Whisper-Small-am-v11
This model is a fine-tuned version of openai/whisper-small on the hearing voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.6920
- Wer: 183.3625
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 OptimizerNames.ADAMW_TORCH 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: 400
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| No log | 0.2222 | 50 | 1.1042 | 338.7729 |
| 1.2376 | 0.4444 | 100 | 0.9183 | 301.9761 |
| 1.2376 | 0.6667 | 150 | 0.8419 | 213.0040 |
| 0.7542 | 0.8889 | 200 | 0.7960 | 188.3506 |
| 0.7542 | 1.1111 | 250 | 0.7673 | 186.8685 |
| 0.676 | 1.3333 | 300 | 0.7388 | 193.2430 |
| 0.676 | 1.5556 | 350 | 0.7148 | 188.5100 |
| 0.6401 | 1.7778 | 400 | 0.6920 | 183.3625 |
Framework versions
- Transformers 4.51.3
- Pytorch 2.6.0+cu124
- Datasets 3.5.1
- Tokenizers 0.21.1