opt-babylm2-rewritten-clean-spacy-earlystop_hierarchical_211_size-origin_adj2-bpe_seed-211_1e-3

This model was trained from scratch on the kanishka/babylm2-rewritten-clean-spacy_hierarchical-adj_211_size-origin_adj2-ablation dataset. It achieves the following results on the evaluation set:

• Loss: 2.6834
• 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
4.096	1.0	2257	3.8270	0.3591
3.4539	2.0	4514	3.3130	0.4080
3.1275	3.0	6771	3.0925	0.4302
2.9224	4.0	9028	2.9899	0.4401
2.8428	5.0	11285	2.9290	0.4465
2.7856	6.0	13542	2.8866	0.4512
2.7408	7.0	15799	2.8610	0.4538
2.7112	8.0	18056	2.8421	0.4561
2.6889	9.0	20313	2.8289	0.4576
2.6668	10.0	22570	2.8148	0.4594
2.6493	11.0	24827	2.8119	0.4599
2.6476	12.0	27084	2.8025	0.4605
2.6366	13.0	29341	2.7954	0.4613
2.6278	14.0	31598	2.7896	0.4621
2.6116	15.0	33855	2.7663	0.4649
2.5666	16.0	36112	2.7427	0.4682
2.518	17.0	38369	2.7196	0.4715
2.462	18.0	40626	2.7015	0.4745
2.3972	19.0	42883	2.6858	0.4772
2.3252	19.9915	45120	2.6834	0.4787

Framework versions

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