
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.906196606765225e-06,
+      "loss": 0.6642,
+      "step": 20454
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.905333305911071e-06,
+      "loss": 0.6409,
+      "step": 20455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.904470077316893e-06,
+      "loss": 0.6115,
+      "step": 20456
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.903606920992936e-06,
+      "loss": 0.6613,
+      "step": 20457
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.902743836949435e-06,
+      "loss": 0.6464,
+      "step": 20458
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.901880825196616e-06,
+      "loss": 0.661,
+      "step": 20459
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.901017885744713e-06,
+      "loss": 0.6553,
+      "step": 20460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.900155018603963e-06,
+      "loss": 0.6234,
+      "step": 20461
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.899292223784583e-06,
+      "loss": 0.5634,
+      "step": 20462
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.898429501296816e-06,
+      "loss": 0.6309,
+      "step": 20463
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.897566851150888e-06,
+      "loss": 0.6015,
+      "step": 20464
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.896704273357022e-06,
+      "loss": 0.6276,
+      "step": 20465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.895841767925447e-06,
+      "loss": 0.62,
+      "step": 20466
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.894979334866393e-06,
+      "loss": 0.6539,
+      "step": 20467
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8941169741900755e-06,
+      "loss": 0.6479,
+      "step": 20468
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.893254685906729e-06,
+      "loss": 0.5766,
+      "step": 20469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.892392470026576e-06,
+      "loss": 0.5971,
+      "step": 20470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.891530326559833e-06,
+      "loss": 0.5896,
+      "step": 20471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.890668255516725e-06,
+      "loss": 0.6777,
+      "step": 20472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.889806256907478e-06,
+      "loss": 0.6099,
+      "step": 20473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.888944330742299e-06,
+      "loss": 0.625,
+      "step": 20474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.88808247703142e-06,
+      "loss": 0.6141,
+      "step": 20475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.887220695785056e-06,
+      "loss": 0.6374,
+      "step": 20476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.886358987013422e-06,
+      "loss": 0.615,
+      "step": 20477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.885497350726736e-06,
+      "loss": 0.6291,
+      "step": 20478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.884635786935217e-06,
+      "loss": 0.6826,
+      "step": 20479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.88377429564907e-06,
+      "loss": 0.6251,
+      "step": 20480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.882912876878518e-06,
+      "loss": 0.6302,
+      "step": 20481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.882051530633777e-06,
+      "loss": 0.615,
+      "step": 20482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.88119025692505e-06,
+      "loss": 0.6638,
+      "step": 20483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.880329055762552e-06,
+      "loss": 0.6429,
+      "step": 20484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8794679271565004e-06,
+      "loss": 0.605,
+      "step": 20485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.878606871117091e-06,
+      "loss": 0.6562,
+      "step": 20486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.877745887654544e-06,
+      "loss": 0.6484,
+      "step": 20487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.876884976779068e-06,
+      "loss": 0.6186,
+      "step": 20488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8760241385008615e-06,
+      "loss": 0.666,
+      "step": 20489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8751633728301375e-06,
+      "loss": 0.5905,
+      "step": 20490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.874302679777102e-06,
+      "loss": 0.5989,
+      "step": 20491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.873442059351951e-06,
+      "loss": 0.6082,
+      "step": 20492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.872581511564897e-06,
+      "loss": 0.6635,
+      "step": 20493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.871721036426145e-06,
+      "loss": 0.5939,
+      "step": 20494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.870860633945889e-06,
+      "loss": 0.603,
+      "step": 20495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.870000304134332e-06,
+      "loss": 0.5993,
+      "step": 20496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8691400470016806e-06,
+      "loss": 0.6029,
+      "step": 20497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8682798625581234e-06,
+      "loss": 0.5999,
+      "step": 20498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.86741975081387e-06,
+      "loss": 0.62,
+      "step": 20499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.866559711779111e-06,
+      "loss": 0.6952,
+      "step": 20500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.865699745464046e-06,
+      "loss": 0.6264,
+      "step": 20501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.864839851878873e-06,
+      "loss": 0.6495,
+      "step": 20502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.86398003103378e-06,
+      "loss": 0.6258,
+      "step": 20503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.863120282938965e-06,
+      "loss": 0.6737,
+      "step": 20504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8622606076046276e-06,
+      "loss": 0.6345,
+      "step": 20505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.861401005040951e-06,
+      "loss": 0.6075,
+      "step": 20506
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.860541475258133e-06,
+      "loss": 0.6312,
+      "step": 20507
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.859682018266364e-06,
+      "loss": 0.6184,
+      "step": 20508
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.858822634075828e-06,
+      "loss": 0.6209,
+      "step": 20509
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.857963322696716e-06,
+      "loss": 0.6805,
+      "step": 20510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.857104084139226e-06,
+      "loss": 0.6105,
+      "step": 20511
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.856244918413534e-06,
+      "loss": 0.6128,
+      "step": 20512
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.85538582552983e-06,
+      "loss": 0.6657,
+      "step": 20513
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.854526805498304e-06,
+      "loss": 0.6607,
+      "step": 20514
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.853667858329133e-06,
+      "loss": 0.6449,
+      "step": 20515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.852808984032501e-06,
+      "loss": 0.6525,
+      "step": 20516
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.851950182618602e-06,
+      "loss": 0.6322,
+      "step": 20517
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.851091454097608e-06,
+      "loss": 0.6542,
+      "step": 20518
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.850232798479702e-06,
+      "loss": 0.6105,
+      "step": 20519
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.849374215775069e-06,
+      "loss": 0.6855,
+      "step": 20520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.848515705993882e-06,
+      "loss": 0.6121,
+      "step": 20521
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.847657269146321e-06,
+      "loss": 0.5787,
+      "step": 20522
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8467989052425715e-06,
+      "loss": 0.6363,
+      "step": 20523
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8459406142928015e-06,
+      "loss": 0.6392,
+      "step": 20524
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.84508239630719e-06,
+      "loss": 0.5958,
+      "step": 20525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.844224251295917e-06,
+      "loss": 0.6715,
+      "step": 20526
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.843366179269149e-06,
+      "loss": 0.6151,
+      "step": 20527
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8425081802370605e-06,
+      "loss": 0.6552,
+      "step": 20528
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.841650254209832e-06,
+      "loss": 0.6423,
+      "step": 20529
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.840792401197628e-06,
+      "loss": 0.6447,
+      "step": 20530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.83993462121062e-06,
+      "loss": 0.5952,
+      "step": 20531
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.839076914258984e-06,
+      "loss": 0.6899,
+      "step": 20532
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.838219280352882e-06,
+      "loss": 0.6146,
+      "step": 20533
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.837361719502484e-06,
+      "loss": 0.6187,
+      "step": 20534
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.836504231717959e-06,
+      "loss": 0.6139,
+      "step": 20535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.835646817009473e-06,
+      "loss": 0.625,
+      "step": 20536
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.834789475387192e-06,
+      "loss": 0.6824,
+      "step": 20537
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.833932206861284e-06,
+      "loss": 0.6383,
+      "step": 20538
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.833075011441907e-06,
+      "loss": 0.6048,
+      "step": 20539
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.832217889139226e-06,
+      "loss": 0.6357,
+      "step": 20540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.831360839963405e-06,
+      "loss": 0.6559,
+      "step": 20541
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.830503863924605e-06,
+      "loss": 0.6193,
+      "step": 20542
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.829646961032985e-06,
+      "loss": 0.6382,
+      "step": 20543
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.82879013129871e-06,
+      "loss": 0.6052,
+      "step": 20544
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.827933374731932e-06,
+      "loss": 0.637,
+      "step": 20545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.827076691342809e-06,
+      "loss": 0.6132,
+      "step": 20546
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.826220081141503e-06,
+      "loss": 0.664,
+      "step": 20547
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.825363544138166e-06,
+      "loss": 0.6092,
+      "step": 20548
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8245070803429575e-06,
+      "loss": 0.6166,
+      "step": 20549
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.823650689766031e-06,
+      "loss": 0.6211,
+      "step": 20550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.822794372417537e-06,
+      "loss": 0.6383,
+      "step": 20551
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.821938128307629e-06,
+      "loss": 0.6437,
+      "step": 20552
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.821081957446459e-06,
+      "loss": 0.6748,
+      "step": 20553
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.820225859844179e-06,
+      "loss": 0.6312,
+      "step": 20554
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.819369835510939e-06,
+      "loss": 0.621,
+      "step": 20555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.818513884456893e-06,
+      "loss": 0.6102,
+      "step": 20556
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.817658006692179e-06,
+      "loss": 0.6626,
+      "step": 20557
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.816802202226951e-06,
+      "loss": 0.6857,
+      "step": 20558
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.815946471071355e-06,
+      "loss": 0.6637,
+      "step": 20559
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.815090813235536e-06,
+      "loss": 0.6397,
+      "step": 20560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8142352287296424e-06,
+      "loss": 0.6202,
+      "step": 20561
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8133797175638133e-06,
+      "loss": 0.6742,
+      "step": 20562
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8125242797481923e-06,
+      "loss": 0.6231,
+      "step": 20563
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8116689152929233e-06,
+      "loss": 0.6346,
+      "step": 20564
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8108136242081485e-06,
+      "loss": 0.6823,
+      "step": 20565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8099584065040062e-06,
+      "loss": 0.6098,
+      "step": 20566
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.809103262190642e-06,
+      "loss": 0.6414,
+      "step": 20567
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8082481912781866e-06,
+      "loss": 0.6136,
+      "step": 20568
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8073931937767804e-06,
+      "loss": 0.6407,
+      "step": 20569
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8065382696965623e-06,
+      "loss": 0.6084,
+      "step": 20570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8056834190476676e-06,
+      "loss": 0.648,
+      "step": 20571
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8048286418402325e-06,
+      "loss": 0.616,
+      "step": 20572
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.803973938084394e-06,
+      "loss": 0.6017,
+      "step": 20573
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8031193077902783e-06,
+      "loss": 0.6433,
+      "step": 20574
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8022647509680234e-06,
+      "loss": 0.6379,
+      "step": 20575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.8014102676277596e-06,
+      "loss": 0.625,
+      "step": 20576
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.800555857779619e-06,
+      "loss": 0.6749,
+      "step": 20577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7997015214337297e-06,
+      "loss": 0.6382,
+      "step": 20578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.798847258600226e-06,
+      "loss": 0.6014,
+      "step": 20579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7979930692892296e-06,
+      "loss": 0.666,
+      "step": 20580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7971389535108703e-06,
+      "loss": 0.6245,
+      "step": 20581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7962849112752764e-06,
+      "loss": 0.6209,
+      "step": 20582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.795430942592572e-06,
+      "loss": 0.6014,
+      "step": 20583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7945770474728827e-06,
+      "loss": 0.6832,
+      "step": 20584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7937232259263356e-06,
+      "loss": 0.6576,
+      "step": 20585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.792869477963047e-06,
+      "loss": 0.5455,
+      "step": 20586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.792015803593143e-06,
+      "loss": 0.616,
+      "step": 20587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7911622028267458e-06,
+      "loss": 0.6536,
+      "step": 20588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7903086756739737e-06,
+      "loss": 0.6512,
+      "step": 20589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7894552221449475e-06,
+      "loss": 0.5535,
+      "step": 20590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7886018422497906e-06,
+      "loss": 0.667,
+      "step": 20591
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.787748535998612e-06,
+      "loss": 0.6146,
+      "step": 20592
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.786895303401533e-06,
+      "loss": 0.6331,
+      "step": 20593
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.78604214446867e-06,
+      "loss": 0.6339,
+      "step": 20594
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7851890592101383e-06,
+      "loss": 0.6578,
+      "step": 20595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7843360476360514e-06,
+      "loss": 0.6689,
+      "step": 20596
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7834831097565274e-06,
+      "loss": 0.6334,
+      "step": 20597
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7826302455816723e-06,
+      "loss": 0.5645,
+      "step": 20598
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.781777455121599e-06,
+      "loss": 0.6217,
+      "step": 20599
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7809247383864213e-06,
+      "loss": 0.6513,
+      "step": 20600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.780072095386248e-06,
+      "loss": 0.6534,
+      "step": 20601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.779219526131188e-06,
+      "loss": 0.5742,
+      "step": 20602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7783670306313535e-06,
+      "loss": 0.6416,
+      "step": 20603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.777514608896844e-06,
+      "loss": 0.6096,
+      "step": 20604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.776662260937771e-06,
+      "loss": 0.6648,
+      "step": 20605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7758099867642396e-06,
+      "loss": 0.6246,
+      "step": 20606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.774957786386354e-06,
+      "loss": 0.6789,
+      "step": 20607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7741056598142188e-06,
+      "loss": 0.6462,
+      "step": 20608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7732536070579408e-06,
+      "loss": 0.6234,
+      "step": 20609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7724016281276145e-06,
+      "loss": 0.6504,
+      "step": 20610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.771549723033344e-06,
+      "loss": 0.6361,
+      "step": 20611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.770697891785231e-06,
+      "loss": 0.6277,
+      "step": 20612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7698461343933745e-06,
+      "loss": 0.6387,
+      "step": 20613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.768994450867873e-06,
+      "loss": 0.6082,
+      "step": 20614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.768142841218829e-06,
+      "loss": 0.6738,
+      "step": 20615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.767291305456331e-06,
+      "loss": 0.6481,
+      "step": 20616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7664398435904792e-06,
+      "loss": 0.6598,
+      "step": 20617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7655884556313683e-06,
+      "loss": 0.6087,
+      "step": 20618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.764737141589092e-06,
+      "loss": 0.5652,
+      "step": 20619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.763885901473746e-06,
+      "loss": 0.5698,
+      "step": 20620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.763034735295423e-06,
+      "loss": 0.6703,
+      "step": 20621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7621836430642113e-06,
+      "loss": 0.6468,
+      "step": 20622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7613326247902026e-06,
+      "loss": 0.6164,
+      "step": 20623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7604816804834875e-06,
+      "loss": 0.6142,
+      "step": 20624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7596308101541555e-06,
+      "loss": 0.6306,
+      "step": 20625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.758780013812299e-06,
+      "loss": 0.6366,
+      "step": 20626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7579292914679956e-06,
+      "loss": 0.6786,
+      "step": 20627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.757078643131337e-06,
+      "loss": 0.6219,
+      "step": 20628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7562280688124074e-06,
+      "loss": 0.6284,
+      "step": 20629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7553775685212933e-06,
+      "loss": 0.6596,
+      "step": 20630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7545271422680785e-06,
+      "loss": 0.6557,
+      "step": 20631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7536767900628478e-06,
+      "loss": 0.5962,
+      "step": 20632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.752826511915677e-06,
+      "loss": 0.5683,
+      "step": 20633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.75197630783665e-06,
+      "loss": 0.6449,
+      "step": 20634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7511261778358487e-06,
+      "loss": 0.6702,
+      "step": 20635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.7502761219233506e-06,
+      "loss": 0.5866,
+      "step": 20636
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7494261401092346e-06,
+      "loss": 0.6358,
+      "step": 20637
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7485762324035834e-06,
+      "loss": 0.6698,
+      "step": 20638
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7477263988164648e-06,
+      "loss": 0.6456,
+      "step": 20639
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7468766393579592e-06,
+      "loss": 0.6127,
+      "step": 20640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.746026954038141e-06,
+      "loss": 0.6078,
+      "step": 20641
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7451773428670855e-06,
+      "loss": 0.6454,
+      "step": 20642
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7443278058548636e-06,
+      "loss": 0.6323,
+      "step": 20643
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.743478343011554e-06,
+      "loss": 0.6553,
+      "step": 20644
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7426289543472185e-06,
+      "loss": 0.6369,
+      "step": 20645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.741779639871933e-06,
+      "loss": 0.6399,
+      "step": 20646
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7409303995957656e-06,
+      "loss": 0.6278,
+      "step": 20647
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7400812335287874e-06,
+      "loss": 0.701,
+      "step": 20648
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7392321416810638e-06,
+      "loss": 0.6286,
+      "step": 20649
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.738383124062668e-06,
+      "loss": 0.6207,
+      "step": 20650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7375341806836573e-06,
+      "loss": 0.647,
+      "step": 20651
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7366853115541e-06,
+      "loss": 0.6251,
+      "step": 20652
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7358365166840627e-06,
+      "loss": 0.6075,
+      "step": 20653
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.734987796083607e-06,
+      "loss": 0.5808,
+      "step": 20654
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7341391497627955e-06,
+      "loss": 0.6684,
+      "step": 20655
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7332905777316954e-06,
+      "loss": 0.6045,
+      "step": 20656
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.732442080000359e-06,
+      "loss": 0.602,
+      "step": 20657
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7315936565788503e-06,
+      "loss": 0.6225,
+      "step": 20658
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7307453074772284e-06,
+      "loss": 0.6789,
+      "step": 20659
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7298970327055517e-06,
+      "loss": 0.6451,
+      "step": 20660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.729048832273877e-06,
+      "loss": 0.6776,
+      "step": 20661
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7282007061922654e-06,
+      "loss": 0.6141,
+      "step": 20662
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.727352654470764e-06,
+      "loss": 0.6129,
+      "step": 20663
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7265046771194322e-06,
+      "loss": 0.5679,
+      "step": 20664
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7256567741483242e-06,
+      "loss": 0.6986,
+      "step": 20665
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7248089455674905e-06,
+      "loss": 0.6739,
+      "step": 20666
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7239611913869867e-06,
+      "loss": 0.611,
+      "step": 20667
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.723113511616865e-06,
+      "loss": 0.6951,
+      "step": 20668
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7222659062671683e-06,
+      "loss": 0.5668,
+      "step": 20669
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7214183753479517e-06,
+      "loss": 0.6469,
+      "step": 20670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.720570918869263e-06,
+      "loss": 0.6869,
+      "step": 20671
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.719723536841149e-06,
+      "loss": 0.631,
+      "step": 20672
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.718876229273658e-06,
+      "loss": 0.6355,
+      "step": 20673
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7180289961768367e-06,
+      "loss": 0.6257,
+      "step": 20674
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.717181837560726e-06,
+      "loss": 0.5925,
+      "step": 20675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7163347534353733e-06,
+      "loss": 0.6017,
+      "step": 20676
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7154877438108193e-06,
+      "loss": 0.6536,
+      "step": 20677
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7146408086971087e-06,
+      "loss": 0.6109,
+      "step": 20678
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.713793948104283e-06,
+      "loss": 0.6467,
+      "step": 20679
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.712947162042385e-06,
+      "loss": 0.5761,
+      "step": 20680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7121004505214476e-06,
+      "loss": 0.6442,
+      "step": 20681
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.711253813551514e-06,
+      "loss": 0.5838,
+      "step": 20682
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7104072511426224e-06,
+      "loss": 0.6643,
+      "step": 20683
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.709560763304808e-06,
+      "loss": 0.6547,
+      "step": 20684
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7087143500481093e-06,
+      "loss": 0.5956,
+      "step": 20685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.707868011382564e-06,
+      "loss": 0.6298,
+      "step": 20686
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.707021747318199e-06,
+      "loss": 0.5679,
+      "step": 20687
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7061755578650548e-06,
+      "loss": 0.5726,
+      "step": 20688
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7053294430331555e-06,
+      "loss": 0.62,
+      "step": 20689
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7044834028325406e-06,
+      "loss": 0.6394,
+      "step": 20690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7036374372732433e-06,
+      "loss": 0.6273,
+      "step": 20691
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.702791546365285e-06,
+      "loss": 0.6235,
+      "step": 20692
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.701945730118699e-06,
+      "loss": 0.6005,
+      "step": 20693
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7010999885435173e-06,
+      "loss": 0.6254,
+      "step": 20694
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7002543216497553e-06,
+      "loss": 0.5897,
+      "step": 20695
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6994087294474523e-06,
+      "loss": 0.5495,
+      "step": 20696
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.698563211946633e-06,
+      "loss": 0.6969,
+      "step": 20697
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.697717769157313e-06,
+      "loss": 0.6544,
+      "step": 20698
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6968724010895228e-06,
+      "loss": 0.6,
+      "step": 20699
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.696027107753286e-06,
+      "loss": 0.6105,
+      "step": 20700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6951818891586167e-06,
+      "loss": 0.6446,
+      "step": 20701
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6943367453155445e-06,
+      "loss": 0.6542,
+      "step": 20702
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.69349167623409e-06,
+      "loss": 0.6015,
+      "step": 20703
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.692646681924267e-06,
+      "loss": 0.6279,
+      "step": 20704
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6918017623960957e-06,
+      "loss": 0.6277,
+      "step": 20705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6909569176595983e-06,
+      "loss": 0.6664,
+      "step": 20706
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6901121477247804e-06,
+      "loss": 0.63,
+      "step": 20707
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.689267452601668e-06,
+      "loss": 0.5869,
+      "step": 20708
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.688422832300277e-06,
+      "loss": 0.6229,
+      "step": 20709
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6875782868306143e-06,
+      "loss": 0.6493,
+      "step": 20710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6867338162026967e-06,
+      "loss": 0.659,
+      "step": 20711
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.685889420426538e-06,
+      "loss": 0.6437,
+      "step": 20712
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6850450995121425e-06,
+      "loss": 0.6081,
+      "step": 20713
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6842008534695284e-06,
+      "loss": 0.6203,
+      "step": 20714
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6833566823087053e-06,
+      "loss": 0.6515,
+      "step": 20715
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.682512586039677e-06,
+      "loss": 0.6258,
+      "step": 20716
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6816685646724537e-06,
+      "loss": 0.6297,
+      "step": 20717
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6808246182170447e-06,
+      "loss": 0.5795,
+      "step": 20718
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.679980746683448e-06,
+      "loss": 0.6687,
+      "step": 20719
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6791369500816765e-06,
+      "loss": 0.6415,
+      "step": 20720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6782932284217377e-06,
+      "loss": 0.6525,
+      "step": 20721
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6774495817136247e-06,
+      "loss": 0.6312,
+      "step": 20722
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6766060099673453e-06,
+      "loss": 0.6837,
+      "step": 20723
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.675762513192904e-06,
+      "loss": 0.6368,
+      "step": 20724
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.674919091400295e-06,
+      "loss": 0.6263,
+      "step": 20725
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.674075744599519e-06,
+      "loss": 0.6552,
+      "step": 20726
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6732324728005832e-06,
+      "loss": 0.5961,
+      "step": 20727
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.672389276013477e-06,
+      "loss": 0.6062,
+      "step": 20728
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6715461542482e-06,
+      "loss": 0.7072,
+      "step": 20729
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6707031075147527e-06,
+      "loss": 0.6043,
+      "step": 20730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.669860135823122e-06,
+      "loss": 0.6562,
+      "step": 20731
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6690172391833055e-06,
+      "loss": 0.6242,
+      "step": 20732
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6681744176053034e-06,
+      "loss": 0.6555,
+      "step": 20733
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6673316710991014e-06,
+      "loss": 0.6676,
+      "step": 20734
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6664889996746935e-06,
+      "loss": 0.6353,
+      "step": 20735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6656464033420725e-06,
+      "loss": 0.6459,
+      "step": 20736
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.664803882111223e-06,
+      "loss": 0.6683,
+      "step": 20737
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6639614359921347e-06,
+      "loss": 0.5954,
+      "step": 20738
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6631190649948057e-06,
+      "loss": 0.63,
+      "step": 20739
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6622767691292127e-06,
+      "loss": 0.6512,
+      "step": 20740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.661434548405346e-06,
+      "loss": 0.5671,
+      "step": 20741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6605924028331953e-06,
+      "loss": 0.6426,
+      "step": 20742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6597503324227367e-06,
+      "loss": 0.63,
+      "step": 20743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.658908337183955e-06,
+      "loss": 0.6708,
+      "step": 20744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.658066417126843e-06,
+      "loss": 0.6509,
+      "step": 20745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.657224572261373e-06,
+      "loss": 0.6453,
+      "step": 20746
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.65638280259753e-06,
+      "loss": 0.6041,
+      "step": 20747
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6555411081452962e-06,
+      "loss": 0.6253,
+      "step": 20748
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6546994889146446e-06,
+      "loss": 0.6494,
+      "step": 20749
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6538579449155585e-06,
+      "loss": 0.6439,
+      "step": 20750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6530164761580145e-06,
+      "loss": 0.6541,
+      "step": 20751
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6521750826519888e-06,
+      "loss": 0.6348,
+      "step": 20752
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.65133376440746e-06,
+      "loss": 0.6516,
+      "step": 20753
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6504925214343978e-06,
+      "loss": 0.6217,
+      "step": 20754
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6496513537427793e-06,
+      "loss": 0.5805,
+      "step": 20755
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.648810261342577e-06,
+      "loss": 0.6438,
+      "step": 20756
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6479692442437632e-06,
+      "loss": 0.6216,
+      "step": 20757
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6471283024563087e-06,
+      "loss": 0.643,
+      "step": 20758
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.646287435990189e-06,
+      "loss": 0.6638,
+      "step": 20759
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.645446644855366e-06,
+      "loss": 0.6363,
+      "step": 20760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6446059290618106e-06,
+      "loss": 0.6721,
+      "step": 20761
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.643765288619492e-06,
+      "loss": 0.6167,
+      "step": 20762
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.642924723538377e-06,
+      "loss": 0.6718,
+      "step": 20763
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6420842338284314e-06,
+      "loss": 0.6365,
+      "step": 20764
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6412438194996226e-06,
+      "loss": 0.6761,
+      "step": 20765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6404034805619084e-06,
+      "loss": 0.6364,
+      "step": 20766
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.639563217025257e-06,
+      "loss": 0.5906,
+      "step": 20767
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.638723028899629e-06,
+      "loss": 0.658,
+      "step": 20768
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.637882916194986e-06,
+      "loss": 0.634,
+      "step": 20769
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6370428789212898e-06,
+      "loss": 0.5941,
+      "step": 20770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6362029170885026e-06,
+      "loss": 0.6109,
+      "step": 20771
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.635363030706578e-06,
+      "loss": 0.6319,
+      "step": 20772
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.634523219785474e-06,
+      "loss": 0.6472,
+      "step": 20773
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.633683484335151e-06,
+      "loss": 0.6768,
+      "step": 20774
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.632843824365563e-06,
+      "loss": 0.6537,
+      "step": 20775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.632004239886666e-06,
+      "loss": 0.6233,
+      "step": 20776
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6311647309084185e-06,
+      "loss": 0.6725,
+      "step": 20777
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6303252974407654e-06,
+      "loss": 0.6597,
+      "step": 20778
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6294859394936644e-06,
+      "loss": 0.646,
+      "step": 20779
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.628646657077065e-06,
+      "loss": 0.6354,
+      "step": 20780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.627807450200921e-06,
+      "loss": 0.6589,
+      "step": 20781
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6269683188751792e-06,
+      "loss": 0.654,
+      "step": 20782
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.626129263109793e-06,
+      "loss": 0.6976,
+      "step": 20783
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6252902829147053e-06,
+      "loss": 0.6707,
+      "step": 20784
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6244513782998647e-06,
+      "loss": 0.6266,
+      "step": 20785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6236125492752184e-06,
+      "loss": 0.6716,
+      "step": 20786
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6227737958507114e-06,
+      "loss": 0.7103,
+      "step": 20787
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6219351180362883e-06,
+      "loss": 0.647,
+      "step": 20788
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6210965158418954e-06,
+      "loss": 0.5422,
+      "step": 20789
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6202579892774705e-06,
+      "loss": 0.6374,
+      "step": 20790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.619419538352956e-06,
+      "loss": 0.6493,
+      "step": 20791
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6185811630782954e-06,
+      "loss": 0.5806,
+      "step": 20792
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6177428634634272e-06,
+      "loss": 0.6508,
+      "step": 20793
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6169046395182915e-06,
+      "loss": 0.6459,
+      "step": 20794
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.616066491252829e-06,
+      "loss": 0.6773,
+      "step": 20795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6152284186769704e-06,
+      "loss": 0.6056,
+      "step": 20796
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6143904218006566e-06,
+      "loss": 0.6327,
+      "step": 20797
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6135525006338213e-06,
+      "loss": 0.6706,
+      "step": 20798
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6127146551864002e-06,
+      "loss": 0.7101,
+      "step": 20799
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.611876885468326e-06,
+      "loss": 0.6538,
+      "step": 20800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.611039191489537e-06,
+      "loss": 0.6212,
+      "step": 20801
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6102015732599572e-06,
+      "loss": 0.6076,
+      "step": 20802
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6093640307895193e-06,
+      "loss": 0.6286,
+      "step": 20803
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6085265640881563e-06,
+      "loss": 0.6062,
+      "step": 20804
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.607689173165796e-06,
+      "loss": 0.6194,
+      "step": 20805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.606851858032366e-06,
+      "loss": 0.6258,
+      "step": 20806
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6060146186977995e-06,
+      "loss": 0.6138,
+      "step": 20807
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.605177455172014e-06,
+      "loss": 0.6555,
+      "step": 20808
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6043403674649393e-06,
+      "loss": 0.5723,
+      "step": 20809
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6035033555864995e-06,
+      "loss": 0.6193,
+      "step": 20810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6026664195466197e-06,
+      "loss": 0.6677,
+      "step": 20811
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6018295593552254e-06,
+      "loss": 0.6283,
+      "step": 20812
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.6009927750222317e-06,
+      "loss": 0.6123,
+      "step": 20813
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.600156066557564e-06,
+      "loss": 0.6244,
+      "step": 20814
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5993194339711424e-06,
+      "loss": 0.6198,
+      "step": 20815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5984828772728843e-06,
+      "loss": 0.5868,
+      "step": 20816
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5976463964727103e-06,
+      "loss": 0.6174,
+      "step": 20817
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.596809991580541e-06,
+      "loss": 0.6208,
+      "step": 20818
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5959736626062858e-06,
+      "loss": 0.6065,
+      "step": 20819
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5951374095598644e-06,
+      "loss": 0.6414,
+      "step": 20820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5943012324511906e-06,
+      "loss": 0.6406,
+      "step": 20821
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5934651312901793e-06,
+      "loss": 0.6241,
+      "step": 20822
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5926291060867426e-06,
+      "loss": 0.6888,
+      "step": 20823
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.591793156850797e-06,
+      "loss": 0.7121,
+      "step": 20824
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.590957283592247e-06,
+      "loss": 0.6442,
+      "step": 20825
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5901214863210055e-06,
+      "loss": 0.628,
+      "step": 20826
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5892857650469815e-06,
+      "loss": 0.6258,
+      "step": 20827
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.588450119780086e-06,
+      "loss": 0.6565,
+      "step": 20828
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.587614550530224e-06,
+      "loss": 0.6766,
+      "step": 20829
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.586779057307306e-06,
+      "loss": 0.5911,
+      "step": 20830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.585943640121232e-06,
+      "loss": 0.6574,
+      "step": 20831
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5851082989819107e-06,
+      "loss": 0.6495,
+      "step": 20832
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.584273033899245e-06,
+      "loss": 0.6516,
+      "step": 20833
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5834378448831385e-06,
+      "loss": 0.5959,
+      "step": 20834
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.582602731943493e-06,
+      "loss": 0.6606,
+      "step": 20835
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5817676950902136e-06,
+      "loss": 0.6715,
+      "step": 20836
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5809327343331946e-06,
+      "loss": 0.6242,
+      "step": 20837
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.580097849682338e-06,
+      "loss": 0.6091,
+      "step": 20838
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5792630411475415e-06,
+      "loss": 0.5857,
+      "step": 20839
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.578428308738705e-06,
+      "loss": 0.597,
+      "step": 20840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.577593652465724e-06,
+      "loss": 0.5746,
+      "step": 20841
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.576759072338498e-06,
+      "loss": 0.6058,
+      "step": 20842
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5759245683669153e-06,
+      "loss": 0.6668,
+      "step": 20843
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5750901405608727e-06,
+      "loss": 0.6669,
+      "step": 20844
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5742557889302653e-06,
+      "loss": 0.6398,
+      "step": 20845
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5734215134849835e-06,
+      "loss": 0.6647,
+      "step": 20846
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.572587314234919e-06,
+      "loss": 0.6204,
+      "step": 20847
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.571753191189966e-06,
+      "loss": 0.6528,
+      "step": 20848
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5709191443600076e-06,
+      "loss": 0.6533,
+      "step": 20849
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5700851737549357e-06,
+      "loss": 0.6056,
+      "step": 20850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.569251279384638e-06,
+      "loss": 0.6634,
+      "step": 20851
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5684174612590016e-06,
+      "loss": 0.6339,
+      "step": 20852
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5675837193879113e-06,
+      "loss": 0.6751,
+      "step": 20853
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5667500537812573e-06,
+      "loss": 0.6296,
+      "step": 20854
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5659164644489152e-06,
+      "loss": 0.6264,
+      "step": 20855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5650829514007733e-06,
+      "loss": 0.6315,
+      "step": 20856
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5642495146467126e-06,
+      "loss": 0.6512,
+      "step": 20857
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5634161541966162e-06,
+      "loss": 0.6136,
+      "step": 20858
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5625828700603624e-06,
+      "loss": 0.5859,
+      "step": 20859
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5617496622478364e-06,
+      "loss": 0.6635,
+      "step": 20860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5609165307689087e-06,
+      "loss": 0.6427,
+      "step": 20861
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.560083475633461e-06,
+      "loss": 0.6175,
+      "step": 20862
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5592504968513697e-06,
+      "loss": 0.6133,
+      "step": 20863
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5584175944325126e-06,
+      "loss": 0.6585,
+      "step": 20864
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5575847683867627e-06,
+      "loss": 0.6255,
+      "step": 20865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.556752018723999e-06,
+      "loss": 0.6699,
+      "step": 20866
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5559193454540875e-06,
+      "loss": 0.6461,
+      "step": 20867
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5550867485869035e-06,
+      "loss": 0.6633,
+      "step": 20868
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.554254228132319e-06,
+      "loss": 0.6454,
+      "step": 20869
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.553421784100204e-06,
+      "loss": 0.6743,
+      "step": 20870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5525894165004294e-06,
+      "loss": 0.5712,
+      "step": 20871
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5517571253428673e-06,
+      "loss": 0.6674,
+      "step": 20872
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5509249106373767e-06,
+      "loss": 0.6283,
+      "step": 20873
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5500927723938307e-06,
+      "loss": 0.6106,
+      "step": 20874
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5492607106220932e-06,
+      "loss": 0.6244,
+      "step": 20875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5484287253320302e-06,
+      "loss": 0.6214,
+      "step": 20876
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5475968165335105e-06,
+      "loss": 0.6637,
+      "step": 20877
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5467649842363884e-06,
+      "loss": 0.6225,
+      "step": 20878
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.545933228450531e-06,
+      "loss": 0.6314,
+      "step": 20879
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5451015491858e-06,
+      "loss": 0.5946,
+      "step": 20880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.544269946452055e-06,
+      "loss": 0.6595,
+      "step": 20881
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5434384202591564e-06,
+      "loss": 0.5778,
+      "step": 20882
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5426069706169664e-06,
+      "loss": 0.6451,
+      "step": 20883
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5417755975353363e-06,
+      "loss": 0.6584,
+      "step": 20884
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5409443010241264e-06,
+      "loss": 0.6458,
+      "step": 20885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5401130810931915e-06,
+      "loss": 0.5892,
+      "step": 20886
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.539281937752389e-06,
+      "loss": 0.6541,
+      "step": 20887
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5384508710115716e-06,
+      "loss": 0.6152,
+      "step": 20888
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.537619880880596e-06,
+      "loss": 0.6332,
+      "step": 20889
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.536788967369309e-06,
+      "loss": 0.6065,
+      "step": 20890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5359581304875645e-06,
+      "loss": 0.6082,
+      "step": 20891
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.535127370245213e-06,
+      "loss": 0.6029,
+      "step": 20892
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5342966866521047e-06,
+      "loss": 0.6599,
+      "step": 20893
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5334660797180887e-06,
+      "loss": 0.6432,
+      "step": 20894
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.532635549453016e-06,
+      "loss": 0.643,
+      "step": 20895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5318050958667262e-06,
+      "loss": 0.5579,
+      "step": 20896
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.53097471896907e-06,
+      "loss": 0.6006,
+      "step": 20897
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.530144418769892e-06,
+      "loss": 0.6545,
+      "step": 20898
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5293141952790357e-06,
+      "loss": 0.6338,
+      "step": 20899
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5284840485063444e-06,
+      "loss": 0.6859,
+      "step": 20900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.527653978461666e-06,
+      "loss": 0.6268,
+      "step": 20901
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5268239851548325e-06,
+      "loss": 0.6295,
+      "step": 20902
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5259940685956906e-06,
+      "loss": 0.653,
+      "step": 20903
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.525164228794078e-06,
+      "loss": 0.6463,
+      "step": 20904
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5243344657598346e-06,
+      "loss": 0.6549,
+      "step": 20905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.523504779502798e-06,
+      "loss": 0.625,
+      "step": 20906
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.522675170032808e-06,
+      "loss": 0.6337,
+      "step": 20907
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.521845637359693e-06,
+      "loss": 0.6454,
+      "step": 20908
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.521016181493294e-06,
+      "loss": 0.7001,
+      "step": 20909
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.520186802443445e-06,
+      "loss": 0.6096,
+      "step": 20910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5193575002199775e-06,
+      "loss": 0.5981,
+      "step": 20911
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5185282748327254e-06,
+      "loss": 0.6525,
+      "step": 20912
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5176991262915227e-06,
+      "loss": 0.6234,
+      "step": 20913
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.516870054606194e-06,
+      "loss": 0.6288,
+      "step": 20914
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.516041059786571e-06,
+      "loss": 0.6291,
+      "step": 20915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5152121418424845e-06,
+      "loss": 0.6348,
+      "step": 20916
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5143833007837614e-06,
+      "loss": 0.6235,
+      "step": 20917
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5135545366202284e-06,
+      "loss": 0.6564,
+      "step": 20918
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5127258493617156e-06,
+      "loss": 0.629,
+      "step": 20919
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5118972390180406e-06,
+      "loss": 0.6215,
+      "step": 20920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.511068705599031e-06,
+      "loss": 0.6135,
+      "step": 20921
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.510240249114514e-06,
+      "loss": 0.6465,
+      "step": 20922
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5094118695743006e-06,
+      "loss": 0.6921,
+      "step": 20923
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.508583566988225e-06,
+      "loss": 0.5994,
+      "step": 20924
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.5077553413661056e-06,
+      "loss": 0.682,
+      "step": 20925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.506927192717756e-06,
+      "loss": 0.618,
+      "step": 20926
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.506099121052999e-06,
+      "loss": 0.6019,
+      "step": 20927
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.505271126381654e-06,
+      "loss": 0.6289,
+      "step": 20928
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5044432087135307e-06,
+      "loss": 0.6175,
+      "step": 20929
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5036153680584517e-06,
+      "loss": 0.6344,
+      "step": 20930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5027876044262343e-06,
+      "loss": 0.556,
+      "step": 20931
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5019599178266862e-06,
+      "loss": 0.6376,
+      "step": 20932
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5011323082696226e-06,
+      "loss": 0.588,
+      "step": 20933
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.50030477576486e-06,
+      "loss": 0.6534,
+      "step": 20934
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4994773203222e-06,
+      "loss": 0.6496,
+      "step": 20935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4986499419514662e-06,
+      "loss": 0.6067,
+      "step": 20936
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4978226406624583e-06,
+      "loss": 0.6042,
+      "step": 20937
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4969954164649877e-06,
+      "loss": 0.6376,
+      "step": 20938
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.496168269368867e-06,
+      "loss": 0.6256,
+      "step": 20939
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4953411993838947e-06,
+      "loss": 0.6635,
+      "step": 20940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4945142065198777e-06,
+      "loss": 0.5964,
+      "step": 20941
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4936872907866314e-06,
+      "loss": 0.6396,
+      "step": 20942
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.492860452193949e-06,
+      "loss": 0.5995,
+      "step": 20943
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.492033690751638e-06,
+      "loss": 0.6863,
+      "step": 20944
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4912070064695035e-06,
+      "loss": 0.6543,
+      "step": 20945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4903803993573406e-06,
+      "loss": 0.6199,
+      "step": 20946
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4895538694249487e-06,
+      "loss": 0.5736,
+      "step": 20947
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.488727416682138e-06,
+      "loss": 0.591,
+      "step": 20948
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4879010411386982e-06,
+      "loss": 0.5662,
+      "step": 20949
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.487074742804428e-06,
+      "loss": 0.6311,
+      "step": 20950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4862485216891307e-06,
+      "loss": 0.6268,
+      "step": 20951
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4854223778025932e-06,
+      "loss": 0.6394,
+      "step": 20952
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.48459631115461e-06,
+      "loss": 0.6324,
+      "step": 20953
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.483770321754987e-06,
+      "loss": 0.6367,
+      "step": 20954
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4829444096135066e-06,
+      "loss": 0.5973,
+      "step": 20955
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4821185747399645e-06,
+      "loss": 0.6208,
+      "step": 20956
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.481292817144155e-06,
+      "loss": 0.686,
+      "step": 20957
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.480467136835861e-06,
+      "loss": 0.5632,
+      "step": 20958
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4796415338248745e-06,
+      "loss": 0.6085,
+      "step": 20959
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.478816008120992e-06,
+      "loss": 0.6272,
+      "step": 20960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4779905597339923e-06,
+      "loss": 0.6508,
+      "step": 20961
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4771651886736646e-06,
+      "loss": 0.6812,
+      "step": 20962
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4763398949497973e-06,
+      "loss": 0.6225,
+      "step": 20963
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4755146785721704e-06,
+      "loss": 0.5988,
+      "step": 20964
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4746895395505674e-06,
+      "loss": 0.5872,
+      "step": 20965
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4738644778947816e-06,
+      "loss": 0.6183,
+      "step": 20966
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.473039493614584e-06,
+      "loss": 0.6309,
+      "step": 20967
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.47221458671976e-06,
+      "loss": 0.6427,
+      "step": 20968
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4713897572200925e-06,
+      "loss": 0.6172,
+      "step": 20969
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.470565005125356e-06,
+      "loss": 0.6031,
+      "step": 20970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.469740330445327e-06,
+      "loss": 0.6029,
+      "step": 20971
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.468915733189794e-06,
+      "loss": 0.6077,
+      "step": 20972
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4680912133685242e-06,
+      "loss": 0.59,
+      "step": 20973
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4672667709912956e-06,
+      "loss": 0.6506,
+      "step": 20974
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4664424060678857e-06,
+      "loss": 0.63,
+      "step": 20975
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.465618118608064e-06,
+      "loss": 0.629,
+      "step": 20976
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.464793908621601e-06,
+      "loss": 0.6481,
+      "step": 20977
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4639697761182812e-06,
+      "loss": 0.6585,
+      "step": 20978
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4631457211078634e-06,
+      "loss": 0.6935,
+      "step": 20979
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4623217436001233e-06,
+      "loss": 0.6588,
+      "step": 20980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.461497843604832e-06,
+      "loss": 0.5883,
+      "step": 20981
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.460674021131751e-06,
+      "loss": 0.6173,
+      "step": 20982
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4598502761906495e-06,
+      "loss": 0.6479,
+      "step": 20983
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4590266087913026e-06,
+      "loss": 0.5968,
+      "step": 20984
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4582030189434667e-06,
+      "loss": 0.6281,
+      "step": 20985
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4573795066569093e-06,
+      "loss": 0.6702,
+      "step": 20986
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.456556071941397e-06,
+      "loss": 0.6566,
+      "step": 20987
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4557327148066867e-06,
+      "loss": 0.6775,
+      "step": 20988
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.454909435262541e-06,
+      "loss": 0.6379,
+      "step": 20989
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4540862333187296e-06,
+      "loss": 0.6691,
+      "step": 20990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4532631089850032e-06,
+      "loss": 0.7043,
+      "step": 20991
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.452440062271124e-06,
+      "loss": 0.5889,
+      "step": 20992
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4516170931868544e-06,
+      "loss": 0.6125,
+      "step": 20993
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.450794201741945e-06,
+      "loss": 0.5912,
+      "step": 20994
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4499713879461515e-06,
+      "loss": 0.6395,
+      "step": 20995
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.449148651809239e-06,
+      "loss": 0.6385,
+      "step": 20996
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.448325993340953e-06,
+      "loss": 0.6869,
+      "step": 20997
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4475034125510543e-06,
+      "loss": 0.6297,
+      "step": 20998
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.446680909449288e-06,
+      "loss": 0.6424,
+      "step": 20999
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.445858484045409e-06,
+      "loss": 0.6202,
+      "step": 21000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4450361363491693e-06,
+      "loss": 0.6632,
+      "step": 21001
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4442138663703182e-06,
+      "loss": 0.6325,
+      "step": 21002
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4433916741186045e-06,
+      "loss": 0.6657,
+      "step": 21003
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.44256955960378e-06,
+      "loss": 0.6445,
+      "step": 21004
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4417475228355855e-06,
+      "loss": 0.6093,
+      "step": 21005
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4409255638237705e-06,
+      "loss": 0.5848,
+      "step": 21006
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.44010368257808e-06,
+      "loss": 0.6372,
+      "step": 21007
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4392818791082583e-06,
+      "loss": 0.6058,
+      "step": 21008
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4384601534240492e-06,
+      "loss": 0.6252,
+      "step": 21009
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.437638505535199e-06,
+      "loss": 0.637,
+      "step": 21010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4368169354514423e-06,
+      "loss": 0.6714,
+      "step": 21011
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4359954431825236e-06,
+      "loss": 0.6114,
+      "step": 21012
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.435174028738183e-06,
+      "loss": 0.6583,
+      "step": 21013
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.434352692128158e-06,
+      "loss": 0.6107,
+      "step": 21014
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4335314333621873e-06,
+      "loss": 0.6517,
+      "step": 21015
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.432710252450012e-06,
+      "loss": 0.6533,
+      "step": 21016
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4318891494013607e-06,
+      "loss": 0.6315,
+      "step": 21017
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4310681242259715e-06,
+      "loss": 0.6275,
+      "step": 21018
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.430247176933581e-06,
+      "loss": 0.6433,
+      "step": 21019
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4294263075339196e-06,
+      "loss": 0.6223,
+      "step": 21020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4286055160367216e-06,
+      "loss": 0.6111,
+      "step": 21021
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.427784802451721e-06,
+      "loss": 0.6745,
+      "step": 21022
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4269641667886434e-06,
+      "loss": 0.5814,
+      "step": 21023
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.42614360905722e-06,
+      "loss": 0.6146,
+      "step": 21024
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4253231292671796e-06,
+      "loss": 0.6074,
+      "step": 21025
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4245027274282504e-06,
+      "loss": 0.5834,
+      "step": 21026
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4236824035501603e-06,
+      "loss": 0.6873,
+      "step": 21027
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.422862157642637e-06,
+      "loss": 0.614,
+      "step": 21028
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4220419897154e-06,
+      "loss": 0.685,
+      "step": 21029
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4212218997781754e-06,
+      "loss": 0.7026,
+      "step": 21030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4204018878406887e-06,
+      "loss": 0.6307,
+      "step": 21031
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4195819539126597e-06,
+      "loss": 0.6366,
+      "step": 21032
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.418762098003812e-06,
+      "loss": 0.6129,
+      "step": 21033
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4179423201238673e-06,
+      "loss": 0.6098,
+      "step": 21034
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4171226202825403e-06,
+      "loss": 0.5856,
+      "step": 21035
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4163029984895515e-06,
+      "loss": 0.5976,
+      "step": 21036
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4154834547546188e-06,
+      "loss": 0.615,
+      "step": 21037
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.41466398908746e-06,
+      "loss": 0.6631,
+      "step": 21038
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.41384460149779e-06,
+      "loss": 0.6341,
+      "step": 21039
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4130252919953265e-06,
+      "loss": 0.6062,
+      "step": 21040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4122060605897777e-06,
+      "loss": 0.6433,
+      "step": 21041
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.411386907290859e-06,
+      "loss": 0.6264,
+      "step": 21042
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4105678321082845e-06,
+      "loss": 0.6046,
+      "step": 21043
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4097488350517626e-06,
+      "loss": 0.6333,
+      "step": 21044
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.408929916131006e-06,
+      "loss": 0.6127,
+      "step": 21045
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.408111075355727e-06,
+      "loss": 0.6635,
+      "step": 21046
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.407292312735625e-06,
+      "loss": 0.6262,
+      "step": 21047
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.406473628280413e-06,
+      "loss": 0.6045,
+      "step": 21048
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4056550219997975e-06,
+      "loss": 0.6702,
+      "step": 21049
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4048364939034827e-06,
+      "loss": 0.6251,
+      "step": 21050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.404018044001175e-06,
+      "loss": 0.6025,
+      "step": 21051
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4031996723025805e-06,
+      "loss": 0.6607,
+      "step": 21052
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4023813788173955e-06,
+      "loss": 0.694,
+      "step": 21053
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4015631635553246e-06,
+      "loss": 0.6282,
+      "step": 21054
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.40074502652607e-06,
+      "loss": 0.6411,
+      "step": 21055
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3999269677393322e-06,
+      "loss": 0.6583,
+      "step": 21056
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3991089872048078e-06,
+      "loss": 0.6292,
+      "step": 21057
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3982910849322004e-06,
+      "loss": 0.6472,
+      "step": 21058
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3974732609312e-06,
+      "loss": 0.631,
+      "step": 21059
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3966555152115068e-06,
+      "loss": 0.6798,
+      "step": 21060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.395837847782816e-06,
+      "loss": 0.6576,
+      "step": 21061
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.395020258654821e-06,
+      "loss": 0.6522,
+      "step": 21062
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3942027478372196e-06,
+      "loss": 0.6703,
+      "step": 21063
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3933853153396977e-06,
+      "loss": 0.6894,
+      "step": 21064
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.39256796117195e-06,
+      "loss": 0.6674,
+      "step": 21065
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3917506853436675e-06,
+      "loss": 0.634,
+      "step": 21066
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3909334878645416e-06,
+      "loss": 0.6179,
+      "step": 21067
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3901163687442582e-06,
+      "loss": 0.6921,
+      "step": 21068
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.38929932799251e-06,
+      "loss": 0.6414,
+      "step": 21069
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.388482365618979e-06,
+      "loss": 0.6258,
+      "step": 21070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.387665481633352e-06,
+      "loss": 0.6191,
+      "step": 21071
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3868486760453157e-06,
+      "loss": 0.648,
+      "step": 21072
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3860319488645544e-06,
+      "loss": 0.636,
+      "step": 21073
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3852153001007506e-06,
+      "loss": 0.6217,
+      "step": 21074
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3843987297635906e-06,
+      "loss": 0.6255,
+      "step": 21075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.383582237862749e-06,
+      "loss": 0.6491,
+      "step": 21076
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3827658244079096e-06,
+      "loss": 0.6288,
+      "step": 21077
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3819494894087523e-06,
+      "loss": 0.6631,
+      "step": 21078
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3811332328749558e-06,
+      "loss": 0.5884,
+      "step": 21079
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.380317054816198e-06,
+      "loss": 0.6004,
+      "step": 21080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.379500955242159e-06,
+      "loss": 0.5747,
+      "step": 21081
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3786849341625072e-06,
+      "loss": 0.6494,
+      "step": 21082
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.377868991586921e-06,
+      "loss": 0.6475,
+      "step": 21083
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3770531275250763e-06,
+      "loss": 0.6205,
+      "step": 21084
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3762373419866432e-06,
+      "loss": 0.6017,
+      "step": 21085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3754216349812972e-06,
+      "loss": 0.6281,
+      "step": 21086
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.37460600651871e-06,
+      "loss": 0.6445,
+      "step": 21087
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3737904566085465e-06,
+      "loss": 0.5806,
+      "step": 21088
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.37297498526048e-06,
+      "loss": 0.6361,
+      "step": 21089
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.372159592484179e-06,
+      "loss": 0.6655,
+      "step": 21090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3713442782893093e-06,
+      "loss": 0.6049,
+      "step": 21091
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3705290426855375e-06,
+      "loss": 0.6346,
+      "step": 21092
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3697138856825353e-06,
+      "loss": 0.6307,
+      "step": 21093
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3688988072899577e-06,
+      "loss": 0.6525,
+      "step": 21094
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.368083807517474e-06,
+      "loss": 0.6617,
+      "step": 21095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.367268886374745e-06,
+      "loss": 0.7749,
+      "step": 21096
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3664540438714357e-06,
+      "loss": 0.6352,
+      "step": 21097
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3656392800172034e-06,
+      "loss": 0.6077,
+      "step": 21098
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.364824594821714e-06,
+      "loss": 0.6434,
+      "step": 21099
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3640099882946198e-06,
+      "loss": 0.6095,
+      "step": 21100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3631954604455807e-06,
+      "loss": 0.5617,
+      "step": 21101
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.362381011284256e-06,
+      "loss": 0.6207,
+      "step": 21102
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3615666408202996e-06,
+      "loss": 0.6282,
+      "step": 21103
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.36075234906337e-06,
+      "loss": 0.6914,
+      "step": 21104
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.359938136023122e-06,
+      "loss": 0.6372,
+      "step": 21105
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3591240017092053e-06,
+      "loss": 0.6638,
+      "step": 21106
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.358309946131273e-06,
+      "loss": 0.6629,
+      "step": 21107
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.357495969298978e-06,
+      "loss": 0.6839,
+      "step": 21108
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3566820712219716e-06,
+      "loss": 0.6501,
+      "step": 21109
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.355868251909904e-06,
+      "loss": 0.6119,
+      "step": 21110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3550545113724253e-06,
+      "loss": 0.63,
+      "step": 21111
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.354240849619178e-06,
+      "loss": 0.5963,
+      "step": 21112
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3534272666598124e-06,
+      "loss": 0.6537,
+      "step": 21113
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3526137625039745e-06,
+      "loss": 0.651,
+      "step": 21114
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3518003371613095e-06,
+      "loss": 0.5845,
+      "step": 21115
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3509869906414617e-06,
+      "loss": 0.6451,
+      "step": 21116
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3501737229540775e-06,
+      "loss": 0.6833,
+      "step": 21117
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.349360534108792e-06,
+      "loss": 0.607,
+      "step": 21118
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3485474241152514e-06,
+      "loss": 0.6596,
+      "step": 21119
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.347734392983095e-06,
+      "loss": 0.5705,
+      "step": 21120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3469214407219617e-06,
+      "loss": 0.5733,
+      "step": 21121
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.346108567341495e-06,
+      "loss": 0.6313,
+      "step": 21122
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3452957728513247e-06,
+      "loss": 0.6234,
+      "step": 21123
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.344483057261092e-06,
+      "loss": 0.6469,
+      "step": 21124
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.343670420580435e-06,
+      "loss": 0.7004,
+      "step": 21125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.342857862818978e-06,
+      "loss": 0.6452,
+      "step": 21126
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3420453839863664e-06,
+      "loss": 0.6409,
+      "step": 21127
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.341232984092232e-06,
+      "loss": 0.7064,
+      "step": 21128
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.340420663146201e-06,
+      "loss": 0.5995,
+      "step": 21129
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.339608421157907e-06,
+      "loss": 0.617,
+      "step": 21130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.338796258136985e-06,
+      "loss": 0.6628,
+      "step": 21131
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.337984174093053e-06,
+      "loss": 0.691,
+      "step": 21132
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3371721690357483e-06,
+      "loss": 0.6374,
+      "step": 21133
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3363602429747e-06,
+      "loss": 0.6774,
+      "step": 21134
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.335548395919528e-06,
+      "loss": 0.6046,
+      "step": 21135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.334736627879859e-06,
+      "loss": 0.6633,
+      "step": 21136
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.333924938865323e-06,
+      "loss": 0.62,
+      "step": 21137
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3331133288855323e-06,
+      "loss": 0.6253,
+      "step": 21138
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3323017979501206e-06,
+      "loss": 0.626,
+      "step": 21139
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.331490346068709e-06,
+      "loss": 0.641,
+      "step": 21140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3306789732509115e-06,
+      "loss": 0.6273,
+      "step": 21141
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.329867679506351e-06,
+      "loss": 0.6706,
+      "step": 21142
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.329056464844651e-06,
+      "loss": 0.6555,
+      "step": 21143
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3282453292754193e-06,
+      "loss": 0.6277,
+      "step": 21144
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.327434272808281e-06,
+      "loss": 0.6529,
+      "step": 21145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3266232954528543e-06,
+      "loss": 0.6656,
+      "step": 21146
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.325812397218747e-06,
+      "loss": 0.6044,
+      "step": 21147
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3250015781155763e-06,
+      "loss": 0.6334,
+      "step": 21148
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3241908381529596e-06,
+      "loss": 0.6546,
+      "step": 21149
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.323380177340498e-06,
+      "loss": 0.6283,
+      "step": 21150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.322569595687816e-06,
+      "loss": 0.6421,
+      "step": 21151
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3217590932045198e-06,
+      "loss": 0.5841,
+      "step": 21152
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3209486699002136e-06,
+      "loss": 0.6304,
+      "step": 21153
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.320138325784511e-06,
+      "loss": 0.6412,
+      "step": 21154
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.319328060867022e-06,
+      "loss": 0.5764,
+      "step": 21155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3185178751573434e-06,
+      "loss": 0.6458,
+      "step": 21156
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3177077686650905e-06,
+      "loss": 0.5925,
+      "step": 21157
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3168977413998683e-06,
+      "loss": 0.6653,
+      "step": 21158
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3160877933712744e-06,
+      "loss": 0.6542,
+      "step": 21159
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3152779245889143e-06,
+      "loss": 0.6249,
+      "step": 21160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.314468135062394e-06,
+      "loss": 0.5904,
+      "step": 21161
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.313658424801306e-06,
+      "loss": 0.5965,
+      "step": 21162
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3128487938152575e-06,
+      "loss": 0.6198,
+      "step": 21163
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.31203924211385e-06,
+      "loss": 0.6716,
+      "step": 21164
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.311229769706674e-06,
+      "loss": 0.5978,
+      "step": 21165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3104203766033305e-06,
+      "loss": 0.6434,
+      "step": 21166
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3096110628134194e-06,
+      "loss": 0.6108,
+      "step": 21167
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3088018283465253e-06,
+      "loss": 0.627,
+      "step": 21168
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3079926732122537e-06,
+      "loss": 0.5663,
+      "step": 21169
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.307183597420197e-06,
+      "loss": 0.6704,
+      "step": 21170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3063746009799426e-06,
+      "loss": 0.6554,
+      "step": 21171
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.305565683901084e-06,
+      "loss": 0.6457,
+      "step": 21172
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3047568461932167e-06,
+      "loss": 0.6356,
+      "step": 21173
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3039480878659204e-06,
+      "loss": 0.644,
+      "step": 21174
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.303139408928793e-06,
+      "loss": 0.6568,
+      "step": 21175
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.302330809391422e-06,
+      "loss": 0.6041,
+      "step": 21176
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.301522289263389e-06,
+      "loss": 0.5992,
+      "step": 21177
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.3007138485542834e-06,
+      "loss": 0.6211,
+      "step": 21178
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2999054872736925e-06,
+      "loss": 0.6679,
+      "step": 21179
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.299097205431191e-06,
+      "loss": 0.6732,
+      "step": 21180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2982890030363723e-06,
+      "loss": 0.6329,
+      "step": 21181
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.297480880098819e-06,
+      "loss": 0.6155,
+      "step": 21182
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2966728366281044e-06,
+      "loss": 0.6378,
+      "step": 21183
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.295864872633817e-06,
+      "loss": 0.6636,
+      "step": 21184
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.295056988125529e-06,
+      "loss": 0.6227,
+      "step": 21185
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.294249183112819e-06,
+      "loss": 0.6163,
+      "step": 21186
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2934414576052733e-06,
+      "loss": 0.613,
+      "step": 21187
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.292633811612461e-06,
+      "loss": 0.618,
+      "step": 21188
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2918262451439585e-06,
+      "loss": 0.6731,
+      "step": 21189
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2910187582093454e-06,
+      "loss": 0.6525,
+      "step": 21190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.29021135081819e-06,
+      "loss": 0.679,
+      "step": 21191
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2894040229800618e-06,
+      "loss": 0.679,
+      "step": 21192
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2885967747045456e-06,
+      "loss": 0.6148,
+      "step": 21193
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2877896060012005e-06,
+      "loss": 0.565,
+      "step": 21194
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.286982516879601e-06,
+      "loss": 0.637,
+      "step": 21195
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2861755073493195e-06,
+      "loss": 0.6294,
+      "step": 21196
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2853685774199163e-06,
+      "loss": 0.6385,
+      "step": 21197
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.28456172710096e-06,
+      "loss": 0.6469,
+      "step": 21198
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2837549564020254e-06,
+      "loss": 0.64,
+      "step": 21199
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2829482653326692e-06,
+      "loss": 0.6538,
+      "step": 21200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2821416539024586e-06,
+      "loss": 0.5938,
+      "step": 21201
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.281335122120959e-06,
+      "loss": 0.6845,
+      "step": 21202
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.280528669997729e-06,
+      "loss": 0.677,
+      "step": 21203
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2797222975423273e-06,
+      "loss": 0.6419,
+      "step": 21204
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2789160047643254e-06,
+      "loss": 0.5889,
+      "step": 21205
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.278109791673274e-06,
+      "loss": 0.6634,
+      "step": 21206
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2773036582787343e-06,
+      "loss": 0.5935,
+      "step": 21207
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2764976045902664e-06,
+      "loss": 0.6382,
+      "step": 21208
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2756916306174224e-06,
+      "loss": 0.5961,
+      "step": 21209
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2748857363697563e-06,
+      "loss": 0.7141,
+      "step": 21210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2740799218568344e-06,
+      "loss": 0.6008,
+      "step": 21211
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.273274187088199e-06,
+      "loss": 0.5838,
+      "step": 21212
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.2724685320734083e-06,
+      "loss": 0.5717,
+      "step": 21213
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.271662956822017e-06,
+      "loss": 0.6485,
+      "step": 21214
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2708574613435685e-06,
+      "loss": 0.6238,
+      "step": 21215
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.270052045647615e-06,
+      "loss": 0.6264,
+      "step": 21216
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2692467097437143e-06,
+      "loss": 0.6244,
+      "step": 21217
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2684414536414045e-06,
+      "loss": 0.6362,
+      "step": 21218
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2676362773502367e-06,
+      "loss": 0.6449,
+      "step": 21219
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2668311808797604e-06,
+      "loss": 0.6251,
+      "step": 21220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2660261642395153e-06,
+      "loss": 0.6043,
+      "step": 21221
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.265221227439045e-06,
+      "loss": 0.6122,
+      "step": 21222
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.264416370487904e-06,
+      "loss": 0.694,
+      "step": 21223
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2636115933956224e-06,
+      "loss": 0.6239,
+      "step": 21224
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2628068961717486e-06,
+      "loss": 0.6398,
+      "step": 21225
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2620022788258244e-06,
+      "loss": 0.6387,
+      "step": 21226
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2611977413673836e-06,
+      "loss": 0.6182,
+      "step": 21227
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.260393283805965e-06,
+      "loss": 0.623,
+      "step": 21228
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2595889061511156e-06,
+      "loss": 0.6074,
+      "step": 21229
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2587846084123633e-06,
+      "loss": 0.6569,
+      "step": 21230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.257980390599247e-06,
+      "loss": 0.5722,
+      "step": 21231
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.257176252721306e-06,
+      "loss": 0.6447,
+      "step": 21232
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2563721947880667e-06,
+      "loss": 0.6122,
+      "step": 21233
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.255568216809062e-06,
+      "loss": 0.6564,
+      "step": 21234
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.254764318793834e-06,
+      "loss": 0.6879,
+      "step": 21235
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2539605007519048e-06,
+      "loss": 0.601,
+      "step": 21236
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2531567626928083e-06,
+      "loss": 0.6094,
+      "step": 21237
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.252353104626076e-06,
+      "loss": 0.6361,
+      "step": 21238
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2515495265612305e-06,
+      "loss": 0.5882,
+      "step": 21239
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2507460285077985e-06,
+      "loss": 0.6124,
+      "step": 21240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2499426104753174e-06,
+      "loss": 0.5776,
+      "step": 21241
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2491392724733018e-06,
+      "loss": 0.6332,
+      "step": 21242
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2483360145112795e-06,
+      "loss": 0.6032,
+      "step": 21243
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.247532836598779e-06,
+      "loss": 0.5747,
+      "step": 21244
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2467297387453143e-06,
+      "loss": 0.6444,
+      "step": 21245
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2459267209604118e-06,
+      "loss": 0.6695,
+      "step": 21246
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2451237832535932e-06,
+      "loss": 0.67,
+      "step": 21247
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2443209256343756e-06,
+      "loss": 0.7109,
+      "step": 21248
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2435181481122837e-06,
+      "loss": 0.5949,
+      "step": 21249
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2427154506968263e-06,
+      "loss": 0.6263,
+      "step": 21250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.241912833397527e-06,
+      "loss": 0.6851,
+      "step": 21251
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.241110296223898e-06,
+      "loss": 0.5782,
+      "step": 21252
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.240307839185458e-06,
+      "loss": 0.592,
+      "step": 21253
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.23950546229172e-06,
+      "loss": 0.7001,
+      "step": 21254
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2387031655521994e-06,
+      "loss": 0.6343,
+      "step": 21255
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2379009489764024e-06,
+      "loss": 0.6413,
+      "step": 21256
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2370988125738443e-06,
+      "loss": 0.6507,
+      "step": 21257
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2362967563540347e-06,
+      "loss": 0.5638,
+      "step": 21258
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2354947803264837e-06,
+      "loss": 0.6065,
+      "step": 21259
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2346928845006987e-06,
+      "loss": 0.6425,
+      "step": 21260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2338910688861903e-06,
+      "loss": 0.6632,
+      "step": 21261
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2330893334924596e-06,
+      "loss": 0.6167,
+      "step": 21262
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.232287678329016e-06,
+      "loss": 0.6422,
+      "step": 21263
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2314861034053624e-06,
+      "loss": 0.5571,
+      "step": 21264
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.230684608731003e-06,
+      "loss": 0.6397,
+      "step": 21265
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2298831943154406e-06,
+      "loss": 0.6619,
+      "step": 21266
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2290818601681818e-06,
+      "loss": 0.6022,
+      "step": 21267
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2282806062987183e-06,
+      "loss": 0.6641,
+      "step": 21268
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2274794327165547e-06,
+      "loss": 0.66,
+      "step": 21269
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2266783394311896e-06,
+      "loss": 0.6218,
+      "step": 21270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2258773264521216e-06,
+      "loss": 0.5942,
+      "step": 21271
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.225076393788846e-06,
+      "loss": 0.6226,
+      "step": 21272
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2242755414508654e-06,
+      "loss": 0.6757,
+      "step": 21273
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2234747694476654e-06,
+      "loss": 0.6157,
+      "step": 21274
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2226740777887435e-06,
+      "loss": 0.6645,
+      "step": 21275
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.221873466483595e-06,
+      "loss": 0.6432,
+      "step": 21276
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2210729355417103e-06,
+      "loss": 0.635,
+      "step": 21277
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2202724849725806e-06,
+      "loss": 0.6533,
+      "step": 21278
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.219472114785701e-06,
+      "loss": 0.6256,
+      "step": 21279
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2186718249905534e-06,
+      "loss": 0.6221,
+      "step": 21280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2178716155966307e-06,
+      "loss": 0.6517,
+      "step": 21281
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.217071486613419e-06,
+      "loss": 0.6636,
+      "step": 21282
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2162714380504044e-06,
+      "loss": 0.6313,
+      "step": 21283
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2154714699170742e-06,
+      "loss": 0.6693,
+      "step": 21284
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2146715822229157e-06,
+      "loss": 0.6388,
+      "step": 21285
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.213871774977405e-06,
+      "loss": 0.6868,
+      "step": 21286
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.213072048190029e-06,
+      "loss": 0.6459,
+      "step": 21287
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.21227240187027e-06,
+      "loss": 0.6406,
+      "step": 21288
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.211472836027608e-06,
+      "loss": 0.6556,
+      "step": 21289
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2106733506715228e-06,
+      "loss": 0.6083,
+      "step": 21290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2098739458114967e-06,
+      "loss": 0.6301,
+      "step": 21291
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.209074621457001e-06,
+      "loss": 0.5846,
+      "step": 21292
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.208275377617517e-06,
+      "loss": 0.6225,
+      "step": 21293
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2074762143025196e-06,
+      "loss": 0.6166,
+      "step": 21294
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.206677131521485e-06,
+      "loss": 0.5802,
+      "step": 21295
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.205878129283886e-06,
+      "loss": 0.6555,
+      "step": 21296
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.205079207599199e-06,
+      "loss": 0.559,
+      "step": 21297
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2042803664768907e-06,
+      "loss": 0.6519,
+      "step": 21298
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2034816059264363e-06,
+      "loss": 0.6301,
+      "step": 21299
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2026829259573044e-06,
+      "loss": 0.6509,
+      "step": 21300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2018843265789647e-06,
+      "loss": 0.6827,
+      "step": 21301
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2010858078008866e-06,
+      "loss": 0.6468,
+      "step": 21302
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.20028736963254e-06,
+      "loss": 0.6231,
+      "step": 21303
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1994890120833844e-06,
+      "loss": 0.6524,
+      "step": 21304
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1986907351628894e-06,
+      "loss": 0.6079,
+      "step": 21305
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1978925388805206e-06,
+      "loss": 0.6281,
+      "step": 21306
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.197094423245739e-06,
+      "loss": 0.5978,
+      "step": 21307
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1962963882680133e-06,
+      "loss": 0.6377,
+      "step": 21308
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1954984339567964e-06,
+      "loss": 0.6378,
+      "step": 21309
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.194700560321554e-06,
+      "loss": 0.6439,
+      "step": 21310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.193902767371745e-06,
+      "loss": 0.5908,
+      "step": 21311
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1931050551168276e-06,
+      "loss": 0.6331,
+      "step": 21312
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.192307423566261e-06,
+      "loss": 0.6624,
+      "step": 21313
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.191509872729506e-06,
+      "loss": 0.6111,
+      "step": 21314
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1907124026160095e-06,
+      "loss": 0.6433,
+      "step": 21315
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.189915013235232e-06,
+      "loss": 0.6253,
+      "step": 21316
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.189117704596627e-06,
+      "loss": 0.5823,
+      "step": 21317
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1883204767096478e-06,
+      "loss": 0.6254,
+      "step": 21318
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.187523329583745e-06,
+      "loss": 0.6394,
+      "step": 21319
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1867262632283745e-06,
+      "loss": 0.5975,
+      "step": 21320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1859292776529803e-06,
+      "loss": 0.6697,
+      "step": 21321
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1851323728670135e-06,
+      "loss": 0.6553,
+      "step": 21322
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1843355488799244e-06,
+      "loss": 0.6856,
+      "step": 21323
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1835388057011584e-06,
+      "loss": 0.6893,
+      "step": 21324
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.182742143340163e-06,
+      "loss": 0.5886,
+      "step": 21325
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1819455618063868e-06,
+      "loss": 0.6352,
+      "step": 21326
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1811490611092667e-06,
+      "loss": 0.6207,
+      "step": 21327
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1803526412582507e-06,
+      "loss": 0.5792,
+      "step": 21328
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.179556302262782e-06,
+      "loss": 0.5904,
+      "step": 21329
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.178760044132301e-06,
+      "loss": 0.6308,
+      "step": 21330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.177963866876248e-06,
+      "loss": 0.6048,
+      "step": 21331
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1771677705040672e-06,
+      "loss": 0.6585,
+      "step": 21332
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1763717550251894e-06,
+      "loss": 0.5941,
+      "step": 21333
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1755758204490563e-06,
+      "loss": 0.6609,
+      "step": 21334
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1747799667851056e-06,
+      "loss": 0.6442,
+      "step": 21335
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.173984194042772e-06,
+      "loss": 0.634,
+      "step": 21336
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1731885022314903e-06,
+      "loss": 0.6253,
+      "step": 21337
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1723928913606984e-06,
+      "loss": 0.5963,
+      "step": 21338
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.171597361439822e-06,
+      "loss": 0.6093,
+      "step": 21339
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1708019124782974e-06,
+      "loss": 0.6338,
+      "step": 21340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1700065444855554e-06,
+      "loss": 0.5978,
+      "step": 21341
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1692112574710254e-06,
+      "loss": 0.6314,
+      "step": 21342
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1684160514441374e-06,
+      "loss": 0.6126,
+      "step": 21343
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.167620926414322e-06,
+      "loss": 0.668,
+      "step": 21344
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.166825882391e-06,
+      "loss": 0.6087,
+      "step": 21345
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1660309193836016e-06,
+      "loss": 0.6541,
+      "step": 21346
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.165236037401551e-06,
+      "loss": 0.6487,
+      "step": 21347
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.164441236454274e-06,
+      "loss": 0.5822,
+      "step": 21348
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.163646516551193e-06,
+      "loss": 0.6264,
+      "step": 21349
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.162851877701734e-06,
+      "loss": 0.648,
+      "step": 21350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1620573199153115e-06,
+      "loss": 0.6139,
+      "step": 21351
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.161262843201349e-06,
+      "loss": 0.6464,
+      "step": 21352
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.160468447569267e-06,
+      "loss": 0.6308,
+      "step": 21353
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.159674133028484e-06,
+      "loss": 0.5582,
+      "step": 21354
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1588798995884172e-06,
+      "loss": 0.68,
+      "step": 21355
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1580857472584856e-06,
+      "loss": 0.6386,
+      "step": 21356
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1572916760480997e-06,
+      "loss": 0.6057,
+      "step": 21357
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1564976859666774e-06,
+      "loss": 0.6132,
+      "step": 21358
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.155703777023632e-06,
+      "loss": 0.6686,
+      "step": 21359
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1549099492283762e-06,
+      "loss": 0.6081,
+      "step": 21360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.154116202590323e-06,
+      "loss": 0.653,
+      "step": 21361
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1533225371188847e-06,
+      "loss": 0.6475,
+      "step": 21362
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.152528952823465e-06,
+      "loss": 0.6309,
+      "step": 21363
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.151735449713478e-06,
+      "loss": 0.6385,
+      "step": 21364
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1509420277983293e-06,
+      "loss": 0.5848,
+      "step": 21365
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1501486870874263e-06,
+      "loss": 0.6125,
+      "step": 21366
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.149355427590177e-06,
+      "loss": 0.6552,
+      "step": 21367
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1485622493159886e-06,
+      "loss": 0.6849,
+      "step": 21368
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1477691522742583e-06,
+      "loss": 0.66,
+      "step": 21369
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1469761364743956e-06,
+      "loss": 0.638,
+      "step": 21370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.146183201925793e-06,
+      "loss": 0.6501,
+      "step": 21371
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1453903486378634e-06,
+      "loss": 0.6443,
+      "step": 21372
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1445975766200044e-06,
+      "loss": 0.6143,
+      "step": 21373
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1438048858816106e-06,
+      "loss": 0.6498,
+      "step": 21374
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1430122764320824e-06,
+      "loss": 0.6268,
+      "step": 21375
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1422197482808226e-06,
+      "loss": 0.5967,
+      "step": 21376
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.141427301437214e-06,
+      "loss": 0.641,
+      "step": 21377
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1406349359106657e-06,
+      "loss": 0.6037,
+      "step": 21378
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1398426517105707e-06,
+      "loss": 0.6617,
+      "step": 21379
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.139050448846316e-06,
+      "loss": 0.6279,
+      "step": 21380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.138258327327296e-06,
+      "loss": 0.6454,
+      "step": 21381
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.137466287162908e-06,
+      "loss": 0.595,
+      "step": 21382
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1366743283625322e-06,
+      "loss": 0.5761,
+      "step": 21383
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.135882450935567e-06,
+      "loss": 0.6551,
+      "step": 21384
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.135090654891402e-06,
+      "loss": 0.5782,
+      "step": 21385
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1342989402394174e-06,
+      "loss": 0.6619,
+      "step": 21386
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1335073069890043e-06,
+      "loss": 0.677,
+      "step": 21387
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1327157551495523e-06,
+      "loss": 0.6317,
+      "step": 21388
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1319242847304355e-06,
+      "loss": 0.643,
+      "step": 21389
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.131132895741047e-06,
+      "loss": 0.6398,
+      "step": 21390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1303415881907705e-06,
+      "loss": 0.6385,
+      "step": 21391
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1295503620889823e-06,
+      "loss": 0.6117,
+      "step": 21392
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.128759217445065e-06,
+      "loss": 0.6272,
+      "step": 21393
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1279681542684026e-06,
+      "loss": 0.6728,
+      "step": 21394
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.127177172568365e-06,
+      "loss": 0.6149,
+      "step": 21395
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1263862723543393e-06,
+      "loss": 0.6179,
+      "step": 21396
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1255954536357013e-06,
+      "loss": 0.5668,
+      "step": 21397
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1248047164218244e-06,
+      "loss": 0.6797,
+      "step": 21398
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1240140607220827e-06,
+      "loss": 0.6235,
+      "step": 21399
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1232234865458567e-06,
+      "loss": 0.6669,
+      "step": 21400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.122432993902509e-06,
+      "loss": 0.5861,
+      "step": 21401
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.121642582801422e-06,
+      "loss": 0.6447,
+      "step": 21402
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.120852253251966e-06,
+      "loss": 0.6308,
+      "step": 21403
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1200620052635057e-06,
+      "loss": 0.5996,
+      "step": 21404
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.119271838845412e-06,
+      "loss": 0.6469,
+      "step": 21405
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.11848175400706e-06,
+      "loss": 0.6424,
+      "step": 21406
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1176917507578054e-06,
+      "loss": 0.6463,
+      "step": 21407
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1169018291070242e-06,
+      "loss": 0.692,
+      "step": 21408
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1161119890640813e-06,
+      "loss": 0.6517,
+      "step": 21409
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1153222306383368e-06,
+      "loss": 0.5749,
+      "step": 21410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1145325538391558e-06,
+      "loss": 0.6134,
+      "step": 21411
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.113742958675905e-06,
+      "loss": 0.6499,
+      "step": 21412
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1129534451579357e-06,
+      "loss": 0.6252,
+      "step": 21413
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1121640132946195e-06,
+      "loss": 0.6517,
+      "step": 21414
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1113746630953146e-06,
+      "loss": 0.6327,
+      "step": 21415
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1105853945693744e-06,
+      "loss": 0.6257,
+      "step": 21416
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.109796207726159e-06,
+      "loss": 0.6019,
+      "step": 21417
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.109007102575029e-06,
+      "loss": 0.5746,
+      "step": 21418
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1082180791253314e-06,
+      "loss": 0.5588,
+      "step": 21419
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.10742913738643e-06,
+      "loss": 0.5703,
+      "step": 21420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1066402773676774e-06,
+      "loss": 0.65,
+      "step": 21421
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.105851499078422e-06,
+      "loss": 0.6245,
+      "step": 21422
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.105062802528017e-06,
+      "loss": 0.6297,
+      "step": 21423
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.104274187725819e-06,
+      "loss": 0.7053,
+      "step": 21424
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1034856546811664e-06,
+      "loss": 0.615,
+      "step": 21425
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.10269720340342e-06,
+      "loss": 0.6359,
+      "step": 21426
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1019088339019254e-06,
+      "loss": 0.6013,
+      "step": 21427
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.101120546186025e-06,
+      "loss": 0.6778,
+      "step": 21428
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.1003323402650677e-06,
+      "loss": 0.5834,
+      "step": 21429
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0995442161484015e-06,
+      "loss": 0.5976,
+      "step": 21430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0987561738453618e-06,
+      "loss": 0.6521,
+      "step": 21431
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0979682133653033e-06,
+      "loss": 0.7011,
+      "step": 21432
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.09718033471756e-06,
+      "loss": 0.5645,
+      "step": 21433
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.096392537911477e-06,
+      "loss": 0.6231,
+      "step": 21434
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.095604822956395e-06,
+      "loss": 0.7076,
+      "step": 21435
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0948171898616498e-06,
+      "loss": 0.6242,
+      "step": 21436
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.094029638636579e-06,
+      "loss": 0.6659,
+      "step": 21437
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.093242169290529e-06,
+      "loss": 0.6773,
+      "step": 21438
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0924547818328267e-06,
+      "loss": 0.6873,
+      "step": 21439
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0916674762728104e-06,
+      "loss": 0.5973,
+      "step": 21440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.09088025261982e-06,
+      "loss": 0.6734,
+      "step": 21441
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.09009311088318e-06,
+      "loss": 0.6327,
+      "step": 21442
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.089306051072225e-06,
+      "loss": 0.6257,
+      "step": 21443
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0885190731962946e-06,
+      "loss": 0.6167,
+      "step": 21444
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0877321772647095e-06,
+      "loss": 0.6495,
+      "step": 21445
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0869453632868053e-06,
+      "loss": 0.6885,
+      "step": 21446
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0861586312719104e-06,
+      "loss": 0.612,
+      "step": 21447
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.085371981229349e-06,
+      "loss": 0.5931,
+      "step": 21448
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.084585413168446e-06,
+      "loss": 0.5685,
+      "step": 21449
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.083798927098538e-06,
+      "loss": 0.6605,
+      "step": 21450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0830125230289386e-06,
+      "loss": 0.6766,
+      "step": 21451
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0822262009689774e-06,
+      "loss": 0.6488,
+      "step": 21452
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.081439960927978e-06,
+      "loss": 0.5966,
+      "step": 21453
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0806538029152554e-06,
+      "loss": 0.6156,
+      "step": 21454
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0798677269401333e-06,
+      "loss": 0.5969,
+      "step": 21455
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0790817330119394e-06,
+      "loss": 0.6158,
+      "step": 21456
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0782958211399826e-06,
+      "loss": 0.6626,
+      "step": 21457
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.077509991333586e-06,
+      "loss": 0.6361,
+      "step": 21458
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.076724243602068e-06,
+      "loss": 0.6706,
+      "step": 21459
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0759385779547392e-06,
+      "loss": 0.6509,
+      "step": 21460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0751529944009174e-06,
+      "loss": 0.6479,
+      "step": 21461
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.074367492949917e-06,
+      "loss": 0.6072,
+      "step": 21462
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.07358207361105e-06,
+      "loss": 0.5998,
+      "step": 21463
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0727967363936307e-06,
+      "loss": 0.6015,
+      "step": 21464
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.072011481306971e-06,
+      "loss": 0.6376,
+      "step": 21465
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0712263083603766e-06,
+      "loss": 0.6313,
+      "step": 21466
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0704412175631594e-06,
+      "loss": 0.6563,
+      "step": 21467
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0696562089246275e-06,
+      "loss": 0.6479,
+      "step": 21468
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0688712824540878e-06,
+      "loss": 0.6432,
+      "step": 21469
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0680864381608478e-06,
+      "loss": 0.6625,
+      "step": 21470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.067301676054215e-06,
+      "loss": 0.6272,
+      "step": 21471
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.066516996143487e-06,
+      "loss": 0.6303,
+      "step": 21472
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0657323984379704e-06,
+      "loss": 0.6855,
+      "step": 21473
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.064947882946969e-06,
+      "loss": 0.5726,
+      "step": 21474
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0641634496797834e-06,
+      "loss": 0.6615,
+      "step": 21475
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0633790986457135e-06,
+      "loss": 0.6922,
+      "step": 21476
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0625948298540632e-06,
+      "loss": 0.6562,
+      "step": 21477
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0618106433141237e-06,
+      "loss": 0.6502,
+      "step": 21478
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0610265390351945e-06,
+      "loss": 0.5785,
+      "step": 21479
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0602425170265747e-06,
+      "loss": 0.5755,
+      "step": 21480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0594585772975594e-06,
+      "loss": 0.616,
+      "step": 21481
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0586747198574408e-06,
+      "loss": 0.6377,
+      "step": 21482
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.057890944715519e-06,
+      "loss": 0.5852,
+      "step": 21483
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0571072518810774e-06,
+      "loss": 0.6959,
+      "step": 21484
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0563236413634135e-06,
+      "loss": 0.5992,
+      "step": 21485
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0555401131718155e-06,
+      "loss": 0.6737,
+      "step": 21486
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0547566673155736e-06,
+      "loss": 0.6878,
+      "step": 21487
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.053973303803979e-06,
+      "loss": 0.6162,
+      "step": 21488
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0531900226463194e-06,
+      "loss": 0.6316,
+      "step": 21489
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0524068238518776e-06,
+      "loss": 0.6127,
+      "step": 21490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0516237074299415e-06,
+      "loss": 0.6791,
+      "step": 21491
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0508406733897954e-06,
+      "loss": 0.6497,
+      "step": 21492
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0500577217407245e-06,
+      "loss": 0.6486,
+      "step": 21493
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0492748524920148e-06,
+      "loss": 0.6441,
+      "step": 21494
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.048492065652939e-06,
+      "loss": 0.6416,
+      "step": 21495
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.047709361232786e-06,
+      "loss": 0.6215,
+      "step": 21496
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.046926739240832e-06,
+      "loss": 0.657,
+      "step": 21497
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.046144199686357e-06,
+      "loss": 0.6425,
+      "step": 21498
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0453617425786396e-06,
+      "loss": 0.6164,
+      "step": 21499
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0445793679269597e-06,
+      "loss": 0.6619,
+      "step": 21500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0437970757405864e-06,
+      "loss": 0.6391,
+      "step": 21501
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.0430148660287994e-06,
+      "loss": 0.617,
+      "step": 21502
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.042232738800871e-06,
+      "loss": 0.6588,
+      "step": 21503
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.041450694066075e-06,
+      "loss": 0.6553,
+      "step": 21504
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0406687318336834e-06,
+      "loss": 0.6872,
+      "step": 21505
