whisper-medium-named-e_v231115
This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.0089
- Wer: 0.0
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: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 5
- training_steps: 120
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 4.4712 | 5.0 | 10 | 2.8412 | 15.0 |
| 1.3896 | 10.0 | 20 | 0.7323 | 95.0 |
| 0.5329 | 15.0 | 30 | 0.5708 | 10.0 |
| 0.4335 | 20.0 | 40 | 0.4734 | 35.0 |
| 0.3684 | 25.0 | 50 | 0.4014 | 15.0 |
| 0.3109 | 30.0 | 60 | 0.3379 | 15.0 |
| 0.2567 | 35.0 | 70 | 0.2760 | 0.0 |
| 0.2065 | 40.0 | 80 | 0.2132 | 0.0 |
| 0.1534 | 45.0 | 90 | 0.1557 | 0.0 |
| 0.105 | 50.0 | 100 | 0.0955 | 0.0 |
| 0.0497 | 55.0 | 110 | 0.0311 | 0.0 |
| 0.0093 | 60.0 | 120 | 0.0089 | 0.0 |
Framework versions
- Transformers 4.36.0.dev0
- Pytorch 2.1.0+cu118
- Datasets 2.14.8.dev0
- Tokenizers 0.15.0
- Downloads last month
- 3