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
5.9736	0.9994	1485	4.4156	0.2928
4.3066	1.9995	2971	3.9060	0.3332
3.6797	2.9996	4457	3.6355	0.3566
3.4925	3.9997	5943	3.4738	0.3709
3.2645	4.9997	7429	3.3761	0.3799
3.1915	5.9998	8915	3.3184	0.3856
3.0832	6.9999	10401	3.2787	0.3894
3.0425	8.0	11887	3.2506	0.3924
2.9847	8.9994	13372	3.2324	0.3942
2.9552	9.9995	14858	3.2234	0.3955
2.9204	10.9996	16344	3.2134	0.3966
2.8999	11.9997	17830	3.2045	0.3980
2.8789	12.9997	19316	3.2006	0.3985
2.8601	13.9998	20802	3.1995	0.3984
2.8494	14.9999	22288	3.1956	0.3990
2.8313	16.0	23774	3.1950	0.3997
2.8307	16.9994	25259	3.1953	0.3998
2.81	17.9995	26745	3.1945	0.4002
2.8189	18.9996	28231	3.1877	0.4007
2.7972	19.9882	29700	3.1956	0.4006

qing-yao/loose_balanced_cf_seed-63_1e-3

loose_balanced_cf_seed-63_1e-3