Whisper Medium Hu - cleaned
This model is a fine-tuned version of openai/whisper-medium on the Common Voice 16.1 hu cleaned dataset. It achieves the following results on the evaluation set:
- Loss: 0.0486
- Wer Ortho: 5.5028
- Wer: 4.9758
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: 6.25e-06
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 16
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.0708 | 0.69 | 200 | 0.0712 | 8.0529 | 7.4029 |
| 0.0279 | 1.37 | 400 | 0.0554 | 6.4034 | 5.8342 |
| 0.0138 | 2.06 | 600 | 0.0487 | 5.6875 | 5.1732 |
| 0.0077 | 2.75 | 800 | 0.0479 | 5.4467 | 4.8837 |
| 0.0037 | 3.43 | 1000 | 0.0486 | 5.5028 | 4.9758 |
Framework versions
- Transformers 4.37.2
- Pytorch 2.1.0+cu121
- Datasets 2.17.0
- Tokenizers 0.15.1
- Downloads last month
- 3