# Whisper Wakeword Lumi

- Base: `vinai/PhoWhisper-base`
- Task: Wakeword classification (binary)
- Current version: **ver1.0.1**

## Folder `ver1.0.1/`:
- `config.json`
- `preprocessor_config.json`
- `model.safetensors`
- `training_args.bin`
- `exports/` (ONNX model)

## How to run:
- Usage: `python inference_whisper_ww.py path/to/audio.wav`
- Expected output:
```
[Positive case]
Wakeword probability: 0.9972
Prediction: WAKEWORD (positive)

[Negative case]
Wakeword probability: 0.0005
Prediction: NON-WAKEWORD (negative)
```