whisper-large-turkish-v5
This model is a fine-tuned version of openai/whisper-large-v3 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.3941
- Wer: 14.9232
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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Use OptimizerNames.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: 1000
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2081 | 0.7553 | 1000 | 0.2191 | 17.8157 |
| 0.1599 | 1.5106 | 2000 | 0.2153 | 16.5870 |
| 0.1025 | 2.2659 | 3000 | 0.2218 | 15.9967 |
| 0.0685 | 3.0211 | 4000 | 0.2349 | 15.6380 |
| 0.0557 | 3.7764 | 5000 | 0.2381 | 16.4826 |
| 0.026 | 4.5317 | 6000 | 0.2690 | 15.5286 |
| 0.0102 | 5.2870 | 7000 | 0.3062 | 15.2361 |
| 0.0053 | 6.0423 | 8000 | 0.3181 | 15.2513 |
| 0.0045 | 6.7976 | 9000 | 0.3256 | 15.6711 |
| 0.0018 | 7.5529 | 10000 | 0.3466 | 15.1267 |
| 0.001 | 8.3082 | 11000 | 0.3717 | 15.3938 |
| 0.0005 | 9.0634 | 12000 | 0.3777 | 15.0377 |
| 0.0006 | 9.8187 | 13000 | 0.3941 | 14.9232 |
Framework versions
- Transformers 4.51.3
- Pytorch 2.7.1+cu128
- Datasets 3.6.0
- Tokenizers 0.21.4