whisper-small-coraa-l
This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.8425
- Wer: 0.3369
- Cer: 0.5704
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: 8
- eval_batch_size: 2
- 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: 100
- num_epochs: 12
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.6087 | 1.0 | 500 | 0.6197 | 0.3478 | 0.5709 |
| 0.192 | 2.0 | 1000 | 0.6408 | 0.3901 | 0.5815 |
| 0.088 | 3.0 | 1500 | 0.7006 | 0.3712 | 0.5775 |
| 0.0431 | 4.0 | 2000 | 0.7459 | 0.3737 | 0.5808 |
| 0.0099 | 5.0 | 2500 | 0.8209 | 0.3467 | 0.5728 |
| 0.004 | 6.0 | 3000 | 0.7809 | 0.3372 | 0.5670 |
| 0.0051 | 7.0 | 3500 | 0.7794 | 0.3423 | 0.5725 |
| 0.0005 | 8.0 | 4000 | 0.8009 | 0.3423 | 0.5709 |
| 0.0005 | 9.0 | 4500 | 0.7871 | 0.3365 | 0.5689 |
| 0.0003 | 10.0 | 5000 | 0.7922 | 0.3405 | 0.5715 |
| 0.0003 | 11.0 | 5500 | 0.8157 | 0.3347 | 0.5683 |
| 0.0003 | 12.0 | 6000 | 0.8297 | 0.3431 | 0.5696 |
Framework versions
- Transformers 4.55.2
- Pytorch 2.7.0+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4