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
5.9705	0.9994	1486	4.4079	0.2941
4.3089	1.9992	2972	3.9092	0.3319
3.703	2.9991	4458	3.6332	0.3553
3.5333	3.9996	5945	3.4678	0.3710
3.3129	4.9994	7431	3.3714	0.3801
3.2404	5.9992	8917	3.3118	0.3860
3.1329	6.9991	10403	3.2677	0.3900
3.0918	7.9996	11890	3.2436	0.3927
3.0319	8.9994	13376	3.2273	0.3947
3.0061	9.9992	14862	3.2162	0.3960
2.9698	10.9991	16348	3.2088	0.3971
2.9485	11.9996	17835	3.2029	0.3982
2.9272	12.9994	19321	3.2025	0.3986
2.9081	13.9992	20807	3.1958	0.3994
2.8982	14.9991	22293	3.1934	0.3992
2.8793	15.9996	23780	3.1888	0.3998
2.8779	16.9994	25266	3.1852	0.4005
2.8592	17.9992	26752	3.1844	0.4004
2.8662	18.9991	28238	3.1828	0.4007
2.8451	19.9962	29720	3.1845	0.4009

qing-yao/swapped-datives_seed-63_1e-3

counterfactual_seed-63_1e-3