Summary of Scalable Training Of Trustworthy and Energy-efficient Predictive Graph Foundation Models For Atomistic Materials Modeling: a Case Study with Hydragnn, by Massimiliano Lupo Pasini et al.
Scalable Training of Trustworthy and Energy-Efficient Predictive Graph Foundation Models for Atomistic Materials Modeling: A Case Study with HydraGNN
by Massimiliano Lupo Pasini, Jong Youl Choi, Kshitij Mehta, Pei Zhang, David Rogers, Jonghyun Bae, Khaled Z. Ibrahim, Ashwin M. Aji, Karl W. Schulz, Jorda Polo, Prasanna Balaprakash
First submitted to arxiv on: 12 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computational Physics (physics.comp-ph)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Our paper introduces scalable, trustworthy, and energy-efficient predictive graph foundation models (GFMs) based on HydraGNN, a multi-headed graph convolutional neural network architecture. HydraGNN enables the computation of graph neural networks (GNNs) at larger scales and with greater data diversity than previously possible. This work discusses optimizations that allowed training GFMs to tens of thousands of GPUs on datasets comprising hundreds of millions of graphs. Our approach employs multi-task learning (MTL) to learn both graph-level and node-level properties of atomistic structures, such as energy and atomic forces. We trained our model using over 154 million atomistic structures and tested its performance on two US-DOE supercomputers: Perlmutter at the National Energy Research Scientific Computing Center and Frontier at Oak Ridge Leadership Computing Facility. The HydraGNN architecture achieves near-linear strong scaling performance using up to 2,000 GPUs on Perlmutter and 16,000 GPUs on Frontier. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary We’ve developed a new way to analyze complex graph data that’s fast, reliable, and energy-efficient. Our method uses something called HydraGNN, which lets us work with really big datasets and lots of data diversity. We also use this method to learn about different properties of atoms and molecules at the same time. To test our approach, we trained it on over 154 million examples and tested it on two powerful supercomputers. Our results show that our method can handle a lot of data quickly and accurately. |
Keywords
* Artificial intelligence * Multi task * Neural network
