WhisperTiny7.5hCommonVoice22.5hAugment
This model is a fine-tuned version of OpenAI/whisper-tiny on the audiofolder dataset. It achieves the following results on the evaluation set:
- Loss: 0.9727
- Wer: 0.4378
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: 1e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 5000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.9792 | 0.4488 | 500 | 1.0101 | 0.4996 |
| 0.7041 | 0.8977 | 1000 | 0.9352 | 0.4617 |
| 0.5032 | 1.3465 | 1500 | 0.9276 | 0.4443 |
| 0.4476 | 1.7953 | 2000 | 0.9268 | 0.4358 |
| 0.2871 | 2.2442 | 2500 | 0.9337 | 0.4326 |
| 0.2773 | 2.6930 | 3000 | 0.9462 | 0.4472 |
| 0.2105 | 3.1418 | 3500 | 0.9559 | 0.4416 |
| 0.1906 | 3.5907 | 4000 | 0.9647 | 0.4371 |
| 0.172 | 4.0395 | 4500 | 0.9694 | 0.4375 |
| 0.1635 | 4.4883 | 5000 | 0.9727 | 0.4378 |
Framework versions
- Transformers 4.48.3
- Pytorch 2.6.0
- Datasets 3.3.0
- Tokenizers 0.21.0