Whisper Medium Shona - Cleaned Data Duration SortaGrad
This model is a fine-tuned version of openai/whisper-base on the Cleaned Google WAXAL Shona dataset. It achieves the following results on the evaluation set:
- Loss: 0.4571
- Wer: 40.7137
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_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: 200
- training_steps: 2000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.8798 | 0.2398 | 200 | 0.9193 | 63.5934 |
| 0.6167 | 0.4796 | 400 | 0.6583 | 50.7509 |
| 0.5866 | 0.7194 | 600 | 0.5679 | 49.3659 |
| 0.5302 | 0.9592 | 800 | 0.5203 | 45.3494 |
| 0.4440 | 1.1990 | 1000 | 0.4967 | 43.8557 |
| 0.4213 | 1.4388 | 1200 | 0.4814 | 42.0742 |
| 0.4077 | 1.6787 | 1400 | 0.4709 | 42.3322 |
| 0.4016 | 1.9185 | 1600 | 0.4619 | 41.7158 |
| 0.3498 | 2.1583 | 1800 | 0.4586 | 40.7164 |
| 0.3882 | 2.3981 | 2000 | 0.4571 | 40.7137 |
Framework versions
- Transformers 5.13.1
- Pytorch 2.12.0+cu130
- Datasets 5.0.0
- Tokenizers 0.22.2