Whisper small model fine-tuned on moroccan darija
This model is a fine-tuned version of openai/whisper-small on the Dvoice darija v2.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.1732
- Wer: 14.1994
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 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
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1141 | 2.3419 | 1000 | 0.2908 | 29.8924 |
| 0.0116 | 4.6838 | 2000 | 0.1821 | 15.6051 |
| 0.0009 | 7.0258 | 3000 | 0.1739 | 13.9249 |
| 0.0004 | 9.3677 | 4000 | 0.1721 | 14.3532 |
| 0.0003 | 11.7096 | 5000 | 0.1732 | 14.1994 |
Framework versions
- Transformers 4.46.0
- Pytorch 2.5.0+cu121
- Datasets 3.0.2
- Tokenizers 0.20.1