taskydata/deberta-v3-base_v_1

Hyperparameters:

• learning rate: 2e-5
• weight decay: 0.01
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• gradient_accumulation_steps:1
• eval steps: 50000
• max_length: 512
• num_epochs: 1
• hidden_dropout_prob: 0.3
• attention_probs_dropout_prob: 0.25

Dataset version:

• taskydata/tasky_or_not/v_1

Checkpoint:

• 455000 steps.

Results on Validation set:

Step	Training Loss	Validation Loss	Accuracy	Precision	Recall	F1
50000	0.0148	0.10890	0.9798	0.9755	0.9843	0.9799
100000	0.0121	0.09090	0.9863	0.9958	0.9767	0.9862
150000	0.0080	0.11800	0.9863	0.9779	0.9950	0.9864
200000	0.0116	0.08965	0.9877	0.9905	0.9848	0.9876
250000	0.0073	3.50100	0.6507	0.5905	0.9830	0.7378
300000	0.0072	0.09807	0.9850	0.9863	0.9870	0.9849
350000	0.0053	0.09830	0.9854	0.9939	0.9870	0.9852
400000	0.0046	0.08130	0.9893	0.9957	0.9828	0.9892
450000	0.0054	0.61280	0.9095	0.5835	0.9888	0.9162
455000	0.0055	0.15790	0.9710	0.9561	0.9874	0.9715

Uploaded Checkpoint:

• 400000