README.md

---
license: mit
tags:
- seismology
- phase-detection
- pytorch
---

# PAW Pretrained Model

Pretrained PAW (Picker for Arrival of Waves) model for seismic phase detection.

## Model Details

- **Architecture:** CNN + LSTM + Transformer hybrid
- **Sampling Rate:** 40 Hz (0.025s per sample)
- **Input:** 5-second waveforms with 0.5s padding on each side
- **Output:** Binary masks indicating phase arrival windows

## Usage

```python
from pawlib import PAW

# Load pretrained model from Hugging Face
model = PAW.from_pretrained('hf://suroRitch/pawlib-pretrained/paw_corrected.pt')

# Make predictions
predictions = model.predict(waveforms)
```

## License

MIT