SageNet-Reverse

Code for "Simulation-based Inference in Searches for the Inflationary Stochastic Gravitational-Wave Backgrounds using Multi-band Conditional Normalizing Flows" (AAS76432, ApJS).

• release/ — frozen, parameterized code that reproduces the paper's results (dataset production → training the two configurations → main table → figures). Start at release/README.md.
• Data, checkpoints, noise files, and posterior samples are distributed on Zenodo (see the paper's Software section).
• The repository root additionally keeps the original 2026-04 normalizing flow library (train.py, test.py, configs/, noise/).

Quick start

cd release
pip install -r requirements.txt
# point paths.json at the Zenodo files, then:
python r55_t2_main.py   # regenerate the main-table npz files

Citation

If you use this code, please cite the paper and the Zenodo data record.