Skip to content

install.rst

Latest commit

 

History

History
88 lines (60 loc) · 2.29 KB

install.rst

File metadata and controls

88 lines (60 loc) · 2.29 KB
Raw

Installation

Python 3.13 is recommended; Python 3.11 or newer is required. ROOT is not required for PycWB >= 1.

Install from PyPI

Create a separate environment, then install PycWB:

conda create -n pycwb -c conda-forge python=3.13 pip \
    nds2-client python-nds2-client lalsuite python-ligo-lw
conda activate pycwb
python -m pip install --pre pycwb
pycwb --version

The --pre flag selects the current 1.x prereleases from PyPI. NDS2 client libraries are supplied through conda-forge.

Postproduction and waveform extras

Production searches need no extras. Install optional components only where they are used:

Extra Needed for
xgboost Postproduction: pycwb post-process and pycwb.modules.postprocess import XGBoost for model training and scoring. It is optional because the Linux wheel also installs NVIDIA's NCCL library (about 360 MB).
root Reading cWB ROOT output with uproot and awkward, for example the ROOT adapter and catalog conversion. PyROOT itself is not needed.
autoencoder The TensorFlow autoencoder glitch score.
pycbc PyCBC waveform integrations. 
python -m pip install --pre 'pycwb[xgboost]'
python -m pip install --pre 'pycwb[xgboost,root]'   # several extras

Install from source

For unreleased changes, activate the environment created above and install the source checkout:

git clone https://git.ligo.org/yumeng.xu/pycwb.git
cd pycwb
python -m pip install .

From this checkout, use for example '.[xgboost]' or '.[root]' to install extras. For an editable development install, see :doc:`dev_setup`.

Next steps

Continue with :doc:`start_here` or :doc:`postproduction`.

  • :ref:`installation_release_channels`: stable and prerelease installation options.

For development and documentation builds, see :doc:`dev_setup`.