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: 1.2572
- Wer: 36.5103
- Cer: 24.3842
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 |
|---|---|---|---|---|---|
| 0.91 | 1.0 | 215 | 0.9039 | 45.0880 | 31.1133 |
| 0.5486 | 2.0 | 430 | 0.8982 | 84.3109 | 51.1766 |
| 0.284 | 3.0 | 645 | 0.9864 | 52.4194 | 33.9480 |
| 0.1557 | 4.0 | 860 | 1.0088 | 52.8592 | 33.5902 |
| 0.0909 | 5.0 | 1075 | 1.0402 | 41.7155 | 28.7601 |
| 0.0734 | 6.0 | 1290 | 1.0860 | 58.2845 | 37.2781 |
| 0.0444 | 7.0 | 1505 | 1.0962 | 51.2463 | 37.2643 |
| 0.0238 | 8.0 | 1720 | 1.1381 | 48.3871 | 32.9434 |
| 0.0213 | 9.0 | 1935 | 1.1168 | 35.4106 | 24.4117 |
| 0.0205 | 10.0 | 2150 | 1.1998 | 38.3431 | 25.5264 |
| 0.0097 | 11.0 | 2365 | 1.1861 | 36.1437 | 24.7970 |
| 0.0041 | 12.0 | 2580 | 1.2100 | 53.8123 | 35.1589 |
| 0.0033 | 13.0 | 2795 | 1.2308 | 37.6100 | 25.8291 |
| 0.004 | 14.0 | 3010 | 1.2381 | 37.0235 | 24.8245 |
| 0.0014 | 15.0 | 3225 | 1.2263 | 35.5572 | 23.1595 |
| 0.0009 | 16.0 | 3440 | 1.2465 | 35.0440 | 23.3246 |
| 0.0006 | 17.0 | 3655 | 1.2452 | 36.3636 | 24.8383 |
| 0.0007 | 18.0 | 3870 | 1.2525 | 35.5572 | 23.6411 |
| 0.0004 | 19.0 | 4085 | 1.2560 | 36.4370 | 24.3292 |
| 0.0004 | 20.0 | 4300 | 1.2572 | 36.5103 | 24.3842 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1