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0398868521129722e-06,
+      "loss": 0.7102,
+      "step": 21506
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0391050549132016e-06,
+      "loss": 0.6461,
+      "step": 21507
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0383233402436475e-06,
+      "loss": 0.6723,
+      "step": 21508
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.037541708113577e-06,
+      "loss": 0.6424,
+      "step": 21509
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.036760158532257e-06,
+      "loss": 0.6104,
+      "step": 21510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.035978691508953e-06,
+      "loss": 0.6199,
+      "step": 21511
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.035197307052935e-06,
+      "loss": 0.631,
+      "step": 21512
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0344160051734605e-06,
+      "loss": 0.5743,
+      "step": 21513
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0336347858797953e-06,
+      "loss": 0.684,
+      "step": 21514
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.032853649181202e-06,
+      "loss": 0.6426,
+      "step": 21515
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0320725950869424e-06,
+      "loss": 0.6566,
+      "step": 21516
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0312916236062773e-06,
+      "loss": 0.5873,
+      "step": 21517
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0305107347484673e-06,
+      "loss": 0.6252,
+      "step": 21518
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0297299285227677e-06,
+      "loss": 0.6283,
+      "step": 21519
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.028949204938436e-06,
+      "loss": 0.6258,
+      "step": 21520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.02816856400473e-06,
+      "loss": 0.6362,
+      "step": 21521
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0273880057309048e-06,
+      "loss": 0.6568,
+      "step": 21522
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0266075301262155e-06,
+      "loss": 0.5914,
+      "step": 21523
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0258271371999192e-06,
+      "loss": 0.6814,
+      "step": 21524
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0250468269612597e-06,
+      "loss": 0.629,
+      "step": 21525
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0242665994194953e-06,
+      "loss": 0.6155,
+      "step": 21526
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.023486454583874e-06,
+      "loss": 0.6814,
+      "step": 21527
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0227063924636457e-06,
+      "loss": 0.6576,
+      "step": 21528
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0219264130680593e-06,
+      "loss": 0.615,
+      "step": 21529
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.021146516406367e-06,
+      "loss": 0.5995,
+      "step": 21530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0203667024878067e-06,
+      "loss": 0.6103,
+      "step": 21531
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0195869713216287e-06,
+      "loss": 0.612,
+      "step": 21532
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.018807322917078e-06,
+      "loss": 0.6239,
+      "step": 21533
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.018027757283396e-06,
+      "loss": 0.6128,
+      "step": 21534
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.017248274429829e-06,
+      "loss": 0.5824,
+      "step": 21535
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.016468874365619e-06,
+      "loss": 0.6339,
+      "step": 21536
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0156895571000023e-06,
+      "loss": 0.6466,
+      "step": 21537
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0149103226422217e-06,
+      "loss": 0.6318,
+      "step": 21538
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0141311710015142e-06,
+      "loss": 0.6454,
+      "step": 21539
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0133521021871192e-06,
+      "loss": 0.713,
+      "step": 21540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0125731162082728e-06,
+      "loss": 0.6554,
+      "step": 21541
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.011794213074215e-06,
+      "loss": 0.6785,
+      "step": 21542
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0110153927941734e-06,
+      "loss": 0.6288,
+      "step": 21543
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0102366553773854e-06,
+      "loss": 0.6511,
+      "step": 21544
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0094580008330843e-06,
+      "loss": 0.6449,
+      "step": 21545
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0086794291705013e-06,
+      "loss": 0.653,
+      "step": 21546
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.007900940398868e-06,
+      "loss": 0.6052,
+      "step": 21547
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.007122534527418e-06,
+      "loss": 0.6352,
+      "step": 21548
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.006344211565372e-06,
+      "loss": 0.5958,
+      "step": 21549
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0055659715219633e-06,
+      "loss": 0.6397,
+      "step": 21550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.004787814406417e-06,
+      "loss": 0.6049,
+      "step": 21551
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0040097402279613e-06,
+      "loss": 0.6338,
+      "step": 21552
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0032317489958184e-06,
+      "loss": 0.6112,
+      "step": 21553
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0024538407192194e-06,
+      "loss": 0.6079,
+      "step": 21554
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0016760154073774e-06,
+      "loss": 0.6521,
+      "step": 21555
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0008982730695193e-06,
+      "loss": 0.6813,
+      "step": 21556
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0001206137148666e-06,
+      "loss": 0.6422,
+      "step": 21557
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.999343037352639e-06,
+      "loss": 0.6094,
+      "step": 21558
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.998565543992058e-06,
+      "loss": 0.6493,
+      "step": 21559
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9977881336423364e-06,
+      "loss": 0.604,
+      "step": 21560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.997010806312695e-06,
+      "loss": 0.6139,
+      "step": 21561
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9962335620123482e-06,
+      "loss": 0.6306,
+      "step": 21562
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9954564007505137e-06,
+      "loss": 0.6443,
+      "step": 21563
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9946793225364036e-06,
+      "loss": 0.6525,
+      "step": 21564
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9939023273792343e-06,
+      "loss": 0.6061,
+      "step": 21565
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9931254152882137e-06,
+      "loss": 0.623,
+      "step": 21566
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9923485862725553e-06,
+      "loss": 0.6417,
+      "step": 21567
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9915718403414684e-06,
+      "loss": 0.5821,
+      "step": 21568
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9907951775041633e-06,
+      "loss": 0.6571,
+      "step": 21569
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.990018597769847e-06,
+      "loss": 0.6204,
+      "step": 21570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9892421011477325e-06,
+      "loss": 0.5625,
+      "step": 21571
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9884656876470175e-06,
+      "loss": 0.5875,
+      "step": 21572
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9876893572769116e-06,
+      "loss": 0.6008,
+      "step": 21573
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.98691311004662e-06,
+      "loss": 0.5848,
+      "step": 21574
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9861369459653433e-06,
+      "loss": 0.6531,
+      "step": 21575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9853608650422873e-06,
+      "loss": 0.6342,
+      "step": 21576
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9845848672866538e-06,
+      "loss": 0.6312,
+      "step": 21577
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9838089527076384e-06,
+      "loss": 0.6762,
+      "step": 21578
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.983033121314444e-06,
+      "loss": 0.6407,
+      "step": 21579
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.982257373116267e-06,
+      "loss": 0.5472,
+      "step": 21580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.981481708122308e-06,
+      "loss": 0.6594,
+      "step": 21581
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9807061263417614e-06,
+      "loss": 0.6647,
+      "step": 21582
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.979930627783826e-06,
+      "loss": 0.6717,
+      "step": 21583
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9791552124576904e-06,
+      "loss": 0.6381,
+      "step": 21584
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9783798803725518e-06,
+      "loss": 0.6298,
+      "step": 21585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9776046315376004e-06,
+      "loss": 0.6384,
+      "step": 21586
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9768294659620313e-06,
+      "loss": 0.6284,
+      "step": 21587
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9760543836550313e-06,
+      "loss": 0.6106,
+      "step": 21588
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9752793846257967e-06,
+      "loss": 0.6097,
+      "step": 21589
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.974504468883508e-06,
+      "loss": 0.6164,
+      "step": 21590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9737296364373557e-06,
+      "loss": 0.5859,
+      "step": 21591
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9729548872965264e-06,
+      "loss": 0.6187,
+      "step": 21592
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.972180221470207e-06,
+      "loss": 0.6544,
+      "step": 21593
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9714056389675818e-06,
+      "loss": 0.6673,
+      "step": 21594
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.970631139797836e-06,
+      "loss": 0.5674,
+      "step": 21595
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.969856723970147e-06,
+      "loss": 0.6384,
+      "step": 21596
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9690823914937006e-06,
+      "loss": 0.5961,
+      "step": 21597
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.968308142377676e-06,
+      "loss": 0.5907,
+      "step": 21598
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.967533976631253e-06,
+      "loss": 0.7038,
+      "step": 21599
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.966759894263612e-06,
+      "loss": 0.6215,
+      "step": 21600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9659858952839316e-06,
+      "loss": 0.6525,
+      "step": 21601
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9652119797013847e-06,
+      "loss": 0.5964,
+      "step": 21602
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9644381475251472e-06,
+      "loss": 0.6213,
+      "step": 21603
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9636643987643963e-06,
+      "loss": 0.5474,
+      "step": 21604
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9628907334283054e-06,
+      "loss": 0.5759,
+      "step": 21605
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.962117151526046e-06,
+      "loss": 0.6507,
+      "step": 21606
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.961343653066795e-06,
+      "loss": 0.6395,
+      "step": 21607
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9605702380597145e-06,
+      "loss": 0.5989,
+      "step": 21608
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9597969065139796e-06,
+      "loss": 0.6226,
+      "step": 21609
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9590236584387576e-06,
+      "loss": 0.5785,
+      "step": 21610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.958250493843218e-06,
+      "loss": 0.6317,
+      "step": 21611
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.957477412736526e-06,
+      "loss": 0.6644,
+      "step": 21612
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.956704415127851e-06,
+      "loss": 0.6165,
+      "step": 21613
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9559315010263522e-06,
+      "loss": 0.6302,
+      "step": 21614
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.955158670441196e-06,
+      "loss": 0.6018,
+      "step": 21615
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9543859233815463e-06,
+      "loss": 0.6527,
+      "step": 21616
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9536132598565637e-06,
+      "loss": 0.6219,
+      "step": 21617
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.952840679875413e-06,
+      "loss": 0.657,
+      "step": 21618
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9520681834472478e-06,
+      "loss": 0.5633,
+      "step": 21619
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.951295770581228e-06,
+      "loss": 0.67,
+      "step": 21620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9505234412865192e-06,
+      "loss": 0.6471,
+      "step": 21621
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9497511955722646e-06,
+      "loss": 0.6659,
+      "step": 21622
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.948979033447632e-06,
+      "loss": 0.6429,
+      "step": 21623
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9482069549217764e-06,
+      "loss": 0.6138,
+      "step": 21624
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9474349600038444e-06,
+      "loss": 0.65,
+      "step": 21625
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.946663048702991e-06,
+      "loss": 0.6164,
+      "step": 21626
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.945891221028375e-06,
+      "loss": 0.6456,
+      "step": 21627
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9451194769891345e-06,
+      "loss": 0.6415,
+      "step": 21628
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9443478165944305e-06,
+      "loss": 0.6965,
+      "step": 21629
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.943576239853412e-06,
+      "loss": 0.5243,
+      "step": 21630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9428047467752195e-06,
+      "loss": 0.6466,
+      "step": 21631
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9420333373690057e-06,
+      "loss": 0.6715,
+      "step": 21632
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.941262011643917e-06,
+      "loss": 0.6415,
+      "step": 21633
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.94049076960909e-06,
+      "loss": 0.6217,
+      "step": 21634
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9397196112736794e-06,
+      "loss": 0.6909,
+      "step": 21635
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9389485366468264e-06,
+      "loss": 0.573,
+      "step": 21636
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.938177545737668e-06,
+      "loss": 0.6477,
+      "step": 21637
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.937406638555348e-06,
+      "loss": 0.6573,
+      "step": 21638
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.93663581510901e-06,
+      "loss": 0.6117,
+      "step": 21639
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9358650754077834e-06,
+      "loss": 0.6328,
+      "step": 21640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9350944194608154e-06,
+      "loss": 0.6438,
+      "step": 21641
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.934323847277243e-06,
+      "loss": 0.5984,
+      "step": 21642
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.933553358866197e-06,
+      "loss": 0.6339,
+      "step": 21643
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9327829542368146e-06,
+      "loss": 0.619,
+      "step": 21644
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9320126333982334e-06,
+      "loss": 0.5829,
+      "step": 21645
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.931242396359577e-06,
+      "loss": 0.6154,
+      "step": 21646
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.930472243129988e-06,
+      "loss": 0.6723,
+      "step": 21647
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.929702173718596e-06,
+      "loss": 0.6651,
+      "step": 21648
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9289321881345257e-06,
+      "loss": 0.6364,
+      "step": 21649
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.92816228638691e-06,
+      "loss": 0.6707,
+      "step": 21650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9273924684848776e-06,
+      "loss": 0.5969,
+      "step": 21651
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9266227344375532e-06,
+      "loss": 0.6419,
+      "step": 21652
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.92585308425406e-06,
+      "loss": 0.6218,
+      "step": 21653
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9250835179435333e-06,
+      "loss": 0.5953,
+      "step": 21654
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.924314035515089e-06,
+      "loss": 0.6907,
+      "step": 21655
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.923544636977853e-06,
+      "loss": 0.6209,
+      "step": 21656
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9227753223409492e-06,
+      "loss": 0.6355,
+      "step": 21657
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.922006091613494e-06,
+      "loss": 0.6203,
+      "step": 21658
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9212369448046075e-06,
+      "loss": 0.6055,
+      "step": 21659
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9204678819234168e-06,
+      "loss": 0.6003,
+      "step": 21660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9196989029790322e-06,
+      "loss": 0.596,
+      "step": 21661
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9189300079805747e-06,
+      "loss": 0.6291,
+      "step": 21662
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9181611969371625e-06,
+      "loss": 0.6445,
+      "step": 21663
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9173924698579037e-06,
+      "loss": 0.6505,
+      "step": 21664
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.916623826751914e-06,
+      "loss": 0.6061,
+      "step": 21665
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.915855267628317e-06,
+      "loss": 0.6502,
+      "step": 21666
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9150867924962124e-06,
+      "loss": 0.6534,
+      "step": 21667
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9143184013647175e-06,
+      "loss": 0.6793,
+      "step": 21668
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.913550094242944e-06,
+      "loss": 0.6386,
+      "step": 21669
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9127818711399945e-06,
+      "loss": 0.676,
+      "step": 21670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9120137320649798e-06,
+      "loss": 0.6212,
+      "step": 21671
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9112456770270137e-06,
+      "loss": 0.6599,
+      "step": 21672
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9104777060351952e-06,
+      "loss": 0.6527,
+      "step": 21673
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9097098190986306e-06,
+      "loss": 0.6216,
+      "step": 21674
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9089420162264294e-06,
+      "loss": 0.5934,
+      "step": 21675
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9081742974276883e-06,
+      "loss": 0.615,
+      "step": 21676
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9074066627115082e-06,
+      "loss": 0.6296,
+      "step": 21677
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.906639112087e-06,
+      "loss": 0.6643,
+      "step": 21678
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.905871645563256e-06,
+      "loss": 0.6249,
+      "step": 21679
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9051042631493775e-06,
+      "loss": 0.6658,
+      "step": 21680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9043369648544662e-06,
+      "loss": 0.6026,
+      "step": 21681
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9035697506876127e-06,
+      "loss": 0.5855,
+      "step": 21682
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9028026206579173e-06,
+      "loss": 0.6425,
+      "step": 21683
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9020355747744744e-06,
+      "loss": 0.659,
+      "step": 21684
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9012686130463797e-06,
+      "loss": 0.6088,
+      "step": 21685
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.900501735482728e-06,
+      "loss": 0.6608,
+      "step": 21686
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8997349420926055e-06,
+      "loss": 0.607,
+      "step": 21687
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8989682328851076e-06,
+      "loss": 0.6222,
+      "step": 21688
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8982016078693252e-06,
+      "loss": 0.6254,
+      "step": 21689
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.897435067054345e-06,
+      "loss": 0.6,
+      "step": 21690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8966686104492582e-06,
+      "loss": 0.7115,
+      "step": 21691
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.895902238063153e-06,
+      "loss": 0.6327,
+      "step": 21692
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.89513594990511e-06,
+      "loss": 0.6274,
+      "step": 21693
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.894369745984219e-06,
+      "loss": 0.6288,
+      "step": 21694
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8936036263095625e-06,
+      "loss": 0.687,
+      "step": 21695
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.892837590890225e-06,
+      "loss": 0.6305,
+      "step": 21696
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.892071639735289e-06,
+      "loss": 0.6133,
+      "step": 21697
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8913057728538375e-06,
+      "loss": 0.6453,
+      "step": 21698
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8905399902549456e-06,
+      "loss": 0.6037,
+      "step": 21699
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8897742919476967e-06,
+      "loss": 0.6231,
+      "step": 21700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8890086779411673e-06,
+      "loss": 0.6486,
+      "step": 21701
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.888243148244435e-06,
+      "loss": 0.5596,
+      "step": 21702
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.887477702866578e-06,
+      "loss": 0.6179,
+      "step": 21703
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.886712341816672e-06,
+      "loss": 0.6638,
+      "step": 21704
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8859470651037867e-06,
+      "loss": 0.6654,
+      "step": 21705
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8851818727369985e-06,
+      "loss": 0.5784,
+      "step": 21706
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8844167647253796e-06,
+      "loss": 0.6409,
+      "step": 21707
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.883651741078001e-06,
+      "loss": 0.6064,
+      "step": 21708
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8828868018039324e-06,
+      "loss": 0.6331,
+      "step": 21709
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8821219469122487e-06,
+      "loss": 0.6497,
+      "step": 21710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8813571764120097e-06,
+      "loss": 0.6652,
+      "step": 21711
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8805924903122863e-06,
+      "loss": 0.5814,
+      "step": 21712
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8798278886221443e-06,
+      "loss": 0.6608,
+      "step": 21713
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.879063371350651e-06,
+      "loss": 0.5875,
+      "step": 21714
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8782989385068694e-06,
+      "loss": 0.6278,
+      "step": 21715
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.877534590099865e-06,
+      "loss": 0.6362,
+      "step": 21716
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.876770326138696e-06,
+      "loss": 0.6604,
+      "step": 21717
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.876006146632425e-06,
+      "loss": 0.5884,
+      "step": 21718
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8752420515901137e-06,
+      "loss": 0.5885,
+      "step": 21719
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8744780410208197e-06,
+      "loss": 0.6667,
+      "step": 21720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8737141149336034e-06,
+      "loss": 0.6795,
+      "step": 21721
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8729502733375238e-06,
+      "loss": 0.5846,
+      "step": 21722
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8721865162416306e-06,
+      "loss": 0.6766,
+      "step": 21723
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.871422843654984e-06,
+      "loss": 0.5824,
+      "step": 21724
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8706592555866364e-06,
+      "loss": 0.6443,
+      "step": 21725
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.869895752045643e-06,
+      "loss": 0.6247,
+      "step": 21726
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.869132333041055e-06,
+      "loss": 0.5822,
+      "step": 21727
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8683689985819264e-06,
+      "loss": 0.6178,
+      "step": 21728
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.867605748677301e-06,
+      "loss": 0.6174,
+      "step": 21729
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8668425833362333e-06,
+      "loss": 0.5799,
+      "step": 21730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.86607950256777e-06,
+      "loss": 0.5896,
+      "step": 21731
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8653165063809584e-06,
+      "loss": 0.6074,
+      "step": 21732
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8645535947848456e-06,
+      "loss": 0.5738,
+      "step": 21733
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.863790767788479e-06,
+      "loss": 0.6327,
+      "step": 21734
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8630280254008967e-06,
+      "loss": 0.6528,
+      "step": 21735
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8622653676311462e-06,
+      "loss": 0.6653,
+      "step": 21736
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8615027944882677e-06,
+      "loss": 0.6064,
+      "step": 21737
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8607403059813054e-06,
+      "loss": 0.5981,
+      "step": 21738
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8599779021192976e-06,
+      "loss": 0.6351,
+      "step": 21739
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.859215582911287e-06,
+      "loss": 0.6694,
+      "step": 21740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8584533483663045e-06,
+      "loss": 0.6764,
+      "step": 21741
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.857691198493393e-06,
+      "loss": 0.6178,
+      "step": 21742
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8569291333015867e-06,
+      "loss": 0.6044,
+      "step": 21743
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8561671527999224e-06,
+      "loss": 0.6293,
+      "step": 21744
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8554052569974366e-06,
+      "loss": 0.6464,
+      "step": 21745
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8546434459031558e-06,
+      "loss": 0.6621,
+      "step": 21746
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.853881719526116e-06,
+      "loss": 0.6717,
+      "step": 21747
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8531200778753477e-06,
+      "loss": 0.6461,
+      "step": 21748
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8523585209598826e-06,
+      "loss": 0.6223,
+      "step": 21749
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.851597048788749e-06,
+      "loss": 0.6243,
+      "step": 21750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8508356613709775e-06,
+      "loss": 0.6704,
+      "step": 21751
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.850074358715591e-06,
+      "loss": 0.642,
+      "step": 21752
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8493131408316167e-06,
+      "loss": 0.6165,
+      "step": 21753
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8485520077280814e-06,
+      "loss": 0.6106,
+      "step": 21754
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8477909594140085e-06,
+      "loss": 0.691,
+      "step": 21755
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8470299958984204e-06,
+      "loss": 0.5842,
+      "step": 21756
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8462691171903447e-06,
+      "loss": 0.589,
+      "step": 21757
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8455083232987946e-06,
+      "loss": 0.6351,
+      "step": 21758
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8447476142327936e-06,
+      "loss": 0.5665,
+      "step": 21759
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.843986990001362e-06,
+      "loss": 0.6417,
+      "step": 21760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8432264506135155e-06,
+      "loss": 0.6096,
+      "step": 21761
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.842465996078274e-06,
+      "loss": 0.6256,
+      "step": 21762
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.841705626404655e-06,
+      "loss": 0.6509,
+      "step": 21763
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8409453416016685e-06,
+      "loss": 0.6203,
+      "step": 21764
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.84018514167833e-06,
+      "loss": 0.6444,
+      "step": 21765
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.839425026643655e-06,
+      "loss": 0.6724,
+      "step": 21766
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8386649965066537e-06,
+      "loss": 0.6237,
+      "step": 21767
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8379050512763396e-06,
+      "loss": 0.6642,
+      "step": 21768
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.837145190961723e-06,
+      "loss": 0.6364,
+      "step": 21769
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8363854155718084e-06,
+      "loss": 0.6744,
+      "step": 21770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.835625725115606e-06,
+      "loss": 0.6466,
+      "step": 21771
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.834866119602124e-06,
+      "loss": 0.6176,
+      "step": 21772
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8341065990403683e-06,
+      "loss": 0.6584,
+      "step": 21773
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.833347163439344e-06,
+      "loss": 0.6527,
+      "step": 21774
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8325878128080585e-06,
+      "loss": 0.6153,
+      "step": 21775
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8318285471555072e-06,
+      "loss": 0.6685,
+      "step": 21776
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8310693664906963e-06,
+      "loss": 0.6452,
+      "step": 21777
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.830310270822627e-06,
+      "loss": 0.633,
+      "step": 21778
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.829551260160299e-06,
+      "loss": 0.6347,
+      "step": 21779
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8287923345127123e-06,
+      "loss": 0.6345,
+      "step": 21780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8280334938888663e-06,
+      "loss": 0.6147,
+      "step": 21781
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8272747382977538e-06,
+      "loss": 0.6099,
+      "step": 21782
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8265160677483717e-06,
+      "loss": 0.6156,
+      "step": 21783
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.825757482249717e-06,
+      "loss": 0.6276,
+      "step": 21784
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8249989818107813e-06,
+      "loss": 0.5926,
+      "step": 21785
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8242405664405604e-06,
+      "loss": 0.5767,
+      "step": 21786
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8234822361480475e-06,
+      "loss": 0.5937,
+      "step": 21787
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.822723990942229e-06,
+      "loss": 0.6468,
+      "step": 21788
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.8219658308320953e-06,
+      "loss": 0.672,
+      "step": 21789
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.821207755826638e-06,
+      "loss": 0.6337,
+      "step": 21790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.820449765934844e-06,
+      "loss": 0.5829,
+      "step": 21791
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8196918611657e-06,
+      "loss": 0.6148,
+      "step": 21792
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8189340415281953e-06,
+      "loss": 0.6452,
+      "step": 21793
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8181763070313073e-06,
+      "loss": 0.617,
+      "step": 21794
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.817418657684026e-06,
+      "loss": 0.6501,
+      "step": 21795
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.816661093495332e-06,
+      "loss": 0.6343,
+      "step": 21796
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.815903614474207e-06,
+      "loss": 0.6334,
+      "step": 21797
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8151462206296334e-06,
+      "loss": 0.6542,
+      "step": 21798
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8143889119705924e-06,
+      "loss": 0.6976,
+      "step": 21799
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8136316885060588e-06,
+      "loss": 0.652,
+      "step": 21800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8128745502450115e-06,
+      "loss": 0.6494,
+      "step": 21801
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.812117497196428e-06,
+      "loss": 0.6299,
+      "step": 21802
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8113605293692858e-06,
+      "loss": 0.6708,
+      "step": 21803
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.810603646772556e-06,
+      "loss": 0.66,
+      "step": 21804
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.80984684941522e-06,
+      "loss": 0.6728,
+      "step": 21805
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8090901373062406e-06,
+      "loss": 0.6219,
+      "step": 21806
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8083335104545938e-06,
+      "loss": 0.6653,
+      "step": 21807
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8075769688692522e-06,
+      "loss": 0.6043,
+      "step": 21808
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.806820512559183e-06,
+      "loss": 0.6141,
+      "step": 21809
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.806064141533359e-06,
+      "loss": 0.6045,
+      "step": 21810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.805307855800743e-06,
+      "loss": 0.6559,
+      "step": 21811
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8045516553703023e-06,
+      "loss": 0.6919,
+      "step": 21812
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.803795540251005e-06,
+      "loss": 0.6509,
+      "step": 21813
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.803039510451815e-06,
+      "loss": 0.6541,
+      "step": 21814
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.802283565981695e-06,
+      "loss": 0.6359,
+      "step": 21815
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8015277068496127e-06,
+      "loss": 0.6435,
+      "step": 21816
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.800771933064521e-06,
+      "loss": 0.6411,
+      "step": 21817
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8000162446353864e-06,
+      "loss": 0.6492,
+      "step": 21818
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7992606415711665e-06,
+      "loss": 0.6507,
+      "step": 21819
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7985051238808203e-06,
+      "loss": 0.6171,
+      "step": 21820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.797749691573307e-06,
+      "loss": 0.6656,
+      "step": 21821
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.796994344657584e-06,
+      "loss": 0.6145,
+      "step": 21822
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7962390831426022e-06,
+      "loss": 0.663,
+      "step": 21823
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7954839070373185e-06,
+      "loss": 0.6629,
+      "step": 21824
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.794728816350688e-06,
+      "loss": 0.6206,
+      "step": 21825
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7939738110916615e-06,
+      "loss": 0.6006,
+      "step": 21826
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7932188912691903e-06,
+      "loss": 0.6461,
+      "step": 21827
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.79246405689223e-06,
+      "loss": 0.6545,
+      "step": 21828
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.791709307969722e-06,
+      "loss": 0.6304,
+      "step": 21829
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7909546445106197e-06,
+      "loss": 0.6642,
+      "step": 21830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7902000665238683e-06,
+      "loss": 0.6232,
+      "step": 21831
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.789445574018417e-06,
+      "loss": 0.6203,
+      "step": 21832
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.788691167003208e-06,
+      "loss": 0.6184,
+      "step": 21833
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7879368454871923e-06,
+      "loss": 0.6186,
+      "step": 21834
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7871826094793043e-06,
+      "loss": 0.6339,
+      "step": 21835
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7864284589884915e-06,
+      "loss": 0.6058,
+      "step": 21836
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.785674394023694e-06,
+      "loss": 0.6936,
+      "step": 21837
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7849204145938535e-06,
+      "loss": 0.6009,
+      "step": 21838
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7841665207079083e-06,
+      "loss": 0.6942,
+      "step": 21839
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7834127123748e-06,
+      "loss": 0.7107,
+      "step": 21840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7826589896034596e-06,
+      "loss": 0.6422,
+      "step": 21841
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.781905352402827e-06,
+      "loss": 0.6029,
+      "step": 21842
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7811518007818404e-06,
+      "loss": 0.6356,
+      "step": 21843
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7803983347494257e-06,
+      "loss": 0.681,
+      "step": 21844
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7796449543145243e-06,
+      "loss": 0.6051,
+      "step": 21845
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7788916594860683e-06,
+      "loss": 0.5988,
+      "step": 21846
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.778138450272985e-06,
+      "loss": 0.6821,
+      "step": 21847
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.777385326684204e-06,
+      "loss": 0.6067,
+      "step": 21848
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.77663228872866e-06,
+      "loss": 0.6236,
+      "step": 21849
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7758793364152727e-06,
+      "loss": 0.6785,
+      "step": 21850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775126469752977e-06,
+      "loss": 0.6,
+      "step": 21851
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7743736887507e-06,
+      "loss": 0.5859,
+      "step": 21852
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7736209934173596e-06,
+      "loss": 0.6619,
+      "step": 21853
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.772868383761884e-06,
+      "loss": 0.6338,
+      "step": 21854
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7721158597931995e-06,
+      "loss": 0.6282,
+      "step": 21855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.771363421520218e-06,
+      "loss": 0.7323,
+      "step": 21856
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7706110689518708e-06,
+      "loss": 0.6579,
+      "step": 21857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7698588020970774e-06,
+      "loss": 0.6361,
+      "step": 21858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.769106620964752e-06,
+      "loss": 0.6299,
+      "step": 21859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7683545255638132e-06,
+      "loss": 0.5851,
+      "step": 21860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.767602515903184e-06,
+      "loss": 0.663,
+      "step": 21861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.76685059199177e-06,
+      "loss": 0.6417,
+      "step": 21862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.766098753838494e-06,
+      "loss": 0.7414,
+      "step": 21863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.765347001452272e-06,
+      "loss": 0.5826,
+      "step": 21864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7645953348420096e-06,
+      "loss": 0.6283,
+      "step": 21865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.763843754016623e-06,
+      "loss": 0.6971,
+      "step": 21866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.763092258985025e-06,
+      "loss": 0.6459,
+      "step": 21867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7623408497561176e-06,
+      "loss": 0.6729,
+      "step": 21868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7615895263388215e-06,
+      "loss": 0.6293,
+      "step": 21869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7608382887420338e-06,
+      "loss": 0.6096,
+      "step": 21870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7600871369746664e-06,
+      "loss": 0.615,
+      "step": 21871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.759336071045627e-06,
+      "loss": 0.6118,
+      "step": 21872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.758585090963817e-06,
+      "loss": 0.6553,
+      "step": 21873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7578341967381363e-06,
+      "loss": 0.6273,
+      "step": 21874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.757083388377498e-06,
+      "loss": 0.6207,
+      "step": 21875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7563326658907974e-06,
+      "loss": 0.6159,
+      "step": 21876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7555820292869352e-06,
+      "loss": 0.6186,
+      "step": 21877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7548314785748155e-06,
+      "loss": 0.6256,
+      "step": 21878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7540810137633313e-06,
+      "loss": 0.661,
+      "step": 21879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7533306348613786e-06,
+      "loss": 0.647,
+      "step": 21880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7525803418778652e-06,
+      "loss": 0.6743,
+      "step": 21881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.751830134821677e-06,
+      "loss": 0.612,
+      "step": 21882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7510800137017114e-06,
+      "loss": 0.6839,
+      "step": 21883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7503299785268645e-06,
+      "loss": 0.5792,
+      "step": 21884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7495800293060247e-06,
+      "loss": 0.644,
+      "step": 21885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.748830166048082e-06,
+      "loss": 0.6819,
+      "step": 21886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7480803887619358e-06,
+      "loss": 0.5393,
+      "step": 21887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7473306974564684e-06,
+      "loss": 0.592,
+      "step": 21888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.746581092140569e-06,
+      "loss": 0.6832,
+      "step": 21889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.74583157282313e-06,
+      "loss": 0.6036,
+      "step": 21890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.745082139513032e-06,
+      "loss": 0.6098,
+      "step": 21891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.744332792219159e-06,
+      "loss": 0.6452,
+      "step": 21892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7435835309504053e-06,
+      "loss": 0.6473,
+      "step": 21893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7428343557156455e-06,
+      "loss": 0.6037,
+      "step": 21894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.742085266523764e-06,
+      "loss": 0.6058,
+      "step": 21895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.741336263383647e-06,
+      "loss": 0.6251,
+      "step": 21896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7405873463041675e-06,
+      "loss": 0.623,
+      "step": 21897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.739838515294205e-06,
+      "loss": 0.6723,
+      "step": 21898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7390897703626475e-06,
+      "loss": 0.6298,
+      "step": 21899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7383411115183624e-06,
+      "loss": 0.6138,
+      "step": 21900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7375925387702296e-06,
+      "loss": 0.6356,
+      "step": 21901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.736844052127127e-06,
+      "loss": 0.5972,
+      "step": 21902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.736095651597923e-06,
+      "loss": 0.6306,
+      "step": 21903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7353473371914917e-06,
+      "loss": 0.6203,
+      "step": 21904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7345991089167135e-06,
+      "loss": 0.5617,
+      "step": 21905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7338509667824496e-06,
+      "loss": 0.6083,
+      "step": 21906
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.733102910797575e-06,
+      "loss": 0.7053,
+      "step": 21907
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7323549409709614e-06,
+      "loss": 0.6449,
+      "step": 21908
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7316070573114696e-06,
+      "loss": 0.5924,
+      "step": 21909
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.730859259827968e-06,
+      "loss": 0.6052,
+      "step": 21910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7301115485293315e-06,
+      "loss": 0.6073,
+      "step": 21911
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7293639234244175e-06,
+      "loss": 0.6102,
+      "step": 21912
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7286163845220892e-06,
+      "loss": 0.644,
+      "step": 21913
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.727868931831217e-06,
+      "loss": 0.6146,
+      "step": 21914
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.727121565360653e-06,
+      "loss": 0.5932,
+      "step": 21915
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.726374285119262e-06,
+      "loss": 0.6636,
+      "step": 21916
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.72562709111591e-06,
+      "loss": 0.6414,
+      "step": 21917
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7248799833594476e-06,
+      "loss": 0.615,
+      "step": 21918
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7241329618587363e-06,
+      "loss": 0.6384,
+      "step": 21919
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7233860266226365e-06,
+      "loss": 0.6183,
+      "step": 21920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.722639177659997e-06,
+      "loss": 0.5726,
+      "step": 21921
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7218924149796723e-06,
+      "loss": 0.6441,
+      "step": 21922
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.721145738590527e-06,
+      "loss": 0.611,
+      "step": 21923
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7203991485014025e-06,
+      "loss": 0.6726,
+      "step": 21924
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.719652644721156e-06,
+      "loss": 0.7045,
+      "step": 21925
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7189062272586396e-06,
+      "loss": 0.576,
+      "step": 21926
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7181598961226973e-06,
+      "loss": 0.618,
+      "step": 21927
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7174136513221784e-06,
+      "loss": 0.6386,
+      "step": 21928
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7166674928659387e-06,
+      "loss": 0.6264,
+      "step": 21929
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7159214207628157e-06,
+      "loss": 0.6123,
+      "step": 21930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.715175435021663e-06,
+      "loss": 0.6268,
+      "step": 21931
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7144295356513174e-06,
+      "loss": 0.6511,
+      "step": 21932
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.713683722660626e-06,
+      "loss": 0.5848,
+      "step": 21933
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7129379960584314e-06,
+      "loss": 0.5886,
+      "step": 21934
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7121923558535745e-06,
+      "loss": 0.6607,
+      "step": 21935
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.711446802054898e-06,
+      "loss": 0.6735,
+      "step": 21936
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7107013346712407e-06,
+      "loss": 0.6258,
+      "step": 21937
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7099559537114383e-06,
+      "loss": 0.6006,
+      "step": 21938
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7092106591843293e-06,
+      "loss": 0.6892,
+      "step": 21939
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7084654510987506e-06,
+      "loss": 0.5839,
+      "step": 21940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.707720329463539e-06,
+      "loss": 0.6516,
+      "step": 21941
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.706975294287527e-06,
+      "loss": 0.6398,
+      "step": 21942
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7062303455795513e-06,
+      "loss": 0.6431,
+      "step": 21943
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.705485483348439e-06,
+      "loss": 0.6358,
+      "step": 21944
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.704740707603024e-06,
+      "loss": 0.6401,
+      "step": 21945
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.703996018352136e-06,
+      "loss": 0.5911,
+      "step": 21946
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7032514156046054e-06,
+      "loss": 0.6392,
+      "step": 21947
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.702506899369258e-06,
+      "loss": 0.6077,
+      "step": 21948
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.701762469654927e-06,
+      "loss": 0.6107,
+      "step": 21949
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7010181264704315e-06,
+      "loss": 0.698,
+      "step": 21950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.700273869824599e-06,
+      "loss": 0.6423,
+      "step": 21951
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6995296997262545e-06,
+      "loss": 0.6294,
+      "step": 21952
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6987856161842197e-06,
+      "loss": 0.6172,
+      "step": 21953
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.698041619207318e-06,
+      "loss": 0.6645,
+      "step": 21954
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6972977088043727e-06,
+      "loss": 0.6747,
+      "step": 21955
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.696553884984198e-06,
+      "loss": 0.6322,
+      "step": 21956
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6958101477556167e-06,
+      "loss": 0.6463,
+      "step": 21957
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6950664971274443e-06,
+      "loss": 0.6643,
+      "step": 21958
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.694322933108501e-06,
+      "loss": 0.6509,
+      "step": 21959
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6935794557076e-06,
+      "loss": 0.6139,
+      "step": 21960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6928360649335606e-06,
+      "loss": 0.6807,
+      "step": 21961
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6920927607951907e-06,
+      "loss": 0.6545,
+      "step": 21962
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6913495433013047e-06,
+      "loss": 0.6044,
+      "step": 21963
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6906064124607157e-06,
+      "loss": 0.5934,
+      "step": 21964
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.689863368282235e-06,
+      "loss": 0.623,
+      "step": 21965
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6891204107746704e-06,
+      "loss": 0.6098,
+      "step": 21966
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.688377539946835e-06,
+      "loss": 0.6439,
+      "step": 21967
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.68763475580753e-06,
+      "loss": 0.6289,
+      "step": 21968
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6868920583655657e-06,
+      "loss": 0.6769,
+      "step": 21969
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.686149447629747e-06,
+      "loss": 0.6245,
+      "step": 21970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6854069236088777e-06,
+      "loss": 0.6359,
+      "step": 21971
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.684664486311763e-06,
+      "loss": 0.6364,
+      "step": 21972
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6839221357472078e-06,
+      "loss": 0.6124,
+      "step": 21973
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.683179871924008e-06,
+      "loss": 0.6377,
+      "step": 21974
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.682437694850966e-06,
+      "loss": 0.6264,
+      "step": 21975
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6816956045368815e-06,
+      "loss": 0.5735,
+      "step": 21976
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6809536009905547e-06,
+      "loss": 0.6032,
+      "step": 21977
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.68021168422078e-06,
+      "loss": 0.5764,
+      "step": 21978
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6794698542363595e-06,
+      "loss": 0.6641,
+      "step": 21979
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6787281110460806e-06,
+      "loss": 0.6001,
+      "step": 21980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.677986454658742e-06,
+      "loss": 0.6098,
+      "step": 21981
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6772448850831368e-06,
+      "loss": 0.6339,
+      "step": 21982
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6765034023280554e-06,
+      "loss": 0.6553,
+      "step": 21983
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6757620064022914e-06,
+      "loss": 0.6095,
+      "step": 21984
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6750206973146374e-06,
+      "loss": 0.6639,
+      "step": 21985
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6742794750738754e-06,
+      "loss": 0.6854,
+      "step": 21986
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.673538339688798e-06,
+      "loss": 0.5896,
+      "step": 21987
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.672797291168191e-06,
+      "loss": 0.6153,
+      "step": 21988
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.672056329520841e-06,
+      "loss": 0.6781,
+      "step": 21989
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6713154547555344e-06,
+      "loss": 0.5565,
+      "step": 21990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.670574666881055e-06,
+      "loss": 0.6429,
+      "step": 21991
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6698339659061835e-06,
+      "loss": 0.668,
+      "step": 21992
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.669093351839701e-06,
+      "loss": 0.6494,
+      "step": 21993
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.668352824690392e-06,
+      "loss": 0.6773,
+      "step": 21994
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.667612384467034e-06,
+      "loss": 0.697,
+      "step": 21995
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6668720311784102e-06,
+      "loss": 0.6859,
+      "step": 21996
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.666131764833293e-06,
+      "loss": 0.6381,
+      "step": 21997
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.665391585440459e-06,
+      "loss": 0.6317,
+      "step": 21998
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6646514930086874e-06,
+      "loss": 0.6373,
+      "step": 21999
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6639114875467507e-06,
+      "loss": 0.6337,
+      "step": 22000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6631715690634253e-06,
+      "loss": 0.6403,
+      "step": 22001
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6624317375674836e-06,
+      "loss": 0.629,
+      "step": 22002
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.661691993067693e-06,
+      "loss": 0.6169,
+      "step": 22003
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6609523355728264e-06,
+      "loss": 0.6748,
+      "step": 22004
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6602127650916545e-06,
+      "loss": 0.6509,
+      "step": 22005
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.659473281632945e-06,
+      "loss": 0.6212,
+      "step": 22006
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6587338852054657e-06,
+      "loss": 0.6639,
+      "step": 22007
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.657994575817985e-06,
+      "loss": 0.6711,
+      "step": 22008
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6572553534792645e-06,
+      "loss": 0.6134,
+      "step": 22009
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.65651621819807e-06,
+      "loss": 0.6751,
+      "step": 22010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.655777169983165e-06,
+      "loss": 0.6055,
+      "step": 22011
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6550382088433125e-06,
+      "loss": 0.6714,
+      "step": 22012
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6542993347872727e-06,
+      "loss": 0.62,
+      "step": 22013
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6535605478238103e-06,
+      "loss": 0.634,
+      "step": 22014
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6528218479616784e-06,
+      "loss": 0.6263,
+      "step": 22015
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.652083235209637e-06,
+      "loss": 0.632,
+      "step": 22016
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6513447095764443e-06,
+      "loss": 0.6193,
+      "step": 22017
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6506062710708556e-06,
+      "loss": 0.6903,
+      "step": 22018
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6498679197016275e-06,
+      "loss": 0.6532,
+      "step": 22019
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.649129655477516e-06,
+      "loss": 0.5941,
+      "step": 22020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.648391478407267e-06,
+      "loss": 0.6391,
+      "step": 22021
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.647653388499638e-06,
+      "loss": 0.5722,
+      "step": 22022
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.646915385763379e-06,
+      "loss": 0.638,
+      "step": 22023
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6461774702072397e-06,
+      "loss": 0.6777,
+      "step": 22024
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6454396418399685e-06,
+      "loss": 0.5726,
+      "step": 22025
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6447019006703177e-06,
+      "loss": 0.5959,
+      "step": 22026
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.643964246707027e-06,
+      "loss": 0.6622,
+      "step": 22027
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.643226679958846e-06,
+      "loss": 0.6628,
+      "step": 22028
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.642489200434518e-06,
+      "loss": 0.6746,
+      "step": 22029
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6417518081427885e-06,
+      "loss": 0.6269,
+      "step": 22030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6410145030923994e-06,
+      "loss": 0.6223,
+      "step": 22031
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.640277285292097e-06,
+      "loss": 0.6722,
+      "step": 22032
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6395401547506127e-06,
+      "loss": 0.6415,
+      "step": 22033
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.638803111476691e-06,
+      "loss": 0.6692,
+      "step": 22034
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6380661554790708e-06,
+      "loss": 0.6334,
+      "step": 22035
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.637329286766489e-06,
+      "loss": 0.6537,
+      "step": 22036
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6365925053476826e-06,
+      "loss": 0.6325,
+      "step": 22037
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6358558112313903e-06,
+      "loss": 0.6488,
+      "step": 22038
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6351192044263385e-06,
+      "loss": 0.6432,
+      "step": 22039
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6343826849412666e-06,
+      "loss": 0.6323,
+      "step": 22040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.633646252784905e-06,
+      "loss": 0.6643,
+      "step": 22041
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.632909907965986e-06,
+      "loss": 0.6164,
+      "step": 22042
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.632173650493238e-06,
+      "loss": 0.6064,
+      "step": 22043
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.631437480375396e-06,
+      "loss": 0.6539,
+      "step": 22044
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6307013976211813e-06,
+      "loss": 0.6536,
+      "step": 22045
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6299654022393227e-06,
+      "loss": 0.6085,
+      "step": 22046
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.629229494238549e-06,
+      "loss": 0.6266,
+      "step": 22047
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6284936736275825e-06,
+      "loss": 0.6597,
+      "step": 22048
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6277579404151477e-06,
+      "loss": 0.6384,
+      "step": 22049
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6270222946099723e-06,
+      "loss": 0.6308,
+      "step": 22050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6262867362207723e-06,
+      "loss": 0.6216,
+      "step": 22051
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.62555126525627e-06,
+      "loss": 0.5933,
+      "step": 22052
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6248158817251857e-06,
+      "loss": 0.6255,
+      "step": 22053
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6240805856362397e-06,
+      "loss": 0.6429,
+      "step": 22054
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.623345376998151e-06,
+      "loss": 0.6106,
+      "step": 22055
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.622610255819632e-06,
+      "loss": 0.627,
+      "step": 22056
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6218752221094015e-06,
+      "loss": 0.6406,
+      "step": 22057
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.621140275876175e-06,
+      "loss": 0.6443,
+      "step": 22058
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6204054171286596e-06,
+      "loss": 0.611,
+      "step": 22059
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6196706458755773e-06,
+      "loss": 0.6938,
+      "step": 22060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6189359621256373e-06,
+      "loss": 0.6137,
+      "step": 22061
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.618201365887546e-06,
+      "loss": 0.6342,
+      "step": 22062
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.617466857170016e-06,
+      "loss": 0.6694,
+      "step": 22063
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.616732435981757e-06,
+      "loss": 0.6134,
+      "step": 22064
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6159981023314695e-06,
+      "loss": 0.651,
+      "step": 22065
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6152638562278687e-06,
+      "loss": 0.6376,
+      "step": 22066
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.61452969767966e-06,
+      "loss": 0.6267,
+      "step": 22067
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.613795626695541e-06,
+      "loss": 0.6155,
+      "step": 22068
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.613061643284218e-06,
+      "loss": 0.6183,
+      "step": 22069
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6123277474543975e-06,
+      "loss": 0.6506,
+      "step": 22070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6115939392147705e-06,
+      "loss": 0.6299,
+      "step": 22071
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6108602185740463e-06,
+      "loss": 0.5991,
+      "step": 22072
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6101265855409254e-06,
+      "loss": 0.6362,
+      "step": 22073
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6093930401240985e-06,
+      "loss": 0.6242,
+      "step": 22074
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.608659582332266e-06,
+      "loss": 0.6786,
+      "step": 22075
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.607926212174128e-06,
+      "loss": 0.6544,
+      "step": 22076
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.6071929296583698e-06,
+      "loss": 0.6003,
+      "step": 22077
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.606459734793694e-06,
+      "loss": 0.7104,
+      "step": 22078
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.605726627588794e-06,
+      "loss": 0.6048,
+      "step": 22079
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6049936080523553e-06,
+      "loss": 0.6275,
+      "step": 22080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.604260676193072e-06,
+      "loss": 0.5871,
+      "step": 22081
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.603527832019638e-06,
+      "loss": 0.646,
+      "step": 22082
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6027950755407318e-06,
+      "loss": 0.6103,
+      "step": 22083
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.60206240676505e-06,
+      "loss": 0.6029,
+      "step": 22084
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.60132982570128e-06,
+      "loss": 0.5876,
+      "step": 22085
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.600597332358101e-06,
+      "loss": 0.5901,
+      "step": 22086
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.599864926744201e-06,
+      "loss": 0.5863,
+      "step": 22087
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.599132608868268e-06,
+      "loss": 0.6224,
+      "step": 22088
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.598400378738972e-06,
+      "loss": 0.652,
+      "step": 22089
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5976682363650065e-06,
+      "loss": 0.605,
+      "step": 22090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.596936181755051e-06,
+      "loss": 0.5973,
+      "step": 22091
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.596204214917779e-06,
+      "loss": 0.6642,
+      "step": 22092
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5954723358618718e-06,
+      "loss": 0.6798,
+      "step": 22093
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5947405445960105e-06,
+      "loss": 0.6369,
+      "step": 22094
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.594008841128861e-06,
+      "loss": 0.652,
+      "step": 22095
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5932772254691085e-06,
+      "loss": 0.6179,
+      "step": 22096
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.592545697625427e-06,
+      "loss": 0.6831,
+      "step": 22097
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.591814257606483e-06,
+      "loss": 0.6213,
+      "step": 22098
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.591082905420954e-06,
+      "loss": 0.5994,
+      "step": 22099
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5903516410775107e-06,
+      "loss": 0.6581,
+      "step": 22100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5896204645848177e-06,
+      "loss": 0.6652,
+      "step": 22101
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.58888937595155e-06,
+      "loss": 0.6072,
+      "step": 22102
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.588158375186377e-06,
+      "loss": 0.6394,
+      "step": 22103
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5874274622979602e-06,
+      "loss": 0.5562,
+      "step": 22104
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5866966372949675e-06,
+      "loss": 0.6228,
+      "step": 22105
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.585965900186067e-06,
+      "loss": 0.6595,
+      "step": 22106
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.585235250979913e-06,
+      "loss": 0.6307,
+      "step": 22107
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.584504689685179e-06,
+      "loss": 0.6361,
+      "step": 22108
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5837742163105238e-06,
+      "loss": 0.6415,
+      "step": 22109
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5830438308646055e-06,
+      "loss": 0.6416,
+      "step": 22110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5823135333560843e-06,
+      "loss": 0.6447,
+      "step": 22111
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5815833237936226e-06,
+      "loss": 0.5791,
+      "step": 22112
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.580853202185869e-06,
+      "loss": 0.6584,
+      "step": 22113
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.58012316854149e-06,
+      "loss": 0.7071,
+      "step": 22114
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.579393222869139e-06,
+      "loss": 0.6958,
+      "step": 22115
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5786633651774652e-06,
+      "loss": 0.6019,
+      "step": 22116
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5779335954751282e-06,
+      "loss": 0.6629,
+      "step": 22117
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5772039137707748e-06,
+      "loss": 0.667,
+      "step": 22118
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5764743200730556e-06,
+      "loss": 0.6207,
+      "step": 22119
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5757448143906303e-06,
+      "loss": 0.6082,
+      "step": 22120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.57501539673214e-06,
+      "loss": 0.6415,
+      "step": 22121
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.574286067106233e-06,
+      "loss": 0.6764,
+      "step": 22122
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5735568255215624e-06,
+      "loss": 0.646,
+      "step": 22123
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5728276719867674e-06,
+      "loss": 0.6169,
+      "step": 22124
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5720986065104912e-06,
+      "loss": 0.6621,
+      "step": 22125
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.57136962910139e-06,
+      "loss": 0.6731,
+      "step": 22126
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5706407397680955e-06,
+      "loss": 0.6163,
+      "step": 22127
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.569911938519253e-06,
+      "loss": 0.6258,
+      "step": 22128
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5691832253635075e-06,
+      "loss": 0.6877,
+      "step": 22129
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.568454600309491e-06,
+      "loss": 0.6042,
+      "step": 22130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.567726063365843e-06,
+      "loss": 0.639,
+      "step": 22131
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.566997614541211e-06,
+      "loss": 0.6961,
+      "step": 22132
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5662692538442213e-06,
+      "loss": 0.6335,
+      "step": 22133
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.565540981283514e-06,
+      "loss": 0.6115,
+      "step": 22134
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.564812796867726e-06,
+      "loss": 0.5503,
+      "step": 22135
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5640847006054847e-06,
+      "loss": 0.6159,
+      "step": 22136
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.563356692505423e-06,
+      "loss": 0.6326,
+      "step": 22137
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.562628772576181e-06,
+      "loss": 0.6019,
+      "step": 22138
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5619009408263793e-06,
+      "loss": 0.6334,
+      "step": 22139
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5611731972646527e-06,
+      "loss": 0.561,
+      "step": 22140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.56044554189963e-06,
+      "loss": 0.6405,
+      "step": 22141
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5597179747399337e-06,
+      "loss": 0.6517,
+      "step": 22142
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5589904957941903e-06,
+      "loss": 0.5968,
+      "step": 22143
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5582631050710326e-06,
+      "loss": 0.6147,
+      "step": 22144
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5575358025790776e-06,
+      "loss": 0.6618,
+      "step": 22145
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.55680858832695e-06,
+      "loss": 0.6191,
+      "step": 22146
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5560814623232755e-06,
+      "loss": 0.6289,
+      "step": 22147
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.555354424576668e-06,
+      "loss": 0.6733,
+      "step": 22148
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.554627475095749e-06,
+      "loss": 0.6516,
+      "step": 22149
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5539006138891464e-06,
+      "loss": 0.6508,
+      "step": 22150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.553173840965467e-06,
+      "loss": 0.6561,
+      "step": 22151
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.552447156333333e-06,
+      "loss": 0.6471,
+      "step": 22152
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5517205600013617e-06,
+      "loss": 0.6254,
+      "step": 22153
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.550994051978163e-06,
+      "loss": 0.6445,
+      "step": 22154
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5502676322723485e-06,
+      "loss": 0.6552,
+      "step": 22155
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5495413008925417e-06,
+      "loss": 0.6205,
+      "step": 22156
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.548815057847346e-06,
+      "loss": 0.5974,
+      "step": 22157
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5480889031453715e-06,
+      "loss": 0.5815,
+      "step": 22158
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.547362836795234e-06,
+      "loss": 0.6657,
+      "step": 22159
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5466368588055347e-06,
+      "loss": 0.6745,
+      "step": 22160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5459109691848805e-06,
+      "loss": 0.6462,
+      "step": 22161
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5451851679418882e-06,
+      "loss": 0.6496,
+      "step": 22162
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.544459455085152e-06,
+      "loss": 0.6329,
+      "step": 22163
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5437338306232807e-06,
+      "loss": 0.6769,
+      "step": 22164
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5430082945648793e-06,
+      "loss": 0.669,
+      "step": 22165
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.542282846918546e-06,
+      "loss": 0.64,
+      "step": 22166
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5415574876928806e-06,
+      "loss": 0.601,
+      "step": 22167
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5408322168964915e-06,
+      "loss": 0.6433,
+      "step": 22168
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5401070345379688e-06,
+      "loss": 0.5776,
+      "step": 22169
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5393819406259145e-06,
+      "loss": 0.6269,
+      "step": 22170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.538656935168928e-06,
+      "loss": 0.6538,
+      "step": 22171
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5379320181756007e-06,
+      "loss": 0.6658,
+      "step": 22172
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5372071896545237e-06,
+      "loss": 0.6787,
+      "step": 22173
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.536482449614304e-06,
+      "loss": 0.659,
+      "step": 22174
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.535757798063522e-06,
+      "loss": 0.578,
+      "step": 22175
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5350332350107732e-06,
+      "loss": 0.5711,
+      "step": 22176
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5343087604646522e-06,
+      "loss": 0.6263,
+      "step": 22177
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.533584374433742e-06,
+      "loss": 0.5834,
+      "step": 22178
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.532860076926634e-06,
+      "loss": 0.6323,
+      "step": 22179
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.532135867951916e-06,
+      "loss": 0.6229,
+      "step": 22180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5314117475181732e-06,
+      "loss": 0.6264,
+      "step": 22181
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.530687715633996e-06,
+      "loss": 0.6463,
+      "step": 22182
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.529963772307962e-06,
+      "loss": 0.6804,
+      "step": 22183
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.529239917548656e-06,
+      "loss": 0.6364,
+      "step": 22184
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5285161513646615e-06,
+      "loss": 0.6292,
+      "step": 22185
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5277924737645598e-06,
+      "loss": 0.5985,
+      "step": 22186
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.527068884756931e-06,
+      "loss": 0.598,
+      "step": 22187
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.526345384350357e-06,
+      "loss": 0.6837,
+      "step": 22188
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.525621972553409e-06,
+      "loss": 0.7108,
+      "step": 22189
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5248986493746688e-06,
+      "loss": 0.5728,
+      "step": 22190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5241754148227103e-06,
+      "loss": 0.6145,
+      "step": 22191
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.523452268906109e-06,
+      "loss": 0.6022,
+      "step": 22192
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.522729211633439e-06,
+      "loss": 0.6757,
+      "step": 22193
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5220062430132764e-06,
+      "loss": 0.6272,
+      "step": 22194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5212833630541866e-06,
+      "loss": 0.627,
+      "step": 22195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.520560571764743e-06,
+      "loss": 0.6542,
+      "step": 22196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5198378691535154e-06,
+      "loss": 0.6345,
+      "step": 22197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.519115255229072e-06,
+      "loss": 0.638,
+      "step": 22198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.51839272999998e-06,
+      "loss": 0.6593,
+      "step": 22199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.517670293474811e-06,
+      "loss": 0.6081,
+      "step": 22200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5169479456621216e-06,
+      "loss": 0.6709,
+      "step": 22201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5162256865704813e-06,
+      "loss": 0.641,
+      "step": 22202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5155035162084518e-06,
+      "loss": 0.6205,
+      "step": 22203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5147814345845966e-06,
+      "loss": 0.6667,
+      "step": 22204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.514059441707476e-06,
+      "loss": 0.6335,
+      "step": 22205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.513337537585654e-06,
+      "loss": 0.6497,
+      "step": 22206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5126157222276824e-06,
+      "loss": 0.663,
+      "step": 22207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.511893995642124e-06,
+      "loss": 0.6187,
+      "step": 22208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5111723578375337e-06,
+      "loss": 0.6345,
+      "step": 22209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5104508088224688e-06,
+      "loss": 0.5993,
+      "step": 22210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.509729348605484e-06,
+      "loss": 0.576,
+      "step": 22211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5090079771951357e-06,
+      "loss": 0.6173,
+      "step": 22212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5082866945999716e-06,
+      "loss": 0.5937,
+      "step": 22213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.507565500828546e-06,
+      "loss": 0.6829,
+      "step": 22214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5068443958894097e-06,
+      "loss": 0.6311,
+      "step": 22215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5061233797911112e-06,
+      "loss": 0.6511,
+      "step": 22216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5054024525421995e-06,
+      "loss": 0.5827,
+      "step": 22217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5046816141512254e-06,
+      "loss": 0.6254,
+      "step": 22218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.50396086462673e-06,
+      "loss": 0.5828,
+      "step": 22219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5032402039772607e-06,
+      "loss": 0.6703,
+      "step": 22220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.502519632211362e-06,
+      "loss": 0.6561,
+      "step": 22221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5017991493375783e-06,
+      "loss": 0.6167,
+      "step": 22222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.501078755364451e-06,
+      "loss": 0.6499,
+      "step": 22223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.500358450300523e-06,
+      "loss": 0.6284,
+      "step": 22224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.49963823415433e-06,
+      "loss": 0.6317,
+      "step": 22225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4989181069344147e-06,
+      "loss": 0.6271,
+      "step": 22226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4981980686493133e-06,
+      "loss": 0.6924,
+      "step": 22227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4974781193075646e-06,
+      "loss": 0.6358,
+      "step": 22228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.496758258917703e-06,
+      "loss": 0.6202,
+      "step": 22229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.496038487488266e-06,
+      "loss": 0.6414,
+      "step": 22230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.495318805027783e-06,
+      "loss": 0.6147,
+      "step": 22231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.494599211544789e-06,
+      "loss": 0.6581,
+      "step": 22232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.493879707047816e-06,
+      "loss": 0.6476,
+      "step": 22233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4931602915453945e-06,
+      "loss": 0.6637,
+      "step": 22234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4924409650460535e-06,
+      "loss": 0.651,
+      "step": 22235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.491721727558325e-06,
+      "loss": 0.6367,
+      "step": 22236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.491002579090731e-06,
+      "loss": 0.6826,
+      "step": 22237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.490283519651799e-06,
+      "loss": 0.6173,
+      "step": 22238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4895645492500576e-06,
+      "loss": 0.6433,
+      "step": 22239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4888456678940276e-06,
+      "loss": 0.5853,
+      "step": 22240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.488126875592237e-06,
+      "loss": 0.6685,
+      "step": 22241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4874081723532027e-06,
+      "loss": 0.6098,
+      "step": 22242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4866895581854476e-06,
+      "loss": 0.6599,
+      "step": 22243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.485971033097493e-06,
+      "loss": 0.6759,
+      "step": 22244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4852525970978556e-06,
+      "loss": 0.6592,
+      "step": 22245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4845342501950565e-06,
+      "loss": 0.6216,
+      "step": 22246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4838159923976123e-06,
+      "loss": 0.6592,
+      "step": 22247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.483097823714036e-06,
+      "loss": 0.6555,
+      "step": 22248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.482379744152843e-06,
+      "loss": 0.666,
+      "step": 22249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4816617537225483e-06,
+      "loss": 0.6547,
+      "step": 22250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.480943852431664e-06,
+      "loss": 0.6276,
+      "step": 22251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4802260402887023e-06,
+      "loss": 0.6172,
+      "step": 22252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.479508317302176e-06,
+      "loss": 0.6444,
+      "step": 22253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4787906834805896e-06,
+      "loss": 0.6169,
+      "step": 22254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.478073138832453e-06,
+      "loss": 0.6547,
+      "step": 22255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.477355683366276e-06,
+      "loss": 0.6275,
+      "step": 22256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4766383170905628e-06,
+      "loss": 0.6294,
+      "step": 22257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.47592104001382e-06,
+      "loss": 0.5406,
+      "step": 22258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.475203852144553e-06,
+      "loss": 0.5928,
+      "step": 22259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.474486753491262e-06,
+      "loss": 0.643,
+      "step": 22260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.47376974406245e-06,
+      "loss": 0.6214,
+      "step": 22261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4730528238666174e-06,
+      "loss": 0.5944,
+      "step": 22262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.472335992912266e-06,
+      "loss": 0.6546,
+      "step": 22263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.471619251207893e-06,
+      "loss": 0.6554,
+      "step": 22264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4709025987620017e-06,
+      "loss": 0.7192,
+      "step": 22265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4701860355830796e-06,
+      "loss": 0.6985,
+      "step": 22266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4694695616796294e-06,
+      "loss": 0.6416,
+      "step": 22267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4687531770601426e-06,
+      "loss": 0.6014,
+      "step": 22268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4680368817331136e-06,
+      "loss": 0.66,
+      "step": 22269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4673206757070356e-06,
+      "loss": 0.6038,
+      "step": 22270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.466604558990403e-06,
+      "loss": 0.629,
+      "step": 22271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4658885315917003e-06,
+      "loss": 0.6575,
+      "step": 22272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4651725935194202e-06,
+      "loss": 0.6218,
+      "step": 22273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.46445674478205e-06,
+      "loss": 0.7432,
+      "step": 22274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.463740985388078e-06,
+      "loss": 0.6407,
+      "step": 22275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4630253153459892e-06,
+      "loss": 0.6303,
+      "step": 22276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4623097346642734e-06,
+      "loss": 0.6608,
+      "step": 22277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.461594243351407e-06,
+      "loss": 0.6079,
+      "step": 22278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.460878841415878e-06,
+      "loss": 0.6496,
+      "step": 22279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4601635288661674e-06,
+      "loss": 0.6169,
+      "step": 22280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4594483057107554e-06,
+      "loss": 0.6198,
+      "step": 22281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.458733171958123e-06,
+      "loss": 0.6657,
+      "step": 22282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4580181276167516e-06,
+      "loss": 0.6354,
+      "step": 22283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4573031726951136e-06,
+      "loss": 0.674,
+      "step": 22284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.456588307201687e-06,
+      "loss": 0.6594,
+      "step": 22285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.455873531144949e-06,
+      "loss": 0.623,
+      "step": 22286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.455158844533373e-06,
+      "loss": 0.5739,
+      "step": 22287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4544442473754336e-06,
+      "loss": 0.6523,
+      "step": 22288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4537297396796058e-06,
+      "loss": 0.6047,
+      "step": 22289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.453015321454355e-06,
+      "loss": 0.6374,
+      "step": 22290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4523009927081556e-06,
+      "loss": 0.6691,
+      "step": 22291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4515867534494743e-06,
+      "loss": 0.612,
+      "step": 22292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4508726036867815e-06,
+      "loss": 0.7,
+      "step": 22293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4501585434285435e-06,
+      "loss": 0.5909,
+      "step": 22294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4494445726832296e-06,
+      "loss": 0.6007,
+      "step": 22295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.448730691459299e-06,
+      "loss": 0.6159,
+      "step": 22296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4480168997652177e-06,
+      "loss": 0.612,
+      "step": 22297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4473031976094507e-06,
+      "loss": 0.6359,
+      "step": 22298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.446589585000457e-06,
+      "loss": 0.6494,
+      "step": 22299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4458760619467004e-06,
+      "loss": 0.6101,
+      "step": 22300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.445162628456641e-06,
+      "loss": 0.6674,
+      "step": 22301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4444492845387323e-06,
+      "loss": 0.5484,
+      "step": 22302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.443736030201439e-06,
+      "loss": 0.6191,
+      "step": 22303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4430228654532075e-06,
+      "loss": 0.6226,
+      "step": 22304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4423097903025026e-06,
+      "loss": 0.5942,
+      "step": 22305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.441596804757779e-06,
+      "loss": 0.6207,
+      "step": 22306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.440883908827483e-06,
+      "loss": 0.6747,
+      "step": 22307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4401711025200714e-06,
+      "loss": 0.6352,
+      "step": 22308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.439458385843997e-06,
+      "loss": 0.698,
+      "step": 22309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.438745758807701e-06,
+      "loss": 0.6504,
+      "step": 22310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4380332214196433e-06,
+      "loss": 0.6037,
+      "step": 22311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.437320773688271e-06,
+      "loss": 0.6445,
+      "step": 22312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.436608415622024e-06,
+      "loss": 0.6752,
+      "step": 22313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4358961472293506e-06,
+      "loss": 0.6456,
+      "step": 22314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4351839685187017e-06,
+      "loss": 0.6458,
+      "step": 22315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.43447187949851e-06,
+      "loss": 0.6444,
+      "step": 22316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4337598801772277e-06,
+      "loss": 0.6468,
+      "step": 22317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.433047970563297e-06,
+      "loss": 0.6373,
+      "step": 22318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.432336150665151e-06,
+      "loss": 0.669,
+      "step": 22319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.431624420491233e-06,
+      "loss": 0.6114,
+      "step": 22320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4309127800499843e-06,
+      "loss": 0.6643,
+      "step": 22321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4302012293498345e-06,
+      "loss": 0.6352,
+      "step": 22322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.429489768399228e-06,
+      "loss": 0.6329,
+      "step": 22323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.428778397206599e-06,
+      "loss": 0.6329,
+      "step": 22324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.428067115780377e-06,
+      "loss": 0.6533,
+      "step": 22325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.427355924128999e-06,
+      "loss": 0.6007,
+      "step": 22326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4266448222608975e-06,
+      "loss": 0.606,
+      "step": 22327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4259338101844965e-06,
+      "loss": 0.6329,
+      "step": 22328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.425222887908235e-06,
+      "loss": 0.6696,
+      "step": 22329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.424512055440541e-06,
+      "loss": 0.6765,
+      "step": 22330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4238013127898363e-06,
+      "loss": 0.6144,
+      "step": 22331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.423090659964551e-06,
+      "loss": 0.6053,
+      "step": 22332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.422380096973114e-06,
+      "loss": 0.6038,
+      "step": 22333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.421669623823941e-06,
+      "loss": 0.6134,
+      "step": 22334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4209592405254634e-06,
+      "loss": 0.6243,
+      "step": 22335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4202489470861046e-06,
+      "loss": 0.7025,
+      "step": 22336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.41953874351428e-06,
+      "loss": 0.6126,
+      "step": 22337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.418828629818414e-06,
+      "loss": 0.6221,
+      "step": 22338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.418118606006927e-06,
+      "loss": 0.6467,
+      "step": 22339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4174086720882295e-06,
+      "loss": 0.6538,
+      "step": 22340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4166988280707472e-06,
+      "loss": 0.6302,
+      "step": 22341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4159890739628955e-06,
+      "loss": 0.6851,
+      "step": 22342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4152794097730846e-06,
+      "loss": 0.6259,
+      "step": 22343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4145698355097314e-06,
+      "loss": 0.6307,
+      "step": 22344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.413860351181251e-06,
+      "loss": 0.6742,
+      "step": 22345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4131509567960467e-06,
+      "loss": 0.6908,
+      "step": 22346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4124416523625384e-06,
+      "loss": 0.6273,
+      "step": 22347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.411732437889135e-06,
+      "loss": 0.5989,
+      "step": 22348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4110233133842396e-06,
+      "loss": 0.7009,
+      "step": 22349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.410314278856264e-06,
+      "loss": 0.6603,
+      "step": 22350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.409605334313616e-06,
+      "loss": 0.603,
+      "step": 22351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.408896479764693e-06,
+      "loss": 0.6201,
+      "step": 22352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.408187715217908e-06,
+      "loss": 0.6561,
+      "step": 22353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4074790406816638e-06,
+      "loss": 0.6748,
+      "step": 22354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.406770456164357e-06,
+      "loss": 0.6553,
+      "step": 22355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.406061961674393e-06,
+      "loss": 0.616,
+      "step": 22356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4053535572201725e-06,
+      "loss": 0.5571,
+      "step": 22357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4046452428100873e-06,
+      "loss": 0.6028,
+      "step": 22358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4039370184525445e-06,
+      "loss": 0.6004,
+      "step": 22359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4032288841559403e-06,
+      "loss": 0.6854,
+      "step": 22360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4025208399286647e-06,
+      "loss": 0.6048,
+      "step": 22361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.401812885779116e-06,
+      "loss": 0.6576,
+      "step": 22362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4011050217156883e-06,
+      "loss": 0.6227,
+      "step": 22363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4003972477467686e-06,
+      "loss": 0.6308,
+      "step": 22364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3996895638807593e-06,
+      "loss": 0.6012,
+      "step": 22365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.398981970126041e-06,
+      "loss": 0.5787,
+      "step": 22366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3982744664910075e-06,
+      "loss": 0.6156,
+      "step": 22367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.3975670529840478e-06,
+      "loss": 0.6472,
+      "step": 22368
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3968597296135464e-06,
+      "loss": 0.6154,
+      "step": 22369
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.396152496387888e-06,
+      "loss": 0.6283,
+      "step": 22370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.395445353315465e-06,
+      "loss": 0.6232,
+      "step": 22371
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3947383004046566e-06,
+      "loss": 0.6244,
+      "step": 22372
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3940313376638446e-06,
+      "loss": 0.6263,
+      "step": 22373
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3933244651014167e-06,
+      "loss": 0.6374,
+      "step": 22374
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.392617682725745e-06,
+      "loss": 0.5795,
+      "step": 22375
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3919109905452133e-06,
+      "loss": 0.6385,
+      "step": 22376
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3912043885682057e-06,
+      "loss": 0.675,
+      "step": 22377
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.390497876803093e-06,
+      "loss": 0.6304,
+      "step": 22378
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3897914552582536e-06,
+      "loss": 0.6883,
+      "step": 22379
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3890851239420663e-06,
+      "loss": 0.6248,
+      "step": 22380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3883788828628997e-06,
+      "loss": 0.6273,
+      "step": 22381
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3876727320291305e-06,
+      "loss": 0.6996,
+      "step": 22382
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.38696667144913e-06,
+      "loss": 0.6274,
+      "step": 22383
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.386260701131271e-06,
+      "loss": 0.572,
+      "step": 22384
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.385554821083923e-06,
+      "loss": 0.6289,
+      "step": 22385
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.384849031315457e-06,
+      "loss": 0.6222,
+      "step": 22386
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3841433318342367e-06,
+      "loss": 0.6211,
+      "step": 22387
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3834377226486317e-06,
+      "loss": 0.6407,
+      "step": 22388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3827322037670062e-06,
+      "loss": 0.6635,
+      "step": 22389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3820267751977276e-06,
+      "loss": 0.6431,
+      "step": 22390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.381321436949158e-06,
+      "loss": 0.6518,
+      "step": 22391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.380616189029663e-06,
+      "loss": 0.5962,
+      "step": 22392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3799110314475982e-06,
+      "loss": 0.6362,
+      "step": 22393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3792059642113284e-06,
+      "loss": 0.5733,
+      "step": 22394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.378500987329212e-06,
+      "loss": 0.6411,
+      "step": 22395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3777961008096082e-06,
+      "loss": 0.6832,
+      "step": 22396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3770913046608736e-06,
+      "loss": 0.6509,
+      "step": 22397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.376386598891367e-06,
+      "loss": 0.6189,
+      "step": 22398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.375681983509438e-06,
+      "loss": 0.6463,
+      "step": 22399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3749774585234442e-06,
+      "loss": 0.6563,
+      "step": 22400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.374273023941739e-06,
+      "loss": 0.6296,
+      "step": 22401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3735686797726732e-06,
+      "loss": 0.5617,
+      "step": 22402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.372864426024598e-06,
+      "loss": 0.6096,
+      "step": 22403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3721602627058673e-06,
+      "loss": 0.652,
+      "step": 22404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.371456189824821e-06,
+      "loss": 0.5452,
+      "step": 22405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3707522073898137e-06,
+      "loss": 0.6268,
+      "step": 22406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.370048315409189e-06,
+      "loss": 0.638,
+      "step": 22407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.369344513891294e-06,
+      "loss": 0.6758,
+      "step": 22408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.368640802844473e-06,
+      "loss": 0.6069,
+      "step": 22409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3679371822770715e-06,
+      "loss": 0.6568,
+      "step": 22410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.367233652197426e-06,
+      "loss": 0.6072,
+      "step": 22411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3665302126138812e-06,
+      "loss": 0.6825,
+      "step": 22412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.365826863534777e-06,
+      "loss": 0.6358,
+      "step": 22413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3651236049684523e-06,
+      "loss": 0.6349,
+      "step": 22414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3644204369232458e-06,
+      "loss": 0.612,
+      "step": 22415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.363717359407498e-06,
+      "loss": 0.5695,
+      "step": 22416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.363014372429536e-06,
+      "loss": 0.6345,
+      "step": 22417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3623114759977015e-06,
+      "loss": 0.6251,
+      "step": 22418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3616086701203243e-06,
+      "loss": 0.607,
+      "step": 22419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.36090595480574e-06,
+      "loss": 0.654,
+      "step": 22420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3602033300622797e-06,
+      "loss": 0.5906,
+      "step": 22421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.359500795898275e-06,
+      "loss": 0.6065,
+      "step": 22422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3587983523220514e-06,
+      "loss": 0.6135,
+      "step": 22423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3580959993419395e-06,
+      "loss": 0.6162,
+      "step": 22424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3573937369662658e-06,
+      "loss": 0.6257,
+      "step": 22425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3566915652033582e-06,
+      "loss": 0.61,
+      "step": 22426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3559894840615437e-06,
+      "loss": 0.6567,
+      "step": 22427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.355287493549141e-06,
+      "loss": 0.6127,
+      "step": 22428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3545855936744746e-06,
+      "loss": 0.593,
+      "step": 22429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3538837844458685e-06,
+      "loss": 0.6475,
+      "step": 22430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.353182065871642e-06,
+      "loss": 0.6336,
+      "step": 22431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3524804379601152e-06,
+      "loss": 0.6114,
+      "step": 22432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.35177890071961e-06,
+      "loss": 0.6693,
+      "step": 22433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.351077454158438e-06,
+      "loss": 0.6743,
+      "step": 22434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.350376098284919e-06,
+      "loss": 0.5762,
+      "step": 22435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3496748331073684e-06,
+      "loss": 0.6325,
+      "step": 22436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3489736586341005e-06,
+      "loss": 0.6144,
+      "step": 22437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.348272574873428e-06,
+      "loss": 0.6393,
+      "step": 22438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3475715818336663e-06,
+      "loss": 0.6093,
+      "step": 22439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.346870679523121e-06,
+      "loss": 0.6372,
+      "step": 22440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3461698679501056e-06,
+      "loss": 0.6523,
+      "step": 22441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3454691471229284e-06,
+      "loss": 0.6266,
+      "step": 22442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.344768517049897e-06,
+      "loss": 0.5784,
+      "step": 22443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3440679777393195e-06,
+      "loss": 0.5868,
+      "step": 22444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3433675291995027e-06,
+      "loss": 0.6074,
+      "step": 22445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3426671714387483e-06,
+      "loss": 0.5582,
+      "step": 22446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3419669044653603e-06,
+      "loss": 0.6158,
+      "step": 22447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.341266728287642e-06,
+      "loss": 0.5925,
+      "step": 22448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.340566642913895e-06,
+      "loss": 0.628,
+      "step": 22449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.339866648352419e-06,
+      "loss": 0.6457,
+      "step": 22450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3391667446115184e-06,
+      "loss": 0.6062,
+      "step": 22451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3384669316994833e-06,
+      "loss": 0.645,
+      "step": 22452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.337767209624614e-06,
+      "loss": 0.6468,
+      "step": 22453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3370675783952077e-06,
+      "loss": 0.5891,
+      "step": 22454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3363680380195598e-06,
+      "loss": 0.6462,
+      "step": 22455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3356685885059628e-06,
+      "loss": 0.5555,
+      "step": 22456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3349692298627125e-06,
+      "loss": 0.6577,
+      "step": 22457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3342699620980957e-06,
+      "loss": 0.6191,
+      "step": 22458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.333570785220406e-06,
+      "loss": 0.658,
+      "step": 22459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.332871699237932e-06,
+      "loss": 0.6249,
+      "step": 22460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3321727041589635e-06,
+      "loss": 0.6534,
+      "step": 22461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.331473799991788e-06,
+      "loss": 0.6746,
+      "step": 22462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3307749867446926e-06,
+      "loss": 0.6586,
+      "step": 22463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.33007626442596e-06,
+      "loss": 0.5821,
+      "step": 22464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3293776330438756e-06,
+      "loss": 0.6648,
+      "step": 22465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3286790926067227e-06,
+      "loss": 0.632,
+      "step": 22466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3279806431227827e-06,
+      "loss": 0.6245,
+      "step": 22467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3272822846003374e-06,
+      "loss": 0.6457,
+      "step": 22468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3265840170476704e-06,
+      "loss": 0.6222,
+      "step": 22469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3258858404730534e-06,
+      "loss": 0.5989,
+      "step": 22470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.325187754884768e-06,
+      "loss": 0.5593,
+      "step": 22471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3244897602910897e-06,
+      "loss": 0.6476,
+      "step": 22472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.323791856700295e-06,
+      "loss": 0.6536,
+      "step": 22473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3230940441206584e-06,
+      "loss": 0.6759,
+      "step": 22474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3223963225604563e-06,
+      "loss": 0.5975,
+      "step": 22475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.321698692027955e-06,
+      "loss": 0.6242,
+      "step": 22476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.321001152531429e-06,
+      "loss": 0.6408,
+      "step": 22477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.320303704079149e-06,
+      "loss": 0.6443,
+      "step": 22478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3196063466793816e-06,
+      "loss": 0.636,
+      "step": 22479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.318909080340398e-06,
+      "loss": 0.648,
+      "step": 22480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3182119050704668e-06,
+      "loss": 0.6003,
+      "step": 22481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3175148208778474e-06,
+      "loss": 0.7212,
+      "step": 22482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.316817827770809e-06,
+      "loss": 0.6538,
+      "step": 22483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3161209257576144e-06,
+      "loss": 0.684,
+      "step": 22484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3154241148465263e-06,
+      "loss": 0.6591,
+      "step": 22485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.314727395045806e-06,
+      "loss": 0.6415,
+      "step": 22486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3140307663637174e-06,
+      "loss": 0.6532,
+      "step": 22487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3133342288085137e-06,
+      "loss": 0.6243,
+      "step": 22488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.312637782388457e-06,
+      "loss": 0.576,
+      "step": 22489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3119414271118036e-06,
+      "loss": 0.6497,
+      "step": 22490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.311245162986809e-06,
+      "loss": 0.5992,
+      "step": 22491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.310548990021734e-06,
+      "loss": 0.6062,
+      "step": 22492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3098529082248245e-06,
+      "loss": 0.6407,
+      "step": 22493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.309156917604337e-06,
+      "loss": 0.6226,
+      "step": 22494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.308461018168523e-06,
+      "loss": 0.6227,
+      "step": 22495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.307765209925633e-06,
+      "loss": 0.6521,
+      "step": 22496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3070694928839176e-06,
+      "loss": 0.587,
+      "step": 22497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3063738670516277e-06,
+      "loss": 0.6371,
+      "step": 22498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3056783324370046e-06,
+      "loss": 0.6442,
+      "step": 22499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3049828890482995e-06,
+      "loss": 0.5823,
+      "step": 22500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3042875368937555e-06,
+      "loss": 0.6145,
+      "step": 22501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3035922759816187e-06,
+      "loss": 0.6072,
+      "step": 22502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.302897106320131e-06,
+      "loss": 0.5834,
+      "step": 22503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3022020279175383e-06,
+      "loss": 0.6202,
+      "step": 22504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.301507040782075e-06,
+      "loss": 0.6522,
+      "step": 22505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3008121449219844e-06,
+      "loss": 0.5962,
+      "step": 22506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3001173403455056e-06,
+      "loss": 0.6297,
+      "step": 22507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2994226270608754e-06,
+      "loss": 0.6817,
+      "step": 22508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2987280050763316e-06,
+      "loss": 0.7229,
+      "step": 22509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2980334744001133e-06,
+      "loss": 0.6711,
+      "step": 22510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2973390350404478e-06,
+      "loss": 0.587,
+      "step": 22511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2966446870055715e-06,
+      "loss": 0.6188,
+      "step": 22512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2959504303037173e-06,
+      "loss": 0.6581,
+      "step": 22513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2952562649431166e-06,
+      "loss": 0.6139,
+      "step": 22514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.294562190931999e-06,
+      "loss": 0.6282,
+      "step": 22515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.293868208278598e-06,
+      "loss": 0.5986,
+      "step": 22516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2931743169911335e-06,
+      "loss": 0.687,
+      "step": 22517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2924805170778373e-06,
+      "loss": 0.6397,
+      "step": 22518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2917868085469353e-06,
+      "loss": 0.6429,
+      "step": 22519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2910931914066503e-06,
+      "loss": 0.6192,
+      "step": 22520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2903996656652085e-06,
+      "loss": 0.6711,
+      "step": 22521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.289706231330834e-06,
+      "loss": 0.6325,
+      "step": 22522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2890128884117425e-06,
+      "loss": 0.5834,
+      "step": 22523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2883196369161574e-06,
+      "loss": 0.6378,
+      "step": 22524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2876264768522983e-06,
+      "loss": 0.6052,
+      "step": 22525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2869334082283844e-06,
+      "loss": 0.6298,
+      "step": 22526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.286240431052631e-06,
+      "loss": 0.6659,
+      "step": 22527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2855475453332587e-06,
+      "loss": 0.6657,
+      "step": 22528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2848547510784757e-06,
+      "loss": 0.607,
+      "step": 22529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2841620482964997e-06,
+      "loss": 0.6376,
+      "step": 22530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2834694369955436e-06,
+      "loss": 0.6043,
+      "step": 22531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.282776917183819e-06,
+      "loss": 0.5894,
+      "step": 22532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.282084488869536e-06,
+      "loss": 0.6428,
+      "step": 22533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.281392152060906e-06,
+      "loss": 0.6959,
+      "step": 22534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.280699906766135e-06,
+      "loss": 0.6152,
+      "step": 22535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2800077529934315e-06,
+      "loss": 0.6421,
+      "step": 22536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2793156907510017e-06,
+      "loss": 0.6081,
+      "step": 22537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2786237200470507e-06,
+      "loss": 0.6493,
+      "step": 22538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2779318408897833e-06,
+      "loss": 0.6287,
+      "step": 22539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.277240053287405e-06,
+      "loss": 0.6239,
+      "step": 22540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.276548357248113e-06,
+      "loss": 0.6134,
+      "step": 22541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.27585675278011e-06,
+      "loss": 0.6476,
+      "step": 22542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2751652398915967e-06,
+      "loss": 0.655,
+      "step": 22543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2744738185907713e-06,
+      "loss": 0.6126,
+      "step": 22544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2737824888858306e-06,
+      "loss": 0.6313,
+      "step": 22545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.273091250784977e-06,
+      "loss": 0.597,
+      "step": 22546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2724001042963963e-06,
+      "loss": 0.6802,
+      "step": 22547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.271709049428289e-06,
+      "loss": 0.705,
+      "step": 22548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2710180861888474e-06,
+      "loss": 0.6118,
+      "step": 22549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.270327214586263e-06,
+      "loss": 0.6575,
+      "step": 22550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.26963643462873e-06,
+      "loss": 0.6835,
+      "step": 22551
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2689457463244345e-06,
+      "loss": 0.6816,
+      "step": 22552
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2682551496815665e-06,
+      "loss": 0.6278,
+      "step": 22553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2675646447083167e-06,
+      "loss": 0.6936,
+      "step": 22554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2668742314128643e-06,
+      "loss": 0.584,
+      "step": 22555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.266183909803403e-06,
+      "loss": 0.6392,
+      "step": 22556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.265493679888119e-06,
+      "loss": 0.5689,
+      "step": 22557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.264803541675187e-06,
+      "loss": 0.6504,
+      "step": 22558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.264113495172795e-06,
+      "loss": 0.6666,
