ExSAIS 2022: Workshop on Extreme Scaling of AI for Science

June 3, 2022
Co-Located with IPDPS 2022
Virtual Workshop

Due to perduring travel limitations, IPDPS 2022 will still be held as a virtual conference.

The ExSAIS organizing committee has planned an exciting online program, consisting in two keynotes, a panel, and a paper session with three flash talks o June 3 (starting at 8:00 AM PST, 11:00 AM EST, 3:00 PM UTC, 5:00 PM CET) with presentations Q/A for each accepted paper and the keynote. We provide below the tentative schedule. 

Please note that IPDPS will be provide the platform for virtual conferencing, and attendees will need to be registered to IPDPS to attend also the workshops, access the papers, and (optional) presentation materials that speakers would like to share before the live event.

To attend ExSAIS, please register at: http://www.ipdps.org


Program

Time Event
8:00 - 8:05Welcome and Introduction
Svitlana Volkova, Robert Rallo, Mahantesh Halappanavar
8:05 - 9:05Keynote 1 - Chair: Svitlana Volkova

Efficient DNN Training at Scale: from Algorithms to Hardware
Gennady Pekhimenko (University of Toronto)
9:05 - 10:35Panel: Scaling AI for Science - Moderator: Svitlana Volkova
 
Panelists: Draguna Vrabie (PNNL), Gina Tourassi (ORNL), Asmeret Naugle (SNL), Timo Bremer (LLNL), Angie Sheffield (NNSA)
10:35 - 10:45Break
10:45 - 11:30

10:45 - 11:00


11:00 - 11:15


11:15 - 11:30
Lighting Talks

Learning to Scale the Summit: AI for Science on a Leadership Supercomputer
Joubert, Messer, Roth, Georgiadou, Lietz, Eisenbach, Yin

Strategies for Integrating Deep Learning Surrogate Models with HPC Simulation Applications
Yin, Wang, Shankar

A Scalable Pipeline for Gigapixel Whole Slide Imaging Analysis on Leadership Class HPC Systems
Dash, Hernandez, Tsaris, Alamudun, Yoon, Wang
11:30 - 12:30Keynote 2 - Chair: Mahantesh Halappanavar

Training Large Language Models: Challenges and Opportunities
Mostofa Patwary (NVIDIA)
12:30 - 12:35Closing
Robert Rallo

Created with Mobirise ‌

Free Website Designer Software