binomial-babylm-base_seed-63_1e-3

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.0697
• Accuracy: 0.4152

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.001
• train_batch_size: 96
• eval_batch_size: 192
• seed: 63
• 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 results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.8571	1.0	608	3.7804	0.3353
3.3717	2.0	1216	3.3271	0.3820
3.2123	3.0	1824	3.2082	0.3947
3.1184	4.0	2432	3.1453	0.4013
3.0517	5.0	3040	3.1102	0.4053
2.9988	6.0	3648	3.0852	0.4083
2.9578	7.0	4256	3.0717	0.4102
2.9109	8.0	4864	3.0589	0.4118
2.8771	9.0	5472	3.0526	0.4132
2.844	10.0	6080	3.0507	0.4138
2.8073	11.0	6688	3.0494	0.4143
2.777	12.0	7296	3.0519	0.4148
2.7456	13.0	7904	3.0537	0.4149
2.7167	14.0	8512	3.0574	0.4151
2.6832	15.0	9120	3.0623	0.4155
2.6509	16.0	9728	3.0697	0.4152

Framework versions

• Transformers 4.57.3
• Pytorch 2.6.0+cu124
• Datasets 4.4.1
• Tokenizers 0.22.1