AnR ML Docker

This repository contains the necessary elements (code and artifacts) to build a Machine Learning container suitable to execute all the ML exercises of the AnR course.

The container is equipped with a few development tools such as vim, tmux, git, ... in order to process and analyse (a.k.a. engineer) data and to diagnose any issue.

!!! Note: Do not clone this repository into a path containing a space !!!*

Prerequisites

• A UNIX-like operating system, preferably Linux. (Ubuntu 20.04 is recommended.)
• The glxinfo command. (It's included with the mesa-utils package on Ubuntu. So, install this on the Linux host before building this repository. On an Ubuntu host, execute sudo apt install mesa-utils to install glxinfo.)
• An operational docker daemon.
• Standard Bash and basic ROS knowledge.
• A Nvidia graphics card capable of running hardware accelerated graphics. Although, any recent AMD or Intel GPU will also work flawlessly.

Known issues

If Nvidia docker returns the following error:

   Docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].

Look at the following site to solve it:

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#install-guide

How to build the container

Linux/Unix

A bash script is provided to build the container, it can be executed by entering the following command:

   $ ./001_build_images.sh

Windows

For Windows, use the premade image at https://drive.google.com/drive/folders/1KqxEocjVeOtsky2f2vomWljWc7ir1reJ?usp=sharing and perform the steps in the next section.

How to start the container

Linux/Unix

To start the container execute the script below:

   $ 003_start_pxl_ml_container.sh

This script will check the available GPU and start the container accordingly.

To use multiple bash shells in the container, It's advised to either work with tmux or execute the script with prefix 005 from the host:

   $ ./005_attach_bash_to_noetic_full_desktop.sh

Pro-tip: Learn to use tmux. It's awesome!

Windows

Open Powershell.

Enter the following command to load the docker image: docker load -i /<path>/<to>/pxl_ml_image.tar

cd 01_ml_docker

While in the 01_ml_docker folder in powershell, execute the following command to run the container:

docker run --privileged -it --rm --name pxl_ml_container --hostname pxl_ml_container -v $PWD/../commands/bin:/home/user/bin -v $PWD/../notebooks:/home/user/notebooks -v $PWD/../data:/home/user/data -v $PWD/../app:/home/user/app  -p 7777:7777 -p 8080:8080 -p 5000:5000 pxl_ml_image:latest bash

When you're in the container execute this command to start jupyter:

start_jupyter

or

jupyter-notebook --ip=0.0.0.0 --allow-root --no-browser --port=7777

When you're done exploring / training / etc, you can edit the app.py file in the app folder to configure your flask backend. In the container, navigate to the app folder and execute python3 app.py

Prebuilt image

You can find a prebuilt version of the image right here.

Important!

Make sure to still clone this repository, even if you use the prebuilt image. You will need the necessary scripts (003..., 004...) to make sure everything works correctly.

Tensorflow 1 version

Use

docker load --input pxl_ml_image_tf1.tar

to import it, and

003_start_pxl_ml_container.sh

to start it.

Tensorflow 2 version

Use

docker load --input pxl_ml_image_tf2.tar

to import it, and

004_start_pxl_ml_container_2.sh

to start it.