Deep Dives: The 2023 Energy HPC Conference Workshops
The 2023 Energy HPC Conference, hosted by the Ken Kennedy Institute at the BioScience Research Collaborative in Houston, TX, featured three post-conference workshops on March 2, 2023 where attendees could choose their own agenda, whether it be through an interest in HPC trends and best practices, Devito code training, or physics-informed machine learning through NVIDIA Modulus.
View the recap of the main conference sessions on February 28 — March 1 here.
Best Practices in HPC Systems Management
The “Best Practices in HPC Systems Management” workshop, organized by conference founder and HPC advisor at Intel, Keith Gray, included panels discussing trends, best practices, and challenges discovered across facilities, as well as sessions dedicated to standardized cross-platform benchmarking, SysAdmin automation, Linux distribution, telemetry, and monitoring. The workshop heard from practitioners and experts from industry, academia, and national labs.
Workshop Speakers:
- Kent Blancett, bp
- Gerard Gorman, Devito Codes
- Fabio Luporini, Devito Codes
- Wade Vinson, HPE
- Keith Gray, Intel
- Tim Osborne, ORNL
- David Baldwin, Shell
- Tommy Minyard, TACC
- Donny Cooper, TotalEnergies
- Russell Jones, Total Energies
Devito Training
Those interested in expanding their technical skill could attend the “Devito Training” workshop organized by Devito Codes co-founders Gerard Gorman and Fabio Luporini, as well as representatives from Chevron and Imperial College London. Attendees learned to develop finite-difference solvers and design a basic FWI (Full‐Waveform Inversion) framework with Devito, a Domain-Specific Language (DSL) and code generation framework. Through a parallel track, attendees also discussed standardized cross-platform benchmarking and learned to develop new benchmarks.
Workshop Speakers:
- Luke Decker, Chevron
- John Washbourne, Chevron
- Gerard Gorman, Devito Codes
- Fabio Luporini, Devito Codes
- Rhodri Nelson, Imperial College London
- Ed Caunt, Imperial College London
- Zoe Leibowitz, Imperial College London
Introduction to Physics-Informed Machine Learning with Modulus
The “Best Practices” and “Devito Training” workshops are returning themes for the Energy HPC Conference, but they were joined by an additional theme in 2023: “Introduction to Physics-Informed Machine Learning with Modulus.” This workshop was organized by NVIDIA representatives including Pavel Dimitrov, Sr Solutions Architect. The workshop introduced NVIDIA Modulus, a machine learning platform that turbocharges quick iterative use cases by building physics-based deep learning models — that are 100,000x faster than traditional methods — and offers high-fidelity simulation results. Attendees learned techniques to solve data-driven and physics-driven problems using Modulus through hands-on application sessions, and they left with an understanding of how the Modulus framework integrates with the overall Omniverse platform.
Workshop Speakers:
- Pavel Dimitrov, NVIDIA
- Ken Hester, NVIDIA
- Fatmir Hoxha, NVIDIA
- Guillaume Barnier, NVIDIA
- Guillaume Thomas Collignon, NVIDIA
- Igor Terentyev, NVIDIA
- Kashif Chauhan, NVIDIA
ICYMI: Recorded Sessions
You can watch the recordings of presentations from each of the three workshops on March 2, as well as the main Energy HPC Conference sessions from February 28 — March 1, 2023, on the Ken Kennedy Institute YouTube Channel. See what you missed or go back to watch key insights from our invited speakers, technical sessions, and workshop discussions.
View the recorded sessions here:
Conference Website: energyhpc.rice.edu
View 2023 Program: https://bit.ly/3Jrv8V3
Save the Date: 2024 Energy HPC Conference | March 5–7, 2024
Author: The Ken Kennedy Institute