Installation

Install a wheel (recommended)

Install the prebuilt wheel that matches your CUDA major version:

[uv] pip install cuspa-cu12  # CUDA 12
[uv] pip install cuspa-cu13  # CUDA 13

Install only one. Both distributions provide the cuspa Python package. Wheels target manylinux_2_28 on x86_64 and aarch64. CUDA 12 wheels target SM 75, 80, 86, 89, and 90; CUDA 13 wheels also target SM 100 and 120.

The optional adapters are extras of the same distribution:

[uv] pip install "cuspa-cu12[geoarrow]"     # or cuspa-cu13
[uv] pip install "cuspa-cu12[geopandas]"
[uv] pip install "cuspa-cu12[spatialdata]"

Verify the installation on a GPU host:

python -c 'import cuspa, cuspa._core; print(cuspa.__version__)'

Requirements

cuspa requires Python 3.12 or newer, a compatible NVIDIA GPU and driver, a CUDA 12 or CUDA 13 runtime, and CuPy or CUDA-enabled PyTorch. Install the CuPy or PyTorch package matching your CUDA environment separately. Aggregation and I/O adapters currently require CuPy.

Build from source

The cuspa distribution is the source distribution. Installing it compiles cuspa for the GPU detected on the build machine, which requires a matching CUDA toolkit that supports the target GPU and a C++17 compiler.

[uv] pip install cuspa

The optional adapters are extras here too:

[uv] pip install "cuspa[geoarrow]"
[uv] pip install "cuspa[geopandas]"
[uv] pip install "cuspa[spatialdata]"

Development install

[uv] pip install -e ".[test]"