Edit model card

smolm-autoreg-bpe-counterfactual-babylm-no_prototypical-seed_211-3e-4

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 3.4074
  • Accuracy: 0.4086

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: 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
3.7439 1.0 18593 3.9121 0.3459
3.438 2.0 37186 3.6178 0.3756
3.2947 3.0 55779 3.4715 0.3901
3.2076 4.0 74372 3.4140 0.3965
3.1477 5.0 92965 3.3983 0.3996
3.1015 6.0 111558 3.3692 0.4021
3.0662 7.0 130151 3.3772 0.4036
3.0315 8.0 148744 3.3735 0.4036
3.0003 9.0 167337 3.3651 0.4057
2.9732 10.0 185930 3.3708 0.4063
2.9496 11.0 204523 3.3636 0.4073
2.9243 12.0 223116 3.3660 0.4085
2.9041 13.0 241709 3.3552 0.4089
2.8866 14.0 260302 3.3649 0.4087
2.8654 15.0 278895 3.3720 0.4086
2.846 16.0 297488 3.3842 0.4086
2.8252 17.0 316081 3.3945 0.4084
2.8084 18.0 334674 3.4002 0.4086
2.7871 19.0 353267 3.3996 0.4087
2.7718 20.0 371860 3.4074 0.4086

Framework versions

  • Transformers 4.35.0
  • Pytorch 2.1.0+cu121
  • Datasets 2.12.0
  • Tokenizers 0.14.1
Downloads last month
1
Safetensors
Model size
97.8M params
Tensor type
F32
·