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
4.3447	0.9479	50	3.7401
4.1121	1.8910	100	3.5651
3.846	2.8341	150	3.3381
3.6293	3.7773	200	3.1652
3.3962	4.7204	250	3.0156
3.3365	5.6635	300	2.8887
3.1526	6.6066	350	2.7780
3.0013	7.5498	400	2.7001
2.9622	8.4929	450	2.6280
2.8684	9.4360	500	2.5671
2.8213	10.3791	550	2.5154
2.7491	11.3223	600	2.4698
2.7324	12.2654	650	2.4360
2.6397	13.2085	700	2.4080
2.6333	14.1517	750	2.3831
2.6342	15.0948	800	2.3621
2.606	16.0379	850	2.3462
2.587	16.9858	900	2.3318
2.5463	17.9289	950	2.3220
2.5925	18.8720	1000	2.3165
2.5653	19.8152	1050	2.3142

ramses64/t5-small-f2inf

t5-small-f2inf