whisper-medium-dora-afrispeech-general
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: 1.0067
- Model Preparation Time: 0.0397
- Wer: 0.107
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: 0.0001
- 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: linear
- lr_scheduler_warmup_steps: 1349
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Wer |
|---|---|---|---|---|---|
| 0.5061 | 1.0 | 1350 | 0.6026 | 0.0397 | 0.12 |
| 0.4024 | 2.0 | 2700 | 0.5541 | 0.0397 | 0.106 |
| 0.2893 | 3.0 | 4050 | 0.5662 | 0.0397 | 0.109 |
| 0.2021 | 4.0 | 5400 | 0.6315 | 0.0397 | 0.107 |
| 0.1596 | 5.0 | 6750 | 1.0067 | 0.0397 | 0.107 |
Framework versions
- PEFT 0.18.1
- Transformers 5.3.0
- Pytorch 2.6.0+cu124
- Datasets 2.19.0
- Tokenizers 0.22.2