babylm-base1m-gpt2

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

• Loss: 3.9386
• Accuracy: 0.3859

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: 198
• training_steps: 19800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
6.1154	0.1033	200	5.0417	0.2987
5.2001	0.2066	400	4.6315	0.3037
4.9324	0.3099	600	4.4495	0.3151
4.6366	0.4132	800	4.3532	0.3213
4.3878	0.5165	1000	4.3022	0.3213
4.3467	0.6198	1200	4.2962	0.3158
4.1575	0.7231	1400	4.2556	0.3160
4.1211	0.8264	1600	4.1590	0.3356
4.0085	0.9298	1800	4.1952	0.3251
3.8398	1.0331	2000	4.1387	0.3313
2.5393	2.0661	4000	3.9741	0.3627
2.1302	3.0992	6000	3.8979	0.3778
1.8989	4.1322	8000	3.7969	0.3872
1.7936	5.1653	10000	3.8104	0.3886
1.6629	6.1983	12000	3.8619	0.3863
1.5851	7.2314	14000	3.8960	0.3865
1.4878	8.2645	16000	3.9092	0.3867
1.4398	9.2975	18000	3.9346	0.3855

Framework versions

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