Overview

• Docker and Deep Learning
  • Getting started with docker.
  • Overview of deep learning and how it works.
  • Introduction to various useful toolsets for deep learning.
• How To

Docker and Deep Learning

Presentation courtesy of Bennett Wineholt.

Abstract:

Scaling up systems for resource-intensive machine learning tasks demands convenient methods to manage computations distributed across multiple servers. Come and learn about both the processes underlying new Deep Learning techniques that have been applied to piloting drones and driving autonomous vehicles as well as the Docker containerization tools used to train these systems at scale.

How To

• Docker and Deep Learning on your laptop
  • Install Docker
  • Start Docker
  • Pull Docker container from Docker Hub
  • Run the Container
  • Run Deep Learning on the Container
  • Shut it Down

Install Docker

• Visit https://docs.docker.com/engine/installation/ and pick your operating system.
  • https://www.docker.com/community-edition

Start Docker

You may need to run the Docker application, or the command

$ sudo service docker start

A reboot may be required since Docker ties tightly into system functions.

Pull Docker container from Docker Hub

• TensorFlow is a widely used learning toolset, so we will be working with the TensorFlow docker container.

  $ docker pull tensorflow/tensorflow
  

Run the Container

Now that we have downloaded a docker container, the next step is to run it:

$ docker run -it -p 8888:8888 tensorflow/tensorflow

Run Deep Learning on the Container

Go to your browser on http://localhost:8888/ (the command printed on container start). Consider reading the MNIST from Scratch jupyter notebook comments to get a good idea of the steps involved.

Use the MNIST from Scratch jupyter notebook to execute the code, running commands with Shift+Enter or Cell -> Run All.

All credit to the authors: https://www.tensorflow.org/get_started/mnist/beginners

• Warning: Step 22 Training is slow. About 10 minutes on 8 core cpu (2GHz i7) 16G RAM
• Warning: Plots in Step 23 seem to be bugged and crash the notebook kernel

Shut it Down

• When finished, exit your browser window.

• In a terminal list running containers:

  $ docker ps
  

  • The previous command will list any / all container id’s presently running on your computer.

  • Take the container_id(s) you want to shut down and use it to stop the containers, then remove them permanently

    $ docker stop container_id
    $ docker rm container_id
    

• Then shut down Docker; based on your OS there may be a service icon, or try the Linux command (your Linux OS may or may not be using service, but it is more likely it is than not):

  $ sudo service docker stop