Whisper large V3 to CV Albanian
This model is a fine-tuned version of openai/whisper-large-v3 on the rishabhjain16/owr_cv_albanian default dataset. It achieves the following results on the evaluation set:
- Loss: 0.7039
- Wer: 29.7352
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: 16
- 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: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0341 | 9.0 | 500 | 0.5451 | 34.7454 |
| 0.0107 | 18.01 | 1000 | 0.5565 | 31.4868 |
| 0.0012 | 27.01 | 1500 | 0.5967 | 30.5092 |
| 0.0001 | 37.0 | 2000 | 0.6311 | 30.7128 |
| 0.0001 | 46.0 | 2500 | 0.6630 | 30.0611 |
| 0.0 | 55.01 | 3000 | 0.6787 | 29.8167 |
| 0.0 | 64.01 | 3500 | 0.6894 | 29.6538 |
| 0.0 | 74.0 | 4000 | 0.6970 | 29.6538 |
| 0.0 | 83.0 | 4500 | 0.7019 | 29.6945 |
| 0.0 | 92.01 | 5000 | 0.7039 | 29.7352 |
Framework versions
- Transformers 4.37.2
- Pytorch 1.14.0a0+44dac51
- Datasets 2.17.1
- Tokenizers 0.15.2