BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:High-performance Artificial Intelligence. Scaling DL on Supercompu
 ters. - Matt Archer\, UIS
DTSTART:20181016T120000Z
DTEND:20181016T130000Z
UID:TALK111418@talks.cam.ac.uk
CONTACT:Chris Richardson
DESCRIPTION:UK researchers can make use of high-performance computing (HPC
 ) clusters at a variety of UK universities. These clusters contain thousan
 ds of nodes connected to multi-petabyte file systems\, that can deliver qu
 adrillions of operations every second. A researcher taking advantage of th
 ese resources could train more sophisticated models\, on more data\, faste
 r than would be otherwise possible.\nThis talk explores the challenges of 
 training large models on a modern distributed-memory cluster. We shall pre
 sent the science applications that motivates and continues to inspire inte
 rest in this\, describe how to setup DL training on an HPC cluster and pre
 sent standard benchmark training results.
LOCATION:JJ Thomson Seminar Room\, Maxwell Centre\, Cavendish Laboratory
END:VEVENT
END:VCALENDAR