BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160054Z
LOCATION:Track 3
DTSTART;TZID=America/New_York:20201118T100000
DTEND;TZID=America/New_York:20201118T103000
UID:submissions.supercomputing.org_SC20_sess180_pap366@linklings.com
SUMMARY:Optimizing Deep Learning Recommender Systems Training on CPU Clust
 er Architectures
DESCRIPTION:Paper\n\nOptimizing Deep Learning Recommender Systems Training
  on CPU Cluster Architectures\n\nKalamkar, Georganas, Srinivasan, Chen, Sh
 iryaev...\n\nDuring the last two years, the goal of many researchers has b
 een to squeeze the last bit of performance out of HPC systems for AI tasks
 . ResNet50 is no longer a representative workload in 2020. Thus, we focus 
 on Recommender Systems, specifically Facebook's DLRM benchmark, which acco
 unt for most of the AI cycles in cloud computing centers. By enabling it t
 o run on latest CPU hardware and software tailored for HPC, we are able to
  achieve up to two orders of magnitude improvement in performance on a sin
 gle socket compared to the reference CPU implementation, and high scaling 
 efficiency up to 64 sockets, while fitting ultra-large datasets. This pape
 r discusses and analyzes novel optimization and parallelization techniques
  for the various operators in DLRM. Several optimizations (e.g., tensor-co
 ntraction accelerated MLPs, framework MPI progression, BFLOAT16 training w
 ith up to 1.8x speed-up) are general and transferable to many other deep l
 earning topologies.\n\nTag: Accelerators, FPGA, and GPUs, Machine Learning
 , Deep Learning and Artificial Intelligence, Scalable Computing\n\nRegistr
 ation Category: Tech Program Reg Pass
END:VEVENT
END:VCALENDAR