whisper-large-v3-chichewa-mix-v1-bible12h-trigram6h-concat
This model is a fine-tuned version of openai/whisper-large-v3 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.3417
- Wer: 73.3432
- Cer: 34.6401
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: 32
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 7500
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.6345 | 2.1552 | 500 | 1.5368 | 89.6490 | 41.0708 |
| 1.1186 | 4.3103 | 1000 | 1.2744 | 78.7730 | 37.2893 |
| 0.9580 | 6.4655 | 1500 | 1.2375 | 74.6719 | 35.8536 |
| 0.9011 | 8.6207 | 2000 | 1.2448 | 76.8701 | 37.6876 |
| 0.8366 | 10.7759 | 2500 | 1.2641 | 74.2126 | 34.2557 |
| 0.6611 | 12.9310 | 3000 | 1.2792 | 74.1142 | 35.3626 |
| 0.7177 | 15.0862 | 3500 | 1.3055 | 73.2448 | 34.7467 |
| 0.6681 | 17.2414 | 4000 | 1.3215 | 74.8196 | 35.3997 |
| 0.6800 | 19.3966 | 4500 | 1.3296 | 74.3110 | 34.6749 |
| 0.6385 | 21.5517 | 5000 | 1.3377 | 73.3432 | 34.3206 |
| 0.6564 | 23.7069 | 5500 | 1.3416 | 74.9016 | 35.4553 |
| 0.6314 | 25.8621 | 6000 | 1.3417 | 73.3432 | 34.6401 |
Framework versions
- Transformers 5.17.0
- Pytorch 2.14.0+cu130
- Datasets 5.0.1
- Tokenizers 0.23.2