Skip to content

Latest commit

 

History

History
141 lines (96 loc) · 5.78 KB

INSTALL.md

File metadata and controls

141 lines (96 loc) · 5.78 KB

Requirements

From version 5.0.0 SCIP is automatically shipped when using PyPI for the following systems:

  • CPython 3.8 / 3.9 / 3.10 / 3.11 / 3.12 for Linux (manylinux2014)
  • CPython 3.8 / 3.9 / 3.10 / 3.11 / 3.12 for MacOS for x86_64 / ARM64
  • CPython 3.8 / 3.9 / 3.10 / 3.11 / 3.12 for Windows 64bit

This work is currently ongoing, and is planned to encompass all Python and OS combinations.

Please note that to use any of these Python / OS combinations without the default SCIP, one must build PySCIPOpt from source and link against your own installation of SCIP. Such a scenario is common if the LP plugin should be Gurobi / Xpress / CPLEX instead of Soplex.

When installing from source or using PyPI with a python version and operating system combination that is not mentioned above, PySCIPOpt requires a working installation of the SCIP Optimization Suite. Please, make sure that your SCIP installation works!

Note that the latest PySCIPOpt version is usually only compatible with the latest major release of the SCIP Optimization Suite. See the table on the README.md page for details.

If installing SCIP from source or using PyPI with a python and operating system that is not mentioned above, and SCIP is not installed in the global path, you need to specify the install location using the environment variable SCIPOPTDIR:

  • on Linux and MacOS:
    export SCIPOPTDIR=<path_to_install_dir>
  • on Windows:
    set SCIPOPTDIR=<path_to_install_dir> (cmd, Cmder, WSL)
    $Env:SCIPOPTDIR = "<path_to_install_dir>" (powershell)

Note: If SCIPOPTDIR is not set, the setup script will automatically attempt to detect a conda environment (via the CONDA_PREFIX environment variable) and use it if available. If no conda environment is detected, it will fall back to searching global installation paths.

SCIPOPTDIR needs to have a subdirectory lib that contains the library, e.g. libscip.so (for Linux) and a subdirectory include that contains the corresponding header files:

SCIPOPTDIR
  > lib
    > libscip.so ...
  > include
    > scip
    > lpi
    > nlpi
    > ...

On MacOS, to ensure that the SCIP dynamic library can be located at runtime by PySCIPOpt, you should add your SCIP installation path to the DYLD_LIBRARY_PATH environment variable by running:

export DYLD_LIBRARY_PATH="<path_to_install_dir>/lib:$DYLD_LIBRARY_PATH"

When building SCIP from source using Windows it is highly recommended to use the Anaconda Python Platform.

Installation from PyPI

python -m pip install pyscipopt

To avoid interfering with system packages, it's best to use a virtual environment.
Warning! This is mandatory in some newer configurations.

python3 -m venv venv source venv/bin/activate  # On Windows use: venv\Scripts\activate pip install pyscipopt
pip install pyscipopt

Remember to activate the environment (source venv/bin/activate) in each terminal session where you use PySCIPOpt.

Please note that if your Python version and OS version are in the combinations at the start of this INSTALL file then pip now automatically installs a pre-built version of SCIP. For these combinations, to use your own installation of SCIP, please see the section on building from source. For unavailable combinations this pip command will automatically search your global installs or custom set paths as above.

On Windows for combinations not listed at the start of this file, you may need to ensure that the scip library can be found at runtime by adjusting your PATH environment variable:

  • on Windows: set PATH=%PATH%;%SCIPOPTDIR%\bin

On Linux and MacOS this is encoded in the generated PySCIPOpt library and therefore not necessary.

Building everything from source

Recommended is to install in a virtual environment (e.g. python3 -m venv <DIR_PATH>). Please note that a globally installed version of PySCIPOpt on your machine might lead to problems.

Furthermore, you need to have the Python development files installed on your system (error message "Python.h not found"):

sudo apt-get install python-dev   # for Python 2, on Linux
sudo apt-get install python3-dev  # for Python 3, on Linux

After setting up SCIPOPTDIR as specified above install pyscipopt

export SCIPOPTDIR=/path/to/scip/install/dir
python -m pip install .

For recompiling the source in the current directory . use

python -m pip install --compile .

Building with debug information

To use debug information in PySCIPOpt you need to build it like this:

python -m pip install --install-option="--debug" .

Be aware that you will need the debug library of the SCIP Optimization Suite for this to work (cmake .. -DCMAKE_BUILD_TYPE=Debug).

Testing new installation

To test your brand-new installation of PySCIPOpt you need some dependencies which can be installed via pip.

pip install -r requirements/test.txt

Tests can be run in the PySCIPOpt directory with: :

py.test # all the available tests
py.test tests/test_name.py # a specific tests/test_name.py (Unix)
py.test -nauto # run tests in parallel using all available cores

Ideally, the status of your tests must be passed or skipped. Running tests with pytest creates the __pycache__ directory in tests and, occasionally, a model file in the working directory. They can be removed harmlessly.

Common errors

  • readline: libreadline.so.6: undefined symbol: PC This is a readline/ncurses compatibility issue that can be fixed like this (when using conda):

    conda install -c conda-forge readline=6.2