How to Build the Open-Source OpenVINO™ Toolkit for Linux* from Source with the Python* API Wrapper

Documentation

Install & Setup

000057448

12/20/2022

To use Python* with the open-source OpenVINO™ toolkit for Linux*, you must install Python* 3.5 or above, and then build the toolkit by specifying the correct Python version in the CMake command line.

If you do not explicitly specify a Python version, CMake picks up the system-level Python version, which is 2.7, and your Python scripts will not work.

Note

The instructions below assume you have already installed Python*.

  1. Build from source.

    Refer to the Building for Linux* instructions for more detailed steps.

    $ git clone https://github.com/openvinotoolkit/openvino.git
    $ cd openvino
    $ git submodule update --init --recursive
    $ chmod +x install_build_dependencies.sh
    $ ./install_build_dependencies.sh
    $ mkdir build && cd build

    $ python3 -m pip install -r ../src/bindings/python/src/compatibility/openvino/requirements-dev.txt

     

    Note

    In the CMake command line below, substitute 3.8 with your version of Python*. The location and version of Python depends on the architecture and operating system.

     

    $ cmake -DCMAKE_BUILD_TYPE=Release \
    -DENABLE_PYTHON=ON \
    -DPYTHON_EXECUTABLE=`which python3.8` \
    -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.8.so \
    -DPYTHON_INCLUDE_DIR=/usr/include/python3.8 ..
    $ make --jobs=$(nproc --all)

     

  2. Export environment variables.

    After the build process finishes, export the environment variables:

    $ export PYTHONPATH=$PYTHONPATH:/~/openvino/bin/intel64/Release/python_api/python3.8/
    $ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:~/openvino/bin/intel64/Release/

     

  3. Test the build.

    Check that you correctly built the Python* wrapper by running the following script to import Core:

    $ python3.8
    >>> from openvino.runtime import Core

     

If you can successfully import Core, you have correctly built the OpenVINO™ toolkit with the Python wrapper.