simple_applications

Simple applications

This directory contains a number of basic applications written in different frameworks targeting the IPU.

TensorFlow 1

• Classifying hand-written digits from the MNIST dataset is a well-known example of a basic machine learning task. An example of its implementation on IPUs can be found in mnist. This example also shows how to use ipu.dataset_benchmark to determine the maximum achievable throughput for a given dataset.

• Classifying images with the ResNet18 model: this example shows how to perform classification by using a pretrained ResNet18 model on the IPU.

TensorFlow 2

• Simple MNIST training example: This example trains a simple 2-layer fully connected model on the MNIST numeral data set.

PyTorch

Complete examples of models

• Classifying hand-written digits from the MNIST dataset is a well-known example of a basic machine learning task.

Pre-trained models

• Hugging Face's BERT is a pre-trained BERT model made available by Hugging Face and which is implemented in PyTorch. This example consists of running one of the pre-trained BERT model on an IPU for an inference session.

Other examples

• PopART's MNIST is an example on how to export a PyTorch model as an ONNX file and reuse this file with Graphcore's PopART.

Popart

• Simple MNIST Examples: Contains 2 simple models, 1 linear and 1 using convolutions trained on the MNIST dataset.