Skip to content

Installation

pip install "pyhuge[runtime]"

Published Linux x86_64 wheels are self-contained: OpenBLAS and the OpenMP runtime are bundled, so no system BLAS package is needed.

Source builds require a C++17 compiler and OpenBLAS development files on Linux. Install them first, for example:

# Ubuntu/Debian
sudo apt-get install build-essential libopenblas-dev

# Fedora/RHEL
sudo dnf install gcc-c++ openblas-devel

Then install from a checkout:

git clone https://github.com/Gatech-Flash/huge.git
cd huge/python-package
pip install -e ".[runtime]"

Native source builds currently support Linux and macOS. Windows source builds are not yet supported; use a Linux environment or container.

Optional extras

pip install -e ".[viz]"   # matplotlib + networkx
pip install -e ".[test]"  # pytest
pip install -e ".[docs]"  # mkdocs
pip install -e ".[dev]"   # common contributor setup

Verify install

python -c "import pyhuge; print(pyhuge.test())"
pyhuge-doctor

runtime=True means core dependencies are available.

OpenMP (multicore solvers)

The C++ core parallelizes its per-column solvers (mb, tiger, and parts of glasso) with OpenMP. Linux source builds request OpenMP by default, while macOS source builds enable it when Homebrew libomp is found. If the Linux toolchain lacks OpenMP, set PYHUGE_NO_OPENMP=1 to build a serial extension.

  • Linux wheel: no additional runtime package is needed.
  • Linux source build: use a compiler with OpenMP support; the package also needs the OpenBLAS development package shown above.
  • macOS: install Homebrew's libomp before building:

bash brew install libomp pip install --force-reinstall --no-binary :all: pyhuge # or for a source checkout: pip install -e . (after brew install libomp)

  • Opt out with PYHUGE_NO_OPENMP=1 pip install ....

Check whether your build is parallel with pyhuge-doctor: the native_core.omp_max_threads field reports the thread count (1 means a serial build).

Apple Silicon / architecture notes

pyhuge does not depend on R architecture, but Python/native wheels must match your interpreter architecture. Check:

python -c 'import platform,sys; print(platform.machine()); print(sys.executable)'

PEP 668 environments (externally-managed)

If pip install shows externally-managed-environment, use a virtual env:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install "pyhuge[runtime]"