Whisper Small Tr - CV 43h large batch
This model is a fine-tuned version of openai/whisper-small on the Common Voice 16.1 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2890
- Wer: 21.0603
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: 64
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.192 | 0.73 | 500 | 0.2638 | 22.1026 |
| 0.1238 | 1.46 | 1000 | 0.2492 | 21.2921 |
| 0.0663 | 2.19 | 1500 | 0.2483 | 20.7799 |
| 0.0656 | 2.92 | 2000 | 0.2445 | 20.3073 |
| 0.0391 | 3.65 | 2500 | 0.2575 | 21.1466 |
| 0.0203 | 4.38 | 3000 | 0.2744 | 20.9956 |
| 0.0125 | 5.11 | 3500 | 0.2841 | 20.9597 |
| 0.0096 | 5.84 | 4000 | 0.2890 | 21.0603 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2