TensorFlow Quick Start

Compute Resources

• Have questions or need help with compute, including activation or issues? Follow this link.

• User Agreement

Docker Usage

• The information on this page assumes that you have a knowledge base of using Docker to create images and push them to a repository for use. If you need to review that information, please see the links below.

• Docker and the RIS Compute Service

• Docker Basics: Building, Tagging, & Pushing A Custom Docker Image

Software Included

• noVNC (https://novnc.com/info.html)

• Python (https://www.python.org)

• TensorFlow (https://www.tensorflow.org)

Interactive GUI Session

• Interactions GUI sessions are done via the Custom noVNC Image application in Open On Demand (OOD).

• You can find out more about OOD here: Compute Quick Start.

• There are two fields beyond the basics that will need information specific to this image.

  • Environment Variables

  • Docker Image

Environment Variables

• This information should be space separated in the field.

PASSWORD=password

• Optional variables

  • GUI display size. This can be changed with the following variables.

    • Width default: 1024

    • Height default: 768

  DISPLAY_WIDTH=<width> DISPLAY_HEIGHT=<height>
  

Docker Image

gcr.io/ris-registry-shared/tensorflow:<tag>

TensorFlow Docker Tag

The <tag> will refer to the version of Tensorflow in the Docker container. Please click here see a current list of supported TensorFlow versions and their corresponding tags.

• Fill out the rest of the fields with the appropriate information (explained in the quick start).

• You will need to select a GPU in the field for doing so since TensorFlow uses GPUs.

• Launch the job through the methods described in the quick start.

• Once in an interactive TensorFlow session using the following command:

> ipython -i --no-banner

Interactive Command-Line Session

• If you wish to use TensorFlow in an interactive command-line session, you can do so with the following command.

bsub -Is -R 'gpuhost' -gpu "num=1" -q general-interactive -a 'docker(gcr.io/ris-registry-shared/tensorflow:<tag>)' /bin/bash

Extend the TensorFlow Image

You may wish to extend the TensorFlow container with additional Python packages. Below is an example Dockerfile which extends the latest RIS-hosted TensorFlow image with the following packages:

• pandas (https://pandas.pydata.org/)

• matplotlib (https://matplotlib.org/)

Dockerfile

FROM gcr.io/ris-registry-shared/tensorflow:latest
RUN pip install pandas matplotlib

• Please see this section for more information on building and pushing your own custom Docker container.

Supported TensorFlow Versions

Version	Docker Image Tag
2.6.0 (Python 3.8.10)	latest, 21.10-tf2-py3