suroRitch/pawlib-pretrained

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

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