learning_rate: 0.001
train_batch_size: 32
eval_batch_size: 64
seed: 63
gradient_accumulation_steps: 8
total_train_batch_size: 256
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_steps: 32000
num_epochs: 20.0
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Accuracy
6.1397	0.9998	1495	4.6645	0.2652
4.6172	1.9997	2990	4.1987	0.3050
4.0314	2.9995	4485	3.9518	0.3255
3.8633	4.0	5981	3.7929	0.3399
3.637	4.9998	7476	3.6933	0.3499
3.561	5.9997	8971	3.6266	0.3569
3.4427	6.9995	10466	3.5855	0.3608
3.4054	8.0	11962	3.5528	0.3644
3.3332	8.9998	13457	3.5351	0.3667
3.3122	9.9997	14952	3.5174	0.3686
3.2653	10.9995	16447	3.5048	0.3702
3.2514	12.0	17943	3.5013	0.3711
3.2181	12.9998	19438	3.4940	0.3720
3.2097	13.9997	20933	3.4829	0.3734
3.1862	14.9995	22428	3.4822	0.3737
3.1818	16.0	23924	3.4781	0.3741
3.1633	16.9998	25419	3.4738	0.3746
3.1605	17.9997	26914	3.4749	0.3752
3.1498	18.9995	28409	3.4698	0.3752
3.1478	19.9967	29900	3.4683	0.3757

qing-yao/random-first_seed-63_1e-3

random_first_noditransitive_seed-63_1e-3