BUILD NEXT GENERATION MACHINE INTELLIGENCE WITH POPLAR®

Introducing the PyTorch API for the IPU.

With PopTorch™ - a simple Python wrapper for PyTorch programs, developers can easily run models directly on Graphcore IPUs with a few lines of extra code.

Learn how to build performant PyTorch applications for training and inference with our latest user guide, tutorials, and code examples.

To find out more about our announcement, read our guest blog with PyTorch.

Full support for TensorFlow 1 & 2 with integrated XLA backend.

The TensorFlow API for IPU includes distributed IPU-specific estimators and optimisers to enhance application performance.

Flexible model building capability with Keras Sequential & Functional API support for TensorFlow 2.

Functional and sequential support for Keras is provided via our TensorFlow 2 integration.

Developers can port many standard Keras models to the IPU, ready for optimization and performance tuning.

Developers can easily import and export models in ONNX format using Poplar Advanced Runtime (PopART™).

PopART uses Python and C++ APIs to allow models built in other ML frameworks to be accelerated on Graphcore’s IPU hardware.

Poplar supports Alibaba Cloud's Heterogeneity-Aware Lowering and Optimization platform (HALO).

HALO is part of Alibaba Cloud's Open Deep Learning API (ODLA), which is designed to rapidly enable heterogeneous computing and meet the growing demand for AI compute acceleration in datacentres.