Installation¶
You need the following to install OpenEquivariance:
A Linux system equipped with an NVIDIA / AMD / Intel graphics card.
Either PyTorch >= 2.10, or JAX>0.5.0 with CUDA, RocM, or XPU support.
To build from source: GCC 9+ and the CUDA / HIP toolkit, or the oneAPI DPC++ compiler (
icpx) for Intel GPUs. The commandc++ --versionshould return >= 9.0; see below for details on setting an alternate compiler.
Note
On Intel GPUs the kernels are generated as SYCL and compiled at runtime
through the oneAPI kernel compiler, so icpx must be on your PATH
(PyTorch locates the SYCL toolchain from it). Tensors passed to the
kernels live on the xpu device rather than cuda. The precompiled
extension and the JAX frontend both remain CUDA/HIP only.
Installation is one easy command, followed by import verification:
pip install openequivariance
python -c "import openequivariance"
OpenEquivariance bundles a precompiled library files for two recent CUDA versions and a HIP version. If you’re building for an architecture we don’t support or need a different CUDA version, you’ll need to build the package from source:
pip install --no-binary openequivariance openequivariance
Set the environment variables CC and CXX before the pip install
to build with a different compiler:
export CC=/path/to/your/gcc
export CXX=/path/to/your/g++
On Intel GPUs, the extension is still compiled on first import, which can take a couple of minutes; subsequent imports use the cached build. Install the extra build dependencies with:
pip install openequivariance[sycl]
First ensure the appropriate JAX Python package is installed in your environment. Then run the following two commands stricly in order:
pip install openequivariance[jax]
pip install openequivariance_extjax --no-build-isolation
Ensure that JAX is installed correctly with RocM support before running, in order,
pip install openequivariance[jax]
JAX_HIP=1 pip install openequivariance_extjax --no-build-isolation
pip install "git+https://github.com/PASSIONLab/OpenEquivariance#subdirectory=openequivariance"
pip install "openequivariance[jax] @ git+https://github.com/PASSIONLab/OpenEquivariance#subdirectory=openequivariance"
pip install "git+https://github.com/PASSIONLab/OpenEquivariance#subdirectory=openequivariance_extjax" --no-build-isolation
# Use the command below for JAX+AMD
# JAX_HIP=1 pip install "git+https://github.com/PASSIONLab/OpenEquivariance#subdirectory=openequivariance_extjax" --no-build-isolation
If you’re using JAX, set the environment variable
OEQ_NOTORCH=1 to avoid a PyTorch import:
export OEQ_NOTORCH=1
python -c "import openequivariance.jax"
Configurations on Major Platforms¶
OpenEquivariance has been tested on both supercomputers and lab clusters. Here are some tested environment configuration files. If you use OpenEquivariance on a major cluster, send us a pull request to add your configuration!
Aurora provides a PyTorch build with XPU support in its frameworks
module; no separate PyTorch install is needed.
module restore
module load frameworks
export CC=icx
export CXX=icpx