Whisper Small Dv - BanUrsus
This model is a fine-tuned version of openai/whisper-small on the Common Voice 13 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2999
- Wer Ortho: 56.6961
- Wer: 10.8095
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 50
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.1219 | 1.63 | 500 | 0.1725 | 63.1729 | 13.4698 |
| 0.0472 | 3.26 | 1000 | 0.1644 | 58.0820 | 11.8076 |
| 0.0288 | 4.89 | 1500 | 0.1815 | 58.2283 | 11.3294 |
| 0.0067 | 6.53 | 2000 | 0.2322 | 59.0919 | 11.2946 |
| 0.0018 | 8.16 | 2500 | 0.2608 | 57.5179 | 11.0217 |
| 0.001 | 9.79 | 3000 | 0.2815 | 57.1558 | 10.8895 |
| 0.0002 | 11.42 | 3500 | 0.2943 | 56.8633 | 10.8634 |
| 0.0002 | 13.05 | 4000 | 0.2999 | 56.6961 | 10.8095 |
Framework versions
- Transformers 4.39.2
- Pytorch 1.13.0+cu117
- Datasets 2.16.1
- Tokenizers 0.15.1