High-performance training and inference for deep equivariant interatomic potentials

Albert Musaelian, Tess Smidt, William C. Witt, Anders Johansson, Chuin Wei Tan, Marc Descoteaux, Mit Kotak, Gabriel de Miranda Nascimento, Seán R. Kavanagh, Laura Zichi, Menghang Wang, Aadit Saluja, Yizhong R. Hu, Boris Kozinsky

2026-01-01

SCID: 54.1/yqu82g6e

Abstract (AI)

The NequIP framework is redesigned for scalable distributed training and PyTorch 2.0 compilation. AOT Inductor inference and optimized Allegro kernels accelerate molecular dynamics by factors of 5–18 on practical system sizes.

Key Findings

Research Object

Research Subject

Publication Details

DOI

10.1039/d5dd00423c

Publication Date

2026-01-01

Journal

Publisher

ISSN

Access Type

Author Information

Authors

Albert Musaelian

Tess Smidt

William C. Witt

Anders Johansson

Chuin Wei Tan

Marc Descoteaux

Mit Kotak

Gabriel de Miranda Nascimento

Seán R. Kavanagh

Laura Zichi

Menghang Wang

Aadit Saluja

Yizhong R. Hu

Boris Kozinsky

Download PDF