Skip to content

Latest commit

 

History

History
74 lines (44 loc) · 4.59 KB

INSTALL.md

File metadata and controls

74 lines (44 loc) · 4.59 KB
Raw

1. Installation using Pre-Built Packages

OpenVINO™ integration with TensorFlow is released in two different versions: one built with CXX11_ABI=0 and the other built with CXX11_ABI=1.

Since TensorFlow packages available in PyPi are built with CXX11_ABI=0 and OpenVINO™ release packages are built with CXX11_ABI=1, binary releases of these packages cannot be installed together. Based on your needs, you can choose one of the two available methods:

  • OpenVINO™ integration with TensorFlow PyPi release alongside PyPi TensorFlow
    • Includes pre-built libraries of OpenVINO™ version 2021.4.1. The users do not have to install OpenVINO™ separately
    • Supports Intel® CPUs, Intel® integrated GPUs, and Intel® Movidius™ Vision Processing Units (VPUs). No VAD-M support
    • Build with CXX11_ABI=0

  • OpenVINO™ integration with TensorFlow package released in Github alongside the Intel® Distribution of OpenVINO™ Toolkit
    • Compatible with OpenVINO™ version 2021.4.1
    • Supports Intel® CPUs, Intel® integrated GPUs, Intel® Movidius™ Vision Processing Units (VPUs), and Intel® Vision Accelerator Design with Movidius™ (VAD-M)
    • Build with CXX11_ABI=1
    • Needs a custom TensorFlow ABI1 package, which is available in Github release

Pre-built packages summary

TensorFlow Pip Package OpenVINO™ integration with TensorFlow Pip Package Supported OpenVINO™ Flavor Supported Hardwares Comments
tensorflow openvino-tensorflow OpenVINO™ built from source CPU,iGPU,MYRIAD OpenVINO™ libraries are built from source and included in the wheel package
tensorflow-abi1 openvino-tensorflow-abi1 Dynamically links to OpenVINO™ binary release CPU,iGPU,MYRIAD,VAD-M OpenVINO™ integration with TensorFlow libraries are dynamically linked to OpenVINO™ binaries

1.1. Install OpenVINO™ integration with TensorFlow alongside PyPi TensorFlow

    pip3 install pip==21.0.1
    pip3 install tensorflow==2.5.1
    pip3 install -U openvino-tensorflow

1.2. Install OpenVINO™ integration with TensorFlow alongside the Intel® Distribution of OpenVINO™ Toolkit

  1. Ensure the following versions are being used for pip and numpy:

     pip3 install pip==21.0.1
 pip3 install numpy==1.20.2

  2. Install TensorFlow based on your Python version. You can build TensorFlow from source with -D_GLIBCXX_USE_CXX11_ABI=1 or follow the insructions below to use the appropriate package:

     pip3.6 install https://github.com/openvinotoolkit/openvino_tensorflow/releases/download/v1.0.1/tensorflow_abi1-2.5.1-cp36-cp36m-manylinux2010_x86_64.whl

 or

 pip3.7 install https://github.com/openvinotoolkit/openvino_tensorflow/releases/download/v1.0.1/tensorflow_abi1-2.5.1-cp37-cp37m-manylinux2010_x86_64.whl

 or

 pip3.8 install https://github.com/openvinotoolkit/openvino_tensorflow/releases/download/v1.0.1/tensorflow_abi1-2.5.1-cp38-cp38-manylinux2010_x86_64.whl

  3. Download & install Intel® Distribution of OpenVINO™ Toolkit 2021.4.1 release along with its dependencies from (https://software.intel.com/en-us/openvino-toolkit/download).

  4. Initialize the OpenVINO™ environment by running the setupvars.sh located in <openvino_install_directory>/bin using the command below:

     source setupvars.sh

  5. Install openvino-tensorflow. Based on your Python version, choose the appropriate package below:

     pip3.6 install https://github.com/openvinotoolkit/openvino_tensorflow/releases/download/v1.0.1/openvino_tensorflow_abi1-1.0.1-cp36-cp36m-linux_x86_64.whl

 or

 pip3.7 install https://github.com/openvinotoolkit/openvino_tensorflow/releases/download/v1.0.1/openvino_tensorflow_abi1-1.0.1-cp37-cp37m-linux_x86_64.whl

 or

 pip3.8 install https://github.com/openvinotoolkit/openvino_tensorflow/releases/download/v1.0.1/openvino_tensorflow_abi1-1.0.1-cp38-cp38-linux_x86_64.whl