dockerfiles

Dockerfiles with Intel® Distribution of OpenVINO™ toolkit

This repository folder contains Dockerfiles to build an docker image with the Intel® Distribution of OpenVINO™ toolkit. You can use Docker CI framework to build an image, please follow Get Started with DockerHub CI for Intel® Distribution of OpenVINO™ toolkit.

1. Supported Operating Systems for Docker image
2. Supported devices and distributions
3. Where to get OpenVINO package
4. How to build
5. Prebuilt images
6. How to run a container

Supported Operating Systems for Docker image

• ubuntu18 folder (Ubuntu* 18.04 LTS)
• ubuntu20 folder (Ubuntu* 20.04 LTS)
• centos7 folder (CentOS* 7)
• centos8 folder (CentOS* 8)
• rhel8 folder (RHEL* 8)
• winserver2019 folder (Windows* Server Core base OS LTSC 2019)

Note: dl-workbench folder contains Dockerfiles for OpenVINO™ Deep Learning Workbench.

Supported devices and distributions

Devices:

• CPU
• GPU
• VPU (NCS2)
• HDDL (VPU HDDL) (Prerequisite: run HDDL daemon on the host machine, follow the configuration guide for HDDL device)

OpenVINO documentation for supported devices.

Distributions:

• runtime: IE core, nGraph, OpenCV, plugins
• data_runtime: runtime image content + DL Streamer runtimes
• dev: IE core, nGraph, OpenCV, plugins, samples, demos, Python dev tools: Model Optimizer, Post training Optimization tool, Accuracy checker, Open Model Zoo tools (downloader, converter)
• data_dev: data_runtime image + dev image + Media SDK, Speech Libraries and End-to-End Speech Demos
• base (only for CPU): IE core, nGraph
• proprietary: data_dev + installer

You can generate Dockerfile with your settings, please follow the DockerHub CI documentation.

• runtime, data_runtime, dev, data_dev distributions based on archive package of OpenVINO product. You can just remove unnecessary parts.
• base distribution is created by OpenVINO™ Deployment Manager.
• proprietary distribution based on installer package of OpenVINO product. You can configure installation COMPONENTS, follow Command-Line Silent Instructions
• proprietary, dev, data_dev distributions images for Ubuntu 18 contain Python TensorFlow* 2. virtual environment in /opt/intel/venv_tf2 folder for Model Optimizer and Model Downloader/Converter, because TensorFlow* 1. (default in the image) and TensorFlow* 2. are not compatible. Please follow this guide to activate and use it. Ubuntu 20 based images have TensorFlow* 2 by default and don't have TensorFlow* 1 environment.

Where to get OpenVINO package

You can get OpenVINO distribution packages (runtime, dev, data_dev) directly from public storage and proprietary package with registration here. For example:

• take data_dev l_openvino_toolkit_data_dev_ubuntu18_p_2021.2.185.tgz package and specify -dist data_dev option for Docker CI docker_openvino.py or take a Dockerfile with data_dev suffix.
• take runtime l_openvino_toolkit_runtime_ubuntu18_p_2021.2.185.tgz package and specify -dist runtime or -dist data_runtime option for Docker CI docker_openvino.py or take a Dockerfile with runtime/data_runtime suffix.
• take proprietary l_openvino_toolkit_p_2021.3.249.tgz package and specify -dist proprietary option for Docker CI docker_openvino.py or take a Dockerfile with proprietary suffix.

How to build

• Base image with CPU only:

You can use Docker CI framework to build an image, please follow Get Started with DockerHub CI for Intel® Distribution of OpenVINO™ toolkit.

python3 docker_openvino.py build --file "dockerfiles/ubuntu18/openvino_c_base_2020.3.dockerfile" -os ubuntu18 -dist base -p 2020.3.341

Or via Docker Engine directly, but you need specify BUILD_ID argument:

docker build --build-arg BUILD_ID=2020.3.341 -t ubuntu18_base_cpu:2020.3.341 - < dockerfiles/ubuntu18/openvino_c_base_2020.3.dockerfile

---

• Dev/data_dev/runtime/data_runtime/proprietary image:

You can use Docker CI framework to build an image, please follow Get Started with DockerHub CI for Intel® Distribution of OpenVINO™ toolkit.

python3 docker_openvino.py build --file "dockerfiles/ubuntu18/openvino_cgvh_dev_2021.dockerfile" -os ubuntu18 -dist dev -p 2021.1

For data_dev/runtime/data_runtime/proprietary distributions, please set appropriate -dist and --file options.

Or via Docker Engine directly, but you need specify package_url argument (see Where to get OpenVINO package section) and OpenCL* version to support GPU: GMMLIB, IGC_CORE, IGC_OPENCL, INTEL_OPENCL, INTEL_OCLOC

docker build --build-arg package_url=https://storage.openvinotoolkit.org/repositories/openvino/packages/2021.1/l_openvino_toolkit_dev_ubuntu18_p_2021.1.110.tgz \
             --build-arg GMMLIB=19.3.2 \
             --build-arg IGC_CORE=1.0.2597 \
             --build-arg IGC_OPENCL=1.0.2597 \
             --build-arg INTEL_OPENCL=19.41.14441 \
             --build-arg INTEL_OCLOC=19.41.14441 \
            -t ubuntu18_dev:2021.1 -f dockerfiles/ubuntu18/openvino_cgvh_dev_2021.dockerfile .

---

• Custom image with CPU, iGPU, VPU support
  To learn more, follow Build custom Intel® Distribution of OpenVINO™ toolkit Docker image.

Prebuilt images

Prebuilt images are available on:

• Docker Hub
• Red Hat* Quay.io
• Red Hat* Ecosystem Catalog

How to run a container

Please follow Run a container section in DockerHub CI getting started guide.

Documentation

• Install Intel® Distribution of OpenVINO™ toolkit for Linux* from a Docker* Image
• Install Intel® Distribution of OpenVINO™ toolkit for Windows* from Docker* Image
• Official Dockerfile reference

---

* Other names and brands may be claimed as the property of others.