Whisper Small ko-Yfreq-E - syp1229
This model is a fine-tuned version of openai/whisper-small on the aihub Y E dialogue dataset. It achieves the following results on the evaluation set:
- Loss: 0.3217
- Cer: 0.0749
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: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 50
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 0.1927 | 0.3 | 100 | 0.3277 | 0.0937 |
| 0.1915 | 0.59 | 200 | 0.3208 | 0.0843 |
| 0.135 | 0.89 | 300 | 0.3242 | 0.0940 |
| 0.062 | 1.19 | 400 | 0.3134 | 0.0819 |
| 0.0512 | 1.48 | 500 | 0.3234 | 0.0827 |
| 0.036 | 1.78 | 600 | 0.3145 | 0.0811 |
| 0.0217 | 2.07 | 700 | 0.3208 | 0.0807 |
| 0.0148 | 2.37 | 800 | 0.3228 | 0.0785 |
| 0.0359 | 2.67 | 900 | 0.3162 | 0.0789 |
| 0.0256 | 2.96 | 1000 | 0.3219 | 0.0784 |
| 0.0054 | 3.26 | 1100 | 0.3224 | 0.0770 |
| 0.0087 | 3.56 | 1200 | 0.3202 | 0.0748 |
| 0.0045 | 3.85 | 1300 | 0.3191 | 0.0755 |
| 0.0095 | 4.15 | 1400 | 0.3165 | 0.0739 |
| 0.0043 | 4.44 | 1500 | 0.3189 | 0.0738 |
| 0.0024 | 4.74 | 1600 | 0.3217 | 0.0749 |
Framework versions
- Transformers 4.34.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.14.5
- Tokenizers 0.13.3