Prerequisites

Throughout this documentation, we assume a Unix-based operating system (e.g. Linux or MacOS). For Windows users, we recommend to use the Windows Subsystem Linux (WSL; see e.g. here).

Below is a list of the most important python packages required to be able to install Magrittetorch.

  • PyTorch (version 2.0.0 or later): for GPU parallelisation and automatic differentiation;

  • Astropy: for units and astronomy related file IO;

  • Anaconda: for managing the required Python packages;

  • Pip: for installing Magrittetorch;

And software required for GPU parallelisation (optional, depending on the GPU vendor):

  • CUDA: for Nvidia GPU’s;

  • ROCm: for AMD GPU’s;

The comprehensive list of required Python packages can be found in the pyproject.toml file.

Once Anaconda, Pip and (optionally) CUDA or ROCm are in place, Magrittetorch can be installed following the quickstart.

Note

Magrittetorch can also be run on CPU. For this, the GPU related dependencies (CUDA, ROCm) are not required. However, computation times will be significantly longer, when compared to compiled C++ code Magritte.