Whisper Small Dv - Peter Gelderbloem
This model is a fine-tuned version of openai/whisper-small on the Common Voice 13 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2863
- Wer Ortho: 57.7129
- Wer: 11.2494
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: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 50
- training_steps: 4000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Wer Ortho |
|---|---|---|---|---|---|
| 0.1248 | 1.63 | 500 | 0.1684 | 12.9881 | 62.0447 |
| 0.0484 | 3.26 | 1000 | 0.1629 | 11.6493 | 58.6113 |
| 0.0315 | 4.89 | 1500 | 0.1878 | 11.7224 | 58.9386 |
| 0.0125 | 6.51 | 2000 | 0.2308 | 11.0895 | 57.2185 |
| 0.0058 | 8.14 | 2500 | 0.2671 | 11.0773 | 57.6224 |
| 0.0049 | 9.77 | 3000 | 0.2843 | 11.2564 | 57.6572 |
| 0.0033 | 11.4 | 3500 | 0.2845 | 11.0982 | 57.1558 |
| 0.0046 | 13.03 | 4000 | 0.2863 | 57.7129 | 11.2494 |
Framework versions
- Transformers 4.31.0.dev0
- Pytorch 1.12.1+cu116
- Datasets 2.4.0
- Tokenizers 0.12.1