Installation

System requirements

PyEPRI can be installed on all plateforms (Linux, MacOs or Windows). However, GPU support is currently only available for systems equipped with an NVIDIA graphics card and a working installation of the CUDA drivers (which excludes MAC systems).

The installation guidelines assume that you have the following installed on your system:

  • ico-req python3 (the Python 3 programming language)

  • ico-req python3-pip (to install Python packages using the pip command)

  • ico-req python3-venv (for the creation of virtual environment)

  • ico-opt python3-tk (recommended, to avoid display issues on some Linux systems)

  • ico-opt an integrated development environment (IDE) suited to Python (for instance Visual Studio Code)

Under a Debian GNU/Linux distribution, one can easily get the required and recommended libraries by typing into a terminal the following apt-get command (requires superuser (root) privilege).

sudo apt update && sudo apt-get install python3 python3-pip python3-venv python3-tk

If you encounter installation difficulties, feel free to reach us by opening a bug issue.

Install latest version from Github

Open a terminal and execute the following steps in order to checkout the current code release, create a virtual environment, and install pyepri from the github repository.

Installation instructions in command lines

##################
# Clone the code #
##################
git clone https://github.com/remy-abergel/pyepri.git
cd pyepri

###################################################
# Create and activate a fresh virtual environment #
###################################################
python3 -m venv ~/.venv/pyepri
source ~/.venv/pyepri/bin/activate

##########################################################
# Install the `pyepri` package from the checked out code #
# (do not forget the . at the end of the command line)   #
##########################################################
pip install -e .

###########################################################
# Optional: enable {torch-cpu, torch-cuda, cupy} backends #
###########################################################

# enable `torch-cpu` backend
pip install -e ".[torch-cpu]"

# enable `torch-cuda` backend (requires a NVIDIA graphics card with CUDA installed)
pip install -e ".[torch-cuda]"

# enable `cupy` backend (requires a NVIDIA graphics card with CUDA installed)
# (please uncomment the appropriate line depending on your CUDA installation)
# pip install -e ".[cupy-cuda12x]" # For CUDA 12.x
# pip install -e ".[cupy-cuda11x]" # For CUDA 11.x

################################################################
# If you want to compile the documentation by yourself, you    #
# must install the [doc] optional dependencies of the package, #
# compilation instructions are provided next                   #
################################################################
pip install -e ".[doc]" # install some optional dependencies
make -C docs html # build the documentation in html format
firefox docs/_build/html/index.html # open the built documentation (you can replace firefox by any other browser)

Note: the instructions above assume that you have git and make installed on your system.

Because this installation was done in editable mode (thanks to the -e option of pip), any further update of the repository (e.g., using the syncing commang git pull) will also update the current installation of the package.

Troubleshooting

  • Mac users are strongly recommended to use bash shell instead of zsh to avoid slow copy-paste issues (type chsh -s /bin/bash in a terminal).

  • Display issues related to matplotlib interactive mode were reported on Linux systems and were solved by installing python3-tk (type sudo apt-get install python3-tk in a terminal).

  • If the installation of the package or one of its optional dependency fails, you may have more chance with miniconda (or conda).

  • If you still encounter difficulties, feel free to open a bug issue.