whisper-small-10DB
This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:
- Cer: 24.3711
- Loss: 0.4070
- Wer: 67.4490
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: 16
- 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: 100
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Cer | Validation Loss | Wer |
|---|---|---|---|---|---|
| 2.4924 | 1.2675 | 100 | 36.3856 | 0.6204 | 85.7319 |
| 0.9743 | 2.5350 | 200 | 29.2916 | 0.3102 | 74.2434 |
| 0.6605 | 3.8025 | 300 | 24.9447 | 0.2807 | 69.4873 |
| 0.3121 | 5.0637 | 400 | 25.4981 | 0.3011 | 68.6844 |
| 0.1779 | 6.3312 | 500 | 24.2302 | 0.3345 | 66.8931 |
| 0.0859 | 7.5987 | 600 | 23.9284 | 0.3719 | 66.3990 |
| 0.0483 | 8.8662 | 700 | 24.3711 | 0.4070 | 67.4490 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.8.3
- Tokenizers 0.22.2