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
4.8361	1.0	495	3.7897
3.538	2.0	990	3.4327
3.2785	3.0	1485	3.2872
3.1342	4.0	1980	3.2071
3.032	5.0	2475	3.1511
2.9514	6.0	2970	3.1132
2.8836	7.0	3465	3.0901
2.8231	8.0	3960	3.0736
2.7678	9.0	4455	3.0633
2.7164	10.0	4950	3.0574
2.6678	11.0	5445	3.0589
2.6213	12.0	5940	3.0614
2.5765	13.0	6435	3.0655
2.5352	14.0	6930	3.0753
2.4983	15.0	7425	3.0819
2.4651	16.0	7920	3.0909
2.4353	17.0	8415	3.1027
2.4096	18.0	8910	3.1099
2.3879	19.0	9405	3.1155
2.3706	20.0	9900	3.1210

fpadovani/cds_np_67

cds_np_67