Whisper Large - 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: 0.2966
- Wer: 0.2467
- Cer: 0.0700
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 |
|---|---|---|---|---|---|
| 0.6329 | 0.61 | 1600 | 0.0878 | 0.3515 | 0.3385 |
| 0.2241 | 1.22 | 3200 | 0.0589 | 0.3045 | 0.2517 |
| 0.1618 | 1.82 | 4800 | 0.0707 | 0.2801 | 0.2645 |
| 0.1109 | 2.43 | 6400 | 0.0774 | 0.2870 | 0.2580 |
| 0.0837 | 3.04 | 8000 | 0.0597 | 0.2900 | 0.2333 |
| 0.045 | 3.65 | 9600 | 0.2966 | 0.2467 | 0.0700 |
Framework versions
- Transformers 4.38.1
- Pytorch 2.2.1
- Datasets 2.17.0
- Tokenizers 0.15.2