opt-babylm2-rewritten-clean-spacy_no-num-adj-earlystop-bpe_seed-211_1e-3

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

• Loss: 2.6884
• 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: 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
4.0839	1.0	2225	3.8319	0.3594
3.4456	2.0	4450	3.3209	0.4073
3.1244	3.0	6675	3.1047	0.4285
2.9574	4.0	8900	3.0003	0.4397
2.8363	5.0	11125	2.9326	0.4464
2.7798	6.0	13350	2.8933	0.4502
2.7383	7.0	15575	2.8661	0.4534
2.706	8.0	17800	2.8462	0.4563
2.6843	9.0	20025	2.8286	0.4581
2.6625	10.0	22250	2.8218	0.4589
2.6482	11.0	24475	2.8130	0.4597
2.6327	12.0	26700	2.8075	0.4604
2.6196	13.0	28925	2.7995	0.4610
2.6254	14.0	31150	2.7951	0.4620
2.6119	15.0	33375	2.7756	0.4640
2.5659	16.0	35600	2.7508	0.4678
2.5146	17.0	37825	2.7269	0.4709
2.4598	18.0	40050	2.7063	0.4740
2.394	19.0	42275	2.6902	0.4770
2.3215	19.9913	44480	2.6884	0.4783

Framework versions

• Transformers 4.48.0
• Pytorch 2.5.1
• Datasets 3.2.0
• Tokenizers 0.21.0