Authors: Xiangyu Zhang (SKLP, Institute of Computing Technology, Chinese Academy of Sciences; University of Chinese Academy of Sciences); Junwei Feng (University of Science and Technology of China); Jielan Li (Shanghai AI Laboratory); Chen Wang and Yan Wang (SKLP, Institute of Computing Technology, Chinese Academy of Sciences; University of Chinese Academy of Sciences); Yusu Zhang (School of Mathematical Sciences, Shanghai Jiao Tong University); Hongxia Hao (Shanghai AI Laboratory); Wei Hu (University of Science and Technology of China); and Weile Jia (SKLP, Institute of Computing Technology, Chinese Academy of Sciences; University of Chinese Academy of Sciences)
Abstract: Machine learning interatomic potentials have enabled molecular dynamics with ab initio accuracy at greatly reduced cost and at large scale. The field is now shifting toward foundation machine learning interatomic potentials with broader transferability, but these models are harder to scale because message-passing layers complicate distributed communication and two-body and three-body features increase graph-construction cost and memory footprint. We present MatterScale, an end-to-end framework for deploying foundation MLIPs on massively parallel molecular dynamics workloads on a next-generation supercomputer. MatterScale combines a Halo operator for nearest-neighbor inter-layer feature exchange, an ML-based multiple-time-step scheme for higher throughput, layer-wise activation checkpointing for higher per-node capacity, and a parallel graph-construction algorithm. With MatterSim, MatterScale scales to 32,768 GPGPUs with 94% weak scaling efficiency and simulates up to 1.13 billion atoms.
Presentation: file