+      "step": 22559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2634235403891257e-06,
+      "loss": 0.6438,
+      "step": 22560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2627336773323516e-06,
+      "loss": 0.6613,
+      "step": 22561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2620439060106603e-06,
+      "loss": 0.6536,
+      "step": 22562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.261354226432231e-06,
+      "loss": 0.6182,
+      "step": 22563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.260664638605232e-06,
+      "loss": 0.6457,
+      "step": 22564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.259975142537846e-06,
+      "loss": 0.6452,
+      "step": 22565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2592857382382473e-06,
+      "loss": 0.6331,
+      "step": 22566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.258596425714603e-06,
+      "loss": 0.6607,
+      "step": 22567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.257907204975094e-06,
+      "loss": 0.7404,
+      "step": 22568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2572180760278916e-06,
+      "loss": 0.6321,
+      "step": 22569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.25652903888116e-06,
+      "loss": 0.6047,
+      "step": 22570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2558400935430736e-06,
+      "loss": 0.6838,
+      "step": 22571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.255151240021801e-06,
+      "loss": 0.6155,
+      "step": 22572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.254462478325503e-06,
+      "loss": 0.6528,
+      "step": 22573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2537738084623527e-06,
+      "loss": 0.6948,
+      "step": 22574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.253085230440516e-06,
+      "loss": 0.6611,
+      "step": 22575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2523967442681506e-06,
+      "loss": 0.6374,
+      "step": 22576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2517083499534233e-06,
+      "loss": 0.6336,
+      "step": 22577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2510200475044963e-06,
+      "loss": 0.6053,
+      "step": 22578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2503318369295278e-06,
+      "loss": 0.625,
+      "step": 22579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.249643718236676e-06,
+      "loss": 0.7078,
+      "step": 22580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2489556914341072e-06,
+      "loss": 0.614,
+      "step": 22581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2482677565299716e-06,
+      "loss": 0.6443,
+      "step": 22582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2475799135324273e-06,
+      "loss": 0.6397,
+      "step": 22583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.246892162449632e-06,
+      "loss": 0.6497,
+      "step": 22584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.246204503289736e-06,
+      "loss": 0.6804,
+      "step": 22585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.245516936060892e-06,
+      "loss": 0.6067,
+      "step": 22586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2448294607712585e-06,
+      "loss": 0.6211,
+      "step": 22587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.24414207742898e-06,
+      "loss": 0.6293,
+      "step": 22588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2434547860422084e-06,
+      "loss": 0.6256,
+      "step": 22589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2427675866190946e-06,
+      "loss": 0.6517,
+      "step": 22590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.242080479167782e-06,
+      "loss": 0.654,
+      "step": 22591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2413934636964165e-06,
+      "loss": 0.685,
+      "step": 22592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.240706540213151e-06,
+      "loss": 0.5988,
+      "step": 22593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.240019708726122e-06,
+      "loss": 0.6056,
+      "step": 22594
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.239332969243476e-06,
+      "loss": 0.6398,
+      "step": 22595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.238646321773359e-06,
+      "loss": 0.6426,
+      "step": 22596
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.237959766323904e-06,
+      "loss": 0.6782,
+      "step": 22597
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2372733029032523e-06,
+      "loss": 0.575,
+      "step": 22598
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.236586931519551e-06,
+      "loss": 0.5739,
+      "step": 22599
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.23590065218093e-06,
+      "loss": 0.6329,
+      "step": 22600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2352144648955287e-06,
+      "loss": 0.609,
+      "step": 22601
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2345283696714848e-06,
+      "loss": 0.5652,
+      "step": 22602
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.233842366516927e-06,
+      "loss": 0.611,
+      "step": 22603
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2331564554399898e-06,
+      "loss": 0.6182,
+      "step": 22604
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.232470636448815e-06,
+      "loss": 0.6632,
+      "step": 22605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.231784909551522e-06,
+      "loss": 0.6998,
+      "step": 22606
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2310992747562465e-06,
+      "loss": 0.6402,
+      "step": 22607
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2304137320711195e-06,
+      "loss": 0.6258,
+      "step": 22608
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2297282815042633e-06,
+      "loss": 0.6514,
+      "step": 22609
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2290429230638054e-06,
+      "loss": 0.6276,
+      "step": 22610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.228357656757879e-06,
+      "loss": 0.6052,
+      "step": 22611
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2276724825946017e-06,
+      "loss": 0.5977,
+      "step": 22612
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2269874005821014e-06,
+      "loss": 0.7034,
+      "step": 22613
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.226302410728496e-06,
+      "loss": 0.621,
+      "step": 22614
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.225617513041909e-06,
+      "loss": 0.5987,
+      "step": 22615
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.224932707530462e-06,
+      "loss": 0.6061,
+      "step": 22616
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.224247994202271e-06,
+      "loss": 0.616,
+      "step": 22617
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.223563373065458e-06,
+      "loss": 0.687,
+      "step": 22618
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.222878844128141e-06,
+      "loss": 0.5905,
+      "step": 22619
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2221944073984314e-06,
+      "loss": 0.6527,
+      "step": 22620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2215100628844444e-06,
+      "loss": 0.6454,
+      "step": 22621
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.220825810594297e-06,
+      "loss": 0.609,
+      "step": 22622
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2201416505360997e-06,
+      "loss": 0.6123,
+      "step": 22623
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2194575827179644e-06,
+      "loss": 0.6124,
+      "step": 22624
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.218773607148005e-06,
+      "loss": 0.6449,
+      "step": 22625
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2180897238343246e-06,
+      "loss": 0.6915,
+      "step": 22626
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.217405932785035e-06,
+      "loss": 0.6489,
+      "step": 22627
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.216722234008244e-06,
+      "loss": 0.6508,
+      "step": 22628
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.216038627512056e-06,
+      "loss": 0.6091,
+      "step": 22629
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.215355113304577e-06,
+      "loss": 0.6623,
+      "step": 22630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2146716913939127e-06,
+      "loss": 0.6099,
+      "step": 22631
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.213988361788162e-06,
+      "loss": 0.6752,
+      "step": 22632
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2133051244954283e-06,
+      "loss": 0.6299,
+      "step": 22633
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2126219795238124e-06,
+      "loss": 0.6383,
+      "step": 22634
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.211938926881415e-06,
+      "loss": 0.6089,
+      "step": 22635
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2112559665763323e-06,
+      "loss": 0.601,
+      "step": 22636
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.210573098616667e-06,
+      "loss": 0.5806,
+      "step": 22637
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2098903230105074e-06,
+      "loss": 0.6341,
+      "step": 22638
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2092076397659535e-06,
+      "loss": 0.6058,
+      "step": 22639
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.208525048891098e-06,
+      "loss": 0.6502,
+      "step": 22640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2078425503940336e-06,
+      "loss": 0.6532,
+      "step": 22641
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.207160144282854e-06,
+      "loss": 0.5841,
+      "step": 22642
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.206477830565651e-06,
+      "loss": 0.6247,
+      "step": 22643
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2057956092505093e-06,
+      "loss": 0.6345,
+      "step": 22644
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.205113480345521e-06,
+      "loss": 0.6477,
+      "step": 22645
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.204431443858773e-06,
+      "loss": 0.6788,
+      "step": 22646
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2037494997983513e-06,
+      "loss": 0.6226,
+      "step": 22647
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2030676481723412e-06,
+      "loss": 0.5995,
+      "step": 22648
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2023858889888304e-06,
+      "loss": 0.6179,
+      "step": 22649
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.201704222255897e-06,
+      "loss": 0.6146,
+      "step": 22650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2010226479816242e-06,
+      "loss": 0.6702,
+      "step": 22651
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2003411661740936e-06,
+      "loss": 0.6301,
+      "step": 22652
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.199659776841385e-06,
+      "loss": 0.6484,
+      "step": 22653
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.198978479991577e-06,
+      "loss": 0.6922,
+      "step": 22654
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.1982972756327516e-06,
+      "loss": 0.5962,
+      "step": 22655
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.1976161637729767e-06,
+      "loss": 0.6359,
+      "step": 22656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1969351444203334e-06,
+      "loss": 0.5991,
+      "step": 22657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1962542175828937e-06,
+      "loss": 0.6816,
+      "step": 22658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.195573383268733e-06,
+      "loss": 0.6069,
+      "step": 22659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1948926414859227e-06,
+      "loss": 0.6434,
+      "step": 22660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1942119922425364e-06,
+      "loss": 0.6488,
+      "step": 22661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1935314355466374e-06,
+      "loss": 0.6817,
+      "step": 22662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1928509714062996e-06,
+      "loss": 0.614,
+      "step": 22663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1921705998295893e-06,
+      "loss": 0.654,
+      "step": 22664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.191490320824573e-06,
+      "loss": 0.6447,
+      "step": 22665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.190810134399317e-06,
+      "loss": 0.634,
+      "step": 22666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1901300405618888e-06,
+      "loss": 0.6273,
+      "step": 22667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1894500393203454e-06,
+      "loss": 0.6286,
+      "step": 22668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1887701306827526e-06,
+      "loss": 0.5944,
+      "step": 22669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.188090314657171e-06,
+      "loss": 0.6329,
+      "step": 22670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1874105912516607e-06,
+      "loss": 0.6834,
+      "step": 22671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1867309604742803e-06,
+      "loss": 0.6821,
+      "step": 22672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1860514223330918e-06,
+      "loss": 0.6593,
+      "step": 22673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1853719768361446e-06,
+      "loss": 0.6111,
+      "step": 22674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.184692623991499e-06,
+      "loss": 0.5988,
+      "step": 22675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1840133638072093e-06,
+      "loss": 0.6024,
+      "step": 22676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1833341962913267e-06,
+      "loss": 0.6528,
+      "step": 22677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.18265512145191e-06,
+      "loss": 0.6638,
+      "step": 22678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.181976139297002e-06,
+      "loss": 0.6761,
+      "step": 22679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1812972498346564e-06,
+      "loss": 0.6201,
+      "step": 22680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1806184530729234e-06,
+      "loss": 0.6363,
+      "step": 22681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1799397490198504e-06,
+      "loss": 0.6245,
+      "step": 22682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.179261137683484e-06,
+      "loss": 0.6064,
+      "step": 22683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.178582619071872e-06,
+      "loss": 0.7355,
+      "step": 22684
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1779041931930543e-06,
+      "loss": 0.6671,
+      "step": 22685
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.177225860055079e-06,
+      "loss": 0.6663,
+      "step": 22686
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.176547619665985e-06,
+      "loss": 0.6081,
+      "step": 22687
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.175869472033817e-06,
+      "loss": 0.6592,
+      "step": 22688
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.175191417166613e-06,
+      "loss": 0.6247,
+      "step": 22689
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.174513455072418e-06,
+      "loss": 0.6646,
+      "step": 22690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.173835585759261e-06,
+      "loss": 0.6278,
+      "step": 22691
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1731578092351825e-06,
+      "loss": 0.6446,
+      "step": 22692
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1724801255082208e-06,
+      "loss": 0.6505,
+      "step": 22693
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.171802534586408e-06,
+      "loss": 0.6567,
+      "step": 22694
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.171125036477779e-06,
+      "loss": 0.6592,
+      "step": 22695
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.17044763119037e-06,
+      "loss": 0.6374,
+      "step": 22696
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.169770318732205e-06,
+      "loss": 0.5711,
+      "step": 22697
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.169093099111318e-06,
+      "loss": 0.5988,
+      "step": 22698
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.168415972335739e-06,
+      "loss": 0.6257,
+      "step": 22699
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.167738938413496e-06,
+      "loss": 0.639,
+      "step": 22700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1670619973526152e-06,
+      "loss": 0.5605,
+      "step": 22701
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1663851491611255e-06,
+      "loss": 0.641,
+      "step": 22702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.165708393847048e-06,
+      "loss": 0.5762,
+      "step": 22703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1650317314184068e-06,
+      "loss": 0.6266,
+      "step": 22704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.164355161883226e-06,
+      "loss": 0.6883,
+      "step": 22705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1636786852495272e-06,
+      "loss": 0.6257,
+      "step": 22706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.16300230152533e-06,
+      "loss": 0.6226,
+      "step": 22707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.162326010718657e-06,
+      "loss": 0.643,
+      "step": 22708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1616498128375217e-06,
+      "loss": 0.5938,
+      "step": 22709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.160973707889943e-06,
+      "loss": 0.6586,
+      "step": 22710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1602976958839383e-06,
+      "loss": 0.6114,
+      "step": 22711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1596217768275205e-06,
+      "loss": 0.5869,
+      "step": 22712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1589459507287056e-06,
+      "loss": 0.6454,
+      "step": 22713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1582702175955074e-06,
+      "loss": 0.6483,
+      "step": 22714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.157594577435934e-06,
+      "loss": 0.6478,
+      "step": 22715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1569190302579968e-06,
+      "loss": 0.6165,
+      "step": 22716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.156243576069705e-06,
+      "loss": 0.6062,
+      "step": 22717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.15556821487907e-06,
+      "loss": 0.593,
+      "step": 22718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.154892946694096e-06,
+      "loss": 0.622,
+      "step": 22719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.154217771522793e-06,
+      "loss": 0.6471,
+      "step": 22720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1535426893731603e-06,
+      "loss": 0.6191,
+      "step": 22721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.152867700253206e-06,
+      "loss": 0.654,
+      "step": 22722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1521928041709305e-06,
+      "loss": 0.6737,
+      "step": 22723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1515180011343373e-06,
+      "loss": 0.632,
+      "step": 22724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1508432911514264e-06,
+      "loss": 0.6248,
+      "step": 22725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.150168674230201e-06,
+      "loss": 0.6328,
+      "step": 22726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1494941503786526e-06,
+      "loss": 0.6435,
+      "step": 22727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.148819719604782e-06,
+      "loss": 0.6423,
+      "step": 22728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1481453819165866e-06,
+      "loss": 0.6006,
+      "step": 22729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1474711373220593e-06,
+      "loss": 0.6238,
+      "step": 22730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1467969858291948e-06,
+      "loss": 0.6083,
+      "step": 22731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1461229274459895e-06,
+      "loss": 0.6565,
+      "step": 22732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1454489621804296e-06,
+      "loss": 0.6784,
+      "step": 22733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1447750900405075e-06,
+      "loss": 0.6528,
+      "step": 22734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1441013110342148e-06,
+      "loss": 0.6336,
+      "step": 22735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1434276251695386e-06,
+      "loss": 0.6637,
+      "step": 22736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.142754032454466e-06,
+      "loss": 0.6224,
+      "step": 22737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.142080532896986e-06,
+      "loss": 0.6457,
+      "step": 22738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1414071265050796e-06,
+      "loss": 0.6574,
+      "step": 22739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1407338132867328e-06,
+      "loss": 0.6214,
+      "step": 22740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1400605932499295e-06,
+      "loss": 0.6189,
+      "step": 22741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1393874664026492e-06,
+      "loss": 0.5981,
+      "step": 22742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1387144327528787e-06,
+      "loss": 0.6293,
+      "step": 22743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.138041492308589e-06,
+      "loss": 0.6111,
+      "step": 22744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1373686450777633e-06,
+      "loss": 0.609,
+      "step": 22745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1366958910683787e-06,
+      "loss": 0.6053,
+      "step": 22746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.136023230288411e-06,
+      "loss": 0.64,
+      "step": 22747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1353506627458365e-06,
+      "loss": 0.6804,
+      "step": 22748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.134678188448631e-06,
+      "loss": 0.6335,
+      "step": 22749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.134005807404763e-06,
+      "loss": 0.627,
+      "step": 22750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.133333519622206e-06,
+      "loss": 0.6919,
+      "step": 22751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1326613251089324e-06,
+      "loss": 0.6489,
+      "step": 22752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.13198922387291e-06,
+      "loss": 0.651,
+      "step": 22753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1313172159221094e-06,
+      "loss": 0.5861,
+      "step": 22754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.130645301264499e-06,
+      "loss": 0.6392,
+      "step": 22755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1299734799080417e-06,
+      "loss": 0.6583,
+      "step": 22756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.129301751860704e-06,
+      "loss": 0.6504,
+      "step": 22757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1286301171304504e-06,
+      "loss": 0.6467,
+      "step": 22758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.127958575725243e-06,
+      "loss": 0.6023,
+      "step": 22759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.127287127653046e-06,
+      "loss": 0.6807,
+      "step": 22760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.126615772921822e-06,
+      "loss": 0.6462,
+      "step": 22761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.125944511539525e-06,
+      "loss": 0.6846,
+      "step": 22762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.125273343514116e-06,
+      "loss": 0.6341,
+      "step": 22763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.124602268853554e-06,
+      "loss": 0.5974,
+      "step": 22764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1239312875657937e-06,
+      "loss": 0.5923,
+      "step": 22765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.123260399658792e-06,
+      "loss": 0.6698,
+      "step": 22766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1225896051405047e-06,
+      "loss": 0.6523,
+      "step": 22767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1219189040188803e-06,
+      "loss": 0.5989,
+      "step": 22768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.121248296301872e-06,
+      "loss": 0.6668,
+      "step": 22769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.120577781997436e-06,
+      "loss": 0.6073,
+      "step": 22770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1199073611135115e-06,
+      "loss": 0.6436,
+      "step": 22771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.119237033658057e-06,
+      "loss": 0.6407,
+      "step": 22772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1185667996390193e-06,
+      "loss": 0.6221,
+      "step": 22773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.117896659064339e-06,
+      "loss": 0.672,
+      "step": 22774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.117226611941965e-06,
+      "loss": 0.6155,
+      "step": 22775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1165566582798435e-06,
+      "loss": 0.6678,
+      "step": 22776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.11588679808591e-06,
+      "loss": 0.5592,
+      "step": 22777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.115217031368115e-06,
+      "loss": 0.6652,
+      "step": 22778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1145473581344e-06,
+      "loss": 0.5601,
+      "step": 22779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.113877778392698e-06,
+      "loss": 0.5952,
+      "step": 22780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.11320829215095e-06,
+      "loss": 0.6233,
+      "step": 22781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.112538899417098e-06,
+      "loss": 0.6203,
+      "step": 22782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.11186960019907e-06,
+      "loss": 0.7006,
+      "step": 22783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.111200394504809e-06,
+      "loss": 0.6187,
+      "step": 22784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1105312823422485e-06,
+      "loss": 0.6642,
+      "step": 22785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1098622637193177e-06,
+      "loss": 0.6234,
+      "step": 22786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.109193338643951e-06,
+      "loss": 0.6908,
+      "step": 22787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1085245071240813e-06,
+      "loss": 0.6223,
+      "step": 22788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1078557691676318e-06,
+      "loss": 0.5778,
+      "step": 22789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1071871247825384e-06,
+      "loss": 0.6593,
+      "step": 22790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.106518573976728e-06,
+      "loss": 0.5788,
+      "step": 22791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1058501167581236e-06,
+      "loss": 0.6172,
+      "step": 22792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1051817531346507e-06,
+      "loss": 0.6374,
+      "step": 22793
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1045134831142387e-06,
+      "loss": 0.6571,
+      "step": 22794
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1038453067048025e-06,
+      "loss": 0.6301,
+      "step": 22795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.103177223914271e-06,
+      "loss": 0.6516,
+      "step": 22796
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.102509234750565e-06,
+      "loss": 0.6809,
+      "step": 22797
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.101841339221601e-06,
+      "loss": 0.6301,
+      "step": 22798
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1011735373352983e-06,
+      "loss": 0.6428,
+      "step": 22799
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1005058290995784e-06,
+      "loss": 0.6115,
+      "step": 22800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0998382145223497e-06,
+      "loss": 0.6684,
+      "step": 22801
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0991706936115375e-06,
+      "loss": 0.6439,
+      "step": 22802
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.098503266375048e-06,
+      "loss": 0.648,
+      "step": 22803
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0978359328207986e-06,
+      "loss": 0.6277,
+      "step": 22804
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0971686929567038e-06,
+      "loss": 0.6596,
+      "step": 22805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.096501546790668e-06,
+      "loss": 0.5724,
+      "step": 22806
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0958344943306007e-06,
+      "loss": 0.5898,
+      "step": 22807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0951675355844213e-06,
+      "loss": 0.5769,
+      "step": 22808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.094500670560027e-06,
+      "loss": 0.5949,
+      "step": 22809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.093833899265327e-06,
+      "loss": 0.6278,
+      "step": 22810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0931672217082323e-06,
+      "loss": 0.5837,
+      "step": 22811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0925006378966373e-06,
+      "loss": 0.6093,
+      "step": 22812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0918341478384486e-06,
+      "loss": 0.6416,
+      "step": 22813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0911677515415763e-06,
+      "loss": 0.6939,
+      "step": 22814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.090501449013911e-06,
+      "loss": 0.6366,
+      "step": 22815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.089835240263357e-06,
+      "loss": 0.6195,
+      "step": 22816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.089169125297815e-06,
+      "loss": 0.6745,
+      "step": 22817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.088503104125178e-06,
+      "loss": 0.6466,
+      "step": 22818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.087837176753342e-06,
+      "loss": 0.6801,
+      "step": 22819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0871713431902106e-06,
+      "loss": 0.6136,
+      "step": 22820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0865056034436703e-06,
+      "loss": 0.554,
+      "step": 22821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.085839957521616e-06,
+      "loss": 0.6507,
+      "step": 22822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0851744054319434e-06,
+      "loss": 0.5792,
+      "step": 22823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0845089471825386e-06,
+      "loss": 0.6236,
+      "step": 22824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.08384358278129e-06,
+      "loss": 0.636,
+      "step": 22825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.083178312236095e-06,
+      "loss": 0.6807,
+      "step": 22826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0825131355548335e-06,
+      "loss": 0.6064,
+      "step": 22827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0818480527453944e-06,
+      "loss": 0.6339,
+      "step": 22828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.081183063815666e-06,
+      "loss": 0.5802,
+      "step": 22829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0805181687735267e-06,
+      "loss": 0.6282,
+      "step": 22830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.079853367626861e-06,
+      "loss": 0.5788,
+      "step": 22831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0791886603835563e-06,
+      "loss": 0.6069,
+      "step": 22832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.078524047051488e-06,
+      "loss": 0.6697,
+      "step": 22833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.077859527638538e-06,
+      "loss": 0.6236,
+      "step": 22834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0771951021525883e-06,
+      "loss": 0.6027,
+      "step": 22835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0765307706015094e-06,
+      "loss": 0.616,
+      "step": 22836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0758665329931784e-06,
+      "loss": 0.7213,
+      "step": 22837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.07520238933548e-06,
+      "loss": 0.6299,
+      "step": 22838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0745383396362793e-06,
+      "loss": 0.6539,
+      "step": 22839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0738743839034526e-06,
+      "loss": 0.6327,
+      "step": 22840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.073210522144874e-06,
+      "loss": 0.62,
+      "step": 22841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0725467543684086e-06,
+      "loss": 0.6382,
+      "step": 22842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0718830805819278e-06,
+      "loss": 0.5574,
+      "step": 22843
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.071219500793308e-06,
+      "loss": 0.6574,
+      "step": 22844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0705560150104074e-06,
+      "loss": 0.5752,
+      "step": 22845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0698926232410964e-06,
+      "loss": 0.6178,
+      "step": 22846
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.069229325493244e-06,
+      "loss": 0.614,
+      "step": 22847
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0685661217747055e-06,
+      "loss": 0.6272,
+      "step": 22848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0679030120933473e-06,
+      "loss": 0.6337,
+      "step": 22849
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0672399964570388e-06,
+      "loss": 0.6111,
+      "step": 22850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0665770748736326e-06,
+      "loss": 0.6615,
+      "step": 22851
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.06591424735099e-06,
+      "loss": 0.6102,
+      "step": 22852
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.065251513896974e-06,
+      "loss": 0.665,
+      "step": 22853
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0645888745194355e-06,
+      "loss": 0.6311,
+      "step": 22854
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0639263292262323e-06,
+      "loss": 0.6041,
+      "step": 22855
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0632638780252266e-06,
+      "loss": 0.6014,
+      "step": 22856
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0626015209242654e-06,
+      "loss": 0.5928,
+      "step": 22857
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0619392579312036e-06,
+      "loss": 0.6853,
+      "step": 22858
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.061277089053897e-06,
+      "loss": 0.6526,
+      "step": 22859
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.060615014300189e-06,
+      "loss": 0.6347,
+      "step": 22860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0599530336779317e-06,
+      "loss": 0.6556,
+      "step": 22861
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0592911471949807e-06,
+      "loss": 0.5625,
+      "step": 22862
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.058629354859175e-06,
+      "loss": 0.6161,
+      "step": 22863
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.057967656678368e-06,
+      "loss": 0.7113,
+      "step": 22864
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.057306052660397e-06,
+      "loss": 0.5866,
+      "step": 22865
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.056644542813111e-06,
+      "loss": 0.5768,
+      "step": 22866
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0559831271443532e-06,
+      "loss": 0.6832,
+      "step": 22867
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.055321805661963e-06,
+      "loss": 0.6477,
+      "step": 22868
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0546605783737837e-06,
+      "loss": 0.6148,
+      "step": 22869
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0539994452876568e-06,
+      "loss": 0.6549,
+      "step": 22870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.053338406411416e-06,
+      "loss": 0.6441,
+      "step": 22871
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0526774617528998e-06,
+      "loss": 0.6396,
+      "step": 22872
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.052016611319947e-06,
+      "loss": 0.6636,
+      "step": 22873
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0513558551203905e-06,
+      "loss": 0.6486,
+      "step": 22874
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0506951931620666e-06,
+      "loss": 0.6178,
+      "step": 22875
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.050034625452808e-06,
+      "loss": 0.6434,
+      "step": 22876
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.049374152000445e-06,
+      "loss": 0.5885,
+      "step": 22877
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.048713772812808e-06,
+      "loss": 0.6611,
+      "step": 22878
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0480534878977277e-06,
+      "loss": 0.6837,
+      "step": 22879
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0473932972630328e-06,
+      "loss": 0.5981,
+      "step": 22880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0467332009165498e-06,
+      "loss": 0.642,
+      "step": 22881
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0460731988661087e-06,
+      "loss": 0.6733,
+      "step": 22882
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.045413291119529e-06,
+      "loss": 0.6863,
+      "step": 22883
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0447534776846377e-06,
+      "loss": 0.6017,
+      "step": 22884
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0440937585692566e-06,
+      "loss": 0.6359,
+      "step": 22885
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.043434133781209e-06,
+      "loss": 0.6705,
+      "step": 22886
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0427746033283147e-06,
+      "loss": 0.6374,
+      "step": 22887
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.042115167218396e-06,
+      "loss": 0.6299,
+      "step": 22888
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0414558254592655e-06,
+      "loss": 0.6514,
+      "step": 22889
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.040796578058745e-06,
+      "loss": 0.6788,
+      "step": 22890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0401374250246477e-06,
+      "loss": 0.6198,
+      "step": 22891
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0394783663647923e-06,
+      "loss": 0.6472,
+      "step": 22892
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.03881940208699e-06,
+      "loss": 0.6104,
+      "step": 22893
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.038160532199057e-06,
+      "loss": 0.6691,
+      "step": 22894
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0375017567088005e-06,
+      "loss": 0.6363,
+      "step": 22895
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0368430756240344e-06,
+      "loss": 0.6514,
+      "step": 22896
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.036184488952566e-06,
+      "loss": 0.6585,
+      "step": 22897
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0355259967022044e-06,
+      "loss": 0.5807,
+      "step": 22898
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0348675988807578e-06,
+      "loss": 0.6538,
+      "step": 22899
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0342092954960347e-06,
+      "loss": 0.6434,
+      "step": 22900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0335510865558337e-06,
+      "loss": 0.6613,
+      "step": 22901
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0328929720679625e-06,
+      "loss": 0.5404,
+      "step": 22902
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.032234952040224e-06,
+      "loss": 0.6563,
+      "step": 22903
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.031577026480419e-06,
+      "loss": 0.6443,
+      "step": 22904
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.030919195396348e-06,
+      "loss": 0.6644,
+      "step": 22905
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0302614587958137e-06,
+      "loss": 0.6322,
+      "step": 22906
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0296038166866094e-06,
+      "loss": 0.5284,
+      "step": 22907
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0289462690765347e-06,
+      "loss": 0.5404,
+      "step": 22908
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0282888159733847e-06,
+      "loss": 0.6231,
+      "step": 22909
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0276314573849552e-06,
+      "loss": 0.6097,
+      "step": 22910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.02697419331904e-06,
+      "loss": 0.6317,
+      "step": 22911
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.026317023783435e-06,
+      "loss": 0.6165,
+      "step": 22912
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0256599487859254e-06,
+      "loss": 0.617,
+      "step": 22913
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0250029683343043e-06,
+      "loss": 0.6551,
+      "step": 22914
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0243460824363616e-06,
+      "loss": 0.6679,
+      "step": 22915
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.023689291099885e-06,
+      "loss": 0.6466,
+      "step": 22916
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0230325943326623e-06,
+      "loss": 0.6044,
+      "step": 22917
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0223759921424824e-06,
+      "loss": 0.6273,
+      "step": 22918
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0217194845371235e-06,
+      "loss": 0.6251,
+      "step": 22919
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0210630715243727e-06,
+      "loss": 0.7112,
+      "step": 22920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.020406753112012e-06,
+      "loss": 0.5967,
+      "step": 22921
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0197505293078246e-06,
+      "loss": 0.5936,
+      "step": 22922
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.019094400119589e-06,
+      "loss": 0.5799,
+      "step": 22923
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0184383655550875e-06,
+      "loss": 0.6414,
+      "step": 22924
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0177824256220945e-06,
+      "loss": 0.5849,
+      "step": 22925
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.017126580328388e-06,
+      "loss": 0.6021,
+      "step": 22926
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.016470829681744e-06,
+      "loss": 0.6035,
+      "step": 22927
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0158151736899377e-06,
+      "loss": 0.6252,
+      "step": 22928
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0151596123607452e-06,
+      "loss": 0.6447,
+      "step": 22929
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0145041457019333e-06,
+      "loss": 0.6611,
+      "step": 22930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0138487737212764e-06,
+      "loss": 0.6475,
+      "step": 22931
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0131934964265452e-06,
+      "loss": 0.6695,
+      "step": 22932
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0125383138255083e-06,
+      "loss": 0.6201,
+      "step": 22933
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0118832259259347e-06,
+      "loss": 0.6227,
+      "step": 22934
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0112282327355925e-06,
+      "loss": 0.5978,
+      "step": 22935
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0105733342622426e-06,
+      "loss": 0.6578,
+      "step": 22936
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0099185305136527e-06,
+      "loss": 0.5649,
+      "step": 22937
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0092638214975856e-06,
+      "loss": 0.6409,
+      "step": 22938
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0086092072218042e-06,
+      "loss": 0.6408,
+      "step": 22939
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0079546876940712e-06,
+      "loss": 0.5812,
+      "step": 22940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.007300262922147e-06,
+      "loss": 0.6281,
+      "step": 22941
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0066459329137865e-06,
+      "loss": 0.6408,
+      "step": 22942
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.0059916976767503e-06,
+      "loss": 0.6041,
+      "step": 22943
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.005337557218795e-06,
+      "loss": 0.6484,
+      "step": 22944
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.004683511547676e-06,
+      "loss": 0.692,
+      "step": 22945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.004029560671148e-06,
+      "loss": 0.579,
+      "step": 22946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.003375704596967e-06,
+      "loss": 0.6532,
+      "step": 22947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0027219433328805e-06,
+      "loss": 0.5985,
+      "step": 22948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.002068276886642e-06,
+      "loss": 0.6789,
+      "step": 22949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0014147052660017e-06,
+      "loss": 0.6106,
+      "step": 22950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0007612284787082e-06,
+      "loss": 0.6257,
+      "step": 22951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0001078465325084e-06,
+      "loss": 0.6508,
+      "step": 22952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9994545594351544e-06,
+      "loss": 0.657,
+      "step": 22953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.998801367194384e-06,
+      "loss": 0.5759,
+      "step": 22954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.998148269817944e-06,
+      "loss": 0.6212,
+      "step": 22955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9974952673135795e-06,
+      "loss": 0.6099,
+      "step": 22956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9968423596890318e-06,
+      "loss": 0.6274,
+      "step": 22957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.996189546952042e-06,
+      "loss": 0.6357,
+      "step": 22958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9955368291103526e-06,
+      "loss": 0.6293,
+      "step": 22959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9948842061716976e-06,
+      "loss": 0.6352,
+      "step": 22960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9942316781438163e-06,
+      "loss": 0.5897,
+      "step": 22961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.993579245034446e-06,
+      "loss": 0.622,
+      "step": 22962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9929269068513215e-06,
+      "loss": 0.6316,
+      "step": 22963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9922746636021782e-06,
+      "loss": 0.5966,
+      "step": 22964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.991622515294751e-06,
+      "loss": 0.6686,
+      "step": 22965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9909704619367663e-06,
+      "loss": 0.6557,
+      "step": 22966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9903185035359585e-06,
+      "loss": 0.6487,
+      "step": 22967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9896666401000573e-06,
+      "loss": 0.6762,
+      "step": 22968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.989014871636792e-06,
+      "loss": 0.6261,
+      "step": 22969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9883631981538885e-06,
+      "loss": 0.6208,
+      "step": 22970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9877116196590763e-06,
+      "loss": 0.6925,
+      "step": 22971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.987060136160076e-06,
+      "loss": 0.63,
+      "step": 22972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9864087476646153e-06,
+      "loss": 0.6451,
+      "step": 22973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9857574541804148e-06,
+      "loss": 0.5971,
+      "step": 22974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.985106255715199e-06,
+      "loss": 0.6423,
+      "step": 22975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9844551522766864e-06,
+      "loss": 0.6539,
+      "step": 22976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.983804143872601e-06,
+      "loss": 0.6056,
+      "step": 22977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.983153230510655e-06,
+      "loss": 0.6679,
+      "step": 22978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.982502412198569e-06,
+      "loss": 0.6521,
+      "step": 22979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9818516889440586e-06,
+      "loss": 0.6316,
+      "step": 22980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9812010607548404e-06,
+      "loss": 0.6287,
+      "step": 22981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.980550527638626e-06,
+      "loss": 0.6374,
+      "step": 22982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.979900089603134e-06,
+      "loss": 0.6124,
+      "step": 22983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.979249746656068e-06,
+      "loss": 0.6462,
+      "step": 22984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9785994988051426e-06,
+      "loss": 0.6215,
+      "step": 22985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9779493460580677e-06,
+      "loss": 0.6468,
+      "step": 22986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9772992884225518e-06,
+      "loss": 0.6209,
+      "step": 22987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9766493259063036e-06,
+      "loss": 0.6795,
+      "step": 22988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.975999458517025e-06,
+      "loss": 0.6128,
+      "step": 22989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9753496862624235e-06,
+      "loss": 0.5921,
+      "step": 22990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9747000091502046e-06,
+      "loss": 0.6292,
+      "step": 22991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9740504271880635e-06,
+      "loss": 0.6472,
+      "step": 22992
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9734009403837106e-06,
+      "loss": 0.5945,
+      "step": 22993
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.972751548744847e-06,
+      "loss": 0.6408,
+      "step": 22994
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9721022522791645e-06,
+      "loss": 0.597,
+      "step": 22995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9714530509943653e-06,
+      "loss": 0.662,
+      "step": 22996
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9708039448981486e-06,
+      "loss": 0.6767,
+      "step": 22997
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.970154933998203e-06,
+      "loss": 0.6383,
+      "step": 22998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.969506018302232e-06,
+      "loss": 0.6632,
+      "step": 22999
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.968857197817927e-06,
+      "loss": 0.6257,
+      "step": 23000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9682084725529773e-06,
+      "loss": 0.6101,
+      "step": 23001
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9675598425150767e-06,
+      "loss": 0.5842,
+      "step": 23002
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9669113077119183e-06,
+      "loss": 0.6199,
+      "step": 23003
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.966262868151182e-06,
+      "loss": 0.6009,
+      "step": 23004
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9656145238405656e-06,
+      "loss": 0.6829,
+      "step": 23005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.964966274787755e-06,
+      "loss": 0.6477,
+      "step": 23006
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9643181210004303e-06,
+      "loss": 0.6632,
+      "step": 23007
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9636700624862803e-06,
+      "loss": 0.6638,
+      "step": 23008
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9630220992529904e-06,
+      "loss": 0.6195,
+      "step": 23009
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.962374231308234e-06,
+      "loss": 0.6599,
+      "step": 23010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9617264586597017e-06,
+      "loss": 0.5401,
+      "step": 23011
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9610787813150744e-06,
+      "loss": 0.6137,
+      "step": 23012
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9604311992820235e-06,
+      "loss": 0.6278,
+      "step": 23013
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9597837125682317e-06,
+      "loss": 0.6463,
+      "step": 23014
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9591363211813772e-06,
+      "loss": 0.6843,
+      "step": 23015
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.958489025129128e-06,
+      "loss": 0.6089,
+      "step": 23016
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9578418244191668e-06,
+      "loss": 0.6933,
+      "step": 23017
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9571947190591666e-06,
+      "loss": 0.6547,
+      "step": 23018
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9565477090567952e-06,
+      "loss": 0.637,
+      "step": 23019
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.955900794419725e-06,
+      "loss": 0.6211,
+      "step": 23020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9552539751556298e-06,
+      "loss": 0.6249,
+      "step": 23021
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9546072512721704e-06,
+      "loss": 0.5906,
+      "step": 23022
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9539606227770214e-06,
+      "loss": 0.5996,
+      "step": 23023
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.95331408967785e-06,
+      "loss": 0.6245,
+      "step": 23024
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9526676519823173e-06,
+      "loss": 0.6416,
+      "step": 23025
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9520213096980888e-06,
+      "loss": 0.6888,
+      "step": 23026
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.951375062832832e-06,
+      "loss": 0.632,
+      "step": 23027
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9507289113941996e-06,
+      "loss": 0.6267,
+      "step": 23028
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.950082855389861e-06,
+      "loss": 0.6496,
+      "step": 23029
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9494368948274755e-06,
+      "loss": 0.6309,
+      "step": 23030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9487910297146974e-06,
+      "loss": 0.6718,
+      "step": 23031
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9481452600591867e-06,
+      "loss": 0.659,
+      "step": 23032
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.947499585868602e-06,
+      "loss": 0.6308,
+      "step": 23033
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9468540071505903e-06,
+      "loss": 0.6798,
+      "step": 23034
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.946208523912816e-06,
+      "loss": 0.6494,
+      "step": 23035
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.945563136162929e-06,
+      "loss": 0.6362,
+      "step": 23036
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.944917843908577e-06,
+      "loss": 0.6237,
+      "step": 23037
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.944272647157415e-06,
+      "loss": 0.6685,
+      "step": 23038
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.943627545917093e-06,
+      "loss": 0.652,
+      "step": 23039
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.942982540195253e-06,
+      "loss": 0.6012,
+      "step": 23040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.942337629999549e-06,
+      "loss": 0.6298,
+      "step": 23041
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.941692815337628e-06,
+      "loss": 0.596,
+      "step": 23042
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9410480962171307e-06,
+      "loss": 0.6145,
+      "step": 23043
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9404034726457034e-06,
+      "loss": 0.6094,
+      "step": 23044
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9397589446309893e-06,
+      "loss": 0.6144,
+      "step": 23045
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9391145121806254e-06,
+      "loss": 0.6825,
+      "step": 23046
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9384701753022574e-06,
+      "loss": 0.6396,
+      "step": 23047
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9378259340035268e-06,
+      "loss": 0.6195,
+      "step": 23048
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.937181788292066e-06,
+      "loss": 0.6067,
+      "step": 23049
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.936537738175517e-06,
+      "loss": 0.6387,
+      "step": 23050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.93589378366151e-06,
+      "loss": 0.5898,
+      "step": 23051
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9352499247576815e-06,
+      "loss": 0.5928,
+      "step": 23052
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9346061614716727e-06,
+      "loss": 0.6172,
+      "step": 23053
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.933962493811108e-06,
+      "loss": 0.5968,
+      "step": 23054
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.93331892178362e-06,
+      "loss": 0.657,
+      "step": 23055
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9326754453968454e-06,
+      "loss": 0.6566,
+      "step": 23056
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9320320646584044e-06,
+      "loss": 0.6622,
+      "step": 23057
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.931388779575927e-06,
+      "loss": 0.6234,
+      "step": 23058
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9307455901570484e-06,
+      "loss": 0.57,
+      "step": 23059
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.930102496409385e-06,
+      "loss": 0.659,
+      "step": 23060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.929459498340566e-06,
+      "loss": 0.6692,
+      "step": 23061
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.928816595958215e-06,
+      "loss": 0.6248,
+      "step": 23062
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9281737892699525e-06,
+      "loss": 0.6219,
+      "step": 23063
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9275310782833965e-06,
+      "loss": 0.6378,
+      "step": 23064
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9268884630061777e-06,
+      "loss": 0.6439,
+      "step": 23065
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.926245943445906e-06,
+      "loss": 0.6284,
+      "step": 23066
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.925603519610201e-06,
+      "loss": 0.6143,
+      "step": 23067
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9249611915066834e-06,
+      "loss": 0.6273,
+      "step": 23068
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9243189591429633e-06,
+      "loss": 0.6167,
+      "step": 23069
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9236768225266557e-06,
+      "loss": 0.6275,
+      "step": 23070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9230347816653803e-06,
+      "loss": 0.6317,
+      "step": 23071
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9223928365667432e-06,
+      "loss": 0.6559,
+      "step": 23072
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9217509872383565e-06,
+      "loss": 0.6373,
+      "step": 23073
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9211092336878344e-06,
+      "loss": 0.6271,
+      "step": 23074
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9204675759227787e-06,
+      "loss": 0.6166,
+      "step": 23075
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9198260139507983e-06,
+      "loss": 0.6391,
+      "step": 23076
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.919184547779508e-06,
+      "loss": 0.6456,
+      "step": 23077
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.918543177416503e-06,
+      "loss": 0.596,
+      "step": 23078
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9179019028693936e-06,
+      "loss": 0.5997,
+      "step": 23079
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.917260724145782e-06,
+      "loss": 0.702,
+      "step": 23080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9166196412532667e-06,
+      "loss": 0.6798,
+      "step": 23081
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.91597865419945e-06,
+      "loss": 0.6183,
+      "step": 23082
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9153377629919367e-06,
+      "loss": 0.5945,
+      "step": 23083
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9146969676383187e-06,
+      "loss": 0.6465,
+      "step": 23084
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.914056268146196e-06,
+      "loss": 0.6401,
+      "step": 23085
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9134156645231685e-06,
+      "loss": 0.6037,
+      "step": 23086
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.912775156776824e-06,
+      "loss": 0.6622,
+      "step": 23087
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.912134744914759e-06,
+      "loss": 0.6172,
+      "step": 23088
+    },
+    {
+      "epoch": 0.8,
+      "eval_loss": 0.5751644968986511,
+      "eval_runtime": 4921.5874,
+      "eval_samples_per_second": 112.086,
+      "eval_steps_per_second": 37.362,
+      "step": 23088
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9114944289445725e-06,
+      "loss": 0.575,
+      "step": 23089
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9108542088738493e-06,
+      "loss": 0.5839,
+      "step": 23090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9102140847101814e-06,
+      "loss": 0.7018,
+      "step": 23091
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.909574056461162e-06,
+      "loss": 0.5891,
+      "step": 23092
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.908934124134375e-06,
+      "loss": 0.638,
+      "step": 23093
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9082942877374054e-06,
+      "loss": 0.6007,
+      "step": 23094
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9076545472778496e-06,
+      "loss": 0.5981,
+      "step": 23095
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9070149027632823e-06,
+      "loss": 0.6246,
+      "step": 23096
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9063753542012909e-06,
+      "loss": 0.6149,
+      "step": 23097
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9057359015994614e-06,
+      "loss": 0.6118,
+      "step": 23098
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.905096544965369e-06,
+      "loss": 0.6525,
+      "step": 23099
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9044572843065956e-06,
+      "loss": 0.6739,
+      "step": 23100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.903818119630726e-06,
+      "loss": 0.6662,
+      "step": 23101
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9031790509453329e-06,
+      "loss": 0.592,
+      "step": 23102
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9025400782579939e-06,
+      "loss": 0.6266,
+      "step": 23103
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9019012015762884e-06,
+      "loss": 0.6286,
+      "step": 23104
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9012624209077857e-06,
+      "loss": 0.6158,
+      "step": 23105
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9006237362600588e-06,
+      "loss": 0.5835,
+      "step": 23106
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8999851476406883e-06,
+      "loss": 0.6896,
+      "step": 23107
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8993466550572392e-06,
+      "loss": 0.6373,
+      "step": 23108
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8987082585172822e-06,
+      "loss": 0.6248,
+      "step": 23109
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.898069958028389e-06,
+      "loss": 0.6183,
+      "step": 23110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8974317535981236e-06,
+      "loss": 0.654,
+      "step": 23111
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8967936452340542e-06,
+      "loss": 0.6279,
+      "step": 23112
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.896155632943747e-06,
+      "loss": 0.6721,
+      "step": 23113
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8955177167347661e-06,
+      "loss": 0.6495,
+      "step": 23114
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8948798966146776e-06,
+      "loss": 0.6775,
+      "step": 23115
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8942421725910377e-06,
+      "loss": 0.6432,
+      "step": 23116
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8936045446714114e-06,
+      "loss": 0.631,
+      "step": 23117
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8929670128633581e-06,
+      "loss": 0.6079,
+      "step": 23118
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8923295771744355e-06,
+      "loss": 0.6129,
+      "step": 23119
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8916922376122027e-06,
+      "loss": 0.5882,
+      "step": 23120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8910549941842172e-06,
+      "loss": 0.6229,
+      "step": 23121
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8904178468980306e-06,
+      "loss": 0.6578,
+      "step": 23122
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.889780795761199e-06,
+      "loss": 0.6707,
+      "step": 23123
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8891438407812757e-06,
+      "loss": 0.6184,
+      "step": 23124
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.888506981965813e-06,
+      "loss": 0.6025,
+      "step": 23125
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8878702193223609e-06,
+      "loss": 0.6379,
+      "step": 23126
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8872335528584707e-06,
+      "loss": 0.6186,
+      "step": 23127
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8865969825816877e-06,
+      "loss": 0.6343,
+      "step": 23128
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8859605084995613e-06,
+      "loss": 0.6657,
+      "step": 23129
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8853241306196368e-06,
+      "loss": 0.6467,
+      "step": 23130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8846878489494603e-06,
+      "loss": 0.6475,
+      "step": 23131
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8840516634965743e-06,
+      "loss": 0.631,
+      "step": 23132
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8834155742685255e-06,
+      "loss": 0.6341,
+      "step": 23133
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.882779581272851e-06,
+      "loss": 0.6435,
+      "step": 23134
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8821436845170927e-06,
+      "loss": 0.6031,
+      "step": 23135
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8815078840087885e-06,
+      "loss": 0.6301,
+      "step": 23136
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8808721797554797e-06,
+      "loss": 0.645,
+      "step": 23137
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.880236571764702e-06,
+      "loss": 0.625,
+      "step": 23138
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.879601060043994e-06,
+      "loss": 0.6367,
+      "step": 23139
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8789656446008841e-06,
+      "loss": 0.6231,
+      "step": 23140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8783303254429109e-06,
+      "loss": 0.6034,
+      "step": 23141
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8776951025776047e-06,
+      "loss": 0.5712,
+      "step": 23142
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8770599760124985e-06,
+      "loss": 0.6155,
+      "step": 23143
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.876424945755122e-06,
+      "loss": 0.614,
+      "step": 23144
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8757900118130057e-06,
+      "loss": 0.6359,
+      "step": 23145
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8751551741936746e-06,
+      "loss": 0.6314,
+      "step": 23146
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8745204329046562e-06,
+      "loss": 0.5968,
+      "step": 23147
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8738857879534766e-06,
+      "loss": 0.6355,
+      "step": 23148
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.873251239347661e-06,
+      "loss": 0.6153,
+      "step": 23149
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.872616787094733e-06,
+      "loss": 0.6542,
+      "step": 23150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8719824312022161e-06,
+      "loss": 0.6351,
+      "step": 23151
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.871348171677626e-06,
+      "loss": 0.6767,
+      "step": 23152
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8707140085284869e-06,
+      "loss": 0.6477,
+      "step": 23153
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8700799417623172e-06,
+      "loss": 0.6509,
+      "step": 23154
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8694459713866342e-06,
+      "loss": 0.597,
+      "step": 23155
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.868812097408954e-06,
+      "loss": 0.6483,
+      "step": 23156
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8681783198367953e-06,
+      "loss": 0.6453,
+      "step": 23157
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8675446386776664e-06,
+      "loss": 0.6227,
+      "step": 23158
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8669110539390832e-06,
+      "loss": 0.5857,
+      "step": 23159
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8662775656285593e-06,
+      "loss": 0.6729,
+      "step": 23160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8656441737536025e-06,
+      "loss": 0.6427,
+      "step": 23161
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8650108783217247e-06,
+      "loss": 0.6564,
+      "step": 23162
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8643776793404366e-06,
+      "loss": 0.6156,
+      "step": 23163
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8637445768172402e-06,
+      "loss": 0.6143,
+      "step": 23164
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8631115707596438e-06,
+      "loss": 0.5633,
+      "step": 23165
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8624786611751522e-06,
+      "loss": 0.6073,
+      "step": 23166
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8618458480712708e-06,
+      "loss": 0.6178,
+      "step": 23167
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8612131314555027e-06,
+      "loss": 0.6522,
+      "step": 23168
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8605805113353502e-06,
+      "loss": 0.6333,
+      "step": 23169
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.859947987718309e-06,
+      "loss": 0.6455,
+      "step": 23170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8593155606118829e-06,
+      "loss": 0.6366,
+      "step": 23171
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.858683230023568e-06,
+      "loss": 0.6079,
+      "step": 23172
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8580509959608617e-06,
+      "loss": 0.6222,
+      "step": 23173
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8574188584312647e-06,
+      "loss": 0.5973,
+      "step": 23174
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.856786817442263e-06,
+      "loss": 0.6574,
+      "step": 23175
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8561548730013567e-06,
+      "loss": 0.6439,
+      "step": 23176
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.855523025116035e-06,
+      "loss": 0.6629,
+      "step": 23177
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8548912737937919e-06,
+      "loss": 0.6485,
+      "step": 23178
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8542596190421146e-06,
+      "loss": 0.6637,
+      "step": 23179
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8536280608684975e-06,
+      "loss": 0.6523,
+      "step": 23180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8529965992804233e-06,
+      "loss": 0.6262,
+      "step": 23181
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8523652342853793e-06,
+      "loss": 0.6224,
+      "step": 23182
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8517339658908528e-06,
+      "loss": 0.6833,
+      "step": 23183
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8511027941043268e-06,
+      "loss": 0.6441,
+      "step": 23184
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8504717189332865e-06,
+      "loss": 0.6645,
+      "step": 23185
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8498407403852158e-06,
+      "loss": 0.5672,
+      "step": 23186
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.84920985846759e-06,
+      "loss": 0.5998,
+      "step": 23187
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8485790731878928e-06,
+      "loss": 0.677,
+      "step": 23188
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8479483845536018e-06,
+      "loss": 0.6462,
+      "step": 23189
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8473177925721951e-06,
+      "loss": 0.6301,
+      "step": 23190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8466872972511485e-06,
+      "loss": 0.5851,
+      "step": 23191
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.846056898597942e-06,
+      "loss": 0.6436,
+      "step": 23192
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8454265966200424e-06,
+      "loss": 0.6477,
+      "step": 23193
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8447963913249257e-06,
+      "loss": 0.6144,
+      "step": 23194
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8441662827200645e-06,
+      "loss": 0.6164,
+      "step": 23195
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.84353627081293e-06,
+      "loss": 0.6376,
+      "step": 23196
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8429063556109895e-06,
+      "loss": 0.6677,
+      "step": 23197
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8422765371217166e-06,
+      "loss": 0.706,
+      "step": 23198
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8416468153525723e-06,
+      "loss": 0.6848,
+      "step": 23199
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8410171903110251e-06,
+      "loss": 0.6355,
+      "step": 23200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8403876620045402e-06,
+      "loss": 0.6429,
+      "step": 23201
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8397582304405815e-06,
+      "loss": 0.638,
+      "step": 23202
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8391288956266119e-06,
+      "loss": 0.6376,
+      "step": 23203
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8384996575700953e-06,
+      "loss": 0.6165,
+      "step": 23204
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.837870516278487e-06,
+      "loss": 0.6612,
+      "step": 23205
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8372414717592491e-06,
+      "loss": 0.6559,
+      "step": 23206
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8366125240198397e-06,
+      "loss": 0.6706,
+      "step": 23207
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.835983673067715e-06,
+      "loss": 0.6166,
+      "step": 23208
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8353549189103315e-06,
+      "loss": 0.5958,
+      "step": 23209
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8347262615551465e-06,
+      "loss": 0.616,
+      "step": 23210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8340977010096083e-06,
+      "loss": 0.6212,
+      "step": 23211
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8334692372811713e-06,
+      "loss": 0.5849,
+      "step": 23212
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8328408703772882e-06,
+      "loss": 0.6788,
+      "step": 23213
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8322126003054074e-06,
+      "loss": 0.593,
+      "step": 23214
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8315844270729788e-06,
+      "loss": 0.6174,
+      "step": 23215
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8309563506874517e-06,
+      "loss": 0.6416,
+      "step": 23216
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.830328371156269e-06,
+      "loss": 0.6105,
+      "step": 23217
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8297004884868785e-06,
+      "loss": 0.6329,
+      "step": 23218
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.829072702686724e-06,
+      "loss": 0.6071,
+      "step": 23219
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8284450137632481e-06,
+      "loss": 0.6456,
+      "step": 23220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8278174217238941e-06,
+      "loss": 0.6465,
+      "step": 23221
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8271899265761052e-06,
+      "loss": 0.612,
+      "step": 23222
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8265625283273158e-06,
+      "loss": 0.6031,
+      "step": 23223
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8259352269849672e-06,
+      "loss": 0.6321,
+      "step": 23224
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8253080225564968e-06,
+      "loss": 0.6407,
+      "step": 23225
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8246809150493405e-06,
+      "loss": 0.6403,
+      "step": 23226
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8240539044709337e-06,
+      "loss": 0.6008,
+      "step": 23227
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8234269908287127e-06,
+      "loss": 0.6471,
+      "step": 23228
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.822800174130106e-06,
+      "loss": 0.6593,
+      "step": 23229
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8221734543825476e-06,
+      "loss": 0.6173,
+      "step": 23230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8215468315934681e-06,
+      "loss": 0.6781,
+      "step": 23231
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8209203057702963e-06,
+      "loss": 0.5725,
+      "step": 23232
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8202938769204614e-06,
+      "loss": 0.6897,
+      "step": 23233
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8196675450513923e-06,
+      "loss": 0.6246,
+      "step": 23234
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8190413101705096e-06,
+      "loss": 0.6588,
+      "step": 23235
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8184151722852439e-06,
+      "loss": 0.5842,
+      "step": 23236
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8177891314030116e-06,
+      "loss": 0.6932,
+      "step": 23237
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8171631875312412e-06,
+      "loss": 0.6301,
+      "step": 23238
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8165373406773558e-06,
+      "loss": 0.6034,
+      "step": 23239
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8159115908487702e-06,
+      "loss": 0.6684,
+      "step": 23240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8152859380529042e-06,
+      "loss": 0.6402,
+      "step": 23241
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.814660382297181e-06,
+      "loss": 0.6522,
+      "step": 23242
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8140349235890075e-06,
+      "loss": 0.5583,
+      "step": 23243
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8134095619358083e-06,
+      "loss": 0.6211,
+      "step": 23244
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.812784297344997e-06,
+      "loss": 0.6539,
+      "step": 23245
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8121591298239827e-06,
+      "loss": 0.5549,
+      "step": 23246
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.81153405938018e-06,
+      "loss": 0.6433,
+      "step": 23247
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.810909086021001e-06,
+      "loss": 0.6248,
+      "step": 23248
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8102842097538498e-06,
+      "loss": 0.6448,
+      "step": 23249
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8096594305861415e-06,
+      "loss": 0.6417,
+      "step": 23250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8090347485252835e-06,
+      "loss": 0.6516,
+      "step": 23251
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8084101635786788e-06,
+      "loss": 0.6135,
+      "step": 23252
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8077856757537326e-06,
+      "loss": 0.5986,
+      "step": 23253
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8071612850578546e-06,
+      "loss": 0.6693,
+      "step": 23254
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8065369914984376e-06,
+      "loss": 0.5671,
+      "step": 23255
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8059127950828914e-06,
+      "loss": 0.6221,
+      "step": 23256
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8052886958186178e-06,
+      "loss": 0.6638,
+      "step": 23257
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8046646937130097e-06,
+      "loss": 0.6228,
+      "step": 23258
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8040407887734702e-06,
+      "loss": 0.5946,
+      "step": 23259
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8034169810073965e-06,
+      "loss": 0.5667,
+      "step": 23260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8027932704221784e-06,
+      "loss": 0.6365,
+      "step": 23261
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.802169657025219e-06,
+      "loss": 0.5909,
+      "step": 23262
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.80154614082391e-06,
+      "loss": 0.6387,
+      "step": 23263
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8009227218256408e-06,
+      "loss": 0.6461,
+      "step": 23264
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8002994000378038e-06,
+      "loss": 0.6372,
+      "step": 23265
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7996761754677938e-06,
+      "loss": 0.6627,
+      "step": 23266
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7990530481229896e-06,
+      "loss": 0.6538,
+      "step": 23267
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7984300180107894e-06,
+      "loss": 0.6603,
+      "step": 23268
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7978070851385788e-06,
+      "loss": 0.6365,
+      "step": 23269
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7971842495137383e-06,
+      "loss": 0.6192,
+      "step": 23270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7965615111436552e-06,
+      "loss": 0.661,
+      "step": 23271
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.795938870035715e-06,
+      "loss": 0.6271,
+      "step": 23272
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7953163261972938e-06,
+      "loss": 0.6516,
+      "step": 23273
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7946938796357781e-06,
+      "loss": 0.6431,
+      "step": 23274
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7940715303585488e-06,
+      "loss": 0.648,
+      "step": 23275
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7934492783729796e-06,
+      "loss": 0.6028,
+      "step": 23276
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7928271236864504e-06,
+      "loss": 0.6001,
+      "step": 23277
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7922050663063406e-06,
+      "loss": 0.6347,
+      "step": 23278
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7915831062400168e-06,
+      "loss": 0.6769,
+      "step": 23279
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7909612434948631e-06,
+      "loss": 0.6115,
+      "step": 23280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.79033947807825e-06,
+      "loss": 0.7098,
+      "step": 23281
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7897178099975443e-06,
+      "loss": 0.642,
+      "step": 23282
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7890962392601208e-06,
+      "loss": 0.6704,
+      "step": 23283
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7884747658733515e-06,
+      "loss": 0.5936,
+      "step": 23284
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7878533898445948e-06,
+      "loss": 0.6536,
+      "step": 23285
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7872321111812286e-06,
+      "loss": 0.6088,
+      "step": 23286
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7866109298906165e-06,
+      "loss": 0.6589,
+      "step": 23287
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7859898459801194e-06,
+      "loss": 0.6035,
+      "step": 23288
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.785368859457104e-06,
+      "loss": 0.5768,
+      "step": 23289
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.784747970328935e-06,
+      "loss": 0.602,
+      "step": 23290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7841271786029656e-06,
+      "loss": 0.6629,
+      "step": 23291
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7835064842865646e-06,
+      "loss": 0.6659,
+      "step": 23292
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7828858873870903e-06,
+      "loss": 0.6292,
+      "step": 23293
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7822653879118955e-06,
+      "loss": 0.6255,
+      "step": 23294
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7816449858683405e-06,
+      "loss": 0.6441,
+      "step": 23295
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7810246812637832e-06,
+      "loss": 0.6109,
+      "step": 23296
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7804044741055692e-06,
+      "loss": 0.6249,
+      "step": 23297
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.779784364401065e-06,
+      "loss": 0.618,
+      "step": 23298
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7791643521576118e-06,
+      "loss": 0.5846,
+      "step": 23299
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7785444373825645e-06,
+      "loss": 0.6613,
+      "step": 23300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7779246200832768e-06,
+      "loss": 0.5901,
+      "step": 23301
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7773049002670906e-06,
+      "loss": 0.6255,
+      "step": 23302
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7766852779413534e-06,
+      "loss": 0.5799,
+      "step": 23303
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7760657531134207e-06,
+      "loss": 0.6231,
+      "step": 23304
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7754463257906284e-06,
+      "loss": 0.6397,
+      "step": 23305
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7748269959803256e-06,
+      "loss": 0.6445,
+      "step": 23306
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7742077636898547e-06,
+      "loss": 0.6417,
+      "step": 23307
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7735886289265546e-06,
+      "loss": 0.6084,
+      "step": 23308
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7729695916977675e-06,
+      "loss": 0.6625,
+      "step": 23309
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7723506520108337e-06,
+      "loss": 0.6416,
+      "step": 23310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7717318098730896e-06,
+      "loss": 0.6052,
+      "step": 23311
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7711130652918752e-06,
+      "loss": 0.6228,
+      "step": 23312
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7704944182745265e-06,
+      "loss": 0.6287,
+      "step": 23313
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7698758688283746e-06,
+      "loss": 0.6086,
+      "step": 23314
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7692574169607556e-06,
+      "loss": 0.6134,
+      "step": 23315
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7686390626790019e-06,
+      "loss": 0.6491,
+      "step": 23316
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7680208059904448e-06,
+      "loss": 0.663,
+      "step": 23317
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7674026469024152e-06,
+      "loss": 0.6044,
+      "step": 23318
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7667845854222432e-06,
+      "loss": 0.6073,
+      "step": 23319
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7661666215572527e-06,
+      "loss": 0.7259,
+      "step": 23320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.765548755314772e-06,
+      "loss": 0.6087,
+      "step": 23321
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.764930986702128e-06,
+      "loss": 0.6586,
+      "step": 23322
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7643133157266458e-06,
+      "loss": 0.6445,
+      "step": 23323
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7636957423956459e-06,
+      "loss": 0.6458,
+      "step": 23324
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7630782667164548e-06,
+      "loss": 0.5655,
+      "step": 23325
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7624608886963878e-06,
+      "loss": 0.5841,
+      "step": 23326
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7618436083427682e-06,
+      "loss": 0.6254,
+      "step": 23327
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.761226425662914e-06,
+      "loss": 0.6466,
+      "step": 23328
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7606093406641422e-06,
+      "loss": 0.5982,
+      "step": 23329
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7599923533537699e-06,
+      "loss": 0.6317,
+      "step": 23330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7593754637391136e-06,
+      "loss": 0.5835,
+      "step": 23331
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.758758671827484e-06,
+      "loss": 0.6668,
+      "step": 23332
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7581419776261954e-06,
+      "loss": 0.6319,
+      "step": 23333
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7575253811425596e-06,
+      "loss": 0.6019,
+      "step": 23334
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7569088823838864e-06,
+      "loss": 0.6784,
+      "step": 23335
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7562924813574866e-06,
+      "loss": 0.608,
+      "step": 23336
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.755676178070671e-06,
+      "loss": 0.6574,
+      "step": 23337
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7550599725307405e-06,
+      "loss": 0.5925,
+      "step": 23338
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7544438647450036e-06,
+      "loss": 0.6462,
+      "step": 23339
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7538278547207654e-06,
+      "loss": 0.5719,
+      "step": 23340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7532119424653293e-06,
+      "loss": 0.6222,
+      "step": 23341
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7525961279859982e-06,
+      "loss": 0.5861,
+      "step": 23342
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7519804112900752e-06,
+      "loss": 0.5617,
+      "step": 23343
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7513647923848565e-06,
+      "loss": 0.6583,
+      "step": 23344
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7507492712776419e-06,
+      "loss": 0.6448,
+      "step": 23345
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7501338479757301e-06,
+      "loss": 0.668,
+      "step": 23346
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7495185224864186e-06,
+      "loss": 0.6043,
+      "step": 23347
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7489032948170003e-06,
+      "loss": 0.6256,
+      "step": 23348
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7482881649747752e-06,
+      "loss": 0.6546,
+      "step": 23349
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7476731329670293e-06,
+      "loss": 0.6011,
+      "step": 23350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.747058198801057e-06,
+      "loss": 0.6351,
+      "step": 23351
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7464433624841504e-06,
+      "loss": 0.6344,
+      "step": 23352
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.745828624023599e-06,
+      "loss": 0.6312,
+      "step": 23353
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7452139834266901e-06,
+      "loss": 0.6453,
+      "step": 23354
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7445994407007149e-06,
+      "loss": 0.5574,
+      "step": 23355
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7439849958529531e-06,
+      "loss": 0.6322,
+      "step": 23356
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7433706488906943e-06,
+      "loss": 0.6044,
+      "step": 23357
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7427563998212204e-06,
+      "loss": 0.5791,
+      "step": 23358
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7421422486518147e-06,
+      "loss": 0.6384,
+      "step": 23359
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7415281953897612e-06,
+      "loss": 0.6664,
+      "step": 23360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7409142400423363e-06,
+      "loss": 0.6031,
+      "step": 23361
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7403003826168208e-06,
+      "loss": 0.6032,
+      "step": 23362
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7396866231204923e-06,
+      "loss": 0.6293,
+      "step": 23363
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7390729615606283e-06,
+      "loss": 0.6128,
+      "step": 23364
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7384593979445042e-06,
+      "loss": 0.6705,
+      "step": 23365
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7378459322793973e-06,
+      "loss": 0.6572,
+      "step": 23366
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7372325645725774e-06,
+      "loss": 0.6662,
+      "step": 23367
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7366192948313166e-06,
+      "loss": 0.6207,
+      "step": 23368
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7360061230628878e-06,
+      "loss": 0.5252,
+      "step": 23369
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7353930492745596e-06,
+      "loss": 0.6652,
+      "step": 23370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7347800734736031e-06,
+      "loss": 0.6275,
+      "step": 23371
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7341671956672857e-06,
+      "loss": 0.6706,
+      "step": 23372
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7335544158628703e-06,
+      "loss": 0.6545,
+      "step": 23373
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7329417340676247e-06,
+      "loss": 0.6509,
+      "step": 23374
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7323291502888118e-06,
+      "loss": 0.603,
+      "step": 23375
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.731716664533697e-06,
+      "loss": 0.6049,
+      "step": 23376
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7311042768095398e-06,
+      "loss": 0.6204,
+      "step": 23377
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7304919871236037e-06,
+      "loss": 0.5924,
+      "step": 23378
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7298797954831436e-06,
+      "loss": 0.5964,
+      "step": 23379
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7292677018954207e-06,
+      "loss": 0.605,
+      "step": 23380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7286557063676922e-06,
+      "loss": 0.656,
+      "step": 23381
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7280438089072126e-06,
+      "loss": 0.6411,
+      "step": 23382
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7274320095212382e-06,
+      "loss": 0.6964,
+      "step": 23383
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7268203082170244e-06,
+      "loss": 0.6209,
+      "step": 23384
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7262087050018194e-06,
+      "loss": 0.6025,
+      "step": 23385
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7255971998828768e-06,
+      "loss": 0.6098,
+      "step": 23386
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7249857928674464e-06,
+      "loss": 0.6386,
+      "step": 23387
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7243744839627762e-06,
+      "loss": 0.6204,
+      "step": 23388
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7237632731761168e-06,
+      "loss": 0.647,
+      "step": 23389
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.723152160514715e-06,
+      "loss": 0.6898,
+      "step": 23390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7225411459858122e-06,
+      "loss": 0.6644,
+      "step": 23391
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7219302295966556e-06,
+      "loss": 0.6495,
+      "step": 23392
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7213194113544873e-06,
+      "loss": 0.6431,
+      "step": 23393
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7207086912665504e-06,
+      "loss": 0.684,
+      "step": 23394
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.720098069340085e-06,
+      "loss": 0.6598,
+      "step": 23395
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7194875455823335e-06,
+      "loss": 0.6487,
+      "step": 23396
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7188771200005304e-06,
+      "loss": 0.6655,
+      "step": 23397
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7182667926019136e-06,
+      "loss": 0.694,
+      "step": 23398
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7176565633937214e-06,
+      "loss": 0.6222,
+      "step": 23399
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.717046432383187e-06,
+      "loss": 0.5871,
+      "step": 23400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7164363995775458e-06,
+      "loss": 0.6459,
+      "step": 23401
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.715826464984033e-06,
+      "loss": 0.6213,
+      "step": 23402
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7152166286098727e-06,
+      "loss": 0.6694,
+      "step": 23403
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7146068904623003e-06,
+      "loss": 0.6248,
+      "step": 23404
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7139972505485447e-06,
+      "loss": 0.6972,
+      "step": 23405
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7133877088758332e-06,
+      "loss": 0.6865,
+      "step": 23406
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7127782654513936e-06,
+      "loss": 0.6275,
+      "step": 23407
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.712168920282453e-06,
+      "loss": 0.6445,
+      "step": 23408
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.711559673376232e-06,
+      "loss": 0.6529,
+      "step": 23409
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.710950524739955e-06,
+      "loss": 0.6333,
+      "step": 23410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7103414743808455e-06,
+      "loss": 0.6718,
+      "step": 23411
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7097325223061257e-06,
+      "loss": 0.5986,
+      "step": 23412
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7091236685230127e-06,
+      "loss": 0.6067,
+      "step": 23413
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7085149130387291e-06,
+      "loss": 0.5752,
+      "step": 23414
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7079062558604876e-06,
+      "loss": 0.6023,
+      "step": 23415
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.707297696995508e-06,
+      "loss": 0.6259,
+      "step": 23416
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7066892364510036e-06,
+      "loss": 0.6115,
+      "step": 23417
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7060808742341895e-06,
+      "loss": 0.5955,
+      "step": 23418
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7054726103522778e-06,
+      "loss": 0.6048,
+      "step": 23419
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7048644448124852e-06,
+      "loss": 0.6321,
+      "step": 23420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7042563776220133e-06,
+      "loss": 0.6432,
+      "step": 23421
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.703648408788078e-06,
+      "loss": 0.6495,
+      "step": 23422
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.703040538317885e-06,
+      "loss": 0.6063,
+      "step": 23423
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7024327662186423e-06,
+      "loss": 0.5986,
+      "step": 23424
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7018250924975577e-06,
+      "loss": 0.6122,
+      "step": 23425
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7012175171618329e-06,
+      "loss": 0.6204,
+      "step": 23426
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.700610040218671e-06,
+      "loss": 0.5719,
+      "step": 23427
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7000026616752763e-06,
+      "loss": 0.587,
+      "step": 23428
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6993953815388485e-06,
+      "loss": 0.6478,
+      "step": 23429
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.69878819981659e-06,
+      "loss": 0.6032,
+      "step": 23430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6981811165157014e-06,
+      "loss": 0.654,
+      "step": 23431
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6975741316433735e-06,
+      "loss": 0.5889,
+      "step": 23432
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6969672452068087e-06,
+      "loss": 0.6146,
+      "step": 23433
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6963604572131997e-06,
+      "loss": 0.6235,
+      "step": 23434
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6957537676697411e-06,
+      "loss": 0.6284,
+      "step": 23435
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.695147176583627e-06,
+      "loss": 0.6589,
+      "step": 23436
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6945406839620515e-06,
+      "loss": 0.6027,
+      "step": 23437
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6939342898122002e-06,
+      "loss": 0.6693,
+      "step": 23438
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.693327994141265e-06,
+      "loss": 0.6004,
+      "step": 23439
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6927217969564336e-06,
+      "loss": 0.6453,
+      "step": 23440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6921156982648946e-06,
+      "loss": 0.6544,
+      "step": 23441
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6915096980738344e-06,
+      "loss": 0.5685,
+      "step": 23442
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6909037963904383e-06,
+      "loss": 0.609,
+      "step": 23443
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6902979932218866e-06,
+      "loss": 0.6029,
+      "step": 23444
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6896922885753641e-06,
+      "loss": 0.6121,
+      "step": 23445
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6890866824580532e-06,
+      "loss": 0.6431,
+      "step": 23446
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6884811748771313e-06,
+      "loss": 0.6744,
+      "step": 23447
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6878757658397805e-06,
+      "loss": 0.5633,
+      "step": 23448
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6872704553531793e-06,
+      "loss": 0.6833,
+      "step": 23449
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6866652434245002e-06,
+      "loss": 0.5688,
+      "step": 23450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6860601300609203e-06,
+      "loss": 0.6955,
+      "step": 23451
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6854551152696152e-06,
+      "loss": 0.6736,
+      "step": 23452
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6848501990577582e-06,
+      "loss": 0.6064,
+      "step": 23453
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6842453814325199e-06,
+      "loss": 0.6413,
+      "step": 23454
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.683640662401076e-06,
+      "loss": 0.643,
+      "step": 23455
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6830360419705882e-06,
+      "loss": 0.641,
+      "step": 23456
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6824315201482299e-06,
+      "loss": 0.6437,
+      "step": 23457
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6818270969411677e-06,
+      "loss": 0.6215,
+      "step": 23458
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6812227723565677e-06,
+      "loss": 0.5828,
+      "step": 23459
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6806185464015957e-06,
+      "loss": 0.6259,
+      "step": 23460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6800144190834168e-06,
+      "loss": 0.6727,
+      "step": 23461
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6794103904091884e-06,
+      "loss": 0.672,
+      "step": 23462
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6788064603860766e-06,
+      "loss": 0.6426,
+      "step": 23463
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6782026290212406e-06,
+      "loss": 0.5593,
+      "step": 23464
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6775988963218393e-06,
+      "loss": 0.6126,
+      "step": 23465
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6769952622950302e-06,
+      "loss": 0.7035,
+      "step": 23466
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6763917269479745e-06,
+      "loss": 0.5812,
+      "step": 23467
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.675788290287822e-06,
+      "loss": 0.6273,
+      "step": 23468
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6751849523217289e-06,
+      "loss": 0.6667,
+      "step": 23469
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6745817130568487e-06,
+      "loss": 0.6877,
+      "step": 23470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6739785725003343e-06,
+      "loss": 0.641,
+      "step": 23471
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.673375530659337e-06,
+      "loss": 0.6067,
+      "step": 23472
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6727725875410083e-06,
+      "loss": 0.6375,
+      "step": 23473
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6721697431524919e-06,
+      "loss": 0.6611,
+      "step": 23474
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6715669975009374e-06,
+      "loss": 0.5521,
+      "step": 23475
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6709643505934925e-06,
+      "loss": 0.6245,
+      "step": 23476
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6703618024373014e-06,
+      "loss": 0.7013,
+      "step": 23477
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6697593530395072e-06,
+      "loss": 0.651,
+      "step": 23478
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6691570024072578e-06,
+      "loss": 0.6488,
+      "step": 23479
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6685547505476872e-06,
+      "loss": 0.6362,
+      "step": 23480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.667952597467939e-06,
+      "loss": 0.5826,
+      "step": 23481
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6673505431751535e-06,
+      "loss": 0.6765,
+      "step": 23482
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.666748587676469e-06,
+      "loss": 0.6205,
+      "step": 23483
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6661467309790224e-06,
+      "loss": 0.6304,
+      "step": 23484
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6655449730899477e-06,
+      "loss": 0.6802,
+      "step": 23485
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6649433140163796e-06,
+      "loss": 0.6496,
+      "step": 23486
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6643417537654548e-06,
+      "loss": 0.6672,
+      "step": 23487
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.663740292344298e-06,
+      "loss": 0.6195,
+      "step": 23488
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6631389297600487e-06,
+      "loss": 0.564,
+      "step": 23489
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6625376660198356e-06,
+      "loss": 0.6709,
+      "step": 23490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.661936501130783e-06,
+      "loss": 0.6428,
+      "step": 23491
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.661335435100021e-06,
+      "loss": 0.5832,
+      "step": 23492
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6607344679346782e-06,
+      "loss": 0.6303,
+      "step": 23493
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6601335996418733e-06,
+      "loss": 0.6052,
+      "step": 23494
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6595328302287362e-06,
+      "loss": 0.6183,
+      "step": 23495
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.65893215970239e-06,
+      "loss": 0.6144,
+      "step": 23496
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6583315880699536e-06,
+      "loss": 0.6528,
+      "step": 23497
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6577311153385478e-06,
+      "loss": 0.7047,
+      "step": 23498
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6571307415152948e-06,
+      "loss": 0.6503,
+      "step": 23499
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6565304666073056e-06,
+      "loss": 0.5955,
+      "step": 23500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6559302906217055e-06,
+      "loss": 0.6543,
+      "step": 23501
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.655330213565609e-06,
+      "loss": 0.6552,
+      "step": 23502
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.654730235446127e-06,
+      "loss": 0.6304,
+      "step": 23503
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6541303562703748e-06,
+      "loss": 0.6502,
+      "step": 23504
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6535305760454667e-06,
+      "loss": 0.6251,
+      "step": 23505
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6529308947785094e-06,
+      "loss": 0.6352,
+      "step": 23506
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.652331312476614e-06,
+      "loss": 0.6412,
+      "step": 23507
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6517318291468954e-06,
+      "loss": 0.6676,
+      "step": 23508
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6511324447964527e-06,
+      "loss": 0.5901,
+      "step": 23509
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.650533159432398e-06,
+      "loss": 0.651,
+      "step": 23510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6499339730618357e-06,
+      "loss": 0.6044,
+      "step": 23511
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6493348856918678e-06,
+      "loss": 0.63,
+      "step": 23512
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6487358973295965e-06,
+      "loss": 0.6528,
+      "step": 23513
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.648137007982129e-06,
+      "loss": 0.6498,
+      "step": 23514
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6475382176565602e-06,
+      "loss": 0.5843,
+      "step": 23515
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.646939526359993e-06,
+      "loss": 0.6256,
+      "step": 23516
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6463409340995252e-06,
+      "loss": 0.6257,
+      "step": 23517
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.645742440882252e-06,
+      "loss": 0.6223,
+      "step": 23518
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6451440467152668e-06,
+      "loss": 0.6072,
+      "step": 23519
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.644545751605674e-06,
+      "loss": 0.6179,
+      "step": 23520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6439475555605578e-06,
+      "loss": 0.6586,
+      "step": 23521
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6433494585870136e-06,
+      "loss": 0.6474,
+      "step": 23522
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6427514606921357e-06,
+      "loss": 0.5896,
+      "step": 23523
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6421535618830087e-06,
+      "loss": 0.6461,
+      "step": 23524
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.641555762166721e-06,
+      "loss": 0.5805,
+      "step": 23525
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6409580615503684e-06,
+      "loss": 0.5799,
+      "step": 23526
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6403604600410294e-06,
+      "loss": 0.6826,
+      "step": 23527
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.639762957645793e-06,
+      "loss": 0.6033,
+      "step": 23528
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6391655543717444e-06,
+      "loss": 0.6405,
+      "step": 23529
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6385682502259627e-06,
+      "loss": 0.6112,
+      "step": 23530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6379710452155285e-06,
+      "loss": 0.5966,
+      "step": 23531
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6373739393475308e-06,
+      "loss": 0.6206,
+      "step": 23532
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6367769326290417e-06,
+      "loss": 0.6624,
+      "step": 23533
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.636180025067141e-06,
+      "loss": 0.6724,
+      "step": 23534
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6355832166689089e-06,
+      "loss": 0.66,
+      "step": 23535
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6349865074414173e-06,
+      "loss": 0.6819,
+      "step": 23536
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6343898973917395e-06,
+      "loss": 0.653,
+      "step": 23537
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6337933865269562e-06,
+      "loss": 0.5999,
+      "step": 23538
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6331969748541331e-06,
+      "loss": 0.6149,
+      "step": 23539
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6326006623803448e-06,
+      "loss": 0.6407,
+      "step": 23540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.632004449112663e-06,
+      "loss": 0.6759,
+      "step": 23541
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6314083350581512e-06,
+      "loss": 0.5859,
+      "step": 23542
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6308123202238769e-06,
+      "loss": 0.6406,
+      "step": 23543
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6302164046169156e-06,
+      "loss": 0.6669,
+      "step": 23544
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6296205882443239e-06,
+      "loss": 0.6578,
+      "step": 23545
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6290248711131707e-06,
+      "loss": 0.6519,
+      "step": 23546
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6284292532305146e-06,
+      "loss": 0.6556,
+      "step": 23547
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.62783373460342e-06,
+      "loss": 0.6341,
+      "step": 23548
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.627238315238947e-06,
+      "loss": 0.6325,
+      "step": 23549
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6266429951441566e-06,
+      "loss": 0.6009,
+      "step": 23550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.626047774326104e-06,
+      "loss": 0.6543,
+      "step": 23551
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6254526527918512e-06,
+      "loss": 0.6506,
+      "step": 23552
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6248576305484497e-06,
+      "loss": 0.6187,
+      "step": 23553
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6242627076029549e-06,
+      "loss": 0.6205,
+      "step": 23554
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.623667883962421e-06,
+      "loss": 0.6089,
+      "step": 23555
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6230731596339e-06,
+      "loss": 0.6579,
+      "step": 23556
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6224785346244443e-06,
+      "loss": 0.6057,
+      "step": 23557
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6218840089411047e-06,
+      "loss": 0.6406,
+      "step": 23558
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.621289582590927e-06,
+      "loss": 0.5979,
+      "step": 23559
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6206952555809607e-06,
+      "loss": 0.6673,
+      "step": 23560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6201010279182506e-06,
+      "loss": 0.6748,
+      "step": 23561
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6195068996098451e-06,
+      "loss": 0.6667,
+      "step": 23562
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6189128706627855e-06,
+      "loss": 0.641,
+      "step": 23563
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6183189410841193e-06,
+      "loss": 0.5954,
+      "step": 23564
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.617725110880881e-06,
+      "loss": 0.5803,
+      "step": 23565
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6171313800601163e-06,
+      "loss": 0.5847,
+      "step": 23566
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6165377486288635e-06,
+      "loss": 0.5924,
+      "step": 23567
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6159442165941607e-06,
+      "loss": 0.6335,
+      "step": 23568
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6153507839630457e-06,
+      "loss": 0.6367,
+      "step": 23569
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6147574507425546e-06,
+      "loss": 0.5806,
+      "step": 23570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6141642169397199e-06,
+      "loss": 0.6587,
+      "step": 23571
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.613571082561577e-06,
+      "loss": 0.6439,
+      "step": 23572
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6129780476151568e-06,
+      "loss": 0.6831,
+      "step": 23573
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6123851121074919e-06,
+      "loss": 0.694,
+      "step": 23574
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6117922760456118e-06,
+      "loss": 0.6327,
+      "step": 23575
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6111995394365488e-06,
+      "loss": 0.6231,
+      "step": 23576
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6106069022873239e-06,
+      "loss": 0.6163,
+      "step": 23577
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.610014364604966e-06,
+      "loss": 0.6103,
+      "step": 23578
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6094219263965028e-06,
+      "loss": 0.654,
+      "step": 23579
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6088295876689553e-06,
+      "loss": 0.6686,
+      "step": 23580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.608237348429349e-06,
+      "loss": 0.6139,
+      "step": 23581
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.607645208684706e-06,
+      "loss": 0.648,
+      "step": 23582
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6070531684420442e-06,
+      "loss": 0.6881,
+      "step": 23583
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6064612277083835e-06,
+      "loss": 0.6412,
+      "step": 23584
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6058693864907426e-06,
+      "loss": 0.6359,
+      "step": 23585
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6052776447961393e-06,
+      "loss": 0.6111,
+      "step": 23586
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6046860026315892e-06,
+      "loss": 0.6415,
+      "step": 23587
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6040944600041086e-06,
+      "loss": 0.6397,
+      "step": 23588
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6035030169207067e-06,
+      "loss": 0.6682,
+      "step": 23589
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6029116733883977e-06,
+      "loss": 0.6274,
+      "step": 23590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6023204294141936e-06,
+      "loss": 0.668,
+      "step": 23591
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6017292850051037e-06,
+      "loss": 0.5841,
+      "step": 23592
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6011382401681375e-06,
+      "loss": 0.6689,
+      "step": 23593
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.600547294910304e-06,
+      "loss": 0.6302,
+      "step": 23594
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5999564492386054e-06,
+      "loss": 0.6923,
+      "step": 23595
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5993657031600495e-06,
+      "loss": 0.5981,
+      "step": 23596
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5987750566816396e-06,
+      "loss": 0.6102,
+      "step": 23597
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.59818450981038e-06,
+      "loss": 0.6336,
+      "step": 23598
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5975940625532705e-06,
+      "loss": 0.6906,
+      "step": 23599
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5970037149173145e-06,
+      "loss": 0.6422,
+      "step": 23600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5964134669095078e-06,
+      "loss": 0.6096,
+      "step": 23601
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5958233185368488e-06,
+      "loss": 0.6126,
+      "step": 23602
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5952332698063367e-06,
+      "loss": 0.6466,
+      "step": 23603
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5946433207249668e-06,
+      "loss": 0.6487,
+      "step": 23604
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5940534712997324e-06,
+      "loss": 0.6622,
+      "step": 23605
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5934637215376303e-06,
+      "loss": 0.6554,
+      "step": 23606
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.592874071445647e-06,
+      "loss": 0.6332,
+      "step": 23607
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.592284521030777e-06,
+      "loss": 0.5895,
+      "step": 23608
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.59169507030001e-06,
+      "loss": 0.6393,
+      "step": 23609
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5911057192603351e-06,
