WhisperTiny15hCommonVoice45hAugment
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: 0.8665
- Wer: 0.4280
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 |
|---|---|---|---|---|
| 1.0258 | 0.2233 | 500 | 1.0545 | 0.4943 |
| 0.8828 | 0.4466 | 1000 | 0.9798 | 0.4687 |
| 0.8023 | 0.6699 | 1500 | 0.9388 | 0.4608 |
| 0.714 | 0.8933 | 2000 | 0.9110 | 0.4420 |
| 0.5927 | 1.1166 | 2500 | 0.8940 | 0.4363 |
| 0.5974 | 1.3399 | 3000 | 0.8872 | 0.4352 |
| 0.5459 | 1.5632 | 3500 | 0.8764 | 0.4319 |
| 0.5245 | 1.7865 | 4000 | 0.8765 | 0.4279 |
| 0.5325 | 2.0098 | 4500 | 0.8709 | 0.4280 |
| 0.479 | 2.2331 | 5000 | 0.8665 | 0.4280 |
Framework versions
- Transformers 4.48.3
- Pytorch 2.6.0
- Datasets 3.3.0
- Tokenizers 0.21.0