Per the Docker docs:
Docker containers wrap a piece of software in a complete filesystem that contains everything needed to run: code, runtime, system tools, system libraries – anything that can be installed on a server. This guarantees that the software will always run the same, regardless of its environment.
Using Docker to run your code consists of the following:
- Install Docker on your computer or a remote system
- Pull the precompiled Docker image from Docker Hub
- Each time you wish to work, run the image as a new container
The tricky part to working with Docker will be accessing your project code while working with a Python process (via Jupyter, Python, or IPython) running in a container.
GPU Support is available for AWS using nvidia-docker
. Start here.
Instructions for installation vary by operating system and version.
OS | Installation Instruction |
Docker System | Shell | Access Jupyter at |
---|---|---|---|---|
Linux | Here | Docker for Linux | bash |
localhost:8888 |
MacOS >= 10.10.3 (Yosemite) |
Here | Docker for Mac | bash |
localhost:8888 |
MacOS >= 10.8 (Mountain Lion) |
Here | Docker Toolbox for Max | Docker Quickstart Terminal | #DOCKERIP:8888 |
Windows 10 Pro, Enterprise, or Education |
Here | Docker for Windows | Windows PowerShell |
localhost:8888 |
Windows 7, 8, 8.1, or 10 Home |
Here | Docker Toolbox for Windows | Docker Quickstart Terminal | #DOCKERIP:8888 |
A precompiled image with all dependencies required for the first term is available on Docker Hub.
Once you have docker working, pull the image using the following command:
docker pull udacity/carnd-term1-starter-kit
In your shell, navigate to the directory of a project, e.g.
$ cd ~/src/CarND-LaneLines-P1
From within this directory, you are going to run a Jupyter server. In order to do this you must attach to the correct port and share a local volume.
The easiest way to share a local volume is via the pwd
command, a shell
command that prints the working directory. This command will be used
differently based on your shell.
If you're using Windows PowerShell
:
docker run -it --rm -p 8888:8888 -v ${pwd}:/src udacity/carnd-term1-starter-kit
If you're using bash
or Docker Quickstart Terminal:
docker run -it --rm -p 8888:8888 -v `pwd`:/src udacity/carnd-term1-starter-kit
Let's break this down.
docker run
is the command a startup and run a Docker container.
-it
forces the container to run in the foreground (interactive mode) and
provides an I/O to the container.
--rm
removes the container once it stops running.
It prevents the buildup of stale containers once you stop them from running.
-p 8888:8888
maps port 8888 on our local machine to port 8888 in the Docker
container, this allows us to access port 8888 in the container
by visiting localhost:8888
.
-v ${pwd}:/src
mounts the pwd (present working directory) to the /src
directory in the container. Basically, this lets us access files
from our local machine on the docker container.
udacity/carnd-term1-starter-kit
is the name of the container to run.
To learn more about Docker visit the docs.
Use the ip address of the container to connect to Jupyter notebook, this can be found by opening another
docker terminal and running this command: docker-machine ip default
. Jupyter server can then be accessed
by going to this page from your browser: http://[ip-of-container]:8888
.