Whisper Small Vi HLV
This model is a fine-tuned version of openai/whisper-small on the Common Voice 17.0 vi + VIVOS dataset. It achieves the following results on the evaluation set:
- Loss: 0.4832
- Model Preparation Time: 0.0063
- Wer Ortho: 27.2121
- Wer: 20.8723
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: 5e-06
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- 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: 100
- training_steps: 1500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Wer Ortho | Wer |
|---|---|---|---|---|---|---|
| 0.3584 | 0.2784 | 250 | 0.4933 | 0.0063 | 28.0569 | 21.0503 |
| 0.3228 | 0.5568 | 500 | 0.4810 | 0.0063 | 27.2121 | 20.1157 |
| 0.3086 | 0.8352 | 750 | 0.4723 | 0.0063 | 27.7901 | 21.4508 |
| 0.2218 | 1.1136 | 1000 | 0.4816 | 0.0063 | 27.6123 | 21.4063 |
| 0.2419 | 1.3920 | 1250 | 0.4818 | 0.0063 | 27.2121 | 21.0503 |
| 0.2283 | 1.6704 | 1500 | 0.4832 | 0.0063 | 27.2121 | 20.8723 |
Framework versions
- Transformers 5.21.0
- Pytorch 2.11.0+cu130
- Datasets 4.8.5
- Tokenizers 0.23.3