TY - JOUR
T1 - dpdata
T2 - A Scalable Python Toolkit for Atomistic Machine Learning Data Sets
AU - 39 authors, inclduing
AU - Zeng, Jinzhe
AU - Peng, Xingliang
AU - Zhuang, Yong-Bin
AU - Wang, Haidi
AU - Yuan, Fengbo
AU - Zhang, Duo
AU - Liu, Renxi
AU - Wang, Yingze
AU - Tuo, Ping
AU - Zhang, Yuzhi
AU - Chen, Yixiao
AU - Li, Yifan
AU - Nguyen, Cao Thang
AU - Huang, Jiameng
AU - Peng, Anyang
AU - Rynik, Marian
AU - Xu, Wei-Hong
AU - Zhang, Zezhong
AU - Zhou, Xu-Yuan
AU - Chen, Tao
AU - Fan, Jiahao
AU - Jiang, Wanrun
AU - Li, Bowen
AU - Li, Denan
AU - Li, Haoxi
AU - Liang, Wenshuo
AU - Liao, Ruihao
AU - Liu, Liping
AU - Luo, Chenxing
AU - Ward, Logan
AU - Wan, Kaiwei
AU - Wang, Junjie
AU - Xiang, Pan
AU - Zhang, Chengqian
AU - Zhang, Jinchao
AU - Zhou, Rui
AU - Zhu, Jia-Xin
AU - Zhang, Linfeng
AU - Wang, Han
PY - 2025/11/10
Y1 - 2025/11/10
N2 - Seamless management of atomistic data sets is a critical prerequisite for the successful development and deployment of machine learning potentials (MLPs). Here, we present dpdata, an open-source Python library designed to streamline every aspect of MLP data handling. Built upon a flexible, plugin-based architecture, dpdata supports reading, writing, and converting between a broad range of file formats─from popular quantum-chemistry packages and molecular-dynamics engines to specialized MLP frameworks. Users may define custom data types, formats, drivers, and minimizers, enabling effortless extension to emerging software. Key utilities include automated train-test splitting, coordinate perturbation for active learning, outlier-energy removal, Δ-learning data set generation, error-metric computation, and unit conversion. Through efficient NumPy-backed storage and system-level operations, dpdata achieves significant memory saving and inference speedups over configuration-by-configuration tools such as ASE. We also highlight practical impact, with dpdata used across published studies, for format conversion, data storage, coordinate perturbation, and utilization in other projects for data processing. © 2025 American Chemical Society.
AB - Seamless management of atomistic data sets is a critical prerequisite for the successful development and deployment of machine learning potentials (MLPs). Here, we present dpdata, an open-source Python library designed to streamline every aspect of MLP data handling. Built upon a flexible, plugin-based architecture, dpdata supports reading, writing, and converting between a broad range of file formats─from popular quantum-chemistry packages and molecular-dynamics engines to specialized MLP frameworks. Users may define custom data types, formats, drivers, and minimizers, enabling effortless extension to emerging software. Key utilities include automated train-test splitting, coordinate perturbation for active learning, outlier-energy removal, Δ-learning data set generation, error-metric computation, and unit conversion. Through efficient NumPy-backed storage and system-level operations, dpdata achieves significant memory saving and inference speedups over configuration-by-configuration tools such as ASE. We also highlight practical impact, with dpdata used across published studies, for format conversion, data storage, coordinate perturbation, and utilization in other projects for data processing. © 2025 American Chemical Society.
UR - https://www.webofscience.com/wos/woscc/full-record/WOS:001596613000001
UR - https://www.scopus.com/pages/publications/105021082353
UR - https://www.scopus.com/pages/publications/105021082353#tab=impact
U2 - 10.1021/acs.jcim.5c01767
DO - 10.1021/acs.jcim.5c01767
M3 - RGC 21 - Publication in refereed journal
SN - 1549-9596
VL - 65
SP - 11497
EP - 11504
JO - Journal of Chemical Information and Modeling
JF - Journal of Chemical Information and Modeling
IS - 21
ER -