Skip to content

Details

Description: In the world of machine learning, a lot of attention is paid to optimizing training. There is a lot less information out there on optimizing prediction. Yet serving models for prediction is where we make our money in ML! The cost of serving predictions may be a major factor in the total return-on-investment for an ML application. In this session we will show you some ways to optimize TensorFlow models for serving predictions, to help you reduce the cost and increase the performance of your ML solution. We will work with both GraphDef and SavedModel formats, learning how to perform optimizations such as freezing, pruning, constant folding, folding batch norms and quantization. The optimization process we discuss will suffice to perform basic serving optimization for a broad array of models.

Speaker Bio: Lukman Ramsey is a Solution Architect for AI and ML in Google Cloud. He joined Google after acting as CTO for a series of New York startups in education technology, AI and speech recognition. Lukman did his PhD work in Computer Science at UC San Diego, doing computer vision research at JPL in the Machine Learning systems group. As an undergrad at MIT in Cognitive Science and Computer Science he worked on multi-layer neural networks before it was called "deep learning." His academic interests have always centered on the intersection of Cognitive Science and Computer Science.

Agenda
5:45 - 6:30 PM Networking/Social/Food
6:30 - 6:35 PM Introduction
6:35 - 7:30 PM Talk and Q&A
7:30 - 8:00 PM Networking/Social/Closing

Related topics

You may also like