Whisper Medium - Denis Musinguzi
This model is a fine-tuned version of openai/whisper-medium on the Common Voice 14.0 dataset. It achieves the following results on the evaluation set:
- Cer: 0.0622
- Loss: 0.2969
- Wer: 0.2355
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
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 10000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Cer | Validation Loss | Wer |
|---|---|---|---|---|---|
| 0.9513 | 0.3 | 800 | 0.0998 | 0.4428 | 0.4067 |
| 0.313 | 0.61 | 1600 | 0.0913 | 0.3519 | 0.3427 |
| 0.2593 | 0.91 | 2400 | 0.0628 | 0.3160 | 0.2689 |
| 0.1887 | 1.22 | 3200 | 0.0633 | 0.3049 | 0.2574 |
| 0.1642 | 1.52 | 4000 | 0.0752 | 0.2906 | 0.2655 |
| 0.1595 | 1.82 | 4800 | 0.0737 | 0.2807 | 0.2617 |
| 0.1288 | 2.13 | 5600 | 0.0643 | 0.2889 | 0.2416 |
| 0.0928 | 2.43 | 6400 | 0.0629 | 0.2860 | 0.2387 |
| 0.0887 | 2.74 | 7200 | 0.0572 | 0.2838 | 0.2309 |
| 0.0836 | 3.04 | 8000 | 0.0575 | 0.2897 | 0.2338 |
| 0.0466 | 3.34 | 8800 | 0.0572 | 0.2968 | 0.2322 |
| 0.045 | 3.65 | 9600 | 0.0622 | 0.2969 | 0.2355 |
Framework versions
- Transformers 4.38.1
- Pytorch 2.2.1
- Datasets 2.17.0
- Tokenizers 0.15.2