opt-babylm2-rewritten-clean-spacy-earlystop_hierarchical_211_size-origin_adj2-bpe_seed-42_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: 2.6852
• Accuracy: 0.4784

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
4.0989	1.0	2257	3.8152	0.3604
3.4438	2.0	4514	3.3071	0.4084
3.1275	3.0	6771	3.0971	0.4290
2.9234	4.0	9028	2.9900	0.4403
2.8435	5.0	11285	2.9322	0.4458
2.7898	6.0	13542	2.8896	0.4508
2.7447	7.0	15799	2.8614	0.4538
2.7128	8.0	18056	2.8415	0.4557
2.6883	9.0	20313	2.8280	0.4577
2.6678	10.0	22570	2.8196	0.4584
2.6501	11.0	24827	2.8080	0.4598
2.6497	12.0	27084	2.8027	0.4607
2.6379	13.0	29341	2.7973	0.4608
2.6276	14.0	31598	2.7943	0.4613
2.6108	15.0	33855	2.7721	0.4642
2.5695	16.0	36112	2.7445	0.4679
2.5192	17.0	38369	2.7229	0.4714
2.4646	18.0	40626	2.7027	0.4744
2.3995	19.0	42883	2.6874	0.4769
2.3253	19.9915	45120	2.6852	0.4784

Framework versions

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