openai/whisper-small
This model is a fine-tuned version of openai/whisper-small on the pphuc25/FrenchMed dataset. It achieves the following results on the evaluation set:
- Loss: 1.3219
- Wer: 50.2199
- Cer: 33.0948
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: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 20
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.9164 | 1.0 | 215 | 0.8750 | 53.6657 | 38.7230 |
| 0.5 | 2.0 | 430 | 0.9152 | 52.4927 | 36.3837 |
| 0.1918 | 3.0 | 645 | 1.0102 | 54.8387 | 36.1222 |
| 0.0742 | 4.0 | 860 | 1.1064 | 54.4721 | 44.0209 |
| 0.0514 | 5.0 | 1075 | 1.1371 | 44.5748 | 32.1178 |
| 0.0395 | 6.0 | 1290 | 1.1607 | 57.3314 | 46.4703 |
| 0.0233 | 7.0 | 1505 | 1.1898 | 41.9355 | 28.8978 |
| 0.0181 | 8.0 | 1720 | 1.2267 | 39.2229 | 26.8474 |
| 0.017 | 9.0 | 1935 | 1.2296 | 39.1496 | 28.0033 |
| 0.0083 | 10.0 | 2150 | 1.2535 | 55.3519 | 36.4800 |
| 0.0013 | 11.0 | 2365 | 1.2872 | 51.6129 | 34.8287 |
| 0.0028 | 12.0 | 2580 | 1.2961 | 37.9765 | 27.2465 |
| 0.0005 | 13.0 | 2795 | 1.3004 | 50.2199 | 33.4251 |
| 0.0008 | 14.0 | 3010 | 1.3092 | 50.1466 | 33.2187 |
| 0.0004 | 15.0 | 3225 | 1.3065 | 51.6129 | 34.2645 |
| 0.0008 | 16.0 | 3440 | 1.3133 | 49.7801 | 32.5857 |
| 0.0005 | 17.0 | 3655 | 1.3164 | 50.2199 | 33.1499 |
| 0.0002 | 18.0 | 3870 | 1.3198 | 50.0 | 32.9847 |
| 0.0002 | 19.0 | 4085 | 1.3213 | 50.2933 | 33.0673 |
| 0.0002 | 20.0 | 4300 | 1.3219 | 50.2199 | 33.0948 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1