Example code and applications for machine learning on Graphcore IPUs
This repository contains sample applications, code examples and tutorials for use with Graphcore IPUs.
If you are interested in finding out more about Graphcore, including getting preview access to IPUs to run these examples, please register your interest here.
Please note we are not currently accepting pull requests or issues on this repository. If you are actively using this repository and want to report any issues, please raise a ticket through the Graphcore support portal: https://www.graphcore.ai/support.
The latest version of the documentation for the Poplar software stack, and other developer resources, is available at https://www.graphcore.ai/developer.
The code presented here requires using Poplar SDK 1.4.x
Please install and enable the Poplar SDK following the instructions in the Getting Started guide for your IPU system. Note that for SDK 1.3 and later you no longer need to source a separate
enable.shfor the Graphcore drivers.
Unless otherwise specified by a LICENSE file in a subdirectory, the LICENSE referenced at the top level applies to the files in this repository.
| Example | Link | | ------- | ---- | | BERT (PopART) | code | | BERT (TensorFlow) | code | | DeepVoice3 (PopART) | code | | CNN Training including ResNet, ResNeXt & EfficientNet (TensorFlow) | code | | CNN Inference including ResNet, MobileNet & EfficientNet (TensorFlow) | code | | CNN Training & Inference including ResNet, ResNeXt & EfficientNet (PyTorch) | code | | ResNet Training (PopART) | code | | ResNext Inference (PopART) | code | | Recommendation using Autoencoders (TensorFlow) | code | | Sales forecasting example (TensorFlow) | code | | Contrastive Divergence VAE using MCMC methods (TensorFlow) | code | | Example reinforcement learning policy model (TensorFlow)| code | | Click through rate: Deep Interest Network (TensorFlow) | code | | Dynamic Sparsity: MNIST RigL (TensorFlow) | code | | Dynamic Sparsity: Autoregressive Language Modelling (TensorFlow) | code |
applications/folder contains example applications written in different frameworks targeting the IPU. See the READMEs in each folder for details on how to use these applications.
code_examples/folder contains small code examples showing you how to use various software features when developing for IPUs. See the READMEs in each folder for details.
tutorials/folder contains tutorials to help you get started using the Poplar SDK and Graphcore tools. Currently, this contains:
tutorials/poplar- A set of tutorials to introduce the Poplar graph programming framework and the PopLibs libraries.
tutorials/pytorch- A tutorial to introduce the PyTorch framework support for the IPU.
tutorials/popvision- A set of tutorials to introduce the PopVision libraries.
The README files for the tutorials are best viewed on GitHub.
utils/folder contains utilities libraries and scripts that are used across the other code examples. This includes:
utils/examples_tests- Common Python helper functions for the repository's unit tests.
utils/benchmarks- Common Python helper functions for running benchmarks on the IPU in different frameworks.