AstroFlow

A GPU-accelerated pipeline for radio single-pulse and fast radio burst searches

AstroFlow is an end-to-end search pipeline for radio single pulses and fast radio bursts (FRBs). It combines CUDA-accelerated dedispersion, RFI mitigation, AI-assisted candidate detection, and publication-ready candidate visualization in a YAML-configured command-line workflow. The pipeline supports FILTERBANK (.fil) and PSRFITS (.fits) observations and is designed for high-throughput offline or near-real-time searches.

Documentation: https://lintian233.github.io/astroflow/

Preview

Installation

Python 3.10 to 3.12 is recommended.

pip install pulseflow

Read the full documentation: https://lintian233.github.io/astroflow/

Source Build

Source builds require CUDA Toolkit 12.6 or newer, GCC/G++ 11.4 or compatible, CMake 3.22 or newer, Conan 2.x, Ninja, and Python 3.10 to 3.12.

git clone https://github.com/lintian233/astroflow
cd astroflow
bash configure.sh
conda activate dev-astroflow-ml
bash build.sh
pip install -e .

Development builds are also published to TestPyPI:

pip install --pre -i https://test.pypi.org/simple/ pulseflow --extra-index-url https://pypi.org/simple

Dataset and Model

See the Dataset and Model page for the public training dataset and released model weights.

Citation

If AstroFlow is useful in your research, please cite:

ASTROFLOW: A Real-Time End-to-End Pipeline for Radio Single-Pulse Searches
https://doi.org/10.3847/1538-4365/ae4a26

Community

Star History