BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260522T150117Z
LOCATION:C2/3/4 Ballroom
DTSTART;TZID=America/Chicago:20181113T083000
DTEND;TZID=America/Chicago:20181113T170000
UID:submissions.supercomputing.org_SC18_sess322_post197@linklings.com
SUMMARY:Multi-Client DeepIO for Large-Scale Deep Learning on HPC Systems
DESCRIPTION:Yue Zhu, Fahim Chowdhury, and Huansong Fu (Florida State Unive
 rsity); Adam Moody, Kathryn Mohror, and Kento Sato (Lawrence Livermore Nat
 ional Laboratory); and Weikuan Yu (Florida State University)\n\nWith the g
 rowth of computation power, leadership High-Performance Computing (HPC) sy
 stems can train larger datasets for Deep neural networks (DNNs) more effic
 iently. On HPC systems, a training dataset is on a parallel file system or
  node-local storage devices. However, not all HPC clusters have node-local
  storage, and large mini-batch sizes stress the read performance of parall
 el systems since the large datasets cannot fit in file system caches. Thus
 , it is a challenge for training DNNs with large datasets on HPC systems.\
 n\nIn prior work, we proposed DeepIO to mitigate the I/O pressure. DeepIO 
 is designed to assist the mini-batch generation of TensorFlow. However, De
 epIO does not support multiple training workers on a single compute node. 
 We address this gap with modification on DeepIO framework, and evaluate mu
 lti-client DeepIO performance against state-of-the-art in-memory file syst
 ems, compare DeepIO and TensorFlow data loading API, and explore the poten
 tial of DeepIO in DNN training.\n\nRegistration Category: Tech Program Reg
  Pass, Exhibits Reg Pass\n\n
END:VEVENT
END:VCALENDAR