Install WholeGraph#

WholeGraph 26.10 is distributed as two packages:

  • libwholegraph contains the native C/CUDA library.

  • pylibwholegraph contains the Python and PyTorch-facing API and depends on libwholegraph.

Review the RAPIDS system requirements before installing. WholeGraph is supported on Linux; Windows users should use WSL2.

Conda#

Install the Python API and native dependency from the 26.10 nightly channel:

conda install \
  -c rapidsai-nightly -c conda-forge \
  pylibwholegraph=26.10 cuda-version=13.3

Native C/CUDA users can install only libwholegraph:

conda install \
  -c rapidsai-nightly -c conda-forge \
  libwholegraph=26.10 cuda-version=13.3

Choose a CUDA version supported by the RAPIDS release and the installed NVIDIA driver.

pip#

RAPIDS wheels use a CUDA-major suffix. For CUDA 13, install the 26.10 nightly Python package with:

python -m pip install \
  --extra-index-url=https://pypi.nvidia.com \
  --extra-index-url=https://pypi.anaconda.org/rapidsai-wheels-nightly/simple \
  "pylibwholegraph-cu13==26.10.*"

Use pylibwholegraph-cu12 on a supported CUDA 12 configuration. Installing pylibwholegraph also installs the matching libwholegraph wheel.

Verify the installation#

pylibwholegraph_smoke_check

Importing the package without starting a distributed environment is also a useful basic check:

python -c "import pylibwholegraph; print(pylibwholegraph.__version__)"

For a development build, see Build WholeGraph from source.