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
4.4886	1.0	495	3.2815
3.039	2.0	990	2.9778
2.8114	3.0	1485	2.8664
2.6915	4.0	1980	2.8042
2.608	5.0	2475	2.7660
2.5421	6.0	2970	2.7402
2.4856	7.0	3465	2.7220
2.435	8.0	3960	2.7119
2.3873	9.0	4455	2.7069
2.3425	10.0	4950	2.7067
2.2996	11.0	5445	2.7088
2.2582	12.0	5940	2.7140
2.219	13.0	6435	2.7213
2.1811	14.0	6930	2.7304
2.1448	15.0	7425	2.7409
2.1127	16.0	7920	2.7512
2.0845	17.0	8415	2.7606
2.0602	18.0	8910	2.7694
2.04	19.0	9405	2.7762
2.0238	20.0	9900	2.7801

fpadovani/cds_o_new_67

cds_o_new_67