[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-120837-en":3,"doc-seo-120837-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},120837,16904993612988,"Olivia Brown","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","TensorMD - Scalable Tensor-Diagram based Machine Learning Interatomic Potential on Heterogeneous Many-Core Processors - paper","TensorMD presents a scalable machine learning interatomic potential for molecular dynamics on heterogeneous many-core processors. By combining physical principles with a tensor-diagram representation, the model delivers more efficient computation and stronger flexibility for integration with other scientific codes. The work also develops portable optimization strategies and provides an extensively optimized TensorMD implementation for the new Sunway supercomputer. The resulting performance enables simulations up to 52 billion atoms with a time-to-solution of 31 ps/step/atom, establishing new HPC+AI+MD records.","arXiv :2310 .08439v2 [physics .comp-ph] 13 Oct 2023  \nTensorMD: Scalable Tensor-Diagram based Machine Learning Interatomic Potential  \non Heterogeneous Many-Core Processors  \nXin Chen, 1, ∗ Yucheng Ouyang,2, ∗ Xin Chen,3 Zhenchuan Chen,2 Rongfen Lin,3 Xingyu Gao, 1 Lifang Wang, 1 Fang Li,3 Yin Liu,2 Honghui Shang,2,† and Haifeng Song 1,‡  \n1 Institute of Applied Physics and Computational Mathematics, Beijing, China  \n2 SKL of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China  \n3 National Research Center of Parallel Computer Engineering and Technology, Beijing, China  \nMolecular dynamics simulations have emerged as a potent tool for investigating the physical properties and kinetic behaviors of materials at the atomic scale, particularly in extreme conditions. Ab initio accuracy is now achievable with machine learning based interatomic potentials. With recent advancements in high-performance computing, highly accurate and large-scale simulations become feasible. This study introduces TensorMD, a new machine learning interatomic potential (MLIP) model that integrates physical principles and tensor diagrams. The tensor formalism provides amore efficient computation and greater flexibility for use with other scientific codes. Additionally, we proposed several portable optimization strategies and developed a highly optimized version for the new Sunway supercomputer. Our optimized TensorMD can achieve unprecedented performance on the new Sunway, enabling simulations of up to 52 billion atoms with a time-to-solution of 31 ps/step/atom, setting new records for HPC + AI + MD.  \nI. INTRODUCTION  \nAtomistic simulation is a powerful theoretical method used to study the physical and chemical properties of materials at the atomic scale. Among these methods, molecular dynamics simulation has become widely used in studying the physical properties of many metal materials and modeling multiphase equations of states. However, the accuracy of molecular dynamics simulation depends on the atomic potential that describes the interaction between atoms.  \nFirst-principles methods are highly accurate but computationally expensive. On the other hand, semiempirical potentials are computationally efficient, but their accuracy is usually not sufficient for quantitative research. In recent years, the development of atomic potentials based on machine learning methods has gained widespread attention. These machine learning interatomic potentials (MLIPs) can approach the accuracy of first-principles methods while remaining computationally efficient, making them suitable for large-scale simulations, even at experimentally observable scales.  \nMLIPs have been used to study the dynamic behavior and extreme physical properties of materials realistically and accurately. For example, Oganov constructed the phase diagram of Uranium[1] in a wide temperaturepressure range with MLIPs. Similarly, Zong et al[2] . studied the martensitic phase transition of Zr based on a modified Behler-Parinello model, while William et al. studied the extreme physical properties of high-density carbon using SNAP potential[3, 4], which directly served the inertial confinement fusion research. By using these  \n∗ These two authors contributed equally  \n† Corresponding [author:shanghonghui@ict.ac.cn](author:shanghonghui@ict.ac.cn)  \n‡ Corresponding author:song˙[haifeng@iapcm.ac.cn](haifeng@iapcm.ac.cn)  \npowerful techniques, scientists can now investigate the behavior of materials at the atomic scale with a high level of accuracy and efficiency, opening up new avenues for research and development. Recently, many traditional first-principles programs, such as VASP[5, 6], have introduced MLIPs based acceleration modules, which can significantly improve the computational efficiency of ab initio molecular dynamics simulations.  \nUntil now, various MLIPs have been proposed, including GAP[7–10], SNAP[3, 11–13], MLFF[5, 6], HDNNP (BP)[14], DP[15–19","cbCaiqEsa1maqiDK","https://ap.wps.com/l/cbCaiqEsa1maqiDK","pdf",814155,1,13,"English","en",105,"# Introduction\n## Molecular dynamics and atomic potentials\n## First-principles vs semiempirical potentials\n## Machine learning interatomic potentials (MLIPs)\n## Prior HPC MLIP work and many-core processors\n## Motivation and contributions of TensorMD","[{\"question\":\"What problem does TensorMD address in molecular dynamics simulations?\",\"answer\":\"TensorMD targets efficient, high-accuracy interatomic potentials for large-scale molecular dynamics, especially on heterogeneous many-core processors. It aims to retain first-principles-like accuracy with ML-based efficiency.\"},{\"question\":\"How does TensorMD represent the interatomic physics?\",\"answer\":\"TensorMD integrates physical principles with tensor diagrams. The tensor formalism is designed to improve computational efficiency and provide flexibility for use with other scientific codes.\"},{\"question\":\"What performance results does the optimized TensorMD achieve on Sunway?\",\"answer\":\"The optimized TensorMD on the new Sunway enables simulations up to 52 billion atoms. It reports a time-to-solution of 31 ps/step/atom, setting new records for HPC+AI+MD.\"}]","TensorMD - Scalable Tensor-Diagram based Machine Learning Interatomic Potential on Heterogeneous Many-Core Processors - paper | PDF",1785732281,33,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"tensormd-scalable-tensor-diagram-based-machine-learning-interatomic-potential-on-heterogeneous-many-core-processors-paper","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/tensormd-scalable-tensor-diagram-based-machine-learning-interatomic-potential-on-heterogeneous-many-core-processors-paper/120837/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does TensorMD address in molecular dynamics simulations?","Question",{"text":75,"@type":76},"TensorMD targets efficient, high-accuracy interatomic potentials for large-scale molecular dynamics, especially on heterogeneous many-core processors. It aims to retain first-principles-like accuracy with ML-based efficiency.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does TensorMD represent the interatomic physics?",{"text":80,"@type":76},"TensorMD integrates physical principles with tensor diagrams. The tensor formalism is designed to improve computational efficiency and provide flexibility for use with other scientific codes.",{"name":82,"@type":73,"acceptedAnswer":83},"What performance results does the optimized TensorMD achieve on Sunway?",{"text":84,"@type":76},"The optimized TensorMD on the new Sunway enables simulations up to 52 billion atoms. It reports a time-to-solution of 31 ps/step/atom, setting new records for HPC+AI+MD.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]