Adding `safetensors` variant of this model

model-00001-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22b80e33d7c4cc3924fb0d348ee31209a713ab25c0f58fa21a95c23b28ee7a20
+size 994058288

model-00002-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e359e05bef7febb83bbc69214ecc0ba9185e24c6c3b36c6909a41422b5eb8f2b
+size 977278024

model-00003-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22258c02668d97019e03082ac098ab9a679dd6a5bc7c526b3a2c548502754c59
+size 69206392

model.safetensors.index.json

@@ -0,0 +1,90 @@
+{
+    "metadata": {
+        "total_size": 2040532992
+    },
+    "weight_map": {
+        "lm_head.weight": "model-00001-of-00003.safetensors",
+        "model.embed_tokens.weight": "model-00001-of-00003.safetensors",
+        "model.layers.0.channel_mixer.l1.weight": "model-00001-of-00003.safetensors",
+        "model.layers.0.channel_mixer.l2.weight": "model-00001-of-00003.safetensors",
+        "model.layers.0.channel_mixer.l3.weight": "model-00001-of-00003.safetensors",
+        "model.layers.0.token_mixer.lrpe.theta": "model-00001-of-00003.safetensors",
+        "model.layers.0.token_mixer.out_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.0.token_mixer.qkvu_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.1.channel_mixer.l1.weight": "model-00001-of-00003.safetensors",
+        "model.layers.1.channel_mixer.l2.weight": "model-00001-of-00003.safetensors",
+        "model.layers.1.channel_mixer.l3.weight": "model-00001-of-00003.safetensors",
+        "model.layers.1.token_mixer.out_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.1.token_mixer.qkvu_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.10.channel_mixer.l1.weight": "model-00002-of-00003.safetensors",
+        "model.layers.10.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.10.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.10.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.10.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.11.channel_mixer.l1.weight": "model-00002-of-00003.safetensors",
+        "model.layers.11.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.11.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.11.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.11.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.12.channel_mixer.l1.weight": "model-00002-of-00003.safetensors",
+        "model.layers.12.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.12.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.12.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.12.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.13.channel_mixer.l1.weight": "model-00002-of-00003.safetensors",
+        "model.layers.13.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.13.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.13.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.13.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.14.channel_mixer.l1.weight": "model-00002-of-00003.safetensors",
+        "model.layers.14.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.14.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.14.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.14.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.15.channel_mixer.l1.weight": "model-00003-of-00003.safetensors",
+        "model.layers.15.channel_mixer.l2.weight": "model-00003-of-00003.safetensors",
+        "model.layers.15.channel_mixer.l3.weight": "model-00003-of-00003.safetensors",
+        "model.layers.15.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.15.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.2.channel_mixer.l1.weight": "model-00001-of-00003.safetensors",
+        "model.layers.2.channel_mixer.l2.weight": "model-00001-of-00003.safetensors",
+        "model.layers.2.channel_mixer.l3.weight": "model-00001-of-00003.safetensors",
+        "model.layers.2.token_mixer.out_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.2.token_mixer.qkvu_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.3.channel_mixer.l1.weight": "model-00001-of-00003.safetensors",
+        "model.layers.3.channel_mixer.l2.weight": "model-00001-of-00003.safetensors",
+        "model.layers.3.channel_mixer.l3.weight": "model-00001-of-00003.safetensors",
+        "model.layers.3.token_mixer.out_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.3.token_mixer.qkvu_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.4.channel_mixer.l1.weight": "model-00001-of-00003.safetensors",
+        "model.layers.4.channel_mixer.l2.weight": "model-00001-of-00003.safetensors",
+        "model.layers.4.channel_mixer.l3.weight": "model-00001-of-00003.safetensors",
+        "model.layers.4.token_mixer.out_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.4.token_mixer.qkvu_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.5.channel_mixer.l1.weight": "model-00001-of-00003.safetensors",
+        "model.layers.5.channel_mixer.l2.weight": "model-00001-of-00003.safetensors",
+        "model.layers.5.channel_mixer.l3.weight": "model-00001-of-00003.safetensors",
+        "model.layers.5.token_mixer.out_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.5.token_mixer.qkvu_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.6.channel_mixer.l1.weight": "model-00001-of-00003.safetensors",
+        "model.layers.6.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.6.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.6.token_mixer.out_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.6.token_mixer.qkvu_proj.weight": "model-00001-of-00003.safetensors",
+        "model.layers.7.channel_mixer.l1.weight": "model-00002-of-00003.safetensors",
+        "model.layers.7.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.7.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.7.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.7.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.8.channel_mixer.l1.weight": "model-00002-of-00003.safetensors",
+        "model.layers.8.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.8.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.8.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.8.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.9.channel_mixer.l1.weight": "model-00002-of-00003.safetensors",
+        "model.layers.9.channel_mixer.l2.weight": "model-00002-of-00003.safetensors",
+        "model.layers.9.channel_mixer.l3.weight": "model-00002-of-00003.safetensors",
+        "model.layers.9.token_mixer.out_proj.weight": "model-00002-of-00003.safetensors",
+        "model.layers.9.token_mixer.qkvu_proj.weight": "model-00002-of-00003.safetensors"
+    }
+}