Journal Article

DPA-2: a large atomic model as a multi-task learner

Luo, Xiaoshan, Huang, Jiameng, Liu, Shi, Zhang, Duo, Liu, Siyuan, Li, Yifan, Zhang, Yuzhi, Dai, Fu-Zhi, York, Darrin M., Li, Bowen, Yang, Yudi, Zhou, Shuo, Peng, Anyang, Liu, Xinzijian, Liu, Jianchuan, Jia, Weile, E, Weinan, Cai, Chun, Zhang, Xiangyu, Lv, Jian, Cheng, Jun, Wang, Xinyan, Chang, Junhan, Jiang, Wanrun, Chen, Mohan, Shan, Yifan, Zeng, Jinzhe, Zhang, Linfeng, Qin, Xuejian, Gong, Fu-Qiang, Zhu, Tong, Wu, Jing, Shi, Mengchao, Zhang, Linshuang, Ke, Guolin, Du, Yiming, Zhong, Zhicheng, Bi, Hangrui, Zhang, Chengqian, Yang, Manyi, Wang, Zhenyu, Yang, Jiyuan, Wang, Han

2024

Overview

The rapid advancements in artificial intelligence (AI) are catalyzing transformative changes in atomic modeling, simulation, and design. AI-driven potential energy models have demonstrated the capability to conduct large-scale, long-duration simulations with the accuracy of ab initio electronic structure methods. However, the model generation process remains a bottleneck for large-scale applications. We propose a shift towards a model-centric ecosystem, wherein a large atomic model (LAM), pre-trained across multiple disciplines, can be efficiently fine-tuned and distilled for various downstream tasks, thereby establishing a new framework for molecular modeling. In this study, we introduce the DPA-2 architecture as a prototype for LAMs. Pre-trained on a diverse array of chemical and materials systems using a multi-task approach, DPA-2 demonstrates superior generalization capabilities across multiple downstream tasks compared to the traditional single-task pre-training and fine-tuning methodologies. Our approach sets the stage for the development and broad application of LAMs in molecular and materials simulation research.

• Publisher: Nature Publishing Group UK,Nature Publishing Group,Nature Portfolio
• Subject: 639/301/1034/1035 / 639/301/1034/1037 / Accuracy / Approximation / Artificial intelligence / Characterization and Evaluation of Materials / Chemistry and Materials Science / Computational Intelligence / Datasets / Developmental stages / Electronic structure / Energy / Labeling / Machine learning / Materials Science / Mathematical and Computational Engineering / Mathematical and Computational Physics / Mathematical Modeling and Industrial Mathematics / Modelling / Molecular modelling / Neural networks / Physical restraints / Potential energy / Simulation / Theoretical / Training