WhisperTiny22.5hCommonVoice7.5hAugment
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: 0.8516
- Wer: 0.4159
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 |
|---|---|---|---|---|
| 0.8335 | 0.4437 | 500 | 1.0273 | 0.5001 |
| 0.7144 | 0.8873 | 1000 | 0.9420 | 0.4956 |
| 0.582 | 1.3310 | 1500 | 0.9039 | 0.4563 |
| 0.5557 | 1.7746 | 2000 | 0.8743 | 0.4367 |
| 0.4823 | 2.2183 | 2500 | 0.8723 | 0.4329 |
| 0.4365 | 2.6619 | 3000 | 0.8576 | 0.4279 |
| 0.3805 | 3.1056 | 3500 | 0.8581 | 0.4205 |
| 0.3766 | 3.5492 | 4000 | 0.8519 | 0.4253 |
| 0.368 | 3.9929 | 4500 | 0.8507 | 0.4171 |
| 0.3373 | 4.4366 | 5000 | 0.8516 | 0.4159 |
Framework versions
- Transformers 4.48.3
- Pytorch 2.6.0
- Datasets 3.3.0
- Tokenizers 0.21.0