Whisper Large Ja-Qve - cportoca
This model is a fine-tuned version of openai/whisper-large on the Quechua_dataset dataset. It achieves the following results on the evaluation set:
- Loss: 0.2409
- Wer: 17.7910
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 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: 500
- training_steps: 8000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2791 | 1.3550 | 1000 | 0.3439 | 32.1619 |
| 0.137 | 2.7100 | 2000 | 0.2366 | 26.9532 |
| 0.0305 | 4.0650 | 3000 | 0.2266 | 21.3367 |
| 0.0142 | 5.4201 | 4000 | 0.2322 | 18.5441 |
| 0.0048 | 6.7751 | 5000 | 0.2285 | 18.4500 |
| 0.0014 | 8.1301 | 6000 | 0.2378 | 18.1362 |
| 0.0007 | 9.4851 | 7000 | 0.2394 | 17.6969 |
| 0.0004 | 10.8401 | 8000 | 0.2409 | 17.7910 |
Framework versions
- Transformers 4.46.3
- Pytorch 2.4.1+cu121
- Datasets 3.1.0
- Tokenizers 0.20.3