learning_rate: 0.0001
train_batch_size: 256
eval_batch_size: 256
seed: 13
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
7.3315	1.0	399	6.7827
6.5649	2.0	798	6.5051
6.3596	3.0	1197	6.3747
6.2374	4.0	1596	6.2833
6.1394	5.0	1995	6.2151
6.0523	6.0	2394	6.1547
5.9717	7.0	2793	6.1051
5.8959	8.0	3192	6.0663
5.8244	9.0	3591	6.0372
5.7569	10.0	3990	6.0158
5.6931	11.0	4389	6.0062
5.6331	12.0	4788	5.9952
5.5764	13.0	5187	5.9912
5.5233	14.0	5586	5.9910
5.4747	15.0	5985	5.9950
5.4307	16.0	6384	6.0011
5.392	17.0	6783	6.0069
5.3601	18.0	7182	6.0111
5.3347	19.0	7581	6.0146
5.3164	20.0	7980	6.0169

fpadovani/wiki_sh1_13

wiki_sh1_13