Whisper Base Hu Final
This model is a fine-tuned version of openai/whisper-base on the Common Voice 16.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.1492
- Wer: 10.7560
- Wer Ortho: 11.8722
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: 2.75e-05
- train_batch_size: 32
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- 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: 3000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Wer Ortho |
|---|---|---|---|---|---|
| 0.2394 | 1.34 | 500 | 0.3043 | 29.5316 | 32.6810 |
| 0.1391 | 2.68 | 1000 | 0.2051 | 18.7726 | 21.4862 |
| 0.0788 | 4.02 | 1500 | 0.1612 | 13.6644 | 15.5877 |
| 0.0243 | 5.35 | 2000 | 0.1480 | 11.4468 | 12.8508 |
| 0.0151 | 6.69 | 2500 | 0.1453 | 10.7857 | 12.0137 |
| 0.0087 | 8.03 | 3000 | 0.1492 | 10.7560 | 11.8722 |
Framework versions
- Transformers 4.36.2
- Pytorch 2.1.0+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0
- Downloads last month
- 10