openai/whisper-base
This model is a fine-tuned version of openai/whisper-base on the pphuc25/FranceMed dataset. It achieves the following results on the evaluation set:
- Loss: 1.5990
- Wer: 44.8680
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 |
|---|---|---|---|---|
| 1.1328 | 1.0 | 215 | 1.0727 | 53.9589 |
| 0.7015 | 2.0 | 430 | 1.1185 | 59.4575 |
| 0.3727 | 3.0 | 645 | 1.1580 | 46.9208 |
| 0.1859 | 4.0 | 860 | 1.2757 | 63.0499 |
| 0.0996 | 5.0 | 1075 | 1.3674 | 43.6950 |
| 0.0619 | 6.0 | 1290 | 1.3916 | 48.2405 |
| 0.0459 | 7.0 | 1505 | 1.4305 | 44.8680 |
| 0.0342 | 8.0 | 1720 | 1.4555 | 47.1408 |
| 0.0323 | 9.0 | 1935 | 1.4792 | 45.0880 |
| 0.0141 | 10.0 | 2150 | 1.5414 | 48.8270 |
| 0.0109 | 11.0 | 2365 | 1.5575 | 44.5748 |
| 0.0064 | 12.0 | 2580 | 1.5499 | 45.5279 |
| 0.0061 | 13.0 | 2795 | 1.5439 | 45.6012 |
| 0.0081 | 14.0 | 3010 | 1.5679 | 46.6276 |
| 0.0013 | 15.0 | 3225 | 1.5714 | 46.9941 |
| 0.0028 | 16.0 | 3440 | 1.5777 | 44.6481 |
| 0.0019 | 17.0 | 3655 | 1.5904 | 45.7478 |
| 0.0006 | 18.0 | 3870 | 1.5931 | 43.9150 |
| 0.0004 | 19.0 | 4085 | 1.5981 | 44.5015 |
| 0.0004 | 20.0 | 4300 | 1.5990 | 44.8680 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1