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: 1.4500
- Wer: 39.8827
- Cer: 27.0125
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 |
|---|---|---|---|---|---|
| 1.0537 | 1.0 | 215 | 1.0242 | 68.4018 | 49.4840 |
| 0.6061 | 2.0 | 430 | 1.0639 | 53.3724 | 32.0077 |
| 0.3104 | 3.0 | 645 | 1.1618 | 54.0323 | 34.5259 |
| 0.1692 | 4.0 | 860 | 1.2151 | 42.4487 | 27.3290 |
| 0.1085 | 5.0 | 1075 | 1.3178 | 58.5044 | 37.6909 |
| 0.0834 | 6.0 | 1290 | 1.3759 | 40.0293 | 26.5171 |
| 0.0523 | 7.0 | 1505 | 1.4465 | 39.9560 | 26.5584 |
| 0.0448 | 8.0 | 1720 | 1.3777 | 43.0352 | 28.1547 |
| 0.0342 | 9.0 | 1935 | 1.4440 | 40.6891 | 28.8152 |
| 0.0292 | 10.0 | 2150 | 1.5108 | 43.3284 | 30.2876 |
| 0.0156 | 11.0 | 2365 | 1.5093 | 39.7361 | 27.3978 |
| 0.0173 | 12.0 | 2580 | 1.5143 | 42.5220 | 27.9070 |
| 0.0088 | 13.0 | 2795 | 1.5168 | 41.2757 | 28.4712 |
| 0.0045 | 14.0 | 3010 | 1.4740 | 38.7097 | 26.3245 |
| 0.003 | 15.0 | 3225 | 1.4854 | 40.0293 | 27.0951 |
| 0.0048 | 16.0 | 3440 | 1.4524 | 40.3226 | 26.9850 |
| 0.0022 | 17.0 | 3655 | 1.4445 | 40.6158 | 27.7694 |
| 0.0005 | 18.0 | 3870 | 1.4494 | 40.3226 | 27.3841 |
| 0.0002 | 19.0 | 4085 | 1.4495 | 41.0557 | 27.8932 |
| 0.0004 | 20.0 | 4300 | 1.4500 | 39.8827 | 27.0125 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1