whisper-large-v3-chichewa-base-only-fixed-pipeline
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: 1.7085
- Wer: 73.7697
- Cer: 33.9617
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 |
|---|---|---|---|---|---|
| 2.5898 | 8.9369 | 500 | 1.4170 | 94.3406 | 47.6542 |
| 1.6746 | 17.8649 | 1000 | 1.2662 | 72.5886 | 34.1932 |
| 1.3350 | 26.7928 | 1500 | 1.3411 | 77.0505 | 37.0600 |
| 1.0047 | 35.7207 | 2000 | 1.4462 | 71.8176 | 32.8964 |
| 0.8172 | 44.6486 | 2500 | 1.5379 | 73.7533 | 34.0288 |
| 0.7438 | 53.5766 | 3000 | 1.6087 | 72.3261 | 33.0493 |
| 0.6622 | 62.5045 | 3500 | 1.6565 | 72.9987 | 33.2345 |
| 0.6374 | 71.4324 | 4000 | 1.6817 | 74.7703 | 35.2283 |
| 0.6423 | 80.3604 | 4500 | 1.7085 | 73.7697 | 33.9617 |
Framework versions
- Transformers 5.17.0
- Pytorch 2.14.0+cu130
- Datasets 5.0.1
- Tokenizers 0.23.2