Installation

DustPy-GPU keeps the upstream dustpy import name, so a fresh virtual environment is recommended to avoid conflicts with an existing DustPy install.

PyPI

For NumPy-only use:

python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install dustpy-gpu

For GPU use, first install the CuPy package matching the local CUDA major version. For example, with CUDA 13:

python -m pip install cupy-cuda13x
python -m pip install dustpy-gpu

DustPy-GPU automatically installs simframe-gpu. Building DustPy-GPU from source also requires a Fortran compiler.

GitHub checkout

For an editable development install:

git clone https://github.com/astroboylrx/dustpy-gpu.git
python -m pip install -e ./dustpy-gpu

Confirm the selected backend after installation:

import dustpy

sim = dustpy.Simulation(backend="auto")
print(sim.backend)