Whisper Tiny chinese - VingeNie
This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 16.1 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2238
- Cer Ortho: 12.1911
- Cer: 9.6411
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: 0.0001
- train_batch_size: 64
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 200
- training_steps: 2400
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer Ortho | Cer |
|---|---|---|---|---|---|
| 0.3743 | 0.4762 | 400 | 0.3739 | 24.9883 | 15.7515 |
| 0.3062 | 0.9524 | 800 | 0.3034 | 17.2255 | 12.8821 |
| 0.2075 | 1.4286 | 1200 | 0.2715 | 14.9716 | 11.5961 |
| 0.1894 | 1.9048 | 1600 | 0.2445 | 13.9330 | 10.5500 |
| 0.1222 | 2.3810 | 2000 | 0.2330 | 12.5011 | 9.9106 |
| 0.1125 | 2.8571 | 2400 | 0.2238 | 12.1911 | 9.6411 |
Framework versions
- Transformers 4.40.1
- Pytorch 2.2.1+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1
- Downloads last month
- 6