<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
whisper-small-ko-speed
This model is a fine-tuned version of openai/whisper-small on the aihub elder over 70 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2785
- Cer: 94.8852
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
- gradient_accumulation_steps: 4
- 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: 500
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 1.1976 | 0.3 | 100 | 0.8230 | 292.9152 |
| 0.603 | 0.59 | 200 | 0.4920 | 18.9046 |
| 0.2918 | 0.89 | 300 | 0.2751 | 13.7068 |
| 0.1954 | 1.19 | 400 | 0.2703 | 35.2460 |
| 0.1999 | 1.48 | 500 | 0.2770 | 46.4190 |
| 0.1885 | 1.78 | 600 | 0.2792 | 25.6631 |
| 0.0669 | 2.07 | 700 | 0.2750 | 40.1412 |
| 0.0669 | 2.37 | 800 | 0.2800 | 85.5634 |
| 0.0518 | 2.67 | 900 | 0.2789 | 94.7962 |
| 0.0582 | 2.96 | 1000 | 0.2785 | 94.8852 |
Framework versions
- Transformers 4.34.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3
