opt-babylm2-rewritten-clean-spacy-earlystop-bpe_seed-211_1e-3

This model was trained from scratch on the kanishka/babylm2-rewritten-clean-spacy dataset. It achieves the following results on the evaluation set:

• Loss: 2.6837
• Accuracy: 0.4789

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: 32
• eval_batch_size: 64
• seed: 211
• 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 results

Training Loss	Epoch	Step	Validation Loss	Accuracy
4.1	1.0	2256	3.8208	0.3601
3.4569	2.0	4512	3.3085	0.4082
3.1304	3.0	6768	3.0990	0.4293
2.9241	4.0	9024	2.9892	0.4405
2.8456	5.0	11280	2.9303	0.4467
2.7882	6.0	13536	2.8887	0.4510
2.7426	7.0	15792	2.8618	0.4543
2.7124	8.0	18048	2.8404	0.4565
2.689	9.0	20304	2.8293	0.4580
2.6674	10.0	22560	2.8172	0.4592
2.651	11.0	24816	2.8067	0.4604
2.6483	12.0	27072	2.8001	0.4613
2.639	13.0	29328	2.7981	0.4619
2.6268	14.0	31584	2.7924	0.4620
2.6116	15.0	33840	2.7665	0.4651
2.5657	16.0	36096	2.7436	0.4687
2.5199	17.0	38352	2.7187	0.4717
2.4626	18.0	40608	2.7018	0.4746
2.3982	19.0	42864	2.6870	0.4773
2.3238	19.9915	45100	2.6837	0.4789

Framework versions

• Transformers 4.48.0
• Pytorch 2.5.1
• Datasets 3.2.0
• Tokenizers 0.21.0