Whisper Small ko-Yfreq - syp1229
This model is a fine-tuned version of openai/whisper-small on the aihub Y dialogue dataset. It achieves the following results on the evaluation set:
- Loss: 4.5884
- Cer: 0.8501
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: 0.002
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 5.7567 | 0.3 | 100 | 6.1420 | 0.9849 |
| 5.5645 | 0.59 | 200 | 5.7215 | 0.9903 |
| 5.1849 | 0.89 | 300 | 5.3951 | 0.9170 |
| 5.0416 | 1.18 | 400 | 5.3100 | 0.8497 |
| 4.5767 | 1.48 | 500 | 5.0656 | 0.9404 |
| 4.6818 | 1.78 | 600 | 5.0057 | 0.8237 |
| 4.4227 | 2.07 | 700 | 4.9162 | 0.8776 |
| 4.3057 | 2.37 | 800 | 4.8226 | 0.8212 |
| 4.3631 | 2.66 | 900 | 4.7700 | 0.8636 |
| 4.1999 | 2.96 | 1000 | 4.6969 | 0.8476 |
| 4.0789 | 3.25 | 1100 | 4.7653 | 0.9222 |
| 4.1997 | 3.55 | 1200 | 4.6345 | 0.8642 |
| 3.8299 | 3.85 | 1300 | 4.6444 | 0.8282 |
| 3.6904 | 4.14 | 1400 | 4.6633 | 0.8451 |
| 3.7531 | 4.44 | 1500 | 4.6527 | 0.9382 |
| 3.8045 | 4.73 | 1600 | 4.5884 | 0.8501 |
Framework versions
- Transformers 4.34.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.14.5
- Tokenizers 0.13.3