1. Introduction

I have always been wondering to learn the Course of Deep Learning. So I started to learn the Udacity course ud730. And the first significant hump I met is the install of docker. From my understanding docker is just a Linux process. But I found the whole docker community is way more complex and interesting. Here in this note, I just want to write down knowledges about Docker when I was installing tensorflow assignment.

1.1 Docker Container From Google Cloud Repository

I was following the instructions from this page:

https://github.com/tensorflow/tensorflow/blob/master/tensorflow/examples/udacity/README.md

To be able to work on the assignments, the tensorflow has to be installed. There are three ways to install. I choose to install a pre-built docker container image. After installing docker on my Linux box, this is the first command I executed:

docker run -p 8888:8888 -it –rm b.gcr.io/tensorflow-udacity/assignments

Here is what the command means:

• docker run doc
• -p 8888:8888 [port mapping]
• -it The -it instructs Docker to allocate a pseudo-TTY connected to the container’s stdin; creating an interactive bash shell in the container. stackoverflow
• –rm [remove when exit]
• b.gcr.io google cloud repository
• docker stop
  • SIGKILL: It will let Kernel kill the process without talking to process
  • SIGTERM: It will ask Process: Would you mind to stop?
  • docker kill —-signal=SIGINT foo
  • docker stop —-time=30 foo Use SIGTERM then SIGKILL
• docker ps
  • docker ps
  • docker ps -a
  • docker ps -l
• /tensorflow-udacity/assignments

This directory contains a Dockerfile. This file has the following content:

FROM b.gcr.io/tensorflow/tensorflow:latest
MAINTAINER Vincent Vanhoucke <vanhoucke@google.com>
RUN pip install scikit-learn
ADD *.ipynb /notebooks/
WORKDIR /notebooks
CMD ["/run_jupyter.sh"]

Imagine docker container is just a process. The “./run_jupyter.sh” is just a process that we asked to run. This will run jupyter ipython command and brings up the ipython process. Which listening on port 8888 is the default port of a notebook server jupyter doc

• The assignment container mount from local directory

  docker run -p 8888:8888 -v “/home/spin/z/tensorflow/tensorflow/examples/udacity:/notebooks” -it –rm b.gcr.io/tensorflow-udacity/assignments

1.2 Hotkey when running iPython notebook

• ipython hotkey
• re-reading stdout,stderr for the existing container

1.3 Drawing on iPython notebook

• You have to start ipython note book: command: ipython notebook
• You have to add this line in the beginning: “%matplotlib inline”

2. Docker Image Login

If we execute the docker command like :

docker run -p 8888:8888 -v ~/Downloads/spin/tensorflow/tensorflow/examples/udacity/:/notebooks -it --rm b.gcr.io/tensorflow-udacity/assignments:0.5.0

We will see it running, and it will mount my udacity example folder on the container. If we want to login to the container, we can do this: First docker ps find out the container running id:

docker ps

Then execute it to login:

docker exec -i -t 52f6e407045a /bin/bash

3. Summary

• Figure out what composes a container? Is it a process command? Is it a binary or shell script?