openai/whisper-tiny
This model is a fine-tuned version of openai/whisper-tiny on the pphuc25/FranceMed dataset. It achieves the following results on the evaluation set:
- Loss: 1.9080
- Wer: 56.1584
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 |
|---|---|---|---|---|
| 1.3284 | 1.0 | 215 | 1.3160 | 94.5748 |
| 0.8758 | 2.0 | 430 | 1.3129 | 84.3842 |
| 0.5092 | 3.0 | 645 | 1.3872 | 57.4047 |
| 0.295 | 4.0 | 860 | 1.4638 | 70.3079 |
| 0.1646 | 5.0 | 1075 | 1.5513 | 69.0616 |
| 0.0939 | 6.0 | 1290 | 1.6665 | 60.6305 |
| 0.0585 | 7.0 | 1505 | 1.7200 | 54.8387 |
| 0.0411 | 8.0 | 1720 | 1.6980 | 54.1789 |
| 0.0295 | 9.0 | 1935 | 1.7985 | 56.0850 |
| 0.023 | 10.0 | 2150 | 1.8064 | 57.1114 |
| 0.0106 | 11.0 | 2365 | 1.8492 | 54.3255 |
| 0.011 | 12.0 | 2580 | 1.8559 | 57.2581 |
| 0.0087 | 13.0 | 2795 | 1.8624 | 59.5308 |
| 0.0069 | 14.0 | 3010 | 1.8581 | 57.1114 |
| 0.0028 | 15.0 | 3225 | 1.8776 | 58.1378 |
| 0.002 | 16.0 | 3440 | 1.8993 | 56.5982 |
| 0.0023 | 17.0 | 3655 | 1.8775 | 56.5249 |
| 0.0011 | 18.0 | 3870 | 1.9031 | 56.8915 |
| 0.0009 | 19.0 | 4085 | 1.9053 | 55.9384 |
| 0.0009 | 20.0 | 4300 | 1.9080 | 56.1584 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1