whisper-small-E30_freq_pause_speed
This model is a fine-tuned version of openai/whisper-small on the aihub old adult freq speed pause changed dataset. It achieves the following results on the evaluation set:
- Loss: 0.1861
- Cer: 4.9225
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use 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: 50
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 0.284 | 0.1289 | 100 | 0.2571 | 6.5731 |
| 0.1676 | 0.2579 | 200 | 0.2290 | 6.2911 |
| 0.1611 | 0.3868 | 300 | 0.2270 | 6.3264 |
| 0.1473 | 0.5158 | 400 | 0.2128 | 5.9915 |
| 0.1198 | 0.6447 | 500 | 0.2088 | 5.7155 |
| 0.1129 | 0.7737 | 600 | 0.2041 | 5.5686 |
| 0.1158 | 0.9026 | 700 | 0.2014 | 5.5275 |
| 0.0537 | 1.0309 | 800 | 0.1902 | 5.0341 |
| 0.049 | 1.1599 | 900 | 0.1930 | 5.0693 |
| 0.0447 | 1.2888 | 1000 | 0.1922 | 4.9812 |
| 0.0405 | 1.4178 | 1100 | 0.1909 | 5.0634 |
| 0.039 | 1.5467 | 1200 | 0.1893 | 4.9166 |
| 0.0403 | 1.6757 | 1300 | 0.1892 | 5.0341 |
| 0.0386 | 1.8046 | 1400 | 0.1871 | 5.1398 |
| 0.0405 | 1.9336 | 1500 | 0.1861 | 4.9225 |
Framework versions
- Transformers 4.47.0.dev0
- Pytorch 2.5.0+cu121
- Datasets 3.1.0
- Tokenizers 0.20.3