smolm-autoreg-bpe-counterfactual-babylm-only_indef_articles_with_pl_nouns_removal-seed_211-1e-3

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: 3.3862
• Accuracy: 0.4127

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: 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
3.5992	1.0	18600	3.8124	0.3592
3.3826	2.0	37200	3.5570	0.3817
3.255	3.0	55800	3.4820	0.3917
3.1751	4.0	74400	3.4194	0.3988
3.1181	5.0	93000	3.3839	0.4022
3.074	6.0	111600	3.3598	0.4055
3.0387	7.0	130200	3.3320	0.4090
3.0113	8.0	148800	3.3243	0.4117
2.9786	9.0	167400	3.3343	0.4103
2.9522	10.0	186000	3.3475	0.4107
2.9315	11.0	204600	3.3211	0.4132
2.9096	12.0	223200	3.3419	0.4125
2.8879	13.0	241800	3.3351	0.4137
2.8675	14.0	260400	3.3329	0.4132
2.8497	15.0	279000	3.3544	0.4124
2.8277	16.0	297600	3.3686	0.4117
2.8093	17.0	316200	3.3650	0.4130
2.7915	18.0	334800	3.3731	0.4126
2.7731	19.0	353400	3.3832	0.4128
2.7504	20.0	372000	3.3862	0.4127

Framework versions

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