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: 1.2159
- Wer: 71.8475
- Cer: 60.7403
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 |
|---|---|---|---|---|---|
| 0.8706 | 1.0 | 215 | 0.8600 | 178.2258 | 102.3256 |
| 0.6025 | 2.0 | 430 | 0.8573 | 114.1496 | 71.0059 |
| 0.3962 | 3.0 | 645 | 0.8968 | 88.9296 | 67.1529 |
| 0.2239 | 4.0 | 860 | 0.9412 | 96.1877 | 60.4651 |
| 7.0487 | 5.0 | 1075 | 6.4281 | 779.6188 | 546.2915 |
| 5.2135 | 6.0 | 1290 | 4.5960 | 95.0147 | 76.2626 |
| 0.2569 | 7.0 | 1505 | 1.0109 | 228.7390 | 163.0246 |
| 0.2082 | 8.0 | 1720 | 1.0642 | 215.1026 | 163.8090 |
| 0.1511 | 9.0 | 1935 | 1.0990 | 162.3167 | 156.4882 |
| 0.0247 | 10.0 | 2150 | 1.1002 | 160.6305 | 182.1660 |
| 0.0155 | 11.0 | 2365 | 1.1253 | 59.9707 | 62.7081 |
| 0.007 | 12.0 | 2580 | 1.1525 | 134.8240 | 131.4298 |
| 0.0042 | 13.0 | 2795 | 1.1656 | 164.0762 | 154.8507 |
| 0.0024 | 14.0 | 3010 | 1.1838 | 118.9883 | 105.3530 |
| 0.0027 | 15.0 | 3225 | 1.1876 | 87.5367 | 71.3362 |
| 0.0015 | 16.0 | 3440 | 1.1978 | 57.2581 | 44.8190 |
| 0.0017 | 17.0 | 3655 | 1.1999 | 72.2874 | 60.8504 |
| 0.0012 | 18.0 | 3870 | 1.2119 | 71.8475 | 60.7541 |
| 0.0011 | 19.0 | 4085 | 1.2144 | 71.8475 | 60.7403 |
| 0.0011 | 20.0 | 4300 | 1.2159 | 71.8475 | 60.7403 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1