whisper-medium-full-clinical-r16
This model is a fine-tuned version of openai/whisper-medium on the afrispeech-200 dataset. It achieves the following results on the evaluation set:
- Loss: 0.7557
- Wer: 22.4040
- Cer: 8.9223
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: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 16
- 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: 200
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 2.5732 | 1.1047 | 2500 | 0.4727 | 26.4687 | 12.7305 |
| 1.4199 | 2.2095 | 5000 | 0.4792 | 24.9217 | 11.0376 |
| 0.5931 | 3.3142 | 7500 | 0.5210 | 23.6145 | 9.9791 |
| 0.3973 | 4.4189 | 10000 | 0.5586 | 24.2720 | 10.2508 |
| 0.1234 | 5.5237 | 12500 | 0.6029 | 23.0924 | 9.3290 |
| 0.0623 | 6.6284 | 15000 | 0.6551 | 22.9377 | 9.1811 |
| 0.0322 | 7.7331 | 17500 | 0.6905 | 22.4311 | 9.0467 |
| 0.0251 | 8.8379 | 20000 | 0.7327 | 22.5123 | 8.9581 |
| 0.0127 | 9.9426 | 22500 | 0.7556 | 22.4233 | 8.9253 |
| 0.0090 | 10.0 | 22630 | 0.7557 | 22.4040 | 8.9223 |
Framework versions
- Transformers 5.5.4
- Pytorch 2.11.0+cu130
- Datasets 2.19.0
- Tokenizers 0.22.2