Intel and Google* have been collaborating to deliver optimized machine learning implementations of compute-intensive TensorFlow* operations. For example, convolution filters that require large matrix multiplications.

In this session, Penporn Koanantakook of Google delivers an overview of the Intel and Google collaboration, which includes the Intel® Extension for TensorFlow* and other key AI developer tools—Intel® oneAPI Deep Neural Network Library (oneDNN) and Intel® Neural Compressor.

This session covers:

• Optimizations that have been implemented, such as operation fusion, primitive caching, and vectorization of int8 and bfloat16 data types.
• A live demonstration of the Intel Neural Compressor automatically quantizing a network to improve performance by 4x with a 0.06% accuracy loss.
• An overview of the PluggableDevice mechanism in TensorFlow, co-architected by Intel and Google to deliver a scalable way for developers to add new device support as plug-in packages.

Note This presentation was current as of TensorFlow v2.8. Starting with TensorFlow v2.9, the oneDNN optimizations are on by default, and no longer require the TF_ENABLE_ONEDNN_OPTS=1 variable setting.

Get all of the following as stand-alone products or as part of AI Tools:

• oneDNN: An open source, cross-platform library that provides implementations of deep learning building blocks that use the same API for CPUs, GPUs, or both.
• Intel Extension for TensorFlow: An end-to-end, open source, machine learning platform.
• Intel Neural Compressor: A unified, low-precision inference interface across multiple deep learning frameworks.