Whisper base zh v2
This model is a fine-tuned version of ZhihCheng/whisper-base-zh on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1522
- Wer: 75.0
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: 4
- 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: 25
- training_steps: 500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.241 | 0.35 | 50 | 0.4000 | 75.0 |
| 0.1766 | 0.71 | 100 | 0.3001 | 75.0 |
| 0.1085 | 1.06 | 150 | 0.2594 | 81.25 |
| 0.059 | 1.42 | 200 | 0.2162 | 81.25 |
| 0.0412 | 1.77 | 250 | 0.1789 | 87.5 |
| 0.0347 | 2.13 | 300 | 0.1431 | 68.75 |
| 0.0172 | 2.48 | 350 | 0.1408 | 68.75 |
| 0.0137 | 2.84 | 400 | 0.1408 | 68.75 |
| 0.0081 | 3.19 | 450 | 0.1458 | 75.0 |
| 0.0074 | 3.55 | 500 | 0.1522 | 75.0 |
Framework versions
- Transformers 4.38.0.dev0
- Pytorch 2.2.0a0+6a974be
- Datasets 2.16.1
- Tokenizers 0.15.1
- Downloads last month
- 3