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
4.5921	1.0	496	3.5129
3.3197	2.0	992	3.2969
3.1648	3.0	1488	3.2212
3.0772	4.0	1984	3.1722
3.0122	5.0	2480	3.1423
2.9591	6.0	2976	3.1217
2.9117	7.0	3472	3.1077
2.8674	8.0	3968	3.0977
2.8258	9.0	4464	3.0949
2.785	10.0	4960	3.0965
2.7468	11.0	5456	3.0959
2.7092	12.0	5952	3.1043
2.6716	13.0	6448	3.1143
2.6365	14.0	6944	3.1229
2.6025	15.0	7440	3.1356
2.5703	16.0	7936	3.1446
2.5419	17.0	8432	3.1555
2.518	18.0	8928	3.1644
2.4981	19.0	9424	3.1716
2.4828	20.0	9920	3.1760

fpadovani/cds_w_67

cds_w_67