openai/whisper-small
This model is a fine-tuned version of openai/whisper-small on the pphuc25/VietMed-split-8-2 dataset. It achieves the following results on the evaluation set:
- Loss: 0.9074
- Wer: 21.2813
- Cer: 17.6054
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: 8
- 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: 100
- num_epochs: 20
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.631 | 1.0 | 569 | 0.6148 | 26.4031 | 20.9932 |
| 0.4198 | 2.0 | 1138 | 0.5805 | 24.4042 | 19.7276 |
| 0.2593 | 3.0 | 1707 | 0.6228 | 24.8545 | 20.0158 |
| 0.1615 | 4.0 | 2276 | 0.6910 | 24.0161 | 19.3631 |
| 0.086 | 5.0 | 2845 | 0.7390 | 24.2065 | 19.7389 |
| 0.0612 | 6.0 | 3414 | 0.7867 | 24.3053 | 19.6096 |
| 0.0467 | 7.0 | 3983 | 0.8099 | 23.6280 | 19.0827 |
| 0.0366 | 8.0 | 4552 | 0.8577 | 23.9868 | 19.4967 |
| 0.0245 | 9.0 | 5121 | 0.8748 | 23.6280 | 19.3119 |
| 0.0166 | 10.0 | 5690 | 0.8653 | 23.1558 | 18.9742 |
| 0.011 | 11.0 | 6259 | 0.8834 | 23.7452 | 19.4160 |
| 0.0139 | 12.0 | 6828 | 0.8843 | 23.3571 | 19.2424 |
| 0.0038 | 13.0 | 7397 | 0.8823 | 22.1600 | 18.0194 |
| 0.0097 | 14.0 | 7966 | 0.8805 | 22.5334 | 18.3371 |
| 0.0022 | 15.0 | 8535 | 0.8891 | 21.7573 | 17.8753 |
| 0.0017 | 16.0 | 9104 | 0.8898 | 21.8854 | 18.0116 |
| 0.0002 | 17.0 | 9673 | 0.8969 | 21.5047 | 17.7807 |
| 0.0016 | 18.0 | 10242 | 0.9140 | 21.3033 | 17.5880 |
| 0.0001 | 19.0 | 10811 | 0.9044 | 21.3070 | 17.6054 |
| 0.0001 | 20.0 | 11380 | 0.9074 | 21.2813 | 17.6054 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1