opt-babylm2-rewritten-clean-spacy_no-num-adj-earlystop-long-bpe_seed-211_1e-3

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

• Loss: 2.6443
• Accuracy: 0.4821

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
• 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
3.0912	1.0	17794	3.1539	0.4227
2.9731	2.0	35588	3.0552	0.4321
2.8935	3.0	53382	2.9870	0.4393
2.8624	4.0	71176	2.9591	0.4424
2.8334	5.0	88970	2.9394	0.4447
2.8155	6.0	106764	2.9166	0.4470
2.7945	7.0	124558	2.9014	0.4484
2.7812	8.0	142352	2.8882	0.4506
2.765	9.0	160146	2.8692	0.4528
2.7404	10.0	177940	2.8527	0.4546
2.7175	11.0	195734	2.8380	0.4560
2.7048	12.0	213528	2.8228	0.4579
2.68	13.0	231322	2.8028	0.4602
2.6498	14.0	249116	2.7846	0.4625
2.6202	15.0	266910	2.7650	0.4652
2.5883	16.0	284704	2.7395	0.4681
2.5396	17.0	302498	2.7147	0.4713
2.4923	18.0	320292	2.6864	0.4754
2.423	19.0	338086	2.6586	0.4793
2.3419	20.0	355880	2.6443	0.4821

Framework versions

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