Whisper Small Co 5.1
This model is a fine-tuned version of openai/whisper-small on the Co audio dataset dataset. It achieves the following results on the evaluation set:
- Loss: 0.0052
- Cer Ortho: 0.0
- Cer: 0.0
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: 1e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 50
- training_steps: 800
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer Ortho | Cer |
|---|---|---|---|---|---|
| 1.847 | 1.2195 | 100 | 0.8785 | 40.6143 | 39.3836 |
| 0.4875 | 2.4390 | 200 | 0.3692 | 22.1843 | 21.2329 |
| 0.2828 | 3.6585 | 300 | 0.1340 | 8.0205 | 6.6781 |
| 0.1403 | 4.8780 | 400 | 0.0556 | 5.6314 | 5.6507 |
| 0.0644 | 6.0976 | 500 | 0.0239 | 2.5597 | 2.2260 |
| 0.0249 | 7.3171 | 600 | 0.0146 | 39.4198 | 39.5548 |
| 0.037 | 8.5366 | 700 | 0.0281 | 1.0239 | 0.8562 |
| 0.0131 | 9.7561 | 800 | 0.0052 | 0.0 | 0.0 |
Framework versions
- Transformers 4.51.1
- Pytorch 2.6.0+cu124
- Datasets 3.5.0
- Tokenizers 0.21.1