Whisper Base Shona - S2S-M Curriculum 3000 steps
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: 0.4155
- Wer: 36.5641
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: 450
- training_steps: 4500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.8726 | 0.5995 | 500 | 0.6528 | 52.2635 |
| 0.3447 | 1.1990 | 1000 | 0.5072 | 43.9481 |
| 0.4441 | 1.7986 | 1500 | 0.4546 | 41.4442 |
| 0.2473 | 2.3981 | 2000 | 0.4379 | 40.0185 |
| 0.3866 | 2.9976 | 2500 | 0.4269 | 40.1108 |
| 0.2246 | 3.5971 | 3000 | 0.4198 | 37.4114 |
| 0.3989 | 4.1966 | 3500 | 0.4188 | 37.5608 |
| 0.2698 | 4.7962 | 4000 | 0.4135 | 37.9871 |
| 0.1525 | 5.3957 | 4500 | 0.4155 | 36.5641 |
Framework versions
- Transformers 5.14.1
- Pytorch 2.13.0+cu130
- Datasets 5.0.1
- Tokenizers 0.22.2