whisper-small-v4
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: 0.2998
- Wer: 37.0027
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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- 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: 5
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.3985 | 0.2458 | 5000 | 0.4006 | 100 |
| 0.3535 | 0.4917 | 10000 | 0.3515 | 41.9359 |
| 0.3329 | 0.7375 | 15000 | 0.3267 | 41.2601 |
| 0.3088 | 0.9834 | 20000 | 0.3122 | 40.8518 |
| 0.2702 | 1.2292 | 25000 | 0.3052 | 39.1540 |
| 0.2682 | 1.4750 | 30000 | 0.2963 | 39.5239 |
| 0.2606 | 1.7208 | 35000 | 0.2891 | 38.2537 |
| 0.2622 | 1.9667 | 40000 | 0.2837 | 38.0593 |
| 0.2164 | 2.2125 | 45000 | 0.2854 | 37.4512 |
| 0.222 | 2.4583 | 50000 | 0.2817 | 37.9314 |
| 0.2233 | 2.7042 | 55000 | 0.2788 | 37.6204 |
| 0.2218 | 2.9500 | 60000 | 0.2755 | 36.9669 |
| 0.1752 | 3.1958 | 65000 | 0.2871 | 36.6628 |
| 0.1802 | 3.4417 | 70000 | 0.2860 | 36.9139 |
| 0.1864 | 3.6875 | 75000 | 0.2818 | 37.8713 |
| 0.1832 | 3.9334 | 80000 | 0.2804 | 38.1352 |
| 0.1396 | 4.1792 | 85000 | 0.3042 | 37.6950 |
| 0.1398 | 4.4250 | 90000 | 0.3042 | 37.7528 |
| 0.1482 | 4.6708 | 95000 | 0.2998 | 37.0027 |
Framework versions
- Transformers 4.48.0.dev0
- Pytorch 2.5.1+cu124
- Datasets 3.1.0
- Tokenizers 0.21.0