learning_rate: 5e-05
train_batch_size: 32
eval_batch_size: 32
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 64
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 500
training_steps: 17000
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
3.1841	0.0588	1000	3.0744
3.1802	0.1176	2000	3.0735
3.1664	0.1765	3000	3.0709
3.1513	0.2353	4000	3.0693
3.2179	0.2941	5000	3.0677
3.0193	0.3529	6000	3.0700
3.0308	0.4118	7000	3.0668
3.1744	0.4706	8000	3.0588
3.1756	0.5294	9000	3.0566
3.1679	0.5882	10000	3.0544
3.1626	0.6471	11000	3.0527
3.1515	0.0588	12000	3.0517
3.1470	0.1176	13000	3.0510
3.1325	0.1765	14000	3.0499
3.1195	0.2353	15000	3.0497
3.1889	0.2941	16000	3.0500
2.9972	0.3529	17000	3.0493