Whisper tiny Japanese
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: 0.7496
- Wer: 24.8747
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: 4
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- 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: 300
- training_steps: 6000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0078 | 9.9055 | 1000 | 0.6305 | 26.4568 |
| 0.0003 | 19.8060 | 2000 | 0.6802 | 24.4361 |
| 0.0001 | 29.7065 | 3000 | 0.7037 | 24.4987 |
| 0.0001 | 39.6070 | 4000 | 0.7204 | 24.4674 |
| 0.0001 | 49.5075 | 5000 | 0.7432 | 24.6711 |
| 0.0001 | 59.4080 | 6000 | 0.7496 | 24.8747 |
Framework versions
- Transformers 5.23.0.dev0
- Pytorch 2.11.0+cu130
- Datasets 5.1.0
- Tokenizers 0.23.3