BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260422T000616Z
LOCATION:401-402
DTSTART;TZID=America/Denver:20231115T133000
DTEND;TZID=America/Denver:20231115T150000
UID:submissions.supercomputing.org_SC23_sess168@linklings.com
SUMMARY:Training in HPC Machine Learning
DESCRIPTION:High Throughput Training of Deep Surrogates from Large Ensembl
 e Runs\n\nRecent years have seen a surge in deep learning approaches to ac
 celerate numerical solvers, which provide faithful but computationally int
 ensive simulations of the physical world. These deep surrogates are genera
 lly trained in a supervised manner from limited amounts of data slowly gen
 erated by the ...\n\n\nLucas Meyer (National Institute for Research in Dig
 ital Science and Technology, Electricité de France); Marc Schouler and Rob
 ert Alexander Caulk (National Institute for Research in Digital Science an
 d Technology); Alejandro Ribes (Electricité de France); and Bruno Raffin (
 National Institute for Research in Digital Science and Technology)\n------
 ---------------\nEasyScale: Elastic Training with Consistent Accuracy and 
 Improved Utilization on GPUs\n\nDistributed synchronized GPU training is c
 ommonly used for deep learning. The resource constraint of using a fixed n
 umber of GPUs makes large-scale training jobs suffer from long queuing tim
 e for resource allocation, and lowers the cluster utilization. Adapting to
  resource elasticity can alleviate t...\n\n\nMingzhen Li (Beihang Universi
 ty), Wencong Xiao (Unaffiliated), Hailong Yang and Biao Sun (Beihang Unive
 rsity), Hanyu Zhao and Shiru Ren (Unaffiliated), Zhongzhi Luan (Beihang Un
 iversity), Xianyan Jia (Unaffiliated), Yi Liu (Beihang University), Yong L
 i and Wei Lin (Unaffiliated), and Depei Qian (Beihang University)\n-------
 --------------\nHanayo: Harnessing Wave-Like Pipeline Parallelism for Enha
 nced Large Model Training Efficiency\n\nLarge-scale language models have b
 ecome increasingly challenging and expensive to train. Among various metho
 ds addressing this issue, Pipeline Parallelism has been widely employed to
  accommodate massive model weights within limited GPU memory. This paper i
 ntroduces Hanayo, a wave-like pipeline paral...\n\n\nZiming Liu, Shenggan 
 Cheng, Haotian Zhou, and Yang You (National University of Singapore)\n\nTa
 g: Artificial Intelligence/Machine Learning\n\nRegistration Category: Tech
  Program Reg Pass\n\nReproducibility Badges: Artifact Available, Artifact 
 Functional, Results Reproduced\n\nSession Chair: Shuaiwen Leon Song (Micro
 soft Corporation)
END:VEVENT
END:VCALENDAR