Google Turns Intel oneDNN AI Libraries as Default on TensorFlow

Google has turned-on Intel oneDNN as a default library on the latest version of its open source artificial intelligence and machine learning software platform, TensorFlow 2.9.

Intel oneAPI Deep Neural Network Library, oneDNN, is also an open source cross-platform performance library of basic deep learning building blocks intended for developers of deep learning applications and frameworks.

oneDNN is part of oneAPI, an open, standards-based, unified programming model for use across CPUs as well as GPUs for machine learning and, in particular, deep learning. Intel released its oneAPI 2022 toolkits about six months ago, providing developers with tools to help improve the productivity and velocity of code development including new capabilities such as a unified compiler implementing C++, SYCL and Fortran, and data parallel Python for CPUs and GPUs.

“Thanks to the years of close engineering collaboration between Intel and Google, optimizations in the oneDNN library are now default for x86 CPU packages in TensorFlow,” said   Wei Li, vice president and general manager of AI and Analytics, Intel.

“This brings significant performance acceleration to the work of millions of TensorFlow developers without the need for them to change any of their code. This is a critical step to deliver faster AI inference and training and will help drive AI Everywhere.”

The setting of oneDNN as default on TensorFlow 2.9 applies to all Linux x86 packages and for CPUs with neural-network-focused hardware features including AVX512_VNNI, AVX512_BF16, and AMX vector and matrix extensions found on 2nd Gen Intel Xeon® processors and newer CPUs.

Prior to the latest release, Google has included experimental support for oneDNN since TensorFlow 2.5. Additional TensorFlow-based applications, including TensorFlow Extended, TensorFlow Hub and TensorFlow Serving also have the oneDNN optimizations.

As announced, Intel-optimized TensorFlow is available both as a standalone component and through the Intel oneAPI AI Analytics Toolkit, and is already being used across a range of industry applications including the Google Health project, animation filmmaking at Laika Studios, language translation at Lilt, and natural language processing at IBM Watson.