Advancing Life & Work

Labs launches HPE Deep Learning Performance Guide

Senior research scientist Sergey Serebryakov and senior research manager Natalia VassilievaSenior research scientist Sergey Serebryakov and senior research manager Natalia Vassilieva

By Curt Hopkins, Managing Editor, Hewlett Packard Labs

At the upcoming NVIDIA GPU Technology Conference, taking place in San Jose March 26-29, Labs’ Senior Research Manager Natalia Vassilieva and Senior Research Scientist Sergey Serebryakov will present a talk titled HPE Deep Learning Cookbook: Recipes to Run Deep Learning Workloads

At the GTC, Vassilieva, who heads AI research for Labs, will debut an important component of Labs’ Deep Learning Cookbook – the Deep Learning Performance Guide – along with new benchmarking results.

The HPE Deep Learning Cookbook is a set of tools to evaluate deep learning workloads and provide a choice of an optimal hardware and software stack for any given workload.  When HPE announced its Deep Learning Cookbook back in the fall, we launched the first component of the Cookbook – the HPE Deep Learning Benchmarking Suite.

Now, the HPE Deep Learning Performance Guide will provide users with an interface with which to look at and explore benchmarking data collected with the first component, our Benchmarking Suite.

“The flexibility of the Performance Guide allows users to slice and dice existing benchmarking data as they want, creating a wide range of plots to visualize different aspects and dependencies of performance,” says Vassilieva. “It’s pretty unique because provides an easy way to do side-by-side comparison of multiple hardware options, deep learning frameworks, runtimes, and models, all in one place. We’ve collected quite a significant amount of performance data and we’re excited to make it available to everyone.”

As explained in the talk abstract, the Deep Learning Cookbook was built on the understanding that, given how deep learning has become a key enabling technology behind the recent revival of artificial intelligence, and with the variety of choices in hardware configurations and software packages, it is hard to pick the most optimal tools. The Deep Learning Cookbook provides a set of open source tools to guide the choice of the best hardware/software environment for a given deep learning task based on extensive benchmarks of reference deep learning workloads and performance modelling.

Attend Vassilieva’s and Serebryakov’s talk on Thursday, Mar 29, 11:00 AM-11:50 AM in Room 211A of San Jose’s McEnery Convention Center..

Read more on HPE's AI offerings,.

About the Author


Managing Editor, Hewlett Packard Labs