opt-babylm2-rewritten-clean-spacy_random-removal-num-adj-earlystop-long-bpe_seed-211_1e-3

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

• Loss: 2.6408
• Accuracy: 0.4824

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
• 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
3.109	1.0	17925	3.1536	0.4224
2.982	2.0	35850	3.0501	0.4330
2.9011	3.0	53775	2.9833	0.4399
2.8706	4.0	71700	2.9563	0.4429
2.8484	5.0	89625	2.9311	0.4458
2.8309	6.0	107550	2.9147	0.4471
2.8091	7.0	125475	2.8978	0.4491
2.787	8.0	143400	2.8819	0.4511
2.7715	9.0	161325	2.8686	0.4524
2.7505	10.0	179250	2.8503	0.4544
2.7302	11.0	197175	2.8350	0.4562
2.7096	12.0	215100	2.8186	0.4581
2.6769	13.0	233025	2.7996	0.4604
2.66	14.0	250950	2.7811	0.4630
2.6287	15.0	268875	2.7616	0.4653
2.5969	16.0	286800	2.7369	0.4682
2.5511	17.0	304725	2.7122	0.4715
2.4985	18.0	322650	2.6844	0.4755
2.4266	19.0	340575	2.6560	0.4794
2.349	20.0	358500	2.6408	0.4824

Framework versions

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