smolm-autoreg-bpe-counterfactual-babylm-only_indef_articles_with_pl_nouns_removal-1e-4

This model was trained from scratch on the kanishka/counterfactual-babylm-only_indef_articles_with_pl_nouns_removal dataset. It achieves the following results on the evaluation set:

• Loss: 3.4138
• Accuracy: 0.4080

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: 0.0001
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
4.0521	1.0	18600	4.2759	0.3096
3.567	2.0	37200	3.7516	0.3623
3.3864	3.0	55800	3.5931	0.3802
3.2901	4.0	74400	3.5232	0.3883
3.2176	5.0	93000	3.4594	0.3939
3.1641	6.0	111600	3.4612	0.3961
3.1229	7.0	130200	3.4155	0.4000
3.0932	8.0	148800	3.4064	0.4015
3.0577	9.0	167400	3.4074	0.4036
3.0285	10.0	186000	3.3945	0.4058
3.0042	11.0	204600	3.3962	0.4052
2.9833	12.0	223200	3.3878	0.4060
2.9614	13.0	241800	3.3943	0.4065
2.9382	14.0	260400	3.3899	0.4072
2.9179	15.0	279000	3.3926	0.4075
2.9009	16.0	297600	3.4043	0.4072
2.8878	17.0	316200	3.3955	0.4079
2.8705	18.0	334800	3.4079	0.4078
2.8533	19.0	353400	3.4119	0.4077
2.8352	20.0	372000	3.4138	0.4080

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1