Whisper Small - Egyptian Arabic
This model is a fine-tuned version of openai/whisper-small on the Egyptian Arabic Speech Recognition dataset. It achieves the following results on the evaluation set:
- Loss: 0.1510
- Wer: 8.3289
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
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1195 | 2.9326 | 1000 | 0.1951 | 19.3987 |
| 0.0139 | 5.8651 | 2000 | 0.1341 | 8.9514 |
| 0.0021 | 8.7977 | 3000 | 0.1413 | 8.7343 |
| 0.001 | 11.7302 | 4000 | 0.1489 | 8.1841 |
| 0.0005 | 14.6628 | 5000 | 0.1510 | 8.3289 |
Framework versions
- Transformers 4.47.0
- Pytorch 2.4.1.post303
- Datasets 3.2.0
- Tokenizers 0.21.0