High-performance training and inference for deep equivariant interatomic potentials

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Tan, Chuin Wei, Descoteaux, Marc L., Kotak, Mit, Nascimento, Gabriel de Miranda, Kavanagh, Seán R., Zichi, Laura, Wang, Menghang, Saluja, Aadit, Hu, Yizhong R., Smidt, Tess, Johansson, Anders, Witt, William C., Kozinsky, Boris, Musaelian, Albert
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913803491344384
author Tan, Chuin Wei
Descoteaux, Marc L.
Kotak, Mit
Nascimento, Gabriel de Miranda
Kavanagh, Seán R.
Zichi, Laura
Wang, Menghang
Saluja, Aadit
Hu, Yizhong R.
Smidt, Tess
Johansson, Anders
Witt, William C.
Kozinsky, Boris
Musaelian, Albert
author_facet Tan, Chuin Wei
Descoteaux, Marc L.
Kotak, Mit
Nascimento, Gabriel de Miranda
Kavanagh, Seán R.
Zichi, Laura
Wang, Menghang
Saluja, Aadit
Hu, Yizhong R.
Smidt, Tess
Johansson, Anders
Witt, William C.
Kozinsky, Boris
Musaelian, Albert
contents Machine learning interatomic potentials, particularly those based on deep equivariant neural networks, have demonstrated state-of-the-art accuracy and computational efficiency in atomistic modeling tasks like molecular dynamics and high-throughput screening. The size of datasets and demands of downstream workflows are growing rapidly, making robust and scalable software essential. This work presents a major overhaul of the NequIP framework focusing on multi-node parallelism, computational performance, and extensibility. The redesigned framework supports distributed training on large datasets and removes barriers preventing full utilization of the PyTorch 2.0 compiler at train time. We demonstrate this acceleration in a case study by training Allegro models on the SPICE 2 dataset of organic molecular systems. For inference, we introduce the first end-to-end infrastructure that uses the PyTorch Ahead-of-Time Inductor compiler for machine learning interatomic potentials. Additionally, we implement a custom kernel for the Allegro model's most expensive operation, the tensor product. Together, these advancements speed up molecular dynamics calculations on system sizes of practical relevance by up to a factor of 18.
format Preprint
id arxiv_https___arxiv_org_abs_2504_16068
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle High-performance training and inference for deep equivariant interatomic potentials
Tan, Chuin Wei
Descoteaux, Marc L.
Kotak, Mit
Nascimento, Gabriel de Miranda
Kavanagh, Seán R.
Zichi, Laura
Wang, Menghang
Saluja, Aadit
Hu, Yizhong R.
Smidt, Tess
Johansson, Anders
Witt, William C.
Kozinsky, Boris
Musaelian, Albert
Computational Physics
Machine Learning
Chemical Physics
Machine learning interatomic potentials, particularly those based on deep equivariant neural networks, have demonstrated state-of-the-art accuracy and computational efficiency in atomistic modeling tasks like molecular dynamics and high-throughput screening. The size of datasets and demands of downstream workflows are growing rapidly, making robust and scalable software essential. This work presents a major overhaul of the NequIP framework focusing on multi-node parallelism, computational performance, and extensibility. The redesigned framework supports distributed training on large datasets and removes barriers preventing full utilization of the PyTorch 2.0 compiler at train time. We demonstrate this acceleration in a case study by training Allegro models on the SPICE 2 dataset of organic molecular systems. For inference, we introduce the first end-to-end infrastructure that uses the PyTorch Ahead-of-Time Inductor compiler for machine learning interatomic potentials. Additionally, we implement a custom kernel for the Allegro model's most expensive operation, the tensor product. Together, these advancements speed up molecular dynamics calculations on system sizes of practical relevance by up to a factor of 18.
title High-performance training and inference for deep equivariant interatomic potentials
topic Computational Physics
Machine Learning
Chemical Physics
url https://arxiv.org/abs/2504.16068