learning_rate: 0.0001
train_batch_size: 256
eval_batch_size: 256
seed: 42
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
7.3524	1.0	399	6.7968
6.563	2.0	798	6.5088
6.3591	3.0	1197	6.3721
6.2363	4.0	1596	6.2806
6.1395	5.0	1995	6.2137
6.0537	6.0	2394	6.1546
5.9739	7.0	2793	6.1077
5.8988	8.0	3192	6.0688
5.8274	9.0	3591	6.0405
5.7603	10.0	3990	6.0212
5.6962	11.0	4389	6.0040
5.636	12.0	4788	5.9978
5.5792	13.0	5187	5.9940
5.5268	14.0	5586	5.9937
5.4777	15.0	5985	5.9968
5.433	16.0	6384	6.0033
5.3945	17.0	6783	6.0066
5.3626	18.0	7182	6.0116
5.3378	19.0	7581	6.0162
5.3188	20.0	7980	6.0189

fpadovani/wiki_sh1_42

wiki_sh1_42