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:20260730T152638Z
LOCATION:
DTSTART;TZID=America/Los_Angeles:20260728T125300
DTEND;TZID=America/Los_Angeles:20260728T125400
UID:dac_DAC 2026_sess315_RESEARCH3050@linklings.com
SUMMARY:Scalable and Adaptive Parallel Training of Graph Transformer on La
 rge Graphs
DESCRIPTION:Jun-Liang Lin and Kamesh Madduri (Pennsylvania State Universit
 y) and Mahmut Kandemir (PSU)\n\nGraph foundation models have demonstrated 
 remarkable adaptability across diverse downstream tasks through large-scal
 e pretraining on graphs. However, existing implementations of the backbone
  model, graph transformers, are typically limited to single-GPU systems, l
 eading to long training times or out-of-memory issues on large graphs. Mor
 eover, parallelizing graph transformer training over the full graph is cha
 llenging, as efficiency depends heavily on both the graph structure and sy
 stem characteristics, such as bandwidth and memory capacity.\n\nIn this wo
 rk, we introduce a distributed training framework for graph transformers, 
 which automatically selects and optimizes parallelization strategies based
  on the graph structure and hardware configuration. With our implementatio
 n of distributed sparse operations, we accelerate sparse graph attention b
 y up to 3.8x and reduce memory consumption by 78% compared to state-of-the
 -art frameworks.  On large graph benchmarks, our proposed framework achiev
 es up to 6x speedup with system scaling up to 8 GPUs. These results demons
 trate that the proposed framework improves the scalability of graph transf
 ormers, bringing them closer to serving as practical graph foundation mode
 ls.\n\nTopics: Quantum\n\nTrack: DES6. Quantum Computing, Systems\n\nSessi
 on Chairs: Hokeun Kim (Arizona State University); BaekGyu Kim (Daegu Gyeon
 gbuk Institute of Science and Technology); Xiang Chen (Peking University);
  Sangyeob Kim (Yonsei University); Robert M. Radway (University of Pennsyl
 vania); Prashanth Mohan (University of Pittsburgh); Xiaoxuan Yang (Univers
 ity of Virginia, Stanford University); Ramtin Zand (University of South Ca
 rolina); Hadi Kamali (University of Central Florida); Jiafeng Xie (Villano
 va University); Prabuddha Chakraborty (University of Maine); Junichiro Kad
 omoto (University of Tokyo); and Yosuke Ueno (RIKEN)\n\n
END:VEVENT
END:VCALENDAR