Whisper small - Denis Musinguzi
This model is a fine-tuned version of openai/whisper-large on the Common Voice 14.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3365
- Wer: 0.2992
- Cer: 0.0886
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: 32
- seed: 42
- 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 |
|---|---|---|---|---|---|
| 1.1439 | 0.3 | 800 | 0.1092 | 0.5335 | 0.4676 |
| 0.3861 | 0.61 | 1600 | 0.1112 | 0.4259 | 0.4185 |
| 0.3195 | 0.91 | 2400 | 0.0818 | 0.3794 | 0.3365 |
| 0.2447 | 1.22 | 3200 | 0.0898 | 0.3637 | 0.3310 |
| 0.2168 | 1.52 | 4000 | 0.0905 | 0.3473 | 0.3250 |
| 0.2099 | 1.82 | 4800 | 0.0874 | 0.3354 | 0.3205 |
| 0.1793 | 2.13 | 5600 | 0.0849 | 0.3376 | 0.3013 |
| 0.1437 | 2.43 | 6400 | 0.0823 | 0.3356 | 0.2985 |
| 0.14 | 2.74 | 7200 | 0.0833 | 0.3322 | 0.2953 |
| 0.1351 | 3.04 | 8000 | 0.0873 | 0.3328 | 0.2979 |
| 0.0994 | 3.34 | 8800 | 0.0699 | 0.3374 | 0.2838 |
| 0.0986 | 3.65 | 9600 | 0.3365 | 0.2992 | 0.0886 |
Framework versions
- Transformers 4.38.1
- Pytorch 2.2.1
- Datasets 2.17.0
- Tokenizers 0.15.2