Trained models for gravitational wave inference with `labrador`

Trained labrador models

This record contains trained models to do amortized gravitational wave inference with labrador, and a jupyter notebook to illustrate their usage.

For demonstration purposes, it also contains data files corresponding to events in GWTC-4.0. These files have been created with cogwheel; refer to its documentation for how to create them.

Note: since these data files are only intended for demonstration, they all have a frequency range of (15, 1024) Hz (even though LVK may have flagged some frequency ranges unusable for some events, or the some signals may have power at higher frequencies), and all are 64s long with the event at the center. Minimal processing has been applied in the form of inpainting loud glitches.

Installation

A conda environment specification file environment.yml has been provided for reproducibility. Create the environment with

conda env create -f environment.ymlconda activate labrador-demo

and run the notebook inference.ipynb within it.

Versions

 The models in v1 of the zenodo record were created with labrador's commit c23b81476139e760195fcc266e087fe7f042f1da

Reference

Javier Roulet, Marco Crisostomi, Lucy M Thomas, Katerina Chatziioannou (2026). labrador: a domain-optimized machine-learning tool for gravitational wave inference [2604.08897]