Presented By O’Reilly and Intel Nervana
Put AI to work
September 17-18, 2017: Training
September 18-20, 2017: Tutorials & Conference
San Francisco, CA

Using GPU acceleration with PyTorch to make your algorithms 2,000% faster

Jeremy Howard (fast.ai)
4:00pm–4:40pm Tuesday, September 19, 2017
Implementing AI
Location: Grand Ballroom Level: Intermediate
Secondary topics:  Algorithms, Tools and frameworks
Average rating: ****.
(4.00, 1 rating)

Prerequisite Knowledge

  • A working knowledge of Python and NumPy

What you'll learn

  • Understand what kinds of algorithms can be accelerated on a GPU and what hardware and software you need to get started
  • Learn how to port a standard NumPy algorithm to run on a GPU, using PyTorch

Description

Although most devs are aware of the benefits of GPU acceleration, many assume that the technique is only applicable to specialist areas like deep learning and that learning to program a GPU takes complex specialist knowledge. It turns out that neither assumption is true. Nearly any nonrecursive algorithm that operates on datasets of 1,000+ items can be accelerated by a GPU. And recent libraries like PyTorch make it nearly as simple to write a GPU-accelerated algorithm as a regular CPU algorithm.

Jeremy Howard explains how to easily harness the power of a GPU using a standard clustering algorithm with PyTorch and demonstrates how to get a 20x speedup over NumPy. Along the way, Jeremy covers the mean-shift clustering algorithm and why it’s important for many data science applications.

Photo of Jeremy Howard

Jeremy Howard

fast.ai

Jeremy Howard is a founding researcher at fast.ai (a research institute dedicated to making deep learning more accessible), a distinguished research scientist at the University of San Francisco, a faculty member at Singularity University, and a young global leader with the World Economic Forum. An entrepreneur, business strategist, developer, and educator, Jeremy has served as the president and chief scientist of the data science platform Kaggle, where he was the top-ranked participant in international machine learning competitions for two years running; the founding CEO of two successful Australian startups, FastMail and Optimal Decisions Group (acquired by Lexis-Nexis); and a management consultant at McKinsey & Company and AT Kearney. Jeremy has contributed to a number of open source projects and created, invested in, mentored, and advised many startups. His most recent startup, Enlitic, was the first company to apply deep learning to medicine and has twice been named one of the world’s top 50 smartest companies by MIT Tech Review. Jeremy is a frequent guest on television and other video; he appeared regularly on Australia’s highest-rated breakfast news program, gave a talk on TED.com (which has over two million views), and has led quite a few data science and web development tutorials and discussions.

Comments on this page are now closed.

Comments

Stefan Beskow | DATA SCIENTIST
09/20/2017 4:03am PDT

Great talk. Is the Jupiter notebook presented in the talk available on GitHub?

Thanks,
Stefan