opt-babylm2-rewritten-clean-spacy-earlystop_hierarchica_size-origin_adj2-bpe_seed-42_1e-3

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

• Loss: 2.6848
• 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
4.0985	1.0	2257	3.8203	0.3600
3.4507	2.0	4514	3.3068	0.4085
3.1258	3.0	6771	3.0913	0.4303
2.9242	4.0	9028	2.9886	0.4404
2.8458	5.0	11285	2.9309	0.4461
2.7883	6.0	13542	2.8888	0.4507
2.7421	7.0	15799	2.8617	0.4541
2.7126	8.0	18056	2.8390	0.4559
2.6885	9.0	20313	2.8266	0.4576
2.667	10.0	22570	2.8134	0.4593
2.6495	11.0	24827	2.8119	0.4598
2.648	12.0	27084	2.8014	0.4607
2.6378	13.0	29341	2.7971	0.4612
2.6289	14.0	31598	2.7918	0.4620
2.6094	15.0	33855	2.7683	0.4646
2.5676	16.0	36112	2.7461	0.4681
2.5182	17.0	38369	2.7220	0.4711
2.4637	18.0	40626	2.7030	0.4742
2.3991	19.0	42883	2.6885	0.4769
2.3251	19.9915	45120	2.6848	0.4783

Framework versions

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