Edit model card

smolm-autoreg-bpe-counterfactual-babylm-indef_articles_with_pl_nouns-removal-3e-4

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

  • Loss: 3.4054
  • 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: 0.0003
  • 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
3.7446 1.0 18601 3.9130 0.3462
3.4295 2.0 37202 3.6268 0.3752
3.2878 3.0 55803 3.5073 0.3878
3.2031 4.0 74404 3.4630 0.3948
3.1443 5.0 93005 3.4077 0.3994
3.0973 6.0 111606 3.3724 0.4028
3.0617 7.0 130207 3.3562 0.4062
3.0252 8.0 148808 3.3648 0.4059
2.994 9.0 167409 3.3582 0.4071
2.9693 10.0 186010 3.3688 0.4075
2.9383 11.0 204611 3.3513 0.4092
2.9188 12.0 223212 3.3659 0.4086
2.8978 13.0 241813 3.3581 0.4097
2.8784 14.0 260414 3.3657 0.4103
2.8592 15.0 279015 3.3693 0.4102
2.8415 16.0 297616 3.3867 0.4092
2.8198 17.0 316217 3.3790 0.4101
2.8013 18.0 334818 3.3924 0.4099
2.7836 19.0 353419 3.4014 0.4098
2.7626 20.0 372020 3.4054 0.4099

Framework versions

  • Transformers 4.36.0
  • Pytorch 2.1.1+cu121
  • Datasets 2.15.0
  • Tokenizers 0.15.0
Downloads last month
2
Safetensors
Model size
97.8M params
Tensor type
F32
·

Dataset used to train kanishka/smolm-autoreg-bpe-counterfactual-babylm-indef_articles_with_pl_nouns-removal-3e-4

Evaluation results

  • Accuracy on kanishka/counterfactual_babylm_aann_indef_articles_with_pl_nouns_removal
    self-reported
    0.410