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
5.2896	1.0	422	4.0651
3.7657	2.0	844	3.7301
3.5513	3.0	1266	3.6160
3.4403	4.0	1688	3.5491
3.3613	5.0	2110	3.5038
3.2972	6.0	2532	3.4731
3.2421	7.0	2954	3.4527
3.1919	8.0	3376	3.4370
3.1446	9.0	3798	3.4284
3.0995	10.0	4220	3.4255
3.0558	11.0	4642	3.4259
3.0135	12.0	5064	3.4277
2.9727	13.0	5486	3.4396
2.9336	14.0	5908	3.4437
2.8958	15.0	6330	3.4524
2.8612	16.0	6752	3.4632
2.831	17.0	7174	3.4734
2.8055	18.0	7596	3.4825
2.784	19.0	8018	3.4882
2.7674	20.0	8440	3.4925

fpadovani/candor_o_30

candor_o_30