opt-babylm2-rewritten-clean-spacy-earlystop_adj-num-strict-swapped-bpe_seed-1024_1e-3

This model was trained from scratch on the kanishka/babylm2-rewritten-clean-spacy_adj-num-strict-swapped dataset. It achieves the following results on the evaluation set:

• Loss: 2.6849
• 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: 1024
• 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
4.0966	1.0	2257	3.8173	0.3603
3.4547	2.0	4514	3.3083	0.4086
3.1304	3.0	6771	3.0937	0.4301
2.9228	4.0	9028	2.9881	0.4404
2.8448	5.0	11285	2.9293	0.4467
2.7889	6.0	13542	2.8876	0.4508
2.7438	7.0	15799	2.8585	0.4542
2.7126	8.0	18056	2.8432	0.4557
2.6902	9.0	20313	2.8259	0.4581
2.668	10.0	22570	2.8183	0.4588
2.6504	11.0	24827	2.8097	0.4597
2.6493	12.0	27084	2.8029	0.4608
2.6386	13.0	29341	2.7987	0.4612
2.6297	14.0	31598	2.7942	0.4619
2.6137	15.0	33855	2.7678	0.4650
2.5683	16.0	36112	2.7435	0.4682
2.5213	17.0	38369	2.7212	0.4714
2.4636	18.0	40626	2.7019	0.4746
2.4011	19.0	42883	2.6866	0.4771
2.3272	19.9914	45120	2.6849	0.4784

Framework versions

• Transformers 4.48.0
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.21.1