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
5.7892	1.0	422	4.8905
4.76	2.0	844	4.7399
4.654	3.0	1266	4.6677
4.5896	4.0	1688	4.6221
4.5425	5.0	2110	4.5925
4.5023	6.0	2532	4.5712
4.4655	7.0	2954	4.5559
4.4306	8.0	3376	4.5437
4.3961	9.0	3798	4.5388
4.3614	10.0	4220	4.5365
4.3258	11.0	4642	4.5365
4.2901	12.0	5064	4.5417
4.2536	13.0	5486	4.5483
4.218	14.0	5908	4.5579
4.1829	15.0	6330	4.5683
4.1497	16.0	6752	4.5812
4.1196	17.0	7174	4.5936
4.0927	18.0	7596	4.6039
4.0703	19.0	8018	4.6130
4.0537	20.0	8440	4.6174

fpadovani/candor_sh1_67

candor_sh1_67