A Continuous-trained Whisper-large-v3-turbo Yue model #JL
This model is a fine-tuned version of openai/whisper-large-v3-turbo on the Common Voice 15.0 & 16.0 & 17.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2660
- Cer: 9.6002
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: 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: 250
- training_steps: 2000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 0.2893 | 0.2347 | 250 | 0.3542 | 14.3001 |
| 0.1931 | 0.4695 | 500 | 0.3283 | 13.7484 |
| 0.1426 | 0.7042 | 750 | 0.3136 | 13.2798 |
| 0.0875 | 0.9390 | 1000 | 0.3004 | 12.9006 |
| 0.0492 | 1.1737 | 1250 | 0.2745 | 10.9732 |
| 0.038 | 1.4085 | 1500 | 0.2749 | 10.4354 |
| 0.0216 | 1.6432 | 1750 | 0.2648 | 9.7879 |
| 0.0134 | 1.8779 | 2000 | 0.2660 | 9.6002 |
Framework versions
- Transformers 4.51.3
- Pytorch 2.7.0+cu126
- Datasets 3.5.1
- Tokenizers 0.21.1