learning_rate: 0.0001
train_batch_size: 256
eval_batch_size: 256
seed: 51
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
7.2649	1.0	399	6.5555
6.3042	2.0	798	6.1980
6.0384	3.0	1197	6.0187
5.8715	4.0	1596	5.8940
5.7374	5.0	1995	5.8007
5.6224	6.0	2394	5.7275
5.5213	7.0	2793	5.6709
5.4308	8.0	3192	5.6304
5.3485	9.0	3591	5.6014
5.2725	10.0	3990	5.5802
5.2018	11.0	4389	5.5682
5.1362	12.0	4788	5.5614
5.0739	13.0	5187	5.5564
5.0174	14.0	5586	5.5584
4.9679	15.0	5985	5.5618
4.9239	16.0	6384	5.5664
4.8867	17.0	6783	5.5708
4.8547	18.0	7182	5.5765
4.8286	19.0	7581	5.5805
4.8091	20.0	7980	5.5829

fpadovani/wiki_np_51

wiki_np_51