Whisper Small Tr - CV 43h - Frozen Encoder
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: 0.2372
- Wer: 20.1528
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
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- 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 |
|---|---|---|---|---|
| 0.2134 | 0.37 | 500 | 0.2738 | 23.3480 |
| 0.1845 | 0.73 | 1000 | 0.2588 | 22.2679 |
| 0.1056 | 1.1 | 1500 | 0.2445 | 21.2688 |
| 0.1009 | 1.46 | 2000 | 0.2414 | 20.7152 |
| 0.0962 | 1.83 | 2500 | 0.2330 | 20.1222 |
| 0.0554 | 2.19 | 3000 | 0.2388 | 20.5230 |
| 0.0578 | 2.56 | 3500 | 0.2388 | 20.3253 |
| 0.0512 | 2.92 | 4000 | 0.2372 | 20.1528 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2