Rangsitra Whisper
This model is a fine-tuned version of openai/whisper-medium on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:
- Cer: 4.0149
- Wer: 14.0500
- Loss: 0.1453
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: 16
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: warmup_stable_decay
- lr_scheduler_warmup_steps: 1025
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Cer | Wer | Validation Loss |
|---|---|---|---|---|---|
| 0.1709 | 1.0 | 2052 | 6.2976 | 22.0505 | 0.1529 |
| 0.1076 | 2.0 | 4104 | 5.5274 | 18.8083 | 0.1377 |
| 0.0509 | 3.0 | 6156 | 4.7389 | 16.4986 | 0.1399 |
| 0.0223 | 4.0 | 8208 | 4.3180 | 14.9002 | 0.1409 |
| 0.004 | 5.0 | 10260 | 4.0149 | 14.0500 | 0.1453 |
Framework versions
- Transformers 4.52.4
- Pytorch 2.8.0+cu128
- Datasets 3.6.0
- Tokenizers 0.21.4
- Downloads last month
- -
Model tree for SmartWhatt/rangsitra-whisper-md-cv
Base model
openai/whisper-mediumDataset used to train SmartWhatt/rangsitra-whisper-md-cv
Evaluation results
- Wer on Common Voice 17.0self-reported14.050