whispertiny_kokoro_25sil_v1_0
This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.3348
- Wer: 7.3443
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: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 10000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0104 | 0.2340 | 1000 | 0.2989 | 10.2020 |
| 0.0108 | 0.4681 | 2000 | 0.3863 | 8.2752 |
| 0.0091 | 0.7021 | 3000 | 0.3104 | 9.0291 |
| 0.0033 | 0.9361 | 4000 | 0.3200 | 8.4892 |
| 0.001 | 1.1701 | 5000 | 0.3119 | 8.1914 |
| 0.001 | 1.4042 | 6000 | 0.3804 | 9.8855 |
| 0.0029 | 1.6382 | 7000 | 0.3756 | 8.1728 |
| 0.0067 | 1.8722 | 8000 | 0.3012 | 7.5026 |
| 0.0 | 2.1062 | 9000 | 0.3309 | 7.4374 |
| 0.0 | 2.3403 | 10000 | 0.3348 | 7.3443 |
Framework versions
- Transformers 4.41.2
- Pytorch 2.3.0+cu121
- Datasets 4.0.0
- Tokenizers 0.19.1