learning_rate: 0.001
train_batch_size: 96
eval_batch_size: 192
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 768
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: 500
num_epochs: 20.0

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.7837	1.0	608	3.6909	0.3447
3.3554	2.0	1216	3.3141	0.3835
3.2025	3.0	1824	3.1991	0.3957
3.1089	4.0	2432	3.1383	0.4021
3.0474	5.0	3040	3.1052	0.4062
2.9929	6.0	3648	3.0808	0.4090
2.9492	7.0	4256	3.0649	0.4108
2.9117	8.0	4864	3.0565	0.4123
2.8725	9.0	5472	3.0508	0.4133
2.8387	10.0	6080	3.0473	0.4144
2.8048	11.0	6688	3.0454	0.4151
2.7698	12.0	7296	3.0486	0.4152
2.7444	13.0	7904	3.0477	0.4161
2.7103	14.0	8512	3.0555	0.4154
2.6783	15.0	9120	3.0599	0.4159
2.645	16.0	9728	3.0685	0.4156

qing-yao/binomial-babylm-base_seed-42_1e-3

binomial-babylm-base_seed-42_1e-3