Whisper small medical
This model is a fine-tuned version of openai/whisper-small on the Medical ASR dataset. It achieves the following results on the evaluation set:
- Loss: 0.0773
- Wer: 10.8376
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: 32
- eval_batch_size: 8
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 50
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.3096 | 0.5917 | 100 | 0.2270 | 22.9544 |
| 0.1338 | 1.1834 | 200 | 0.1209 | 18.8269 |
| 0.0762 | 1.7751 | 300 | 0.0904 | 20.7338 |
| 0.0279 | 2.3669 | 400 | 0.0846 | 13.6133 |
| 0.0275 | 2.9586 | 500 | 0.0783 | 13.8788 |
| 0.0123 | 3.5503 | 600 | 0.0750 | 15.7374 |
| 0.0072 | 4.1420 | 700 | 0.0764 | 10.6203 |
| 0.0045 | 4.7337 | 800 | 0.0764 | 11.0065 |
| 0.0035 | 5.3254 | 900 | 0.0769 | 11.1272 |
| 0.0035 | 5.9172 | 1000 | 0.0773 | 10.8376 |
Framework versions
- Transformers 5.20.2
- Pytorch 2.11.0+cu130
- Datasets 5.1.0
- Tokenizers 0.23.3