kanishka/smolm-autoreg-bpe-counterfactual_babylm_indef_articles_with_pl_nouns_removal_new-1e-3

smolm-autoreg-bpe-counterfactual_babylm_aann_indef_articles_with_pl_nouns_removal_new-1e-3

This model was trained from scratch on the kanishka/counterfactual_babylm_aann_indef_articles_with_pl_nouns_removal_new dataset. It achieves the following results on the evaluation set:

• Loss: 3.4004
• Accuracy: 0.4117

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: 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
3.5965	1.0	18600	3.7932	0.3590
3.376	2.0	37200	3.5949	0.3809
3.247	3.0	55800	3.4625	0.3933
3.1633	4.0	74400	3.4094	0.3999
3.1084	5.0	93000	3.3589	0.4061
3.0663	6.0	111600	3.3638	0.4077
3.0305	7.0	130200	3.3580	0.4081
2.994	8.0	148800	3.3293	0.4100
2.9664	9.0	167400	3.3262	0.4114
2.942	10.0	186000	3.3377	0.4105
2.9136	11.0	204600	3.3401	0.4118
2.8886	12.0	223200	3.3339	0.4125
2.8701	13.0	241800	3.3341	0.4137
2.8515	14.0	260400	3.3494	0.4125
2.8292	15.0	279000	3.3648	0.4116
2.8094	16.0	297600	3.3643	0.4128
2.7851	17.0	316200	3.3658	0.4125
2.7685	18.0	334800	3.3846	0.4120
2.7454	19.0	353400	3.3961	0.4116
2.7269	20.0	372000	3.4004	0.4117

Framework versions

• Transformers 4.38.0
• Pytorch 2.3.1+cu121
• Datasets 2.16.1
• Tokenizers 0.15.2