Distributed Equivariant Graph Neural Networks for Large-Scale Electronic Structure Prediction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kaniselvan, Manasa, Maeder, Alexander, Xia, Chen Hao, Ziogas, Alexandros Nikolaos, Luisier, Mathieu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915374270775296
author Kaniselvan, Manasa
Maeder, Alexander
Xia, Chen Hao
Ziogas, Alexandros Nikolaos
Luisier, Mathieu
author_facet Kaniselvan, Manasa
Maeder, Alexander
Xia, Chen Hao
Ziogas, Alexandros Nikolaos
Luisier, Mathieu
contents Equivariant Graph Neural Networks (eGNNs) trained on density-functional theory (DFT) data can potentially perform electronic structure prediction at unprecedented scales, enabling investigation of the electronic properties of materials with extended defects, interfaces, or exhibiting disordered phases. However, as interactions between atomic orbitals typically extend over 10+ angstroms, the graph representations required for this task tend to be densely connected, and the memory requirements to perform training and inference on these large structures can exceed the limits of modern GPUs. Here we present a distributed eGNN implementation which leverages direct GPU communication and introduce a partitioning strategy of the input graph to reduce the number of embedding exchanges between GPUs. Our implementation shows strong scaling up to 128 GPUs, and weak scaling up to 512 GPUs with 87% parallel efficiency for structures with 3,000 to 190,000 atoms on the Alps supercomputer.
format Preprint
id arxiv_https___arxiv_org_abs_2507_03840
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Distributed Equivariant Graph Neural Networks for Large-Scale Electronic Structure Prediction
Kaniselvan, Manasa
Maeder, Alexander
Xia, Chen Hao
Ziogas, Alexandros Nikolaos
Luisier, Mathieu
Machine Learning
Materials Science
Distributed, Parallel, and Cluster Computing
Computational Physics
Equivariant Graph Neural Networks (eGNNs) trained on density-functional theory (DFT) data can potentially perform electronic structure prediction at unprecedented scales, enabling investigation of the electronic properties of materials with extended defects, interfaces, or exhibiting disordered phases. However, as interactions between atomic orbitals typically extend over 10+ angstroms, the graph representations required for this task tend to be densely connected, and the memory requirements to perform training and inference on these large structures can exceed the limits of modern GPUs. Here we present a distributed eGNN implementation which leverages direct GPU communication and introduce a partitioning strategy of the input graph to reduce the number of embedding exchanges between GPUs. Our implementation shows strong scaling up to 128 GPUs, and weak scaling up to 512 GPUs with 87% parallel efficiency for structures with 3,000 to 190,000 atoms on the Alps supercomputer.
title Distributed Equivariant Graph Neural Networks for Large-Scale Electronic Structure Prediction
topic Machine Learning
Materials Science
Distributed, Parallel, and Cluster Computing
Computational Physics
url https://arxiv.org/abs/2507.03840