BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240626T180033Z
LOCATION:Level 2 Lobby
DTSTART;TZID=America/Los_Angeles:20240626T180000
DTEND;TZID=America/Los_Angeles:20240626T190000
UID:dac_DAC 2024_sess237_RESEARCH1240@linklings.com
SUMMARY:EffiPipe: Towards Energy-Efficient Large-scale Model Training on C
 ommodity GPUs
DESCRIPTION:Work-in-Progress Poster\n\nZijie Tian, Shuo Wang, Yuhao Zhang,
  Youyou Lu, and Jiwu Shu (Tsinghua University)\n\nWith the continuous evol
 ution of large models, large model\ntraining has become increasingly criti
 cal. However, large-\nscale model training typically requires a substantia
 l energy\nconsumption, which adds to the cost of training these mod-\nels.
  We present EffiPipe, an energy-efficient GPU scheduling\nsystem for large
 -scale model training tasks. EffiPipe con-\nducts fine-grained scheduling 
 of operators. Incorporating\ndynamic frequency adjustment for both computi
 ng and mem-\nory, and taking into account distributed model training sce-\
 nnarios.Compared to existing works, we can reduce power\nconsumption by 20
 -30% while ensuring performance is main-\ntained\n\nTopic: AI, Autonomous 
 Systems, Cloud, Design, EDA, Embedded Systems, IP, Security
END:VEVENT
END:VCALENDAR