BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250822T115805Z
LOCATION:Campussaal - Plenary Room
DTSTART;TZID=Europe/Stockholm:20250616T102000
DTEND;TZID=Europe/Stockholm:20250616T105000
UID:submissions.pasc-conference.org_PASC25_sess149_pos110@linklings.com
SUMMARY:P11 - Enabling Lattice QCD Normalizing Flows in HPC Infrastructure
 s
DESCRIPTION:Matteo Bunino (CERN), Isabel Campos Plasencia (IFCA/CSIC), Jav
 ad Komijani and Marina Marinkovic (ETH Zurich), Gaurav Sinha Ray (IFCA/CSI
 C), Rakesh Sarma (Forschungszentrum Jülich), and Jarl Sondre Saether (CERN
 )\n\nThe Horizon Europe project interTwin aims at developing a prototype f
 or a multidisciplinary Digital Twin Engine, applicable across a whole spec
 trum of scientific disciplines: High Energy Physics (HEP), Environment, Cl
 imate, etc. As part of this effort we explore the extent to which Machine 
 Learning (ML) methods can speed up Lattice Gauge Theory Simulations in cha
 llenging areas of the parameter space where Monte Carlo methods suffer fro
 m severe critical slowing down. The overall goal is progressing towards de
 signing the digital twin of a HEP detector, where Lattice QCD simulations 
 could provide future realistic simulations of the Standard model. We are e
 xploiting the advantages of the tools developed in the project interTwin, 
 notably intertwinai, to scale up and support the deployment of our simulat
 ions in HPC systems, while enabling as well several code features. The itw
 inai toolkit provides functionalities for distributed machine learning on 
 HPC, supporting different distributed frameworks (DeepSpeed, Horovod, and 
 PyTorch DistributedDataParallel) implementing different communication prot
 ocols across different GPUs, suited to different infrastructures. Furtherm
 ore, itwinai also offers a profiling feature based on the PyTorch profilin
 g backend, enabling it to identify communication and computation shares. T
 his profiler will enable the identification of bottlenecks, and hence opti
 mize the code to improve performance.\n\nSession Chair: Chris Cantwell (Im
 perial College London)\n\n
END:VEVENT
END:VCALENDAR