opt-babylm2-clean-spacy-earlystop_no-multi-adj-strict-bpe_seed-1024_1e-3

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

• Loss: 2.6983
• Accuracy: 0.4771

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
4.0297	1.0	2152	3.8585	0.3570
3.3906	2.0	4304	3.3435	0.4046
3.0817	3.0	6456	3.1232	0.4269
2.9257	4.0	8608	3.0129	0.4385
2.8295	5.0	10760	2.9414	0.4454
2.7596	6.0	12912	2.9054	0.4488
2.7142	7.0	15064	2.8777	0.4523
2.6808	8.0	17216	2.8535	0.4549
2.6562	9.0	19368	2.8391	0.4567
2.6344	10.0	21520	2.8294	0.4577
2.6175	11.0	23672	2.8214	0.4591
2.601	12.0	25824	2.8116	0.4598
2.5881	13.0	27976	2.8084	0.4603
2.5985	14.0	30128	2.8016	0.4608
2.5886	15.0	32280	2.7969	0.4613
2.5566	16.0	34432	2.7678	0.4649
2.5056	17.0	36584	2.7411	0.4693
2.4459	18.0	38736	2.7160	0.4727
2.3774	19.0	40888	2.7009	0.4754
2.2986	19.9911	43020	2.6983	0.4771

Framework versions

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