smolm-autoreg-bpe-counterfactual-babylm-only_indef_articles_with_pl_nouns_removal-3e-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: 3.4202
• Accuracy: 0.4089

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: 0.0003
• 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
3.7444	1.0	18600	3.9303	0.3452
3.4367	2.0	37200	3.6375	0.3746
3.29	3.0	55800	3.5155	0.3880
3.2081	4.0	74400	3.4751	0.3940
3.1438	5.0	93000	3.4190	0.3983
3.0947	6.0	111600	3.3905	0.4022
3.0569	7.0	130200	3.3832	0.4029
3.029	8.0	148800	3.3740	0.4051
2.9953	9.0	167400	3.3781	0.4059
2.9667	10.0	186000	3.3879	0.4069
2.9426	11.0	204600	3.3766	0.4071
2.9217	12.0	223200	3.3644	0.4085
2.8993	13.0	241800	3.3694	0.4082
2.8758	14.0	260400	3.3866	0.4088
2.8544	15.0	279000	3.3849	0.4087
2.8363	16.0	297600	3.4028	0.4086
2.8222	17.0	316200	3.3990	0.4089
2.8018	18.0	334800	3.3984	0.4096
2.7834	19.0	353400	3.4143	0.4091
2.7626	20.0	372000	3.4202	0.4089

Framework versions

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