Whisper medium to Kaggle Albanian
This model is a fine-tuned version of openai/whisper-medium on the rishabhjain16/kaggle_albanian default dataset. It achieves the following results on the evaluation set:
- Loss: 0.3393
- Wer: 26.91
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: 5500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Step | Validation Loss | Wer |
|---|---|---|---|
| 0.4716 | 500 | 0.5687 | 45.1093 |
| 0.3883 | 1000 | 0.4522 | 36.6402 |
| 0.3695 | 1500 | 0.4099 | 33.9459 |
| 0.3375 | 2000 | 0.3783 | 31.4906 |
| 0.2146 | 2500 | 0.3662 | 29.5184 |
| 0.2010 | 3000 | 0.3638 | 29.7474 |
| 0.1799 | 3500 | 0.3519 | 28.6020 |
| 0.1791 | 4000 | 0.3455 | 28.3978 |
| 0.0621 | 4500 | 0.3592 | 27.1452 |
| 0.0679 | 5000 | 0.3561 | 27.1079 |
| 0.0687 | 5500 | 0.3560 | 26.9186 |
Framework versions
- Transformers 4.37.2
- Pytorch 1.14.0a0+44dac51
- Datasets 2.17.1
- Tokenizers 0.15.2