+      "loss": 0.6511,
+      "step": 23610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5905164679187412e-06,
+      "loss": 0.5737,
+      "step": 23611
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5899273162822094e-06,
+      "loss": 0.5967,
+      "step": 23612
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5893382643577294e-06,
+      "loss": 0.6198,
+      "step": 23613
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5887493121522824e-06,
+      "loss": 0.6306,
+      "step": 23614
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5881604596728539e-06,
+      "loss": 0.5959,
+      "step": 23615
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.587571706926424e-06,
+      "loss": 0.5802,
+      "step": 23616
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5869830539199748e-06,
+      "loss": 0.6078,
+      "step": 23617
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.586394500660482e-06,
+      "loss": 0.6116,
+      "step": 23618
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5858060471549263e-06,
+      "loss": 0.6744,
+      "step": 23619
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5852176934102824e-06,
+      "loss": 0.6588,
+      "step": 23620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5846294394335292e-06,
+      "loss": 0.6082,
+      "step": 23621
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5840412852316389e-06,
+      "loss": 0.6195,
+      "step": 23622
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.583453230811588e-06,
+      "loss": 0.657,
+      "step": 23623
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5828652761803432e-06,
+      "loss": 0.6051,
+      "step": 23624
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5822774213448777e-06,
+      "loss": 0.6245,
+      "step": 23625
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5816896663121628e-06,
+      "loss": 0.5916,
+      "step": 23626
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.581102011089166e-06,
+      "loss": 0.6301,
+      "step": 23627
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5805144556828544e-06,
+      "loss": 0.6616,
+      "step": 23628
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5799270001001977e-06,
+      "loss": 0.7073,
+      "step": 23629
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5793396443481556e-06,
+      "loss": 0.6699,
+      "step": 23630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.578752388433693e-06,
+      "loss": 0.6478,
+      "step": 23631
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5781652323637753e-06,
+      "loss": 0.6287,
+      "step": 23632
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5775781761453612e-06,
+      "loss": 0.641,
+      "step": 23633
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5769912197854131e-06,
+      "loss": 0.6066,
+      "step": 23634
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.576404363290891e-06,
+      "loss": 0.6342,
+      "step": 23635
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5758176066687492e-06,
+      "loss": 0.6001,
+      "step": 23636
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5752309499259454e-06,
+      "loss": 0.6428,
+      "step": 23637
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5746443930694367e-06,
+      "loss": 0.635,
+      "step": 23638
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.574057936106177e-06,
+      "loss": 0.6811,
+      "step": 23639
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5734715790431188e-06,
+      "loss": 0.6526,
+      "step": 23640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5728853218872175e-06,
+      "loss": 0.6562,
+      "step": 23641
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5722991646454189e-06,
+      "loss": 0.6003,
+      "step": 23642
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.571713107324675e-06,
+      "loss": 0.6564,
+      "step": 23643
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5711271499319336e-06,
+      "loss": 0.6149,
+      "step": 23644
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5705412924741414e-06,
+      "loss": 0.6405,
+      "step": 23645
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5699555349582474e-06,
+      "loss": 0.6336,
+      "step": 23646
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5693698773911959e-06,
+      "loss": 0.6228,
+      "step": 23647
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5687843197799269e-06,
+      "loss": 0.5782,
+      "step": 23648
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5681988621313847e-06,
+      "loss": 0.615,
+      "step": 23649
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.567613504452513e-06,
+      "loss": 0.6057,
+      "step": 23650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5670282467502497e-06,
+      "loss": 0.5652,
+      "step": 23651
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.566443089031533e-06,
+      "loss": 0.6017,
+      "step": 23652
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5658580313033056e-06,
+      "loss": 0.6231,
+      "step": 23653
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5652730735724974e-06,
+      "loss": 0.642,
+      "step": 23654
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5646882158460474e-06,
+      "loss": 0.6056,
+      "step": 23655
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5641034581308889e-06,
+      "loss": 0.61,
+      "step": 23656
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5635188004339563e-06,
+      "loss": 0.6932,
+      "step": 23657
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5629342427621797e-06,
+      "loss": 0.6384,
+      "step": 23658
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5623497851224934e-06,
+      "loss": 0.6259,
+      "step": 23659
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5617654275218209e-06,
+      "loss": 0.6066,
+      "step": 23660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5611811699670943e-06,
+      "loss": 0.5846,
+      "step": 23661
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5605970124652404e-06,
+      "loss": 0.6158,
+      "step": 23662
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5600129550231845e-06,
+      "loss": 0.583,
+      "step": 23663
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5594289976478516e-06,
+      "loss": 0.655,
+      "step": 23664
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.558845140346168e-06,
+      "loss": 0.6449,
+      "step": 23665
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5582613831250504e-06,
+      "loss": 0.659,
+      "step": 23666
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5576777259914234e-06,
+      "loss": 0.7006,
+      "step": 23667
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5570941689522058e-06,
+      "loss": 0.6724,
+      "step": 23668
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5565107120143186e-06,
+      "loss": 0.6869,
+      "step": 23669
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5559273551846788e-06,
+      "loss": 0.6285,
+      "step": 23670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5553440984702005e-06,
+      "loss": 0.6602,
+      "step": 23671
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.554760941877801e-06,
+      "loss": 0.6476,
+      "step": 23672
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5541778854143929e-06,
+      "loss": 0.6277,
+      "step": 23673
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.55359492908689e-06,
+      "loss": 0.6888,
+      "step": 23674
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5530120729022035e-06,
+      "loss": 0.6759,
+      "step": 23675
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5524293168672478e-06,
+      "loss": 0.6746,
+      "step": 23676
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.551846660988926e-06,
+      "loss": 0.6337,
+      "step": 23677
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5512641052741494e-06,
+      "loss": 0.6316,
+      "step": 23678
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5506816497298238e-06,
+      "loss": 0.6035,
+      "step": 23679
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5500992943628568e-06,
+      "loss": 0.6137,
+      "step": 23680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5495170391801506e-06,
+      "loss": 0.6322,
+      "step": 23681
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.548934884188613e-06,
+      "loss": 0.6366,
+      "step": 23682
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.54835282939514e-06,
+      "loss": 0.6093,
+      "step": 23683
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5477708748066357e-06,
+      "loss": 0.6348,
+      "step": 23684
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.54718902043e-06,
+      "loss": 0.6239,
+      "step": 23685
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.546607266272132e-06,
+      "loss": 0.6207,
+      "step": 23686
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5460256123399275e-06,
+      "loss": 0.6295,
+      "step": 23687
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5454440586402864e-06,
+      "loss": 0.6147,
+      "step": 23688
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5448626051800985e-06,
+      "loss": 0.6226,
+      "step": 23689
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.544281251966261e-06,
+      "loss": 0.5993,
+      "step": 23690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5436999990056646e-06,
+      "loss": 0.5381,
+      "step": 23691
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5431188463052028e-06,
+      "loss": 0.6244,
+      "step": 23692
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5425377938717644e-06,
+      "loss": 0.6639,
+      "step": 23693
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5419568417122421e-06,
+      "loss": 0.6511,
+      "step": 23694
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.541375989833519e-06,
+      "loss": 0.654,
+      "step": 23695
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5407952382424829e-06,
+      "loss": 0.6135,
+      "step": 23696
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5402145869460228e-06,
+      "loss": 0.6346,
+      "step": 23697
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5396340359510164e-06,
+      "loss": 0.6394,
+      "step": 23698
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.539053585264353e-06,
+      "loss": 0.6086,
+      "step": 23699
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5384732348929143e-06,
+      "loss": 0.674,
+      "step": 23700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5378929848435776e-06,
+      "loss": 0.6511,
+      "step": 23701
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5373128351232236e-06,
+      "loss": 0.6952,
+      "step": 23702
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5367327857387337e-06,
+      "loss": 0.6474,
+      "step": 23703
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5361528366969781e-06,
+      "loss": 0.6551,
+      "step": 23704
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5355729880048398e-06,
+      "loss": 0.5565,
+      "step": 23705
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5349932396691936e-06,
+      "loss": 0.6147,
+      "step": 23706
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5344135916969084e-06,
+      "loss": 0.6663,
+      "step": 23707
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5338340440948595e-06,
+      "loss": 0.6086,
+      "step": 23708
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5332545968699198e-06,
+      "loss": 0.6659,
+      "step": 23709
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5326752500289522e-06,
+      "loss": 0.6664,
+      "step": 23710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5320960035788334e-06,
+      "loss": 0.6437,
+      "step": 23711
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5315168575264317e-06,
+      "loss": 0.5965,
+      "step": 23712
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5309378118786066e-06,
+      "loss": 0.5973,
+      "step": 23713
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5303588666422287e-06,
+      "loss": 0.6823,
+      "step": 23714
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.529780021824162e-06,
+      "loss": 0.6205,
+      "step": 23715
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5292012774312626e-06,
+      "loss": 0.6395,
+      "step": 23716
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5286226334704013e-06,
+      "loss": 0.6278,
+      "step": 23717
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5280440899484372e-06,
+      "loss": 0.6048,
+      "step": 23718
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5274656468722249e-06,
+      "loss": 0.6379,
+      "step": 23719
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5268873042486243e-06,
+      "loss": 0.5879,
+      "step": 23720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5263090620844968e-06,
+      "loss": 0.6561,
+      "step": 23721
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5257309203866898e-06,
+      "loss": 0.6062,
+      "step": 23722
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.525152879162065e-06,
+      "loss": 0.621,
+      "step": 23723
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5245749384174758e-06,
+      "loss": 0.6752,
+      "step": 23724
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5239970981597697e-06,
+      "loss": 0.594,
+      "step": 23725
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5234193583958e-06,
+      "loss": 0.5786,
+      "step": 23726
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5228417191324197e-06,
+      "loss": 0.5866,
+      "step": 23727
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5222641803764692e-06,
+      "loss": 0.6572,
+      "step": 23728
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5216867421348035e-06,
+      "loss": 0.6192,
+      "step": 23729
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5211094044142694e-06,
+      "loss": 0.6454,
+      "step": 23730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5205321672217066e-06,
+      "loss": 0.6553,
+      "step": 23731
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5199550305639621e-06,
+      "loss": 0.5977,
+      "step": 23732
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5193779944478793e-06,
+      "loss": 0.6001,
+      "step": 23733
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5188010588802949e-06,
+      "loss": 0.6895,
+      "step": 23734
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5182242238680567e-06,
+      "loss": 0.6318,
+      "step": 23735
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.517647489417997e-06,
+      "loss": 0.6254,
+      "step": 23736
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5170708555369574e-06,
+      "loss": 0.5948,
+      "step": 23737
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5164943222317763e-06,
+      "loss": 0.6653,
+      "step": 23738
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.515917889509283e-06,
+      "loss": 0.5969,
+      "step": 23739
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.515341557376314e-06,
+      "loss": 0.5986,
+      "step": 23740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.51476532583971e-06,
+      "loss": 0.6857,
+      "step": 23741
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5141891949062927e-06,
+      "loss": 0.5995,
+      "step": 23742
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5136131645828977e-06,
+      "loss": 0.6712,
+      "step": 23743
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5130372348763567e-06,
+      "loss": 0.6457,
+      "step": 23744
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5124614057934928e-06,
+      "loss": 0.6262,
+      "step": 23745
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5118856773411327e-06,
+      "loss": 0.6713,
+      "step": 23746
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5113100495261112e-06,
+      "loss": 0.688,
+      "step": 23747
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5107345223552438e-06,
+      "loss": 0.6206,
+      "step": 23748
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5101590958353584e-06,
+      "loss": 0.6184,
+      "step": 23749
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5095837699732796e-06,
+      "loss": 0.6002,
+      "step": 23750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5090085447758218e-06,
+      "loss": 0.5846,
+      "step": 23751
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5084334202498075e-06,
+      "loss": 0.6898,
+      "step": 23752
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.507858396402061e-06,
+      "loss": 0.6131,
+      "step": 23753
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5072834732393938e-06,
+      "loss": 0.6614,
+      "step": 23754
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5067086507686245e-06,
+      "loss": 0.6736,
+      "step": 23755
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5061339289965704e-06,
+      "loss": 0.6,
+      "step": 23756
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5055593079300412e-06,
+      "loss": 0.6209,
+      "step": 23757
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5049847875758494e-06,
+      "loss": 0.6525,
+      "step": 23758
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.504410367940814e-06,
+      "loss": 0.6448,
+      "step": 23759
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5038360490317382e-06,
+      "loss": 0.6132,
+      "step": 23760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5032618308554337e-06,
+      "loss": 0.5856,
+      "step": 23761
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5026877134187113e-06,
+      "loss": 0.6666,
+      "step": 23762
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5021136967283722e-06,
+      "loss": 0.6277,
+      "step": 23763
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5015397807912235e-06,
+      "loss": 0.6237,
+      "step": 23764
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5009659656140762e-06,
+      "loss": 0.6412,
+      "step": 23765
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.500392251203725e-06,
+      "loss": 0.597,
+      "step": 23766
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4998186375669765e-06,
+      "loss": 0.6139,
+      "step": 23767
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4992451247106344e-06,
+      "loss": 0.6434,
+      "step": 23768
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4986717126414907e-06,
+      "loss": 0.6179,
+      "step": 23769
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.498098401366347e-06,
+      "loss": 0.5717,
+      "step": 23770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4975251908920053e-06,
+      "loss": 0.6289,
+      "step": 23771
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4969520812252559e-06,
+      "loss": 0.6101,
+      "step": 23772
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4963790723728966e-06,
+      "loss": 0.6509,
+      "step": 23773
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4958061643417232e-06,
+      "loss": 0.6456,
+      "step": 23774
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4952333571385224e-06,
+      "loss": 0.6269,
+      "step": 23775
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4946606507700866e-06,
+      "loss": 0.5851,
+      "step": 23776
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4940880452432128e-06,
+      "loss": 0.7139,
+      "step": 23777
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4935155405646828e-06,
+      "loss": 0.645,
+      "step": 23778
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4929431367412872e-06,
+      "loss": 0.6499,
+      "step": 23779
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4923708337798148e-06,
+      "loss": 0.6637,
+      "step": 23780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4917986316870459e-06,
+      "loss": 0.6143,
+      "step": 23781
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4912265304697637e-06,
+      "loss": 0.6482,
+      "step": 23782
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4906545301347608e-06,
+      "loss": 0.635,
+      "step": 23783
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4900826306888093e-06,
+      "loss": 0.6012,
+      "step": 23784
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4895108321386942e-06,
+      "loss": 0.6001,
+      "step": 23785
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4889391344911952e-06,
+      "loss": 0.6064,
+      "step": 23786
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4883675377530882e-06,
+      "loss": 0.5504,
+      "step": 23787
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4877960419311487e-06,
+      "loss": 0.6783,
+      "step": 23788
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4872246470321594e-06,
+      "loss": 0.6204,
+      "step": 23789
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.486653353062889e-06,
+      "loss": 0.6794,
+      "step": 23790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.486082160030111e-06,
+      "loss": 0.6871,
+      "step": 23791
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4855110679406026e-06,
+      "loss": 0.6235,
+      "step": 23792
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4849400768011291e-06,
+      "loss": 0.6681,
+      "step": 23793
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4843691866184606e-06,
+      "loss": 0.6064,
+      "step": 23794
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.483798397399372e-06,
+      "loss": 0.6501,
+      "step": 23795
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4832277091506241e-06,
+      "loss": 0.6305,
+      "step": 23796
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4826571218789875e-06,
+      "loss": 0.6688,
+      "step": 23797
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4820866355912234e-06,
+      "loss": 0.6817,
+      "step": 23798
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4815162502940972e-06,
+      "loss": 0.6536,
+      "step": 23799
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4809459659943714e-06,
+      "loss": 0.6768,
+      "step": 23800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4803757826988086e-06,
+      "loss": 0.6491,
+      "step": 23801
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4798057004141675e-06,
+      "loss": 0.5981,
+      "step": 23802
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4792357191472106e-06,
+      "loss": 0.6165,
+      "step": 23803
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4786658389046904e-06,
+      "loss": 0.596,
+      "step": 23804
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4780960596933669e-06,
+      "loss": 0.6726,
+      "step": 23805
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4775263815199947e-06,
+      "loss": 0.6532,
+      "step": 23806
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4769568043913284e-06,
+      "loss": 0.6397,
+      "step": 23807
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4763873283141206e-06,
+      "loss": 0.6064,
+      "step": 23808
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4758179532951255e-06,
+      "loss": 0.6457,
+      "step": 23809
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4752486793410904e-06,
+      "loss": 0.5699,
+      "step": 23810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.474679506458765e-06,
+      "loss": 0.6715,
+      "step": 23811
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4741104346548985e-06,
+      "loss": 0.6092,
+      "step": 23812
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.473541463936239e-06,
+      "loss": 0.5936,
+      "step": 23813
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4729725943095309e-06,
+      "loss": 0.6358,
+      "step": 23814
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.472403825781522e-06,
+      "loss": 0.6156,
+      "step": 23815
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4718351583589508e-06,
+      "loss": 0.6447,
+      "step": 23816
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4712665920485614e-06,
+      "loss": 0.6522,
+      "step": 23817
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4706981268570963e-06,
+      "loss": 0.6,
+      "step": 23818
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4701297627912947e-06,
+      "loss": 0.6222,
+      "step": 23819
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4695614998578945e-06,
+      "loss": 0.6205,
+      "step": 23820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4689933380636357e-06,
+      "loss": 0.6128,
+      "step": 23821
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4684252774152508e-06,
+      "loss": 0.6668,
+      "step": 23822
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4678573179194778e-06,
+      "loss": 0.6249,
+      "step": 23823
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4672894595830478e-06,
+      "loss": 0.6311,
+      "step": 23824
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4667217024126967e-06,
+      "loss": 0.5693,
+      "step": 23825
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4661540464151535e-06,
+      "loss": 0.6349,
+      "step": 23826
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4655864915971519e-06,
+      "loss": 0.64,
+      "step": 23827
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.465019037965416e-06,
+      "loss": 0.6766,
+      "step": 23828
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4644516855266767e-06,
+      "loss": 0.6608,
+      "step": 23829
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4638844342876602e-06,
+      "loss": 0.612,
+      "step": 23830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4633172842550914e-06,
+      "loss": 0.6715,
+      "step": 23831
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.462750235435696e-06,
+      "loss": 0.7011,
+      "step": 23832
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4621832878361975e-06,
+      "loss": 0.6816,
+      "step": 23833
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4616164414633149e-06,
+      "loss": 0.6216,
+      "step": 23834
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4610496963237709e-06,
+      "loss": 0.6172,
+      "step": 23835
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.460483052424284e-06,
+      "loss": 0.6496,
+      "step": 23836
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4599165097715727e-06,
+      "loss": 0.5721,
+      "step": 23837
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.459350068372355e-06,
+      "loss": 0.6726,
+      "step": 23838
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4587837282333484e-06,
+      "loss": 0.6314,
+      "step": 23839
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4582174893612632e-06,
+      "loss": 0.5748,
+      "step": 23840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4576513517628144e-06,
+      "loss": 0.6417,
+      "step": 23841
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4570853154447162e-06,
+      "loss": 0.6347,
+      "step": 23842
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4565193804136769e-06,
+      "loss": 0.6122,
+      "step": 23843
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4559535466764096e-06,
+      "loss": 0.6502,
+      "step": 23844
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4553878142396228e-06,
+      "loss": 0.6467,
+      "step": 23845
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4548221831100196e-06,
+      "loss": 0.6237,
+      "step": 23846
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4542566532943092e-06,
+      "loss": 0.6639,
+      "step": 23847
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4536912247991975e-06,
+      "loss": 0.6349,
+      "step": 23848
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4531258976313868e-06,
+      "loss": 0.5806,
+      "step": 23849
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4525606717975804e-06,
+      "loss": 0.6306,
+      "step": 23850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4519955473044822e-06,
+      "loss": 0.6339,
+      "step": 23851
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4514305241587877e-06,
+      "loss": 0.5798,
+      "step": 23852
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4508656023671986e-06,
+      "loss": 0.6788,
+      "step": 23853
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4503007819364124e-06,
+      "loss": 0.6518,
+      "step": 23854
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449736062873126e-06,
+      "loss": 0.604,
+      "step": 23855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4491714451840343e-06,
+      "loss": 0.6104,
+      "step": 23856
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4486069288758354e-06,
+      "loss": 0.6073,
+      "step": 23857
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4480425139552157e-06,
+      "loss": 0.6016,
+      "step": 23858
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4474782004288702e-06,
+      "loss": 0.6468,
+      "step": 23859
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44691398830349e-06,
+      "loss": 0.6607,
+      "step": 23860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446349877585763e-06,
+      "loss": 0.5978,
+      "step": 23861
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4457858682823821e-06,
+      "loss": 0.6368,
+      "step": 23862
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4452219604000284e-06,
+      "loss": 0.6026,
+      "step": 23863
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4446581539453907e-06,
+      "loss": 0.6136,
+      "step": 23864
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4440944489251519e-06,
+      "loss": 0.6518,
+      "step": 23865
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4435308453459973e-06,
+      "loss": 0.6236,
+      "step": 23866
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4429673432146085e-06,
+      "loss": 0.6346,
+      "step": 23867
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.442403942537669e-06,
+      "loss": 0.605,
+      "step": 23868
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4418406433218546e-06,
+      "loss": 0.6052,
+      "step": 23869
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4412774455738444e-06,
+      "loss": 0.6007,
+      "step": 23870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4407143493003173e-06,
+      "loss": 0.6072,
+      "step": 23871
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4401513545079493e-06,
+      "loss": 0.6469,
+      "step": 23872
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.439588461203415e-06,
+      "loss": 0.6599,
+      "step": 23873
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4390256693933902e-06,
+      "loss": 0.6373,
+      "step": 23874
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.438462979084544e-06,
+      "loss": 0.5979,
+      "step": 23875
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4379003902835497e-06,
+      "loss": 0.6342,
+      "step": 23876
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4373379029970779e-06,
+      "loss": 0.5868,
+      "step": 23877
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4367755172317965e-06,
+      "loss": 0.6402,
+      "step": 23878
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4362132329943735e-06,
+      "loss": 0.6201,
+      "step": 23879
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4356510502914788e-06,
+      "loss": 0.6351,
+      "step": 23880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.435088969129772e-06,
+      "loss": 0.6594,
+      "step": 23881
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.434526989515921e-06,
+      "loss": 0.6669,
+      "step": 23882
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4339651114565877e-06,
+      "loss": 0.7147,
+      "step": 23883
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4334033349584332e-06,
+      "loss": 0.5992,
+      "step": 23884
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4328416600281202e-06,
+      "loss": 0.6012,
+      "step": 23885
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4322800866723085e-06,
+      "loss": 0.6044,
+      "step": 23886
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4317186148976514e-06,
+      "loss": 0.6276,
+      "step": 23887
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4311572447108102e-06,
+      "loss": 0.6732,
+      "step": 23888
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4305959761184397e-06,
+      "loss": 0.6119,
+      "step": 23889
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4300348091271932e-06,
+      "loss": 0.6827,
+      "step": 23890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4294737437437256e-06,
+      "loss": 0.6599,
+      "step": 23891
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4289127799746915e-06,
+      "loss": 0.6265,
+      "step": 23892
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4283519178267358e-06,
+      "loss": 0.6933,
+      "step": 23893
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4277911573065117e-06,
+      "loss": 0.641,
+      "step": 23894
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4272304984206674e-06,
+      "loss": 0.6369,
+      "step": 23895
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.42666994117585e-06,
+      "loss": 0.6179,
+      "step": 23896
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.426109485578705e-06,
+      "loss": 0.6336,
+      "step": 23897
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4255491316358816e-06,
+      "loss": 0.6582,
+      "step": 23898
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4249888793540167e-06,
+      "loss": 0.6414,
+      "step": 23899
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4244287287397563e-06,
+      "loss": 0.6386,
+      "step": 23900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4238686797997425e-06,
+      "loss": 0.577,
+      "step": 23901
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4233087325406136e-06,
+      "loss": 0.6264,
+      "step": 23902
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4227488869690087e-06,
+      "loss": 0.5958,
+      "step": 23903
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4221891430915692e-06,
+      "loss": 0.6118,
+      "step": 23904
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4216295009149251e-06,
+      "loss": 0.6385,
+      "step": 23905
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4210699604457156e-06,
+      "loss": 0.6439,
+      "step": 23906
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4205105216905747e-06,
+      "loss": 0.6475,
+      "step": 23907
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4199511846561343e-06,
+      "loss": 0.6094,
+      "step": 23908
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.419391949349026e-06,
+      "loss": 0.6234,
+      "step": 23909
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4188328157758824e-06,
+      "loss": 0.6629,
+      "step": 23910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4182737839433303e-06,
+      "loss": 0.6034,
+      "step": 23911
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4177148538579978e-06,
+      "loss": 0.7097,
+      "step": 23912
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4171560255265114e-06,
+      "loss": 0.6545,
+      "step": 23913
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4165972989554988e-06,
+      "loss": 0.6249,
+      "step": 23914
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.416038674151583e-06,
+      "loss": 0.7104,
+      "step": 23915
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4154801511213889e-06,
+      "loss": 0.609,
+      "step": 23916
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4149217298715356e-06,
+      "loss": 0.6843,
+      "step": 23917
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4143634104086458e-06,
+      "loss": 0.5758,
+      "step": 23918
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4138051927393381e-06,
+      "loss": 0.6432,
+      "step": 23919
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4132470768702323e-06,
+      "loss": 0.5854,
+      "step": 23920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.412689062807946e-06,
+      "loss": 0.6914,
+      "step": 23921
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4121311505590918e-06,
+      "loss": 0.5988,
+      "step": 23922
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4115733401302866e-06,
+      "loss": 0.6345,
+      "step": 23923
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4110156315281453e-06,
+      "loss": 0.5423,
+      "step": 23924
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4104580247592758e-06,
+      "loss": 0.6468,
+      "step": 23925
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.409900519830293e-06,
+      "loss": 0.5911,
+      "step": 23926
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.40934311674781e-06,
+      "loss": 0.5857,
+      "step": 23927
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4087858155184274e-06,
+      "loss": 0.6259,
+      "step": 23928
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.408228616148758e-06,
+      "loss": 0.6565,
+      "step": 23929
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4076715186454094e-06,
+      "loss": 0.6777,
+      "step": 23930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4071145230149797e-06,
+      "loss": 0.5936,
+      "step": 23931
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4065576292640792e-06,
+      "loss": 0.566,
+      "step": 23932
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4060008373993118e-06,
+      "loss": 0.5878,
+      "step": 23933
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.405444147427273e-06,
+      "loss": 0.623,
+      "step": 23934
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4048875593545663e-06,
+      "loss": 0.6267,
+      "step": 23935
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4043310731877923e-06,
+      "loss": 0.579,
+      "step": 23936
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4037746889335435e-06,
+      "loss": 0.6399,
+      "step": 23937
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4032184065984212e-06,
+      "loss": 0.6092,
+      "step": 23938
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.402662226189022e-06,
+      "loss": 0.6273,
+      "step": 23939
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4021061477119357e-06,
+      "loss": 0.6354,
+      "step": 23940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4015501711737578e-06,
+      "loss": 0.6354,
+      "step": 23941
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4009942965810818e-06,
+      "loss": 0.624,
+      "step": 23942
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4004385239404917e-06,
+      "loss": 0.6455,
+      "step": 23943
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.399882853258584e-06,
+      "loss": 0.6295,
+      "step": 23944
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3993272845419458e-06,
+      "loss": 0.6096,
+      "step": 23945
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.398771817797161e-06,
+      "loss": 0.6596,
+      "step": 23946
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3982164530308162e-06,
+      "loss": 0.7171,
+      "step": 23947
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3976611902494997e-06,
+      "loss": 0.6215,
+      "step": 23948
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3971060294597872e-06,
+      "loss": 0.5899,
+      "step": 23949
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3965509706682668e-06,
+      "loss": 0.664,
+      "step": 23950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.395996013881521e-06,
+      "loss": 0.6226,
+      "step": 23951
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3954411591061245e-06,
+      "loss": 0.6868,
+      "step": 23952
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3948864063486578e-06,
+      "loss": 0.6,
+      "step": 23953
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3943317556156998e-06,
+      "loss": 0.651,
+      "step": 23954
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.393777206913821e-06,
+      "loss": 0.6445,
+      "step": 23955
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3932227602496029e-06,
+      "loss": 0.6175,
+      "step": 23956
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3926684156296188e-06,
+      "loss": 0.6803,
+      "step": 23957
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.392114173060436e-06,
+      "loss": 0.6467,
+      "step": 23958
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.39156003254863e-06,
+      "loss": 0.6734,
+      "step": 23959
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3910059941007703e-06,
+      "loss": 0.671,
+      "step": 23960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3904520577234215e-06,
+      "loss": 0.6077,
+      "step": 23961
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3898982234231574e-06,
+      "loss": 0.6415,
+      "step": 23962
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3893444912065423e-06,
+      "loss": 0.6408,
+      "step": 23963
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3887908610801392e-06,
+      "loss": 0.6176,
+      "step": 23964
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.388237333050515e-06,
+      "loss": 0.6291,
+      "step": 23965
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.387683907124232e-06,
+      "loss": 0.681,
+      "step": 23966
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3871305833078464e-06,
+      "loss": 0.5863,
+      "step": 23967
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3865773616079258e-06,
+      "loss": 0.5991,
+      "step": 23968
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3860242420310287e-06,
+      "loss": 0.632,
+      "step": 23969
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3854712245837087e-06,
+      "loss": 0.6289,
+      "step": 23970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.384918309272525e-06,
+      "loss": 0.6287,
+      "step": 23971
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3843654961040354e-06,
+      "loss": 0.5779,
+      "step": 23972
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.383812785084786e-06,
+      "loss": 0.6499,
+      "step": 23973
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3832601762213383e-06,
+      "loss": 0.6553,
+      "step": 23974
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3827076695202446e-06,
+      "loss": 0.573,
+      "step": 23975
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3821552649880487e-06,
+      "loss": 0.6038,
+      "step": 23976
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3816029626313055e-06,
+      "loss": 0.6173,
+      "step": 23977
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3810507624565627e-06,
+      "loss": 0.6342,
+      "step": 23978
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.380498664470361e-06,
+      "loss": 0.6678,
+      "step": 23979
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.379946668679255e-06,
+      "loss": 0.591,
+      "step": 23980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3793947750897873e-06,
+      "loss": 0.6592,
+      "step": 23981
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3788429837084971e-06,
+      "loss": 0.6339,
+      "step": 23982
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3782912945419314e-06,
+      "loss": 0.5841,
+      "step": 23983
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.377739707596627e-06,
+      "loss": 0.6351,
+      "step": 23984
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3771882228791234e-06,
+      "loss": 0.6501,
+      "step": 23985
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3766368403959662e-06,
+      "loss": 0.6322,
+      "step": 23986
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3760855601536848e-06,
+      "loss": 0.637,
+      "step": 23987
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3755343821588185e-06,
+      "loss": 0.66,
+      "step": 23988
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3749833064179053e-06,
+      "loss": 0.6248,
+      "step": 23989
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.374432332937472e-06,
+      "loss": 0.5907,
+      "step": 23990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3738814617240536e-06,
+      "loss": 0.6078,
+      "step": 23991
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3733306927841871e-06,
+      "loss": 0.6284,
+      "step": 23992
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.372780026124395e-06,
+      "loss": 0.5953,
+      "step": 23993
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.37222946175121e-06,
+      "loss": 0.6653,
+      "step": 23994
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.37167899967116e-06,
+      "loss": 0.6043,
+      "step": 23995
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3711286398907698e-06,
+      "loss": 0.6298,
+      "step": 23996
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.370578382416562e-06,
+      "loss": 0.5899,
+      "step": 23997
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3700282272550681e-06,
+      "loss": 0.5905,
+      "step": 23998
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3694781744128038e-06,
+      "loss": 0.6117,
+      "step": 23999
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3689282238962942e-06,
+      "loss": 0.6275,
+      "step": 24000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3683783757120605e-06,
+      "loss": 0.683,
+      "step": 24001
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3678286298666187e-06,
+      "loss": 0.5866,
+      "step": 24002
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.367278986366486e-06,
+      "loss": 0.6688,
+      "step": 24003
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3667294452181857e-06,
+      "loss": 0.6468,
+      "step": 24004
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.366180006428226e-06,
+      "loss": 0.6407,
+      "step": 24005
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3656306700031253e-06,
+      "loss": 0.5994,
+      "step": 24006
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3650814359493968e-06,
+      "loss": 0.6335,
+      "step": 24007
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3645323042735492e-06,
+      "loss": 0.572,
+      "step": 24008
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3639832749820935e-06,
+      "loss": 0.6543,
+      "step": 24009
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3634343480815437e-06,
+      "loss": 0.6924,
+      "step": 24010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3628855235784045e-06,
+      "loss": 0.6178,
+      "step": 24011
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3623368014791827e-06,
+      "loss": 0.6865,
+      "step": 24012
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3617881817903866e-06,
+      "loss": 0.6684,
+      "step": 24013
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3612396645185166e-06,
+      "loss": 0.6224,
+      "step": 24014
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3606912496700775e-06,
+      "loss": 0.6825,
+      "step": 24015
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3601429372515761e-06,
+      "loss": 0.5939,
+      "step": 24016
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3595947272695076e-06,
+      "loss": 0.5982,
+      "step": 24017
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.359046619730373e-06,
+      "loss": 0.5657,
+      "step": 24018
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3584986146406742e-06,
+      "loss": 0.6237,
+      "step": 24019
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3579507120069047e-06,
+      "loss": 0.6734,
+      "step": 24020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.357402911835558e-06,
+      "loss": 0.6667,
+      "step": 24021
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3568552141331382e-06,
+      "loss": 0.6495,
+      "step": 24022
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3563076189061308e-06,
+      "loss": 0.6277,
+      "step": 24023
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.355760126161031e-06,
+      "loss": 0.6255,
+      "step": 24024
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.355212735904332e-06,
+      "loss": 0.6471,
+      "step": 24025
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3546654481425202e-06,
+      "loss": 0.6007,
+      "step": 24026
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3541182628820826e-06,
+      "loss": 0.5953,
+      "step": 24027
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.353571180129516e-06,
+      "loss": 0.6562,
+      "step": 24028
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3530241998912974e-06,
+      "loss": 0.6123,
+      "step": 24029
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3524773221739152e-06,
+      "loss": 0.6172,
+      "step": 24030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3519305469838563e-06,
+      "loss": 0.6479,
+      "step": 24031
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3513838743275987e-06,
+      "loss": 0.6645,
+      "step": 24032
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.350837304211623e-06,
+      "loss": 0.6712,
+      "step": 24033
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3502908366424171e-06,
+      "loss": 0.6343,
+      "step": 24034
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.349744471626453e-06,
+      "loss": 0.6117,
+      "step": 24035
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3491982091702116e-06,
+      "loss": 0.5969,
+      "step": 24036
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3486520492801702e-06,
+      "loss": 0.6569,
+      "step": 24037
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3481059919628004e-06,
+      "loss": 0.6164,
+      "step": 24038
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3475600372245768e-06,
+      "loss": 0.6001,
+      "step": 24039
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3470141850719786e-06,
+      "loss": 0.6077,
+      "step": 24040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.346468435511471e-06,
+      "loss": 0.6492,
+      "step": 24041
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3459227885495274e-06,
+      "loss": 0.6075,
+      "step": 24042
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3453772441926184e-06,
+      "loss": 0.6518,
+      "step": 24043
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3448318024472073e-06,
+      "loss": 0.6082,
+      "step": 24044
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.344286463319765e-06,
+      "loss": 0.6112,
+      "step": 24045
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.343741226816756e-06,
+      "loss": 0.6213,
+      "step": 24046
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3431960929446454e-06,
+      "loss": 0.5726,
+      "step": 24047
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3426510617098976e-06,
+      "loss": 0.5874,
+      "step": 24048
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3421061331189723e-06,
+      "loss": 0.6225,
+      "step": 24049
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3415613071783296e-06,
+      "loss": 0.5861,
+      "step": 24050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.341016583894431e-06,
+      "loss": 0.5974,
+      "step": 24051
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.340471963273734e-06,
+      "loss": 0.6161,
+      "step": 24052
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3399274453226973e-06,
+      "loss": 0.6179,
+      "step": 24053
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3393830300477772e-06,
+      "loss": 0.6366,
+      "step": 24054
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3388387174554252e-06,
+      "loss": 0.5865,
+      "step": 24055
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.338294507552096e-06,
+      "loss": 0.6363,
+      "step": 24056
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3377504003442433e-06,
+      "loss": 0.6579,
+      "step": 24057
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3372063958383164e-06,
+      "loss": 0.7104,
+      "step": 24058
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3366624940407668e-06,
+      "loss": 0.6262,
+      "step": 24059
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3361186949580451e-06,
+      "loss": 0.6238,
+      "step": 24060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3355749985965938e-06,
+      "loss": 0.6332,
+      "step": 24061
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.335031404962861e-06,
+      "loss": 0.6359,
+      "step": 24062
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3344879140632917e-06,
+      "loss": 0.6407,
+      "step": 24063
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3339445259043314e-06,
+      "loss": 0.6144,
+      "step": 24064
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3334012404924201e-06,
+      "loss": 0.6593,
+      "step": 24065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3328580578340034e-06,
+      "loss": 0.6607,
+      "step": 24066
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.332314977935516e-06,
+      "loss": 0.6421,
+      "step": 24067
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3317720008033997e-06,
+      "loss": 0.6593,
+      "step": 24068
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3312291264440914e-06,
+      "loss": 0.5603,
+      "step": 24069
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.330686354864027e-06,
+      "loss": 0.5601,
+      "step": 24070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3301436860696437e-06,
+      "loss": 0.6238,
+      "step": 24071
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3296011200673764e-06,
+      "loss": 0.6351,
+      "step": 24072
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3290586568636533e-06,
+      "loss": 0.6159,
+      "step": 24073
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.328516296464909e-06,
+      "loss": 0.6504,
+      "step": 24074
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.327974038877573e-06,
+      "loss": 0.5933,
+      "step": 24075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3274318841080757e-06,
+      "loss": 0.5915,
+      "step": 24076
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3268898321628433e-06,
+      "loss": 0.6404,
+      "step": 24077
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3263478830483068e-06,
+      "loss": 0.6502,
+      "step": 24078
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.325806036770886e-06,
+      "loss": 0.6536,
+      "step": 24079
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3252642933370063e-06,
+      "loss": 0.6354,
+      "step": 24080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3247226527530932e-06,
+      "loss": 0.641,
+      "step": 24081
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3241811150255668e-06,
+      "loss": 0.6313,
+      "step": 24082
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3236396801608487e-06,
+      "loss": 0.585,
+      "step": 24083
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3230983481653593e-06,
+      "loss": 0.625,
+      "step": 24084
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3225571190455145e-06,
+      "loss": 0.6111,
+      "step": 24085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3220159928077304e-06,
+      "loss": 0.6475,
+      "step": 24086
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.321474969458425e-06,
+      "loss": 0.6458,
+      "step": 24087
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3209340490040135e-06,
+      "loss": 0.6176,
+      "step": 24088
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3203932314509061e-06,
+      "loss": 0.6586,
+      "step": 24089
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3198525168055198e-06,
+      "loss": 0.6087,
+      "step": 24090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3193119050742597e-06,
+      "loss": 0.6198,
+      "step": 24091
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3187713962635384e-06,
+      "loss": 0.6766,
+      "step": 24092
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3182309903797642e-06,
+      "loss": 0.6816,
+      "step": 24093
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3176906874293437e-06,
+      "loss": 0.5963,
+      "step": 24094
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3171504874186837e-06,
+      "loss": 0.6335,
+      "step": 24095
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3166103903541904e-06,
+      "loss": 0.6112,
+      "step": 24096
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3160703962422639e-06,
+      "loss": 0.6264,
+      "step": 24097
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3155305050893075e-06,
+      "loss": 0.6431,
+      "step": 24098
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3149907169017228e-06,
+      "loss": 0.5885,
+      "step": 24099
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3144510316859106e-06,
+      "loss": 0.5922,
+      "step": 24100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3139114494482674e-06,
+      "loss": 0.6423,
+      "step": 24101
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3133719701951953e-06,
+      "loss": 0.6525,
+      "step": 24102
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3128325939330844e-06,
+      "loss": 0.6312,
+      "step": 24103
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.312293320668332e-06,
+      "loss": 0.6247,
+      "step": 24104
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3117541504073328e-06,
+      "loss": 0.6198,
+      "step": 24105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3112150831564786e-06,
+      "loss": 0.6484,
+      "step": 24106
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3106761189221628e-06,
+      "loss": 0.6171,
+      "step": 24107
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.310137257710772e-06,
+      "loss": 0.6745,
+      "step": 24108
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.309598499528697e-06,
+      "loss": 0.6922,
+      "step": 24109
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3090598443823243e-06,
+      "loss": 0.5624,
+      "step": 24110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3085212922780422e-06,
+      "loss": 0.6379,
+      "step": 24111
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.307982843222234e-06,
+      "loss": 0.6354,
+      "step": 24112
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3074444972212885e-06,
+      "loss": 0.6382,
+      "step": 24113
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3069062542815815e-06,
+      "loss": 0.6086,
+      "step": 24114
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3063681144094964e-06,
+      "loss": 0.6237,
+      "step": 24115
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3058300776114163e-06,
+      "loss": 0.6208,
+      "step": 24116
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3052921438937182e-06,
+      "loss": 0.6646,
+      "step": 24117
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3047543132627816e-06,
+      "loss": 0.6303,
+      "step": 24118
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3042165857249834e-06,
+      "loss": 0.6175,
+      "step": 24119
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3036789612866952e-06,
+      "loss": 0.6973,
+      "step": 24120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3031414399542952e-06,
+      "loss": 0.6029,
+      "step": 24121
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.302604021734154e-06,
+      "loss": 0.5859,
+      "step": 24122
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3020667066326442e-06,
+      "loss": 0.601,
+      "step": 24123
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3015294946561373e-06,
+      "loss": 0.6435,
+      "step": 24124
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3009923858110041e-06,
+      "loss": 0.6364,
+      "step": 24125
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.300455380103608e-06,
+      "loss": 0.6487,
+      "step": 24126
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2999184775403173e-06,
+      "loss": 0.605,
+      "step": 24127
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2993816781275003e-06,
+      "loss": 0.6226,
+      "step": 24128
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2988449818715187e-06,
+      "loss": 0.6217,
+      "step": 24129
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2983083887787374e-06,
+      "loss": 0.6662,
+      "step": 24130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2977718988555198e-06,
+      "loss": 0.627,
+      "step": 24131
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2972355121082226e-06,
+      "loss": 0.617,
+      "step": 24132
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.296699228543208e-06,
+      "loss": 0.6768,
+      "step": 24133
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.296163048166833e-06,
+      "loss": 0.6976,
+      "step": 24134
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2956269709854563e-06,
+      "loss": 0.6177,
+      "step": 24135
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2950909970054326e-06,
+      "loss": 0.623,
+      "step": 24136
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.294555126233119e-06,
+      "loss": 0.6216,
+      "step": 24137
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.294019358674864e-06,
+      "loss": 0.5928,
+      "step": 24138
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2934836943370233e-06,
+      "loss": 0.6104,
+      "step": 24139
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2929481332259475e-06,
+      "loss": 0.6164,
+      "step": 24140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.292412675347986e-06,
+      "loss": 0.6356,
+      "step": 24141
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.291877320709487e-06,
+      "loss": 0.6709,
+      "step": 24142
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2913420693168e-06,
+      "loss": 0.6067,
+      "step": 24143
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2908069211762676e-06,
+      "loss": 0.5791,
+      "step": 24144
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2902718762942368e-06,
+      "loss": 0.6342,
+      "step": 24145
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2897369346770494e-06,
+      "loss": 0.6143,
+      "step": 24146
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2892020963310504e-06,
+      "loss": 0.6228,
+      "step": 24147
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2886673612625789e-06,
+      "loss": 0.6321,
+      "step": 24148
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2881327294779788e-06,
+      "loss": 0.7298,
+      "step": 24149
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.287598200983583e-06,
+      "loss": 0.6665,
+      "step": 24150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2870637757857318e-06,
+      "loss": 0.6324,
+      "step": 24151
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2865294538907624e-06,
+      "loss": 0.6768,
+      "step": 24152
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2859952353050075e-06,
+      "loss": 0.6274,
+      "step": 24153
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2854611200348033e-06,
+      "loss": 0.6117,
+      "step": 24154
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2849271080864834e-06,
+      "loss": 0.6248,
+      "step": 24155
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.284393199466375e-06,
+      "loss": 0.6376,
+      "step": 24156
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2838593941808108e-06,
+      "loss": 0.6564,
+      "step": 24157
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2833256922361192e-06,
+      "loss": 0.6375,
+      "step": 24158
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2827920936386273e-06,
+      "loss": 0.6101,
+      "step": 24159
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2822585983946633e-06,
+      "loss": 0.6349,
+      "step": 24160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2817252065105545e-06,
+      "loss": 0.6463,
+      "step": 24161
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2811919179926192e-06,
+      "loss": 0.5903,
+      "step": 24162
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2806587328471832e-06,
+      "loss": 0.6142,
+      "step": 24163
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.280125651080567e-06,
+      "loss": 0.5856,
+      "step": 24164
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2795926726990926e-06,
+      "loss": 0.6336,
+      "step": 24165
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2790597977090792e-06,
+      "loss": 0.6582,
+      "step": 24166
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2785270261168458e-06,
+      "loss": 0.6093,
+      "step": 24167
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2779943579287046e-06,
+      "loss": 0.6256,
+      "step": 24168
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2774617931509758e-06,
+      "loss": 0.6292,
+      "step": 24169
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2769293317899666e-06,
+      "loss": 0.6364,
+      "step": 24170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2763969738519988e-06,
+      "loss": 0.6025,
+      "step": 24171
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2758647193433816e-06,
+      "loss": 0.6516,
+      "step": 24172
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.275332568270421e-06,
+      "loss": 0.6702,
+      "step": 24173
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.274800520639431e-06,
+      "loss": 0.6061,
+      "step": 24174
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2742685764567197e-06,
+      "loss": 0.6196,
+      "step": 24175
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.273736735728589e-06,
+      "loss": 0.7116,
+      "step": 24176
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2732049984613493e-06,
+      "loss": 0.596,
+      "step": 24177
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2726733646613066e-06,
+      "loss": 0.6513,
+      "step": 24178
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.272141834334759e-06,
+      "loss": 0.6008,
+      "step": 24179
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2716104074880098e-06,
+      "loss": 0.6486,
+      "step": 24180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2710790841273647e-06,
+      "loss": 0.6621,
+      "step": 24181
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2705478642591128e-06,
+      "loss": 0.6478,
+      "step": 24182
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.270016747889562e-06,
+      "loss": 0.6252,
+      "step": 24183
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.269485735025009e-06,
+      "loss": 0.596,
+      "step": 24184
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2689548256717433e-06,
+      "loss": 0.6365,
+      "step": 24185
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2684240198360643e-06,
+      "loss": 0.5932,
+      "step": 24186
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2678933175242647e-06,
+      "loss": 0.6631,
+      "step": 24187
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.267362718742633e-06,
+      "loss": 0.6384,
+      "step": 24188
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.266832223497465e-06,
+      "loss": 0.6468,
+      "step": 24189
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2663018317950504e-06,
+      "loss": 0.6469,
+      "step": 24190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2657715436416728e-06,
+      "loss": 0.6464,
+      "step": 24191
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.265241359043624e-06,
+      "loss": 0.6325,
+      "step": 24192
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2647112780071892e-06,
+      "loss": 0.6484,
+      "step": 24193
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2641813005386482e-06,
+      "loss": 0.577,
+      "step": 24194
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2636514266442923e-06,
+      "loss": 0.6816,
+      "step": 24195
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2631216563304015e-06,
+      "loss": 0.6269,
+      "step": 24196
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2625919896032535e-06,
+      "loss": 0.6393,
+      "step": 24197
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2620624264691305e-06,
+      "loss": 0.6259,
+      "step": 24198
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2615329669343123e-06,
+      "loss": 0.6132,
+      "step": 24199
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2610036110050716e-06,
+      "loss": 0.6031,
+      "step": 24200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.26047435868769e-06,
+      "loss": 0.5774,
+      "step": 24201
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2599452099884423e-06,
+      "loss": 0.6257,
+      "step": 24202
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2594161649135982e-06,
+      "loss": 0.6368,
+      "step": 24203
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2588872234694327e-06,
+      "loss": 0.6009,
+      "step": 24204
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2583583856622195e-06,
+      "loss": 0.6253,
+      "step": 24205
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2578296514982203e-06,
+      "loss": 0.5915,
+      "step": 24206
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2573010209837122e-06,
+      "loss": 0.6131,
+      "step": 24207
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2567724941249627e-06,
+      "loss": 0.5988,
+      "step": 24208
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.256244070928232e-06,
+      "loss": 0.5841,
+      "step": 24209
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2557157513997898e-06,
+      "loss": 0.5967,
+      "step": 24210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.255187535545901e-06,
+      "loss": 0.6536,
+      "step": 24211
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2546594233728215e-06,
+      "loss": 0.6766,
+      "step": 24212
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.25413141488682e-06,
+      "loss": 0.6127,
+      "step": 24213
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2536035100941568e-06,
+      "loss": 0.6303,
+      "step": 24214
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2530757090010848e-06,
+      "loss": 0.6264,
+      "step": 24215
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2525480116138667e-06,
+      "loss": 0.6147,
+      "step": 24216
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2520204179387585e-06,
+      "loss": 0.6423,
+      "step": 24217
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.251492927982011e-06,
+      "loss": 0.5567,
+      "step": 24218
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2509655417498834e-06,
+      "loss": 0.5954,
+      "step": 24219
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2504382592486286e-06,
+      "loss": 0.6627,
+      "step": 24220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2499110804844949e-06,
+      "loss": 0.6115,
+      "step": 24221
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.249384005463734e-06,
+      "loss": 0.6733,
+      "step": 24222
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2488570341925977e-06,
+      "loss": 0.6746,
+      "step": 24223
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2483301666773273e-06,
+      "loss": 0.683,
+      "step": 24224
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2478034029241758e-06,
+      "loss": 0.6071,
+      "step": 24225
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2472767429393884e-06,
+      "loss": 0.6115,
+      "step": 24226
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2467501867292054e-06,
+      "loss": 0.6057,
+      "step": 24227
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2462237342998718e-06,
+      "loss": 0.6569,
+      "step": 24228
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2456973856576327e-06,
+      "loss": 0.631,
+      "step": 24229
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2451711408087196e-06,
+      "loss": 0.7085,
+      "step": 24230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2446449997593824e-06,
+      "loss": 0.6308,
+      "step": 24231
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2441189625158522e-06,
+      "loss": 0.6157,
+      "step": 24232
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.243593029084368e-06,
+      "loss": 0.6364,
+      "step": 24233
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2430671994711662e-06,
+      "loss": 0.6784,
+      "step": 24234
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2425414736824804e-06,
+      "loss": 0.6472,
+      "step": 24235
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2420158517245418e-06,
+      "loss": 0.617,
+      "step": 24236
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2414903336035856e-06,
+      "loss": 0.6311,
+      "step": 24237
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.24096491932584e-06,
+      "loss": 0.6306,
+      "step": 24238
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2404396088975357e-06,
+      "loss": 0.6325,
+      "step": 24239
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2399144023249033e-06,
+      "loss": 0.6656,
+      "step": 24240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2393892996141654e-06,
+      "loss": 0.6339,
+      "step": 24241
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2388643007715483e-06,
+      "loss": 0.6261,
+      "step": 24242
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2383394058032794e-06,
+      "loss": 0.6238,
+      "step": 24243
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.237814614715579e-06,
+      "loss": 0.6203,
+      "step": 24244
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.237289927514671e-06,
+      "loss": 0.5981,
+      "step": 24245
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2367653442067783e-06,
+      "loss": 0.6007,
+      "step": 24246
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.236240864798116e-06,
+      "loss": 0.6212,
+      "step": 24247
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2357164892949037e-06,
+      "loss": 0.6594,
+      "step": 24248
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2351922177033604e-06,
+      "loss": 0.6131,
+      "step": 24249
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2346680500297003e-06,
+      "loss": 0.6472,
+      "step": 24250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2341439862801386e-06,
+      "loss": 0.63,
+      "step": 24251
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.23362002646089e-06,
+      "loss": 0.6635,
+      "step": 24252
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2330961705781652e-06,
+      "loss": 0.6523,
+      "step": 24253
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2325724186381739e-06,
+      "loss": 0.5973,
+      "step": 24254
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2320487706471284e-06,
+      "loss": 0.6923,
+      "step": 24255
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2315252266112355e-06,
+      "loss": 0.6819,
+      "step": 24256
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.231001786536704e-06,
+      "loss": 0.693,
+      "step": 24257
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2304784504297408e-06,
+      "loss": 0.6584,
+      "step": 24258
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.229955218296548e-06,
+      "loss": 0.6664,
+      "step": 24259
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2294320901433288e-06,
+      "loss": 0.6112,
+      "step": 24260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.228909065976287e-06,
+      "loss": 0.6616,
+      "step": 24261
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2283861458016244e-06,
+      "loss": 0.6654,
+      "step": 24262
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2278633296255404e-06,
+      "loss": 0.666,
+      "step": 24263
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2273406174542358e-06,
+      "loss": 0.6592,
+      "step": 24264
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2268180092939019e-06,
+      "loss": 0.6144,
+      "step": 24265
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2262955051507398e-06,
+      "loss": 0.69,
+      "step": 24266
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.225773105030944e-06,
+      "loss": 0.6414,
+      "step": 24267
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2252508089407068e-06,
+      "loss": 0.6785,
+      "step": 24268
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2247286168862204e-06,
+      "loss": 0.6326,
+      "step": 24269
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2242065288736804e-06,
+      "loss": 0.6504,
+      "step": 24270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2236845449092705e-06,
+      "loss": 0.6473,
+      "step": 24271
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2231626649991834e-06,
+      "loss": 0.6081,
+      "step": 24272
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2226408891496044e-06,
+      "loss": 0.6392,
+      "step": 24273
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2221192173667217e-06,
+      "loss": 0.5665,
+      "step": 24274
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2215976496567207e-06,
+      "loss": 0.6779,
+      "step": 24275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2210761860257848e-06,
+      "loss": 0.6684,
+      "step": 24276
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2205548264800947e-06,
+      "loss": 0.6135,
+      "step": 24277
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2200335710258337e-06,
+      "loss": 0.6365,
+      "step": 24278
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.219512419669181e-06,
+      "loss": 0.6069,
+      "step": 24279
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.218991372416316e-06,
+      "loss": 0.6627,
+      "step": 24280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2184704292734161e-06,
+      "loss": 0.6574,
+      "step": 24281
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2179495902466597e-06,
+      "loss": 0.6284,
+      "step": 24282
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2174288553422186e-06,
+      "loss": 0.5837,
+      "step": 24283
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.216908224566269e-06,
+      "loss": 0.6678,
+      "step": 24284
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2163876979249823e-06,
+      "loss": 0.6662,
+      "step": 24285
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2158672754245304e-06,
+      "loss": 0.6495,
+      "step": 24286
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2153469570710851e-06,
+      "loss": 0.6541,
+      "step": 24287
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.214826742870816e-06,
+      "loss": 0.6845,
+      "step": 24288
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.214306632829887e-06,
+      "loss": 0.5556,
+      "step": 24289
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2137866269544663e-06,
+      "loss": 0.6915,
+      "step": 24290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2132667252507203e-06,
+      "loss": 0.6715,
+      "step": 24291
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2127469277248115e-06,
+      "loss": 0.627,
+      "step": 24292
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2122272343829077e-06,
+      "loss": 0.6065,
+      "step": 24293
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2117076452311628e-06,
+      "loss": 0.6369,
+      "step": 24294
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2111881602757424e-06,
+      "loss": 0.6591,
+      "step": 24295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2106687795228034e-06,
+      "loss": 0.6544,
+      "step": 24296
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2101495029785048e-06,
+      "loss": 0.6245,
+      "step": 24297
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2096303306490031e-06,
+      "loss": 0.682,
+      "step": 24298
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2091112625404555e-06,
+      "loss": 0.6259,
+      "step": 24299
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2085922986590126e-06,
+      "loss": 0.6446,
+      "step": 24300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2080734390108296e-06,
+      "loss": 0.6684,
+      "step": 24301
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.207554683602058e-06,
+      "loss": 0.6322,
+      "step": 24302
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2070360324388474e-06,
+      "loss": 0.628,
+      "step": 24303
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2065174855273477e-06,
+      "loss": 0.6356,
+      "step": 24304
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2059990428737101e-06,
+      "loss": 0.621,
+      "step": 24305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2054807044840745e-06,
+      "loss": 0.6601,
+      "step": 24306
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2049624703645913e-06,
+      "loss": 0.6667,
+      "step": 24307
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2044443405214047e-06,
+      "loss": 0.553,
+      "step": 24308
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.203926314960655e-06,
+      "loss": 0.5913,
+      "step": 24309
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2034083936884877e-06,
+      "loss": 0.6194,
+      "step": 24310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.202890576711042e-06,
+      "loss": 0.6099,
+      "step": 24311
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2023728640344568e-06,
+      "loss": 0.6694,
+      "step": 24312
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2018552556648689e-06,
+      "loss": 0.6734,
+      "step": 24313
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2013377516084169e-06,
+      "loss": 0.6533,
+      "step": 24314
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.200820351871237e-06,
+      "loss": 0.6292,
+      "step": 24315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.200303056459462e-06,
+      "loss": 0.6261,
+      "step": 24316
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1997858653792283e-06,
+      "loss": 0.6386,
+      "step": 24317
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.199268778636663e-06,
+      "loss": 0.6302,
+      "step": 24318
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.198751796237899e-06,
+      "loss": 0.6069,
+      "step": 24319
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1982349181890673e-06,
+      "loss": 0.652,
+      "step": 24320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1977181444962948e-06,
+      "loss": 0.6637,
+      "step": 24321
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.197201475165708e-06,
+      "loss": 0.6174,
+      "step": 24322
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.196684910203435e-06,
+      "loss": 0.6737,
+      "step": 24323
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1961684496155978e-06,
+      "loss": 0.584,
+      "step": 24324
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1956520934083192e-06,
+      "loss": 0.6653,
+      "step": 24325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1951358415877234e-06,
+      "loss": 0.5839,
+      "step": 24326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1946196941599309e-06,
+      "loss": 0.6845,
+      "step": 24327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.19410365113106e-06,
+      "loss": 0.6328,
+      "step": 24328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1935877125072325e-06,
+      "loss": 0.672,
+      "step": 24329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1930718782945605e-06,
+      "loss": 0.5969,
+      "step": 24330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1925561484991622e-06,
+      "loss": 0.6471,
+      "step": 24331
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1920405231271514e-06,
+      "loss": 0.5968,
+      "step": 24332
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1915250021846437e-06,
+      "loss": 0.5953,
+      "step": 24333
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1910095856777492e-06,
+      "loss": 0.6057,
+      "step": 24334
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1904942736125814e-06,
+      "loss": 0.5979,
+      "step": 24335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.189979065995246e-06,
+      "loss": 0.64,
+      "step": 24336
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1894639628318538e-06,
+      "loss": 0.7045,
+      "step": 24337
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1889489641285113e-06,
+      "loss": 0.62,
+      "step": 24338
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1884340698913255e-06,
+      "loss": 0.6318,
+      "step": 24339
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1879192801264006e-06,
+      "loss": 0.6793,
+      "step": 24340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1874045948398416e-06,
+      "loss": 0.6151,
+      "step": 24341
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.186890014037747e-06,
+      "loss": 0.6295,
+      "step": 24342
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1863755377262199e-06,
+      "loss": 0.6488,
+      "step": 24343
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1858611659113594e-06,
+      "loss": 0.6703,
+      "step": 24344
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1853468985992656e-06,
+      "loss": 0.667,
+      "step": 24345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1848327357960353e-06,
+      "loss": 0.6174,
+      "step": 24346
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1843186775077653e-06,
+      "loss": 0.6007,
+      "step": 24347
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.183804723740548e-06,
+      "loss": 0.6228,
+      "step": 24348
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1832908745004778e-06,
+      "loss": 0.7143,
+      "step": 24349
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1827771297936475e-06,
+      "loss": 0.6188,
+      "step": 24350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1822634896261487e-06,
+      "loss": 0.6232,
+      "step": 24351
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.18174995400407e-06,
+      "loss": 0.607,
+      "step": 24352
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1812365229335033e-06,
+      "loss": 0.6471,
+      "step": 24353
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.18072319642053e-06,
+      "loss": 0.6328,
+      "step": 24354
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1802099744712414e-06,
+      "loss": 0.6408,
+      "step": 24355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.17969685709172e-06,
+      "loss": 0.619,
+      "step": 24356
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.17918384428805e-06,
+      "loss": 0.6372,
+      "step": 24357
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1786709360663162e-06,
+      "loss": 0.6508,
+      "step": 24358
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1781581324325954e-06,
+      "loss": 0.6657,
+      "step": 24359
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1776454333929698e-06,
+      "loss": 0.5927,
+      "step": 24360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1771328389535175e-06,
+      "loss": 0.639,
+      "step": 24361
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1766203491203165e-06,
+      "loss": 0.6453,
+      "step": 24362
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1761079638994434e-06,
+      "loss": 0.6123,
+      "step": 24363
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.175595683296974e-06,
+      "loss": 0.6614,
+      "step": 24364
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1750835073189782e-06,
+      "loss": 0.6309,
+      "step": 24365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1745714359715322e-06,
+      "loss": 0.6478,
+      "step": 24366
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1740594692607055e-06,
+      "loss": 0.641,
+      "step": 24367
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1735476071925688e-06,
+      "loss": 0.6949,
+      "step": 24368
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.173035849773191e-06,
+      "loss": 0.5463,
+      "step": 24369
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.172524197008641e-06,
+      "loss": 0.7055,
+      "step": 24370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1720126489049811e-06,
+      "loss": 0.5965,
+      "step": 24371
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1715012054682796e-06,
+      "loss": 0.6194,
+      "step": 24372
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1709898667045982e-06,
+      "loss": 0.686,
+      "step": 24373
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1704786326200013e-06,
+      "loss": 0.6096,
+      "step": 24374
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1699675032205504e-06,
+      "loss": 0.6329,
+      "step": 24375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1694564785123064e-06,
+      "loss": 0.6715,
+      "step": 24376
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1689455585013243e-06,
+      "loss": 0.6317,
+      "step": 24377
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1684347431936637e-06,
+      "loss": 0.6296,
+      "step": 24378
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1679240325953812e-06,
+      "loss": 0.6317,
+      "step": 24379
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1674134267125314e-06,
+      "loss": 0.6614,
+      "step": 24380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.16690292555117e-06,
+      "loss": 0.5873,
+      "step": 24381
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1663925291173495e-06,
+      "loss": 0.6043,
+      "step": 24382
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1658822374171186e-06,
+      "loss": 0.6005,
+      "step": 24383
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.165372050456529e-06,
+      "loss": 0.6612,
+      "step": 24384
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1648619682416297e-06,
+      "loss": 0.6257,
+      "step": 24385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1643519907784673e-06,
+      "loss": 0.5986,
+      "step": 24386
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1638421180730908e-06,
+      "loss": 0.6452,
+      "step": 24387
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1633323501315442e-06,
+      "loss": 0.6528,
+      "step": 24388
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1628226869598703e-06,
+      "loss": 0.657,
+      "step": 24389
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1623131285641121e-06,
+      "loss": 0.6099,
+      "step": 24390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1618036749503115e-06,
+      "loss": 0.648,
+      "step": 24391
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.161294326124508e-06,
+      "loss": 0.664,
+      "step": 24392
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1607850820927413e-06,
+      "loss": 0.6387,
+      "step": 24393
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.160275942861052e-06,
+      "loss": 0.6728,
+      "step": 24394
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1597669084354712e-06,
+      "loss": 0.6259,
+      "step": 24395
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.159257978822037e-06,
+      "loss": 0.5875,
+      "step": 24396
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1587491540267836e-06,
+      "loss": 0.6532,
+      "step": 24397
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1582404340557419e-06,
+      "loss": 0.6229,
+      "step": 24398
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1577318189149457e-06,
+      "loss": 0.5809,
+      "step": 24399
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1572233086104257e-06,
+      "loss": 0.6778,
+      "step": 24400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1567149031482084e-06,
+      "loss": 0.6216,
+      "step": 24401
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1562066025343222e-06,
+      "loss": 0.6753,
+      "step": 24402
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1556984067747935e-06,
+      "loss": 0.6153,
+      "step": 24403
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1551903158756494e-06,
+      "loss": 0.6401,
+      "step": 24404
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1546823298429122e-06,
+      "loss": 0.5983,
+      "step": 24405
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1541744486826067e-06,
+      "loss": 0.6501,
+      "step": 24406
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1536666724007528e-06,
+      "loss": 0.6371,
+      "step": 24407
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.15315900100337e-06,
+      "loss": 0.6045,
+      "step": 24408
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.152651434496479e-06,
+      "loss": 0.6419,
+      "step": 24409
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1521439728860973e-06,
+      "loss": 0.5726,
+      "step": 24410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.151636616178241e-06,
+      "loss": 0.6286,
+      "step": 24411
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.151129364378928e-06,
+      "loss": 0.5647,
+      "step": 24412
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1506222174941684e-06,
+      "loss": 0.6896,
+      "step": 24413
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1501151755299765e-06,
+      "loss": 0.606,
+      "step": 24414
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1496082384923646e-06,
+      "loss": 0.6495,
+      "step": 24415
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1491014063873428e-06,
+      "loss": 0.65,
+      "step": 24416
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1485946792209225e-06,
+      "loss": 0.6386,
+      "step": 24417
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.148088056999107e-06,
+      "loss": 0.6278,
+      "step": 24418
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1475815397279067e-06,
+      "loss": 0.5798,
+      "step": 24419
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1470751274133263e-06,
+      "loss": 0.6338,
+      "step": 24420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1465688200613656e-06,
+      "loss": 0.6448,
+      "step": 24421
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.146062617678032e-06,
+      "loss": 0.6705,
+      "step": 24422
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1455565202693297e-06,
+      "loss": 0.6488,
+      "step": 24423
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.145050527841254e-06,
+      "loss": 0.5875,
+      "step": 24424
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.144544640399805e-06,
+      "loss": 0.6534,
+      "step": 24425
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1440388579509842e-06,
+      "loss": 0.6113,
+      "step": 24426
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1435331805007833e-06,
+      "loss": 0.603,
+      "step": 24427
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1430276080551972e-06,
+      "loss": 0.615,
+      "step": 24428
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.142522140620228e-06,
+      "loss": 0.6394,
+      "step": 24429
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.142016778201861e-06,
+      "loss": 0.6682,
+      "step": 24430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1415115208060913e-06,
+      "loss": 0.6223,
+      "step": 24431
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1410063684389106e-06,
+      "loss": 0.7128,
+      "step": 24432
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1405013211063032e-06,
+      "loss": 0.6352,
+      "step": 24433
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1399963788142575e-06,
+      "loss": 0.6275,
+      "step": 24434
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.139491541568768e-06,
+      "loss": 0.5998,
+      "step": 24435
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1389868093758128e-06,
+      "loss": 0.7071,
+      "step": 24436
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1384821822413772e-06,
+      "loss": 0.6259,
+      "step": 24437
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1379776601714475e-06,
+      "loss": 0.6759,
+      "step": 24438
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1374732431720027e-06,
+      "loss": 0.636,
+      "step": 24439
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1369689312490196e-06,
+      "loss": 0.622,
+      "step": 24440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.136464724408487e-06,
+      "loss": 0.6424,
+      "step": 24441
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1359606226563746e-06,
+      "loss": 0.6208,
+      "step": 24442
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1354566259986631e-06,
+      "loss": 0.6118,
+      "step": 24443
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.134952734441328e-06,
+      "loss": 0.6376,
+      "step": 24444
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1344489479903409e-06,
+      "loss": 0.6412,
+      "step": 24445
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.133945266651675e-06,
+      "loss": 0.647,
+      "step": 24446
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1334416904313073e-06,
+      "loss": 0.5926,
+      "step": 24447
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1329382193352023e-06,
+      "loss": 0.6966,
+      "step": 24448
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1324348533693318e-06,
+      "loss": 0.6685,
+      "step": 24449
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1319315925396656e-06,
+      "loss": 0.5982,
+      "step": 24450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1314284368521667e-06,
+      "loss": 0.6162,
+      "step": 24451
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1309253863128e-06,
+      "loss": 0.6702,
+      "step": 24452
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1304224409275365e-06,
+      "loss": 0.6361,
+      "step": 24453
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1299196007023338e-06,
+      "loss": 0.5743,
+      "step": 24454
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1294168656431536e-06,
+      "loss": 0.6174,
+      "step": 24455
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1289142357559612e-06,
+      "loss": 0.6173,
+      "step": 24456
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1284117110467097e-06,
+      "loss": 0.6423,
+      "step": 24457
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1279092915213585e-06,
+      "loss": 0.6446,
+      "step": 24458
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1274069771858697e-06,
+      "loss": 0.6442,
+      "step": 24459
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.126904768046192e-06,
+      "loss": 0.6153,
+      "step": 24460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1264026641082837e-06,
+      "loss": 0.6171,
+      "step": 24461
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.125900665378099e-06,
+      "loss": 0.5959,
+      "step": 24462
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1253987718615866e-06,
+      "loss": 0.628,
+      "step": 24463
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.124896983564695e-06,
+      "loss": 0.5986,
+      "step": 24464
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1243953004933805e-06,
+      "loss": 0.688,
+      "step": 24465
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1238937226535861e-06,
+      "loss": 0.6376,
+      "step": 24466
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1233922500512594e-06,
+      "loss": 0.5917,
+      "step": 24467
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.122890882692349e-06,
+      "loss": 0.6242,
+      "step": 24468
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1223896205827956e-06,
+      "loss": 0.6678,
+      "step": 24469
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1218884637285398e-06,
+      "loss": 0.6484,
+      "step": 24470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1213874121355327e-06,
+      "loss": 0.6509,
+      "step": 24471
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1208864658097062e-06,
+      "loss": 0.6309,
+      "step": 24472
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1203856247570033e-06,
+      "loss": 0.6094,
+      "step": 24473
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1198848889833636e-06,
+      "loss": 0.646,
+      "step": 24474
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.119384258494719e-06,
+      "loss": 0.7011,
+      "step": 24475
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.118883733297007e-06,
+      "loss": 0.6119,
+      "step": 24476
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1183833133961674e-06,
+      "loss": 0.5991,
+      "step": 24477
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1178829987981256e-06,
+      "loss": 0.6388,
+      "step": 24478
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1173827895088197e-06,
+      "loss": 0.6033,
+      "step": 24479
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1168826855341752e-06,
+      "loss": 0.6246,
+      "step": 24480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.116382686880124e-06,
+      "loss": 0.6179,
+      "step": 24481
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.115882793552594e-06,
+      "loss": 0.6346,
+      "step": 24482
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.115383005557511e-06,
+      "loss": 0.6547,
+      "step": 24483
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1148833229008028e-06,
+      "loss": 0.6308,
+      "step": 24484
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1143837455883932e-06,
+      "loss": 0.6337,
+      "step": 24485
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1138842736262035e-06,
+      "loss": 0.6816,
+      "step": 24486
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1133849070201564e-06,
+      "loss": 0.6544,
+      "step": 24487
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1128856457761728e-06,
+      "loss": 0.599,
+      "step": 24488
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1123864899001723e-06,
+      "loss": 0.6351,
+      "step": 24489
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1118874393980728e-06,
+      "loss": 0.6357,
+      "step": 24490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1113884942757935e-06,
+      "loss": 0.6294,
+      "step": 24491
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1108896545392455e-06,
+      "loss": 0.6272,
+      "step": 24492
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1103909201943452e-06,
+      "loss": 0.6511,
+      "step": 24493
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1098922912470067e-06,
+      "loss": 0.622,
+      "step": 24494
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1093937677031408e-06,
+      "loss": 0.6951,
+      "step": 24495
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1088953495686584e-06,
+      "loss": 0.6238,
+      "step": 24496
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1083970368494702e-06,
+      "loss": 0.6061,
+      "step": 24497
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1078988295514825e-06,
+      "loss": 0.6148,
+      "step": 24498
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1074007276806009e-06,
+      "loss": 0.6568,
+      "step": 24499
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1069027312427349e-06,
+      "loss": 0.6871,
+      "step": 24500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1064048402437855e-06,
+      "loss": 0.6704,
+      "step": 24501
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1059070546896577e-06,
+      "loss": 0.5918,
+      "step": 24502
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.105409374586255e-06,
+      "loss": 0.6152,
+      "step": 24503
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1049117999394732e-06,
+      "loss": 0.5823,
+      "step": 24504
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1044143307552146e-06,
+      "loss": 0.6627,
+      "step": 24505
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1039169670393768e-06,
+      "loss": 0.6319,
+      "step": 24506
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1034197087978583e-06,
+      "loss": 0.6392,
+      "step": 24507
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.102922556036552e-06,
+      "loss": 0.6188,
+      "step": 24508
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1024255087613566e-06,
+      "loss": 0.6802,
+      "step": 24509
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1019285669781598e-06,
+      "loss": 0.5946,
+      "step": 24510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1014317306928557e-06,
+      "loss": 0.6112,
+      "step": 24511
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.100934999911335e-06,
+      "loss": 0.594,
+      "step": 24512
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1004383746394876e-06,
+      "loss": 0.6558,
+      "step": 24513
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.099941854883202e-06,
+      "loss": 0.6509,
+      "step": 24514
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0994454406483657e-06,
+      "loss": 0.6438,
+      "step": 24515
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0989491319408607e-06,
+      "loss": 0.6633,
+      "step": 24516
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0984529287665736e-06,
+      "loss": 0.6165,
+      "step": 24517
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0979568311313883e-06,
+      "loss": 0.6392,
+      "step": 24518
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0974608390411857e-06,
+      "loss": 0.6905,
+      "step": 24519
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0969649525018456e-06,
+      "loss": 0.5623,
+      "step": 24520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.096469171519252e-06,
+      "loss": 0.609,
+      "step": 24521
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0959734960992762e-06,
+      "loss": 0.6139,
+      "step": 24522
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0954779262477988e-06,
+      "loss": 0.5772,
+      "step": 24523
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.094982461970695e-06,
+      "loss": 0.6264,
+      "step": 24524
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.094487103273838e-06,
+      "loss": 0.7355,
+      "step": 24525
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.093991850163103e-06,
+      "loss": 0.6183,
+      "step": 24526
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0934967026443622e-06,
+      "loss": 0.6673,
+      "step": 24527
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0930016607234829e-06,
+      "loss": 0.6095,
+      "step": 24528
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.092506724406337e-06,
+      "loss": 0.6568,
+      "step": 24529
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.092011893698791e-06,
+      "loss": 0.5843,
+      "step": 24530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0915171686067128e-06,
+      "loss": 0.6277,
+      "step": 24531
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0910225491359682e-06,
+      "loss": 0.5909,
+      "step": 24532
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.090528035292423e-06,
+      "loss": 0.5951,
+      "step": 24533
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0900336270819367e-06,
+      "loss": 0.6377,
+      "step": 24534
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0895393245103725e-06,
+      "loss": 0.6266,
+      "step": 24535
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.089045127583591e-06,
+      "loss": 0.6611,
+      "step": 24536
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0885510363074536e-06,
+      "loss": 0.6333,
+      "step": 24537
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0880570506878152e-06,
+      "loss": 0.529,
+      "step": 24538
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0875631707305367e-06,
+      "loss": 0.6005,
+      "step": 24539
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0870693964414692e-06,
+      "loss": 0.6401,
+      "step": 24540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0865757278264687e-06,
+      "loss": 0.6387,
+      "step": 24541
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0860821648913888e-06,
+      "loss": 0.6824,
+      "step": 24542
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.08558870764208e-06,
+      "loss": 0.6772,
+      "step": 24543
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.085095356084397e-06,
+      "loss": 0.6862,
+      "step": 24544
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0846021102241834e-06,
+      "loss": 0.653,
+      "step": 24545
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0841089700672903e-06,
+      "loss": 0.6314,
+      "step": 24546
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.083615935619563e-06,
+      "loss": 0.6826,
+      "step": 24547
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0831230068868493e-06,
+      "loss": 0.6069,
+      "step": 24548
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.082630183874992e-06,
+      "loss": 0.6578,
+      "step": 24549
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0821374665898354e-06,
+      "loss": 0.6193,
+      "step": 24550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0816448550372194e-06,
+      "loss": 0.645,
+      "step": 24551
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0811523492229848e-06,
+      "loss": 0.6194,
+      "step": 24552
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0806599491529711e-06,
+      "loss": 0.6536,
+      "step": 24553
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0801676548330175e-06,
+      "loss": 0.6539,
+      "step": 24554
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0796754662689602e-06,
+      "loss": 0.6531,
+      "step": 24555
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0791833834666365e-06,
+      "loss": 0.6559,
+      "step": 24556
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0786914064318766e-06,
+      "loss": 0.5959,
+      "step": 24557
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0781995351705143e-06,
+      "loss": 0.5987,
+      "step": 24558
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0777077696883843e-06,
+      "loss": 0.5774,
+      "step": 24559
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0772161099913148e-06,
+      "loss": 0.5961,
+      "step": 24560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0767245560851358e-06,
+      "loss": 0.6437,
+      "step": 24561
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0762331079756772e-06,
+      "loss": 0.6177,
+      "step": 24562
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0757417656687619e-06,
+      "loss": 0.6387,
+      "step": 24563
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0752505291702165e-06,
+      "loss": 0.6174,
+      "step": 24564
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.074759398485866e-06,
+      "loss": 0.6329,
+      "step": 24565
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0742683736215343e-06,
+      "loss": 0.6712,
+      "step": 24566
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0737774545830415e-06,
+      "loss": 0.594,
+      "step": 24567
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.073286641376211e-06,
+      "loss": 0.6578,
+      "step": 24568
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0727959340068571e-06,
+      "loss": 0.6227,
+      "step": 24569
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0723053324808008e-06,
+      "loss": 0.6311,
+      "step": 24570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0718148368038583e-06,
+      "loss": 0.5826,
+      "step": 24571
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0713244469818452e-06,
+      "loss": 0.6384,
+      "step": 24572
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0708341630205754e-06,
+      "loss": 0.6424,
+      "step": 24573
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0703439849258646e-06,
+      "loss": 0.6213,
+      "step": 24574
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0698539127035191e-06,
+      "loss": 0.6498,
+      "step": 24575
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.069363946359353e-06,
+      "loss": 0.6561,
+      "step": 24576
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0688740858991741e-06,
+      "loss": 0.6696,
+      "step": 24577
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0683843313287911e-06,
+      "loss": 0.655,
+      "step": 24578
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0678946826540104e-06,
+      "loss": 0.6576,
+      "step": 24579
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0674051398806395e-06,
+      "loss": 0.6544,
+      "step": 24580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0669157030144773e-06,
+      "loss": 0.6531,
+      "step": 24581
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0664263720613311e-06,
+      "loss": 0.5799,
+      "step": 24582
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0659371470270009e-06,
+      "loss": 0.6567,
+      "step": 24583
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0654480279172874e-06,
+      "loss": 0.6409,
+      "step": 24584
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0649590147379907e-06,
+      "loss": 0.6347,
+      "step": 24585
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0644701074949082e-06,
+      "loss": 0.6511,
+      "step": 24586
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0639813061938353e-06,
+      "loss": 0.6914,
+      "step": 24587
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0634926108405675e-06,
+      "loss": 0.5978,
+      "step": 24588
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0630040214408998e-06,
+      "loss": 0.6573,
+      "step": 24589
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0625155380006247e-06,
+      "loss": 0.614,
+      "step": 24590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.062027160525534e-06,
+      "loss": 0.6257,
+      "step": 24591
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0615388890214207e-06,
+      "loss": 0.6225,
+      "step": 24592
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0610507234940681e-06,
+      "loss": 0.6153,
+      "step": 24593
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0605626639492671e-06,
+      "loss": 0.6396,
+      "step": 24594
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0600747103928043e-06,
+      "loss": 0.6472,
+      "step": 24595
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.059586862830464e-06,
+      "loss": 0.6471,
+      "step": 24596
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0590991212680324e-06,
+      "loss": 0.6259,
+      "step": 24597
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0586114857112927e-06,
+      "loss": 0.6876,
+      "step": 24598
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0581239561660218e-06,
+      "loss": 0.6027,
+      "step": 24599
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0576365326380045e-06,
+      "loss": 0.6007,
+      "step": 24600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0571492151330165e-06,
+      "loss": 0.6338,
+      "step": 24601
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0566620036568386e-06,
+      "loss": 0.6256,
+      "step": 24602
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0561748982152475e-06,
+      "loss": 0.6481,
+      "step": 24603
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.055687898814015e-06,
+      "loss": 0.6068,
+      "step": 24604
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0552010054589178e-06,
+      "loss": 0.5791,
+      "step": 24605
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0547142181557267e-06,
+      "loss": 0.6076,
+      "step": 24606
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0542275369102163e-06,
+      "loss": 0.6163,
+      "step": 24607
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0537409617281536e-06,
+      "loss": 0.6922,
+      "step": 24608
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.05325449261531e-06,
+      "loss": 0.6833,
+      "step": 24609
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0527681295774516e-06,
+      "loss": 0.6535,
+      "step": 24610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0522818726203442e-06,
+      "loss": 0.6351,
+      "step": 24611
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0517957217497553e-06,
+      "loss": 0.5835,
+      "step": 24612
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.051309676971447e-06,
+      "loss": 0.6092,
+      "step": 24613
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0508237382911823e-06,
+      "loss": 0.6421,
+      "step": 24614
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0503379057147257e-06,
+      "loss": 0.6326,
+      "step": 24615
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0498521792478323e-06,
+      "loss": 0.6562,
+      "step": 24616
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0493665588962632e-06,
+      "loss": 0.6085,
+      "step": 24617
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.048881044665776e-06,
+      "loss": 0.6981,
+      "step": 24618
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0483956365621273e-06,
+      "loss": 0.6968,
+      "step": 24619
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0479103345910725e-06,
+      "loss": 0.6678,
+      "step": 24620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0474251387583666e-06,
+      "loss": 0.5955,
+      "step": 24621
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.046940049069759e-06,
+      "loss": 0.597,
+      "step": 24622
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0464550655310035e-06,
+      "loss": 0.6051,
+      "step": 24623
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0459701881478501e-06,
+      "loss": 0.6117,
+      "step": 24624
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0454854169260442e-06,
+      "loss": 0.6265,
+      "step": 24625
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.045000751871338e-06,
+      "loss": 0.6347,
+      "step": 24626
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0445161929894776e-06,
+      "loss": 0.6262,
+      "step": 24627
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0440317402862044e-06,
+      "loss": 0.6042,
+      "step": 24628
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0435473937672646e-06,
+      "loss": 0.6401,
+      "step": 24629
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0430631534384027e-06,
+      "loss": 0.662,
+      "step": 24630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.042579019305353e-06,
+      "loss": 0.635,
+      "step": 24631
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.042094991373863e-06,
+      "loss": 0.5887,
+      "step": 24632
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0416110696496695e-06,
+      "loss": 0.5704,
+      "step": 24633
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0411272541385088e-06,
+      "loss": 0.6269,
+      "step": 24634
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.040643544846116e-06,
+      "loss": 0.6071,
+      "step": 24635
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0401599417782304e-06,
+      "loss": 0.6603,
+      "step": 24636
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0396764449405793e-06,
+      "loss": 0.6782,
+      "step": 24637
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0391930543389006e-06,
+      "loss": 0.6556,
+      "step": 24638
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.038709769978925e-06,
+      "loss": 0.6953,
+      "step": 24639
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0382265918663803e-06,
+      "loss": 0.6648,
+      "step": 24640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0377435200069952e-06,
+      "loss": 0.6035,
+      "step": 24641
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0372605544064994e-06,
+      "loss": 0.587,
+      "step": 24642
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0367776950706155e-06,
+      "loss": 0.7101,
+      "step": 24643
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0362949420050716e-06,
+      "loss": 0.5753,
+      "step": 24644
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0358122952155914e-06,
+      "loss": 0.578,
+      "step": 24645
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0353297547078955e-06,
+      "loss": 0.6242,
+      "step": 24646
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0348473204877052e-06,
+      "loss": 0.5931,
+      "step": 24647
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0343649925607424e-06,
+      "loss": 0.6402,
+      "step": 24648
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0338827709327215e-06,
+      "loss": 0.6013,
+      "step": 24649
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0334006556093634e-06,
+      "loss": 0.6386,
+      "step": 24650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.032918646596386e-06,
+      "loss": 0.603,
+      "step": 24651
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0324367438994986e-06,
+      "loss": 0.6616,
+      "step": 24652
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0319549475244184e-06,
+      "loss": 0.6507,
+      "step": 24653
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0314732574768593e-06,
+      "loss": 0.5625,
+      "step": 24654
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0309916737625248e-06,
+      "loss": 0.6469,
+      "step": 24655
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0305101963871333e-06,
+      "loss": 0.5889,
+      "step": 24656
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0300288253563905e-06,
+      "loss": 0.6173,
+      "step": 24657
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.029547560676002e-06,
+      "loss": 0.6953,
+      "step": 24658
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0290664023516738e-06,
+      "loss": 0.6424,
+      "step": 24659
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0285853503891153e-06,
+      "loss": 0.602,
+      "step": 24660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0281044047940214e-06,
