Skip to content

Docker

Jump to bottom
Ethan edited this page Oct 13, 2022 · 11 revisions

To run notebooks inside a Linux-based Docker container, please use the Dockerfile by following the steps below.

Step 0: Clone the Notebooks Repo

git clone https://github.com/openvinotoolkit/openvino_notebooks.git
cd openvino_notebooks

Step 1: Build the Docker Image

docker build -t openvino_notebooks .

Step 2: Run the Docker Image

  • To run an image on CPU, use the following command:
docker run -it -p 8888:8888 openvino_notebooks
  • To run inference on GPU, the host system is required to be Windows 10 21H2 or Windows 11, and have Intel iGPU drivers with version 30.0.100.9684 or above installed. Use the following command (available since 2022.1.0 release):
docker run -it --rm --device /dev/dxg --volume /usr/lib/wsl:/usr/lib/wsl -p 8888:8888 openvino_notebooks
  • If your host system is Ubuntu* 20.04 then inference run inside the docker image is available for CPU, GPU, Myriad* (NCS2), HDDL targets. Default run:
docker run -it --device /dev/dri:/dev/dri --device-cgroup-rule='c 189:* rmw' -v /dev/bus/usb:/dev/bus/usb --rm -p 8888:8888 openvino_notebooks
  • If you need GPU accelerator only, run the image with the following command:
docker run -it --device /dev/dri:/dev/dri --rm -p 8888:8888 openvino_notebooks
  • If you need Myriad* (NCS2) accelerator only, run the image with the following command:
docker run -it --device-cgroup-rule='c 189:* rmw' -v /dev/bus/usb:/dev/bus/usb --rm -p 8888:8888 openvino_notebooks

NOTE: If your host system is Ubuntu 20, follow the Configuration Guide for the Intel® Graphics Compute Runtime for OpenCL™ on Ubuntu* 20.04.

NOTE: If you plan to run model training notebooks, allocate additional memory by running docker run -it -p 8888:8888 --shm-size 8G openvino_notebooks

Step 3: Open Browser

Copy the URL printed in the terminal and open in a browser. If it is a remote machine, replace 127.0.0.1 with the correct IP address. Jupyter_URL

The Dockerfile can be used to run a local image on Windows, Linux or macOS. It is also compatible with Open Data Hub and Red Hat OpenShift Data Science. The base layer is a UBI 8-based image provided by Project Thoth.

NOTE: When running the container on Windows and macOS, only CPU devices can be used. To access the iGPU, please install the notebooks locally following the instructions on the main README

© Copyright 2018-2023, OpenVINO team

Clone this wiki locally