smolm-autoreg-bpe-counterfactual_babylm_indef_articles_with_pl_nouns_removal_new-seed_211-1e-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: 3.4117
• Accuracy: 0.4099

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
3.6063	1.0	18600	3.7970	0.3581
3.3804	2.0	37200	3.5813	0.3798
3.2565	3.0	55800	3.4691	0.3916
3.1729	4.0	74400	3.4057	0.3994
3.1189	5.0	93000	3.3898	0.4013
3.0726	6.0	111600	3.3687	0.4048
3.039	7.0	130200	3.3482	0.4067
3.0048	8.0	148800	3.3683	0.4065
2.9745	9.0	167400	3.3460	0.4084
2.9572	10.0	186000	3.3252	0.4113
2.9298	11.0	204600	3.3407	0.4103
2.9096	12.0	223200	3.3573	0.4100
2.8897	13.0	241800	3.3583	0.4096
2.8634	14.0	260400	3.3634	0.4100
2.849	15.0	279000	3.3750	0.4101
2.8253	16.0	297600	3.3725	0.4105
2.8153	17.0	316200	3.3877	0.4104
2.7909	18.0	334800	3.4038	0.4098
2.7719	19.0	353400	3.4056	0.4100
2.7505	20.0	372000	3.4117	0.4099

Framework versions

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