opt-babylm2-rewritten-clean-spacy-earlystop_hierarchica_size-color_adj2-bpe_seed-42_1e-3

This model was trained from scratch on the kanishka/babylm2-rewritten-clean-spacy_hierarchical-adj_size-color_adj2-ablation dataset. It achieves the following results on the evaluation set:

• Loss: 2.6851
• Accuracy: 0.4783

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
4.0969	1.0	2256	3.8070	0.3614
3.4446	2.0	4512	3.3003	0.4094
3.1276	3.0	6768	3.0918	0.4301
2.923	4.0	9024	2.9832	0.4408
2.8441	5.0	11280	2.9216	0.4471
2.7868	6.0	13536	2.8837	0.4515
2.7439	7.0	15792	2.8574	0.4543
2.7127	8.0	18048	2.8372	0.4567
2.6875	9.0	20304	2.8226	0.4584
2.6689	10.0	22560	2.8152	0.4595
2.6489	11.0	24816	2.8068	0.4603
2.6473	12.0	27072	2.8009	0.4610
2.6379	13.0	29328	2.7934	0.4619
2.6283	14.0	31584	2.7925	0.4621
2.6112	15.0	33840	2.7649	0.4653
2.568	16.0	36096	2.7397	0.4684
2.5179	17.0	38352	2.7208	0.4714
2.4626	18.0	40608	2.7015	0.4743
2.3968	19.0	42864	2.6875	0.4770
2.3253	19.9912	45100	2.6851	0.4783

Framework versions

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