Whisper Large Indonesian
This model is a fine-tuned version of openai/whisper-large on the mozilla-foundation/common_voice_11_0, magic_data, titml id dataset. It achieves the following results on the evaluation set:
- Loss: 0.2034
- Wer: 6.2483
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-06
- train_batch_size: 12
- eval_batch_size: 12
- 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 |
|---|---|---|---|---|
| 0.1516 | 0.5 | 1000 | 0.1730 | 6.5664 |
| 0.1081 | 1.0 | 2000 | 0.1638 | 6.3682 |
| 0.0715 | 1.49 | 3000 | 0.1803 | 6.2713 |
| 0.1009 | 1.99 | 4000 | 0.1796 | 6.2667 |
| 0.0387 | 2.49 | 5000 | 0.2054 | 6.4927 |
| 0.0494 | 2.99 | 6000 | 0.2034 | 6.2483 |
| 0.0259 | 3.48 | 7000 | 0.2226 | 6.3497 |
| 0.0265 | 3.98 | 8000 | 0.2274 | 6.4004 |
| 0.0232 | 4.48 | 9000 | 0.2443 | 6.5618 |
| 0.015 | 4.98 | 10000 | 0.2413 | 6.4927 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.13.0+cu117
- Datasets 2.7.1.dev0
- Tokenizers 0.13.2