learning_rate: 0.0001
train_batch_size: 256
eval_batch_size: 256
seed: 67
optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 20
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
6.9863	1.0	399	5.9816
5.6184	2.0	798	5.3611
5.168	3.0	1197	5.0639
4.9038	4.0	1596	4.8716
4.7112	5.0	1995	4.7365
4.5567	6.0	2394	4.6368
4.4271	7.0	2793	4.5616
4.3154	8.0	3192	4.5030
4.2166	9.0	3591	4.4596
4.1296	10.0	3990	4.4295
4.0523	11.0	4389	4.4067
3.9838	12.0	4788	4.3896
3.9219	13.0	5187	4.3781
3.8667	14.0	5586	4.3692
3.8168	15.0	5985	4.3677
3.7728	16.0	6384	4.3674
3.7344	17.0	6783	4.3653
3.7016	18.0	7182	4.3660
3.6741	19.0	7581	4.3677
3.6531	20.0	7980	4.3688

fpadovani/wiki_o_67

wiki_o_67