+      "loss": 0.5939,
+      "step": 24661
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0276235655721012e-06,
+      "loss": 0.6664,
+      "step": 24662
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0271428327290555e-06,
+      "loss": 0.6034,
+      "step": 24663
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.026662206270581e-06,
+      "loss": 0.6349,
+      "step": 24664
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0261816862023787e-06,
+      "loss": 0.6811,
+      "step": 24665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.025701272530143e-06,
+      "loss": 0.6411,
+      "step": 24666
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.025220965259569e-06,
+      "loss": 0.5957,
+      "step": 24667
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.024740764396357e-06,
+      "loss": 0.6948,
+      "step": 24668
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0242606699461966e-06,
+      "loss": 0.6325,
+      "step": 24669
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0237806819147788e-06,
+      "loss": 0.6294,
+      "step": 24670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0233008003077982e-06,
+      "loss": 0.6661,
+      "step": 24671
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0228210251309412e-06,
+      "loss": 0.7067,
+      "step": 24672
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0223413563898944e-06,
+      "loss": 0.6416,
+      "step": 24673
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.021861794090352e-06,
+      "loss": 0.6408,
+      "step": 24674
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0213823382379927e-06,
+      "loss": 0.6439,
+      "step": 24675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0209029888385036e-06,
+      "loss": 0.6761,
+      "step": 24676
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0204237458975708e-06,
+      "loss": 0.6445,
+      "step": 24677
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.019944609420872e-06,
+      "loss": 0.642,
+      "step": 24678
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0194655794140872e-06,
+      "loss": 0.6181,
+      "step": 24679
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0189866558829032e-06,
+      "loss": 0.6033,
+      "step": 24680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0185078388329905e-06,
+      "loss": 0.6798,
+      "step": 24681
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0180291282700295e-06,
+      "loss": 0.6445,
+      "step": 24682
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0175505241996974e-06,
+      "loss": 0.6625,
+      "step": 24683
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0170720266276647e-06,
+      "loss": 0.6793,
+      "step": 24684
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.016593635559604e-06,
+      "loss": 0.6188,
+      "step": 24685
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0161153510011945e-06,
+      "loss": 0.6853,
+      "step": 24686
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0156371729580993e-06,
+      "loss": 0.6328,
+      "step": 24687
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0151591014359918e-06,
+      "loss": 0.6266,
+      "step": 24688
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0146811364405407e-06,
+      "loss": 0.6113,
+      "step": 24689
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0142032779774092e-06,
+      "loss": 0.6583,
+      "step": 24690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0137255260522628e-06,
+      "loss": 0.5799,
+      "step": 24691
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0132478806707713e-06,
+      "loss": 0.6683,
+      "step": 24692
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0127703418385937e-06,
+      "loss": 0.6094,
+      "step": 24693
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.012292909561392e-06,
+      "loss": 0.614,
+      "step": 24694
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0118155838448297e-06,
+      "loss": 0.6435,
+      "step": 24695
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0113383646945617e-06,
+      "loss": 0.6533,
+      "step": 24696
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0108612521162453e-06,
+      "loss": 0.6248,
+      "step": 24697
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0103842461155456e-06,
+      "loss": 0.6321,
+      "step": 24698
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0099073466981091e-06,
+      "loss": 0.635,
+      "step": 24699
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0094305538695937e-06,
+      "loss": 0.6395,
+      "step": 24700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0089538676356546e-06,
+      "loss": 0.6475,
+      "step": 24701
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.00847728800194e-06,
+      "loss": 0.5902,
+      "step": 24702
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0080008149740984e-06,
+      "loss": 0.5833,
+      "step": 24703
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0075244485577863e-06,
+      "loss": 0.6404,
+      "step": 24704
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0070481887586459e-06,
+      "loss": 0.63,
+      "step": 24705
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0065720355823238e-06,
+      "loss": 0.6058,
+      "step": 24706
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0060959890344702e-06,
+      "loss": 0.6549,
+      "step": 24707
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0056200491207246e-06,
+      "loss": 0.6045,
+      "step": 24708
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0051442158467283e-06,
+      "loss": 0.6337,
+      "step": 24709
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0046684892181303e-06,
+      "loss": 0.6135,
+      "step": 24710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0041928692405645e-06,
+      "loss": 0.6157,
+      "step": 24711
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003717355919671e-06,
+      "loss": 0.7028,
+      "step": 24712
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0032419492610911e-06,
+      "loss": 0.6539,
+      "step": 24713
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.002766649270458e-06,
+      "loss": 0.6621,
+      "step": 24714
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0022914559534048e-06,
+      "loss": 0.5894,
+      "step": 24715
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0018163693155726e-06,
+      "loss": 0.6183,
+      "step": 24716
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0013413893625879e-06,
+      "loss": 0.6104,
+      "step": 24717
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0008665161000853e-06,
+      "loss": 0.6171,
+      "step": 24718
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0003917495336956e-06,
+      "loss": 0.6826,
+      "step": 24719
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.999170896690436e-07,
+      "loss": 0.6284,
+      "step": 24720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.994425365117587e-07,
+      "loss": 0.614,
+      "step": 24721
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.989680900674714e-07,
+      "loss": 0.65,
+      "step": 24722
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.984937503418024e-07,
+      "loss": 0.6198,
+      "step": 24723
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.980195173403772e-07,
+      "loss": 0.6285,
+      "step": 24724
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.975453910688193e-07,
+      "loss": 0.6635,
+      "step": 24725
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.970713715327473e-07,
+      "loss": 0.6579,
+      "step": 24726
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.965974587377835e-07,
+      "loss": 0.6057,
+      "step": 24727
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.961236526895457e-07,
+      "loss": 0.6665,
+      "step": 24728
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.956499533936525e-07,
+      "loss": 0.6508,
+      "step": 24729
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.951763608557208e-07,
+      "loss": 0.5992,
+      "step": 24730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.947028750813625e-07,
+      "loss": 0.6486,
+      "step": 24731
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.942294960761944e-07,
+      "loss": 0.7002,
+      "step": 24732
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.937562238458277e-07,
+      "loss": 0.6117,
+      "step": 24733
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.93283058395873e-07,
+      "loss": 0.6867,
+      "step": 24734
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.928099997319417e-07,
+      "loss": 0.6499,
+      "step": 24735
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.923370478596438e-07,
+      "loss": 0.6031,
+      "step": 24736
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.918642027845837e-07,
+      "loss": 0.5838,
+      "step": 24737
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.913914645123678e-07,
+      "loss": 0.6402,
+      "step": 24738
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.90918833048603e-07,
+      "loss": 0.676,
+      "step": 24739
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.904463083988912e-07,
+      "loss": 0.649,
+      "step": 24740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.89973890568836e-07,
+      "loss": 0.593,
+      "step": 24741
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.895015795640406e-07,
+      "loss": 0.6527,
+      "step": 24742
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.890293753900992e-07,
+      "loss": 0.6249,
+      "step": 24743
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.885572780526143e-07,
+      "loss": 0.5923,
+      "step": 24744
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.880852875571823e-07,
+      "loss": 0.6317,
+      "step": 24745
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.876134039093988e-07,
+      "loss": 0.6503,
+      "step": 24746
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.871416271148604e-07,
+      "loss": 0.615,
+      "step": 24747
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.866699571791593e-07,
+      "loss": 0.6838,
+      "step": 24748
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.86198394107888e-07,
+      "loss": 0.6344,
+      "step": 24749
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.857269379066358e-07,
+      "loss": 0.5869,
+      "step": 24750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.852555885809945e-07,
+      "loss": 0.6226,
+      "step": 24751
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.847843461365513e-07,
+      "loss": 0.6232,
+      "step": 24752
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.843132105788944e-07,
+      "loss": 0.6656,
+      "step": 24753
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.838421819136112e-07,
+      "loss": 0.6217,
+      "step": 24754
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.833712601462819e-07,
+      "loss": 0.657,
+      "step": 24755
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.82900445282493e-07,
+      "loss": 0.6254,
+      "step": 24756
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.824297373278268e-07,
+      "loss": 0.6456,
+      "step": 24757
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.819591362878621e-07,
+      "loss": 0.6188,
+      "step": 24758
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.814886421681802e-07,
+      "loss": 0.5853,
+      "step": 24759
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.810182549743608e-07,
+      "loss": 0.6454,
+      "step": 24760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.805479747119773e-07,
+      "loss": 0.6098,
+      "step": 24761
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.800778013866064e-07,
+      "loss": 0.5959,
+      "step": 24762
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.796077350038247e-07,
+      "loss": 0.6043,
+      "step": 24763
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.791377755692034e-07,
+      "loss": 0.6433,
+      "step": 24764
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.786679230883155e-07,
+      "loss": 0.6613,
+      "step": 24765
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.781981775667327e-07,
+      "loss": 0.6245,
+      "step": 24766
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.777285390100221e-07,
+      "loss": 0.6693,
+      "step": 24767
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.77259007423752e-07,
+      "loss": 0.6291,
+      "step": 24768
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.76789582813491e-07,
+      "loss": 0.6254,
+      "step": 24769
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.763202651848035e-07,
+      "loss": 0.6232,
+      "step": 24770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.758510545432543e-07,
+      "loss": 0.6308,
+      "step": 24771
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.753819508944085e-07,
+      "loss": 0.6005,
+      "step": 24772
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.749129542438241e-07,
+      "loss": 0.6421,
+      "step": 24773
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.744440645970633e-07,
+      "loss": 0.7088,
+      "step": 24774
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.73975281959686e-07,
+      "loss": 0.5766,
+      "step": 24775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.735066063372488e-07,
+      "loss": 0.7131,
+      "step": 24776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.7303803773531e-07,
+      "loss": 0.6393,
+      "step": 24777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.725695761594256e-07,
+      "loss": 0.5966,
+      "step": 24778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.721012216151472e-07,
+      "loss": 0.64,
+      "step": 24779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.716329741080288e-07,
+      "loss": 0.5694,
+      "step": 24780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.71164833643623e-07,
+      "loss": 0.6466,
+      "step": 24781
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.706968002274797e-07,
+      "loss": 0.6149,
+      "step": 24782
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.702288738651477e-07,
+      "loss": 0.6745,
+      "step": 24783
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.697610545621772e-07,
+      "loss": 0.5783,
+      "step": 24784
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.692933423241101e-07,
+      "loss": 0.6206,
+      "step": 24785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.688257371564947e-07,
+      "loss": 0.6281,
+      "step": 24786
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.683582390648748e-07,
+      "loss": 0.5873,
+      "step": 24787
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.678908480547932e-07,
+      "loss": 0.5923,
+      "step": 24788
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.67423564131792e-07,
+      "loss": 0.6038,
+      "step": 24789
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.669563873014108e-07,
+      "loss": 0.6748,
+      "step": 24790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.66489317569188e-07,
+      "loss": 0.6127,
+      "step": 24791
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.660223549406611e-07,
+      "loss": 0.572,
+      "step": 24792
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.65555499421368e-07,
+      "loss": 0.6837,
+      "step": 24793
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.650887510168427e-07,
+      "loss": 0.6368,
+      "step": 24794
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.646221097326213e-07,
+      "loss": 0.6059,
+      "step": 24795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.641555755742328e-07,
+      "loss": 0.688,
+      "step": 24796
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.6368914854721e-07,
+      "loss": 0.5705,
+      "step": 24797
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.632228286570832e-07,
+      "loss": 0.6962,
+      "step": 24798
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.627566159093815e-07,
+      "loss": 0.659,
+      "step": 24799
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.622905103096324e-07,
+      "loss": 0.6544,
+      "step": 24800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.61824511863364e-07,
+      "loss": 0.6332,
+      "step": 24801
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.613586205760983e-07,
+      "loss": 0.6164,
+      "step": 24802
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.608928364533588e-07,
+      "loss": 0.6495,
+      "step": 24803
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.604271595006698e-07,
+      "loss": 0.6028,
+      "step": 24804
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.599615897235514e-07,
+      "loss": 0.6201,
+      "step": 24805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.59496127127525e-07,
+      "loss": 0.6444,
+      "step": 24806
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.590307717181103e-07,
+      "loss": 0.6652,
+      "step": 24807
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.585655235008195e-07,
+      "loss": 0.6052,
+      "step": 24808
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.58100382481173e-07,
+      "loss": 0.6792,
+      "step": 24809
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.57635348664684e-07,
+      "loss": 0.6157,
+      "step": 24810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.571704220568668e-07,
+      "loss": 0.6536,
+      "step": 24811
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.567056026632337e-07,
+      "loss": 0.5903,
+      "step": 24812
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.562408904892972e-07,
+      "loss": 0.635,
+      "step": 24813
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.557762855405627e-07,
+      "loss": 0.6585,
+      "step": 24814
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.553117878225425e-07,
+      "loss": 0.6012,
+      "step": 24815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.548473973407423e-07,
+      "loss": 0.6168,
+      "step": 24816
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.543831141006676e-07,
+      "loss": 0.6199,
+      "step": 24817
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.539189381078251e-07,
+      "loss": 0.7044,
+      "step": 24818
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.534548693677181e-07,
+      "loss": 0.6862,
+      "step": 24819
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.529909078858457e-07,
+      "loss": 0.6173,
+      "step": 24820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.525270536677112e-07,
+      "loss": 0.5814,
+      "step": 24821
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.520633067188134e-07,
+      "loss": 0.5801,
+      "step": 24822
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.515996670446504e-07,
+      "loss": 0.5689,
+      "step": 24823
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.511361346507197e-07,
+      "loss": 0.6329,
+      "step": 24824
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.506727095425183e-07,
+      "loss": 0.5842,
+      "step": 24825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.502093917255384e-07,
+      "loss": 0.6392,
+      "step": 24826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.497461812052744e-07,
+      "loss": 0.6374,
+      "step": 24827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.492830779872175e-07,
+      "loss": 0.6544,
+      "step": 24828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.48820082076859e-07,
+      "loss": 0.6234,
+      "step": 24829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.483571934796887e-07,
+      "loss": 0.6569,
+      "step": 24830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.478944122011958e-07,
+      "loss": 0.6063,
+      "step": 24831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.474317382468634e-07,
+      "loss": 0.6362,
+      "step": 24832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.469691716221796e-07,
+      "loss": 0.6422,
+      "step": 24833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.465067123326277e-07,
+      "loss": 0.6035,
+      "step": 24834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.460443603836922e-07,
+      "loss": 0.6547,
+      "step": 24835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.455821157808543e-07,
+      "loss": 0.6446,
+      "step": 24836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.451199785295961e-07,
+      "loss": 0.634,
+      "step": 24837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.446579486353924e-07,
+      "loss": 0.6555,
+      "step": 24838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.441960261037241e-07,
+      "loss": 0.7614,
+      "step": 24839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.437342109400683e-07,
+      "loss": 0.6125,
+      "step": 24840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.43272503149899e-07,
+      "loss": 0.6599,
+      "step": 24841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.428109027386911e-07,
+      "loss": 0.6323,
+      "step": 24842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.423494097119201e-07,
+      "loss": 0.6153,
+      "step": 24843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.418880240750517e-07,
+      "loss": 0.6267,
+      "step": 24844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.414267458335602e-07,
+      "loss": 0.7123,
+      "step": 24845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.409655749929136e-07,
+      "loss": 0.6401,
+      "step": 24846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.405045115585787e-07,
+      "loss": 0.6278,
+      "step": 24847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.400435555360243e-07,
+      "loss": 0.6758,
+      "step": 24848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.395827069307162e-07,
+      "loss": 0.6275,
+      "step": 24849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.391219657481143e-07,
+      "loss": 0.6014,
+      "step": 24850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.386613319936843e-07,
+      "loss": 0.6534,
+      "step": 24851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.382008056728864e-07,
+      "loss": 0.6633,
+      "step": 24852
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.377403867911816e-07,
+      "loss": 0.6669,
+      "step": 24853
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.372800753540301e-07,
+      "loss": 0.6029,
+      "step": 24854
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.368198713668853e-07,
+      "loss": 0.6159,
+      "step": 24855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.363597748352061e-07,
+      "loss": 0.5878,
+      "step": 24856
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.358997857644492e-07,
+      "loss": 0.6076,
+      "step": 24857
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.354399041600625e-07,
+      "loss": 0.6283,
+      "step": 24858
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.349801300275052e-07,
+      "loss": 0.6752,
+      "step": 24859
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.345204633722272e-07,
+      "loss": 0.6388,
+      "step": 24860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.340609041996751e-07,
+      "loss": 0.5739,
+      "step": 24861
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.336014525152992e-07,
+      "loss": 0.5939,
+      "step": 24862
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.331421083245495e-07,
+      "loss": 0.5901,
+      "step": 24863
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.326828716328651e-07,
+      "loss": 0.5933,
+      "step": 24864
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.322237424456981e-07,
+      "loss": 0.6254,
+      "step": 24865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.31764720768491e-07,
+      "loss": 0.67,
+      "step": 24866
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.313058066066827e-07,
+      "loss": 0.6047,
+      "step": 24867
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.308469999657166e-07,
+      "loss": 0.7025,
+      "step": 24868
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.303883008510328e-07,
+      "loss": 0.6381,
+      "step": 24869
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.299297092680659e-07,
+      "loss": 0.7041,
+      "step": 24870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.294712252222571e-07,
+      "loss": 0.6592,
+      "step": 24871
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.290128487190442e-07,
+      "loss": 0.663,
+      "step": 24872
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.285545797638562e-07,
+      "loss": 0.6241,
+      "step": 24873
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.280964183621288e-07,
+      "loss": 0.606,
+      "step": 24874
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.276383645192966e-07,
+      "loss": 0.613,
+      "step": 24875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.271804182407839e-07,
+      "loss": 0.6152,
+      "step": 24876
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.267225795320279e-07,
+      "loss": 0.6245,
+      "step": 24877
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.262648483984537e-07,
+      "loss": 0.5919,
+      "step": 24878
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.258072248454875e-07,
+      "loss": 0.649,
+      "step": 24879
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.253497088785546e-07,
+      "loss": 0.6416,
+      "step": 24880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.248923005030818e-07,
+      "loss": 0.6618,
+      "step": 24881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.244349997244884e-07,
+      "loss": 0.5935,
+      "step": 24882
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.239778065481996e-07,
+      "loss": 0.6683,
+      "step": 24883
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.235207209796382e-07,
+      "loss": 0.5927,
+      "step": 24884
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.230637430242173e-07,
+      "loss": 0.6856,
+      "step": 24885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.226068726873594e-07,
+      "loss": 0.6318,
+      "step": 24886
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.221501099744801e-07,
+      "loss": 0.6119,
+      "step": 24887
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.216934548909929e-07,
+      "loss": 0.6372,
+      "step": 24888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.212369074423144e-07,
+      "loss": 0.6559,
+      "step": 24889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.207804676338594e-07,
+      "loss": 0.6547,
+      "step": 24890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.203241354710358e-07,
+      "loss": 0.6487,
+      "step": 24891
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.198679109592546e-07,
+      "loss": 0.6435,
+      "step": 24892
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.194117941039282e-07,
+      "loss": 0.5671,
+      "step": 24893
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.189557849104581e-07,
+      "loss": 0.6577,
+      "step": 24894
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.184998833842573e-07,
+      "loss": 0.626,
+      "step": 24895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.180440895307297e-07,
+      "loss": 0.6486,
+      "step": 24896
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.175884033552762e-07,
+      "loss": 0.5795,
+      "step": 24897
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.171328248633016e-07,
+      "loss": 0.6355,
+      "step": 24898
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.166773540602092e-07,
+      "loss": 0.6748,
+      "step": 24899
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.162219909513925e-07,
+      "loss": 0.6142,
+      "step": 24900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.157667355422583e-07,
+      "loss": 0.6656,
+      "step": 24901
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.153115878382024e-07,
+      "loss": 0.6306,
+      "step": 24902
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.148565478446181e-07,
+      "loss": 0.619,
+      "step": 24903
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.144016155669022e-07,
+      "loss": 0.6145,
+      "step": 24904
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.139467910104504e-07,
+      "loss": 0.62,
+      "step": 24905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.134920741806496e-07,
+      "loss": 0.643,
+      "step": 24906
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.130374650828966e-07,
+      "loss": 0.6093,
+      "step": 24907
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.125829637225813e-07,
+      "loss": 0.6492,
+      "step": 24908
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.121285701050886e-07,
+      "loss": 0.581,
+      "step": 24909
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.116742842358083e-07,
+      "loss": 0.627,
+      "step": 24910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.112201061201275e-07,
+      "loss": 0.6625,
+      "step": 24911
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.107660357634263e-07,
+      "loss": 0.6339,
+      "step": 24912
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.103120731710935e-07,
+      "loss": 0.6172,
+      "step": 24913
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.098582183485116e-07,
+      "loss": 0.6596,
+      "step": 24914
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.094044713010585e-07,
+      "loss": 0.5722,
+      "step": 24915
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.089508320341156e-07,
+      "loss": 0.5972,
+      "step": 24916
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.084973005530595e-07,
+      "loss": 0.6518,
+      "step": 24917
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.080438768632671e-07,
+      "loss": 0.6384,
+      "step": 24918
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.075905609701185e-07,
+      "loss": 0.658,
+      "step": 24919
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.071373528789851e-07,
+      "loss": 0.6392,
+      "step": 24920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.066842525952402e-07,
+      "loss": 0.6498,
+      "step": 24921
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.062312601242595e-07,
+      "loss": 0.6379,
+      "step": 24922
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.057783754714078e-07,
+      "loss": 0.6462,
+      "step": 24923
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.053255986420562e-07,
+      "loss": 0.6505,
+      "step": 24924
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.048729296415782e-07,
+      "loss": 0.621,
+      "step": 24925
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.044203684753361e-07,
+      "loss": 0.6278,
+      "step": 24926
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.039679151486957e-07,
+      "loss": 0.6161,
+      "step": 24927
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.035155696670239e-07,
+      "loss": 0.6355,
+      "step": 24928
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.030633320356796e-07,
+      "loss": 0.6624,
+      "step": 24929
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.026112022600264e-07,
+      "loss": 0.5683,
+      "step": 24930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.0215918034543e-07,
+      "loss": 0.6385,
+      "step": 24931
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.017072662972415e-07,
+      "loss": 0.6657,
+      "step": 24932
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.012554601208245e-07,
+      "loss": 0.6846,
+      "step": 24933
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.008037618215348e-07,
+      "loss": 0.6362,
+      "step": 24934
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.003521714047258e-07,
+      "loss": 0.6102,
+      "step": 24935
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.999006888757511e-07,
+      "loss": 0.637,
+      "step": 24936
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.994493142399685e-07,
+      "loss": 0.6456,
+      "step": 24937
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.989980475027249e-07,
+      "loss": 0.666,
+      "step": 24938
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.985468886693727e-07,
+      "loss": 0.6331,
+      "step": 24939
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.98095837745262e-07,
+      "loss": 0.6271,
+      "step": 24940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.976448947357375e-07,
+      "loss": 0.6294,
+      "step": 24941
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.97194059646146e-07,
+      "loss": 0.6304,
+      "step": 24942
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.967433324818365e-07,
+      "loss": 0.6623,
+      "step": 24943
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.962927132481491e-07,
+      "loss": 0.611,
+      "step": 24944
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.958422019504287e-07,
+      "loss": 0.6263,
+      "step": 24945
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.953917985940164e-07,
+      "loss": 0.6769,
+      "step": 24946
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.949415031842501e-07,
+      "loss": 0.6752,
+      "step": 24947
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.944913157264689e-07,
+      "loss": 0.5851,
+      "step": 24948
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.940412362260154e-07,
+      "loss": 0.6686,
+      "step": 24949
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.935912646882194e-07,
+      "loss": 0.6401,
+      "step": 24950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.931414011184191e-07,
+      "loss": 0.6909,
+      "step": 24951
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.926916455219503e-07,
+      "loss": 0.6821,
+      "step": 24952
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.922419979041407e-07,
+      "loss": 0.6601,
+      "step": 24953
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.917924582703208e-07,
+      "loss": 0.5935,
+      "step": 24954
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.913430266258282e-07,
+      "loss": 0.6604,
+      "step": 24955
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.908937029759835e-07,
+      "loss": 0.6088,
+      "step": 24956
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.904444873261175e-07,
+      "loss": 0.616,
+      "step": 24957
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.899953796815586e-07,
+      "loss": 0.6563,
+      "step": 24958
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.895463800476256e-07,
+      "loss": 0.6766,
+      "step": 24959
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.890974884296433e-07,
+      "loss": 0.6338,
+      "step": 24960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.886487048329395e-07,
+      "loss": 0.609,
+      "step": 24961
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.882000292628301e-07,
+      "loss": 0.648,
+      "step": 24962
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.877514617246352e-07,
+      "loss": 0.656,
+      "step": 24963
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.873030022236751e-07,
+      "loss": 0.6718,
+      "step": 24964
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.868546507652643e-07,
+      "loss": 0.6469,
+      "step": 24965
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.864064073547197e-07,
+      "loss": 0.6463,
+      "step": 24966
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.859582719973559e-07,
+      "loss": 0.6759,
+      "step": 24967
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.855102446984865e-07,
+      "loss": 0.6238,
+      "step": 24968
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.850623254634228e-07,
+      "loss": 0.6112,
+      "step": 24969
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.846145142974771e-07,
+      "loss": 0.6518,
+      "step": 24970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.841668112059565e-07,
+      "loss": 0.6046,
+      "step": 24971
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.837192161941688e-07,
+      "loss": 0.6574,
+      "step": 24972
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.83271729267422e-07,
+      "loss": 0.6662,
+      "step": 24973
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.828243504310219e-07,
+      "loss": 0.6637,
+      "step": 24974
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.823770796902731e-07,
+      "loss": 0.6704,
+      "step": 24975
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.819299170504791e-07,
+      "loss": 0.5545,
+      "step": 24976
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.81482862516938e-07,
+      "loss": 0.6091,
+      "step": 24977
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.810359160949533e-07,
+      "loss": 0.6455,
+      "step": 24978
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.805890777898229e-07,
+      "loss": 0.5674,
+      "step": 24979
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.80142347606846e-07,
+      "loss": 0.5538,
+      "step": 24980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.796957255513194e-07,
+      "loss": 0.6661,
+      "step": 24981
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.792492116285345e-07,
+      "loss": 0.6252,
+      "step": 24982
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.788028058437892e-07,
+      "loss": 0.633,
+      "step": 24983
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.783565082023748e-07,
+      "loss": 0.5778,
+      "step": 24984
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.779103187095827e-07,
+      "loss": 0.632,
+      "step": 24985
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.774642373707032e-07,
+      "loss": 0.6629,
+      "step": 24986
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.770182641910274e-07,
+      "loss": 0.5808,
+      "step": 24987
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.76572399175839e-07,
+      "loss": 0.6566,
+      "step": 24988
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.76126642330426e-07,
+      "loss": 0.6869,
+      "step": 24989
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.75680993660073e-07,
+      "loss": 0.6672,
+      "step": 24990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.752354531700646e-07,
+      "loss": 0.5979,
+      "step": 24991
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.747900208656824e-07,
+      "loss": 0.6236,
+      "step": 24992
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.743446967522096e-07,
+      "loss": 0.6779,
+      "step": 24993
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.738994808349233e-07,
+      "loss": 0.6514,
+      "step": 24994
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.734543731191025e-07,
+      "loss": 0.6153,
+      "step": 24995
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.730093736100254e-07,
+      "loss": 0.6583,
+      "step": 24996
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.725644823129686e-07,
+      "loss": 0.6381,
+      "step": 24997
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.721196992332049e-07,
+      "loss": 0.6132,
+      "step": 24998
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.716750243760119e-07,
+      "loss": 0.5923,
+      "step": 24999
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.712304577466569e-07,
+      "loss": 0.5515,
+      "step": 25000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.707859993504119e-07,
+      "loss": 0.6451,
+      "step": 25001
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.703416491925487e-07,
+      "loss": 0.6219,
+      "step": 25002
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.698974072783328e-07,
+      "loss": 0.5886,
+      "step": 25003
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.694532736130346e-07,
+      "loss": 0.6472,
+      "step": 25004
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.690092482019185e-07,
+      "loss": 0.6062,
+      "step": 25005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.685653310502473e-07,
+      "loss": 0.629,
+      "step": 25006
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.681215221632855e-07,
+      "loss": 0.607,
+      "step": 25007
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.676778215462944e-07,
+      "loss": 0.5592,
+      "step": 25008
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.672342292045355e-07,
+      "loss": 0.5845,
+      "step": 25009
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.667907451432678e-07,
+      "loss": 0.6381,
+      "step": 25010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.663473693677504e-07,
+      "loss": 0.6267,
+      "step": 25011
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.659041018832382e-07,
+      "loss": 0.5758,
+      "step": 25012
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.654609426949878e-07,
+      "loss": 0.5656,
+      "step": 25013
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.65017891808253e-07,
+      "loss": 0.6451,
+      "step": 25014
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.645749492282862e-07,
+      "loss": 0.6114,
+      "step": 25015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.641321149603398e-07,
+      "loss": 0.6722,
+      "step": 25016
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.636893890096665e-07,
+      "loss": 0.5704,
+      "step": 25017
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.632467713815107e-07,
+      "loss": 0.6619,
+      "step": 25018
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.628042620811228e-07,
+      "loss": 0.6538,
+      "step": 25019
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.623618611137497e-07,
+      "loss": 0.6012,
+      "step": 25020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.619195684846349e-07,
+      "loss": 0.5902,
+      "step": 25021
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.614773841990231e-07,
+      "loss": 0.6305,
+      "step": 25022
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.610353082621603e-07,
+      "loss": 0.6401,
+      "step": 25023
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.605933406792821e-07,
+      "loss": 0.6158,
+      "step": 25024
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.601514814556311e-07,
+      "loss": 0.6135,
+      "step": 25025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.597097305964463e-07,
+      "loss": 0.6397,
+      "step": 25026
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.592680881069659e-07,
+      "loss": 0.6637,
+      "step": 25027
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.588265539924246e-07,
+      "loss": 0.6453,
+      "step": 25028
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.583851282580602e-07,
+      "loss": 0.6535,
+      "step": 25029
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.579438109091032e-07,
+      "loss": 0.6189,
+      "step": 25030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.57502601950786e-07,
+      "loss": 0.6237,
+      "step": 25031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.57061501388341e-07,
+      "loss": 0.6141,
+      "step": 25032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.566205092269985e-07,
+      "loss": 0.6429,
+      "step": 25033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.561796254719857e-07,
+      "loss": 0.6893,
+      "step": 25034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.557388501285324e-07,
+      "loss": 0.6605,
+      "step": 25035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.552981832018603e-07,
+      "loss": 0.7093,
+      "step": 25036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.548576246971963e-07,
+      "loss": 0.6666,
+      "step": 25037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.54417174619765e-07,
+      "loss": 0.6197,
+      "step": 25038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.539768329747866e-07,
+      "loss": 0.5966,
+      "step": 25039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.535365997674838e-07,
+      "loss": 0.619,
+      "step": 25040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.530964750030734e-07,
+      "loss": 0.6568,
+      "step": 25041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.526564586867758e-07,
+      "loss": 0.6589,
+      "step": 25042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.522165508238078e-07,
+      "loss": 0.6382,
+      "step": 25043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.51776751419383e-07,
+      "loss": 0.5796,
+      "step": 25044
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.513370604787185e-07,
+      "loss": 0.6832,
+      "step": 25045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.508974780070279e-07,
+      "loss": 0.6729,
+      "step": 25046
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.504580040095189e-07,
+      "loss": 0.6405,
+      "step": 25047
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.500186384914055e-07,
+      "loss": 0.7035,
+      "step": 25048
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.495793814578957e-07,
+      "loss": 0.6242,
+      "step": 25049
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.491402329141974e-07,
+      "loss": 0.6884,
+      "step": 25050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.487011928655175e-07,
+      "loss": 0.6446,
+      "step": 25051
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.482622613170633e-07,
+      "loss": 0.6178,
+      "step": 25052
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.478234382740358e-07,
+      "loss": 0.5501,
+      "step": 25053
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.473847237416377e-07,
+      "loss": 0.6608,
+      "step": 25054
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.469461177250726e-07,
+      "loss": 0.6319,
+      "step": 25055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.465076202295397e-07,
+      "loss": 0.6414,
+      "step": 25056
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.460692312602392e-07,
+      "loss": 0.607,
+      "step": 25057
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.456309508223681e-07,
+      "loss": 0.6304,
+      "step": 25058
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.451927789211212e-07,
+      "loss": 0.6205,
+      "step": 25059
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.447547155616953e-07,
+      "loss": 0.6247,
+      "step": 25060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.44316760749283e-07,
+      "loss": 0.5837,
+      "step": 25061
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.43878914489078e-07,
+      "loss": 0.6429,
+      "step": 25062
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.434411767862704e-07,
+      "loss": 0.5877,
+      "step": 25063
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.43003547646053e-07,
+      "loss": 0.6482,
+      "step": 25064
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.425660270736102e-07,
+      "loss": 0.5694,
+      "step": 25065
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.421286150741315e-07,
+      "loss": 0.5995,
+      "step": 25066
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.416913116528025e-07,
+      "loss": 0.6286,
+      "step": 25067
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.412541168148092e-07,
+      "loss": 0.6224,
+      "step": 25068
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.408170305653329e-07,
+      "loss": 0.6168,
+      "step": 25069
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.403800529095596e-07,
+      "loss": 0.6521,
+      "step": 25070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.399431838526662e-07,
+      "loss": 0.6353,
+      "step": 25071
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.39506423399834e-07,
+      "loss": 0.5859,
+      "step": 25072
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.390697715562412e-07,
+      "loss": 0.6193,
+      "step": 25073
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.386332283270649e-07,
+      "loss": 0.6533,
+      "step": 25074
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.381967937174796e-07,
+      "loss": 0.605,
+      "step": 25075
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.377604677326646e-07,
+      "loss": 0.6159,
+      "step": 25076
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.373242503777867e-07,
+      "loss": 0.5921,
+      "step": 25077
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.36888141658021e-07,
+      "loss": 0.6104,
+      "step": 25078
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.364521415785376e-07,
+      "loss": 0.6436,
+      "step": 25079
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.360162501445068e-07,
+      "loss": 0.6892,
+      "step": 25080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.355804673610946e-07,
+      "loss": 0.6223,
+      "step": 25081
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.351447932334711e-07,
+      "loss": 0.6489,
+      "step": 25082
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.34709227766799e-07,
+      "loss": 0.5965,
+      "step": 25083
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.342737709662419e-07,
+      "loss": 0.6367,
+      "step": 25084
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.338384228369645e-07,
+      "loss": 0.6298,
+      "step": 25085
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.334031833841272e-07,
+      "loss": 0.5768,
+      "step": 25086
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.329680526128924e-07,
+      "loss": 0.6514,
+      "step": 25087
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.325330305284185e-07,
+      "loss": 0.6857,
+      "step": 25088
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.3209811713586e-07,
+      "loss": 0.6577,
+      "step": 25089
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.316633124403772e-07,
+      "loss": 0.6127,
+      "step": 25090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.312286164471229e-07,
+      "loss": 0.601,
+      "step": 25091
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.307940291612526e-07,
+      "loss": 0.638,
+      "step": 25092
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.303595505879169e-07,
+      "loss": 0.6285,
+      "step": 25093
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.299251807322717e-07,
+      "loss": 0.5812,
+      "step": 25094
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.294909195994605e-07,
+      "loss": 0.6381,
+      "step": 25095
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.290567671946359e-07,
+      "loss": 0.6628,
+      "step": 25096
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.28622723522945e-07,
+      "loss": 0.6111,
+      "step": 25097
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.281887885895334e-07,
+      "loss": 0.648,
+      "step": 25098
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.277549623995462e-07,
+      "loss": 0.6887,
+      "step": 25099
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.273212449581281e-07,
+      "loss": 0.5723,
+      "step": 25100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.268876362704192e-07,
+      "loss": 0.639,
+      "step": 25101
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.264541363415623e-07,
+      "loss": 0.6027,
+      "step": 25102
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.260207451766933e-07,
+      "loss": 0.6149,
+      "step": 25103
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.255874627809546e-07,
+      "loss": 0.6611,
+      "step": 25104
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.251542891594844e-07,
+      "loss": 0.6033,
+      "step": 25105
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.247212243174141e-07,
+      "loss": 0.5807,
+      "step": 25106
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.242882682598819e-07,
+      "loss": 0.6043,
+      "step": 25107
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.238554209920202e-07,
+      "loss": 0.6343,
+      "step": 25108
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.234226825189562e-07,
+      "loss": 0.6343,
+      "step": 25109
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.229900528458278e-07,
+      "loss": 0.6375,
+      "step": 25110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.225575319777623e-07,
+      "loss": 0.6125,
+      "step": 25111
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.221251199198854e-07,
+      "loss": 0.5819,
+      "step": 25112
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.216928166773252e-07,
+      "loss": 0.6463,
+      "step": 25113
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.212606222552089e-07,
+      "loss": 0.6699,
+      "step": 25114
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.208285366586565e-07,
+      "loss": 0.6629,
+      "step": 25115
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.203965598927944e-07,
+      "loss": 0.6577,
+      "step": 25116
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.199646919627447e-07,
+      "loss": 0.6059,
+      "step": 25117
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.195329328736256e-07,
+      "loss": 0.6655,
+      "step": 25118
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.191012826305556e-07,
+      "loss": 0.6755,
+      "step": 25119
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.186697412386569e-07,
+      "loss": 0.6269,
+      "step": 25120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.182383087030387e-07,
+      "loss": 0.6176,
+      "step": 25121
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.178069850288217e-07,
+      "loss": 0.6211,
+      "step": 25122
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.173757702211215e-07,
+      "loss": 0.6472,
+      "step": 25123
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.169446642850454e-07,
+      "loss": 0.6214,
+      "step": 25124
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.165136672257057e-07,
+      "loss": 0.5915,
+      "step": 25125
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.160827790482162e-07,
+      "loss": 0.649,
+      "step": 25126
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.156519997576806e-07,
+      "loss": 0.6471,
+      "step": 25127
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.152213293592093e-07,
+      "loss": 0.6641,
+      "step": 25128
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.147907678579103e-07,
+      "loss": 0.6264,
+      "step": 25129
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.143603152588841e-07,
+      "loss": 0.6599,
+      "step": 25130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.139299715672356e-07,
+      "loss": 0.6275,
+      "step": 25131
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.134997367880704e-07,
+      "loss": 0.6257,
+      "step": 25132
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.130696109264824e-07,
+      "loss": 0.6194,
+      "step": 25133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.126395939875786e-07,
+      "loss": 0.581,
+      "step": 25134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.12209685976455e-07,
+      "loss": 0.6307,
+      "step": 25135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.117798868982074e-07,
+      "loss": 0.6455,
+      "step": 25136
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.11350196757933e-07,
+      "loss": 0.6057,
+      "step": 25137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.109206155607264e-07,
+      "loss": 0.6437,
+      "step": 25138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.104911433116769e-07,
+      "loss": 0.6504,
+      "step": 25139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.100617800158805e-07,
+      "loss": 0.6328,
+      "step": 25140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.096325256784299e-07,
+      "loss": 0.6191,
+      "step": 25141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.092033803044097e-07,
+      "loss": 0.6558,
+      "step": 25142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.087743438989093e-07,
+      "loss": 0.625,
+      "step": 25143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.08345416467019e-07,
+      "loss": 0.6744,
+      "step": 25144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.07916598013817e-07,
+      "loss": 0.6331,
+      "step": 25145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.074878885443926e-07,
+      "loss": 0.5944,
+      "step": 25146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.070592880638306e-07,
+      "loss": 0.6305,
+      "step": 25147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.06630796577208e-07,
+      "loss": 0.6192,
+      "step": 25148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.062024140896074e-07,
+      "loss": 0.6607,
+      "step": 25149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.057741406061092e-07,
+      "loss": 0.6934,
+      "step": 25150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.05345976131785e-07,
+      "loss": 0.5944,
+      "step": 25151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.049179206717173e-07,
+      "loss": 0.6061,
+      "step": 25152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.04489974230982e-07,
+      "loss": 0.5696,
+      "step": 25153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.040621368146484e-07,
+      "loss": 0.6676,
+      "step": 25154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.036344084277913e-07,
+      "loss": 0.5923,
+      "step": 25155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.032067890754824e-07,
+      "loss": 0.6105,
+      "step": 25156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.027792787627886e-07,
+      "loss": 0.5836,
+      "step": 25157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.023518774947825e-07,
+      "loss": 0.6127,
+      "step": 25158
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.019245852765323e-07,
+      "loss": 0.6118,
+      "step": 25159
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.014974021130994e-07,
+      "loss": 0.6202,
+      "step": 25160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.010703280095511e-07,
+      "loss": 0.6098,
+      "step": 25161
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.00643362970952e-07,
+      "loss": 0.6309,
+      "step": 25162
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.002165070023616e-07,
+      "loss": 0.6879,
+      "step": 25163
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.997897601088422e-07,
+      "loss": 0.5789,
+      "step": 25164
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.993631222954545e-07,
+      "loss": 0.6417,
+      "step": 25165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.989365935672544e-07,
+      "loss": 0.6435,
+      "step": 25166
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.985101739293033e-07,
+      "loss": 0.6607,
+      "step": 25167
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.980838633866516e-07,
+      "loss": 0.6613,
+      "step": 25168
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.976576619443566e-07,
+      "loss": 0.6407,
+      "step": 25169
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.972315696074706e-07,
+      "loss": 0.6247,
+      "step": 25170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.968055863810464e-07,
+      "loss": 0.6345,
+      "step": 25171
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.963797122701333e-07,
+      "loss": 0.6539,
+      "step": 25172
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.959539472797839e-07,
+      "loss": 0.6481,
+      "step": 25173
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.95528291415042e-07,
+      "loss": 0.6351,
+      "step": 25174
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.951027446809556e-07,
+      "loss": 0.6697,
+      "step": 25175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.946773070825698e-07,
+      "loss": 0.6293,
+      "step": 25176
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.942519786249304e-07,
+      "loss": 0.6168,
+      "step": 25177
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.938267593130778e-07,
+      "loss": 0.5963,
+      "step": 25178
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.934016491520579e-07,
+      "loss": 0.6633,
+      "step": 25179
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.929766481469059e-07,
+      "loss": 0.7079,
+      "step": 25180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.925517563026619e-07,
+      "loss": 0.6099,
+      "step": 25181
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.921269736243642e-07,
+      "loss": 0.6792,
+      "step": 25182
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.917023001170498e-07,
+      "loss": 0.5728,
+      "step": 25183
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.912777357857537e-07,
+      "loss": 0.6483,
+      "step": 25184
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.908532806355107e-07,
+      "loss": 0.6523,
+      "step": 25185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.904289346713489e-07,
+      "loss": 0.6449,
+      "step": 25186
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.900046978983034e-07,
+      "loss": 0.6838,
+      "step": 25187
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.895805703214022e-07,
+      "loss": 0.6281,
+      "step": 25188
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.891565519456757e-07,
+      "loss": 0.6858,
+      "step": 25189
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.887326427761488e-07,
+      "loss": 0.6153,
+      "step": 25190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.883088428178509e-07,
+      "loss": 0.6537,
+      "step": 25191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.878851520758024e-07,
+      "loss": 0.6767,
+      "step": 25192
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.874615705550293e-07,
+      "loss": 0.5937,
+      "step": 25193
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.87038098260553e-07,
+      "loss": 0.6836,
+      "step": 25194
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.866147351973952e-07,
+      "loss": 0.5687,
+      "step": 25195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.86191481370574e-07,
+      "loss": 0.6158,
+      "step": 25196
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.857683367851098e-07,
+      "loss": 0.6281,
+      "step": 25197
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.853453014460167e-07,
+      "loss": 0.6258,
+      "step": 25198
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.849223753583113e-07,
+      "loss": 0.62,
+      "step": 25199
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.844995585270088e-07,
+      "loss": 0.6267,
+      "step": 25200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.840768509571217e-07,
+      "loss": 0.6152,
+      "step": 25201
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.836542526536616e-07,
+      "loss": 0.6474,
+      "step": 25202
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.832317636216402e-07,
+      "loss": 0.6597,
+      "step": 25203
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.828093838660644e-07,
+      "loss": 0.6673,
+      "step": 25204
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.823871133919436e-07,
+      "loss": 0.6923,
+      "step": 25205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.819649522042838e-07,
+      "loss": 0.666,
+      "step": 25206
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.815429003080898e-07,
+      "loss": 0.6004,
+      "step": 25207
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.811209577083656e-07,
+      "loss": 0.5576,
+      "step": 25208
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.806991244101169e-07,
+      "loss": 0.6444,
+      "step": 25209
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.802774004183411e-07,
+      "loss": 0.5508,
+      "step": 25210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.798557857380396e-07,
+      "loss": 0.6161,
+      "step": 25211
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.794342803742105e-07,
+      "loss": 0.6578,
+      "step": 25212
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.790128843318523e-07,
+      "loss": 0.6675,
+      "step": 25213
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.785915976159619e-07,
+      "loss": 0.6069,
+      "step": 25214
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.781704202315343e-07,
+      "loss": 0.6435,
+      "step": 25215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.777493521835599e-07,
+      "loss": 0.6043,
+      "step": 25216
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.773283934770337e-07,
+      "loss": 0.6411,
+      "step": 25217
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.76907544116946e-07,
+      "loss": 0.6698,
+      "step": 25218
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.764868041082873e-07,
+      "loss": 0.6098,
+      "step": 25219
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.760661734560448e-07,
+      "loss": 0.6202,
+      "step": 25220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.756456521652078e-07,
+      "loss": 0.6261,
+      "step": 25221
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.752252402407589e-07,
+      "loss": 0.6168,
+      "step": 25222
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.748049376876832e-07,
+      "loss": 0.625,
+      "step": 25223
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.743847445109665e-07,
+      "loss": 0.6034,
+      "step": 25224
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.739646607155882e-07,
+      "loss": 0.6537,
+      "step": 25225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.73544686306531e-07,
+      "loss": 0.6655,
+      "step": 25226
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.731248212887721e-07,
+      "loss": 0.6324,
+      "step": 25227
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.727050656672907e-07,
+      "loss": 0.6346,
+      "step": 25228
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.722854194470631e-07,
+      "loss": 0.6243,
+      "step": 25229
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.71865882633065e-07,
+      "loss": 0.6397,
+      "step": 25230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.714464552302702e-07,
+      "loss": 0.6153,
+      "step": 25231
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.71027137243654e-07,
+      "loss": 0.6178,
+      "step": 25232
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.706079286781843e-07,
+      "loss": 0.6097,
+      "step": 25233
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.701888295388327e-07,
+      "loss": 0.6208,
+      "step": 25234
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.697698398305676e-07,
+      "loss": 0.6297,
+      "step": 25235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.693509595583582e-07,
+      "loss": 0.6134,
+      "step": 25236
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.689321887271695e-07,
+      "loss": 0.6103,
+      "step": 25237
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.685135273419686e-07,
+      "loss": 0.5886,
+      "step": 25238
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.68094975407716e-07,
+      "loss": 0.6477,
+      "step": 25239
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.676765329293767e-07,
+      "loss": 0.5875,
+      "step": 25240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.67258199911911e-07,
+      "loss": 0.6517,
+      "step": 25241
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.668399763602786e-07,
+      "loss": 0.6786,
+      "step": 25242
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.664218622794373e-07,
+      "loss": 0.6363,
+      "step": 25243
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.660038576743478e-07,
+      "loss": 0.6558,
+      "step": 25244
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.655859625499618e-07,
+      "loss": 0.6178,
+      "step": 25245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.651681769112362e-07,
+      "loss": 0.6464,
+      "step": 25246
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.647505007631239e-07,
+      "loss": 0.6369,
+      "step": 25247
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.643329341105765e-07,
+      "loss": 0.6511,
+      "step": 25248
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.639154769585467e-07,
+      "loss": 0.6505,
+      "step": 25249
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.634981293119825e-07,
+      "loss": 0.6679,
+      "step": 25250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.630808911758314e-07,
+      "loss": 0.5803,
+      "step": 25251
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.626637625550415e-07,
+      "loss": 0.6194,
+      "step": 25252
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.622467434545577e-07,
+      "loss": 0.6578,
+      "step": 25253
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 7.618298338793239e-07,
+      "loss": 0.6176,
+      "step": 25254
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.614130338342851e-07,
+      "loss": 0.6479,
+      "step": 25255
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.609963433243828e-07,
+      "loss": 0.6065,
+      "step": 25256
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.605797623545542e-07,
+      "loss": 0.6396,
+      "step": 25257
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.601632909297407e-07,
+      "loss": 0.6185,
+      "step": 25258
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.597469290548798e-07,
+      "loss": 0.6201,
+      "step": 25259
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.593306767349074e-07,
+      "loss": 0.6409,
+      "step": 25260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.589145339747606e-07,
+      "loss": 0.6447,
+      "step": 25261
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.584985007793721e-07,
+      "loss": 0.6925,
+      "step": 25262
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.580825771536737e-07,
+      "loss": 0.6702,
+      "step": 25263
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.57666763102598e-07,
+      "loss": 0.6366,
+      "step": 25264
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.572510586310733e-07,
+      "loss": 0.6176,
+      "step": 25265
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.5683546374403e-07,
+      "loss": 0.5864,
+      "step": 25266
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.564199784463955e-07,
+      "loss": 0.6644,
+      "step": 25267
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.560046027430957e-07,
+      "loss": 0.5989,
+      "step": 25268
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.555893366390544e-07,
+      "loss": 0.6826,
+      "step": 25269
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.551741801391954e-07,
+      "loss": 0.5797,
+      "step": 25270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.547591332484416e-07,
+      "loss": 0.6075,
+      "step": 25271
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.543441959717135e-07,
+      "loss": 0.6056,
+      "step": 25272
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.539293683139303e-07,
+      "loss": 0.6551,
+      "step": 25273
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.535146502800128e-07,
+      "loss": 0.6214,
+      "step": 25274
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.531000418748746e-07,
+      "loss": 0.6454,
+      "step": 25275
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.526855431034319e-07,
+      "loss": 0.586,
+      "step": 25276
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.522711539706007e-07,
+      "loss": 0.6408,
+      "step": 25277
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.518568744812926e-07,
+      "loss": 0.7003,
+      "step": 25278
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.514427046404215e-07,
+      "loss": 0.665,
+      "step": 25279
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.510286444528981e-07,
+      "loss": 0.5687,
+      "step": 25280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.50614693923627e-07,
+      "loss": 0.6593,
+      "step": 25281
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.502008530575211e-07,
+      "loss": 0.6218,
+      "step": 25282
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.497871218594843e-07,
+      "loss": 0.6276,
+      "step": 25283
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.493735003344227e-07,
+      "loss": 0.6337,
+      "step": 25284
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.489599884872401e-07,
+      "loss": 0.6311,
+      "step": 25285
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.485465863228413e-07,
+      "loss": 0.5871,
+      "step": 25286
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.481332938461239e-07,
+      "loss": 0.6304,
+      "step": 25287
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.477201110619903e-07,
+      "loss": 0.6163,
+      "step": 25288
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.4730703797534e-07,
+      "loss": 0.6905,
+      "step": 25289
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.46894074591068e-07,
+      "loss": 0.6095,
+      "step": 25290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.464812209140748e-07,
+      "loss": 0.5589,
+      "step": 25291
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.460684769492499e-07,
+      "loss": 0.6495,
+      "step": 25292
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.456558427014893e-07,
+      "loss": 0.6564,
+      "step": 25293
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.452433181756868e-07,
+      "loss": 0.5753,
+      "step": 25294
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.44830903376731e-07,
+      "loss": 0.6532,
+      "step": 25295
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.444185983095131e-07,
+      "loss": 0.6639,
+      "step": 25296
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.440064029789229e-07,
+      "loss": 0.6038,
+      "step": 25297
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.43594317389843e-07,
+      "loss": 0.6368,
+      "step": 25298
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.431823415471629e-07,
+      "loss": 0.6058,
+      "step": 25299
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.427704754557664e-07,
+      "loss": 0.5832,
+      "step": 25300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.423587191205351e-07,
+      "loss": 0.6794,
+      "step": 25301
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.419470725463529e-07,
+      "loss": 0.587,
+      "step": 25302
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.415355357381015e-07,
+      "loss": 0.6108,
+      "step": 25303
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.41124108700656e-07,
+      "loss": 0.6335,
+      "step": 25304
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.407127914388978e-07,
+      "loss": 0.669,
+      "step": 25305
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.40301583957701e-07,
+      "loss": 0.6228,
+      "step": 25306
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.398904862619438e-07,
+      "loss": 0.6263,
+      "step": 25307
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.39479498356499e-07,
+      "loss": 0.6384,
+      "step": 25308
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.390686202462405e-07,
+      "loss": 0.6965,
+      "step": 25309
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.386578519360365e-07,
+      "loss": 0.6049,
+      "step": 25310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.3824719343076e-07,
+      "loss": 0.6332,
+      "step": 25311
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.378366447352791e-07,
+      "loss": 0.6093,
+      "step": 25312
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.374262058544624e-07,
+      "loss": 0.6289,
+      "step": 25313
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.370158767931734e-07,
+      "loss": 0.5854,
+      "step": 25314
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.366056575562818e-07,
+      "loss": 0.6186,
+      "step": 25315
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.36195548148646e-07,
+      "loss": 0.6248,
+      "step": 25316
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.35785548575132e-07,
+      "loss": 0.6425,
+      "step": 25317
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.35375658840598e-07,
+      "loss": 0.6006,
+      "step": 25318
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.349658789499071e-07,
+      "loss": 0.6572,
+      "step": 25319
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.34556208907915e-07,
+      "loss": 0.6185,
+      "step": 25320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.341466487194814e-07,
+      "loss": 0.6708,
+      "step": 25321
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.337371983894592e-07,
+      "loss": 0.6205,
+      "step": 25322
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.333278579227043e-07,
+      "loss": 0.6085,
+      "step": 25323
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.329186273240707e-07,
+      "loss": 0.6054,
+      "step": 25324
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.325095065984089e-07,
+      "loss": 0.6508,
+      "step": 25325
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.321004957505707e-07,
+      "loss": 0.65,
+      "step": 25326
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.316915947854065e-07,
+      "loss": 0.5879,
+      "step": 25327
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.312828037077613e-07,
+      "loss": 0.654,
+      "step": 25328
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.308741225224836e-07,
+      "loss": 0.6919,
+      "step": 25329
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.304655512344183e-07,
+      "loss": 0.6122,
+      "step": 25330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.300570898484105e-07,
+      "loss": 0.6359,
+      "step": 25331
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.296487383693019e-07,
+      "loss": 0.6624,
+      "step": 25332
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.292404968019373e-07,
+      "loss": 0.6562,
+      "step": 25333
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.288323651511508e-07,
+      "loss": 0.6117,
+      "step": 25334
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.284243434217864e-07,
+      "loss": 0.6415,
+      "step": 25335
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.280164316186788e-07,
+      "loss": 0.6217,
+      "step": 25336
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.276086297466667e-07,
+      "loss": 0.6462,
+      "step": 25337
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.272009378105827e-07,
+      "loss": 0.6697,
+      "step": 25338
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.26793355815264e-07,
+      "loss": 0.62,
+      "step": 25339
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.263858837655391e-07,
+      "loss": 0.6,
+      "step": 25340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.259785216662407e-07,
+      "loss": 0.5916,
+      "step": 25341
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.255712695221984e-07,
+      "loss": 0.6026,
+      "step": 25342
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.251641273382403e-07,
+      "loss": 0.6168,
+      "step": 25343
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.247570951191951e-07,
+      "loss": 0.6255,
+      "step": 25344
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.243501728698898e-07,
+      "loss": 0.6344,
+      "step": 25345
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.239433605951452e-07,
+      "loss": 0.6092,
+      "step": 25346
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.235366582997849e-07,
+      "loss": 0.6411,
+      "step": 25347
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.231300659886342e-07,
+      "loss": 0.6395,
+      "step": 25348
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.227235836665103e-07,
+      "loss": 0.6543,
+      "step": 25349
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.223172113382371e-07,
+      "loss": 0.6039,
+      "step": 25350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.219109490086274e-07,
+      "loss": 0.6118,
+      "step": 25351
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.215047966824995e-07,
+      "loss": 0.6329,
+      "step": 25352
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.210987543646719e-07,
+      "loss": 0.6406,
+      "step": 25353
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.206928220599552e-07,
+      "loss": 0.6692,
+      "step": 25354
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.202869997731621e-07,
+      "loss": 0.6125,
+      "step": 25355
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.198812875091076e-07,
+      "loss": 0.6458,
+      "step": 25356
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.194756852725993e-07,
+      "loss": 0.5764,
+      "step": 25357
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.190701930684463e-07,
+      "loss": 0.6592,
+      "step": 25358
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.186648109014571e-07,
+      "loss": 0.6523,
+      "step": 25359
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.18259538776438e-07,
+      "loss": 0.6091,
+      "step": 25360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.178543766981904e-07,
+      "loss": 0.6077,
+      "step": 25361
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.174493246715242e-07,
+      "loss": 0.6243,
+      "step": 25362
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.170443827012386e-07,
+      "loss": 0.6727,
+      "step": 25363
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.166395507921331e-07,
+      "loss": 0.6038,
+      "step": 25364
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.162348289490117e-07,
+      "loss": 0.6368,
+      "step": 25365
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.158302171766696e-07,
+      "loss": 0.6606,
+      "step": 25366
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.154257154799016e-07,
+      "loss": 0.5698,
+      "step": 25367
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.150213238635117e-07,
+      "loss": 0.6292,
+      "step": 25368
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.146170423322873e-07,
+      "loss": 0.6535,
+      "step": 25369
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.142128708910235e-07,
+      "loss": 0.6701,
+      "step": 25370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.138088095445139e-07,
+      "loss": 0.6426,
+      "step": 25371
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.134048582975473e-07,
+      "loss": 0.6177,
+      "step": 25372
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.130010171549118e-07,
+      "loss": 0.6341,
+      "step": 25373
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.125972861214003e-07,
+      "loss": 0.6274,
+      "step": 25374
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.121936652017958e-07,
+      "loss": 0.6221,
+      "step": 25375
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.117901544008843e-07,
+      "loss": 0.6285,
+      "step": 25376
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.113867537234509e-07,
+      "loss": 0.6404,
+      "step": 25377
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.109834631742763e-07,
+      "loss": 0.6486,
+      "step": 25378
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.105802827581421e-07,
+      "loss": 0.6785,
+      "step": 25379
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.101772124798334e-07,
+      "loss": 0.6438,
+      "step": 25380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.097742523441242e-07,
+      "loss": 0.6868,
+      "step": 25381
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.093714023557918e-07,
+      "loss": 0.6337,
+      "step": 25382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.089686625196169e-07,
+      "loss": 0.6283,
+      "step": 25383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.085660328403709e-07,
+      "loss": 0.6459,
+      "step": 25384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.081635133228248e-07,
+      "loss": 0.6438,
+      "step": 25385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.077611039717591e-07,
+      "loss": 0.5878,
+      "step": 25386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.073588047919388e-07,
+      "loss": 0.6468,
+      "step": 25387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.069566157881346e-07,
+      "loss": 0.6258,
+      "step": 25388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.065545369651161e-07,
+      "loss": 0.6656,
+      "step": 25389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.061525683276493e-07,
+      "loss": 0.5662,
+      "step": 25390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.057507098804994e-07,
+      "loss": 0.6464,
+      "step": 25391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.053489616284348e-07,
+      "loss": 0.5951,
+      "step": 25392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.049473235762139e-07,
+      "loss": 0.6562,
+      "step": 25393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.045457957286017e-07,
+      "loss": 0.633,
+      "step": 25394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.041443780903589e-07,
+      "loss": 0.6297,
+      "step": 25395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.03743070666243e-07,
+      "loss": 0.6434,
+      "step": 25396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.033418734610098e-07,
+      "loss": 0.6031,
+      "step": 25397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.029407864794235e-07,
+      "loss": 0.6155,
+      "step": 25398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.025398097262326e-07,
+      "loss": 0.6106,
+      "step": 25399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.021389432061942e-07,
+      "loss": 0.638,
+      "step": 25400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.017381869240625e-07,
+      "loss": 0.6377,
+      "step": 25401
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.013375408845846e-07,
+      "loss": 0.6761,
+      "step": 25402
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.009370050925113e-07,
+      "loss": 0.6484,
+      "step": 25403
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.005365795525965e-07,
+      "loss": 0.6598,
+      "step": 25404
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.001362642695831e-07,
+      "loss": 0.6475,
+      "step": 25405
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.997360592482183e-07,
+      "loss": 0.6376,
+      "step": 25406
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.993359644932496e-07,
+      "loss": 0.6503,
+      "step": 25407
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.989359800094164e-07,
+      "loss": 0.611,
+      "step": 25408
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.985361058014617e-07,
+      "loss": 0.6331,
+      "step": 25409
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.981363418741316e-07,
+      "loss": 0.6645,
+      "step": 25410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.97736688232159e-07,
+      "loss": 0.6241,
+      "step": 25411
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.973371448802879e-07,
+      "loss": 0.6673,
+      "step": 25412
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.969377118232512e-07,
+      "loss": 0.6293,
+      "step": 25413
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.965383890657862e-07,
+      "loss": 0.6476,
+      "step": 25414
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.96139176612628e-07,
+      "loss": 0.6804,
+      "step": 25415
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.957400744685094e-07,
+      "loss": 0.6382,
+      "step": 25416
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.953410826381613e-07,
+      "loss": 0.6284,
+      "step": 25417
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.949422011263174e-07,
+      "loss": 0.6353,
+      "step": 25418
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.945434299377029e-07,
+      "loss": 0.6476,
+      "step": 25419
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.941447690770475e-07,
+      "loss": 0.6453,
+      "step": 25420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.937462185490773e-07,
+      "loss": 0.5936,
+      "step": 25421
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.933477783585185e-07,
+      "loss": 0.6392,
+      "step": 25422
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.929494485100952e-07,
+      "loss": 0.6715,
+      "step": 25423
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.925512290085313e-07,
+      "loss": 0.6459,
+      "step": 25424
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.92153119858544e-07,
+      "loss": 0.6378,
+      "step": 25425
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.917551210648577e-07,
+      "loss": 0.5975,
+      "step": 25426
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.913572326321883e-07,
+      "loss": 0.5736,
+      "step": 25427
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.909594545652542e-07,
+      "loss": 0.685,
+      "step": 25428
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.905617868687719e-07,
+      "loss": 0.5336,
+      "step": 25429
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.901642295474575e-07,
+      "loss": 0.6161,
+      "step": 25430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.897667826060217e-07,
+      "loss": 0.6303,
+      "step": 25431
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.893694460491785e-07,
+      "loss": 0.6201,
+      "step": 25432
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.889722198816384e-07,
+      "loss": 0.5966,
+      "step": 25433
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.885751041081112e-07,
+      "loss": 0.6443,
+      "step": 25434
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.881780987333053e-07,
+      "loss": 0.6133,
+      "step": 25435
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.877812037619291e-07,
+      "loss": 0.6374,
+      "step": 25436
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.873844191986845e-07,
+      "loss": 0.6635,
+      "step": 25437
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.869877450482787e-07,
+      "loss": 0.6174,
+      "step": 25438
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.865911813154146e-07,
+      "loss": 0.6038,
+      "step": 25439
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.861947280047932e-07,
+      "loss": 0.5863,
+      "step": 25440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.857983851211147e-07,
+      "loss": 0.6292,
+      "step": 25441
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.854021526690813e-07,
+      "loss": 0.6945,
+      "step": 25442
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.850060306533868e-07,
+      "loss": 0.6125,
+      "step": 25443
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.846100190787286e-07,
+      "loss": 0.6702,
+      "step": 25444
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.842141179498029e-07,
+      "loss": 0.6351,
+      "step": 25445
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.838183272713039e-07,
+      "loss": 0.6194,
+      "step": 25446
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.834226470479233e-07,
+      "loss": 0.5707,
+      "step": 25447
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.830270772843539e-07,
+      "loss": 0.577,
+      "step": 25448
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.826316179852832e-07,
+      "loss": 0.6552,
+      "step": 25449
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.822362691554018e-07,
+      "loss": 0.6449,
+      "step": 25450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.818410307993961e-07,
+      "loss": 0.6328,
+      "step": 25451
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.814459029219522e-07,
+      "loss": 0.6422,
+      "step": 25452
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.810508855277554e-07,
+      "loss": 0.593,
+      "step": 25453
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.806559786214906e-07,
+      "loss": 0.6255,
+      "step": 25454
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.802611822078365e-07,
+      "loss": 0.6442,
+      "step": 25455
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.798664962914758e-07,
+      "loss": 0.6278,
+      "step": 25456
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.794719208770873e-07,
+      "loss": 0.6156,
+      "step": 25457
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.790774559693503e-07,
+      "loss": 0.6372,
+      "step": 25458
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.786831015729412e-07,
+      "loss": 0.5904,
+      "step": 25459
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.782888576925362e-07,
+      "loss": 0.6616,
+      "step": 25460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.778947243328083e-07,
+      "loss": 0.6802,
+      "step": 25461
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.775007014984303e-07,
+      "loss": 0.6573,
+      "step": 25462
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.771067891940741e-07,
+      "loss": 0.6932,
+      "step": 25463
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.767129874244105e-07,
+      "loss": 0.6367,
+      "step": 25464
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.763192961941089e-07,
+      "loss": 0.6109,
+      "step": 25465
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.759257155078391e-07,
+      "loss": 0.6112,
+      "step": 25466
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.755322453702617e-07,
+      "loss": 0.6235,
+      "step": 25467
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.751388857860464e-07,
+      "loss": 0.6054,
+      "step": 25468
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.747456367598549e-07,
+      "loss": 0.683,
+      "step": 25469
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.7435249829635e-07,
+      "loss": 0.652,
+      "step": 25470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.739594704001939e-07,
+      "loss": 0.6328,
+      "step": 25471
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.735665530760472e-07,
+      "loss": 0.6368,
+      "step": 25472
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.731737463285648e-07,
+      "loss": 0.658,
+      "step": 25473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.727810501624066e-07,
+      "loss": 0.6112,
+      "step": 25474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.723884645822287e-07,
+      "loss": 0.6177,
+      "step": 25475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.719959895926842e-07,
+      "loss": 0.6016,
+      "step": 25476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.716036251984293e-07,
+      "loss": 0.6077,
+      "step": 25477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.712113714041124e-07,
+      "loss": 0.6403,
+      "step": 25478
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.708192282143855e-07,
+      "loss": 0.5909,
+      "step": 25479
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.704271956338981e-07,
+      "loss": 0.6898,
+      "step": 25480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.700352736672977e-07,
+      "loss": 0.6364,
+      "step": 25481
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.696434623192328e-07,
+      "loss": 0.5972,
+      "step": 25482
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.692517615943483e-07,
+      "loss": 0.7274,
+      "step": 25483
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.688601714972853e-07,
+      "loss": 0.6349,
+      "step": 25484
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.684686920326899e-07,
+      "loss": 0.6564,
+      "step": 25485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.68077323205203e-07,
+      "loss": 0.6546,
+      "step": 25486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.676860650194639e-07,
+      "loss": 0.6683,
+      "step": 25487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.672949174801124e-07,
+      "loss": 0.6095,
+      "step": 25488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.66903880591786e-07,
+      "loss": 0.7046,
+      "step": 25489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.665129543591198e-07,
+      "loss": 0.6077,
+      "step": 25490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.6612213878675e-07,
+      "loss": 0.6212,
+      "step": 25491
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.657314338793086e-07,
+      "loss": 0.6436,
+      "step": 25492
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.653408396414296e-07,
+      "loss": 0.6633,
+      "step": 25493
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.649503560777426e-07,
+      "loss": 0.6053,
+      "step": 25494
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.645599831928806e-07,
+      "loss": 0.6186,
+      "step": 25495
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.641697209914666e-07,
+      "loss": 0.6357,
+      "step": 25496
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.637795694781313e-07,
+      "loss": 0.6388,
+      "step": 25497
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.633895286574998e-07,
+      "loss": 0.6143,
+      "step": 25498
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.629995985341963e-07,
+      "loss": 0.6055,
+      "step": 25499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.626097791128427e-07,
+      "loss": 0.7086,
+      "step": 25500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.622200703980652e-07,
+      "loss": 0.5988,
+      "step": 25501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.618304723944791e-07,
+      "loss": 0.5994,
+      "step": 25502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.614409851067049e-07,
+      "loss": 0.6405,
+      "step": 25503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.610516085393615e-07,
+      "loss": 0.6429,
+      "step": 25504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.606623426970649e-07,
+      "loss": 0.6141,
+      "step": 25505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.602731875844314e-07,
+      "loss": 0.618,
+      "step": 25506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.598841432060754e-07,
+      "loss": 0.5665,
+      "step": 25507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.594952095666063e-07,
+      "loss": 0.6809,
+      "step": 25508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.591063866706371e-07,
+      "loss": 0.6431,
+      "step": 25509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.587176745227775e-07,
+      "loss": 0.6217,
+      "step": 25510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.583290731276381e-07,
+      "loss": 0.6589,
+      "step": 25511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.579405824898233e-07,
+      "loss": 0.5732,
+      "step": 25512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.575522026139436e-07,
+      "loss": 0.6722,
+      "step": 25513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.571639335045988e-07,
+      "loss": 0.6035,
+      "step": 25514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.567757751663939e-07,
+      "loss": 0.6306,
+      "step": 25515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.563877276039321e-07,
+      "loss": 0.6491,
+      "step": 25516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.55999790821813e-07,
+      "loss": 0.6308,
+      "step": 25517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.556119648246361e-07,
+      "loss": 0.6262,
+      "step": 25518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.552242496170025e-07,
+      "loss": 0.6523,
+      "step": 25519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.548366452035049e-07,
+      "loss": 0.6355,
+      "step": 25520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.544491515887397e-07,
+      "loss": 0.6362,
+      "step": 25521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.540617687773032e-07,
+      "loss": 0.6181,
+      "step": 25522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.536744967737862e-07,
+      "loss": 0.6326,
+      "step": 25523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.532873355827818e-07,
+      "loss": 0.6098,
+      "step": 25524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.529002852088817e-07,
+      "loss": 0.6403,
+      "step": 25525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.525133456566701e-07,
+      "loss": 0.5951,
+      "step": 25526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.521265169307378e-07,
+      "loss": 0.6179,
+      "step": 25527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.517397990356711e-07,
+      "loss": 0.6134,
+      "step": 25528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.513531919760552e-07,
+      "loss": 0.6513,
+      "step": 25529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.50966695756472e-07,
+      "loss": 0.6279,
+      "step": 25530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.505803103815078e-07,
+      "loss": 0.6543,
+      "step": 25531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.501940358557391e-07,
+      "loss": 0.6039,
+      "step": 25532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.498078721837476e-07,
+      "loss": 0.6489,
+      "step": 25533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.494218193701118e-07,
+      "loss": 0.6329,
+      "step": 25534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.490358774194094e-07,
+      "loss": 0.638,
+      "step": 25535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.486500463362167e-07,
+      "loss": 0.6725,
+      "step": 25536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.482643261251064e-07,
+      "loss": 0.5998,
+      "step": 25537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.47878716790652e-07,
+      "loss": 0.6821,
+      "step": 25538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.474932183374261e-07,
+      "loss": 0.6144,
+      "step": 25539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.471078307699985e-07,
+      "loss": 0.5891,
+      "step": 25540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.46722554092939e-07,
+      "loss": 0.6116,
+      "step": 25541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.463373883108182e-07,
+      "loss": 0.661,
+      "step": 25542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.459523334281981e-07,
+      "loss": 0.6156,
+      "step": 25543
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.45567389449645e-07,
+      "loss": 0.626,
+      "step": 25544
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.451825563797254e-07,
+      "loss": 0.6142,
+      "step": 25545
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.447978342229999e-07,
+      "loss": 0.6231,
+      "step": 25546
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.444132229840305e-07,
+      "loss": 0.6124,
+      "step": 25547
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.44028722667378e-07,
+      "loss": 0.6476,
+      "step": 25548
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.436443332775999e-07,
+      "loss": 0.6307,
+      "step": 25549
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.432600548192535e-07,
+      "loss": 0.6217,
+      "step": 25550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.428758872968977e-07,
+      "loss": 0.5846,
+      "step": 25551
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.424918307150807e-07,
+      "loss": 0.6698,
+      "step": 25552
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.421078850783635e-07,
+      "loss": 0.6117,
+      "step": 25553
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.417240503912958e-07,
+      "loss": 0.5705,
+      "step": 25554
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.413403266584262e-07,
+      "loss": 0.5795,
+      "step": 25555
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.409567138843053e-07,
+      "loss": 0.6509,
+      "step": 25556
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.405732120734842e-07,
+      "loss": 0.6448,
+      "step": 25557
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.401898212305036e-07,
+      "loss": 0.6788,
+      "step": 25558
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.398065413599153e-07,
+      "loss": 0.6242,
+      "step": 25559
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.394233724662624e-07,
+      "loss": 0.6168,
+      "step": 25560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.390403145540847e-07,
+      "loss": 0.6373,
+      "step": 25561
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.386573676279262e-07,
+      "loss": 0.6205,
+      "step": 25562
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.382745316923278e-07,
+      "loss": 0.5781,
+      "step": 25563
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.378918067518247e-07,
+      "loss": 0.6388,
+      "step": 25564
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.375091928109589e-07,
+      "loss": 0.6197,
+      "step": 25565
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.371266898742667e-07,
+      "loss": 0.6916,
+      "step": 25566
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.367442979462802e-07,
+      "loss": 0.6022,
+      "step": 25567
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.363620170315355e-07,
+      "loss": 0.6557,
+      "step": 25568
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.359798471345646e-07,
+      "loss": 0.6249,
+      "step": 25569
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.355977882598951e-07,
+      "loss": 0.6427,
+      "step": 25570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.352158404120612e-07,
+      "loss": 0.6462,
+      "step": 25571
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.348340035955924e-07,
+      "loss": 0.6797,
+      "step": 25572
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.344522778150109e-07,
+      "loss": 0.6377,
+      "step": 25573
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.340706630748461e-07,
+      "loss": 0.6811,
+      "step": 25574
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.336891593796224e-07,
+      "loss": 0.6222,
+      "step": 25575
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.333077667338594e-07,
+      "loss": 0.6685,
+      "step": 25576
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.329264851420836e-07,
+      "loss": 0.7011,
+      "step": 25577
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.325453146088145e-07,
+      "loss": 0.6288,
+      "step": 25578
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.321642551385698e-07,
+      "loss": 0.6253,
+      "step": 25579
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.317833067358681e-07,
+      "loss": 0.6067,
+      "step": 25580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.314024694052278e-07,
+      "loss": 0.5988,
+      "step": 25581
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.310217431511589e-07,
+      "loss": 0.5846,
+      "step": 25582
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.30641127978181e-07,
+      "loss": 0.6613,
+      "step": 25583
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.30260623890806e-07,
+      "loss": 0.6265,
+      "step": 25584
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.298802308935437e-07,
+      "loss": 0.576,
+      "step": 25585
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.294999489909026e-07,
+      "loss": 0.6367,
+      "step": 25586
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.29119778187397e-07,
+      "loss": 0.6185,
+      "step": 25587
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.287397184875255e-07,
+      "loss": 0.5698,
+      "step": 25588
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.283597698958022e-07,
+      "loss": 0.6336,
+      "step": 25589
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.279799324167302e-07,
+      "loss": 0.6633,
+      "step": 25590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.276002060548103e-07,
+      "loss": 0.6295,
+      "step": 25591
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.272205908145467e-07,
+      "loss": 0.621,
+      "step": 25592
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.268410867004404e-07,
+      "loss": 0.6016,
+      "step": 25593
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.264616937169865e-07,
+      "loss": 0.5446,
+      "step": 25594
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.260824118686892e-07,
+      "loss": 0.6492,
+      "step": 25595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.25703241160045e-07,
+      "loss": 0.6336,
+      "step": 25596
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.253241815955446e-07,
+      "loss": 0.6413,
+      "step": 25597
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.249452331796879e-07,
+      "loss": 0.6397,
+      "step": 25598
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.245663959169623e-07,
+      "loss": 0.6143,
+      "step": 25599
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.24187669811861e-07,
+      "loss": 0.6001,
+      "step": 25600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.23809054868878e-07,
+      "loss": 0.6252,
+      "step": 25601
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.234305510924988e-07,
+      "loss": 0.627,
+      "step": 25602
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.230521584872118e-07,
+      "loss": 0.6637,
+      "step": 25603
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.226738770575047e-07,
+      "loss": 0.6092,
+      "step": 25604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.222957068078594e-07,
+      "loss": 0.6382,
+      "step": 25605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.219176477427602e-07,
+      "loss": 0.6132,
+      "step": 25606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.215396998666933e-07,
+      "loss": 0.5941,
+      "step": 25607
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.211618631841355e-07,
+      "loss": 0.6484,
+      "step": 25608
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.207841376995682e-07,
+      "loss": 0.6089,
+      "step": 25609
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.204065234174717e-07,
+      "loss": 0.6589,
+      "step": 25610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.200290203423187e-07,
+      "loss": 0.6246,
+      "step": 25611
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.19651628478587e-07,
+      "loss": 0.597,
+      "step": 25612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.192743478307528e-07,
+      "loss": 0.6214,
+      "step": 25613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.188971784032882e-07,
+      "loss": 0.6313,
+      "step": 25614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.185201202006641e-07,
+      "loss": 0.6344,
+      "step": 25615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.181431732273524e-07,
+      "loss": 0.6452,
+      "step": 25616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.177663374878207e-07,
+      "loss": 0.5987,
+      "step": 25617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.173896129865364e-07,
+      "loss": 0.6005,
+      "step": 25618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.170129997279695e-07,
+      "loss": 0.6408,
+      "step": 25619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.166364977165817e-07,
+      "loss": 0.6019,
+      "step": 25620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.162601069568385e-07,
+      "loss": 0.7069,
+      "step": 25621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.158838274532042e-07,
+      "loss": 0.6628,
+      "step": 25622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.155076592101361e-07,
+      "loss": 0.6185,
+      "step": 25623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.151316022320941e-07,
+      "loss": 0.6093,
+      "step": 25624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.147556565235424e-07,
+      "loss": 0.6802,
+      "step": 25625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.14379822088933e-07,
+      "loss": 0.6757,
+      "step": 25626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.140040989327245e-07,
+      "loss": 0.6332,
+      "step": 25627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.136284870593711e-07,
+      "loss": 0.6657,
+      "step": 25628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.13252986473325e-07,
+      "loss": 0.5919,
+      "step": 25629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.128775971790379e-07,
+      "loss": 0.6482,
+      "step": 25630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.125023191809643e-07,
+      "loss": 0.5886,
+      "step": 25631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.121271524835493e-07,
+      "loss": 0.6969,
+      "step": 25632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.117520970912427e-07,
+      "loss": 0.6723,
+      "step": 25633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.113771530084933e-07,
+      "loss": 0.6055,
+      "step": 25634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.110023202397419e-07,
+      "loss": 0.6506,
+      "step": 25635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.106275987894339e-07,
+      "loss": 0.6429,
+      "step": 25636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.102529886620167e-07,
+      "loss": 0.6118,
+      "step": 25637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.098784898619269e-07,
+      "loss": 0.6479,
+      "step": 25638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.095041023936055e-07,
+      "loss": 0.6124,
+      "step": 25639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.091298262614942e-07,
+      "loss": 0.6558,
+      "step": 25640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.087556614700263e-07,
+      "loss": 0.6401,
+      "step": 25641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.083816080236394e-07,
+      "loss": 0.6129,
+      "step": 25642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.080076659267708e-07,
+      "loss": 0.6294,
+      "step": 25643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.076338351838518e-07,
+      "loss": 0.6335,
+      "step": 25644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.072601157993141e-07,
+      "loss": 0.6726,
+      "step": 25645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.068865077775921e-07,
+      "loss": 0.6426,
+      "step": 25646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.06513011123111e-07,
+      "loss": 0.6364,
+      "step": 25647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.061396258402996e-07,
+      "loss": 0.6206,
+      "step": 25648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.057663519335888e-07,
+      "loss": 0.607,
+      "step": 25649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.053931894074005e-07,
+      "loss": 0.6324,
+      "step": 25650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.050201382661602e-07,
+      "loss": 0.6237,
+      "step": 25651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.046471985142922e-07,
+      "loss": 0.6848,
+      "step": 25652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.042743701562149e-07,
+      "loss": 0.6383,
+      "step": 25653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.039016531963504e-07,
+      "loss": 0.6536,
+      "step": 25654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.035290476391198e-07,
+      "loss": 0.6592,
+      "step": 25655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.031565534889372e-07,
+      "loss": 0.61,
+      "step": 25656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.027841707502214e-07,
+      "loss": 0.6253,
+      "step": 25657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.024118994273887e-07,
+      "loss": 0.6435,
+      "step": 25658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.02039739524849e-07,
+      "loss": 0.6429,
+      "step": 25659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.016676910470165e-07,
+      "loss": 0.648,
+      "step": 25660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.012957539983022e-07,
+      "loss": 0.628,
+      "step": 25661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.009239283831169e-07,
+      "loss": 0.6271,
+      "step": 25662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.005522142058696e-07,
+      "loss": 0.6378,
+      "step": 25663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.00180611470964e-07,
+      "loss": 0.63,
+      "step": 25664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.998091201828093e-07,
+      "loss": 0.6611,
+      "step": 25665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.994377403458084e-07,
+      "loss": 0.6444,
+      "step": 25666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.990664719643646e-07,
+      "loss": 0.6645,
+      "step": 25667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.986953150428798e-07,
+      "loss": 0.6516,
+      "step": 25668
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.983242695857561e-07,
+      "loss": 0.6347,
+      "step": 25669
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.979533355973898e-07,
+      "loss": 0.6057,
+      "step": 25670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.97582513082181e-07,
+      "loss": 0.6338,
+      "step": 25671
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.972118020445261e-07,
+      "loss": 0.5813,
+      "step": 25672
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.968412024888182e-07,
+      "loss": 0.6765,
+      "step": 25673
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.964707144194548e-07,
+      "loss": 0.6371,
+      "step": 25674
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.96100337840827e-07,
+      "loss": 0.6045,
+      "step": 25675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.957300727573245e-07,
+      "loss": 0.6486,
+      "step": 25676
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.953599191733383e-07,
+      "loss": 0.6571,
+      "step": 25677
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.949898770932572e-07,
+      "loss": 0.6265,
+      "step": 25678
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.946199465214686e-07,
+      "loss": 0.6543,
+      "step": 25679
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.94250127462358e-07,
+      "loss": 0.6711,
+      "step": 25680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.938804199203118e-07,
+      "loss": 0.5772,
+      "step": 25681
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.935108238997112e-07,
+      "loss": 0.6391,
+      "step": 25682
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.93141339404939e-07,
+      "loss": 0.6228,
+      "step": 25683
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.927719664403753e-07,
+      "loss": 0.6256,
+      "step": 25684
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.924027050104008e-07,
+      "loss": 0.6207,
+      "step": 25685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.920335551193934e-07,
+      "loss": 0.6405,
+      "step": 25686
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.916645167717294e-07,
+      "loss": 0.6877,
+      "step": 25687
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.912955899717842e-07,
+      "loss": 0.6397,
+      "step": 25688
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.909267747239312e-07,
+      "loss": 0.6398,
+      "step": 25689
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.905580710325432e-07,
+      "loss": 0.6513,
+      "step": 25690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.901894789019936e-07,
+      "loss": 0.605,
+      "step": 25691
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.898209983366498e-07,
+      "loss": 0.636,
+      "step": 25692
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.894526293408842e-07,
+      "loss": 0.6234,
+      "step": 25693
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.890843719190609e-07,
+      "loss": 0.5711,
+      "step": 25694
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.887162260755463e-07,
+      "loss": 0.624,
+      "step": 25695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.883481918147072e-07,
+      "loss": 0.6634,
+      "step": 25696
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.879802691409053e-07,
+      "loss": 0.6539,
+      "step": 25697
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.876124580585041e-07,
+      "loss": 0.6227,
+      "step": 25698
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.872447585718655e-07,
+      "loss": 0.6575,
+      "step": 25699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.868771706853449e-07,
+      "loss": 0.6261,
+      "step": 25700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.865096944033044e-07,
+      "loss": 0.6546,
+      "step": 25701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.861423297300994e-07,
+      "loss": 0.6566,
+      "step": 25702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.857750766700854e-07,
+      "loss": 0.6151,
+      "step": 25703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.854079352276165e-07,
+      "loss": 0.6379,
+      "step": 25704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.850409054070483e-07,
+      "loss": 0.6048,
+      "step": 25705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.846739872127283e-07,
+      "loss": 0.6045,
+      "step": 25706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.843071806490086e-07,
+      "loss": 0.6652,
+      "step": 25707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.83940485720238e-07,
+      "loss": 0.6398,
+      "step": 25708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.835739024307652e-07,
+      "loss": 0.645,
+      "step": 25709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.832074307849345e-07,
+      "loss": 0.6685,
+      "step": 25710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.828410707870935e-07,
+      "loss": 0.675,
+      "step": 25711
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.824748224415832e-07,
+      "loss": 0.6159,
+      "step": 25712
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.821086857527481e-07,
+      "loss": 0.6534,
+      "step": 25713
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.817426607249277e-07,
