learning_rate: 0.0001
train_batch_size: 256
eval_batch_size: 256
seed: 30
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
4.8295	1.0	495	3.7844
3.5371	2.0	990	3.4360
3.2797	3.0	1485	3.2933
3.1349	4.0	1980	3.2079
3.0332	5.0	2475	3.1528
2.9527	6.0	2970	3.1139
2.8841	7.0	3465	3.0916
2.824	8.0	3960	3.0747
2.7682	9.0	4455	3.0630
2.7163	10.0	4950	3.0596
2.6677	11.0	5445	3.0574
2.6216	12.0	5940	3.0626
2.576	13.0	6435	3.0706
2.5352	14.0	6930	3.0741
2.4975	15.0	7425	3.0858
2.4642	16.0	7920	3.0938
2.435	17.0	8415	3.1037
2.4088	18.0	8910	3.1112
2.3869	19.0	9405	3.1176
2.3697	20.0	9900	3.1222

fpadovani/cds_np_30

cds_np_30