Jupyter Notebook Docker Server for TensorFlow

This is a stand alone Jupyter notebook server docker configuration, with an integration with TensorFlow CPU version.

Quick Start

Before git clone or download this project, please make sure that docker service is running on local computer.

Build Docker Image

$ docker-compose build

The command will build a docker image based on Ubuntu 16.04 with Python 3.5+ and TensorFlow/Jupyter, etc.

NOTE During the installation, there could be a possible warning like the following, but I would suggest to ignore it since the upgrade on pip itself has some knonw issues on jupyter.

You are using pip version 8.1.1, however version 10.0.1 is available.
You should consider upgrading via the 'pip install --upgrade pip' command.

Run Server

$ docker-compose up

If success, some random link should print out in the terminal like this:

http://4e2a38a20392:8888/?token=6aff5dc5 ....

Then you can copy this url and replace the prefix with http://localhost:8888/?token=... to open it in your browser

TIPS You can change the default port 8888 by replacing it all in the docker-compose.yml file.

Stop

To turn it off, just go to the same project folder:

$ docker-compose down

Extra Python 3 Libraries

If "module not found", just go Dockerfile to add more RUN pip3 install --upgrade <the missing module name> as need.

Content

All your work in this stand alone server will be saved into the content folder, by default. Check docker-composer.yml for detail.

Support

Please feel free to contact me by linkedin https://www.linkedin.com/in/rockfordwei/