+      "loss": 0.6677,
+      "step": 25714
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.81376747362461e-07,
+      "loss": 0.6827,
+      "step": 25715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.810109456696888e-07,
+      "loss": 0.6538,
+      "step": 25716
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.806452556509478e-07,
+      "loss": 0.679,
+      "step": 25717
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.802796773105723e-07,
+      "loss": 0.6429,
+      "step": 25718
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.799142106528965e-07,
+      "loss": 0.6861,
+      "step": 25719
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.795488556822537e-07,
+      "loss": 0.6155,
+      "step": 25720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.79183612402977e-07,
+      "loss": 0.6601,
+      "step": 25721
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.788184808193975e-07,
+      "loss": 0.6607,
+      "step": 25722
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.784534609358405e-07,
+      "loss": 0.6316,
+      "step": 25723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.780885527566371e-07,
+      "loss": 0.6394,
+      "step": 25724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.777237562861115e-07,
+      "loss": 0.6111,
+      "step": 25725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.773590715285915e-07,
+      "loss": 0.5919,
+      "step": 25726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.769944984883979e-07,
+      "loss": 0.6492,
+      "step": 25727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.766300371698574e-07,
+      "loss": 0.6121,
+      "step": 25728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.762656875772866e-07,
+      "loss": 0.5878,
+      "step": 25729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.759014497150062e-07,
+      "loss": 0.6208,
+      "step": 25730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.755373235873363e-07,
+      "loss": 0.6682,
+      "step": 25731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.751733091985934e-07,
+      "loss": 0.6901,
+      "step": 25732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.74809406553094e-07,
+      "loss": 0.6075,
+      "step": 25733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.744456156551526e-07,
+      "loss": 0.6428,
+      "step": 25734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.74081936509081e-07,
+      "loss": 0.616,
+      "step": 25735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.737183691191917e-07,
+      "loss": 0.5731,
+      "step": 25736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.733549134897953e-07,
+      "loss": 0.6594,
+      "step": 25737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.729915696252008e-07,
+      "loss": 0.6627,
+      "step": 25738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.726283375297171e-07,
+      "loss": 0.6037,
+      "step": 25739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.722652172076514e-07,
+      "loss": 0.634,
+      "step": 25740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.719022086633064e-07,
+      "loss": 0.6092,
+      "step": 25741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.715393119009871e-07,
+      "loss": 0.6555,
+      "step": 25742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.711765269249969e-07,
+      "loss": 0.6827,
+      "step": 25743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.708138537396357e-07,
+      "loss": 0.6405,
+      "step": 25744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.704512923492045e-07,
+      "loss": 0.5995,
+      "step": 25745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.700888427580031e-07,
+      "loss": 0.7173,
+      "step": 25746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.69726504970326e-07,
+      "loss": 0.6303,
+      "step": 25747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.693642789904708e-07,
+      "loss": 0.5947,
+      "step": 25748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.690021648227306e-07,
+      "loss": 0.599,
+      "step": 25749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.686401624714011e-07,
+      "loss": 0.6534,
+      "step": 25750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.682782719407731e-07,
+      "loss": 0.652,
+      "step": 25751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.679164932351389e-07,
+      "loss": 0.5661,
+      "step": 25752
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.675548263587848e-07,
+      "loss": 0.6345,
+      "step": 25753
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.671932713160011e-07,
+      "loss": 0.5846,
+      "step": 25754
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.66831828111073e-07,
+      "loss": 0.6441,
+      "step": 25755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.664704967482859e-07,
+      "loss": 0.6204,
+      "step": 25756
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.661092772319254e-07,
+      "loss": 0.627,
+      "step": 25757
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.657481695662747e-07,
+      "loss": 0.5896,
+      "step": 25758
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.653871737556127e-07,
+      "loss": 0.6036,
+      "step": 25759
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.650262898042203e-07,
+      "loss": 0.6066,
+      "step": 25760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.646655177163762e-07,
+      "loss": 0.6562,
+      "step": 25761
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.643048574963583e-07,
+      "loss": 0.6378,
+      "step": 25762
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.63944309148442e-07,
+      "loss": 0.6386,
+      "step": 25763
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.635838726769049e-07,
+      "loss": 0.6051,
+      "step": 25764
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.632235480860171e-07,
+      "loss": 0.6179,
+      "step": 25765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.628633353800506e-07,
+      "loss": 0.6142,
+      "step": 25766
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.625032345632775e-07,
+      "loss": 0.6284,
+      "step": 25767
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.621432456399679e-07,
+      "loss": 0.6,
+      "step": 25768
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.617833686143881e-07,
+      "loss": 0.5961,
+      "step": 25769
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.614236034908083e-07,
+      "loss": 0.6551,
+      "step": 25770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.610639502734894e-07,
+      "loss": 0.6023,
+      "step": 25771
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.607044089666969e-07,
+      "loss": 0.6196,
+      "step": 25772
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.603449795746962e-07,
+      "loss": 0.6608,
+      "step": 25773
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.599856621017452e-07,
+      "loss": 0.6432,
+      "step": 25774
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.59626456552107e-07,
+      "loss": 0.6241,
+      "step": 25775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.592673629300404e-07,
+      "loss": 0.6158,
+      "step": 25776
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.58908381239801e-07,
+      "loss": 0.6005,
+      "step": 25777
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.585495114856443e-07,
+      "loss": 0.6726,
+      "step": 25778
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.581907536718278e-07,
+      "loss": 0.6558,
+      "step": 25779
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.578321078026039e-07,
+      "loss": 0.6445,
+      "step": 25780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.574735738822245e-07,
+      "loss": 0.6431,
+      "step": 25781
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.571151519149421e-07,
+      "loss": 0.6116,
+      "step": 25782
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.56756841905004e-07,
+      "loss": 0.6036,
+      "step": 25783
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.563986438566582e-07,
+      "loss": 0.6106,
+      "step": 25784
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.560405577741535e-07,
+      "loss": 0.6708,
+      "step": 25785
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.556825836617341e-07,
+      "loss": 0.6378,
+      "step": 25786
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.553247215236468e-07,
+      "loss": 0.6977,
+      "step": 25787
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.549669713641314e-07,
+      "loss": 0.6068,
+      "step": 25788
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.546093331874292e-07,
+      "loss": 0.6238,
+      "step": 25789
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.542518069977843e-07,
+      "loss": 0.6401,
+      "step": 25790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5389439279943e-07,
+      "loss": 0.5721,
+      "step": 25791
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.535370905966087e-07,
+      "loss": 0.6449,
+      "step": 25792
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.531799003935567e-07,
+      "loss": 0.6515,
+      "step": 25793
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.528228221945054e-07,
+      "loss": 0.5563,
+      "step": 25794
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.524658560036899e-07,
+      "loss": 0.6255,
+      "step": 25795
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.52109001825345e-07,
+      "loss": 0.599,
+      "step": 25796
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.517522596636971e-07,
+      "loss": 0.6201,
+      "step": 25797
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.513956295229783e-07,
+      "loss": 0.5871,
+      "step": 25798
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.510391114074199e-07,
+      "loss": 0.6519,
+      "step": 25799
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.506827053212427e-07,
+      "loss": 0.6491,
+      "step": 25800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.503264112686768e-07,
+      "loss": 0.6689,
+      "step": 25801
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.499702292539455e-07,
+      "loss": 0.6641,
+      "step": 25802
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.496141592812698e-07,
+      "loss": 0.6166,
+      "step": 25803
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.492582013548742e-07,
+      "loss": 0.6044,
+      "step": 25804
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.489023554789785e-07,
+      "loss": 0.6845,
+      "step": 25805
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.485466216578005e-07,
+      "loss": 0.6799,
+      "step": 25806
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.481909998955582e-07,
+      "loss": 0.7005,
+      "step": 25807
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.478354901964688e-07,
+      "loss": 0.6456,
+      "step": 25808
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.47480092564745e-07,
+      "loss": 0.6083,
+      "step": 25809
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.471248070046031e-07,
+      "loss": 0.632,
+      "step": 25810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.467696335202566e-07,
+      "loss": 0.6711,
+      "step": 25811
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.46414572115912e-07,
+      "loss": 0.7065,
+      "step": 25812
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.460596227957826e-07,
+      "loss": 0.5789,
+      "step": 25813
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.457047855640762e-07,
+      "loss": 0.6269,
+      "step": 25814
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.453500604249973e-07,
+      "loss": 0.5582,
+      "step": 25815
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.449954473827546e-07,
+      "loss": 0.6377,
+      "step": 25816
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.446409464415525e-07,
+      "loss": 0.6085,
+      "step": 25817
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.442865576055922e-07,
+      "loss": 0.618,
+      "step": 25818
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.439322808790759e-07,
+      "loss": 0.5865,
+      "step": 25819
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.435781162662068e-07,
+      "loss": 0.6773,
+      "step": 25820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.432240637711772e-07,
+      "loss": 0.6191,
+      "step": 25821
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.428701233981914e-07,
+      "loss": 0.6437,
+      "step": 25822
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.425162951514451e-07,
+      "loss": 0.6074,
+      "step": 25823
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.421625790351303e-07,
+      "loss": 0.6124,
+      "step": 25824
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.418089750534428e-07,
+      "loss": 0.6045,
+      "step": 25825
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.414554832105756e-07,
+      "loss": 0.6393,
+      "step": 25826
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.411021035107167e-07,
+      "loss": 0.6986,
+      "step": 25827
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.407488359580582e-07,
+      "loss": 0.6447,
+      "step": 25828
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.403956805567901e-07,
+      "loss": 0.63,
+      "step": 25829
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.400426373110968e-07,
+      "loss": 0.6441,
+      "step": 25830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.396897062251638e-07,
+      "loss": 0.5785,
+      "step": 25831
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.39336887303179e-07,
+      "loss": 0.6232,
+      "step": 25832
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.389841805493201e-07,
+      "loss": 0.5984,
+      "step": 25833
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.386315859677737e-07,
+      "loss": 0.6177,
+      "step": 25834
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.3827910356272e-07,
+      "loss": 0.6214,
+      "step": 25835
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.379267333383342e-07,
+      "loss": 0.594,
+      "step": 25836
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.375744752987976e-07,
+      "loss": 0.6069,
+      "step": 25837
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.37222329448287e-07,
+      "loss": 0.6467,
+      "step": 25838
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.368702957909733e-07,
+      "loss": 0.6284,
+      "step": 25839
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.365183743310343e-07,
+      "loss": 0.6555,
+      "step": 25840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.361665650726422e-07,
+      "loss": 0.6459,
+      "step": 25841
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.358148680199671e-07,
+      "loss": 0.6609,
+      "step": 25842
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.354632831771778e-07,
+      "loss": 0.6621,
+      "step": 25843
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.351118105484466e-07,
+      "loss": 0.6544,
+      "step": 25844
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.347604501379344e-07,
+      "loss": 0.6457,
+      "step": 25845
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.344092019498126e-07,
+      "loss": 0.6619,
+      "step": 25846
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.340580659882444e-07,
+      "loss": 0.6587,
+      "step": 25847
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.337070422573909e-07,
+      "loss": 0.6156,
+      "step": 25848
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.333561307614177e-07,
+      "loss": 0.6508,
+      "step": 25849
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.330053315044814e-07,
+      "loss": 0.6379,
+      "step": 25850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.326546444907399e-07,
+      "loss": 0.6842,
+      "step": 25851
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.323040697243587e-07,
+      "loss": 0.6065,
+      "step": 25852
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.319536072094866e-07,
+      "loss": 0.683,
+      "step": 25853
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.316032569502816e-07,
+      "loss": 0.6893,
+      "step": 25854
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.312530189508991e-07,
+      "loss": 0.6259,
+      "step": 25855
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.309028932154892e-07,
+      "loss": 0.6945,
+      "step": 25856
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.30552879748203e-07,
+      "loss": 0.6038,
+      "step": 25857
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.302029785531926e-07,
+      "loss": 0.576,
+      "step": 25858
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.298531896346048e-07,
+      "loss": 0.5775,
+      "step": 25859
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.295035129965875e-07,
+      "loss": 0.6178,
+      "step": 25860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.291539486432873e-07,
+      "loss": 0.5827,
+      "step": 25861
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.288044965788464e-07,
+      "loss": 0.5917,
+      "step": 25862
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.284551568074081e-07,
+      "loss": 0.7071,
+      "step": 25863
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.281059293331181e-07,
+      "loss": 0.6477,
+      "step": 25864
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.27756814160113e-07,
+      "loss": 0.6174,
+      "step": 25865
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.27407811292533e-07,
+      "loss": 0.6139,
+      "step": 25866
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.270589207345178e-07,
+      "loss": 0.5989,
+      "step": 25867
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.267101424902011e-07,
+      "loss": 0.6232,
+      "step": 25868
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.263614765637182e-07,
+      "loss": 0.6566,
+      "step": 25869
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.26012922959207e-07,
+      "loss": 0.657,
+      "step": 25870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.256644816807954e-07,
+      "loss": 0.6106,
+      "step": 25871
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.253161527326167e-07,
+      "loss": 0.6589,
+      "step": 25872
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.249679361188021e-07,
+      "loss": 0.6046,
+      "step": 25873
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.24619831843477e-07,
+      "loss": 0.624,
+      "step": 25874
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.242718399107693e-07,
+      "loss": 0.6633,
+      "step": 25875
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.239239603248081e-07,
+      "loss": 0.7111,
+      "step": 25876
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.235761930897143e-07,
+      "loss": 0.6289,
+      "step": 25877
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.232285382096125e-07,
+      "loss": 0.6401,
+      "step": 25878
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.22880995688626e-07,
+      "loss": 0.6317,
+      "step": 25879
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.225335655308728e-07,
+      "loss": 0.6217,
+      "step": 25880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.221862477404716e-07,
+      "loss": 0.5649,
+      "step": 25881
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.218390423215435e-07,
+      "loss": 0.6446,
+      "step": 25882
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.214919492782033e-07,
+      "loss": 0.6733,
+      "step": 25883
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.211449686145664e-07,
+      "loss": 0.6524,
+      "step": 25884
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.207981003347484e-07,
+      "loss": 0.6319,
+      "step": 25885
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.204513444428583e-07,
+      "loss": 0.5774,
+      "step": 25886
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.201047009430082e-07,
+      "loss": 0.593,
+      "step": 25887
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.197581698393117e-07,
+      "loss": 0.6107,
+      "step": 25888
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.194117511358732e-07,
+      "loss": 0.6313,
+      "step": 25889
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.190654448368016e-07,
+      "loss": 0.6795,
+      "step": 25890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.187192509462035e-07,
+      "loss": 0.6542,
+      "step": 25891
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.183731694681815e-07,
+      "loss": 0.5801,
+      "step": 25892
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.180272004068398e-07,
+      "loss": 0.5815,
+      "step": 25893
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.176813437662798e-07,
+      "loss": 0.6143,
+      "step": 25894
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.173355995506036e-07,
+      "loss": 0.62,
+      "step": 25895
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.16989967763909e-07,
+      "loss": 0.606,
+      "step": 25896
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.166444484102961e-07,
+      "loss": 0.6287,
+      "step": 25897
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.162990414938585e-07,
+      "loss": 0.6178,
+      "step": 25898
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.159537470186926e-07,
+      "loss": 0.6546,
+      "step": 25899
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.15608564988892e-07,
+      "loss": 0.632,
+      "step": 25900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1526349540855e-07,
+      "loss": 0.6231,
+      "step": 25901
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.149185382817568e-07,
+      "loss": 0.6766,
+      "step": 25902
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.145736936126056e-07,
+      "loss": 0.6426,
+      "step": 25903
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.142289614051799e-07,
+      "loss": 0.6058,
+      "step": 25904
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.138843416635708e-07,
+      "loss": 0.596,
+      "step": 25905
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.135398343918618e-07,
+      "loss": 0.6186,
+      "step": 25906
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.131954395941385e-07,
+      "loss": 0.6078,
+      "step": 25907
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.128511572744843e-07,
+      "loss": 0.6287,
+      "step": 25908
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.125069874369826e-07,
+      "loss": 0.6661,
+      "step": 25909
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1216293008571e-07,
+      "loss": 0.6739,
+      "step": 25910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.11818985224749e-07,
+      "loss": 0.5998,
+      "step": 25911
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.114751528581763e-07,
+      "loss": 0.6331,
+      "step": 25912
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.111314329900685e-07,
+      "loss": 0.6766,
+      "step": 25913
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.107878256245025e-07,
+      "loss": 0.628,
+      "step": 25914
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.104443307655493e-07,
+      "loss": 0.62,
+      "step": 25915
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.101009484172836e-07,
+      "loss": 0.6053,
+      "step": 25916
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.097576785837754e-07,
+      "loss": 0.6443,
+      "step": 25917
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.094145212690948e-07,
+      "loss": 0.6495,
+      "step": 25918
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.090714764773108e-07,
+      "loss": 0.6481,
+      "step": 25919
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.087285442124923e-07,
+      "loss": 0.6156,
+      "step": 25920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.083857244787027e-07,
+      "loss": 0.6313,
+      "step": 25921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.080430172800066e-07,
+      "loss": 0.6426,
+      "step": 25922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.077004226204674e-07,
+      "loss": 0.6538,
+      "step": 25923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.073579405041473e-07,
+      "loss": 0.5875,
+      "step": 25924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.070155709351077e-07,
+      "loss": 0.6243,
+      "step": 25925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.066733139174096e-07,
+      "loss": 0.6149,
+      "step": 25926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.063311694551054e-07,
+      "loss": 0.6435,
+      "step": 25927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.059891375522552e-07,
+      "loss": 0.5963,
+      "step": 25928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.056472182129146e-07,
+      "loss": 0.6334,
+      "step": 25929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.05305411441136e-07,
+      "loss": 0.6883,
+      "step": 25930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.049637172409727e-07,
+      "loss": 0.6271,
+      "step": 25931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.046221356164771e-07,
+      "loss": 0.6193,
+      "step": 25932
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.042806665716959e-07,
+      "loss": 0.6527,
+      "step": 25933
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.039393101106805e-07,
+      "loss": 0.6341,
+      "step": 25934
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.035980662374762e-07,
+      "loss": 0.648,
+      "step": 25935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.032569349561312e-07,
+      "loss": 0.6398,
+      "step": 25936
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.029159162706876e-07,
+      "loss": 0.6415,
+      "step": 25937
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.025750101851912e-07,
+      "loss": 0.6441,
+      "step": 25938
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.022342167036808e-07,
+      "loss": 0.6274,
+      "step": 25939
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.018935358301991e-07,
+      "loss": 0.6244,
+      "step": 25940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.015529675687847e-07,
+      "loss": 0.6527,
+      "step": 25941
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.012125119234745e-07,
+      "loss": 0.652,
+      "step": 25942
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.008721688983065e-07,
+      "loss": 0.63,
+      "step": 25943
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.005319384973173e-07,
+      "loss": 0.6343,
+      "step": 25944
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.001918207245371e-07,
+      "loss": 0.6338,
+      "step": 25945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.998518155840004e-07,
+      "loss": 0.6252,
+      "step": 25946
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.995119230797373e-07,
+      "loss": 0.639,
+      "step": 25947
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.991721432157792e-07,
+      "loss": 0.6532,
+      "step": 25948
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.98832475996155e-07,
+      "loss": 0.6263,
+      "step": 25949
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.984929214248902e-07,
+      "loss": 0.6179,
+      "step": 25950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.981534795060116e-07,
+      "loss": 0.5916,
+      "step": 25951
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.978141502435419e-07,
+      "loss": 0.6373,
+      "step": 25952
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.974749336415074e-07,
+      "loss": 0.6314,
+      "step": 25953
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.971358297039275e-07,
+      "loss": 0.6063,
+      "step": 25954
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.967968384348232e-07,
+      "loss": 0.6739,
+      "step": 25955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.964579598382158e-07,
+      "loss": 0.6466,
+      "step": 25956
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.961191939181198e-07,
+      "loss": 0.641,
+      "step": 25957
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.957805406785532e-07,
+      "loss": 0.6582,
+      "step": 25958
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.954420001235305e-07,
+      "loss": 0.6234,
+      "step": 25959
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.951035722570674e-07,
+      "loss": 0.595,
+      "step": 25960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.947652570831751e-07,
+      "loss": 0.6229,
+      "step": 25961
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.94427054605866e-07,
+      "loss": 0.595,
+      "step": 25962
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.940889648291469e-07,
+      "loss": 0.6242,
+      "step": 25963
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.93750987757029e-07,
+      "loss": 0.6121,
+      "step": 25964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.93413123393518e-07,
+      "loss": 0.5855,
+      "step": 25965
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.930753717426207e-07,
+      "loss": 0.6697,
+      "step": 25966
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.927377328083405e-07,
+      "loss": 0.6569,
+      "step": 25967
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.924002065946831e-07,
+      "loss": 0.5902,
+      "step": 25968
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.920627931056465e-07,
+      "loss": 0.6626,
+      "step": 25969
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.917254923452341e-07,
+      "loss": 0.6213,
+      "step": 25970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.913883043174439e-07,
+      "loss": 0.6977,
+      "step": 25971
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.910512290262736e-07,
+      "loss": 0.6007,
+      "step": 25972
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.907142664757203e-07,
+      "loss": 0.6285,
+      "step": 25973
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.903774166697783e-07,
+      "loss": 0.6039,
+      "step": 25974
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.900406796124413e-07,
+      "loss": 0.6344,
+      "step": 25975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.897040553077025e-07,
+      "loss": 0.587,
+      "step": 25976
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.893675437595535e-07,
+      "loss": 0.6184,
+      "step": 25977
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.890311449719821e-07,
+      "loss": 0.6023,
+      "step": 25978
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.886948589489792e-07,
+      "loss": 0.5906,
+      "step": 25979
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8835868569453e-07,
+      "loss": 0.6377,
+      "step": 25980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.880226252126208e-07,
+      "loss": 0.6366,
+      "step": 25981
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.876866775072353e-07,
+      "loss": 0.6026,
+      "step": 25982
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.873508425823581e-07,
+      "loss": 0.6318,
+      "step": 25983
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.870151204419704e-07,
+      "loss": 0.6481,
+      "step": 25984
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.866795110900535e-07,
+      "loss": 0.681,
+      "step": 25985
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.863440145305842e-07,
+      "loss": 0.6472,
+      "step": 25986
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.860086307675404e-07,
+      "loss": 0.6813,
+      "step": 25987
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.856733598049013e-07,
+      "loss": 0.5799,
+      "step": 25988
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.85338201646639e-07,
+      "loss": 0.6093,
+      "step": 25989
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.850031562967284e-07,
+      "loss": 0.6121,
+      "step": 25990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.846682237591438e-07,
+      "loss": 0.6422,
+      "step": 25991
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.843334040378522e-07,
+      "loss": 0.6109,
+      "step": 25992
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.83998697136826e-07,
+      "loss": 0.7125,
+      "step": 25993
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.836641030600331e-07,
+      "loss": 0.6657,
+      "step": 25994
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.833296218114413e-07,
+      "loss": 0.5976,
+      "step": 25995
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.829952533950145e-07,
+      "loss": 0.6774,
+      "step": 25996
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.82660997814719e-07,
+      "loss": 0.6188,
+      "step": 25997
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.823268550745164e-07,
+      "loss": 0.6209,
+      "step": 25998
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.81992825178369e-07,
+      "loss": 0.5847,
+      "step": 25999
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.816589081302359e-07,
+      "loss": 0.6949,
+      "step": 26000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.813251039340782e-07,
+      "loss": 0.6546,
+      "step": 26001
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.80991412593852e-07,
+      "loss": 0.6301,
+      "step": 26002
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.80657834113516e-07,
+      "loss": 0.65,
+      "step": 26003
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.803243684970215e-07,
+      "loss": 0.6604,
+      "step": 26004
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.799910157483245e-07,
+      "loss": 0.658,
+      "step": 26005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.796577758713772e-07,
+      "loss": 0.6242,
+      "step": 26006
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.793246488701298e-07,
+      "loss": 0.6295,
+      "step": 26007
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.789916347485324e-07,
+      "loss": 0.6457,
+      "step": 26008
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.786587335105352e-07,
+      "loss": 0.6313,
+      "step": 26009
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.783259451600808e-07,
+      "loss": 0.5967,
+      "step": 26010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.77993269701118e-07,
+      "loss": 0.5775,
+      "step": 26011
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.776607071375905e-07,
+      "loss": 0.6347,
+      "step": 26012
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.773282574734406e-07,
+      "loss": 0.6917,
+      "step": 26013
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.769959207126119e-07,
+      "loss": 0.6135,
+      "step": 26014
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7666369685904323e-07,
+      "loss": 0.669,
+      "step": 26015
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.763315859166717e-07,
+      "loss": 0.6239,
+      "step": 26016
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7599958788943744e-07,
+      "loss": 0.6696,
+      "step": 26017
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7566770278127614e-07,
+      "loss": 0.6637,
+      "step": 26018
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.753359305961225e-07,
+      "loss": 0.5719,
+      "step": 26019
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7500427133790996e-07,
+      "loss": 0.614,
+      "step": 26020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7467272501057313e-07,
+      "loss": 0.6729,
+      "step": 26021
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.743412916180379e-07,
+      "loss": 0.6429,
+      "step": 26022
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.740099711642376e-07,
+      "loss": 0.6319,
+      "step": 26023
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.736787636531004e-07,
+      "loss": 0.6015,
+      "step": 26024
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.733476690885508e-07,
+      "loss": 0.6392,
+      "step": 26025
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.730166874745168e-07,
+      "loss": 0.6467,
+      "step": 26026
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7268581881492193e-07,
+      "loss": 0.6301,
+      "step": 26027
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.723550631136886e-07,
+      "loss": 0.5268,
+      "step": 26028
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.720244203747382e-07,
+      "loss": 0.647,
+      "step": 26029
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.716938906019908e-07,
+      "loss": 0.6664,
+      "step": 26030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7136347379936554e-07,
+      "loss": 0.5911,
+      "step": 26031
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.710331699707815e-07,
+      "loss": 0.625,
+      "step": 26032
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.707029791201534e-07,
+      "loss": 0.6309,
+      "step": 26033
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7037290125139576e-07,
+      "loss": 0.6814,
+      "step": 26034
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.700429363684222e-07,
+      "loss": 0.6777,
+      "step": 26035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.697130844751441e-07,
+      "loss": 0.661,
+      "step": 26036
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.693833455754737e-07,
+      "loss": 0.6226,
+      "step": 26037
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6905371967332134e-07,
+      "loss": 0.6413,
+      "step": 26038
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6872420677259277e-07,
+      "loss": 0.6491,
+      "step": 26039
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6839480687719594e-07,
+      "loss": 0.6532,
+      "step": 26040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.680655199910378e-07,
+      "loss": 0.6301,
+      "step": 26041
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.677363461180173e-07,
+      "loss": 0.6364,
+      "step": 26042
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.674072852620426e-07,
+      "loss": 0.6424,
+      "step": 26043
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.670783374270149e-07,
+      "loss": 0.6032,
+      "step": 26044
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6674950261683116e-07,
+      "loss": 0.6223,
+      "step": 26045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.664207808353915e-07,
+      "loss": 0.6086,
+      "step": 26046
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.660921720865952e-07,
+      "loss": 0.6481,
+      "step": 26047
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.657636763743345e-07,
+      "loss": 0.5807,
+      "step": 26048
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.654352937025064e-07,
+      "loss": 0.5928,
+      "step": 26049
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6510702407500664e-07,
+      "loss": 0.6348,
+      "step": 26050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.647788674957232e-07,
+      "loss": 0.697,
+      "step": 26051
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.644508239685475e-07,
+      "loss": 0.6346,
+      "step": 26052
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6412289349737183e-07,
+      "loss": 0.7038,
+      "step": 26053
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6379507608607987e-07,
+      "loss": 0.643,
+      "step": 26054
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.634673717385618e-07,
+      "loss": 0.6324,
+      "step": 26055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6313978045870344e-07,
+      "loss": 0.605,
+      "step": 26056
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.62812302250385e-07,
+      "loss": 0.6522,
+      "step": 26057
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.624849371174911e-07,
+      "loss": 0.6324,
+      "step": 26058
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.621576850639053e-07,
+      "loss": 0.683,
+      "step": 26059
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.618305460935035e-07,
+      "loss": 0.6379,
+      "step": 26060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.61503520210167e-07,
+      "loss": 0.6659,
+      "step": 26061
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6117660741777373e-07,
+      "loss": 0.6174,
+      "step": 26062
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6084980772019727e-07,
+      "loss": 0.6123,
+      "step": 26063
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.605231211213135e-07,
+      "loss": 0.6239,
+      "step": 26064
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.60196547624997e-07,
+      "loss": 0.6603,
+      "step": 26065
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5987008723511586e-07,
+      "loss": 0.6569,
+      "step": 26066
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.595437399555447e-07,
+      "loss": 0.5961,
+      "step": 26067
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5921750579015157e-07,
+      "loss": 0.6286,
+      "step": 26068
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5889138474280337e-07,
+      "loss": 0.6561,
+      "step": 26069
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.58565376817367e-07,
+      "loss": 0.6507,
+      "step": 26070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5823948201771054e-07,
+      "loss": 0.6226,
+      "step": 26071
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.579137003476919e-07,
+      "loss": 0.6882,
+      "step": 26072
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5758803181117806e-07,
+      "loss": 0.6202,
+      "step": 26073
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.572624764120315e-07,
+      "loss": 0.6363,
+      "step": 26074
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5693703415410796e-07,
+      "loss": 0.6261,
+      "step": 26075
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.566117050412677e-07,
+      "loss": 0.5893,
+      "step": 26076
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5628648907736994e-07,
+      "loss": 0.603,
+      "step": 26077
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.55961386266266e-07,
+      "loss": 0.6046,
+      "step": 26078
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.556363966118149e-07,
+      "loss": 0.6597,
+      "step": 26079
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5531152011786816e-07,
+      "loss": 0.6509,
+      "step": 26080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.549867567882771e-07,
+      "loss": 0.6815,
+      "step": 26081
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5466210662689193e-07,
+      "loss": 0.6346,
+      "step": 26082
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5433756963756403e-07,
+      "loss": 0.621,
+      "step": 26083
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.54013145824137e-07,
+      "loss": 0.6451,
+      "step": 26084
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.536888351904611e-07,
+      "loss": 0.5863,
+      "step": 26085
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.533646377403822e-07,
+      "loss": 0.5907,
+      "step": 26086
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.530405534777393e-07,
+      "loss": 0.5878,
+      "step": 26087
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.527165824063784e-07,
+      "loss": 0.6315,
+      "step": 26088
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.523927245301418e-07,
+      "loss": 0.649,
+      "step": 26089
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5206897985286546e-07,
+      "loss": 0.6253,
+      "step": 26090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5174534837838737e-07,
+      "loss": 0.6031,
+      "step": 26091
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5142183011054995e-07,
+      "loss": 0.6257,
+      "step": 26092
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.510984250531847e-07,
+      "loss": 0.689,
+      "step": 26093
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.507751332101273e-07,
+      "loss": 0.6789,
+      "step": 26094
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.504519545852115e-07,
+      "loss": 0.5909,
+      "step": 26095
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.5012888918226636e-07,
+      "loss": 0.6073,
+      "step": 26096
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4980593700512445e-07,
+      "loss": 0.5325,
+      "step": 26097
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.494830980576137e-07,
+      "loss": 0.6092,
+      "step": 26098
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4916037234356223e-07,
+      "loss": 0.5915,
+      "step": 26099
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.488377598667981e-07,
+      "loss": 0.5763,
+      "step": 26100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.485152606311427e-07,
+      "loss": 0.6411,
+      "step": 26101
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.481928746404229e-07,
+      "loss": 0.6008,
+      "step": 26102
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.478706018984591e-07,
+      "loss": 0.6613,
+      "step": 26103
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4754844240907257e-07,
+      "loss": 0.6284,
+      "step": 26104
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4722639617608255e-07,
+      "loss": 0.6349,
+      "step": 26105
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4690446320331036e-07,
+      "loss": 0.6349,
+      "step": 26106
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.465826434945686e-07,
+      "loss": 0.6489,
+      "step": 26107
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4626093705367413e-07,
+      "loss": 0.5967,
+      "step": 26108
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.459393438844428e-07,
+      "loss": 0.6525,
+      "step": 26109
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4561786399068605e-07,
+      "loss": 0.6474,
+      "step": 26110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.452964973762164e-07,
+      "loss": 0.5917,
+      "step": 26111
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4497524404484405e-07,
+      "loss": 0.6287,
+      "step": 26112
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.446541040003771e-07,
+      "loss": 0.6456,
+      "step": 26113
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4433307724662254e-07,
+      "loss": 0.675,
+      "step": 26114
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.440121637873873e-07,
+      "loss": 0.647,
+      "step": 26115
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4369136362647615e-07,
+      "loss": 0.6517,
+      "step": 26116
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.4337067676769264e-07,
+      "loss": 0.6493,
+      "step": 26117
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.430501032148404e-07,
+      "loss": 0.5956,
+      "step": 26118
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.427296429717187e-07,
+      "loss": 0.6276,
+      "step": 26119
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.424092960421256e-07,
+      "loss": 0.6157,
+      "step": 26120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.420890624298613e-07,
+      "loss": 0.6878,
+      "step": 26121
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.417689421387217e-07,
+      "loss": 0.6193,
+      "step": 26122
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4144893517250263e-07,
+      "loss": 0.6384,
+      "step": 26123
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.411290415349989e-07,
+      "loss": 0.5376,
+      "step": 26124
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.408092612300019e-07,
+      "loss": 0.6022,
+      "step": 26125
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.40489594261303e-07,
+      "loss": 0.6504,
+      "step": 26126
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4017004063269363e-07,
+      "loss": 0.6593,
+      "step": 26127
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.398506003479608e-07,
+      "loss": 0.6593,
+      "step": 26128
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3953127341089365e-07,
+      "loss": 0.5569,
+      "step": 26129
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3921205982527694e-07,
+      "loss": 0.5248,
+      "step": 26130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.388929595948965e-07,
+      "loss": 0.5579,
+      "step": 26131
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3857397272353385e-07,
+      "loss": 0.596,
+      "step": 26132
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.382550992149714e-07,
+      "loss": 0.6473,
+      "step": 26133
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.379363390729907e-07,
+      "loss": 0.6896,
+      "step": 26134
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3761769230137196e-07,
+      "loss": 0.6224,
+      "step": 26135
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.372991589038922e-07,
+      "loss": 0.6449,
+      "step": 26136
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3698073888432613e-07,
+      "loss": 0.642,
+      "step": 26137
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.366624322464519e-07,
+      "loss": 0.5853,
+      "step": 26138
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.36344238994042e-07,
+      "loss": 0.6575,
+      "step": 26139
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3602615913086787e-07,
+      "loss": 0.6435,
+      "step": 26140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.357081926607032e-07,
+      "loss": 0.6588,
+      "step": 26141
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3539033958731824e-07,
+      "loss": 0.6189,
+      "step": 26142
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3507259991447895e-07,
+      "loss": 0.6029,
+      "step": 26143
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.347549736459533e-07,
+      "loss": 0.5914,
+      "step": 26144
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.344374607855073e-07,
+      "loss": 0.6323,
+      "step": 26145
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3412006133690564e-07,
+      "loss": 0.5856,
+      "step": 26146
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.33802775303912e-07,
+      "loss": 0.6373,
+      "step": 26147
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3348560269028895e-07,
+      "loss": 0.6509,
+      "step": 26148
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3316854349979456e-07,
+      "loss": 0.5958,
+      "step": 26149
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3285159773618914e-07,
+      "loss": 0.6604,
+      "step": 26150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.325347654032297e-07,
+      "loss": 0.613,
+      "step": 26151
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3221804650467435e-07,
+      "loss": 0.653,
+      "step": 26152
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.319014410442779e-07,
+      "loss": 0.6276,
+      "step": 26153
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3158494902579396e-07,
+      "loss": 0.6407,
+      "step": 26154
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3126857045297286e-07,
+      "loss": 0.6389,
+      "step": 26155
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.309523053295683e-07,
+      "loss": 0.6639,
+      "step": 26156
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.306361536593295e-07,
+      "loss": 0.7004,
+      "step": 26157
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.303201154460035e-07,
+      "loss": 0.6358,
+      "step": 26158
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.300041906933394e-07,
+      "loss": 0.6593,
+      "step": 26159
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.29688379405081e-07,
+      "loss": 0.7056,
+      "step": 26160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2937268158497305e-07,
+      "loss": 0.6298,
+      "step": 26161
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2905709723675916e-07,
+      "loss": 0.6651,
+      "step": 26162
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2874162636418084e-07,
+      "loss": 0.6124,
+      "step": 26163
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2842626897097726e-07,
+      "loss": 0.6376,
+      "step": 26164
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2811102506088997e-07,
+      "loss": 0.6313,
+      "step": 26165
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.277958946376548e-07,
+      "loss": 0.6273,
+      "step": 26166
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2748087770500767e-07,
+      "loss": 0.6447,
+      "step": 26167
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.271659742666834e-07,
+      "loss": 0.6499,
+      "step": 26168
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.268511843264167e-07,
+      "loss": 0.6281,
+      "step": 26169
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.265365078879391e-07,
+      "loss": 0.5999,
+      "step": 26170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2622194495498314e-07,
+      "loss": 0.6391,
+      "step": 26171
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2590749553127585e-07,
+      "loss": 0.6163,
+      "step": 26172
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2559315962054535e-07,
+      "loss": 0.667,
+      "step": 26173
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.252789372265209e-07,
+      "loss": 0.6121,
+      "step": 26174
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.249648283529262e-07,
+      "loss": 0.6311,
+      "step": 26175
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2465083300348595e-07,
+      "loss": 0.6486,
+      "step": 26176
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.243369511819251e-07,
+      "loss": 0.6774,
+      "step": 26177
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2402318289196054e-07,
+      "loss": 0.6065,
+      "step": 26178
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.237095281373149e-07,
+      "loss": 0.6765,
+      "step": 26179
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2339598692170747e-07,
+      "loss": 0.5784,
+      "step": 26180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2308255924885523e-07,
+      "loss": 0.6098,
+      "step": 26181
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2276924512247295e-07,
+      "loss": 0.5878,
+      "step": 26182
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2245604454627885e-07,
+      "loss": 0.6469,
+      "step": 26183
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.221429575239822e-07,
+      "loss": 0.6727,
+      "step": 26184
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2182998405929766e-07,
+      "loss": 0.5908,
+      "step": 26185
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.215171241559346e-07,
+      "loss": 0.6718,
+      "step": 26186
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2120437781760224e-07,
+      "loss": 0.6561,
+      "step": 26187
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.208917450480099e-07,
+      "loss": 0.6477,
+      "step": 26188
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.205792258508645e-07,
+      "loss": 0.6199,
+      "step": 26189
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2026682022986986e-07,
+      "loss": 0.6612,
+      "step": 26190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.199545281887296e-07,
+      "loss": 0.6075,
+      "step": 26191
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1964234973114747e-07,
+      "loss": 0.6273,
+      "step": 26192
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1933028486082493e-07,
+      "loss": 0.6429,
+      "step": 26193
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1901833358146127e-07,
+      "loss": 0.5855,
+      "step": 26194
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.187064958967557e-07,
+      "loss": 0.6312,
+      "step": 26195
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.183947718104053e-07,
+      "loss": 0.612,
+      "step": 26196
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1808316132610385e-07,
+      "loss": 0.6066,
+      "step": 26197
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1777166444754936e-07,
+      "loss": 0.6914,
+      "step": 26198
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.174602811784323e-07,
+      "loss": 0.6173,
+      "step": 26199
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1714901152244633e-07,
+      "loss": 0.6498,
+      "step": 26200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.168378554832819e-07,
+      "loss": 0.6211,
+      "step": 26201
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1652681306462715e-07,
+      "loss": 0.6686,
+      "step": 26202
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.162158842701691e-07,
+      "loss": 0.6535,
+      "step": 26203
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1590506910359597e-07,
+      "loss": 0.6439,
+      "step": 26204
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.155943675685914e-07,
+      "loss": 0.6581,
+      "step": 26205
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1528377966884024e-07,
+      "loss": 0.6183,
+      "step": 26206
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.149733054080263e-07,
+      "loss": 0.5764,
+      "step": 26207
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.146629447898276e-07,
+      "loss": 0.618,
+      "step": 26208
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1435269781792465e-07,
+      "loss": 0.6761,
+      "step": 26209
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1404256449599666e-07,
+      "loss": 0.6742,
+      "step": 26210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1373254482772076e-07,
+      "loss": 0.6274,
+      "step": 26211
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.134226388167717e-07,
+      "loss": 0.6661,
+      "step": 26212
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1311284646682656e-07,
+      "loss": 0.6609,
+      "step": 26213
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.128031677815536e-07,
+      "loss": 0.6303,
+      "step": 26214
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.124936027646276e-07,
+      "loss": 0.6062,
+      "step": 26215
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.121841514197178e-07,
+      "loss": 0.6162,
+      "step": 26216
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.118748137504935e-07,
+      "loss": 0.5954,
+      "step": 26217
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.11565589760623e-07,
+      "loss": 0.6014,
+      "step": 26218
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.112564794537732e-07,
+      "loss": 0.6263,
+      "step": 26219
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1094748283360576e-07,
+      "loss": 0.6445,
+      "step": 26220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.106385999037865e-07,
+      "loss": 0.651,
+      "step": 26221
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.10329830667977e-07,
+      "loss": 0.6167,
+      "step": 26222
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1002117512983775e-07,
+      "loss": 0.6838,
+      "step": 26223
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0971263329303126e-07,
+      "loss": 0.646,
+      "step": 26224
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.094042051612113e-07,
+      "loss": 0.6508,
+      "step": 26225
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.090958907380371e-07,
+      "loss": 0.6625,
+      "step": 26226
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.087876900271637e-07,
+      "loss": 0.6663,
+      "step": 26227
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0847960303224464e-07,
+      "loss": 0.6514,
+      "step": 26228
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.081716297569338e-07,
+      "loss": 0.6369,
+      "step": 26229
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.078637702048827e-07,
+      "loss": 0.6357,
+      "step": 26230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.075560243797394e-07,
+      "loss": 0.5953,
+      "step": 26231
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.072483922851544e-07,
+      "loss": 0.5998,
+      "step": 26232
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0694087392477486e-07,
+      "loss": 0.6476,
+      "step": 26233
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0663346930224557e-07,
+      "loss": 0.6482,
+      "step": 26234
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0632617842121357e-07,
+      "loss": 0.5975,
+      "step": 26235
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0601900128532046e-07,
+      "loss": 0.6345,
+      "step": 26236
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0571193789820886e-07,
+      "loss": 0.6464,
+      "step": 26237
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0540498826351804e-07,
+      "loss": 0.6824,
+      "step": 26238
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.050981523848885e-07,
+      "loss": 0.6819,
+      "step": 26239
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0479143026595835e-07,
+      "loss": 0.6122,
+      "step": 26240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0448482191036364e-07,
+      "loss": 0.6505,
+      "step": 26241
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0417832732174036e-07,
+      "loss": 0.6749,
+      "step": 26242
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0387194650372106e-07,
+      "loss": 0.6615,
+      "step": 26243
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.035656794599385e-07,
+      "loss": 0.6348,
+      "step": 26244
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.032595261940253e-07,
+      "loss": 0.5794,
+      "step": 26245
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0295348670960967e-07,
+      "loss": 0.6618,
+      "step": 26246
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.026475610103209e-07,
+      "loss": 0.5858,
+      "step": 26247
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.023417490997883e-07,
+      "loss": 0.6105,
+      "step": 26248
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0203605098163234e-07,
+      "loss": 0.6227,
+      "step": 26249
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0173046665948125e-07,
+      "loss": 0.6369,
+      "step": 26250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.014249961369576e-07,
+      "loss": 0.6398,
+      "step": 26251
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0111963941768306e-07,
+      "loss": 0.6647,
+      "step": 26252
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0081439650527687e-07,
+      "loss": 0.5924,
+      "step": 26253
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.005092674033617e-07,
+      "loss": 0.6291,
+      "step": 26254
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0020425211555024e-07,
+      "loss": 0.6161,
+      "step": 26255
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.998993506454618e-07,
+      "loss": 0.6671,
+      "step": 26256
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9959456299671015e-07,
+      "loss": 0.6602,
+      "step": 26257
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9928988917291023e-07,
+      "loss": 0.6052,
+      "step": 26258
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.989853291776724e-07,
+      "loss": 0.6745,
+      "step": 26259
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9868088301461163e-07,
+      "loss": 0.6006,
+      "step": 26260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.983765506873327e-07,
+      "loss": 0.6425,
+      "step": 26261
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.980723321994451e-07,
+      "loss": 0.6014,
+      "step": 26262
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.97768227554558e-07,
+      "loss": 0.6349,
+      "step": 26263
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.974642367562742e-07,
+      "loss": 0.666,
+      "step": 26264
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.971603598081997e-07,
+      "loss": 0.6166,
+      "step": 26265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.968565967139382e-07,
+      "loss": 0.6666,
+      "step": 26266
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9655294747708907e-07,
+      "loss": 0.6768,
+      "step": 26267
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.962494121012517e-07,
+      "loss": 0.6948,
+      "step": 26268
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.959459905900276e-07,
+      "loss": 0.5998,
+      "step": 26269
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.956426829470128e-07,
+      "loss": 0.6254,
+      "step": 26270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.953394891758033e-07,
+      "loss": 0.6633,
+      "step": 26271
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.950364092799952e-07,
+      "loss": 0.5759,
+      "step": 26272
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9473344326317995e-07,
+      "loss": 0.6157,
+      "step": 26273
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.944305911289492e-07,
+      "loss": 0.596,
+      "step": 26274
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9412785288089673e-07,
+      "loss": 0.6307,
+      "step": 26275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.938252285226085e-07,
+      "loss": 0.6765,
+      "step": 26276
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9352271805767506e-07,
+      "loss": 0.6054,
+      "step": 26277
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.932203214896835e-07,
+      "loss": 0.6173,
+      "step": 26278
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9291803882221534e-07,
+      "loss": 0.6335,
+      "step": 26279
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.926158700588578e-07,
+      "loss": 0.6159,
+      "step": 26280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9231381520319353e-07,
+      "loss": 0.5929,
+      "step": 26281
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9201187425879974e-07,
+      "loss": 0.6394,
+      "step": 26282
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.917100472292623e-07,
+      "loss": 0.6251,
+      "step": 26283
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9140833411815625e-07,
+      "loss": 0.6534,
+      "step": 26284
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.911067349290587e-07,
+      "loss": 0.6412,
+      "step": 26285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.908052496655468e-07,
+      "loss": 0.6292,
+      "step": 26286
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9050387833119427e-07,
+      "loss": 0.5879,
+      "step": 26287
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.902026209295717e-07,
+      "loss": 0.6604,
+      "step": 26288
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.899014774642573e-07,
+      "loss": 0.6404,
+      "step": 26289
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8960044793881604e-07,
+      "loss": 0.5972,
+      "step": 26290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8929953235681826e-07,
+      "loss": 0.5746,
+      "step": 26291
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.889987307218335e-07,
+      "loss": 0.6593,
+      "step": 26292
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8869804303742544e-07,
+      "loss": 0.6416,
+      "step": 26293
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.883974693071602e-07,
+      "loss": 0.619,
+      "step": 26294
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.880970095346026e-07,
+      "loss": 0.6229,
+      "step": 26295
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8779666372331435e-07,
+      "loss": 0.5669,
+      "step": 26296
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.874964318768548e-07,
+      "loss": 0.598,
+      "step": 26297
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.871963139987878e-07,
+      "loss": 0.6268,
+      "step": 26298
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8689631009266595e-07,
+      "loss": 0.7189,
+      "step": 26299
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.865964201620487e-07,
+      "loss": 0.5905,
+      "step": 26300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8629664421049427e-07,
+      "loss": 0.6493,
+      "step": 26301
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8599698224155214e-07,
+      "loss": 0.6619,
+      "step": 26302
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.856974342587794e-07,
+      "loss": 0.5718,
+      "step": 26303
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.853980002657254e-07,
+      "loss": 0.68,
+      "step": 26304
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8509868026594066e-07,
+      "loss": 0.6571,
+      "step": 26305
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.847994742629713e-07,
+      "loss": 0.6918,
+      "step": 26306
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.84500382260371e-07,
+      "loss": 0.6698,
+      "step": 26307
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8420140426168153e-07,
+      "loss": 0.6203,
+      "step": 26308
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.839025402704477e-07,
+      "loss": 0.6275,
+      "step": 26309
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8360379029021456e-07,
+      "loss": 0.6217,
+      "step": 26310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.833051543245225e-07,
+      "loss": 0.6863,
+      "step": 26311
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8300663237691214e-07,
+      "loss": 0.6353,
+      "step": 26312
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.827082244509261e-07,
+      "loss": 0.6622,
+      "step": 26313
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8240993055009835e-07,
+      "loss": 0.644,
+      "step": 26314
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8211175067796704e-07,
+      "loss": 0.6915,
+      "step": 26315
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8181368483806935e-07,
+      "loss": 0.6246,
+      "step": 26316
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8151573303393587e-07,
+      "loss": 0.663,
+      "step": 26317
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8121789526910036e-07,
+      "loss": 0.6726,
+      "step": 26318
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.809201715470956e-07,
+      "loss": 0.6494,
+      "step": 26319
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8062256187144984e-07,
+      "loss": 0.5864,
+      "step": 26320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8032506624569256e-07,
+      "loss": 0.6826,
+      "step": 26321
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.800276846733519e-07,
+      "loss": 0.5707,
+      "step": 26322
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.797304171579508e-07,
+      "loss": 0.652,
+      "step": 26323
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7943326370301403e-07,
+      "loss": 0.6774,
+      "step": 26324
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7913622431206886e-07,
+      "loss": 0.6347,
+      "step": 26325
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7883929898863247e-07,
+      "loss": 0.6359,
+      "step": 26326
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.785424877362265e-07,
+      "loss": 0.5802,
+      "step": 26327
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.782457905583725e-07,
+      "loss": 0.6411,
+      "step": 26328
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7794920745858554e-07,
+      "loss": 0.6357,
+      "step": 26329
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.776527384403805e-07,
+      "loss": 0.5923,
+      "step": 26330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7735638350727797e-07,
+      "loss": 0.588,
+      "step": 26331
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7706014266278624e-07,
+      "loss": 0.6356,
+      "step": 26332
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.767640159104191e-07,
+      "loss": 0.6373,
+      "step": 26333
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.764680032536905e-07,
+      "loss": 0.6159,
+      "step": 26334
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.761721046961053e-07,
+      "loss": 0.6258,
+      "step": 26335
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7587632024117415e-07,
+      "loss": 0.6504,
+      "step": 26336
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.755806498924053e-07,
+      "loss": 0.6074,
+      "step": 26337
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7528509365330143e-07,
+      "loss": 0.7038,
+      "step": 26338
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7498965152736764e-07,
+      "loss": 0.6513,
+      "step": 26339
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7469432351811e-07,
+      "loss": 0.6483,
+      "step": 26340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7439910962902447e-07,
+      "loss": 0.6841,
+      "step": 26341
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7410400986361397e-07,
+      "loss": 0.6158,
+      "step": 26342
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.738090242253789e-07,
+      "loss": 0.6116,
+      "step": 26343
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7351415271781435e-07,
+      "loss": 0.6639,
+      "step": 26344
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7321939534441745e-07,
+      "loss": 0.6282,
+      "step": 26345
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7292475210868207e-07,
+      "loss": 0.6212,
+      "step": 26346
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7263022301410213e-07,
+      "loss": 0.6342,
+      "step": 26347
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7233580806416926e-07,
+      "loss": 0.6511,
+      "step": 26348
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7204150726237396e-07,
+      "loss": 0.6217,
+      "step": 26349
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.717473206122069e-07,
+      "loss": 0.5933,
+      "step": 26350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7145324811715623e-07,
+      "loss": 0.6438,
+      "step": 26351
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7115928978070594e-07,
+      "loss": 0.6759,
+      "step": 26352
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7086544560634206e-07,
+      "loss": 0.6254,
+      "step": 26353
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7057171559755076e-07,
+      "loss": 0.616,
+      "step": 26354
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.702780997578126e-07,
+      "loss": 0.6275,
+      "step": 26355
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6998459809060916e-07,
+      "loss": 0.6298,
+      "step": 26356
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6969121059942104e-07,
+      "loss": 0.6215,
+      "step": 26357
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6939793728772544e-07,
+      "loss": 0.6743,
+      "step": 26358
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.691047781589996e-07,
+      "loss": 0.6457,
+      "step": 26359
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.688117332167196e-07,
+      "loss": 0.6377,
+      "step": 26360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6851880246436047e-07,
+      "loss": 0.6816,
+      "step": 26361
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6822598590539495e-07,
+      "loss": 0.6377,
+      "step": 26362
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.679332835432958e-07,
+      "loss": 0.612,
+      "step": 26363
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6764069538153144e-07,
+      "loss": 0.6404,
+      "step": 26364
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.673482214235702e-07,
+      "loss": 0.6001,
+      "step": 26365
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.670558616728825e-07,
+      "loss": 0.6378,
+      "step": 26366
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.667636161329324e-07,
+      "loss": 0.6525,
+      "step": 26367
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.66471484807186e-07,
+      "loss": 0.6749,
+      "step": 26368
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6617946769910816e-07,
+      "loss": 0.6415,
+      "step": 26369
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.658875648121585e-07,
+      "loss": 0.5854,
+      "step": 26370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.655957761497986e-07,
+      "loss": 0.6333,
+      "step": 26371
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.653041017154879e-07,
+      "loss": 0.6264,
+      "step": 26372
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.650125415126859e-07,
+      "loss": 0.6503,
+      "step": 26373
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.647210955448488e-07,
+      "loss": 0.61,
+      "step": 26374
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.644297638154315e-07,
+      "loss": 0.5688,
+      "step": 26375
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6413854632788904e-07,
+      "loss": 0.6819,
+      "step": 26376
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.63847443085672e-07,
+      "loss": 0.6398,
+      "step": 26377
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6355645409223427e-07,
+      "loss": 0.5934,
+      "step": 26378
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6326557935102427e-07,
+      "loss": 0.653,
+      "step": 26379
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6297481886549136e-07,
+      "loss": 0.6227,
+      "step": 26380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6268417263908504e-07,
+      "loss": 0.6221,
+      "step": 26381
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6239364067524707e-07,
+      "loss": 0.5897,
+      "step": 26382
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.621032229774246e-07,
+      "loss": 0.6301,
+      "step": 26383
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.618129195490594e-07,
+      "loss": 0.609,
+      "step": 26384
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6152273039359533e-07,
+      "loss": 0.6408,
+      "step": 26385
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6123265551447183e-07,
+      "loss": 0.6451,
+      "step": 26386
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6094269491512956e-07,
+      "loss": 0.6122,
+      "step": 26387
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.606528485990046e-07,
+      "loss": 0.5782,
+      "step": 26388
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6036311656953315e-07,
+      "loss": 0.5995,
+      "step": 26389
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.600734988301513e-07,
+      "loss": 0.6458,
+      "step": 26390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.5978399538429186e-07,
+      "loss": 0.5397,
+      "step": 26391
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.594946062353888e-07,
+      "loss": 0.6117,
+      "step": 26392
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.592053313868737e-07,
+      "loss": 0.6688,
+      "step": 26393
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.58916170842174e-07,
+      "loss": 0.6618,
+      "step": 26394
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.5862712460471903e-07,
+      "loss": 0.6407,
+      "step": 26395
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.5833819267793613e-07,
+      "loss": 0.6591,
+      "step": 26396
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.580493750652514e-07,
+      "loss": 0.6225,
+      "step": 26397
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.5776067177008656e-07,
+      "loss": 0.6193,
+      "step": 26398
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.574720827958689e-07,
+      "loss": 0.6768,
+      "step": 26399
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.571836081460167e-07,
+      "loss": 0.6045,
+      "step": 26400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.568952478239507e-07,
+      "loss": 0.5935,
+      "step": 26401
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.566070018330892e-07,
+      "loss": 0.6265,
+      "step": 26402
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.563188701768516e-07,
+      "loss": 0.5644,
+      "step": 26403
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.5603085285865314e-07,
+      "loss": 0.6633,
+      "step": 26404
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.557429498819087e-07,
+      "loss": 0.6867,
+      "step": 26405
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.554551612500323e-07,
+      "loss": 0.6589,
+      "step": 26406
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.551674869664334e-07,
+      "loss": 0.6103,
+      "step": 26407
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.5487992703452603e-07,
+      "loss": 0.6194,
+      "step": 26408
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.545924814577184e-07,
+      "loss": 0.6843,
+      "step": 26409
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5430515023941903e-07,
+      "loss": 0.6244,
+      "step": 26410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.540179333830329e-07,
+      "loss": 0.5972,
+      "step": 26411
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5373083089196627e-07,
+      "loss": 0.6641,
+      "step": 26412
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5344384276962296e-07,
+      "loss": 0.5931,
+      "step": 26413
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.531569690194059e-07,
+      "loss": 0.6219,
+      "step": 26414
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.528702096447167e-07,
+      "loss": 0.7088,
+      "step": 26415
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5258356464895615e-07,
+      "loss": 0.6013,
+      "step": 26416
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.522970340355203e-07,
+      "loss": 0.6063,
+      "step": 26417
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5201061780780754e-07,
+      "loss": 0.6035,
+      "step": 26418
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5172431596921296e-07,
+      "loss": 0.5984,
+      "step": 26419
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5143812852313275e-07,
+      "loss": 0.6302,
+      "step": 26420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.511520554729575e-07,
+      "loss": 0.6432,
+      "step": 26421
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5086609682208337e-07,
+      "loss": 0.6577,
+      "step": 26422
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.505802525738955e-07,
+      "loss": 0.6082,
+      "step": 26423
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5029452273178664e-07,
+      "loss": 0.5632,
+      "step": 26424
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5000890729914197e-07,
+      "loss": 0.6318,
+      "step": 26425
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4972340627934866e-07,
+      "loss": 0.6185,
+      "step": 26426
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4943801967579295e-07,
+      "loss": 0.6409,
+      "step": 26427
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.491527474918588e-07,
+      "loss": 0.595,
+      "step": 26428
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.488675897309257e-07,
+      "loss": 0.6569,
+      "step": 26429
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4858254639637657e-07,
+      "loss": 0.6346,
+      "step": 26430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4829761749158976e-07,
+      "loss": 0.6559,
+      "step": 26431
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4801280301994366e-07,
+      "loss": 0.6497,
+      "step": 26432
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4772810298481673e-07,
+      "loss": 0.6513,
+      "step": 26433
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4744351738958406e-07,
+      "loss": 0.664,
+      "step": 26434
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4715904623761844e-07,
+      "loss": 0.6267,
+      "step": 26435
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4687468953229276e-07,
+      "loss": 0.6092,
+      "step": 26436
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.465904472769788e-07,
+      "loss": 0.6136,
+      "step": 26437
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.46306319475046e-07,
+      "loss": 0.597,
+      "step": 26438
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.460223061298651e-07,
+      "loss": 0.5952,
+      "step": 26439
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.457384072448022e-07,
+      "loss": 0.5924,
+      "step": 26440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4545462282322253e-07,
+      "loss": 0.6215,
+      "step": 26441
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4517095286849213e-07,
+      "loss": 0.6317,
+      "step": 26442
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.448873973839728e-07,
+      "loss": 0.6327,
+      "step": 26443
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4460395637302633e-07,
+      "loss": 0.5581,
+      "step": 26444
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.443206298390156e-07,
+      "loss": 0.6499,
+      "step": 26445
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4403741778529785e-07,
+      "loss": 0.6325,
+      "step": 26446
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.437543202152316e-07,
+      "loss": 0.6022,
+      "step": 26447
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4347133713217295e-07,
+      "loss": 0.6037,
+      "step": 26448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4318846853947596e-07,
+      "loss": 0.661,
+      "step": 26449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.429057144404968e-07,
+      "loss": 0.609,
+      "step": 26450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4262307483858614e-07,
+      "loss": 0.588,
+      "step": 26451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4234054973709687e-07,
+      "loss": 0.5837,
+      "step": 26452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4205813913937626e-07,
+      "loss": 0.6199,
+      "step": 26453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.417758430487739e-07,
+      "loss": 0.6347,
+      "step": 26454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4149366146863703e-07,
+      "loss": 0.6195,
+      "step": 26455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4121159440230977e-07,
+      "loss": 0.651,
+      "step": 26456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4092964185313825e-07,
+      "loss": 0.6153,
+      "step": 26457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.406478038244665e-07,
+      "loss": 0.6265,
+      "step": 26458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.403660803196318e-07,
+      "loss": 0.6422,
+      "step": 26459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.400844713419771e-07,
+      "loss": 0.5482,
+      "step": 26460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3980297689484077e-07,
+      "loss": 0.6197,
+      "step": 26461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.395215969815602e-07,
+      "loss": 0.6553,
+      "step": 26462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3924033160547153e-07,
+      "loss": 0.6172,
+      "step": 26463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.389591807699111e-07,
+      "loss": 0.6213,
+      "step": 26464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3867814447821056e-07,
+      "loss": 0.606,
+      "step": 26465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.383972227337007e-07,
+      "loss": 0.6701,
+      "step": 26466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.381164155397143e-07,
+      "loss": 0.5846,
+      "step": 26467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.37835722899581e-07,
+      "loss": 0.6495,
+      "step": 26468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.375551448166281e-07,
+      "loss": 0.6407,
+      "step": 26469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3727468129418184e-07,
+      "loss": 0.6008,
+      "step": 26470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3699433233556623e-07,
+      "loss": 0.6486,
+      "step": 26471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.367140979441086e-07,
+      "loss": 0.6241,
+      "step": 26472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.364339781231274e-07,
+      "loss": 0.6879,
+      "step": 26473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3615397287594664e-07,
+      "loss": 0.6547,
+      "step": 26474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.358740822058859e-07,
+      "loss": 0.593,
+      "step": 26475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3559430611626253e-07,
+      "loss": 0.6321,
+      "step": 26476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.35314644610395e-07,
+      "loss": 0.6495,
+      "step": 26477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3503509769159834e-07,
+      "loss": 0.611,
+      "step": 26478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.347556653631845e-07,
+      "loss": 0.6296,
+      "step": 26479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.344763476284707e-07,
+      "loss": 0.6485,
+      "step": 26480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.341971444907677e-07,
+      "loss": 0.6056,
+      "step": 26481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.339180559533828e-07,
+      "loss": 0.5849,
+      "step": 26482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.336390820196278e-07,
+      "loss": 0.6441,
+      "step": 26483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.333602226928112e-07,
+      "loss": 0.6613,
+      "step": 26484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.330814779762348e-07,
+      "loss": 0.6042,
+      "step": 26485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.32802847873207e-07,
+      "loss": 0.6277,
+      "step": 26486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3252433238703085e-07,
+      "loss": 0.6067,
+      "step": 26487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3224593152100804e-07,
+      "loss": 0.6244,
+      "step": 26488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.319676452784404e-07,
+      "loss": 0.6499,
+      "step": 26489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.316894736626264e-07,
+      "loss": 0.6456,
+      "step": 26490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3141141667686226e-07,
+      "loss": 0.612,
+      "step": 26491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.311334743244465e-07,
+      "loss": 0.6147,
+      "step": 26492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3085564660867763e-07,
+      "loss": 0.6149,
+      "step": 26493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.305779335328441e-07,
+      "loss": 0.5989,
+      "step": 26494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3030033510024205e-07,
+      "loss": 0.6729,
+      "step": 26495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3002285131416235e-07,
+      "loss": 0.6057,
+      "step": 26496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2974548217789224e-07,
+      "loss": 0.6412,
+      "step": 26497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.294682276947225e-07,
+      "loss": 0.6087,
+      "step": 26498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.291910878679427e-07,
+      "loss": 0.6466,
+      "step": 26499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.289140627008347e-07,
+      "loss": 0.5982,
+      "step": 26500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.286371521966847e-07,
+      "loss": 0.6764,
+      "step": 26501
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.283603563587756e-07,
+      "loss": 0.6277,
+      "step": 26502
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.280836751903882e-07,
+      "loss": 0.67,
+      "step": 26503
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2780710869480425e-07,
+      "loss": 0.6702,
+      "step": 26504
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.275306568753034e-07,
+      "loss": 0.6206,
+      "step": 26505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.272543197351619e-07,
+      "loss": 0.6344,
+      "step": 26506
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2697809727765593e-07,
+      "loss": 0.6395,
+      "step": 26507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.26701989506063e-07,
+      "loss": 0.6425,
+      "step": 26508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.264259964236516e-07,
+      "loss": 0.6594,
+      "step": 26509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.26150118033699e-07,
+      "loss": 0.6464,
+      "step": 26510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.258743543394749e-07,
+      "loss": 0.6457,
+      "step": 26511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2559870534424776e-07,
+      "loss": 0.6055,
+      "step": 26512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2532317105128495e-07,
+      "loss": 0.6559,
+      "step": 26513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2504775146385616e-07,
+      "loss": 0.5936,
+      "step": 26514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.247724465852231e-07,
+      "loss": 0.633,
+      "step": 26515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2449725641865326e-07,
+      "loss": 0.6382,
+      "step": 26516
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2422218096740845e-07,
+      "loss": 0.6428,
+      "step": 26517
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.239472202347493e-07,
+      "loss": 0.6112,
+      "step": 26518
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2367237422393563e-07,
+      "loss": 0.6363,
+      "step": 26519
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2339764293822686e-07,
+      "loss": 0.666,
+      "step": 26520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2312302638087934e-07,
+      "loss": 0.6838,
+      "step": 26521
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2284852455514937e-07,
+      "loss": 0.6085,
+      "step": 26522
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.225741374642943e-07,
+      "loss": 0.6571,
+      "step": 26523
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2229986511156277e-07,
+      "loss": 0.7027,
+      "step": 26524
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2202570750020866e-07,
+      "loss": 0.632,
+      "step": 26525
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2175166463348396e-07,
+      "loss": 0.7007,
+      "step": 26526
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2147773651463375e-07,
+      "loss": 0.6358,
+      "step": 26527
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2120392314690996e-07,
+      "loss": 0.6207,
+      "step": 26528
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2093022453355883e-07,
+      "loss": 0.6765,
+      "step": 26529
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.206566406778222e-07,
+      "loss": 0.6182,
+      "step": 26530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.203831715829464e-07,
+      "loss": 0.6144,
+      "step": 26531
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.201098172521722e-07,
+      "loss": 0.6329,
+      "step": 26532
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1983657768874023e-07,
+      "loss": 0.6391,
+      "step": 26533
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.195634528958935e-07,
+      "loss": 0.6459,
+      "step": 26534
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1929044287686615e-07,
+      "loss": 0.6254,
+      "step": 26535
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1901754763489667e-07,
+      "loss": 0.6439,
+      "step": 26536
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1874476717322246e-07,
+      "loss": 0.6943,
+      "step": 26537
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1847210149507426e-07,
+      "loss": 0.6413,
+      "step": 26538
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1819955060368614e-07,
+      "loss": 0.613,
+      "step": 26539
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1792711450229107e-07,
+      "loss": 0.6296,
+      "step": 26540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.176547931941176e-07,
+      "loss": 0.5829,
+      "step": 26541
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.173825866823943e-07,
+      "loss": 0.6327,
+      "step": 26542
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.171104949703496e-07,
+      "loss": 0.6358,
+      "step": 26543
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.168385180612088e-07,
+      "loss": 0.6658,
+      "step": 26544
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1656665595819593e-07,
+      "loss": 0.6801,
+      "step": 26545
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1629490866453615e-07,
+      "loss": 0.5962,
+      "step": 26546
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.160232761834492e-07,
+      "loss": 0.6187,
+      "step": 26547
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1575175851815687e-07,
+      "loss": 0.6456,
+      "step": 26548
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1548035567187885e-07,
+      "loss": 0.516,
+      "step": 26549
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.152090676478314e-07,
+      "loss": 0.6668,
+      "step": 26550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1493789444923093e-07,
+      "loss": 0.6219,
+      "step": 26551
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1466683607929484e-07,
+      "loss": 0.6216,
+      "step": 26552
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.143958925412349e-07,
+      "loss": 0.6091,
+      "step": 26553
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1412506383826426e-07,
+      "loss": 0.6441,
+      "step": 26554
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1385434997359356e-07,
+      "loss": 0.6287,
+      "step": 26555
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1358375095043246e-07,
+      "loss": 0.6597,
+      "step": 26556
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1331326677198846e-07,
+      "loss": 0.6694,
+      "step": 26557
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.130428974414712e-07,
+      "loss": 0.6378,
+      "step": 26558
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1277264296208365e-07,
+      "loss": 0.6622,
+      "step": 26559
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.125025033370299e-07,
+      "loss": 0.6154,
+      "step": 26560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1223247856951634e-07,
+      "loss": 0.6249,
+      "step": 26561
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.119625686627392e-07,
+      "loss": 0.5727,
+      "step": 26562
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.116927736199016e-07,
+      "loss": 0.6344,
+      "step": 26563
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.114230934442031e-07,
+      "loss": 0.6423,
+      "step": 26564
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1115352813883893e-07,
+      "loss": 0.652,
+      "step": 26565
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1088407770700655e-07,
+      "loss": 0.6464,
+      "step": 26566
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.106147421519012e-07,
+      "loss": 0.6524,
+      "step": 26567
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.103455214767137e-07,
+      "loss": 0.6278,
+      "step": 26568
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.10076415684637e-07,
+      "loss": 0.5889,
+      "step": 26569
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0980742477886407e-07,
+      "loss": 0.5839,
+      "step": 26570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0953854876258125e-07,
+      "loss": 0.6355,
+      "step": 26571
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.092697876389772e-07,
+      "loss": 0.6243,
+      "step": 26572
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0900114141124036e-07,
+      "loss": 0.6303,
+      "step": 26573
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.087326100825527e-07,
+      "loss": 0.595,
+      "step": 26574
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0846419365609834e-07,
+      "loss": 0.6682,
+      "step": 26575
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.081958921350625e-07,
+      "loss": 0.6323,
+      "step": 26576
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0792770552262485e-07,
+      "loss": 0.6462,
+      "step": 26577
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0765963382196393e-07,
+      "loss": 0.6807,
+      "step": 26578
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.073916770362595e-07,
+      "loss": 0.6351,
+      "step": 26579
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0712383516868783e-07,
+      "loss": 0.582,
+      "step": 26580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0685610822242307e-07,
+      "loss": 0.6259,
+      "step": 26581
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0658849620064267e-07,
+      "loss": 0.6517,
+      "step": 26582
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.063209991065175e-07,
+      "loss": 0.6286,
+      "step": 26583
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.060536169432193e-07,
+      "loss": 0.6361,
+      "step": 26584
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0578634971392016e-07,
+      "loss": 0.6789,
+      "step": 26585
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0551919742178524e-07,
+      "loss": 0.6024,
+      "step": 26586
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0525216006998314e-07,
+      "loss": 0.65,
+      "step": 26587
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.049852376616824e-07,
+      "loss": 0.6463,
+      "step": 26588
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.04718430200045e-07,
+      "loss": 0.6303,
+      "step": 26589
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0445173768823497e-07,
+      "loss": 0.6284,
+      "step": 26590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0418516012941546e-07,
+      "loss": 0.6005,
+      "step": 26591
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.03918697526745e-07,
+      "loss": 0.6259,
+      "step": 26592
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.036523498833843e-07,
+      "loss": 0.6414,
+      "step": 26593
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.03386117202491e-07,
+      "loss": 0.6459,
+      "step": 26594
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0311999948722137e-07,
+      "loss": 0.6122,
+      "step": 26595
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.028539967407329e-07,
+      "loss": 0.6243,
+      "step": 26596
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.025881089661753e-07,
+      "loss": 0.6361,
+      "step": 26597
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.023223361667027e-07,
+      "loss": 0.6259,
+      "step": 26598
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0205667834546706e-07,
+      "loss": 0.6101,
+      "step": 26599
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.017911355056169e-07,
+      "loss": 0.5891,
+      "step": 26600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.015257076503009e-07,
+      "loss": 0.6347,
+      "step": 26601
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0126039478266753e-07,
+      "loss": 0.66,
+      "step": 26602
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0099519690585997e-07,
+      "loss": 0.6103,
+      "step": 26603
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0073011402302454e-07,
+      "loss": 0.6281,
+      "step": 26604
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0046514613730315e-07,
+      "loss": 0.6588,
+      "step": 26605
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0020029325183667e-07,
+      "loss": 0.6593,
+      "step": 26606
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.999355553697658e-07,
+      "loss": 0.6484,
+      "step": 26607
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.996709324942315e-07,
+      "loss": 0.6576,
+      "step": 26608
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.994064246283679e-07,
+      "loss": 0.6112,
+      "step": 26609
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.991420317753124e-07,
+      "loss": 0.6243,
+      "step": 26610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.988777539381993e-07,
+      "loss": 0.6611,
+      "step": 26611
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9861359112016375e-07,
+      "loss": 0.6046,
+      "step": 26612
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9834954332433554e-07,
+      "loss": 0.62,
+      "step": 26613
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.980856105538466e-07,
+      "loss": 0.597,
+      "step": 26614
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9782179281182543e-07,
+      "loss": 0.6455,
+      "step": 26615
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9755809010139967e-07,
+      "loss": 0.6715,
+      "step": 26616
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.972945024256968e-07,
+      "loss": 0.6389,
+      "step": 26617
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9703102978784205e-07,
+      "loss": 0.6372,
+      "step": 26618
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.967676721909585e-07,
+      "loss": 0.6323,
+      "step": 26619
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9650442963816914e-07,
+      "loss": 0.5856,
+      "step": 26620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9624130213259493e-07,
+      "loss": 0.6003,
+      "step": 26621
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9597828967735443e-07,
+      "loss": 0.633,
+      "step": 26622
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.957153922755673e-07,
+      "loss": 0.6261,
+      "step": 26623
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.954526099303501e-07,
+      "loss": 0.6773,
+      "step": 26624
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9518994264481795e-07,
+      "loss": 0.5909,
+      "step": 26625
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.949273904220873e-07,
+      "loss": 0.6213,
+      "step": 26626
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.94664953265269e-07,
+      "loss": 0.6663,
+      "step": 26627
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.944026311774739e-07,
+      "loss": 0.6098,
+      "step": 26628
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.94140424161814e-07,
+      "loss": 0.6756,
+      "step": 26629
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.938783322213967e-07,
+      "loss": 0.6671,
+      "step": 26630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9361635535933077e-07,
+      "loss": 0.6221,
+      "step": 26631
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9335449357872246e-07,
+      "loss": 0.5555,
+      "step": 26632
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.93092746882675e-07,
+      "loss": 0.6573,
+      "step": 26633
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.928311152742913e-07,
+      "loss": 0.6227,
+      "step": 26634
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9256959875667455e-07,
+      "loss": 0.602,
+      "step": 26635
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.923081973329256e-07,
+      "loss": 0.641,
+      "step": 26636
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9204691100614304e-07,
+      "loss": 0.6271,
+      "step": 26637
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.917857397794266e-07,
+      "loss": 0.6959,
+      "step": 26638
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9152468365586936e-07,
+      "loss": 0.7219,
+      "step": 26639
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.912637426385678e-07,
+      "loss": 0.6439,
+      "step": 26640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.910029167306172e-07,
+      "loss": 0.6569,
+      "step": 26641
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9074220593510947e-07,
+      "loss": 0.6005,
+      "step": 26642
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.904816102551333e-07,
+      "loss": 0.6412,
+      "step": 26643
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.902211296937829e-07,
+      "loss": 0.6062,
+      "step": 26644
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8996076425414244e-07,
+      "loss": 0.5877,
+      "step": 26645
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.897005139393005e-07,
+      "loss": 0.6274,
+      "step": 26646
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.894403787523425e-07,
+      "loss": 0.652,
+      "step": 26647
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8918035869635264e-07,
+      "loss": 0.5843,
+      "step": 26648
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.889204537744139e-07,
+      "loss": 0.6328,
+      "step": 26649
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8866066398960835e-07,
+      "loss": 0.6359,
+      "step": 26650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.884009893450146e-07,
+      "loss": 0.6184,
+      "step": 26651
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.881414298437124e-07,
+      "loss": 0.5781,
+      "step": 26652
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.878819854887793e-07,
+      "loss": 0.6125,
+      "step": 26653
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.876226562832918e-07,
+      "loss": 0.6428,
+      "step": 26654
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.87363442230324e-07,
+      "loss": 0.588,
+      "step": 26655
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.87104343332949e-07,
+      "loss": 0.5938,
+      "step": 26656
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8684535959423777e-07,
+      "loss": 0.6228,
+      "step": 26657
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8658649101726333e-07,
+      "loss": 0.6356,
+      "step": 26658
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.863277376050921e-07,
+      "loss": 0.589,
+      "step": 26659
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.86069099360794e-07,
+      "loss": 0.6389,
+      "step": 26660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8581057628743636e-07,
+      "loss": 0.6473,
+      "step": 26661
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.855521683880813e-07,
+      "loss": 0.6556,
+      "step": 26662
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.852938756657953e-07,
+      "loss": 0.6493,
+      "step": 26663
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.85035698123638e-07,
+      "loss": 0.5576,
+      "step": 26664
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8477763576467253e-07,
+      "loss": 0.6742,
+      "step": 26665
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8451968859195876e-07,
+      "loss": 0.6564,
+      "step": 26666
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.842618566085542e-07,
+      "loss": 0.6267,
+      "step": 26667
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.840041398175153e-07,
+      "loss": 0.6289,
+      "step": 26668
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8374653822189846e-07,
+      "loss": 0.5885,
+      "step": 26669
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8348905182475683e-07,
+      "loss": 0.5788,
+      "step": 26670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8323168062914355e-07,
+      "loss": 0.6056,
+      "step": 26671
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8297442463811164e-07,
+      "loss": 0.6601,
+      "step": 26672
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8271728385471095e-07,
+      "loss": 0.6374,
+      "step": 26673
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.824602582819869e-07,
+      "loss": 0.621,
+      "step": 26674
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.822033479229902e-07,
+      "loss": 0.5917,
+      "step": 26675
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.819465527807663e-07,
+      "loss": 0.6079,
+      "step": 26676
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8168987285835946e-07,
+      "loss": 0.6069,
+      "step": 26677
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8143330815881167e-07,
+      "loss": 0.6219,
+      "step": 26678
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.811768586851682e-07,
+      "loss": 0.6055,
+      "step": 26679
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.809205244404656e-07,
+      "loss": 0.6294,
+      "step": 26680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.806643054277458e-07,
+      "loss": 0.6368,
+      "step": 26681
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.804082016500453e-07,
+      "loss": 0.6043,
+      "step": 26682
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.8015221311040064e-07,
+      "loss": 0.6253,
+      "step": 26683
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7989633981184706e-07,
+      "loss": 0.652,
+      "step": 26684
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.796405817574199e-07,
+      "loss": 0.6438,
+      "step": 26685
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7938493895014907e-07,
+      "loss": 0.6544,
+      "step": 26686
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.791294113930665e-07,
+      "loss": 0.6349,
+      "step": 26687
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.788739990892009e-07,
+      "loss": 0.6423,
+      "step": 26688
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7861870204158206e-07,
+      "loss": 0.6426,
+      "step": 26689
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7836352025323535e-07,
+      "loss": 0.6695,
+      "step": 26690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.781084537271883e-07,
+      "loss": 0.6016,
+      "step": 26691
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.77853502466463e-07,
+      "loss": 0.6199,
+      "step": 26692
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7759866647408373e-07,
+      "loss": 0.5942,
+      "step": 26693
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7734394575307024e-07,
+      "loss": 0.6206,
+      "step": 26694
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7708934030644343e-07,
+      "loss": 0.5986,
+      "step": 26695
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.76834850137222e-07,
+      "loss": 0.6177,
+      "step": 26696
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.765804752484247e-07,
+      "loss": 0.5922,
+      "step": 26697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7632621564306463e-07,
+      "loss": 0.6503,
+      "step": 26698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7607207132415824e-07,
+      "loss": 0.6137,
+      "step": 26699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7581804229471874e-07,
+      "loss": 0.6298,
+      "step": 26700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7556412855775707e-07,
+      "loss": 0.5922,
+      "step": 26701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7531033011628405e-07,
+      "loss": 0.6168,
+      "step": 26702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7505664697330956e-07,
+      "loss": 0.6349,
+      "step": 26703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7480307913184015e-07,
+      "loss": 0.5508,
+      "step": 26704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7454962659488214e-07,
+      "loss": 0.6183,
+      "step": 26705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7429628936544104e-07,
+      "loss": 0.6545,
+      "step": 26706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.740430674465211e-07,
+      "loss": 0.6798,
+      "step": 26707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.737899608411232e-07,
+      "loss": 0.6476,
+      "step": 26708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.735369695522505e-07,
+      "loss": 0.6015,
+      "step": 26709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7328409358290065e-07,
+      "loss": 0.5936,
+      "step": 26710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.730313329360712e-07,
+      "loss": 0.5695,
+      "step": 26711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.727786876147609e-07,
+      "loss": 0.6225,
+      "step": 26712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.725261576219629e-07,
+      "loss": 0.6565,
+      "step": 26713
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7227374296067365e-07,
+      "loss": 0.6155,
+      "step": 26714
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.720214436338864e-07,
+      "loss": 0.6238,
+      "step": 26715
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7176925964458865e-07,
+      "loss": 0.5993,
+      "step": 26716
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.715171909957737e-07,
+      "loss": 0.674,
+      "step": 26717
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.712652376904279e-07,
+      "loss": 0.6038,
+      "step": 26718
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7101339973154004e-07,
+      "loss": 0.686,
+      "step": 26719
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7076167712209664e-07,
+      "loss": 0.6362,
+      "step": 26720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.705100698650809e-07,
+      "loss": 0.6282,
+      "step": 26721
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.702585779634748e-07,
+      "loss": 0.6564,
+      "step": 26722
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.700072014202637e-07,
+      "loss": 0.6727,
+      "step": 26723
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.697559402384231e-07,
+      "loss": 0.6192,
+      "step": 26724
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6950479442093503e-07,
+      "loss": 0.6596,
+      "step": 26725
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6925376397077816e-07,
+      "loss": 0.6289,
+      "step": 26726
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.690028488909269e-07,
+      "loss": 0.6985,
+      "step": 26727
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.687520491843565e-07,
+      "loss": 0.6098,
+      "step": 26728
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.685013648540413e-07,
+      "loss": 0.5937,
+      "step": 26729
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.682507959029512e-07,
+      "loss": 0.6792,
+      "step": 26730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6800034233405935e-07,
+      "loss": 0.6434,
+      "step": 26731
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.677500041503356e-07,
+      "loss": 0.6107,
+      "step": 26732
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6749978135474643e-07,
+      "loss": 0.5445,
+      "step": 26733
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6724967395025836e-07,
+      "loss": 0.6218,
+      "step": 26734
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6699968193983796e-07,
+      "loss": 0.6062,
+      "step": 26735
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6674980532644725e-07,
+      "loss": 0.5972,
+      "step": 26736
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6650004411305165e-07,
+      "loss": 0.634,
+      "step": 26737
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.66250398302611e-07,
+      "loss": 0.6357,
+      "step": 26738
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.660008678980852e-07,
+      "loss": 0.6695,
+      "step": 26739
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6575145290243185e-07,
+      "loss": 0.6125,
+      "step": 26740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.655021533186097e-07,
+      "loss": 0.6438,
+      "step": 26741
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.652529691495731e-07,
+      "loss": 0.6118,
+      "step": 26742
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.650039003982763e-07,
+      "loss": 0.6336,
+      "step": 26743
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.647549470676758e-07,
+      "loss": 0.667,
+      "step": 26744
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6450610916071815e-07,
+      "loss": 0.6449,
+      "step": 26745
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.642573866803566e-07,
+      "loss": 0.5996,
+      "step": 26746
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6400877962954094e-07,
+      "loss": 0.6919,
+      "step": 26747
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6376028801121446e-07,
+      "loss": 0.6542,
+      "step": 26748
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6351191182832803e-07,
+      "loss": 0.5726,
+      "step": 26749
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.632636510838249e-07,
+      "loss": 0.6262,
+      "step": 26750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6301550578064715e-07,
+      "loss": 0.6271,
+      "step": 26751
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.627674759217391e-07,
+      "loss": 0.6293,
+      "step": 26752
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6251956151003956e-07,
+      "loss": 0.6618,
+      "step": 26753
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.622717625484883e-07,
+      "loss": 0.6871,
+      "step": 26754
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6202407904002303e-07,
+      "loss": 0.6293,
+      "step": 26755
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.617765109875836e-07,
+      "loss": 0.6337,
+      "step": 26756
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6152905839409993e-07,
+      "loss": 0.6018,
+      "step": 26757
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.612817212625096e-07,
+      "loss": 0.6325,
+      "step": 26758
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6103449959574476e-07,
+      "loss": 0.5942,
+      "step": 26759
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.607873933967331e-07,
+      "loss": 0.6508,
+      "step": 26760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6054040266840884e-07,
+      "loss": 0.6323,
+      "step": 26761
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.602935274136997e-07,
+      "loss": 0.6102,
+      "step": 26762
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6004676763553007e-07,
+      "loss": 0.666,
+      "step": 26763
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5980012333682636e-07,
+      "loss": 0.6129,
+      "step": 26764
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5955359452051634e-07,
+      "loss": 0.6224,
+      "step": 26765
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.593071811895165e-07,
