Installation

is available for to .

Note

We do not recommend installation as a root user on your system . Please setup a virtual environment, e.g., via Anaconda or Miniconda, or create a Docker image.

Quick Start

PyTorch
Your OS
Package
CUDA
Run:

Installation via Anaconda

You can now install via Anaconda for all major OS, and CUDA combinations 🤗. If you have not yet installed , install it via conda install as described in its official documentation. Given that you have installed (>=1.11.0), simply run

conda install pyg -c pyg

Warning

Conda packages are currently not available for Windows and M1/M2/M3 macs.

If conda does not pick up the correct CUDA version of , you can enforce it as follows:

conda install pyg=*=*cu* -c pyg

Installation via PyPi

From PyG 2.3 onwards, you can install and use without any external library required except for . For this, simply run:

pip install torch_geometric

Additional Libraries

If you want to utilize the full set of features from , there exists several additional libraries you may want to install:

These packages come with their own CPU and GPU kernel implementations based on the PyTorch C++/CUDA/hip(ROCm) extension interface. For a basic usage of , these dependencies are fully optional. We recommend to start with a minimal installation, and install additional dependencies once you start to actually need them.

Installation from Wheels

For ease of installation of these extensions, we provide pip wheels for these packages for all major OS, and CUDA combinations, see here:

Warning

Wheels are currently not available for M1/M2/M3 macs. Please install the extension packages from source.

  1. Ensure that at least 1.11.0 is installed:

    python -c "import torch; print(torch.__version__)"
    >>> 2.2.0
    
  2. Find the CUDA version was installed with:

    python -c "import torch; print(torch.version.cuda)"
    >>> 12.1
    
  3. Install the relevant packages:

    pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html
    

    where ${TORCH} and ${CUDA} should be replaced by the specific and CUDA versions, respectively:

    • 2.2: ${TORCH}=2.2.0 and ${CUDA}=cpu|cu118|cu121

    • 2.1: ${TORCH}=2.1.0 and ${CUDA}=cpu|cu118|cu121

    • 2.0: ${TORCH}=2.0.0 and ${CUDA}=cpu|cu117|cu118

    • 1.13: ${TORCH}=1.13.0 and ${CUDA}=cpu|cu116|cu117

    For example, for 2.2.* and CUDA 12.1, type:

    pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.2.0+cu121.html
    

    For example, for 2.1.* and CUDA 11.8, type:

    pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.1.0+cu118.html
    

Note: Binaries of older versions are also provided for 1.4.0, 1.5.0, 1.6.0, 1.7.0/1.7.1, 1.8.0/1.8.1, 1.9.0, 1.10.0/1.10.1/1.10.2, 1.11.0, 1.12.0/1.12.1, 1.13.0/1.13.1, and 2.0.0 (following the same procedure). For older versions, you need to explicitly specify the latest supported version number or install via pip install --no-index in order to prevent a manual installation from source. You can look up the latest supported version number here.

ROCm: The external pyg-rocm-build repository provides wheels and detailed instructions on how to install for ROCm. If you have any questions about it, please open an issue here.

Installation from Source

In case a specific version is not supported by our wheels, you can alternatively install them from source:

  1. Ensure that your CUDA is setup correctly (optional):

    1. Check if is installed with CUDA support:

      python -c "import torch; print(torch.cuda.is_available())"
      >>> True
      
    2. Add CUDA to $PATH and $CPATH (note that your actual CUDA path may vary from /usr/local/cuda):

      export PATH=/usr/local/cuda/bin:$PATH
      echo $PATH
      >>> /usr/local/cuda/bin:...
      
      export CPATH=/usr/local/cuda/include:$CPATH
      echo $CPATH
      >>> /usr/local/cuda/include:...
      
    3. Add CUDA to $LD_LIBRARY_PATH on Linux and to $DYLD_LIBRARY_PATH on macOS (note that your actual CUDA path may vary from /usr/local/cuda):

      export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
      echo $LD_LIBRARY_PATH
      >>> /usr/local/cuda/lib64:...
      
      export DYLD_LIBRARY_PATH=/usr/local/cuda/lib:$DYLD_LIBRARY_PATH
      echo $DYLD_LIBRARY_PATH
      >>> /usr/local/cuda/lib:...
      
    4. Verify that nvcc is accessible from terminal:

      nvcc --version
      >>> 11.8
      
    5. Ensure that and system CUDA versions match:

      python -c "import torch; print(torch.version.cuda)"
      >>> 11.8
      
      nvcc --version
      >>> 11.8
      
  2. Install the relevant packages:

    pip install --verbose git+https://github.com/pyg-team/pyg-lib.git
    pip install --verbose torch_scatter
    pip install --verbose torch_sparse
    pip install --verbose torch_cluster
    pip install --verbose torch_spline_conv
    

In rare cases, CUDA or path problems can prevent a successful installation. pip may even signal a successful installation, but execution simply crashes with Segmentation fault (core dumped). We collected common installation errors in the Frequently Asked Questions subsection. In case the FAQ does not help you in solving your problem, please create an issue. Before, please verify that your CUDA is set up correctly by following the official installation guide.

Frequently Asked Questions

  1. undefined symbol: **make_function_schema**: This issue signals (1) a version conflict between your installed version and the ${TORCH} version specified to install the extension packages, or (2) a version conflict between the installed CUDA version of and the ${CUDA} version specified to install the extension packages. Please verify that your version and its CUDA version match with your installation command:

    python -c "import torch; print(torch.__version__)"
    python -c "import torch; print(torch.version.cuda)"
    nvcc --version
    

    For re-installation, ensure that you do not run into any caching issues by using the pip --force-reinstall --no-cache-dir flags. In addition, the pip --verbose option may help to track down any issues during installation. If you still do not find any success in installation, please try to install the extension packages from source.