learning_rate: 5e-06
train_batch_size: 4
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 20
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
4.6007	0.9479	50	4.4553
4.3901	1.8910	100	3.8501
3.9927	2.8341	150	3.3790
3.6562	3.7773	200	3.1353
3.4484	4.7204	250	2.9598
3.352	5.6635	300	2.8255
3.1997	6.6066	350	2.7154
3.0431	7.5498	400	2.6390
3.0088	8.4929	450	2.5868
2.9281	9.4360	500	2.5419
2.8857	10.3791	550	2.5115
2.8598	11.3223	600	2.4742
2.828	12.2654	650	2.4441
2.7331	13.2085	700	2.4207
2.7396	14.1517	750	2.4025
2.7336	15.0948	800	2.3858
2.7193	16.0379	850	2.3726
2.7096	16.9858	900	2.3626
2.6839	17.9289	950	2.3562
2.6633	18.8720	1000	2.3512
2.6655	19.8152	1050	2.3495

ramses64/t5-small-toinf

t5-small-toinf