README.md

Example 4

< Go Back

Jupyter + TensorFlow in Docker

Jupyter and TensorFlow are mainstays of the machine-learning and data analytics world.

Let's define a Jupyter image with TensorFlow support.

Command	Arguments	Description
FROM	<image name>	Base image for this image
RUN	<shell-command>	Run this command when building the image
WORKDIR	<directory>	Set the current working directory
ARG	<name>[=optional-default]	Set build argument ARG notebook_dir=/tmp/notebooks
ENV	<name>[=optional-default]	Set environment variable
CMD	<shell-command>	Run this command when starting the container

Complete the Dockerfile skeleton with the necessary packages and their dependencies:

• apt-get: python3, python3-pip
• pip: jupyterlab, pandas, pymysql, tensorflow

Set up the image CMD to launch jupyter when the container starts.

CMD ["jupyter", "notebook", "--ip=0.0.0.0", "--no-browser"]

Building the image

We will build the image with the current user's UID and GID

$ docker build \
	--rm \
	--build-arg USER_NAME=`whoami` \
	--build-arg USER_UID=`id -u` \
	--build-arg USER_GID=`id -g` \
	-t workshop-docker/example4 \
	.

Running the image

Complete the command below:

1. Publish the container port 8888 to the host port 8888.
2. Bind-mount the ./notebooks directory to the container's /tmp/notebooks directory.

Run the container and try accessing the jupyter notebook from within.

$ docker run \
	-it \
	-p <container-port>:<host-port> \
	-u `id -u`:`id -g` \
	-v <host-directory>:<container-directory> \
	--rm \
	workshop-docker/example4

< Go Back | Solution