+      "loss": 0.6344,
+      "step": 26766
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.590608833467545e-07,
+      "loss": 0.5977,
+      "step": 26767
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5881470099514807e-07,
+      "loss": 0.6106,
+      "step": 26768
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.585686341376159e-07,
+      "loss": 0.6064,
+      "step": 26769
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5832268277707463e-07,
+      "loss": 0.6446,
+      "step": 26770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.58076846916443e-07,
+      "loss": 0.6207,
+      "step": 26771
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.57831126558632e-07,
+      "loss": 0.6571,
+      "step": 26772
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5758552170655706e-07,
+      "loss": 0.5959,
+      "step": 26773
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5734003236313145e-07,
+      "loss": 0.6187,
+      "step": 26774
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.570946585312639e-07,
+      "loss": 0.5932,
+      "step": 26775
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.568494002138633e-07,
+      "loss": 0.6123,
+      "step": 26776
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5660425741383943e-07,
+      "loss": 0.6655,
+      "step": 26777
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.563592301340945e-07,
+      "loss": 0.6069,
+      "step": 26778
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.561143183775405e-07,
+      "loss": 0.5975,
+      "step": 26779
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.558695221470753e-07,
+      "loss": 0.6653,
+      "step": 26780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5562484144560306e-07,
+      "loss": 0.6121,
+      "step": 26781
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.55380276276026e-07,
+      "loss": 0.667,
+      "step": 26782
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.551358266412429e-07,
+      "loss": 0.6695,
+      "step": 26783
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5489149254415036e-07,
+      "loss": 0.6468,
+      "step": 26784
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5464727398764823e-07,
+      "loss": 0.6121,
+      "step": 26785
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5440317097462974e-07,
+      "loss": 0.6088,
+      "step": 26786
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5415918350799043e-07,
+      "loss": 0.6309,
+      "step": 26787
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5391531159062345e-07,
+      "loss": 0.6985,
+      "step": 26788
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5367155522541765e-07,
+      "loss": 0.6558,
+      "step": 26789
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5342791441526406e-07,
+      "loss": 0.7036,
+      "step": 26790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.531843891630548e-07,
+      "loss": 0.658,
+      "step": 26791
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5294097947167304e-07,
+      "loss": 0.6679,
+      "step": 26792
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.526976853440055e-07,
+      "loss": 0.5742,
+      "step": 26793
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.524545067829387e-07,
+      "loss": 0.6039,
+      "step": 26794
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5221144379135255e-07,
+      "loss": 0.5983,
+      "step": 26795
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.519684963721314e-07,
+      "loss": 0.6353,
+      "step": 26796
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5172566452815626e-07,
+      "loss": 0.7335,
+      "step": 26797
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.514829482623038e-07,
+      "loss": 0.6161,
+      "step": 26798
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5124034757745275e-07,
+      "loss": 0.6488,
+      "step": 26799
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5099786247648085e-07,
+      "loss": 0.6491,
+      "step": 26800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.507554929622613e-07,
+      "loss": 0.6393,
+      "step": 26801
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.505132390376663e-07,
+      "loss": 0.647,
+      "step": 26802
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5027110070557246e-07,
+      "loss": 0.6696,
+      "step": 26803
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5002907796884747e-07,
+      "loss": 0.6266,
+      "step": 26804
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.497871708303612e-07,
+      "loss": 0.6285,
+      "step": 26805
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.495453792929836e-07,
+      "loss": 0.662,
+      "step": 26806
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.49303703359578e-07,
+      "loss": 0.6136,
+      "step": 26807
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.49062143033012e-07,
+      "loss": 0.6073,
+      "step": 26808
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.488206983161501e-07,
+      "loss": 0.7008,
+      "step": 26809
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4857936921185433e-07,
+      "loss": 0.6218,
+      "step": 26810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.483381557229847e-07,
+      "loss": 0.6173,
+      "step": 26811
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4809705785240337e-07,
+      "loss": 0.6468,
+      "step": 26812
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4785607560296686e-07,
+      "loss": 0.6412,
+      "step": 26813
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.476152089775319e-07,
+      "loss": 0.6466,
+      "step": 26814
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.473744579789572e-07,
+      "loss": 0.6159,
+      "step": 26815
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4713382261009497e-07,
+      "loss": 0.6424,
+      "step": 26816
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.468933028737985e-07,
+      "loss": 0.6636,
+      "step": 26817
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.466528987729211e-07,
+      "loss": 0.6288,
+      "step": 26818
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.464126103103104e-07,
+      "loss": 0.6525,
+      "step": 26819
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4617243748881635e-07,
+      "loss": 0.6741,
+      "step": 26820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4593238031128676e-07,
+      "loss": 0.5968,
+      "step": 26821
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4569243878056704e-07,
+      "loss": 0.6463,
+      "step": 26822
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4545261289950385e-07,
+      "loss": 0.6297,
+      "step": 26823
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4521290267093935e-07,
+      "loss": 0.6368,
+      "step": 26824
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4497330809771566e-07,
+      "loss": 0.5876,
+      "step": 26825
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4473382918267395e-07,
+      "loss": 0.6188,
+      "step": 26826
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.444944659286519e-07,
+      "loss": 0.5795,
+      "step": 26827
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.442552183384894e-07,
+      "loss": 0.584,
+      "step": 26828
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4401608641502314e-07,
+      "loss": 0.6807,
+      "step": 26829
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4377707016108755e-07,
+      "loss": 0.6376,
+      "step": 26830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.435381695795158e-07,
+      "loss": 0.6491,
+      "step": 26831
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4329938467314127e-07,
+      "loss": 0.608,
+      "step": 26832
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4306071544479395e-07,
+      "loss": 0.5697,
+      "step": 26833
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.428221618973059e-07,
+      "loss": 0.611,
+      "step": 26834
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.425837240335027e-07,
+      "loss": 0.6687,
+      "step": 26835
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4234540185621435e-07,
+      "loss": 0.6758,
+      "step": 26836
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.42107195368263e-07,
+      "loss": 0.6176,
+      "step": 26837
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.418691045724753e-07,
+      "loss": 0.6547,
+      "step": 26838
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.416311294716733e-07,
+      "loss": 0.6126,
+      "step": 26839
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.413932700686794e-07,
+      "loss": 0.605,
+      "step": 26840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4115552636631236e-07,
+      "loss": 0.6131,
+      "step": 26841
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.409178983673921e-07,
+      "loss": 0.6462,
+      "step": 26842
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.406803860747342e-07,
+      "loss": 0.6185,
+      "step": 26843
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.404429894911575e-07,
+      "loss": 0.6374,
+      "step": 26844
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.402057086194731e-07,
+      "loss": 0.6336,
+      "step": 26845
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.399685434624976e-07,
+      "loss": 0.6307,
+      "step": 26846
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3973149402304216e-07,
+      "loss": 0.6556,
+      "step": 26847
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.394945603039156e-07,
+      "loss": 0.6059,
+      "step": 26848
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.39257742307929e-07,
+      "loss": 0.6043,
+      "step": 26849
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3902104003788783e-07,
+      "loss": 0.6397,
+      "step": 26850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3878445349660105e-07,
+      "loss": 0.6487,
+      "step": 26851
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3854798268687195e-07,
+      "loss": 0.6391,
+      "step": 26852
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3831162761150606e-07,
+      "loss": 0.6218,
+      "step": 26853
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3807538827330333e-07,
+      "loss": 0.6737,
+      "step": 26854
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.37839264675066e-07,
+      "loss": 0.5954,
+      "step": 26855
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3760325681959296e-07,
+      "loss": 0.6044,
+      "step": 26856
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3736736470968302e-07,
+      "loss": 0.6436,
+      "step": 26857
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3713158834813287e-07,
+      "loss": 0.6449,
+      "step": 26858
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3689592773773807e-07,
+      "loss": 0.5965,
+      "step": 26859
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.366603828812919e-07,
+      "loss": 0.6692,
+      "step": 26860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.364249537815877e-07,
+      "loss": 0.6626,
+      "step": 26861
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3618964044141768e-07,
+      "loss": 0.648,
+      "step": 26862
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.359544428635696e-07,
+      "loss": 0.6762,
+      "step": 26863
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3571936105083347e-07,
+      "loss": 0.6461,
+      "step": 26864
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3548439500599707e-07,
+      "loss": 0.6155,
+      "step": 26865
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.352495447318437e-07,
+      "loss": 0.6418,
+      "step": 26866
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3501481023116002e-07,
+      "loss": 0.6292,
+      "step": 26867
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3478019150672938e-07,
+      "loss": 0.6745,
+      "step": 26868
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3454568856133175e-07,
+      "loss": 0.6146,
+      "step": 26869
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3431130139774827e-07,
+      "loss": 0.5839,
+      "step": 26870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.340770300187589e-07,
+      "loss": 0.5825,
+      "step": 26871
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3384287442714038e-07,
+      "loss": 0.6219,
+      "step": 26872
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3360883462566818e-07,
+      "loss": 0.6576,
+      "step": 26873
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3337491061711681e-07,
+      "loss": 0.6364,
+      "step": 26874
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.331411024042618e-07,
+      "loss": 0.6429,
+      "step": 26875
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3290740998987426e-07,
+      "loss": 0.6403,
+      "step": 26876
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3267383337672533e-07,
+      "loss": 0.6687,
+      "step": 26877
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3244037256758278e-07,
+      "loss": 0.6589,
+      "step": 26878
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3220702756521662e-07,
+      "loss": 0.631,
+      "step": 26879
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.319737983723913e-07,
+      "loss": 0.5994,
+      "step": 26880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3174068499187352e-07,
+      "loss": 0.6647,
+      "step": 26881
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.315076874264266e-07,
+      "loss": 0.6655,
+      "step": 26882
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.31274805678815e-07,
+      "loss": 0.6032,
+      "step": 26883
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.310420397517965e-07,
+      "loss": 0.6236,
+      "step": 26884
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3080938964813227e-07,
+      "loss": 0.6266,
+      "step": 26885
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3057685537058117e-07,
+      "loss": 0.6369,
+      "step": 26886
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.303444369218999e-07,
+      "loss": 0.5797,
+      "step": 26887
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3011213430484403e-07,
+      "loss": 0.6665,
+      "step": 26888
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2987994752216914e-07,
+      "loss": 0.641,
+      "step": 26889
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2964787657662524e-07,
+      "loss": 0.6547,
+      "step": 26890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2941592147096458e-07,
+      "loss": 0.6916,
+      "step": 26891
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2918408220793942e-07,
+      "loss": 0.6543,
+      "step": 26892
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2895235879029643e-07,
+      "loss": 0.6217,
+      "step": 26893
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2872075122078342e-07,
+      "loss": 0.6334,
+      "step": 26894
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2848925950214817e-07,
+      "loss": 0.6395,
+      "step": 26895
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2825788363713186e-07,
+      "loss": 0.7169,
+      "step": 26896
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2802662362848115e-07,
+      "loss": 0.638,
+      "step": 26897
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2779547947893498e-07,
+      "loss": 0.6254,
+      "step": 26898
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.275644511912367e-07,
+      "loss": 0.6243,
+      "step": 26899
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2733353876812302e-07,
+      "loss": 0.6861,
+      "step": 26900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2710274221233396e-07,
+      "loss": 0.7436,
+      "step": 26901
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2687206152660401e-07,
+      "loss": 0.6301,
+      "step": 26902
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2664149671366763e-07,
+      "loss": 0.5994,
+      "step": 26903
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2641104777626045e-07,
+      "loss": 0.6289,
+      "step": 26904
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2618071471711466e-07,
+      "loss": 0.684,
+      "step": 26905
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2595049753896038e-07,
+      "loss": 0.7002,
+      "step": 26906
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2572039624452647e-07,
+      "loss": 0.6177,
+      "step": 26907
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.25490410836543e-07,
+      "loss": 0.5911,
+      "step": 26908
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2526054131773444e-07,
+      "loss": 0.6797,
+      "step": 26909
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2503078769082643e-07,
+      "loss": 0.6294,
+      "step": 26910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2480114995854562e-07,
+      "loss": 0.6211,
+      "step": 26911
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.245716281236121e-07,
+      "loss": 0.6256,
+      "step": 26912
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2434222218874813e-07,
+      "loss": 0.6675,
+      "step": 26913
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2411293215667373e-07,
+      "loss": 0.6022,
+      "step": 26914
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2388375803010566e-07,
+      "loss": 0.6582,
+      "step": 26915
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2365469981176392e-07,
+      "loss": 0.5902,
+      "step": 26916
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.234257575043619e-07,
+      "loss": 0.617,
+      "step": 26917
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2319693111061636e-07,
+      "loss": 0.6936,
+      "step": 26918
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.229682206332373e-07,
+      "loss": 0.6328,
+      "step": 26919
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2273962607493816e-07,
+      "loss": 0.6664,
+      "step": 26920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2251114743843005e-07,
+      "loss": 0.5389,
+      "step": 26921
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2228278472641973e-07,
+      "loss": 0.6833,
+      "step": 26922
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2205453794161614e-07,
+      "loss": 0.6395,
+      "step": 26923
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2182640708672708e-07,
+      "loss": 0.6548,
+      "step": 26924
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.215983921644538e-07,
+      "loss": 0.671,
+      "step": 26925
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.213704931775007e-07,
+      "loss": 0.6475,
+      "step": 26926
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2114271012857125e-07,
+      "loss": 0.6099,
+      "step": 26927
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2091504302036438e-07,
+      "loss": 0.6035,
+      "step": 26928
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2068749185558126e-07,
+      "loss": 0.6153,
+      "step": 26929
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2046005663691972e-07,
+      "loss": 0.6293,
+      "step": 26930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2023273736707428e-07,
+      "loss": 0.6776,
+      "step": 26931
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.200055340487406e-07,
+      "loss": 0.6453,
+      "step": 26932
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1977844668461422e-07,
+      "loss": 0.6395,
+      "step": 26933
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.195514752773853e-07,
+      "loss": 0.6808,
+      "step": 26934
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1932461982974607e-07,
+      "loss": 0.6673,
+      "step": 26935
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1909788034438772e-07,
+      "loss": 0.6076,
+      "step": 26936
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.188712568239959e-07,
+      "loss": 0.6604,
+      "step": 26937
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1864474927125735e-07,
+      "loss": 0.6053,
+      "step": 26938
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1841835768885988e-07,
+      "loss": 0.6168,
+      "step": 26939
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1819208207948695e-07,
+      "loss": 0.5862,
+      "step": 26940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1796592244581971e-07,
+      "loss": 0.6213,
+      "step": 26941
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.177398787905427e-07,
+      "loss": 0.6094,
+      "step": 26942
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1751395111633268e-07,
+      "loss": 0.5953,
+      "step": 26943
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1728813942586967e-07,
+      "loss": 0.5967,
+      "step": 26944
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.170624437218305e-07,
+      "loss": 0.6129,
+      "step": 26945
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1683686400689298e-07,
+      "loss": 0.6197,
+      "step": 26946
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.166114002837283e-07,
+      "loss": 0.6668,
+      "step": 26947
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1638605255501433e-07,
+      "loss": 0.6237,
+      "step": 26948
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1616082082341784e-07,
+      "loss": 0.5887,
+      "step": 26949
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1593570509161222e-07,
+      "loss": 0.6021,
+      "step": 26950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1571070536226536e-07,
+      "loss": 0.6427,
+      "step": 26951
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.154858216380451e-07,
+      "loss": 0.6088,
+      "step": 26952
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.152610539216182e-07,
+      "loss": 0.651,
+      "step": 26953
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1503640221564926e-07,
+      "loss": 0.6357,
+      "step": 26954
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1481186652280162e-07,
+      "loss": 0.6302,
+      "step": 26955
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.145874468457376e-07,
+      "loss": 0.5978,
+      "step": 26956
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1436314318711849e-07,
+      "loss": 0.6152,
+      "step": 26957
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1413895554960318e-07,
+      "loss": 0.6505,
+      "step": 26958
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1391488393584959e-07,
+      "loss": 0.6299,
+      "step": 26959
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1369092834851447e-07,
+      "loss": 0.7024,
+      "step": 26960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1346708879025346e-07,
+      "loss": 0.5895,
+      "step": 26961
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1324336526371892e-07,
+      "loss": 0.6285,
+      "step": 26962
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1301975777156536e-07,
+      "loss": 0.6615,
+      "step": 26963
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.12796266316444e-07,
+      "loss": 0.6175,
+      "step": 26964
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1257289090100274e-07,
+      "loss": 0.6688,
+      "step": 26965
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1234963152789166e-07,
+      "loss": 0.6085,
+      "step": 26966
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1212648819975644e-07,
+      "loss": 0.644,
+      "step": 26967
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1190346091924496e-07,
+      "loss": 0.6186,
+      "step": 26968
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.116805496889973e-07,
+      "loss": 0.6887,
+      "step": 26969
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1145775451166028e-07,
+      "loss": 0.6066,
+      "step": 26970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1123507538987508e-07,
+      "loss": 0.6109,
+      "step": 26971
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1101251232628073e-07,
+      "loss": 0.6297,
+      "step": 26972
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.107900653235151e-07,
+      "loss": 0.6575,
+      "step": 26973
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.105677343842183e-07,
+      "loss": 0.6304,
+      "step": 26974
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.1034551951102268e-07,
+      "loss": 0.6037,
+      "step": 26975
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.10123420706565e-07,
+      "loss": 0.7023,
+      "step": 26976
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.0990143797347984e-07,
+      "loss": 0.5515,
+      "step": 26977
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.096795713143973e-07,
+      "loss": 0.6801,
+      "step": 26978
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.0945782073194754e-07,
+      "loss": 0.5708,
+      "step": 26979
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.0923618622876173e-07,
+      "loss": 0.6239,
+      "step": 26980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.090146678074645e-07,
+      "loss": 0.6256,
+      "step": 26981
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.087932654706859e-07,
+      "loss": 0.638,
+      "step": 26982
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.0857197922104834e-07,
+      "loss": 0.6223,
+      "step": 26983
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.0835080906117633e-07,
+      "loss": 0.6265,
+      "step": 26984
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.0812975499369115e-07,
+      "loss": 0.6262,
+      "step": 26985
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.0790881702121623e-07,
+      "loss": 0.6839,
+      "step": 26986
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0768799514636728e-07,
+      "loss": 0.6646,
+      "step": 26987
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.074672893717644e-07,
+      "loss": 0.6784,
+      "step": 26988
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0724669970002664e-07,
+      "loss": 0.6259,
+      "step": 26989
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0702622613376632e-07,
+      "loss": 0.6441,
+      "step": 26990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0680586867559692e-07,
+      "loss": 0.6272,
+      "step": 26991
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0658562732813415e-07,
+      "loss": 0.6434,
+      "step": 26992
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0636550209398588e-07,
+      "loss": 0.66,
+      "step": 26993
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0614549297576447e-07,
+      "loss": 0.6333,
+      "step": 26994
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0592559997607785e-07,
+      "loss": 0.6611,
+      "step": 26995
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0570582309753285e-07,
+      "loss": 0.6519,
+      "step": 26996
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0548616234273399e-07,
+      "loss": 0.6135,
+      "step": 26997
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0526661771428925e-07,
+      "loss": 0.6424,
+      "step": 26998
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0504718921479648e-07,
+      "loss": 0.6502,
+      "step": 26999
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0482787684686035e-07,
+      "loss": 0.6301,
+      "step": 27000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0460868061308203e-07,
+      "loss": 0.6161,
+      "step": 27001
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.043896005160584e-07,
+      "loss": 0.5902,
+      "step": 27002
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0417063655838732e-07,
+      "loss": 0.662,
+      "step": 27003
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0395178874266565e-07,
+      "loss": 0.573,
+      "step": 27004
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0373305707148684e-07,
+      "loss": 0.6433,
+      "step": 27005
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0351444154744437e-07,
+      "loss": 0.6237,
+      "step": 27006
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0329594217313287e-07,
+      "loss": 0.6391,
+      "step": 27007
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0307755895114023e-07,
+      "loss": 0.6767,
+      "step": 27008
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0285929188405551e-07,
+      "loss": 0.6403,
+      "step": 27009
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0264114097446884e-07,
+      "loss": 0.6763,
+      "step": 27010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0242310622496374e-07,
+      "loss": 0.5961,
+      "step": 27011
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0220518763812592e-07,
+      "loss": 0.6442,
+      "step": 27012
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0198738521654105e-07,
+      "loss": 0.6475,
+      "step": 27013
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.017696989627904e-07,
+      "loss": 0.6161,
+      "step": 27014
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0155212887945418e-07,
+      "loss": 0.6449,
+      "step": 27015
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0133467496911364e-07,
+      "loss": 0.6335,
+      "step": 27016
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0111733723434555e-07,
+      "loss": 0.6338,
+      "step": 27017
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0090011567772572e-07,
+      "loss": 0.6334,
+      "step": 27018
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0068301030183313e-07,
+      "loss": 0.6204,
+      "step": 27019
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0046602110923796e-07,
+      "loss": 0.6652,
+      "step": 27020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0024914810251483e-07,
+      "loss": 0.6553,
+      "step": 27021
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0003239128423614e-07,
+      "loss": 0.6311,
+      "step": 27022
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9981575065696978e-07,
+      "loss": 0.6656,
+      "step": 27023
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9959922622328375e-07,
+      "loss": 0.6516,
+      "step": 27024
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.993828179857471e-07,
+      "loss": 0.65,
+      "step": 27025
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9916652594692552e-07,
+      "loss": 0.6146,
+      "step": 27026
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9895035010938257e-07,
+      "loss": 0.6671,
+      "step": 27027
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9873429047568283e-07,
+      "loss": 0.7014,
+      "step": 27028
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9851834704838535e-07,
+      "loss": 0.6532,
+      "step": 27029
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9830251983005145e-07,
+      "loss": 0.5848,
+      "step": 27030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9808680882324128e-07,
+      "loss": 0.6328,
+      "step": 27031
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9787121403051057e-07,
+      "loss": 0.6906,
+      "step": 27032
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.976557354544184e-07,
+      "loss": 0.6437,
+      "step": 27033
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9744037309751607e-07,
+      "loss": 0.623,
+      "step": 27034
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9722512696235818e-07,
+      "loss": 0.6588,
+      "step": 27035
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9700999705149715e-07,
+      "loss": 0.6625,
+      "step": 27036
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9679498336748315e-07,
+      "loss": 0.6704,
+      "step": 27037
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9658008591286638e-07,
+      "loss": 0.6504,
+      "step": 27038
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9636530469019478e-07,
+      "loss": 0.583,
+      "step": 27039
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9615063970201297e-07,
+      "loss": 0.5704,
+      "step": 27040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9593609095086674e-07,
+      "loss": 0.6224,
+      "step": 27041
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.957216584393018e-07,
+      "loss": 0.5598,
+      "step": 27042
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9550734216985724e-07,
+      "loss": 0.6289,
+      "step": 27043
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9529314214507767e-07,
+      "loss": 0.6635,
+      "step": 27044
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9507905836750107e-07,
+      "loss": 0.6803,
+      "step": 27045
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9486509083966432e-07,
+      "loss": 0.6682,
+      "step": 27046
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9465123956410537e-07,
+      "loss": 0.6021,
+      "step": 27047
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9443750454336109e-07,
+      "loss": 0.6452,
+      "step": 27048
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9422388577996277e-07,
+      "loss": 0.5757,
+      "step": 27049
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9401038327644505e-07,
+      "loss": 0.6353,
+      "step": 27050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9379699703534037e-07,
+      "loss": 0.6546,
+      "step": 27051
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.935837270591756e-07,
+      "loss": 0.6855,
+      "step": 27052
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9337057335048204e-07,
+      "loss": 0.6499,
+      "step": 27053
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9315753591178432e-07,
+      "loss": 0.6378,
+      "step": 27054
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9294461474561043e-07,
+      "loss": 0.6129,
+      "step": 27055
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9273180985448504e-07,
+      "loss": 0.6851,
+      "step": 27056
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.925191212409294e-07,
+      "loss": 0.6541,
+      "step": 27057
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9230654890746714e-07,
+      "loss": 0.646,
+      "step": 27058
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.920940928566162e-07,
+      "loss": 0.6338,
+      "step": 27059
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9188175309089675e-07,
+      "loss": 0.6227,
+      "step": 27060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.916695296128268e-07,
+      "loss": 0.6132,
+      "step": 27061
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9145742242492217e-07,
+      "loss": 0.6594,
+      "step": 27062
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9124543152969856e-07,
+      "loss": 0.6672,
+      "step": 27063
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9103355692966842e-07,
+      "loss": 0.6569,
+      "step": 27064
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9082179862734306e-07,
+      "loss": 0.6397,
+      "step": 27065
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9061015662523384e-07,
+      "loss": 0.6418,
+      "step": 27066
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9039863092584986e-07,
+      "loss": 0.6358,
+      "step": 27067
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9018722153170023e-07,
+      "loss": 0.611,
+      "step": 27068
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8997592844529066e-07,
+      "loss": 0.6204,
+      "step": 27069
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8976475166912588e-07,
+      "loss": 0.6439,
+      "step": 27070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8955369120570944e-07,
+      "loss": 0.6767,
+      "step": 27071
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8934274705754486e-07,
+      "loss": 0.6247,
+      "step": 27072
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8913191922713236e-07,
+      "loss": 0.6255,
+      "step": 27073
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.889212077169722e-07,
+      "loss": 0.6324,
+      "step": 27074
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8871061252956235e-07,
+      "loss": 0.6628,
+      "step": 27075
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.885001336673986e-07,
+      "loss": 0.6699,
+      "step": 27076
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8828977113297786e-07,
+      "loss": 0.6003,
+      "step": 27077
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8807952492879367e-07,
+      "loss": 0.6901,
+      "step": 27078
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8786939505733847e-07,
+      "loss": 0.6563,
+      "step": 27079
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8765938152110473e-07,
+      "loss": 0.6391,
+      "step": 27080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8744948432258157e-07,
+      "loss": 0.6457,
+      "step": 27081
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8723970346425813e-07,
+      "loss": 0.6387,
+      "step": 27082
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8703003894862016e-07,
+      "loss": 0.6628,
+      "step": 27083
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8682049077815456e-07,
+      "loss": 0.6258,
+      "step": 27084
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8661105895534604e-07,
+      "loss": 0.6408,
+      "step": 27085
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8640174348267815e-07,
+      "loss": 0.6082,
+      "step": 27086
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8619254436263224e-07,
+      "loss": 0.6373,
+      "step": 27087
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8598346159768742e-07,
+      "loss": 0.6652,
+      "step": 27088
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8577449519032288e-07,
+      "loss": 0.6102,
+      "step": 27089
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8556564514301655e-07,
+      "loss": 0.6014,
+      "step": 27090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.853569114582454e-07,
+      "loss": 0.6043,
+      "step": 27091
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8514829413848522e-07,
+      "loss": 0.6143,
+      "step": 27092
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8493979318620625e-07,
+      "loss": 0.6365,
+      "step": 27093
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8473140860388207e-07,
+      "loss": 0.628,
+      "step": 27094
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.845231403939829e-07,
+      "loss": 0.686,
+      "step": 27095
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.843149885589801e-07,
+      "loss": 0.5894,
+      "step": 27096
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8410695310133842e-07,
+      "loss": 0.6882,
+      "step": 27097
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8389903402352805e-07,
+      "loss": 0.6768,
+      "step": 27098
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8369123132801038e-07,
+      "loss": 0.6689,
+      "step": 27099
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.834835450172512e-07,
+      "loss": 0.6243,
+      "step": 27100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8327597509371186e-07,
+      "loss": 0.6193,
+      "step": 27101
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.830685215598549e-07,
+      "loss": 0.6533,
+      "step": 27102
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.828611844181394e-07,
+      "loss": 0.6124,
+      "step": 27103
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.826539636710234e-07,
+      "loss": 0.6242,
+      "step": 27104
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8244685932096274e-07,
+      "loss": 0.6526,
+      "step": 27105
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8223987137041432e-07,
+      "loss": 0.6344,
+      "step": 27106
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8203299982183175e-07,
+      "loss": 0.6873,
+      "step": 27107
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.818262446776675e-07,
+      "loss": 0.5985,
+      "step": 27108
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.81619605940373e-07,
+      "loss": 0.588,
+      "step": 27109
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8141308361239952e-07,
+      "loss": 0.6808,
+      "step": 27110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8120667769619404e-07,
+      "loss": 0.6546,
+      "step": 27111
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.810003881942035e-07,
+      "loss": 0.6812,
+      "step": 27112
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8079421510887485e-07,
+      "loss": 0.6135,
+      "step": 27113
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8058815844265277e-07,
+      "loss": 0.6309,
+      "step": 27114
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8038221819797863e-07,
+      "loss": 0.6346,
+      "step": 27115
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8017639437729607e-07,
+      "loss": 0.6544,
+      "step": 27116
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7997068698304422e-07,
+      "loss": 0.6369,
+      "step": 27117
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7976509601766222e-07,
+      "loss": 0.6052,
+      "step": 27118
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7955962148358707e-07,
+      "loss": 0.6245,
+      "step": 27119
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.793542633832568e-07,
+      "loss": 0.6059,
+      "step": 27120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7914902171910388e-07,
+      "loss": 0.6406,
+      "step": 27121
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7894389649356304e-07,
+      "loss": 0.6664,
+      "step": 27122
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7873888770906678e-07,
+      "loss": 0.6745,
+      "step": 27123
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7853399536804427e-07,
+      "loss": 0.7024,
+      "step": 27124
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.783292194729247e-07,
+      "loss": 0.6356,
+      "step": 27125
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.781245600261372e-07,
+      "loss": 0.622,
+      "step": 27126
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7792001703010763e-07,
+      "loss": 0.5979,
+      "step": 27127
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7771559048726295e-07,
+      "loss": 0.6256,
+      "step": 27128
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7751128040002342e-07,
+      "loss": 0.7081,
+      "step": 27129
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7730708677081375e-07,
+      "loss": 0.6026,
+      "step": 27130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.771030096020543e-07,
+      "loss": 0.6616,
+      "step": 27131
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7689904889616416e-07,
+      "loss": 0.6405,
+      "step": 27132
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7669520465556256e-07,
+      "loss": 0.5755,
+      "step": 27133
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7649147688266643e-07,
+      "loss": 0.6344,
+      "step": 27134
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.762878655798894e-07,
+      "loss": 0.6504,
+      "step": 27135
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7608437074964734e-07,
+      "loss": 0.6176,
+      "step": 27136
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7588099239435274e-07,
+      "loss": 0.641,
+      "step": 27137
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.756777305164159e-07,
+      "loss": 0.6233,
+      "step": 27138
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.754745851182482e-07,
+      "loss": 0.6308,
+      "step": 27139
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7527155620225777e-07,
+      "loss": 0.6213,
+      "step": 27140
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7506864377085043e-07,
+      "loss": 0.6299,
+      "step": 27141
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7486584782643313e-07,
+      "loss": 0.5855,
+      "step": 27142
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.746631683714095e-07,
+      "loss": 0.6606,
+      "step": 27143
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.744606054081832e-07,
+      "loss": 0.6496,
+      "step": 27144
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7425815893915677e-07,
+      "loss": 0.6083,
+      "step": 27145
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7405582896672934e-07,
+      "loss": 0.6436,
+      "step": 27146
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7385361549329904e-07,
+      "loss": 0.6592,
+      "step": 27147
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7365151852126505e-07,
+      "loss": 0.6565,
+      "step": 27148
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7344953805302212e-07,
+      "loss": 0.6507,
+      "step": 27149
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.732476740909661e-07,
+      "loss": 0.5943,
+      "step": 27150
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7304592663748954e-07,
+      "loss": 0.6802,
+      "step": 27151
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7284429569498494e-07,
+      "loss": 0.6924,
+      "step": 27152
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7264278126584266e-07,
+      "loss": 0.6353,
+      "step": 27153
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7244138335245185e-07,
+      "loss": 0.6904,
+      "step": 27154
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7224010195719954e-07,
+      "loss": 0.6203,
+      "step": 27155
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7203893708247378e-07,
+      "loss": 0.6341,
+      "step": 27156
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7183788873065933e-07,
+      "loss": 0.6365,
+      "step": 27157
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7163695690413873e-07,
+      "loss": 0.6268,
+      "step": 27158
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7143614160529454e-07,
+      "loss": 0.5908,
+      "step": 27159
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7123544283650817e-07,
+      "loss": 0.6028,
+      "step": 27160
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7103486060015994e-07,
+      "loss": 0.6124,
+      "step": 27161
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7083439489862576e-07,
+      "loss": 0.6355,
+      "step": 27162
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7063404573428587e-07,
+      "loss": 0.6299,
+      "step": 27163
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.704338131095118e-07,
+      "loss": 0.611,
+      "step": 27164
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7023369702668048e-07,
+      "loss": 0.662,
+      "step": 27165
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7003369748816223e-07,
+      "loss": 0.5802,
+      "step": 27166
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6983381449632962e-07,
+      "loss": 0.637,
+      "step": 27167
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6963404805355298e-07,
+      "loss": 0.625,
+      "step": 27168
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6943439816220154e-07,
+      "loss": 0.6617,
+      "step": 27169
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.692348648246389e-07,
+      "loss": 0.6198,
+      "step": 27170
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6903544804323324e-07,
+      "loss": 0.6363,
+      "step": 27171
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6883614782034929e-07,
+      "loss": 0.6004,
+      "step": 27172
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6863696415834853e-07,
+      "loss": 0.6482,
+      "step": 27173
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6843789705959347e-07,
+      "loss": 0.6265,
+      "step": 27174
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6823894652644447e-07,
+      "loss": 0.5618,
+      "step": 27175
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6804011256125962e-07,
+      "loss": 0.6205,
+      "step": 27176
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6784139516639709e-07,
+      "loss": 0.5948,
+      "step": 27177
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.676427943442127e-07,
+      "loss": 0.6202,
+      "step": 27178
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.674443100970602e-07,
+      "loss": 0.5855,
+      "step": 27179
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.672459424272943e-07,
+      "loss": 0.6094,
+      "step": 27180
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6704769133726652e-07,
+      "loss": 0.6777,
+      "step": 27181
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6684955682932712e-07,
+      "loss": 0.6543,
+      "step": 27182
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.666515389058243e-07,
+      "loss": 0.6209,
+      "step": 27183
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6645363756910727e-07,
+      "loss": 0.5951,
+      "step": 27184
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.662558528215219e-07,
+      "loss": 0.6956,
+      "step": 27185
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.66058184665413e-07,
+      "loss": 0.6261,
+      "step": 27186
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6586063310312539e-07,
+      "loss": 0.6013,
+      "step": 27187
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6566319813699826e-07,
+      "loss": 0.627,
+      "step": 27188
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6546587976937534e-07,
+      "loss": 0.6702,
+      "step": 27189
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6526867800259583e-07,
+      "loss": 0.6515,
+      "step": 27190
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6507159283899565e-07,
+      "loss": 0.6645,
+      "step": 27191
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6487462428091404e-07,
+      "loss": 0.6428,
+      "step": 27192
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6467777233068583e-07,
+      "loss": 0.6434,
+      "step": 27193
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6448103699064355e-07,
+      "loss": 0.6489,
+      "step": 27194
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.642844182631209e-07,
+      "loss": 0.652,
+      "step": 27195
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.640879161504483e-07,
+      "loss": 0.5956,
+      "step": 27196
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.63891530654956e-07,
+      "loss": 0.6962,
+      "step": 27197
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6369526177897221e-07,
+      "loss": 0.5834,
+      "step": 27198
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6349910952482505e-07,
+      "loss": 0.6568,
+      "step": 27199
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6330307389483823e-07,
+      "loss": 0.6465,
+      "step": 27200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6310715489133654e-07,
+      "loss": 0.6778,
+      "step": 27201
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.629113525166448e-07,
+      "loss": 0.6174,
+      "step": 27202
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6271566677308114e-07,
+      "loss": 0.6324,
+      "step": 27203
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6252009766296817e-07,
+      "loss": 0.619,
+      "step": 27204
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.623246451886251e-07,
+      "loss": 0.6056,
+      "step": 27205
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.621293093523668e-07,
+      "loss": 0.6755,
+      "step": 27206
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.619340901565114e-07,
+      "loss": 0.6331,
+      "step": 27207
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6173898760337259e-07,
+      "loss": 0.592,
+      "step": 27208
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6154400169526186e-07,
+      "loss": 0.634,
+      "step": 27209
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6134913243449402e-07,
+      "loss": 0.6402,
+      "step": 27210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6115437982337835e-07,
+      "loss": 0.6033,
+      "step": 27211
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6095974386422298e-07,
+      "loss": 0.6465,
+      "step": 27212
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.607652245593372e-07,
+      "loss": 0.6555,
+      "step": 27213
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.605708219110269e-07,
+      "loss": 0.6737,
+      "step": 27214
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6037653592159365e-07,
+      "loss": 0.6569,
+      "step": 27215
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6018236659334663e-07,
+      "loss": 0.6554,
+      "step": 27216
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5998831392858405e-07,
+      "loss": 0.6476,
+      "step": 27217
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5979437792960738e-07,
+      "loss": 0.5886,
+      "step": 27218
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5960055859871705e-07,
+      "loss": 0.6029,
+      "step": 27219
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5940685593821003e-07,
+      "loss": 0.666,
+      "step": 27220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5921326995038122e-07,
+      "loss": 0.626,
+      "step": 27221
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5901980063752988e-07,
+      "loss": 0.6524,
+      "step": 27222
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5882644800194747e-07,
+      "loss": 0.6342,
+      "step": 27223
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5863321204592662e-07,
+      "loss": 0.6523,
+      "step": 27224
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5844009277175887e-07,
+      "loss": 0.5684,
+      "step": 27225
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5824709018173233e-07,
+      "loss": 0.72,
+      "step": 27226
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5805420427813634e-07,
+      "loss": 0.6342,
+      "step": 27227
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5786143506325902e-07,
+      "loss": 0.6208,
+      "step": 27228
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.576687825393841e-07,
+      "loss": 0.6587,
+      "step": 27229
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5747624670879647e-07,
+      "loss": 0.652,
+      "step": 27230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.572838275737787e-07,
+      "loss": 0.6413,
+      "step": 27231
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.570915251366123e-07,
+      "loss": 0.6232,
+      "step": 27232
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.568993393995766e-07,
+      "loss": 0.684,
+      "step": 27233
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5670727036495193e-07,
+      "loss": 0.6378,
+      "step": 27234
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.565153180350143e-07,
+      "loss": 0.6868,
+      "step": 27235
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5632348241203853e-07,
+      "loss": 0.5763,
+      "step": 27236
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5613176349830173e-07,
+      "loss": 0.656,
+      "step": 27237
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5594016129607426e-07,
+      "loss": 0.6258,
+      "step": 27238
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5574867580762875e-07,
+      "loss": 0.6556,
+      "step": 27239
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5555730703523675e-07,
+      "loss": 0.638,
+      "step": 27240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.553660549811664e-07,
+      "loss": 0.6066,
+      "step": 27241
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5517491964768373e-07,
+      "loss": 0.6234,
+      "step": 27242
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5498390103705795e-07,
+      "loss": 0.6709,
+      "step": 27243
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5479299915155066e-07,
+      "loss": 0.5819,
+      "step": 27244
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5460221399342557e-07,
+      "loss": 0.6269,
+      "step": 27245
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5441154556494753e-07,
+      "loss": 0.6293,
+      "step": 27246
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5422099386837475e-07,
+      "loss": 0.6502,
+      "step": 27247
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5403055890596764e-07,
+      "loss": 0.6474,
+      "step": 27248
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5384024067998326e-07,
+      "loss": 0.6467,
+      "step": 27249
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5365003919267761e-07,
+      "loss": 0.6418,
+      "step": 27250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5345995444630557e-07,
+      "loss": 0.6646,
+      "step": 27251
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.532699864431242e-07,
+      "loss": 0.6483,
+      "step": 27252
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5308013518538166e-07,
+      "loss": 0.7064,
+      "step": 27253
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5289040067533067e-07,
+      "loss": 0.6146,
+      "step": 27254
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.527007829152216e-07,
+      "loss": 0.6221,
+      "step": 27255
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5251128190730047e-07,
+      "loss": 0.6285,
+      "step": 27256
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5232189765381544e-07,
+      "loss": 0.6242,
+      "step": 27257
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5213263015701252e-07,
+      "loss": 0.6706,
+      "step": 27258
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5194347941913434e-07,
+      "loss": 0.609,
+      "step": 27259
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5175444544242357e-07,
+      "loss": 0.6515,
+      "step": 27260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5156552822912284e-07,
+      "loss": 0.6144,
+      "step": 27261
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.513767277814715e-07,
+      "loss": 0.5802,
+      "step": 27262
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5118804410170551e-07,
+      "loss": 0.6577,
+      "step": 27263
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.509994771920664e-07,
+      "loss": 0.6093,
+      "step": 27264
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5081102705478578e-07,
+      "loss": 0.6797,
+      "step": 27265
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.506226936921007e-07,
+      "loss": 0.6231,
+      "step": 27266
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.504344771062438e-07,
+      "loss": 0.6018,
+      "step": 27267
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5024637729944446e-07,
+      "loss": 0.6071,
+      "step": 27268
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.500583942739342e-07,
+      "loss": 0.624,
+      "step": 27269
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.4987052803194345e-07,
+      "loss": 0.548,
+      "step": 27270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.4968277857569713e-07,
+      "loss": 0.6405,
+      "step": 27271
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.494951459074212e-07,
+      "loss": 0.6323,
+      "step": 27272
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.493076300293439e-07,
+      "loss": 0.6614,
+      "step": 27273
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.4912023094368344e-07,
+      "loss": 0.6907,
+      "step": 27274
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4893294865266472e-07,
+      "loss": 0.5816,
+      "step": 27275
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.487457831585082e-07,
+      "loss": 0.6111,
+      "step": 27276
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4855873446343093e-07,
+      "loss": 0.6537,
+      "step": 27277
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4837180256965456e-07,
+      "loss": 0.6658,
+      "step": 27278
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4818498747939058e-07,
+      "loss": 0.588,
+      "step": 27279
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4799828919485616e-07,
+      "loss": 0.6767,
+      "step": 27280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4781170771826615e-07,
+      "loss": 0.5846,
+      "step": 27281
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.47625243051831e-07,
+      "loss": 0.6786,
+      "step": 27282
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4743889519776122e-07,
+      "loss": 0.6055,
+      "step": 27283
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4725266415826834e-07,
+      "loss": 0.5877,
+      "step": 27284
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4706654993555724e-07,
+      "loss": 0.5401,
+      "step": 27285
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4688055253183732e-07,
+      "loss": 0.5993,
+      "step": 27286
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.466946719493123e-07,
+      "loss": 0.663,
+      "step": 27287
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4650890819018604e-07,
+      "loss": 0.6753,
+      "step": 27288
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4632326125666117e-07,
+      "loss": 0.5775,
+      "step": 27289
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4613773115094044e-07,
+      "loss": 0.6325,
+      "step": 27290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.45952317875222e-07,
+      "loss": 0.6512,
+      "step": 27291
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4576702143170308e-07,
+      "loss": 0.6038,
+      "step": 27292
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4558184182258182e-07,
+      "loss": 0.6331,
+      "step": 27293
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4539677905005433e-07,
+      "loss": 0.6402,
+      "step": 27294
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4521183311631437e-07,
+      "loss": 0.639,
+      "step": 27295
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4502700402355462e-07,
+      "loss": 0.6705,
+      "step": 27296
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4484229177396557e-07,
+      "loss": 0.6322,
+      "step": 27297
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.446576963697388e-07,
+      "loss": 0.6149,
+      "step": 27298
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4447321781306144e-07,
+      "loss": 0.6857,
+      "step": 27299
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4428885610612064e-07,
+      "loss": 0.6348,
+      "step": 27300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4410461125110464e-07,
+      "loss": 0.6341,
+      "step": 27301
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4392048325019613e-07,
+      "loss": 0.6715,
+      "step": 27302
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4373647210557673e-07,
+      "loss": 0.6265,
+      "step": 27303
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4355257781943134e-07,
+      "loss": 0.5706,
+      "step": 27304
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4336880039393708e-07,
+      "loss": 0.589,
+      "step": 27305
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.431851398312756e-07,
+      "loss": 0.6401,
+      "step": 27306
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4300159613362285e-07,
+      "loss": 0.6098,
+      "step": 27307
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4281816930315606e-07,
+      "loss": 0.6313,
+      "step": 27308
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4263485934204792e-07,
+      "loss": 0.6509,
+      "step": 27309
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.424516662524744e-07,
+      "loss": 0.5734,
+      "step": 27310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.42268590036605e-07,
+      "loss": 0.573,
+      "step": 27311
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.420856306966123e-07,
+      "loss": 0.6241,
+      "step": 27312
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4190278823466462e-07,
+      "loss": 0.5688,
+      "step": 27313
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4172006265293025e-07,
+      "loss": 0.644,
+      "step": 27314
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4153745395357522e-07,
+      "loss": 0.633,
+      "step": 27315
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4135496213876444e-07,
+      "loss": 0.6838,
+      "step": 27316
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4117258721066173e-07,
+      "loss": 0.6167,
+      "step": 27317
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4099032917142986e-07,
+      "loss": 0.6181,
+      "step": 27318
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4080818802322926e-07,
+      "loss": 0.6397,
+      "step": 27319
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4062616376822046e-07,
+      "loss": 0.5975,
+      "step": 27320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4044425640855953e-07,
+      "loss": 0.6546,
+      "step": 27321
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4026246594640358e-07,
+      "loss": 0.6763,
+      "step": 27322
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.400807923839098e-07,
+      "loss": 0.6162,
+      "step": 27323
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3989923572323094e-07,
+      "loss": 0.5743,
+      "step": 27324
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3971779596651968e-07,
+      "loss": 0.6318,
+      "step": 27325
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3953647311592766e-07,
+      "loss": 0.5899,
+      "step": 27326
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3935526717360425e-07,
+      "loss": 0.6955,
+      "step": 27327
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3917417814169777e-07,
+      "loss": 0.641,
+      "step": 27328
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3899320602235422e-07,
+      "loss": 0.6392,
+      "step": 27329
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3881235081772194e-07,
+      "loss": 0.6097,
+      "step": 27330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.386316125299436e-07,
+      "loss": 0.5944,
+      "step": 27331
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3845099116116201e-07,
+      "loss": 0.6228,
+      "step": 27332
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3827048671351871e-07,
+      "loss": 0.6339,
+      "step": 27333
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3809009918915318e-07,
+      "loss": 0.6396,
+      "step": 27334
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3790982859020584e-07,
+      "loss": 0.6403,
+      "step": 27335
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3772967491881283e-07,
+      "loss": 0.6664,
+      "step": 27336
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.375496381771102e-07,
+      "loss": 0.6611,
+      "step": 27337
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.37369718367234e-07,
+      "loss": 0.6218,
+      "step": 27338
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3718991549131477e-07,
+      "loss": 0.616,
+      "step": 27339
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3701022955148525e-07,
+      "loss": 0.6777,
+      "step": 27340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3683066054987703e-07,
+      "loss": 0.6556,
+      "step": 27341
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3665120848861846e-07,
+      "loss": 0.654,
+      "step": 27342
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3647187336983668e-07,
+      "loss": 0.6312,
+      "step": 27343
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3629265519565782e-07,
+      "loss": 0.5841,
+      "step": 27344
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.361135539682079e-07,
+      "loss": 0.6948,
+      "step": 27345
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.359345696896097e-07,
+      "loss": 0.6234,
+      "step": 27346
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3575570236198487e-07,
+      "loss": 0.6146,
+      "step": 27347
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3557695198745391e-07,
+      "loss": 0.5771,
+      "step": 27348
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3539831856813845e-07,
+      "loss": 0.5985,
+      "step": 27349
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.352198021061535e-07,
+      "loss": 0.6075,
+      "step": 27350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.350414026036162e-07,
+      "loss": 0.6573,
+      "step": 27351
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.348631200626438e-07,
+      "loss": 0.5941,
+      "step": 27352
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.346849544853479e-07,
+      "loss": 0.604,
+      "step": 27353
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.345069058738413e-07,
+      "loss": 0.5983,
+      "step": 27354
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3432897423023562e-07,
+      "loss": 0.6616,
+      "step": 27355
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3415115955664026e-07,
+      "loss": 0.5747,
+      "step": 27356
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3397346185516246e-07,
+      "loss": 0.6284,
+      "step": 27357
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3379588112791053e-07,
+      "loss": 0.6057,
+      "step": 27358
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3361841737698832e-07,
+      "loss": 0.6002,
+      "step": 27359
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3344107060450195e-07,
+      "loss": 0.6241,
+      "step": 27360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3326384081255306e-07,
+      "loss": 0.6599,
+      "step": 27361
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3308672800324218e-07,
+      "loss": 0.594,
+      "step": 27362
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3290973217866987e-07,
+      "loss": 0.6246,
+      "step": 27363
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3273285334093444e-07,
+      "loss": 0.631,
+      "step": 27364
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.325560914921331e-07,
+      "loss": 0.6736,
+      "step": 27365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3237944663436198e-07,
+      "loss": 0.6446,
+      "step": 27366
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3220291876971602e-07,
+      "loss": 0.5744,
+      "step": 27367
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3202650790028692e-07,
+      "loss": 0.6221,
+      "step": 27368
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3185021402816524e-07,
+      "loss": 0.6923,
+      "step": 27369
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.316740371554437e-07,
+      "loss": 0.661,
+      "step": 27370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3149797728420954e-07,
+      "loss": 0.6555,
+      "step": 27371
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3132203441655112e-07,
+      "loss": 0.6419,
+      "step": 27372
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.311462085545534e-07,
+      "loss": 0.6625,
+      "step": 27373
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.309704997003025e-07,
+      "loss": 0.6675,
+      "step": 27374
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.30794907855879e-07,
+      "loss": 0.5569,
+      "step": 27375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.306194330233679e-07,
+      "loss": 0.6538,
+      "step": 27376
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.304440752048475e-07,
+      "loss": 0.6899,
+      "step": 27377
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3026883440239724e-07,
+      "loss": 0.6805,
+      "step": 27378
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3009371061809662e-07,
+      "loss": 0.5782,
+      "step": 27379
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2991870385402062e-07,
+      "loss": 0.6331,
+      "step": 27380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2974381411224313e-07,
+      "loss": 0.6164,
+      "step": 27381
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2956904139483806e-07,
+      "loss": 0.577,
+      "step": 27382
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.293943857038793e-07,
+      "loss": 0.6226,
+      "step": 27383
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.292198470414352e-07,
+      "loss": 0.6171,
+      "step": 27384
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2904542540957853e-07,
+      "loss": 0.5969,
+      "step": 27385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2887112081037434e-07,
+      "loss": 0.5821,
+      "step": 27386
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2869693324588871e-07,
+      "loss": 0.6242,
+      "step": 27387
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2852286271818893e-07,
+      "loss": 0.6292,
+      "step": 27388
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2834890922933885e-07,
+      "loss": 0.654,
+      "step": 27389
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2817507278139906e-07,
+      "loss": 0.6717,
+      "step": 27390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2800135337643348e-07,
+      "loss": 0.636,
+      "step": 27391
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2782775101649824e-07,
+      "loss": 0.6451,
+      "step": 27392
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.276542657036539e-07,
+      "loss": 0.6184,
+      "step": 27393
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2748089743995661e-07,
+      "loss": 0.6174,
+      "step": 27394
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.273076462274625e-07,
+      "loss": 0.6491,
+      "step": 27395
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2713451206822547e-07,
+      "loss": 0.5982,
+      "step": 27396
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.269614949642972e-07,
+      "loss": 0.5853,
+      "step": 27397
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2678859491773055e-07,
+      "loss": 0.6209,
+      "step": 27398
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2661581193057494e-07,
+      "loss": 0.6495,
+      "step": 27399
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.264431460048776e-07,
+      "loss": 0.604,
+      "step": 27400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2627059714268807e-07,
+      "loss": 0.5604,
+      "step": 27401
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2609816534605135e-07,
+      "loss": 0.5694,
+      "step": 27402
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2592585061701024e-07,
+      "loss": 0.6444,
+      "step": 27403
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2575365295760977e-07,
+      "loss": 0.6154,
+      "step": 27404
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2558157236989055e-07,
+      "loss": 0.6483,
+      "step": 27405
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2540960885589203e-07,
+      "loss": 0.5934,
+      "step": 27406
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2523776241765483e-07,
+      "loss": 0.675,
+      "step": 27407
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.250660330572151e-07,
+      "loss": 0.6075,
+      "step": 27408
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2489442077661007e-07,
+      "loss": 0.6159,
+      "step": 27409
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.247229255778737e-07,
+      "loss": 0.6658,
+      "step": 27410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.245515474630399e-07,
+      "loss": 0.6498,
+      "step": 27411
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2438028643413812e-07,
+      "loss": 0.6079,
+      "step": 27412
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2420914249320127e-07,
+      "loss": 0.6277,
+      "step": 27413
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2403811564225988e-07,
+      "loss": 0.5776,
+      "step": 27414
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.238672058833379e-07,
+      "loss": 0.642,
+      "step": 27415
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2369641321846482e-07,
+      "loss": 0.6389,
+      "step": 27416
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2352573764966346e-07,
+      "loss": 0.6467,
+      "step": 27417
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2335517917895778e-07,
+      "loss": 0.5883,
+      "step": 27418
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.231847378083717e-07,
+      "loss": 0.6859,
+      "step": 27419
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2301441353992472e-07,
+      "loss": 0.6549,
+      "step": 27420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.228442063756352e-07,
+      "loss": 0.6006,
+      "step": 27421
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.226741163175238e-07,
+      "loss": 0.6373,
+      "step": 27422
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2250414336760442e-07,
+      "loss": 0.6215,
+      "step": 27423
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2233428752789322e-07,
+      "loss": 0.6529,
+      "step": 27424
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.221645488004053e-07,
+      "loss": 0.5862,
+      "step": 27425
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2199492718715233e-07,
+      "loss": 0.6405,
+      "step": 27426
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2182542269014496e-07,
+      "loss": 0.6722,
+      "step": 27427
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.216560353113927e-07,
+      "loss": 0.6299,
+      "step": 27428
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2148676505290502e-07,
+      "loss": 0.6224,
+      "step": 27429
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.21317611916687e-07,
+      "loss": 0.6487,
+      "step": 27430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2114857590474592e-07,
+      "loss": 0.5789,
+      "step": 27431
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2097965701908576e-07,
+      "loss": 0.6338,
+      "step": 27432
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2081085526170932e-07,
+      "loss": 0.6359,
+      "step": 27433
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2064217063461615e-07,
+      "loss": 0.6406,
+      "step": 27434
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2047360313980794e-07,
+      "loss": 0.6244,
+      "step": 27435
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.20305152779282e-07,
+      "loss": 0.6436,
+      "step": 27436
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2013681955503786e-07,
+      "loss": 0.5753,
+      "step": 27437
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1996860346906948e-07,
+      "loss": 0.5927,
+      "step": 27438
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1980050452337078e-07,
+      "loss": 0.6866,
+      "step": 27439
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1963252271993574e-07,
+      "loss": 0.6499,
+      "step": 27440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1946465806075724e-07,
+      "loss": 0.6138,
+      "step": 27441
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1929691054782256e-07,
+      "loss": 0.6724,
+      "step": 27442
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1912928018312231e-07,
+      "loss": 0.6525,
+      "step": 27443
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1896176696864492e-07,
+      "loss": 0.6677,
+      "step": 27444
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1879437090637436e-07,
+      "loss": 0.5987,
+      "step": 27445
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1862709199829681e-07,
+      "loss": 0.6263,
+      "step": 27446
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1845993024639512e-07,
+      "loss": 0.6326,
+      "step": 27447
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1829288565264996e-07,
+      "loss": 0.5902,
+      "step": 27448
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1812595821904415e-07,
+      "loss": 0.6388,
+      "step": 27449
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1795914794755614e-07,
+      "loss": 0.6471,
+      "step": 27450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1779245484016321e-07,
+      "loss": 0.6141,
+      "step": 27451
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1762587889884048e-07,
+      "loss": 0.6608,
+      "step": 27452
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1745942012556633e-07,
+      "loss": 0.6909,
+      "step": 27453
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1729307852231032e-07,
+      "loss": 0.6531,
+      "step": 27454
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1712685409104641e-07,
+      "loss": 0.6078,
+      "step": 27455
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1696074683374748e-07,
+      "loss": 0.6757,
+      "step": 27456
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1679475675237972e-07,
+      "loss": 0.6374,
+      "step": 27457
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1662888384891269e-07,
+      "loss": 0.6258,
+      "step": 27458
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1646312812531257e-07,
+      "loss": 0.6221,
+      "step": 27459
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1629748958354448e-07,
+      "loss": 0.6418,
+      "step": 27460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1613196822557238e-07,
+      "loss": 0.64,
+      "step": 27461
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1596656405336026e-07,
+      "loss": 0.607,
+      "step": 27462
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1580127706886767e-07,
+      "loss": 0.6184,
+      "step": 27463
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1563610727405417e-07,
+      "loss": 0.6843,
+      "step": 27464
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1547105467087815e-07,
+      "loss": 0.5965,
+      "step": 27465
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1530611926129698e-07,
+      "loss": 0.5925,
+      "step": 27466
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1514130104726573e-07,
+      "loss": 0.633,
+      "step": 27467
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1497660003073952e-07,
+      "loss": 0.6137,
+      "step": 27468
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.14812016213669e-07,
+      "loss": 0.6052,
+      "step": 27469
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1464754959800817e-07,
+      "loss": 0.6549,
+      "step": 27470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1448320018570547e-07,
+      "loss": 0.6085,
+      "step": 27471
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1431896797870823e-07,
+      "loss": 0.58,
+      "step": 27472
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1415485297896711e-07,
+      "loss": 0.6481,
+      "step": 27473
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1399085518842502e-07,
+      "loss": 0.6199,
+      "step": 27474
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1382697460902703e-07,
+      "loss": 0.6577,
+      "step": 27475
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1366321124271717e-07,
+      "loss": 0.6114,
+      "step": 27476
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1349956509143501e-07,
+      "loss": 0.6347,
+      "step": 27477
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.133360361571223e-07,
+      "loss": 0.6431,
+      "step": 27478
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1317262444171751e-07,
+      "loss": 0.6424,
+      "step": 27479
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1300932994715908e-07,
+      "loss": 0.6216,
+      "step": 27480
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1284615267538102e-07,
+      "loss": 0.583,
+      "step": 27481
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1268309262832067e-07,
+      "loss": 0.6307,
+      "step": 27482
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.125201498079076e-07,
+      "loss": 0.6861,
+      "step": 27483
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1235732421607693e-07,
+      "loss": 0.6337,
+      "step": 27484
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1219461585475822e-07,
+      "loss": 0.6282,
+      "step": 27485
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1203202472588105e-07,
+      "loss": 0.6172,
+      "step": 27486
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1186955083137163e-07,
+      "loss": 0.6509,
+      "step": 27487
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1170719417315845e-07,
+      "loss": 0.6664,
+      "step": 27488
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1154495475316329e-07,
+      "loss": 0.6402,
+      "step": 27489
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.113828325733124e-07,
+      "loss": 0.6236,
+      "step": 27490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1122082763552754e-07,
+      "loss": 0.6631,
+      "step": 27491
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1105893994172945e-07,
+      "loss": 0.6075,
+      "step": 27492
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1089716949383655e-07,
+      "loss": 0.6035,
+      "step": 27493
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1073551629376844e-07,
+      "loss": 0.6421,
+      "step": 27494
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1057398034344024e-07,
+      "loss": 0.6694,
+      "step": 27495
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.104125616447671e-07,
+      "loss": 0.6461,
+      "step": 27496
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1025126019966414e-07,
+      "loss": 0.6592,
+      "step": 27497
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1009007601004318e-07,
+      "loss": 0.5916,
+      "step": 27498
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0992900907781379e-07,
+      "loss": 0.6134,
+      "step": 27499
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0976805940488888e-07,
+      "loss": 0.685,
+      "step": 27500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0960722699317362e-07,
+      "loss": 0.6375,
+      "step": 27501
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0944651184457645e-07,
+      "loss": 0.6422,
+      "step": 27502
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0928591396100252e-07,
+      "loss": 0.6527,
+      "step": 27503
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0912543334435589e-07,
+      "loss": 0.5891,
+      "step": 27504
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0896506999653945e-07,
+      "loss": 0.6963,
+      "step": 27505
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0880482391945501e-07,
+      "loss": 0.6222,
+      "step": 27506
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0864469511500109e-07,
+      "loss": 0.6603,
+      "step": 27507
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0848468358507613e-07,
+      "loss": 0.6343,
+      "step": 27508
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0832478933157865e-07,
+      "loss": 0.5697,
+      "step": 27509
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0816501235640376e-07,
+      "loss": 0.6164,
+      "step": 27510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0800535266144663e-07,
+      "loss": 0.6759,
+      "step": 27511
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0784581024859909e-07,
+      "loss": 0.6823,
+      "step": 27512
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0768638511975293e-07,
+      "loss": 0.6146,
+      "step": 27513
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0752707727679779e-07,
+      "loss": 0.5619,
+      "step": 27514
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0736788672162434e-07,
+      "loss": 0.6709,
+      "step": 27515
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0720881345611778e-07,
+      "loss": 0.6381,
+      "step": 27516
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0704985748216656e-07,
+      "loss": 0.6076,
+      "step": 27517
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0689101880165365e-07,
+      "loss": 0.5543,
+      "step": 27518
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0673229741646196e-07,
+      "loss": 0.6856,
+      "step": 27519
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0657369332847334e-07,
+      "loss": 0.566,
+      "step": 27520
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.064152065395696e-07,
+      "loss": 0.7188,
+      "step": 27521
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0625683705162925e-07,
+      "loss": 0.5428,
+      "step": 27522
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0609858486652969e-07,
+      "loss": 0.6742,
+      "step": 27523
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0594044998614716e-07,
+      "loss": 0.6181,
+      "step": 27524
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0578243241235686e-07,
+      "loss": 0.6794,
+      "step": 27525
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0562453214703172e-07,
+      "loss": 0.6508,
+      "step": 27526
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0546674919204359e-07,
+      "loss": 0.623,
+      "step": 27527
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0530908354926428e-07,
+      "loss": 0.6509,
+      "step": 27528
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0515153522056343e-07,
+      "loss": 0.6695,
+      "step": 27529
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0499410420780732e-07,
+      "loss": 0.6277,
+      "step": 27530
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0483679051286333e-07,
+      "loss": 0.6483,
+      "step": 27531
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0467959413759777e-07,
+      "loss": 0.6523,
+      "step": 27532
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0452251508387134e-07,
+      "loss": 0.6293,
+      "step": 27533
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0436555335355036e-07,
+      "loss": 0.6271,
+      "step": 27534
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0420870894849334e-07,
+      "loss": 0.6332,
+      "step": 27535
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0405198187055877e-07,
+      "loss": 0.6012,
+      "step": 27536
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0389537212160739e-07,
+      "loss": 0.6757,
+      "step": 27537
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.037388797034955e-07,
+      "loss": 0.6706,
+      "step": 27538
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0358250461807717e-07,
+      "loss": 0.6177,
+      "step": 27539
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0342624686720759e-07,
+      "loss": 0.6072,
+      "step": 27540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0327010645273972e-07,
+      "loss": 0.6566,
+      "step": 27541
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0311408337652318e-07,
+      "loss": 0.6214,
+      "step": 27542
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0295817764040983e-07,
+      "loss": 0.7029,
+      "step": 27543
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0280238924624597e-07,
+      "loss": 0.6777,
+      "step": 27544
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0264671819588013e-07,
+      "loss": 0.6872,
+      "step": 27545
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0249116449115859e-07,
+      "loss": 0.6245,
+      "step": 27546
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0233572813392434e-07,
+      "loss": 0.6265,
+      "step": 27547
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0218040912602034e-07,
+      "loss": 0.6811,
+      "step": 27548
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0202520746928734e-07,
+      "loss": 0.6041,
+      "step": 27549
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0187012316556832e-07,
+      "loss": 0.6141,
+      "step": 27550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0171515621669847e-07,
+      "loss": 0.6282,
+      "step": 27551
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0156030662451855e-07,
+      "loss": 0.6066,
+      "step": 27552
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0140557439086263e-07,
+      "loss": 0.6526,
+      "step": 27553
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0125095951756481e-07,
+      "loss": 0.6038,
+      "step": 27554
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0109646200645806e-07,
+      "loss": 0.6267,
+      "step": 27555
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0094208185937538e-07,
+      "loss": 0.6647,
+      "step": 27556
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0078781907814749e-07,
+      "loss": 0.5978,
+      "step": 27557
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0063367366460185e-07,
+      "loss": 0.6751,
+      "step": 27558
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.00479645620567e-07,
+      "loss": 0.6393,
+      "step": 27559
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0032573494786924e-07,
+      "loss": 0.6052,
+      "step": 27560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0017194164833267e-07,
+      "loss": 0.6141,
+      "step": 27561
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0001826572378137e-07,
+      "loss": 0.6364,
+      "step": 27562
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.986470717603614e-08,
+      "loss": 0.67,
+      "step": 27563
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.971126600691994e-08,
+      "loss": 0.6098,
+      "step": 27564
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.955794221825021e-08,
+      "loss": 0.6425,
+      "step": 27565
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.94047358118444e-08,
+      "loss": 0.6395,
+      "step": 27566
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.925164678951993e-08,
+      "loss": 0.5767,
+      "step": 27567
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.909867515309201e-08,
+      "loss": 0.6115,
+      "step": 27568
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.894582090437476e-08,
+      "loss": 0.6314,
+      "step": 27569
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.879308404517895e-08,
+      "loss": 0.6374,
+      "step": 27570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.864046457731646e-08,
+      "loss": 0.6751,
+      "step": 27571
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.848796250259585e-08,
+      "loss": 0.6296,
+      "step": 27572
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.833557782282677e-08,
+      "loss": 0.6064,
+      "step": 27573
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.818331053981444e-08,
+      "loss": 0.6054,
+      "step": 27574
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.803116065536522e-08,
+      "loss": 0.622,
+      "step": 27575
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.787912817128209e-08,
+      "loss": 0.6282,
+      "step": 27576
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.772721308936917e-08,
+      "loss": 0.6444,
+      "step": 27577
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.757541541142612e-08,
+      "loss": 0.6282,
+      "step": 27578
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.742373513925263e-08,
+      "loss": 0.6222,
+      "step": 27579
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.727217227464836e-08,
+      "loss": 0.6567,
+      "step": 27580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.712072681940854e-08,
+      "loss": 0.6602,
+      "step": 27581
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.696939877533063e-08,
+      "loss": 0.6543,
+      "step": 27582
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.681818814420763e-08,
+      "loss": 0.6453,
+      "step": 27583
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.666709492783256e-08,
+      "loss": 0.6829,
+      "step": 27584
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.65161191279973e-08,
+      "loss": 0.6063,
+      "step": 27585
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.636526074649045e-08,
+      "loss": 0.6979,
+      "step": 27586
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.621451978510165e-08,
+      "loss": 0.6438,
+      "step": 27587
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.606389624561952e-08,
+      "loss": 0.6621,
+      "step": 27588
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.591339012982703e-08,
+      "loss": 0.6441,
+      "step": 27589
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.576300143951056e-08,
+      "loss": 0.6975,
+      "step": 27590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.561273017645201e-08,
+      "loss": 0.578,
+      "step": 27591
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.546257634243439e-08,
+      "loss": 0.6463,
+      "step": 27592
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.531253993923739e-08,
+      "loss": 0.67,
+      "step": 27593
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.51626209686396e-08,
+      "loss": 0.6662,
+      "step": 27594
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.501281943241847e-08,
+      "loss": 0.6501,
+      "step": 27595
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.486313533234925e-08,
+      "loss": 0.6162,
+      "step": 27596
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.471356867020942e-08,
+      "loss": 0.6011,
+      "step": 27597
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.456411944776866e-08,
+      "loss": 0.6438,
+      "step": 27598
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.441478766680223e-08,
+      "loss": 0.6272,
+      "step": 27599
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.426557332907871e-08,
+      "loss": 0.6357,
+      "step": 27600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.411647643636779e-08,
+      "loss": 0.6237,
+      "step": 27601
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.396749699043583e-08,
+      "loss": 0.5898,
+      "step": 27602
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.381863499305143e-08,
+      "loss": 0.6243,
+      "step": 27603
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.366989044597874e-08,
+      "loss": 0.6118,
+      "step": 27604
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.352126335097966e-08,
+      "loss": 0.5948,
+      "step": 27605
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.337275370981946e-08,
+      "loss": 0.5545,
+      "step": 27606
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.322436152425674e-08,
+      "loss": 0.6371,
+      "step": 27607
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.307608679605119e-08,
+      "loss": 0.6376,
+      "step": 27608
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.292792952696028e-08,
+      "loss": 0.6559,
+      "step": 27609
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.277988971874263e-08,
+      "loss": 0.6231,
+      "step": 27610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.263196737315128e-08,
+      "loss": 0.6022,
+      "step": 27611
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.248416249194148e-08,
+      "loss": 0.6135,
+      "step": 27612
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.233647507686405e-08,
+      "loss": 0.5879,
+      "step": 27613
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.218890512967094e-08,
+      "loss": 0.6402,
+      "step": 27614
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.204145265211184e-08,
+      "loss": 0.6571,
+      "step": 27615
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.189411764593536e-08,
+      "loss": 0.6137,
+      "step": 27616
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.17469001128879e-08,
+      "loss": 0.6014,
+      "step": 27617
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.15998000547158e-08,
+      "loss": 0.6027,
+      "step": 27618
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.145281747316104e-08,
+      "loss": 0.6362,
+      "step": 27619
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.130595236996776e-08,
+      "loss": 0.6243,
+      "step": 27620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.115920474687678e-08,
+      "loss": 0.6755,
+      "step": 27621
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.101257460562895e-08,
+      "loss": 0.6403,
+      "step": 27622
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.086606194796176e-08,
+      "loss": 0.6367,
+      "step": 27623
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.071966677561273e-08,
+      "loss": 0.6083,
+      "step": 27624
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.0573389090316e-08,
+      "loss": 0.6934,
+      "step": 27625
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.042722889380795e-08,
+      "loss": 0.639,
+      "step": 27626
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.028118618782167e-08,
+      "loss": 0.5551,
+      "step": 27627
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.013526097408687e-08,
+      "loss": 0.6433,
+      "step": 27628
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.998945325433439e-08,
+      "loss": 0.6025,
+      "step": 27629
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.984376303029397e-08,
+      "loss": 0.6284,
+      "step": 27630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.969819030369198e-08,
+      "loss": 0.5856,
+      "step": 27631
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.955273507625484e-08,
+      "loss": 0.6303,
+      "step": 27632
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.940739734970561e-08,
+      "loss": 0.6419,
+      "step": 27633
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.926217712576957e-08,
+      "loss": 0.6055,
+      "step": 27634
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.911707440616756e-08,
+      "loss": 0.6779,
+      "step": 27635
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.897208919261934e-08,
+      "loss": 0.6186,
+      "step": 27636
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.882722148684575e-08,
+      "loss": 0.5826,
+      "step": 27637
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.868247129056207e-08,
+      "loss": 0.6795,
+      "step": 27638
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.853783860548581e-08,
+      "loss": 0.6276,
+      "step": 27639
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.839332343333117e-08,
+      "loss": 0.6055,
+      "step": 27640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.824892577581234e-08,
+      "loss": 0.6729,
+      "step": 27641
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.810464563464128e-08,
+      "loss": 0.6687,
+      "step": 27642
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.796048301152771e-08,
+      "loss": 0.6291,
+      "step": 27643
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.78164379081814e-08,
+      "loss": 0.5719,
+      "step": 27644
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.767251032630985e-08,
+      "loss": 0.6111,
+      "step": 27645
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.752870026762062e-08,
+      "loss": 0.6218,
+      "step": 27646
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.738500773381786e-08,
+      "loss": 0.6449,
+      "step": 27647
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.724143272660579e-08,
+      "loss": 0.5905,
+      "step": 27648
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.709797524768526e-08,
+      "loss": 0.6221,
+      "step": 27649
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.695463529875936e-08,
+      "loss": 0.5964,
+      "step": 27650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.681141288152451e-08,
+      "loss": 0.6294,
+      "step": 27651
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.666830799768155e-08,
+      "loss": 0.6132,
+      "step": 27652
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.652532064892693e-08,
+      "loss": 0.6809,
+      "step": 27653
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.638245083695485e-08,
+      "loss": 0.6442,
+      "step": 27654
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.623969856345948e-08,
+      "loss": 0.625,
+      "step": 27655
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.609706383013395e-08,
+      "loss": 0.6053,
+      "step": 27656
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.595454663866686e-08,
+      "loss": 0.6028,
+      "step": 27657
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.581214699075136e-08,
+      "loss": 0.6602,
+      "step": 27658
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.566986488807494e-08,
+      "loss": 0.6227,
+      "step": 27659
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.552770033232294e-08,
+      "loss": 0.6028,
+      "step": 27660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.538565332518289e-08,
+      "loss": 0.6322,
+      "step": 27661
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.52437238683379e-08,
+      "loss": 0.6119,
+      "step": 27662
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.510191196346995e-08,
+      "loss": 0.6192,
+      "step": 27663
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.496021761226103e-08,
+      "loss": 0.6326,
+      "step": 27664
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.481864081639312e-08,
+      "loss": 0.6016,
+      "step": 27665
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.467718157754157e-08,
+      "loss": 0.6499,
+      "step": 27666
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.453583989738501e-08,
+      "loss": 0.6105,
+      "step": 27667
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.439461577759989e-08,
+      "loss": 0.6364,
+      "step": 27668
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.42535092198582e-08,
+      "loss": 0.64,
+      "step": 27669
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.411252022583639e-08,
+      "loss": 0.6286,
+      "step": 27670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.397164879720421e-08,
+      "loss": 0.6371,
+      "step": 27671
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.383089493563146e-08,
+      "loss": 0.6102,
+      "step": 27672
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.36902586427868e-08,
+      "loss": 0.6001,
+      "step": 27673
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.354973992033999e-08,
+      "loss": 0.6725,
+      "step": 27674
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.340933876995416e-08,
+      "loss": 0.6085,
+      "step": 27675
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.326905519329575e-08,
+      "loss": 0.6408,
+      "step": 27676
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.312888919202677e-08,
+      "loss": 0.6356,
+      "step": 27677
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.298884076781033e-08,
+      "loss": 0.6165,
+      "step": 27678
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.284890992230621e-08,
+      "loss": 0.6519,
+      "step": 27679
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.27090966571742e-08,
+      "loss": 0.6105,
+      "step": 27680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.256940097406962e-08,
+      "loss": 0.6436,
+      "step": 27681
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.242982287465118e-08,
+      "loss": 0.607,
+      "step": 27682
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.229036236057308e-08,
+      "loss": 0.6014,
+      "step": 27683
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.215101943348847e-08,
+      "loss": 0.652,
+      "step": 27684
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.201179409505045e-08,
+      "loss": 0.6247,
+      "step": 27685
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.187268634690881e-08,
+      "loss": 0.6242,
+      "step": 27686
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.173369619071114e-08,
+      "loss": 0.5815,
+      "step": 27687
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.159482362810834e-08,
+      "loss": 0.6416,
+      "step": 27688
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.145606866074685e-08,
+      "loss": 0.6042,
+      "step": 27689
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.131743129026981e-08,
+      "loss": 0.6412,
+      "step": 27690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.117891151832147e-08,
+      "loss": 0.6514,
+      "step": 27691
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.104050934654495e-08,
+      "loss": 0.6766,
+      "step": 27692
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.090222477658005e-08,
+      "loss": 0.622,
+      "step": 27693
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.076405781006769e-08,
+      "loss": 0.6005,
+      "step": 27694
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.062600844864543e-08,
+      "loss": 0.5978,
+      "step": 27695
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.048807669394975e-08,
+      "loss": 0.6614,
+      "step": 27696
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.035026254761602e-08,
+      "loss": 0.6136,
+      "step": 27697
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.021256601127959e-08,
+      "loss": 0.6329,
+      "step": 27698
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.007498708657025e-08,
+      "loss": 0.6313,
+      "step": 27699
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.993752577512004e-08,
+      "loss": 0.6616,
+      "step": 27700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.9800182078561e-08,
+      "loss": 0.6608,
+      "step": 27701
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.966295599851959e-08,
+      "loss": 0.6703,
+      "step": 27702
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.95258475366234e-08,
+      "loss": 0.666,
+      "step": 27703
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.93888566944978e-08,
+      "loss": 0.589,
+      "step": 27704
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.925198347376484e-08,
+      "loss": 0.6703,
+      "step": 27705
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.911522787605097e-08,
+      "loss": 0.6399,
+      "step": 27706
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.897858990297713e-08,
+      "loss": 0.6285,
+      "step": 27707
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.884206955616092e-08,
+      "loss": 0.6306,
+      "step": 27708
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.870566683722214e-08,
+      "loss": 0.623,
+      "step": 27709
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.85693817477795e-08,
+      "loss": 0.6556,
+      "step": 27710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.843321428944617e-08,
+      "loss": 0.6339,
+      "step": 27711
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.829716446383862e-08,
+      "loss": 0.6729,
+      "step": 27712
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.816123227256889e-08,
+      "loss": 0.5872,
+      "step": 27713
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.802541771724903e-08,
+      "loss": 0.5997,
+      "step": 27714
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.788972079948998e-08,
+      "loss": 0.5869,
+      "step": 27715
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.775414152089933e-08,
+      "loss": 0.6283,
+      "step": 27716
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.761867988308358e-08,
+      "loss": 0.6275,
+      "step": 27717
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.748333588765256e-08,
+      "loss": 0.6191,
+      "step": 27718
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.734810953620719e-08,
+      "loss": 0.6369,
+      "step": 27719
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.721300083035287e-08,
+      "loss": 0.6188,
+      "step": 27720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.707800977169055e-08,
+      "loss": 0.5962,
+      "step": 27721
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.694313636182115e-08,
+      "loss": 0.6763,
+      "step": 27722
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.68083806023423e-08,
+      "loss": 0.5901,
+      "step": 27723
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.667374249485493e-08,
+      "loss": 0.6276,
+      "step": 27724
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.653922204095221e-08,
+      "loss": 0.6472,
+      "step": 27725
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.640481924223064e-08,
+      "loss": 0.6295,
+      "step": 27726
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.62705341002834e-08,
+      "loss": 0.6553,
+      "step": 27727
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.613636661670254e-08,
+      "loss": 0.6314,
+      "step": 27728
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.60023167930779e-08,
+      "loss": 0.6229,
+      "step": 27729
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.586838463100154e-08,
+      "loss": 0.6244,
+      "step": 27730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.573457013205887e-08,
+      "loss": 0.6701,
+      "step": 27731
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.560087329783639e-08,
+      "loss": 0.6029,
+      "step": 27732
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.54672941299206e-08,
+      "loss": 0.6341,
+      "step": 27733
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.533383262989469e-08,
+      "loss": 0.6428,
+      "step": 27734
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.520048879934073e-08,
+      "loss": 0.6069,
+      "step": 27735
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.506726263983965e-08,
+      "loss": 0.6354,
+      "step": 27736
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.493415415297245e-08,
+      "loss": 0.6342,
+      "step": 27737
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.480116334031451e-08,
+      "loss": 0.6289,
+      "step": 27738
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.466829020344568e-08,
+      "loss": 0.6341,
+      "step": 27739
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.453553474393915e-08,
+      "loss": 0.6326,
+      "step": 27740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.440289696336922e-08,
+      "loss": 0.6709,
+      "step": 27741
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.427037686330907e-08,
+      "loss": 0.6312,
+      "step": 27742
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.413797444533077e-08,
+      "loss": 0.6623,
+      "step": 27743
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.400568971100198e-08,
+      "loss": 0.5829,
+      "step": 27744
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.387352266189251e-08,
+      "loss": 0.6524,
+      "step": 27745
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.374147329956893e-08,
+      "loss": 0.5741,
+      "step": 27746
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.360954162559774e-08,
+      "loss": 0.664,
+      "step": 27747
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.347772764154104e-08,
+      "loss": 0.6565,
+      "step": 27748
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.334603134896312e-08,
+      "loss": 0.5974,
+      "step": 27749
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.321445274942607e-08,
+      "loss": 0.5768,
+      "step": 27750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.308299184448974e-08,
+      "loss": 0.6469,
+      "step": 27751
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.295164863571069e-08,
+      "loss": 0.6068,
+      "step": 27752
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.282042312464876e-08,
+      "loss": 0.624,
+      "step": 27753
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.268931531285828e-08,
+      "loss": 0.5857,
+      "step": 27754
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.255832520189354e-08,
+      "loss": 0.6361,
+      "step": 27755
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.242745279330777e-08,
+      "loss": 0.6607,
+      "step": 27756
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.229669808865414e-08,
+      "loss": 0.6263,
+      "step": 27757
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.216606108948143e-08,
+      "loss": 0.6764,
+      "step": 27758
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.203554179733951e-08,
+      "loss": 0.6554,
+      "step": 27759
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.190514021377382e-08,
+      "loss": 0.5833,
+      "step": 27760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.177485634033199e-08,
+      "loss": 0.6634,
+      "step": 27761
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.164469017855946e-08,
+      "loss": 0.6152,
+      "step": 27762
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.151464172999834e-08,
+      "loss": 0.6073,
+      "step": 27763
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.13847109961907e-08,
+      "loss": 0.6942,
+      "step": 27764
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.125489797867646e-08,
+      "loss": 0.6175,
+      "step": 27765
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.112520267899547e-08,
+      "loss": 0.6224,
+      "step": 27766
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.099562509868541e-08,
+      "loss": 0.6187,
+      "step": 27767
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.086616523928281e-08,
+      "loss": 0.6042,
+      "step": 27768
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.073682310232199e-08,
+      "loss": 0.5683,
+      "step": 27769
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.06075986893373e-08,
+      "loss": 0.6399,
+      "step": 27770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.047849200185974e-08,
+      "loss": 0.6067,
+      "step": 27771
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.03495030414203e-08,
+      "loss": 0.6455,
+      "step": 27772
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.022063180954886e-08,
+      "loss": 0.6301,
+      "step": 27773
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.009187830777308e-08,
+      "loss": 0.6163,
+      "step": 27774
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.996324253761955e-08,
+      "loss": 0.624,
+      "step": 27775
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.983472450061257e-08,
+      "loss": 0.656,
+      "step": 27776
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.970632419827761e-08,
+      "loss": 0.6433,
+      "step": 27777
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.957804163213567e-08,
+      "loss": 0.6012,
+      "step": 27778
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.944987680370774e-08,
+      "loss": 0.6604,
+      "step": 27779
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.932182971451485e-08,
+      "loss": 0.5883,
+      "step": 27780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.919390036607354e-08,
+      "loss": 0.6369,
+      "step": 27781
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.90660887599004e-08,
+      "loss": 0.6559,
+      "step": 27782
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.893839489751197e-08,
+      "loss": 0.6328,
+      "step": 27783
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.88108187804215e-08,
+      "loss": 0.6194,
+      "step": 27784
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.868336041014224e-08,
+      "loss": 0.6739,
+      "step": 27785
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.855601978818516e-08,
+      "loss": 0.6719,
+      "step": 27786
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.842879691605908e-08,
+      "loss": 0.6187,
+      "step": 27787
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.83016917952739e-08,
+      "loss": 0.6645,
+      "step": 27788
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.81747044273351e-08,
+      "loss": 0.6075,
+      "step": 27789
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.804783481374922e-08,
+      "loss": 0.6154,
+      "step": 27790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.792108295601951e-08,
+      "loss": 0.6569,
+      "step": 27791
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.779444885565146e-08,
+      "loss": 0.6866,
+      "step": 27792
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.766793251414272e-08,
+      "loss": 0.6301,
+      "step": 27793
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.754153393299545e-08,
+      "loss": 0.6299,
+      "step": 27794
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.741525311370845e-08,
+      "loss": 0.6557,
+      "step": 27795
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.728909005777828e-08,
+      "loss": 0.6161,
+      "step": 27796
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.716304476670044e-08,
+      "loss": 0.577,
+      "step": 27797
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.703711724197037e-08,
+      "loss": 0.5468,
+      "step": 27798
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.691130748508023e-08,
+      "loss": 0.6421,
+      "step": 27799
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.678561549752216e-08,
+      "loss": 0.5799,
+      "step": 27800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.666004128078607e-08,
+      "loss": 0.6488,
+      "step": 27801
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.653458483636078e-08,
+      "loss": 0.6466,
+      "step": 27802
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.640924616573396e-08,
+      "loss": 0.6426,
+      "step": 27803
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.628402527039224e-08,
+      "loss": 0.6157,
+      "step": 27804
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.615892215181885e-08,
+      "loss": 0.6487,
+      "step": 27805
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.603393681149706e-08,
+      "loss": 0.7066,
+      "step": 27806
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.590906925091123e-08,
+      "loss": 0.6318,
+      "step": 27807
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.578431947153907e-08,
+      "loss": 0.6091,
+      "step": 27808
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.565968747486051e-08,
+      "loss": 0.6729,
+      "step": 27809
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.553517326235436e-08,
+      "loss": 0.6339,
+      "step": 27810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.541077683549502e-08,
+      "loss": 0.6493,
+      "step": 27811
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.528649819575794e-08,
+      "loss": 0.6002,
+      "step": 27812
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.516233734461751e-08,
+      "loss": 0.6804,
+      "step": 27813
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.50382942835448e-08,
+      "loss": 0.651,
+      "step": 27814
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.491436901401082e-08,
+      "loss": 0.6848,
+      "step": 27815
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.479056153748665e-08,
+      "loss": 0.6349,
+      "step": 27816
