opt-babylm2-rewritten-clean-spacy-earlystop_hierarchical_211_age-origin_adj1-bpe_seed-211_1e-3

This model was trained from scratch on the kanishka/babylm2-rewritten-clean-spacy_hierarchical-adj_211_age-origin_adj1-ablation dataset. It achieves the following results on the evaluation set:

• Loss: 2.6825
• Accuracy: 0.4787

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
4.0937	1.0	2256	3.8119	0.3608
3.4513	2.0	4512	3.3015	0.4094
3.1289	3.0	6768	3.0841	0.4310
2.9222	4.0	9024	2.9813	0.4413
2.8416	5.0	11280	2.9218	0.4472
2.7864	6.0	13536	2.8815	0.4520
2.7435	7.0	15792	2.8545	0.4546
2.7114	8.0	18048	2.8351	0.4569
2.6873	9.0	20304	2.8203	0.4585
2.6676	10.0	22560	2.8122	0.4598
2.6483	11.0	24816	2.8032	0.4608
2.6488	12.0	27072	2.7977	0.4616
2.6363	13.0	29328	2.7915	0.4621
2.6275	14.0	31584	2.7895	0.4623
2.6115	15.0	33840	2.7641	0.4653
2.5674	16.0	36096	2.7405	0.4686
2.5189	17.0	38352	2.7183	0.4719
2.4623	18.0	40608	2.7001	0.4747
2.3973	19.0	42864	2.6871	0.4772
2.3245	20.0	45120	2.6825	0.4787

Framework versions

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