babylm-base9m-gpt2

This model is a fine-tuned version of on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.8877
• Accuracy: 0.4846

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• 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: 182
• training_steps: 18200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
5.1725	0.0926	200	4.7183	0.3455
4.3947	0.1852	400	4.2783	0.3653
4.1458	0.2778	600	4.1072	0.3731
4.0969	0.3704	800	4.0114	0.3750
4.0248	0.4630	1000	3.9462	0.3779
3.9036	0.5556	1200	3.8976	0.3813
3.8369	0.6481	1400	3.8476	0.3838
3.8524	0.7407	1600	3.7954	0.3909
3.7137	0.8333	1800	3.7376	0.3964
3.6625	0.9259	2000	3.6851	0.4016
3.4353	1.8519	4000	3.3886	0.4282
3.159	2.7778	6000	3.2123	0.4455
2.9746	3.7037	8000	3.0870	0.4600
2.8211	4.6296	10000	2.9982	0.4708
2.7117	5.5556	12000	2.9491	0.4764
2.6895	6.4815	14000	2.9184	0.4807
2.6446	7.4074	16000	2.8997	0.4833
2.6014	8.3333	18000	2.8879	0.4845

Framework versions

• Transformers 4.50.3
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.4