Installation

Installation#

Requirements#

ASTRA KernelKit is a Python package with requirements:

  • An NVIDIA CUDA-enabled GPU with a compute capability 3.0 or higher.

  • CuPy v12.0 or later.

  • Python 3.10 or later.

This package furthermore relies on a few Python dependencies, such as numpy. For detailed information on the dependencies, see the setup.cfg file.

1. Install cupy first#

It is better to install CuPy yourself, rather than pulling it in as a dependency, since this leaves you with more flexibility towards the CUDA version and installation method. We recommend following CuPy instructions. KernelKit does not require installation of additional packages (such as cuDNN), and is not dependent on a specific CUDA version.

2. Then install astra-kernelkit#

Option 1: Using PyPI#

pip install astra-kernelkit

Option 2: using Conda-Forge#

If you are using an Anaconda environment, you can install ASTRA KernelKit from the conda-forge channel:

conda install -c conda-forge astra-kernelkit

There are no particular advantages to a conda package.

Option 3: from source#

Installation from source is recommended if you want to contribute to the development.

git clone https://github.com/adriaangraas/astra-kernelkit.git
cd astra-kernelkit
pip install -e .
# alternatively: use "pip install -e .[dev]" for development
Contents