opt-babylm2-rewritten-clean-spacy-earlystop_hierarchical_211_age-origin_both-bpe_seed-211_1e-3

This model was trained from scratch on the kanishka/babylm2-rewritten-clean-spacy_hierarchical-adj_211_age-origin_both-ablation dataset. It achieves the following results on the evaluation set:

• Loss: 2.6835
• Accuracy: 0.4786

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
4.0958	1.0	2256	3.8124	0.3605
3.4517	2.0	4512	3.2981	0.4100
3.1265	3.0	6768	3.0846	0.4309
2.9213	4.0	9024	2.9795	0.4415
2.8426	5.0	11280	2.9195	0.4476
2.7855	6.0	13536	2.8805	0.4518
2.7414	7.0	15792	2.8528	0.4549
2.7128	8.0	18048	2.8364	0.4567
2.6875	9.0	20304	2.8219	0.4585
2.669	10.0	22560	2.8101	0.4600
2.6499	11.0	24816	2.8022	0.4606
2.6491	12.0	27072	2.7955	0.4617
2.637	13.0	29328	2.7922	0.4619
2.6284	14.0	31584	2.7894	0.4623
2.6114	15.0	33840	2.7655	0.4651
2.5688	16.0	36096	2.7414	0.4685
2.5189	17.0	38352	2.7192	0.4718
2.4626	18.0	40608	2.7007	0.4747
2.3995	19.0	42864	2.6868	0.4772
2.3244	20.0	45120	2.6835	0.4786

Framework versions

• Transformers 4.48.0
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.21.4