real-voices-800-10-epochs
This model is a fine-tuned version of openai/whisper-small on the real-voices-800 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2401
- Wer: 12.6441
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: 8
- eval_batch_size: 4
- seed: 42
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| No log | 0 | 0 | 0.6890 | 41.0325 |
| 0.2728 | 1.0 | 100 | 0.1977 | 14.0149 |
| 0.0829 | 2.0 | 200 | 0.1845 | 13.0028 |
| 0.0421 | 3.0 | 300 | 0.1917 | 12.7466 |
| 0.0233 | 4.0 | 400 | 0.1987 | 12.6313 |
| 0.0138 | 5.0 | 500 | 0.2179 | 12.5673 |
| 0.0074 | 6.0 | 600 | 0.2184 | 13.0028 |
| 0.0032 | 7.0 | 700 | 0.2318 | 12.9644 |
| 0.0018 | 8.0 | 800 | 0.2361 | 12.8363 |
| 0.0014 | 9.0 | 900 | 0.2390 | 12.7338 |
| 0.0013 | 10.0 | 1000 | 0.2401 | 12.6441 |
Framework versions
- Transformers 4.49.0
- Pytorch 2.5.1+cu121
- Datasets 3.4.1
- Tokenizers 0.21.0