whisper-small-taiwanese-asr-v2
This model is a fine-tuned version of openai/whisper-small on the gacky1601/Taiwanese_ASR dataset. It achieves the following results on the evaluation set:
- Loss: 0.2297
- Wer: 6.0497
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: 6
- eval_batch_size: 6
- 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 |
|---|---|---|---|---|
| 0.083 | 2.2676 | 1000 | 0.1953 | 8.1731 |
| 0.0444 | 4.5351 | 2000 | 0.1973 | 6.9444 |
| 0.0294 | 6.8027 | 3000 | 0.1984 | 6.5171 |
| 0.0334 | 9.0703 | 4000 | 0.2099 | 6.3034 |
| 0.0011 | 11.3379 | 5000 | 0.2229 | 6.3835 |
| 0.0001 | 13.6054 | 6000 | 0.2200 | 6.2099 |
| 0.0001 | 15.8730 | 7000 | 0.2297 | 6.0497 |
| 0.0001 | 18.1406 | 8000 | 0.2317 | 6.0764 |
| 0.0001 | 20.4082 | 9000 | 0.2375 | 6.3969 |
| 0.0 | 22.6757 | 10000 | 0.2379 | 6.4904 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.5.1+cu124
- Datasets 3.2.0
- Tokenizers 0.20.3