whisper-small-E10_Yspeed
This model is a fine-tuned version of openai/whisper-small on the aihub old50 adult freq speed pause changed dataset. It achieves the following results on the evaluation set:
- Loss: 0.2268
- Cer: 6.0150
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 |
|---|---|---|---|---|
| 0.2535 | 0.1289 | 100 | 0.2842 | 7.1194 |
| 0.1493 | 0.2579 | 200 | 0.2564 | 7.1135 |
| 0.1186 | 0.3868 | 300 | 0.2586 | 7.2133 |
| 0.1192 | 0.5158 | 400 | 0.2443 | 6.6436 |
| 0.1055 | 0.6447 | 500 | 0.2439 | 6.5907 |
| 0.0979 | 0.7737 | 600 | 0.2365 | 6.5907 |
| 0.0938 | 0.9026 | 700 | 0.2361 | 6.3851 |
| 0.0409 | 1.0309 | 800 | 0.2237 | 6.0973 |
| 0.0302 | 1.1599 | 900 | 0.2298 | 5.9915 |
| 0.0366 | 1.2888 | 1000 | 0.2287 | 6.0679 |
| 0.0335 | 1.4178 | 1100 | 0.2287 | 6.2148 |
| 0.0376 | 1.5467 | 1200 | 0.2299 | 6.1795 |
| 0.037 | 1.6757 | 1300 | 0.2284 | 6.1266 |
| 0.0326 | 1.8046 | 1400 | 0.2266 | 6.0738 |
| 0.0375 | 1.9336 | 1500 | 0.2268 | 6.0150 |
Framework versions
- Transformers 4.47.0.dev0
- Pytorch 2.5.0+cu121
- Datasets 3.1.0
- Tokenizers 0.20.3