Whisper tiny Japanese English
This model is a fine-tuned version of openai/whisper-tiny on the Japanese English dataset. It achieves the following results on the evaluation set:
- Loss: 0.5073
- Wer: 20.8647
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: 3.75e-05
- train_batch_size: 32
- eval_batch_size: 64
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 50
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2209 | 1.9608 | 100 | 0.4809 | 22.3371 |
| 0.0601 | 3.9216 | 200 | 0.4812 | 21.4286 |
| 0.0342 | 5.8824 | 300 | 0.4979 | 22.2431 |
| 0.0187 | 7.8431 | 400 | 0.5164 | 21.4286 |
| 0.0145 | 9.8039 | 500 | 0.5096 | 21.4756 |
| 0.0084 | 11.7647 | 600 | 0.5095 | 21.3503 |
| 0.0071 | 13.7255 | 700 | 0.5233 | 21.3503 |
| 0.0059 | 15.6863 | 800 | 0.5151 | 21.0526 |
| 0.0055 | 17.6471 | 900 | 0.5088 | 20.6454 |
| 0.0061 | 19.6078 | 1000 | 0.5073 | 20.8647 |
Framework versions
- Transformers 5.23.0.dev0
- Pytorch 2.11.0+cu130
- Datasets 5.1.0
- Tokenizers 0.23.3