test_seed-42_1e-3

This model was trained from scratch on the qing-yao/slightly-cleaner-babylm dataset. It achieves the following results on the evaluation set:

• Loss: 3.0025
• Accuracy: 0.4223

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: 42
• 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
6.1409	1.0	1824	4.2361	0.3072
4.0561	2.0	3648	3.7116	0.3490
3.6138	3.0	5472	3.4533	0.3723
3.3863	4.0	7296	3.3195	0.3850
3.2557	5.0	9120	3.2404	0.3930
3.188	6.0	10944	3.1924	0.3973
3.1185	7.0	12768	3.1594	0.4009
3.0753	8.0	14592	3.1404	0.4029
3.047	9.0	16416	3.1230	0.4046
3.0232	10.0	18240	3.1120	0.4060
3.008	11.0	20064	3.1057	0.4074
2.9609	12.0	21888	3.1000	0.4079
2.954	13.0	23712	3.0922	0.4087
2.953	14.0	25536	3.0897	0.4089
2.952	15.0	27360	3.0875	0.4093
2.9527	16.0	29184	3.0876	0.4098
2.9549	17.0	31008	3.0856	0.4099
2.9073	18.0	32832	3.0625	0.4127
2.8458	19.0	34656	3.0216	0.4183
2.7073	19.9894	36460	3.0025	0.4223

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.1