4. Next steps

For more information about working with TensorFlow 1 on IPUs refer to the following:

Documentation

• The IPU Programmer’s Guide provides an introduction to the IPU architecture, programming model and tools available.

• User guide and API reference for running TensorFlow 1 on the IPU.

• Technical notes covering a range of topics for optimising models for the IPU.

Running applications in Docker

• You can run TensorFlow 1 applications in Docker on a Linux machine using one or more physical IPU devices. Refer to Using IPUs from Docker for more information.

Tutorials and simple examples

All tutorials are available from https://github.com/graphcore/tutorials.

Note

You need to checkout the branch of the tutorials repository that corresponds to the version of the Poplar SDK you are using. The links below link to the SDK 3.0 branch, which is the last release branch for TensorFlow 1 tutorials.

• Tutorials for TensorFlow 1.

• Simple examples for TensorFlow 1.

• Feature examples for TensorFlow 1.

Application and code examples

• All examples are available from https://github.com/graphcore/examples.

  Note

  You need to checkout the tagged version of the examples repository that corresponds to the version of the Poplar SDK you are using. There is a list of the tagged versions. The last tagged version of the examples repository for TensorFlow 1 is v3.0.0.

• Browse the Graphcore Model Garden. You can filter by model type and framework.

Other support

• You can use the tags “ipu”, “poplar” and “tensorflow” when asking questions or looking for answers on StackOverflow.

• Support is available from the Graphcore customer engineering team via the Graphcore support portal.

• For general help, discussions and announcements, please join our Graphcore Slack Community.