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.466687185543663e-08,
+      "loss": 0.6171,
+      "step": 27817
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.454329996932963e-08,
+      "loss": 0.6191,
+      "step": 27818
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.441984588062999e-08,
+      "loss": 0.6452,
+      "step": 27819
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.429650959080213e-08,
+      "loss": 0.6121,
+      "step": 27820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.41732911013071e-08,
+      "loss": 0.6119,
+      "step": 27821
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.405019041360816e-08,
+      "loss": 0.6616,
+      "step": 27822
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.392720752916193e-08,
+      "loss": 0.682,
+      "step": 27823
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.380434244942724e-08,
+      "loss": 0.6596,
+      "step": 27824
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.368159517586293e-08,
+      "loss": 0.6342,
+      "step": 27825
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.355896570992225e-08,
+      "loss": 0.646,
+      "step": 27826
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.343645405305964e-08,
+      "loss": 0.6303,
+      "step": 27827
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.331406020672726e-08,
+      "loss": 0.694,
+      "step": 27828
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.319178417237725e-08,
+      "loss": 0.6623,
+      "step": 27829
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.306962595145849e-08,
+      "loss": 0.5928,
+      "step": 27830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.29475855454198e-08,
+      "loss": 0.6202,
+      "step": 27831
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.28256629557078e-08,
+      "loss": 0.6772,
+      "step": 27832
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.270385818376801e-08,
+      "loss": 0.5809,
+      "step": 27833
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.258217123104593e-08,
+      "loss": 0.6494,
+      "step": 27834
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.246060209898152e-08,
+      "loss": 0.6418,
+      "step": 27835
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.233915078901809e-08,
+      "loss": 0.6065,
+      "step": 27836
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.221781730259558e-08,
+      "loss": 0.6626,
+      "step": 27837
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.209660164115171e-08,
+      "loss": 0.6148,
+      "step": 27838
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.197550380612538e-08,
+      "loss": 0.6152,
+      "step": 27839
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.185452379894985e-08,
+      "loss": 0.6555,
+      "step": 27840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.173366162106175e-08,
+      "loss": 0.6241,
+      "step": 27841
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.161291727389218e-08,
+      "loss": 0.6548,
+      "step": 27842
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.149229075887331e-08,
+      "loss": 0.6023,
+      "step": 27843
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.137178207743622e-08,
+      "loss": 0.5622,
+      "step": 27844
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.125139123100976e-08,
+      "loss": 0.5847,
+      "step": 27845
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.113111822101947e-08,
+      "loss": 0.6975,
+      "step": 27846
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.101096304889309e-08,
+      "loss": 0.6282,
+      "step": 27847
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.089092571605393e-08,
+      "loss": 0.5718,
+      "step": 27848
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.07710062239264e-08,
+      "loss": 0.6025,
+      "step": 27849
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.065120457393158e-08,
+      "loss": 0.7128,
+      "step": 27850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.053152076749169e-08,
+      "loss": 0.6363,
+      "step": 27851
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.041195480602225e-08,
+      "loss": 0.6046,
+      "step": 27852
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.029250669094322e-08,
+      "loss": 0.5973,
+      "step": 27853
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.017317642367015e-08,
+      "loss": 0.6702,
+      "step": 27854
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.005396400561859e-08,
+      "loss": 0.613,
+      "step": 27855
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.99348694382007e-08,
+      "loss": 0.6198,
+      "step": 27856
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.981589272282984e-08,
+      "loss": 0.6126,
+      "step": 27857
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.969703386091486e-08,
+      "loss": 0.6202,
+      "step": 27858
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.957829285386685e-08,
+      "loss": 0.5952,
+      "step": 27859
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.945966970309358e-08,
+      "loss": 0.6021,
+      "step": 27860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.934116440999948e-08,
+      "loss": 0.6287,
+      "step": 27861
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9222776975991214e-08,
+      "loss": 0.6727,
+      "step": 27862
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.91045074024732e-08,
+      "loss": 0.637,
+      "step": 27863
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.898635569084543e-08,
+      "loss": 0.5125,
+      "step": 27864
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8868321842510123e-08,
+      "loss": 0.6381,
+      "step": 27865
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8750405858866156e-08,
+      "loss": 0.6512,
+      "step": 27866
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8632607741312406e-08,
+      "loss": 0.6651,
+      "step": 27867
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.851492749124443e-08,
+      "loss": 0.6756,
+      "step": 27868
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.839736511005889e-08,
+      "loss": 0.6477,
+      "step": 27869
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8279920599148e-08,
+      "loss": 0.6073,
+      "step": 27870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.816259395990509e-08,
+      "loss": 0.5935,
+      "step": 27871
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8045385193721274e-08,
+      "loss": 0.6186,
+      "step": 27872
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.792829430198543e-08,
+      "loss": 0.5859,
+      "step": 27873
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7811321286087575e-08,
+      "loss": 0.5552,
+      "step": 27874
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.769446614741325e-08,
+      "loss": 0.5935,
+      "step": 27875
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.757772888734803e-08,
+      "loss": 0.6643,
+      "step": 27876
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.746110950727524e-08,
+      "loss": 0.6505,
+      "step": 27877
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.734460800857933e-08,
+      "loss": 0.6557,
+      "step": 27878
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.72282243926392e-08,
+      "loss": 0.6055,
+      "step": 27879
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7111958660837075e-08,
+      "loss": 0.5953,
+      "step": 27880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6995810814551856e-08,
+      "loss": 0.6402,
+      "step": 27881
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6879780855158e-08,
+      "loss": 0.612,
+      "step": 27882
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6763868784032174e-08,
+      "loss": 0.5903,
+      "step": 27883
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.664807460254884e-08,
+      "loss": 0.638,
+      "step": 27884
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.653239831208135e-08,
+      "loss": 0.6077,
+      "step": 27885
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.641683991400082e-08,
+      "loss": 0.6085,
+      "step": 27886
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.630139940967727e-08,
+      "loss": 0.612,
+      "step": 27887
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.618607680047961e-08,
+      "loss": 0.6271,
+      "step": 27888
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.607087208777562e-08,
+      "loss": 0.6411,
+      "step": 27889
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5955785272929774e-08,
+      "loss": 0.6178,
+      "step": 27890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.584081635730654e-08,
+      "loss": 0.6116,
+      "step": 27891
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.572596534227148e-08,
+      "loss": 0.6452,
+      "step": 27892
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.561123222918463e-08,
+      "loss": 0.617,
+      "step": 27893
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5496617019404894e-08,
+      "loss": 0.6126,
+      "step": 27894
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5382119714293416e-08,
+      "loss": 0.623,
+      "step": 27895
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.526774031520687e-08,
+      "loss": 0.6193,
+      "step": 27896
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.515347882350197e-08,
+      "loss": 0.642,
+      "step": 27897
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5039335240532063e-08,
+      "loss": 0.6984,
+      "step": 27898
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.492530956765163e-08,
+      "loss": 0.6223,
+      "step": 27899
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.481140180621291e-08,
+      "loss": 0.6056,
+      "step": 27900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.469761195756485e-08,
+      "loss": 0.6412,
+      "step": 27901
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.458394002305745e-08,
+      "loss": 0.6875,
+      "step": 27902
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4470386004038535e-08,
+      "loss": 0.6032,
+      "step": 27903
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.435694990185481e-08,
+      "loss": 0.6533,
+      "step": 27904
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.424363171785074e-08,
+      "loss": 0.6665,
+      "step": 27905
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.413043145336972e-08,
+      "loss": 0.64,
+      "step": 27906
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.401734910975398e-08,
+      "loss": 0.6549,
+      "step": 27907
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.390438468834358e-08,
+      "loss": 0.6611,
+      "step": 27908
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3791538190479665e-08,
+      "loss": 0.6075,
+      "step": 27909
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.367880961749894e-08,
+      "loss": 0.6104,
+      "step": 27910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3566198970737e-08,
+      "loss": 0.6571,
+      "step": 27911
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.345370625153168e-08,
+      "loss": 0.6592,
+      "step": 27912
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3341331461214116e-08,
+      "loss": 0.6909,
+      "step": 27913
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.32290746011177e-08,
+      "loss": 0.5804,
+      "step": 27914
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.311693567257248e-08,
+      "loss": 0.6735,
+      "step": 27915
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.300491467691071e-08,
+      "loss": 0.624,
+      "step": 27916
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2893011615458013e-08,
+      "loss": 0.6425,
+      "step": 27917
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.278122648954109e-08,
+      "loss": 0.6372,
+      "step": 27918
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.266955930048778e-08,
+      "loss": 0.6081,
+      "step": 27919
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2558010049618134e-08,
+      "loss": 0.617,
+      "step": 27920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.244657873825776e-08,
+      "loss": 0.5647,
+      "step": 27921
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.233526536772671e-08,
+      "loss": 0.6838,
+      "step": 27922
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.222406993934504e-08,
+      "loss": 0.6281,
+      "step": 27923
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.211299245443169e-08,
+      "loss": 0.6079,
+      "step": 27924
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.200203291430228e-08,
+      "loss": 0.5762,
+      "step": 27925
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1891191320273536e-08,
+      "loss": 0.5933,
+      "step": 27926
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.178046767365885e-08,
+      "loss": 0.6546,
+      "step": 27927
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1669861975771616e-08,
+      "loss": 0.6169,
+      "step": 27928
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.15593742279219e-08,
+      "loss": 0.6008,
+      "step": 27929
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.144900443142087e-08,
+      "loss": 0.6376,
+      "step": 27930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.133875258757748e-08,
+      "loss": 0.6372,
+      "step": 27931
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1228618697698463e-08,
+      "loss": 0.5421,
+      "step": 27932
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.111860276308833e-08,
+      "loss": 0.6869,
+      "step": 27933
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.100870478505382e-08,
+      "loss": 0.5973,
+      "step": 27934
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.089892476489611e-08,
+      "loss": 0.6478,
+      "step": 27935
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.078926270391748e-08,
+      "loss": 0.6287,
+      "step": 27936
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0679718603418024e-08,
+      "loss": 0.5933,
+      "step": 27937
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.057029246469669e-08,
+      "loss": 0.6301,
+      "step": 27938
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0460984289049106e-08,
+      "loss": 0.664,
+      "step": 27939
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.035179407777424e-08,
+      "loss": 0.6587,
+      "step": 27940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0242721832164385e-08,
+      "loss": 0.6424,
+      "step": 27941
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.013376755351407e-08,
+      "loss": 0.6562,
+      "step": 27942
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.002493124311558e-08,
+      "loss": 0.6795,
+      "step": 27943
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.991621290225679e-08,
+      "loss": 0.63,
+      "step": 27944
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9807612532227766e-08,
+      "loss": 0.5794,
+      "step": 27945
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9699130134318596e-08,
+      "loss": 0.613,
+      "step": 27946
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.959076570981158e-08,
+      "loss": 0.6235,
+      "step": 27947
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.948251925999459e-08,
+      "loss": 0.6557,
+      "step": 27948
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9374390786149914e-08,
+      "loss": 0.6571,
+      "step": 27949
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.926638028955877e-08,
+      "loss": 0.619,
+      "step": 27950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.915848777150234e-08,
+      "loss": 0.6253,
+      "step": 27951
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9050713233259604e-08,
+      "loss": 0.601,
+      "step": 27952
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.894305667610955e-08,
+      "loss": 0.6529,
+      "step": 27953
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8835518101326693e-08,
+      "loss": 0.6586,
+      "step": 27954
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.872809751018892e-08,
+      "loss": 0.6297,
+      "step": 27955
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8620794903966316e-08,
+      "loss": 0.6246,
+      "step": 27956
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.851361028393231e-08,
+      "loss": 0.6794,
+      "step": 27957
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8406543651360324e-08,
+      "loss": 0.6395,
+      "step": 27958
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8299595007516023e-08,
+      "loss": 0.6243,
+      "step": 27959
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8192764353668374e-08,
+      "loss": 0.6119,
+      "step": 27960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.808605169108638e-08,
+      "loss": 0.5834,
+      "step": 27961
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.797945702103235e-08,
+      "loss": 0.634,
+      "step": 27962
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7872980344771945e-08,
+      "loss": 0.6923,
+      "step": 27963
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.776662166356638e-08,
+      "loss": 0.6475,
+      "step": 27964
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.766038097867687e-08,
+      "loss": 0.5735,
+      "step": 27965
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.755425829136351e-08,
+      "loss": 0.639,
+      "step": 27966
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7448253602885296e-08,
+      "loss": 0.6549,
+      "step": 27967
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7342366914496784e-08,
+      "loss": 0.6488,
+      "step": 27968
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.723659822745474e-08,
+      "loss": 0.6723,
+      "step": 27969
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7130947543013724e-08,
+      "loss": 0.5904,
+      "step": 27970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.702541486242496e-08,
+      "loss": 0.633,
+      "step": 27971
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.692000018694187e-08,
+      "loss": 0.64,
+      "step": 27972
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6814703517812364e-08,
+      "loss": 0.6184,
+      "step": 27973
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.670952485628433e-08,
+      "loss": 0.6437,
+      "step": 27974
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6604464203606756e-08,
+      "loss": 0.5621,
+      "step": 27975
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.649952156102422e-08,
+      "loss": 0.5994,
+      "step": 27976
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.639469692978127e-08,
+      "loss": 0.6075,
+      "step": 27977
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.628999031112136e-08,
+      "loss": 0.6286,
+      "step": 27978
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.618540170628349e-08,
+      "loss": 0.6564,
+      "step": 27979
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.608093111651002e-08,
+      "loss": 0.5915,
+      "step": 27980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.597657854303883e-08,
+      "loss": 0.6829,
+      "step": 27981
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.587234398710783e-08,
+      "loss": 0.6298,
+      "step": 27982
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.576822744995158e-08,
+      "loss": 0.606,
+      "step": 27983
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.566422893280575e-08,
+      "loss": 0.6504,
+      "step": 27984
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5560348436901603e-08,
+      "loss": 0.603,
+      "step": 27985
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.545658596347258e-08,
+      "loss": 0.6325,
+      "step": 27986
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.53529415137488e-08,
+      "loss": 0.6283,
+      "step": 27987
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.524941508895819e-08,
+      "loss": 0.6052,
+      "step": 27988
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.514600669032865e-08,
+      "loss": 0.6607,
+      "step": 27989
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.504271631908586e-08,
+      "loss": 0.5756,
+      "step": 27990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.49395439764555e-08,
+      "loss": 0.6097,
+      "step": 27991
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4836489663658836e-08,
+      "loss": 0.6691,
+      "step": 27992
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4733553381919316e-08,
+      "loss": 0.613,
+      "step": 27993
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.463073513245708e-08,
+      "loss": 0.6267,
+      "step": 27994
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.452803491649227e-08,
+      "loss": 0.6372,
+      "step": 27995
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4425452735240574e-08,
+      "loss": 0.6244,
+      "step": 27996
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.432298858991879e-08,
+      "loss": 0.6316,
+      "step": 27997
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.422064248174151e-08,
+      "loss": 0.6934,
+      "step": 27998
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.411841441192333e-08,
+      "loss": 0.6313,
+      "step": 27999
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.401630438167548e-08,
+      "loss": 0.6593,
+      "step": 28000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3914312392208116e-08,
+      "loss": 0.5834,
+      "step": 28001
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.381243844473249e-08,
+      "loss": 0.6082,
+      "step": 28002
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.37106825404543e-08,
+      "loss": 0.6565,
+      "step": 28003
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3609044680580356e-08,
+      "loss": 0.6597,
+      "step": 28004
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3507524866316376e-08,
+      "loss": 0.6416,
+      "step": 28005
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3406123098864714e-08,
+      "loss": 0.6556,
+      "step": 28006
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.330483937942997e-08,
+      "loss": 0.681,
+      "step": 28007
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.320367370921119e-08,
+      "loss": 0.6184,
+      "step": 28008
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3102626089408515e-08,
+      "loss": 0.6215,
+      "step": 28009
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3001696521219884e-08,
+      "loss": 0.6856,
+      "step": 28010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2900885005840995e-08,
+      "loss": 0.6326,
+      "step": 28011
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2800191544468686e-08,
+      "loss": 0.6804,
+      "step": 28012
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2699616138295316e-08,
+      "loss": 0.6491,
+      "step": 28013
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2599158788515505e-08,
+      "loss": 0.6594,
+      "step": 28014
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2498819496317176e-08,
+      "loss": 0.6196,
+      "step": 28015
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.239859826289272e-08,
+      "loss": 0.6553,
+      "step": 28016
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.229849508942896e-08,
+      "loss": 0.6362,
+      "step": 28017
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.219850997711272e-08,
+      "loss": 0.6584,
+      "step": 28018
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.209864292712973e-08,
+      "loss": 0.6153,
+      "step": 28019
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1998893940665696e-08,
+      "loss": 0.6541,
+      "step": 28020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1899263018899685e-08,
+      "loss": 0.6139,
+      "step": 28021
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.179975016301629e-08,
+      "loss": 0.6004,
+      "step": 28022
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.170035537419348e-08,
+      "loss": 0.6588,
+      "step": 28023
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1601078653610295e-08,
+      "loss": 0.6467,
+      "step": 28024
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.150192000244469e-08,
+      "loss": 0.6285,
+      "step": 28025
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1402879421870154e-08,
+      "loss": 0.6561,
+      "step": 28026
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.130395691306244e-08,
+      "loss": 0.6889,
+      "step": 28027
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1205152477193924e-08,
+      "loss": 0.6133,
+      "step": 28028
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1106466115437006e-08,
+      "loss": 0.6648,
+      "step": 28029
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1007897828960754e-08,
+      "loss": 0.7149,
+      "step": 28030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.090944761893423e-08,
+      "loss": 0.6398,
+      "step": 28031
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.081111548652428e-08,
+      "loss": 0.634,
+      "step": 28032
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0712901432896636e-08,
+      "loss": 0.6297,
+      "step": 28033
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0614805459215925e-08,
+      "loss": 0.5892,
+      "step": 28034
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0516827566646764e-08,
+      "loss": 0.62,
+      "step": 28035
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0418967756348236e-08,
+      "loss": 0.6094,
+      "step": 28036
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.032122602948163e-08,
+      "loss": 0.6546,
+      "step": 28037
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0223602387207125e-08,
+      "loss": 0.6542,
+      "step": 28038
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.012609683068047e-08,
+      "loss": 0.603,
+      "step": 28039
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0028709361057406e-08,
+      "loss": 0.606,
+      "step": 28040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9931439979493667e-08,
+      "loss": 0.6178,
+      "step": 28041
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.983428868714278e-08,
+      "loss": 0.6587,
+      "step": 28042
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.973725548515606e-08,
+      "loss": 0.6033,
+      "step": 28043
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.964034037468367e-08,
+      "loss": 0.644,
+      "step": 28044
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.954354335687582e-08,
+      "loss": 0.6112,
+      "step": 28045
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9446864432878264e-08,
+      "loss": 0.707,
+      "step": 28046
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9350303603838954e-08,
+      "loss": 0.6319,
+      "step": 28047
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.925386087090144e-08,
+      "loss": 0.6231,
+      "step": 28048
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9157536235210346e-08,
+      "loss": 0.6581,
+      "step": 28049
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.906132969790699e-08,
+      "loss": 0.6171,
+      "step": 28050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.896524126013268e-08,
+      "loss": 0.6901,
+      "step": 28051
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.886927092302539e-08,
+      "loss": 0.6054,
+      "step": 28052
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8773418687724215e-08,
+      "loss": 0.6684,
+      "step": 28053
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8677684555364917e-08,
+      "loss": 0.6433,
+      "step": 28054
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.858206852708324e-08,
+      "loss": 0.7306,
+      "step": 28055
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.848657060401273e-08,
+      "loss": 0.6014,
+      "step": 28056
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.83911907872847e-08,
+      "loss": 0.5954,
+      "step": 28057
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.829592907803048e-08,
+      "loss": 0.5734,
+      "step": 28058
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.820078547737915e-08,
+      "loss": 0.6428,
+      "step": 28059
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.810575998645982e-08,
+      "loss": 0.6075,
+      "step": 28060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8010852606399365e-08,
+      "loss": 0.6521,
+      "step": 28061
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.791606333832132e-08,
+      "loss": 0.6249,
+      "step": 28062
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7821392183350344e-08,
+      "loss": 0.6241,
+      "step": 28063
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7726839142609997e-08,
+      "loss": 0.6429,
+      "step": 28064
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.763240421721937e-08,
+      "loss": 0.6431,
+      "step": 28065
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7538087408299784e-08,
+      "loss": 0.6297,
+      "step": 28066
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.744388871696814e-08,
+      "loss": 0.6198,
+      "step": 28067
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7349808144342413e-08,
+      "loss": 0.6143,
+      "step": 28068
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7255845691536174e-08,
+      "loss": 0.6496,
+      "step": 28069
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.71620013596663e-08,
+      "loss": 0.6039,
+      "step": 28070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7068275149843017e-08,
+      "loss": 0.6474,
+      "step": 28071
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.697466706317876e-08,
+      "loss": 0.5835,
+      "step": 28072
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.688117710078376e-08,
+      "loss": 0.6108,
+      "step": 28073
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.678780526376602e-08,
+      "loss": 0.6638,
+      "step": 28074
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6694551553231315e-08,
+      "loss": 0.6495,
+      "step": 28075
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6601415970286546e-08,
+      "loss": 0.6149,
+      "step": 28076
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.650839851603638e-08,
+      "loss": 0.6211,
+      "step": 28077
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.641549919158216e-08,
+      "loss": 0.5779,
+      "step": 28078
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.632271799802634e-08,
+      "loss": 0.6317,
+      "step": 28079
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6230054936469136e-08,
+      "loss": 0.6456,
+      "step": 28080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.613751000800969e-08,
+      "loss": 0.5978,
+      "step": 28081
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.604508321374267e-08,
+      "loss": 0.5957,
+      "step": 28082
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.59527745547672e-08,
+      "loss": 0.6236,
+      "step": 28083
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.586058403217574e-08,
+      "loss": 0.6278,
+      "step": 28084
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5768511647061856e-08,
+      "loss": 0.631,
+      "step": 28085
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.567655740051579e-08,
+      "loss": 0.6197,
+      "step": 28086
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.558472129363e-08,
+      "loss": 0.599,
+      "step": 28087
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.54930033274925e-08,
+      "loss": 0.6759,
+      "step": 28088
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5401403503191324e-08,
+      "loss": 0.693,
+      "step": 28089
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5309921821811146e-08,
+      "loss": 0.6258,
+      "step": 28090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.521855828443665e-08,
+      "loss": 0.6131,
+      "step": 28091
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.512731289215254e-08,
+      "loss": 0.613,
+      "step": 28092
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5036185646039056e-08,
+      "loss": 0.6357,
+      "step": 28093
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.494517654717755e-08,
+      "loss": 0.559,
+      "step": 28094
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4854285596647166e-08,
+      "loss": 0.656,
+      "step": 28095
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.476351279552481e-08,
+      "loss": 0.6059,
+      "step": 28096
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.467285814488741e-08,
+      "loss": 0.6555,
+      "step": 28097
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.458232164580966e-08,
+      "loss": 0.618,
+      "step": 28098
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.449190329936403e-08,
+      "loss": 0.6266,
+      "step": 28099
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4401603106624106e-08,
+      "loss": 0.6288,
+      "step": 28100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4311421068659035e-08,
+      "loss": 0.6979,
+      "step": 28101
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.422135718653907e-08,
+      "loss": 0.6442,
+      "step": 28102
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.413141146133225e-08,
+      "loss": 0.5877,
+      "step": 28103
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.404158389410328e-08,
+      "loss": 0.6353,
+      "step": 28104
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3951874485919076e-08,
+      "loss": 0.6046,
+      "step": 28105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.386228323784213e-08,
+      "loss": 0.6332,
+      "step": 28106
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.377281015093492e-08,
+      "loss": 0.5879,
+      "step": 28107
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.368345522625993e-08,
+      "loss": 0.6426,
+      "step": 28108
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.359421846487298e-08,
+      "loss": 0.6429,
+      "step": 28109
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.350509986783546e-08,
+      "loss": 0.6812,
+      "step": 28110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.341609943620206e-08,
+      "loss": 0.6233,
+      "step": 28111
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.33272171710286e-08,
+      "loss": 0.6389,
+      "step": 28112
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.32384530733687e-08,
+      "loss": 0.5806,
+      "step": 28113
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.314980714427596e-08,
+      "loss": 0.6903,
+      "step": 28114
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3061279384799524e-08,
+      "loss": 0.6197,
+      "step": 28115
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2972869795989684e-08,
+      "loss": 0.6441,
+      "step": 28116
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.288457837889558e-08,
+      "loss": 0.6756,
+      "step": 28117
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.279640513456306e-08,
+      "loss": 0.6547,
+      "step": 28118
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.270835006403683e-08,
+      "loss": 0.639,
+      "step": 28119
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.262041316836384e-08,
+      "loss": 0.557,
+      "step": 28120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2532594448583256e-08,
+      "loss": 0.5904,
+      "step": 28121
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.244489390573757e-08,
+      "loss": 0.5806,
+      "step": 28122
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.235731154086708e-08,
+      "loss": 0.6929,
+      "step": 28123
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.226984735500871e-08,
+      "loss": 0.6341,
+      "step": 28124
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.218250134920164e-08,
+      "loss": 0.6335,
+      "step": 28125
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2095273524480605e-08,
+      "loss": 0.6005,
+      "step": 28126
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2008163881879215e-08,
+      "loss": 0.6604,
+      "step": 28127
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.192117242242998e-08,
+      "loss": 0.6241,
+      "step": 28128
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.18342991471654e-08,
+      "loss": 0.56,
+      "step": 28129
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.174754405711578e-08,
+      "loss": 0.625,
+      "step": 28130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.166090715330805e-08,
+      "loss": 0.6351,
+      "step": 28131
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1574388436770296e-08,
+      "loss": 0.5887,
+      "step": 28132
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.148798790852836e-08,
+      "loss": 0.614,
+      "step": 28133
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.140170556960698e-08,
+      "loss": 0.6372,
+      "step": 28134
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.131554142102866e-08,
+      "loss": 0.5924,
+      "step": 28135
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.122949546381482e-08,
+      "loss": 0.6266,
+      "step": 28136
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1143567698985745e-08,
+      "loss": 0.6004,
+      "step": 28137
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.105775812756062e-08,
+      "loss": 0.6235,
+      "step": 28138
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.097206675055642e-08,
+      "loss": 0.5666,
+      "step": 28139
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.088649356898899e-08,
+      "loss": 0.6149,
+      "step": 28140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.080103858387306e-08,
+      "loss": 0.6079,
+      "step": 28141
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0715701796222295e-08,
+      "loss": 0.6331,
+      "step": 28142
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.063048320704809e-08,
+      "loss": 0.6403,
+      "step": 28143
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0545382817360744e-08,
+      "loss": 0.5923,
+      "step": 28144
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0460400628169465e-08,
+      "loss": 0.6324,
+      "step": 28145
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.037553664048121e-08,
+      "loss": 0.6284,
+      "step": 28146
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.029079085530407e-08,
+      "loss": 0.6388,
+      "step": 28147
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0206163273639455e-08,
+      "loss": 0.5888,
+      "step": 28148
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.012165389649435e-08,
+      "loss": 0.6667,
+      "step": 28149
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0037262724867957e-08,
+      "loss": 0.6239,
+      "step": 28150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9952989759761684e-08,
+      "loss": 0.651,
+      "step": 28151
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.986883500217475e-08,
+      "loss": 0.6098,
+      "step": 28152
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9784798453106332e-08,
+      "loss": 0.5797,
+      "step": 28153
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.97008801135501e-08,
+      "loss": 0.6528,
+      "step": 28154
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9617079984503028e-08,
+      "loss": 0.6272,
+      "step": 28155
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9533398066957652e-08,
+      "loss": 0.6359,
+      "step": 28156
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9449834361907624e-08,
+      "loss": 0.6321,
+      "step": 28157
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9366388870342157e-08,
+      "loss": 0.6392,
+      "step": 28158
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.928306159325045e-08,
+      "loss": 0.6362,
+      "step": 28159
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.919985253162172e-08,
+      "loss": 0.5915,
+      "step": 28160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9116761686441842e-08,
+      "loss": 0.6651,
+      "step": 28161
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9033789058695584e-08,
+      "loss": 0.5789,
+      "step": 28162
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8950934649366603e-08,
+      "loss": 0.6223,
+      "step": 28163
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8868198459439668e-08,
+      "loss": 0.6748,
+      "step": 28164
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8785580489891774e-08,
+      "loss": 0.5843,
+      "step": 28165
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.870308074170547e-08,
+      "loss": 0.6524,
+      "step": 28166
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.862069921585775e-08,
+      "loss": 0.5715,
+      "step": 28167
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8538435913325614e-08,
+      "loss": 0.6201,
+      "step": 28168
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8456290835083834e-08,
+      "loss": 0.6673,
+      "step": 28169
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8374263982107188e-08,
+      "loss": 0.6236,
+      "step": 28170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8292355355367118e-08,
+      "loss": 0.6176,
+      "step": 28171
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8210564955836182e-08,
+      "loss": 0.6415,
+      "step": 28172
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.812889278448361e-08,
+      "loss": 0.578,
+      "step": 28173
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8047338842276396e-08,
+      "loss": 0.6248,
+      "step": 28174
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.796590313018377e-08,
+      "loss": 0.6093,
+      "step": 28175
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7884585649169404e-08,
+      "loss": 0.6385,
+      "step": 28176
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7803386400198085e-08,
+      "loss": 0.6322,
+      "step": 28177
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7722305384232372e-08,
+      "loss": 0.6489,
+      "step": 28178
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7641342602234833e-08,
+      "loss": 0.5864,
+      "step": 28179
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7560498055163586e-08,
+      "loss": 0.6309,
+      "step": 28180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7479771743976756e-08,
+      "loss": 0.5959,
+      "step": 28181
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7399163669634686e-08,
+      "loss": 0.5952,
+      "step": 28182
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.731867383308884e-08,
+      "loss": 0.5874,
+      "step": 28183
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7238302235296223e-08,
+      "loss": 0.6611,
+      "step": 28184
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.715804887720941e-08,
+      "loss": 0.5486,
+      "step": 28185
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.707791375977986e-08,
+      "loss": 0.642,
+      "step": 28186
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.699789688395793e-08,
+      "loss": 0.6064,
+      "step": 28187
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6917998250690636e-08,
+      "loss": 0.6259,
+      "step": 28188
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6838217860927217e-08,
+      "loss": 0.6322,
+      "step": 28189
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.675855571561248e-08,
+      "loss": 0.6303,
+      "step": 28190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6679011815691213e-08,
+      "loss": 0.6373,
+      "step": 28191
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6599586162107115e-08,
+      "loss": 0.6035,
+      "step": 28192
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6520278755800545e-08,
+      "loss": 0.6698,
+      "step": 28193
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6441089597714077e-08,
+      "loss": 0.6287,
+      "step": 28194
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6362018688783632e-08,
+      "loss": 0.5971,
+      "step": 28195
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6283066029948457e-08,
+      "loss": 0.5887,
+      "step": 28196
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.620423162214447e-08,
+      "loss": 0.6369,
+      "step": 28197
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6125515466306483e-08,
+      "loss": 0.6916,
+      "step": 28198
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.604691756336708e-08,
+      "loss": 0.6508,
+      "step": 28199
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5968437914258847e-08,
+      "loss": 0.6292,
+      "step": 28200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5890076519912155e-08,
+      "loss": 0.5955,
+      "step": 28201
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5811833381256256e-08,
+      "loss": 0.5989,
+      "step": 28202
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.573370849921819e-08,
+      "loss": 0.6607,
+      "step": 28203
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5655701874723883e-08,
+      "loss": 0.6538,
+      "step": 28204
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5577813508699258e-08,
+      "loss": 0.6581,
+      "step": 28205
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5500043402068018e-08,
+      "loss": 0.6811,
+      "step": 28206
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.542239155575055e-08,
+      "loss": 0.6678,
+      "step": 28207
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5344857970668325e-08,
+      "loss": 0.6011,
+      "step": 28208
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5267442647741724e-08,
+      "loss": 0.6624,
+      "step": 28209
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5190145587887792e-08,
+      "loss": 0.6347,
+      "step": 28210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5112966792022466e-08,
+      "loss": 0.6316,
+      "step": 28211
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.503590626106056e-08,
+      "loss": 0.6581,
+      "step": 28212
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4958963995918016e-08,
+      "loss": 0.6597,
+      "step": 28213
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4882139997504107e-08,
+      "loss": 0.6303,
+      "step": 28214
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.480543426673143e-08,
+      "loss": 0.6101,
+      "step": 28215
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4728846804508154e-08,
+      "loss": 0.5769,
+      "step": 28216
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.465237761174466e-08,
+      "loss": 0.5982,
+      "step": 28217
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4576026689345776e-08,
+      "loss": 0.6865,
+      "step": 28218
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.449979403821634e-08,
+      "loss": 0.6093,
+      "step": 28219
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.442367965926118e-08,
+      "loss": 0.6093,
+      "step": 28220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4347683553382906e-08,
+      "loss": 0.6699,
+      "step": 28221
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.427180572148191e-08,
+      "loss": 0.6841,
+      "step": 28222
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4196046164457477e-08,
+      "loss": 0.5661,
+      "step": 28223
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.412040488320888e-08,
+      "loss": 0.6354,
+      "step": 28224
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4044881878633185e-08,
+      "loss": 0.6161,
+      "step": 28225
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.396947715162412e-08,
+      "loss": 0.6634,
+      "step": 28226
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3894190703076525e-08,
+      "loss": 0.5843,
+      "step": 28227
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3819022533883017e-08,
+      "loss": 0.6183,
+      "step": 28228
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3743972644936218e-08,
+      "loss": 0.6369,
+      "step": 28229
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3669041037123196e-08,
+      "loss": 0.6155,
+      "step": 28230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.359422771133435e-08,
+      "loss": 0.6255,
+      "step": 28231
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3519532668456748e-08,
+      "loss": 0.5967,
+      "step": 28232
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3444955909374124e-08,
+      "loss": 0.6319,
+      "step": 28233
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.337049743497244e-08,
+      "loss": 0.6119,
+      "step": 28234
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.329615724613432e-08,
+      "loss": 0.6175,
+      "step": 28235
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3221935343741286e-08,
+      "loss": 0.5878,
+      "step": 28236
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3147831728672632e-08,
+      "loss": 0.5933,
+      "step": 28237
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3073846401806542e-08,
+      "loss": 0.6492,
+      "step": 28238
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2999979364021208e-08,
+      "loss": 0.5737,
+      "step": 28239
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.292623061619259e-08,
+      "loss": 0.5918,
+      "step": 28240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2852600159194437e-08,
+      "loss": 0.6253,
+      "step": 28241
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2779087993899384e-08,
+      "loss": 0.6438,
+      "step": 28242
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2705694121178955e-08,
+      "loss": 0.6781,
+      "step": 28243
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.263241854190468e-08,
+      "loss": 0.6094,
+      "step": 28244
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2559261256944742e-08,
+      "loss": 0.5883,
+      "step": 28245
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2486222267166237e-08,
+      "loss": 0.6497,
+      "step": 28246
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.241330157343513e-08,
+      "loss": 0.6007,
+      "step": 28247
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.234049917661518e-08,
+      "loss": 0.5793,
+      "step": 28248
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2267815077571253e-08,
+      "loss": 0.646,
+      "step": 28249
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2195249277163766e-08,
+      "loss": 0.6486,
+      "step": 28250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2122801776254254e-08,
+      "loss": 0.6219,
+      "step": 28251
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2050472575700922e-08,
+      "loss": 0.6404,
+      "step": 28252
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1978261676360858e-08,
+      "loss": 0.6614,
+      "step": 28253
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1906169079091154e-08,
+      "loss": 0.6267,
+      "step": 28254
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1834194784746688e-08,
+      "loss": 0.6183,
+      "step": 28255
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1762338794178993e-08,
+      "loss": 0.6172,
+      "step": 28256
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1690601108241838e-08,
+      "loss": 0.5989,
+      "step": 28257
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1618981727785648e-08,
+      "loss": 0.6468,
+      "step": 28258
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1547480653659748e-08,
+      "loss": 0.5957,
+      "step": 28259
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1476097886710122e-08,
+      "loss": 0.6419,
+      "step": 28260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.140483342778499e-08,
+      "loss": 0.5833,
+      "step": 28261
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1333687277727e-08,
+      "loss": 0.6629,
+      "step": 28262
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.126265943738326e-08,
+      "loss": 0.6332,
+      "step": 28263
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1191749907591986e-08,
+      "loss": 0.6001,
+      "step": 28264
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1120958689196948e-08,
+      "loss": 0.6187,
+      "step": 28265
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.105028578303636e-08,
+      "loss": 0.669,
+      "step": 28266
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.097973118994845e-08,
+      "loss": 0.6471,
+      "step": 28267
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0909294910768097e-08,
+      "loss": 0.6517,
+      "step": 28268
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.083897694633241e-08,
+      "loss": 0.631,
+      "step": 28269
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0768777297474064e-08,
+      "loss": 0.6601,
+      "step": 28270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0698695965025716e-08,
+      "loss": 0.6119,
+      "step": 28271
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.062873294981893e-08,
+      "loss": 0.592,
+      "step": 28272
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.055888825268193e-08,
+      "loss": 0.6639,
+      "step": 28273
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.048916187444294e-08,
+      "loss": 0.6188,
+      "step": 28274
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0419553815929083e-08,
+      "loss": 0.642,
+      "step": 28275
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.035006407796636e-08,
+      "loss": 0.6065,
+      "step": 28276
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0280692661377443e-08,
+      "loss": 0.6581,
+      "step": 28277
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0211439566986123e-08,
+      "loss": 0.6707,
+      "step": 28278
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0142304795611745e-08,
+      "loss": 0.6306,
+      "step": 28279
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.007328834807476e-08,
+      "loss": 0.6442,
+      "step": 28280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0004390225194516e-08,
+      "loss": 0.6149,
+      "step": 28281
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.993561042778702e-08,
+      "loss": 0.7116,
+      "step": 28282
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9866948956667186e-08,
+      "loss": 0.5911,
+      "step": 28283
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9798405812648803e-08,
+      "loss": 0.6028,
+      "step": 28284
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.972998099654677e-08,
+      "loss": 0.6076,
+      "step": 28285
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9661674509170447e-08,
+      "loss": 0.6745,
+      "step": 28286
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9593486351329184e-08,
+      "loss": 0.6041,
+      "step": 28287
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.952541652383344e-08,
+      "loss": 0.6608,
+      "step": 28288
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9457465027489243e-08,
+      "loss": 0.5779,
+      "step": 28289
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.938963186310261e-08,
+      "loss": 0.5874,
+      "step": 28290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9321917031476233e-08,
+      "loss": 0.6641,
+      "step": 28291
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9254320533415026e-08,
+      "loss": 0.6896,
+      "step": 28292
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.918684236971946e-08,
+      "loss": 0.6146,
+      "step": 28293
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9119482541190005e-08,
+      "loss": 0.6478,
+      "step": 28294
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9052241048624907e-08,
+      "loss": 0.6505,
+      "step": 28295
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8985117892820205e-08,
+      "loss": 0.5918,
+      "step": 28296
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.891811307457414e-08,
+      "loss": 0.5845,
+      "step": 28297
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8851226594679416e-08,
+      "loss": 0.6209,
+      "step": 28298
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8784458453929845e-08,
+      "loss": 0.6682,
+      "step": 28299
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.871780865311812e-08,
+      "loss": 0.6415,
+      "step": 28300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8651277193031392e-08,
+      "loss": 0.6546,
+      "step": 28301
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.858486407446125e-08,
+      "loss": 0.5951,
+      "step": 28302
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8518569298192622e-08,
+      "loss": 0.6298,
+      "step": 28303
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.845239286501488e-08,
+      "loss": 0.6364,
+      "step": 28304
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8386334775709614e-08,
+      "loss": 0.5829,
+      "step": 28305
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8320395031061755e-08,
+      "loss": 0.6717,
+      "step": 28306
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.825457363185179e-08,
+      "loss": 0.7132,
+      "step": 28307
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8188870578861317e-08,
+      "loss": 0.661,
+      "step": 28308
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.81232858728686e-08,
+      "loss": 0.6154,
+      "step": 28309
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.805781951465191e-08,
+      "loss": 0.6504,
+      "step": 28310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7992471504986175e-08,
+      "loss": 0.6004,
+      "step": 28311
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7927241844647447e-08,
+      "loss": 0.6056,
+      "step": 28312
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.786213053440955e-08,
+      "loss": 0.5955,
+      "step": 28313
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7797137575041868e-08,
+      "loss": 0.6749,
+      "step": 28314
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7732262967318227e-08,
+      "loss": 0.6553,
+      "step": 28315
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.766750671200468e-08,
+      "loss": 0.6761,
+      "step": 28316
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.760286880987172e-08,
+      "loss": 0.5783,
+      "step": 28317
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7538349261684294e-08,
+      "loss": 0.6433,
+      "step": 28318
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7473948068207346e-08,
+      "loss": 0.6567,
+      "step": 28319
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7409665230205818e-08,
+      "loss": 0.5975,
+      "step": 28320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7345500748439104e-08,
+      "loss": 0.6318,
+      "step": 28321
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7281454623671034e-08,
+      "loss": 0.6654,
+      "step": 28322
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7217526856658785e-08,
+      "loss": 0.5984,
+      "step": 28323
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7153717448160634e-08,
+      "loss": 0.5916,
+      "step": 28324
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7090026398933756e-08,
+      "loss": 0.6907,
+      "step": 28325
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7026453709734216e-08,
+      "loss": 0.6716,
+      "step": 28326
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6962999381313627e-08,
+      "loss": 0.5919,
+      "step": 28327
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6899663414424727e-08,
+      "loss": 0.6743,
+      "step": 28328
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6836445809820244e-08,
+      "loss": 0.6611,
+      "step": 28329
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.677334656824736e-08,
+      "loss": 0.6144,
+      "step": 28330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6710365690455475e-08,
+      "loss": 0.574,
+      "step": 28331
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.664750317719177e-08,
+      "loss": 0.6406,
+      "step": 28332
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.65847590292012e-08,
+      "loss": 0.6129,
+      "step": 28333
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6522133247226514e-08,
+      "loss": 0.6183,
+      "step": 28334
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.645962583201155e-08,
+      "loss": 0.6566,
+      "step": 28335
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6397236784296834e-08,
+      "loss": 0.6053,
+      "step": 28336
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.633496610482177e-08,
+      "loss": 0.5867,
+      "step": 28337
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.627281379432577e-08,
+      "loss": 0.62,
+      "step": 28338
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.621077985354491e-08,
+      "loss": 0.6214,
+      "step": 28339
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6148864283214162e-08,
+      "loss": 0.654,
+      "step": 28340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6087067084068485e-08,
+      "loss": 0.5842,
+      "step": 28341
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6025388256840637e-08,
+      "loss": 0.6561,
+      "step": 28342
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5963827802262244e-08,
+      "loss": 0.5718,
+      "step": 28343
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.590238572106162e-08,
+      "loss": 0.6059,
+      "step": 28344
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5841062013968178e-08,
+      "loss": 0.6731,
+      "step": 28345
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5779856681709116e-08,
+      "loss": 0.6346,
+      "step": 28346
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5718769725010518e-08,
+      "loss": 0.6199,
+      "step": 28347
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.565780114459514e-08,
+      "loss": 0.6037,
+      "step": 28348
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.559695094118796e-08,
+      "loss": 0.6799,
+      "step": 28349
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.55362191155084e-08,
+      "loss": 0.6496,
+      "step": 28350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5475605668276992e-08,
+      "loss": 0.6223,
+      "step": 28351
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.541511060021317e-08,
+      "loss": 0.5902,
+      "step": 28352
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5354733912035237e-08,
+      "loss": 0.6326,
+      "step": 28353
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.529447560445596e-08,
+      "loss": 0.6131,
+      "step": 28354
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5234335678191435e-08,
+      "loss": 0.5918,
+      "step": 28355
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5174314133955536e-08,
+      "loss": 0.6246,
+      "step": 28356
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5114410972457692e-08,
+      "loss": 0.6214,
+      "step": 28357
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.505462619441067e-08,
+      "loss": 0.6603,
+      "step": 28358
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.499495980052168e-08,
+      "loss": 0.6727,
+      "step": 28359
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4935411791497933e-08,
+      "loss": 0.6372,
+      "step": 28360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4875982168046643e-08,
+      "loss": 0.6553,
+      "step": 28361
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4816670930871691e-08,
+      "loss": 0.6398,
+      "step": 28362
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4757478080675846e-08,
+      "loss": 0.6283,
+      "step": 28363
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4698403618162994e-08,
+      "loss": 0.6309,
+      "step": 28364
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.463944754403146e-08,
+      "loss": 0.5699,
+      "step": 28365
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4580609858979577e-08,
+      "loss": 0.6547,
+      "step": 28366
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4521890563707896e-08,
+      "loss": 0.6059,
+      "step": 28367
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4463289658910306e-08,
+      "loss": 0.6154,
+      "step": 28368
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4404807145281808e-08,
+      "loss": 0.603,
+      "step": 28369
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4346443023516287e-08,
+      "loss": 0.6367,
+      "step": 28370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4288197294305418e-08,
+      "loss": 0.6105,
+      "step": 28371
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4230069958340865e-08,
+      "loss": 0.6117,
+      "step": 28372
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4172061016309857e-08,
+      "loss": 0.6797,
+      "step": 28373
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4114170468901845e-08,
+      "loss": 0.6099,
+      "step": 28374
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4056398316801833e-08,
+      "loss": 0.5442,
+      "step": 28375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3998744560695942e-08,
+      "loss": 0.5611,
+      "step": 28376
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3941209201268069e-08,
+      "loss": 0.6712,
+      "step": 28377
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.388379223919878e-08,
+      "loss": 0.6357,
+      "step": 28378
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.382649367516975e-08,
+      "loss": 0.6748,
+      "step": 28379
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3769313509860438e-08,
+      "loss": 0.6127,
+      "step": 28380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.371225174394808e-08,
+      "loss": 0.6596,
+      "step": 28381
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.36553083781088e-08,
+      "loss": 0.5797,
+      "step": 28382
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3598483413020947e-08,
+      "loss": 0.6329,
+      "step": 28383
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3541776849353982e-08,
+      "loss": 0.6609,
+      "step": 28384
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3485188687782925e-08,
+      "loss": 0.6355,
+      "step": 28385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3428718928978346e-08,
+      "loss": 0.5823,
+      "step": 28386
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3372367573608603e-08,
+      "loss": 0.6316,
+      "step": 28387
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3316134622343157e-08,
+      "loss": 0.5929,
+      "step": 28388
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3260020075848145e-08,
+      "loss": 0.5873,
+      "step": 28389
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.320402393478859e-08,
+      "loss": 0.6726,
+      "step": 28390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3148146199829515e-08,
+      "loss": 0.6671,
+      "step": 28391
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3092386871631502e-08,
+      "loss": 0.6697,
+      "step": 28392
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3036745950857355e-08,
+      "loss": 0.654,
+      "step": 28393
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2981223438165436e-08,
+      "loss": 0.592,
+      "step": 28394
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2925819334216327e-08,
+      "loss": 0.6436,
+      "step": 28395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.287053363966395e-08,
+      "loss": 0.6043,
+      "step": 28396
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2815366355165559e-08,
+      "loss": 0.6192,
+      "step": 28397
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2760317481375072e-08,
+      "loss": 0.6514,
+      "step": 28398
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2705387018944193e-08,
+      "loss": 0.6532,
+      "step": 28399
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.265057496852573e-08,
+      "loss": 0.6437,
+      "step": 28400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2595881330768056e-08,
+      "loss": 0.5935,
+      "step": 28401
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.254130610632176e-08,
+      "loss": 0.6027,
+      "step": 28402
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2486849295830773e-08,
+      "loss": 0.6469,
+      "step": 28403
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2432510899943461e-08,
+      "loss": 0.6122,
+      "step": 28404
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2378290919302649e-08,
+      "loss": 0.6037,
+      "step": 28405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2324189354552263e-08,
+      "loss": 0.6983,
+      "step": 28406
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2270206206331791e-08,
+      "loss": 0.6443,
+      "step": 28407
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2216341475284055e-08,
+      "loss": 0.5957,
+      "step": 28408
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2162595162045209e-08,
+      "loss": 0.6354,
+      "step": 28409
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2108967267253636e-08,
+      "loss": 0.6387,
+      "step": 28410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2055457791545488e-08,
+      "loss": 0.6473,
+      "step": 28411
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2002066735554708e-08,
+      "loss": 0.692,
+      "step": 28412
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1948794099914118e-08,
+      "loss": 0.6302,
+      "step": 28413
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1895639885254328e-08,
+      "loss": 0.6444,
+      "step": 28414
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1842604092208166e-08,
+      "loss": 0.6533,
+      "step": 28415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1789686721402904e-08,
+      "loss": 0.5641,
+      "step": 28416
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1736887773465821e-08,
+      "loss": 0.6041,
+      "step": 28417
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1684207249023083e-08,
+      "loss": 0.6452,
+      "step": 28418
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1631645148699745e-08,
+      "loss": 0.6316,
+      "step": 28419
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1579201473117529e-08,
+      "loss": 0.6548,
+      "step": 28420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1526876222900384e-08,
+      "loss": 0.666,
+      "step": 28421
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1474669398667815e-08,
+      "loss": 0.6517,
+      "step": 28422
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1422581001037104e-08,
+      "loss": 0.6103,
+      "step": 28423
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1370611030628863e-08,
+      "loss": 0.6297,
+      "step": 28424
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1318759488057052e-08,
+      "loss": 0.6349,
+      "step": 28425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.126702637393673e-08,
+      "loss": 0.601,
+      "step": 28426
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.121541168888296e-08,
+      "loss": 0.6587,
+      "step": 28427
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1163915433505256e-08,
+      "loss": 0.6372,
+      "step": 28428
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1112537608415353e-08,
+      "loss": 0.6093,
+      "step": 28429
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1061278214221648e-08,
+      "loss": 0.6176,
+      "step": 28430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.101013725153366e-08,
+      "loss": 0.6362,
+      "step": 28431
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0959114720956455e-08,
+      "loss": 0.6633,
+      "step": 28432
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0908210623095106e-08,
+      "loss": 0.5725,
+      "step": 28433
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0857424958552465e-08,
+      "loss": 0.6476,
+      "step": 28434
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0806757727932492e-08,
+      "loss": 0.6465,
+      "step": 28435
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.075620893183471e-08,
+      "loss": 0.591,
+      "step": 28436
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0705778570857528e-08,
+      "loss": 0.6364,
+      "step": 28437
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0655466645600465e-08,
+      "loss": 0.5904,
+      "step": 28438
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0605273156659712e-08,
+      "loss": 0.6101,
+      "step": 28439
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0555198104629238e-08,
+      "loss": 0.6134,
+      "step": 28440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0505241490104124e-08,
+      "loss": 0.6702,
+      "step": 28441
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0455403313677226e-08,
+      "loss": 0.6621,
+      "step": 28442
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0405683575936964e-08,
+      "loss": 0.653,
+      "step": 28443
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0356082277473978e-08,
+      "loss": 0.6502,
+      "step": 28444
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0306599418877794e-08,
+      "loss": 0.5884,
+      "step": 28445
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0257235000733502e-08,
+      "loss": 0.5735,
+      "step": 28446
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0207989023626185e-08,
+      "loss": 0.6047,
+      "step": 28447
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0158861488140937e-08,
+      "loss": 0.5816,
+      "step": 28448
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.010985239485951e-08,
+      "loss": 0.6137,
+      "step": 28449
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0060961744362552e-08,
+      "loss": 0.6513,
+      "step": 28450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.001218953723071e-08,
+      "loss": 0.5989,
+      "step": 28451
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.96353577404241e-09,
+      "loss": 0.6278,
+      "step": 28452
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.915000455372747e-09,
+      "loss": 0.6385,
+      "step": 28453
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.866583581797928e-09,
+      "loss": 0.6278,
+      "step": 28454
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.818285153893048e-09,
+      "loss": 0.6273,
+      "step": 28455
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.77010517223098e-09,
+      "loss": 0.6738,
+      "step": 28456
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.722043637381273e-09,
+      "loss": 0.6673,
+      "step": 28457
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.674100549913468e-09,
+      "loss": 0.6325,
+      "step": 28458
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.62627591039822e-09,
+      "loss": 0.6233,
+      "step": 28459
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.578569719400632e-09,
+      "loss": 0.691,
+      "step": 28460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.53098197748803e-09,
+      "loss": 0.6054,
+      "step": 28461
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.483512685223294e-09,
+      "loss": 0.5654,
+      "step": 28462
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.436161843169312e-09,
+      "loss": 0.6365,
+      "step": 28463
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.388929451887852e-09,
+      "loss": 0.6672,
+      "step": 28464
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.341815511938467e-09,
+      "loss": 0.5744,
+      "step": 28465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.294820023880713e-09,
+      "loss": 0.6426,
+      "step": 28466
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.247942988270808e-09,
+      "loss": 0.6044,
+      "step": 28467
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.201184405664976e-09,
+      "loss": 0.6365,
+      "step": 28468
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.154544276617217e-09,
+      "loss": 0.5471,
+      "step": 28469
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.108022601681533e-09,
+      "loss": 0.6986,
+      "step": 28470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.061619381408593e-09,
+      "loss": 0.6622,
+      "step": 28471
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.01533461634796e-09,
+      "loss": 0.584,
+      "step": 28472
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.969168307050302e-09,
+      "loss": 0.649,
+      "step": 28473
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.923120454060742e-09,
+      "loss": 0.6319,
+      "step": 28474
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.877191057927725e-09,
+      "loss": 0.5914,
+      "step": 28475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.831380119193045e-09,
+      "loss": 0.5833,
+      "step": 28476
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.785687638401818e-09,
+      "loss": 0.6956,
+      "step": 28477
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.740113616095835e-09,
+      "loss": 0.5905,
+      "step": 28478
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.694658052813554e-09,
+      "loss": 0.6448,
+      "step": 28479
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.649320949096763e-09,
+      "loss": 0.659,
+      "step": 28480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.604102305480588e-09,
+      "loss": 0.5929,
+      "step": 28481
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.55900212250127e-09,
+      "loss": 0.6072,
+      "step": 28482
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.514020400695044e-09,
+      "loss": 0.6375,
+      "step": 28483
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.469157140593708e-09,
+      "loss": 0.6021,
+      "step": 28484
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.424412342731281e-09,
+      "loss": 0.7119,
+      "step": 28485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.379786007635116e-09,
+      "loss": 0.5886,
+      "step": 28486
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.33527813583701e-09,
+      "loss": 0.6677,
+      "step": 28487
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.29088872786432e-09,
+      "loss": 0.6196,
+      "step": 28488
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.246617784241073e-09,
+      "loss": 0.6259,
+      "step": 28489
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.202465305494622e-09,
+      "loss": 0.5653,
+      "step": 28490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.158431292147883e-09,
+      "loss": 0.6389,
+      "step": 28491
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.114515744722663e-09,
+      "loss": 0.6559,
+      "step": 28492
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.070718663739652e-09,
+      "loss": 0.6219,
+      "step": 28493
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.027040049718438e-09,
+      "loss": 0.6306,
+      "step": 28494
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.983479903176384e-09,
+      "loss": 0.6344,
+      "step": 28495
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.940038224629742e-09,
+      "loss": 0.5985,
+      "step": 28496
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.89671501459366e-09,
+      "loss": 0.6606,
+      "step": 28497
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.853510273583276e-09,
+      "loss": 0.667,
+      "step": 28498
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.810424002108185e-09,
+      "loss": 0.6514,
+      "step": 28499
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.767456200681311e-09,
+      "loss": 0.6523,
+      "step": 28500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.724606869812245e-09,
+      "loss": 0.6321,
+      "step": 28501
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.68187601000725e-09,
+      "loss": 0.5889,
+      "step": 28502
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.639263621774806e-09,
+      "loss": 0.643,
+      "step": 28503
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.596769705618956e-09,
+      "loss": 0.6249,
+      "step": 28504
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.55439426204374e-09,
+      "loss": 0.6627,
+      "step": 28505
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.51213729155098e-09,
+      "loss": 0.5699,
+      "step": 28506
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.469998794643607e-09,
+      "loss": 0.6144,
+      "step": 28507
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.427978771819e-09,
+      "loss": 0.5973,
+      "step": 28508
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.38607722357676e-09,
+      "loss": 0.6161,
+      "step": 28509
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.344294150414266e-09,
+      "loss": 0.6234,
+      "step": 28510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.302629552824458e-09,
+      "loss": 0.5813,
+      "step": 28511
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.261083431303606e-09,
+      "loss": 0.6375,
+      "step": 28512
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.219655786343538e-09,
+      "loss": 0.6223,
+      "step": 28513
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.178346618434973e-09,
+      "loss": 0.6338,
+      "step": 28514
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.1371559280686285e-09,
+      "loss": 0.6116,
+      "step": 28515
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.096083715731894e-09,
+      "loss": 0.6955,
+      "step": 28516
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.055129981912157e-09,
+      "loss": 0.5827,
+      "step": 28517
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.014294727095694e-09,
+      "loss": 0.6374,
+      "step": 28518
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.9735779517643436e-09,
+      "loss": 0.6246,
+      "step": 28519
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.9329796564043814e-09,
+      "loss": 0.6136,
+      "step": 28520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.8924998414943155e-09,
+      "loss": 0.6548,
+      "step": 28521
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.8521385075148716e-09,
+      "loss": 0.6128,
+      "step": 28522
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.8118956549445555e-09,
+      "loss": 0.606,
+      "step": 28523
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.771771284261874e-09,
+      "loss": 0.6236,
+      "step": 28524
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.731765395939782e-09,
+      "loss": 0.696,
+      "step": 28525
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.691877990454565e-09,
+      "loss": 0.6131,
+      "step": 28526
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.652109068279178e-09,
+      "loss": 0.6248,
+      "step": 28527
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.612458629885465e-09,
+      "loss": 0.5928,
+      "step": 28528
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.572926675741942e-09,
+      "loss": 0.6532,
+      "step": 28529
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.5335132063171216e-09,
+      "loss": 0.6587,
+      "step": 28530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.494218222080628e-09,
+      "loss": 0.6133,
+      "step": 28531
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.455041723496536e-09,
+      "loss": 0.62,
+      "step": 28532
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.4159837110300274e-09,
+      "loss": 0.6456,
+      "step": 28533
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.377044185144066e-09,
+      "loss": 0.7008,
+      "step": 28534
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.3382231462993935e-09,
+      "loss": 0.5842,
+      "step": 28535
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.299520594957864e-09,
+      "loss": 0.6262,
+      "step": 28536
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.260936531577999e-09,
+      "loss": 0.6123,
+      "step": 28537
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.2224709566161e-09,
+      "loss": 0.6397,
+      "step": 28538
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.184123870529579e-09,
+      "loss": 0.5864,
+      "step": 28539
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.145895273772517e-09,
+      "loss": 0.5948,
+      "step": 28540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.107785166797886e-09,
+      "loss": 0.6085,
+      "step": 28541
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.069793550057546e-09,
+      "loss": 0.6461,
+      "step": 28542
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.031920424002247e-09,
+      "loss": 0.6111,
+      "step": 28543
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.994165789081629e-09,
+      "loss": 0.6773,
+      "step": 28544
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.956529645742004e-09,
+      "loss": 0.6717,
+      "step": 28545
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.91901199443079e-09,
+      "loss": 0.6506,
+      "step": 28546
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.881612835590966e-09,
+      "loss": 0.6277,
+      "step": 28547
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.844332169668843e-09,
+      "loss": 0.6311,
+      "step": 28548
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.807169997104067e-09,
+      "loss": 0.6407,
+      "step": 28549
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.770126318338509e-09,
+      "loss": 0.5476,
+      "step": 28550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.733201133810706e-09,
+      "loss": 0.6124,
+      "step": 28551
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.696394443958087e-09,
+      "loss": 0.6276,
+      "step": 28552
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.6597062492180775e-09,
+      "loss": 0.6335,
+      "step": 28553
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.623136550024777e-09,
+      "loss": 0.6818,
+      "step": 28554
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.586685346812281e-09,
+      "loss": 0.6185,
+      "step": 28555
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.550352640012468e-09,
+      "loss": 0.5974,
+      "step": 28556
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.514138430056104e-09,
+      "loss": 0.6019,
+      "step": 28557
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.478042717373955e-09,
+      "loss": 0.6146,
+      "step": 28558
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.442065502391236e-09,
+      "loss": 0.6531,
+      "step": 28559
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.406206785536494e-09,
+      "loss": 0.5844,
+      "step": 28560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.370466567234944e-09,
+      "loss": 0.6434,
+      "step": 28561
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.334844847909582e-09,
+      "loss": 0.6174,
+      "step": 28562
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.299341627982291e-09,
+      "loss": 0.6235,
+      "step": 28563
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.263956907876067e-09,
+      "loss": 0.6161,
+      "step": 28564
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.228690688007243e-09,
+      "loss": 0.6418,
+      "step": 28565
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.193542968796594e-09,
+      "loss": 0.6165,
+      "step": 28566
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.158513750660455e-09,
+      "loss": 0.6049,
+      "step": 28567
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.1236030340129365e-09,
+      "loss": 0.7095,
+      "step": 28568
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.088810819269263e-09,
+      "loss": 0.5931,
+      "step": 28569
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.054137106841328e-09,
+      "loss": 0.6026,
+      "step": 28570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.019581897139913e-09,
+      "loss": 0.631,
+      "step": 28571
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.985145190574692e-09,
+      "loss": 0.6307,
+      "step": 28572
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.950826987555335e-09,
+      "loss": 0.6616,
+      "step": 28573
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.916627288487075e-09,
+      "loss": 0.6727,
+      "step": 28574
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.882546093776252e-09,
+      "loss": 0.6269,
+      "step": 28575
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.84858340382699e-09,
+      "loss": 0.6522,
+      "step": 28576
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8147392190411865e-09,
+      "loss": 0.6094,
+      "step": 28577
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7810135398207444e-09,
+      "loss": 0.6305,
+      "step": 28578
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.747406366565344e-09,
+      "loss": 0.6214,
+      "step": 28579
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7139176996735535e-09,
+      "loss": 0.6174,
+      "step": 28580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.680547539542834e-09,
+      "loss": 0.6193,
+      "step": 28581
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.647295886568426e-09,
+      "loss": 0.5999,
+      "step": 28582
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.614162741143346e-09,
+      "loss": 0.6143,
+      "step": 28583
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.581148103662836e-09,
+      "loss": 0.7111,
+      "step": 28584
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.548251974516582e-09,
+      "loss": 0.6237,
+      "step": 28585
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.515474354095384e-09,
+      "loss": 0.6208,
+      "step": 28586
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.482815242786709e-09,
+      "loss": 0.5909,
+      "step": 28587
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4502746409791355e-09,
+      "loss": 0.6381,
+      "step": 28588
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4178525490590204e-09,
+      "loss": 0.598,
+      "step": 28589
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.385548967408282e-09,
+      "loss": 0.6797,
+      "step": 28590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.353363896412166e-09,
+      "loss": 0.6255,
+      "step": 28591
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.32129733645148e-09,
+      "loss": 0.6151,
+      "step": 28592
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.28934928790592e-09,
+      "loss": 0.6652,
+      "step": 28593
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.257519751155182e-09,
+      "loss": 0.6358,
+      "step": 28594
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.225808726575631e-09,
+      "loss": 0.596,
+      "step": 28595
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.194216214544744e-09,
+      "loss": 0.6228,
+      "step": 28596
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.162742215435555e-09,
+      "loss": 0.654,
+      "step": 28597
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.131386729622211e-09,
+      "loss": 0.6188,
+      "step": 28598
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.100149757475524e-09,
+      "loss": 0.6751,
+      "step": 28599
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.069031299367421e-09,
+      "loss": 0.6582,
+      "step": 28600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0380313556653836e-09,
+      "loss": 0.5732,
+      "step": 28601
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0071499267368975e-09,
+      "loss": 0.6465,
+      "step": 28602
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.976387012949445e-09,
+      "loss": 0.6191,
+      "step": 28603
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.94574261466607e-09,
+      "loss": 0.6245,
+      "step": 28604
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.915216732252036e-09,
+      "loss": 0.614,
+      "step": 28605
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8848093660681655e-09,
+      "loss": 0.5709,
+      "step": 28606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.854520516474169e-09,
+      "loss": 0.6392,
+      "step": 28607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.824350183829761e-09,
+      "loss": 0.5395,
+      "step": 28608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.794298368494653e-09,
+      "loss": 0.675,
+      "step": 28609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.764365070821896e-09,
+      "loss": 0.692,
+      "step": 28610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7345502911678714e-09,
+      "loss": 0.6121,
+      "step": 28611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7048540298867396e-09,
+      "loss": 0.5862,
+      "step": 28612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6752762873304425e-09,
+      "loss": 0.6142,
+      "step": 28613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.64581706384759e-09,
+      "loss": 0.6963,
+      "step": 28614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.616476359790122e-09,
+      "loss": 0.5877,
+      "step": 28615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.587254175505539e-09,
+      "loss": 0.6714,
+      "step": 28616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.55815051133801e-09,
+      "loss": 0.6902,
+      "step": 28617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5291653676350348e-09,
+      "loss": 0.6493,
+      "step": 28618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.500298744739672e-09,
+      "loss": 0.6612,
+      "step": 28619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.47155064299276e-09,
+      "loss": 0.6021,
+      "step": 28620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4429210627373587e-09,
+      "loss": 0.6461,
+      "step": 28621
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.414410004310975e-09,
+      "loss": 0.5731,
+      "step": 28622
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3860174680533375e-09,
+      "loss": 0.5746,
+      "step": 28623
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3577434542997335e-09,
+      "loss": 0.6081,
+      "step": 28624
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.32958796338656e-09,
+      "loss": 0.623,
+      "step": 28625
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3015509956457746e-09,
+      "loss": 0.6713,
+      "step": 28626
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.273632551411554e-09,
+      "loss": 0.5934,
+      "step": 28627
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2458326310147447e-09,
+      "loss": 0.6637,
+      "step": 28628
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2181512347839726e-09,
+      "loss": 0.6706,
+      "step": 28629
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.190588363048974e-09,
+      "loss": 0.6112,
+      "step": 28630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1631440161350445e-09,
+      "loss": 0.5838,
+      "step": 28631
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1358181943674793e-09,
+      "loss": 0.6212,
+      "step": 28632
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1086108980715735e-09,
+      "loss": 0.6271,
+      "step": 28633
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.081522127569292e-09,
+      "loss": 0.6253,
+      "step": 28634
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.054551883181489e-09,
+      "loss": 0.6663,
+      "step": 28635
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0277001652279094e-09,
+      "loss": 0.6728,
+      "step": 28636
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0009669740271865e-09,
+      "loss": 0.6655,
+      "step": 28637
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9743523098968443e-09,
+      "loss": 0.608,
+      "step": 28638
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9478561731510758e-09,
+      "loss": 0.6169,
+      "step": 28639
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.921478564106295e-09,
+      "loss": 0.6687,
+      "step": 28640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8952194830722535e-09,
+      "loss": 0.6502,
+      "step": 28641
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.869078930362035e-09,
+      "loss": 0.646,
+      "step": 28642
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8430569062853908e-09,
+      "loss": 0.6432,
+      "step": 28643
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8171534111509635e-09,
+      "loss": 0.6218,
+      "step": 28644
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7913684452662848e-09,
+      "loss": 0.6199,
+      "step": 28645
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7657020089355556e-09,
+      "loss": 0.6383,
+      "step": 28646
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7401541024640875e-09,
+      "loss": 0.6576,
+      "step": 28647
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7147247261549714e-09,
+      "loss": 0.6248,
+      "step": 28648
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6894138803090773e-09,
+      "loss": 0.6772,
+      "step": 28649
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.664221565227276e-09,
+      "loss": 0.6666,
+      "step": 28650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6391477812082176e-09,
+      "loss": 0.6641,
+      "step": 28651
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6141925285483315e-09,
+      "loss": 0.6788,
+      "step": 28652
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5893558075440474e-09,
+      "loss": 0.6388,
+      "step": 28653
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5646376184895737e-09,
+      "loss": 0.6209,
+      "step": 28654
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.54003796167801e-09,
+      "loss": 0.6283,
+      "step": 28655
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.515556837400235e-09,
+      "loss": 0.6351,
+      "step": 28656
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4911942459493467e-09,
+      "loss": 0.5735,
+      "step": 28657
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4669501876106725e-09,
+      "loss": 0.5949,
+      "step": 28658
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4428246626739814e-09,
+      "loss": 0.6183,
+      "step": 28659
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4188176714246005e-09,
+      "loss": 0.5981,
+      "step": 28660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3949292141478564e-09,
+      "loss": 0.6508,
+      "step": 28661
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3711592911246363e-09,
+      "loss": 0.6371,
+      "step": 28662
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3475079026391566e-09,
+      "loss": 0.6085,
+      "step": 28663
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3239750489711943e-09,
+      "loss": 0.6187,
+      "step": 28664
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.300560730399415e-09,
+      "loss": 0.606,
+      "step": 28665
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2772649472024842e-09,
+      "loss": 0.6011,
+      "step": 28666
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.254087699654628e-09,
+      "loss": 0.6457,
+      "step": 28667
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.231028988032291e-09,
+      "loss": 0.6428,
+      "step": 28668
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2080888126085888e-09,
+      "loss": 0.6482,
+      "step": 28669
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.185267173654415e-09,
+      "loss": 0.6547,
+      "step": 28670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.162564071441775e-09,
+      "loss": 0.5887,
+      "step": 28671
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1399795062393424e-09,
+      "loss": 0.6694,
+      "step": 28672
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.117513478314681e-09,
+      "loss": 0.5973,
+      "step": 28673
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0951659879342447e-09,
+      "loss": 0.6609,
+      "step": 28674
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.072937035363376e-09,
+      "loss": 0.655,
+      "step": 28675
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0508266208651985e-09,
+      "loss": 0.6723,
+      "step": 28676
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0288347447017243e-09,
+      "loss": 0.6416,
+      "step": 28677
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.006961407133856e-09,
+      "loss": 0.6365,
+      "step": 28678
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.985206608421386e-09,
+      "loss": 0.6049,
+      "step": 28679
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.963570348821886e-09,
+      "loss": 0.6399,
+      "step": 28680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.942052628590707e-09,
+      "loss": 0.693,
+      "step": 28681
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.920653447985421e-09,
+      "loss": 0.6242,
+      "step": 28682
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8993728072580486e-09,
+      "loss": 0.6172,
+      "step": 28683
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.87821070666061e-09,
+      "loss": 0.5951,
+      "step": 28684
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8571671464451268e-09,
+      "loss": 0.6157,
+      "step": 28685
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8362421268613984e-09,
+      "loss": 0.6534,
+      "step": 28686
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8154356481558944e-09,
+      "loss": 0.6181,
+      "step": 28687
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7947477105761945e-09,
+      "loss": 0.6849,
+      "step": 28688
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7741783143676583e-09,
+      "loss": 0.6354,
+      "step": 28689
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7537274597745347e-09,
+      "loss": 0.6414,
+      "step": 28690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7333951470377419e-09,
+      "loss": 0.5934,
+      "step": 28691
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.713181376400419e-09,
+      "loss": 0.6266,
+      "step": 28692
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6930861481001537e-09,
+      "loss": 0.6248,
+      "step": 28693
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6731094623778644e-09,
+      "loss": 0.6811,
+      "step": 28694
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6532513194678079e-09,
+      "loss": 0.6344,
+      "step": 28695
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.633511719606462e-09,
+      "loss": 0.626,
+      "step": 28696
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6138906630280838e-09,
+      "loss": 0.6582,
+      "step": 28697
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5943881499647095e-09,
+      "loss": 0.5711,
+      "step": 28698
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5750041806483763e-09,
+      "loss": 0.6322,
+      "step": 28699
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5557387553077895e-09,
+      "loss": 0.601,
+      "step": 28700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.536591874172766e-09,
+      "loss": 0.6372,
+      "step": 28701
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5175635374686804e-09,
+      "loss": 0.6818,
+      "step": 28702
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4986537454231288e-09,
+      "loss": 0.5859,
+      "step": 28703
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4798624982581556e-09,
+      "loss": 0.6722,
+      "step": 28704
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.461189796198026e-09,
+      "loss": 0.5881,
+      "step": 28705
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4426356394636742e-09,
+      "loss": 0.6628,
+      "step": 28706
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4242000282760349e-09,
+      "loss": 0.6286,
+      "step": 28707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4058829628516012e-09,
+      "loss": 0.6182,
+      "step": 28708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3876844434090874e-09,
+      "loss": 0.6701,
+      "step": 28709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3696044701638766e-09,
+      "loss": 0.6355,
+      "step": 28710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.351643043330242e-09,
+      "loss": 0.6449,
+      "step": 28711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.333800163121346e-09,
+      "loss": 0.6292,
+      "step": 28712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3160758297481313e-09,
+      "loss": 0.6979,
+      "step": 28713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2984700434215403e-09,
+      "loss": 0.6725,
+      "step": 28714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2809828043502948e-09,
+      "loss": 0.6442,
+      "step": 28715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2636141127420066e-09,
+      "loss": 0.6123,
+      "step": 28716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2463639688009565e-09,
+      "loss": 0.5718,
+      "step": 28717
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2292323727336463e-09,
+      "loss": 0.6361,
+      "step": 28718
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2122193247410263e-09,
+      "loss": 0.6426,
+      "step": 28719
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1953248250273775e-09,
+      "loss": 0.6524,
+      "step": 28720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1785488737914297e-09,
+      "loss": 0.6431,
+      "step": 28721
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.161891471231913e-09,
+      "loss": 0.6966,
+      "step": 28722
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.145352617546447e-09,
+      "loss": 0.5618,
+      "step": 28723
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1289323129326513e-09,
+      "loss": 0.5374,
+      "step": 28724
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1126305575825946e-09,
+      "loss": 0.6334,
+      "step": 28725
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0964473516916762e-09,
+      "loss": 0.6645,
+      "step": 28726
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0803826954508544e-09,
+      "loss": 0.6324,
+      "step": 28727
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0644365890510878e-09,
+      "loss": 0.599,
+      "step": 28728
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0486090326811138e-09,
+      "loss": 0.7716,
+      "step": 28729
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0329000265285605e-09,
+      "loss": 0.6405,
+      "step": 28730
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0173095707799451e-09,
+      "loss": 0.5958,
+      "step": 28731
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.001837665619565e-09,
+      "loss": 0.6329,
+      "step": 28732
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.864843112317169e-10,
+      "loss": 0.642,
+      "step": 28733
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.712495077973671e-10,
+      "loss": 0.6313,
+      "step": 28734
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.561332554974824e-10,
+      "loss": 0.5864,
+      "step": 28735
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.411355545119182e-10,
+      "loss": 0.6124,
+      "step": 28736
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.26256405018311e-10,
+      "loss": 0.6473,
+      "step": 28737
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.114958071920754e-10,
+      "loss": 0.6214,
+      "step": 28738
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.968537612097372e-10,
+      "loss": 0.664,
+      "step": 28739
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.823302672433809e-10,
+      "loss": 0.6402,
+      "step": 28740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.679253254673114e-10,
+      "loss": 0.6387,
+      "step": 28741
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.536389360491726e-10,
+      "loss": 0.5808,
+      "step": 28742
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.394710991610489e-10,
+      "loss": 0.5806,
+      "step": 28743
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.25421814970584e-10,
+      "loss": 0.5862,
+      "step": 28744
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.114910836432011e-10,
+      "loss": 0.6365,
+      "step": 28745
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.976789053443235e-10,
+      "loss": 0.611,
+      "step": 28746
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.839852802382642e-10,
+      "loss": 0.6484,
+      "step": 28747
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.704102084871156e-10,
+      "loss": 0.6471,
+      "step": 28748
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.569536902518604e-10,
+      "loss": 0.6016,
+      "step": 28749
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.436157256912602e-10,
+      "loss": 0.6156,
+      "step": 28750
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.303963149651872e-10,
+      "loss": 0.6755,
+      "step": 28751
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.172954582290725e-10,
+      "loss": 0.6814,
+      "step": 28752
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.043131556383476e-10,
+      "loss": 0.5977,
+      "step": 28753
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.914494073484435e-10,
+      "loss": 0.5721,
+      "step": 28754
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.787042135092403e-10,
+      "loss": 0.6453,
+      "step": 28755
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.660775742739489e-10,
+      "loss": 0.6442,
+      "step": 28756
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.535694897913392e-10,
+      "loss": 0.6279,
+      "step": 28757
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.411799602090707e-10,
+      "loss": 0.6535,
+      "step": 28758
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.289089856759135e-10,
+      "loss": 0.6782,
+      "step": 28759
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.167565663361963e-10,
+      "loss": 0.5885,
+      "step": 28760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.047227023331381e-10,
+      "loss": 0.6498,
+      "step": 28761
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.928073938110679e-10,
+      "loss": 0.6681,
+      "step": 28762
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.810106409109839e-10,
+      "loss": 0.6478,
+      "step": 28763
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.693324437716641e-10,
+      "loss": 0.6136,
+      "step": 28764
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.577728025329965e-10,
+      "loss": 0.5998,
+      "step": 28765
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.463317173304284e-10,
+      "loss": 0.6426,
+      "step": 28766
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.350091883016273e-10,
+      "loss": 0.6322,
+      "step": 28767
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.238052155787099e-10,
+      "loss": 0.5805,
+      "step": 28768
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.127197992960132e-10,
+      "loss": 0.5958,
+      "step": 28769
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.017529395845433e-10,
+      "loss": 0.6706,
+      "step": 28770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.909046365730863e-10,
+      "loss": 0.5906,
+      "step": 28771
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.801748903926484e-10,
+      "loss": 0.6555,
+      "step": 28772
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.695637011686848e-10,
+      "loss": 0.6681,
+      "step": 28773
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.590710690277611e-10,
+      "loss": 0.6782,
+      "step": 28774
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.4869699409311187e-10,
+      "loss": 0.6125,
+      "step": 28775
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.384414764901923e-10,
+      "loss": 0.6571,
+      "step": 28776
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2830451633779637e-10,
+      "loss": 0.6149,
+      "step": 28777
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.182861137580485e-10,
+      "loss": 0.6139,
+      "step": 28778
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.083862688686324e-10,
+      "loss": 0.6196,
+      "step": 28779
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.986049817872317e-10,
+      "loss": 0.6527,
+      "step": 28780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8894225263041987e-10,
+      "loss": 0.6965,
+      "step": 28781
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.793980815125498e-10,
+      "loss": 0.5885,
+      "step": 28782
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.69972468545754e-10,
+      "loss": 0.6283,
+      "step": 28783
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.6066541384327524e-10,
+      "loss": 0.6303,
+      "step": 28784
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.5147691751391543e-10,
+      "loss": 0.6404,
+      "step": 28785
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.424069796675866e-10,
+      "loss": 0.6848,
+      "step": 28786
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.3345560041198046e-10,
+      "loss": 0.6464,
+      "step": 28787
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2462277985256807e-10,
+      "loss": 0.6492,
+      "step": 28788
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.159085180948207e-10,
+      "loss": 0.6338,
+      "step": 28789
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.073128152419891e-10,
+      "loss": 0.6349,
+      "step": 28790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.988356713951035e-10,
+      "loss": 0.6242,
+      "step": 28791
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9047708665519425e-10,
+      "loss": 0.6125,
+      "step": 28792
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.822370611210712e-10,
+      "loss": 0.5709,
+      "step": 28793
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7411559489043393e-10,
+      "loss": 0.667,
+      "step": 28794
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.661126880609821e-10,
+      "loss": 0.625,
+      "step": 28795
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.582283407259745e-10,
+      "loss": 0.6663,
+      "step": 28796
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5046255297977993e-10,
+      "loss": 0.6069,
+      "step": 28797
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4281532491343685e-10,
+      "loss": 0.6668,
+      "step": 28798
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3528665661909367e-10,
+      "loss": 0.6523,
+      "step": 28799
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2787654818445804e-10,
+      "loss": 0.6314,
+      "step": 28800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.205849996983478e-10,
+      "loss": 0.6234,
+      "step": 28801
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.134120112462501e-10,
+      "loss": 0.6408,
+      "step": 28802
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0635758291476238e-10,
+      "loss": 0.5751,
+      "step": 28803
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9942171478715134e-10,
+      "loss": 0.5748,
+      "step": 28804
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9260440694446325e-10,
+      "loss": 0.6396,
+      "step": 28805
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8590565946885465e-10,
+      "loss": 0.6516,
+      "step": 28806
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.793254724380411e-10,
+      "loss": 0.6594,
+      "step": 28807
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.728638459319587e-10,
+      "loss": 0.6384,
+      "step": 28808
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6652078002721284e-10,
+      "loss": 0.6512,
+      "step": 28809
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.602962747970782e-10,
+      "loss": 0.6262,
+      "step": 28810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5419033031705e-10,
+      "loss": 0.6943,
+      "step": 28811
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4820294665818247e-10,
+      "loss": 0.7248,
+      "step": 28812
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4233412389264012e-10,
+      "loss": 0.6348,
+      "step": 28813
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.36583862090367e-10,
+      "loss": 0.6072,
+      "step": 28814
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.309521613179765e-10,
+      "loss": 0.5981,
+      "step": 28815
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.254390216431922e-10,
+      "loss": 0.593,
+      "step": 28816
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2004444313151732e-10,
+      "loss": 0.638,
+      "step": 28817
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1476842584623448e-10,
+      "loss": 0.6279,
+      "step": 28818
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0961096985062647e-10,
+      "loss": 0.655,
+      "step": 28819
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0457207520575551e-10,
+      "loss": 0.5785,
+      "step": 28820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.965174197046346e-11,
+      "loss": 0.6772,
+      "step": 28821
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.48499702035921e-11,
+      "loss": 0.6137,
+      "step": 28822
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.016675996287305e-11,
+      "loss": 0.6692,
+      "step": 28823
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.560211130270724e-11,
+      "loss": 0.662,
+      "step": 28824
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.11560242786058e-11,
+      "loss": 0.6618,
+      "step": 28825
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.682849894163902e-11,
+      "loss": 0.6765,
+      "step": 28826
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.261953534398736e-11,
+      "loss": 0.6334,
+      "step": 28827
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.852913353450063e-11,
+      "loss": 0.6774,
+      "step": 28828
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.455729356424911e-11,
+      "loss": 0.6406,
+      "step": 28829
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.070401547764171e-11,
+      "loss": 0.6408,
+      "step": 28830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.696929932130779e-11,
+      "loss": 0.6537,
+      "step": 28831
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.3353145139656284e-11,
+      "loss": 0.6322,
+      "step": 28832
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.9855552975985875e-11,
+      "loss": 0.669,
+      "step": 28833
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.6476522870264605e-11,
+      "loss": 0.6104,
+      "step": 28834
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3216054864680946e-11,
+      "loss": 0.6696,
+      "step": 28835
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0074148995872254e-11,
+      "loss": 0.6365,
+      "step": 28836
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.705080530269634e-11,
+      "loss": 0.657,
+      "step": 28837
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.414602382068033e-11,
+      "loss": 0.66,
+      "step": 28838
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1359804583130924e-11,
+      "loss": 0.5967,
+      "step": 28839
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8692147623354817e-11,
+      "loss": 0.5877,
+      "step": 28840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.614305297465869e-11,
+      "loss": 0.6345,
+      "step": 28841
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3712520664798123e-11,
+      "loss": 0.6491,
+      "step": 28842
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.140055072485936e-11,
+      "loss": 0.6044,
+      "step": 28843
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9207143181487752e-11,
+      "loss": 0.608,
+      "step": 28844
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.713229805910821e-11,
+      "loss": 0.6408,
+      "step": 28845
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5176015384366083e-11,
+      "loss": 0.6499,
+      "step": 28846
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3338295180576055e-11,
+      "loss": 0.6278,
+      "step": 28847
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1619137467722142e-11,
+      "loss": 0.6448,
+      "step": 28848
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0018542268008801e-11,
+      "loss": 0.6321,
+      "step": 28849
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.536509598089383e-12,
+      "loss": 0.6547,
+      "step": 28850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.1730394790581194e-12,
+      "loss": 0.6632,
+      "step": 28851
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.928131923127467e-12,
+      "loss": 0.6338,
+      "step": 28852
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.801786948060994e-12,
+      "loss": 0.6432,
+      "step": 28853
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.794004566071152e-12,
+      "loss": 0.5524,
+      "step": 28854
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9047847893703963e-12,
+      "loss": 0.664,
+      "step": 28855
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1341276268405097e-12,
+      "loss": 0.5892,
+      "step": 28856
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4820330906939462e-12,
+      "loss": 0.6121,
+      "step": 28857
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.485011864818206e-13,
+      "loss": 0.6341,
+      "step": 28858
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.33531920865471e-13,
+      "loss": 0.6557,
+      "step": 28859
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.371252993960127e-13,
+      "loss": 0.5615,
+      "step": 28860
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 0.5751461982727051,
+      "eval_runtime": 4924.422,
+      "eval_samples_per_second": 112.021,
+      "eval_steps_per_second": 37.341,
+      "step": 28860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.928132540411468e-14,
+      "loss": 0.6164,
+      "step": 28861
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 0.599,
+      "step": 28862
+    },
+    {
+      "epoch": 1.0,
+      "step": 28862,
+      "total_flos": 1.3321809067702747e+19,
+      "train_loss": 0.6358378202117544,
+      "train_runtime": 173828.4743,
+      "train_samples_per_second": 23.91,
+      "train_steps_per_second": 0.166
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 28862,
+  "num_train_epochs": 1,
+  "save_steps": 5772,
+  "total_flos": 1.3321809067702747e+19,
+  "trial_name": null,
+  "trial_params": null
+}