2022-10-24 SciML GPU Bootcamp (hbcp1w22)

Registration

Please apply via the www.openhackathons.org https://www.openhackathons.org/s/siteevent/a0C5e000005UPv0EAG/se000141 registration page with your official e-mail address to prove your affiliation.
The final participants will be selected and informed after the registration deadline has passed.

Contents

Deep Learning (DL) has revolutionized the way of performing classification, pattern recognition, and regression tasks in various application areas. Scientific applications solving linear and non-linear equations with demanding accuracy and computational performance requirements can use a class of DL networks, called Physics-Informed Neural Networks (PINN). In fact, PINNs are specifically designed to integrate scientific computing equations, such as Ordinary Differential Equations (ODE), Partial Differential Equations (PDE), non-linear and integral-differential equations into the DL network training. This workshop introduce Scientific Machine Learning (SciML) with PINN and hands-on experience with a PDE solver NVIDIA Modulus, a neural network framework that blends the power of physics in the form of governing partial differential equations (PDEs) with data to build high-fidelity, parameterized surrogate models with near-real-time latency. This online Bootcamp is a hands-on learning experience where you will be guided through step-by-step instructions with teaching assistants on hand to help throughout.

The Bootcamp is co-organised by HLRS, LRZ, OpenACC.org and NVIDIA for EuroCC@GCS, the German National Competence Centre for High-Performance Computing.

Preliminary Agenda (all times in CEST)

Day 1 - Monday, October 24, 2022

• 09:00 AM - 09:15 AM: Welcome (Moderator and Host)
• 09:15 AM - 10:15 AM: Connecting to a Cluster
• 10:15 AM - 11:15 AM: Invited Talk by Dr. Niki Loppi, NVIDIA
  • Title: Physics-Informed Neural Networks with NVIDIA Modulus Application to External Flow Problems
• 11:15 AM - 11:30 AM: Break
• 11:30 AM - 12:30 PM: Invited Talk by Dr. Alexander Heinlein, Delft University of Technology (TU Delft)
  • Title: Neural networks with physical constraints, domain decomposition-based training strategies, and model order reduction

Day 2 - Tuesday, October 25, 2022

• 09:00 AM - 09:05 AM: Welcome (Moderator and Host)
• 09:05 AM - 09:35 AM: Introduction: Data Driven vs PINN Approach (Lecture)
• 09:35 AM - 10:05 AM: What is NVIDIA Modulus? (Lecture)
• 10:05 AM - 11:35 AM: Lab 1: Solving Partial Differential Equations using Modulus
• 11:35 AM - 12:20 AM: Lab 2: Solving Transient Problems and Inverse Problems using Modulus
• 12:20 AM - 01:20 PM: Break
• 01:20 PM - 02:05 PM: Lab 2 (Continued)
• 02:05 PM - 05:05 PM: Mini Challenge
• 05:05 PM - 05:30 PM: Wrap up and Q&A

Prerequisites

Basic experience with Python. No GPU programming or AI knowledge is required.

Hands-On

Attendees will be given access to a GPU cluster for the duration of the Bootcamp.

Language

English

Lecturers

Dr. Niki Andreas Loppi (NVIDIA)
Dr. Flavio C. Cunha Galeazzo, Dr. Khatuna Kakhiani, Dr. Lorenzo Zanon (HLRS)
PD Dr. Juan Durillo Barrionuevo, Maja Piskac (LRZ)
Dr. Mozhgan Kabiri Chimeh, Dr. Paul Graham, Dr. Gunter Roth (NVIDIA)

Course Webpages by the Co-Organisers

• openhackathons.org
• HLRS

Prices and Eligibility

The course is open and free of charge for participants from academia and industry.

Legal Notices

This course is offered in cooperation by HLRS, LRZ, OpenACC.org and NVIDIA. Registration is done via www.gpuhackathons.org hosted by OpenACC-Standard.org. Your registration data will be transferred to these partners. For legal notes see Privacy Policy.