whisper-medium-chinese-tw-minnan-soap-demucs
This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.1568
- Wer: 171.4230
- Cer: 114.9043
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: 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: 500
- training_steps: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.0787 | 4.6512 | 1000 | 1.7614 | 167.7583 | 101.5008 |
| 0.0167 | 9.3023 | 2000 | 1.9160 | 247.4659 | 158.0093 |
| 0.007 | 13.9535 | 3000 | 2.0324 | 213.8402 | 152.7759 |
| 0.0023 | 18.6047 | 4000 | 2.0963 | 212.0663 | 141.3053 |
| 0.0014 | 23.2558 | 5000 | 2.1568 | 171.4230 | 114.9043 |
Framework versions
- Transformers 4.51.3
- Pytorch 2.6.0+cu124
- Datasets 3.5.0
- Tokenizers 0.21.1