DLHub: Simplifying publication, discovery, and use of machine learning models in science

Li, Zhuozhao; Chard, Ryan; Ward, Logan; Chard, Kyle; Skluzacek, Tyler J.; Babuji, Yadu; Woodard, Anna; Tuecke, Steven; Blaiszik, Ben; Franklin, Michael J.; Foster, Ian

doi:10.1016/j.jpdc.2020.08.006

DLHub: Simplifying publication, discovery, and use of machine learning models in science

Journal Article · 27 August 2020 · Journal of Parallel and Distributed Computing

DOI:https://doi.org/10.1016/j.jpdc.2020.08.006· OSTI ID:1837199

Li, Zhuozhao ^[1]; Chard, Ryan ^[2]; Ward, Logan ^[2]; Chard, Kyle ^[3]; Skluzacek, Tyler J. ^[2]; Babuji, Yadu ^[4]; Woodard, Anna ^[2]; Tuecke, Steven ^[4]; Blaiszik, Ben ^[4]; Franklin, Michael J. ^[2]; Foster, Ian ^[3]

Univ. of Chicago, IL (United States); Argonne National Lab. (ANL), Argonne, IL (United States)
Argonne National Lab. (ANL), Argonne, IL (United States)
Univ. of Chicago, IL (United States); Argonne National Lab. (ANL), Argonne, IL (United States); Univ. of Chicago, IL (United States). Globus
Argonne National Lab. (ANL), Argonne, IL (United States); Univ. of Chicago, IL (United States). Globus

Machine Learning (ML) has become a critical tool enabling new methods of analysis and driving deeper understanding of phenomena across scientific disciplines. There is a growing need for "learning systems" to support various phases in the ML lifecycle. While others have focused on supporting model development, training, and inference, few have focused on the unique challenges inherent in science, such as the need to publish and share models and to serve them on a range of available computing resources. In this paper, we present the Data and Learning Hub for science (DLHub), a learning system designed to support these use cases. Specifically, DLHub enables publication of models, with descriptive metadata, persistent identifiers, and flexible access control. It packages arbitrary models into portable servable containers, and enables low-latency, distributed serving of these models on heterogeneous compute resources. In this work, we show that DLHub supports low-latency model inference comparable to other model serving systems including TensorFlow Serving, SageMaker, and Clipper, and improved performance, by up to 95%, with batching and memoization enabled. We also show that DLHub can scale to concurrently serve models on 500 containers. Finally, we describe five case studies that highlight the use of DLHub for scientific applications.

View Accepted Manuscript (DOE)

Research Organization:: Argonne National Laboratory (ANL), Argonne, IL (United States)

Sponsoring Organization:: Defense Advanced Research Projects Agency (DARPA); National Science Foundation (NSF); USDOE; USDOE Laboratory Directed Research and Development (LDRD) Program; USDOE Office of Fossil Energy (FE)

Grant/Contract Number:: AC02-06CH11357

OSTI ID:: 1837199

Alternate ID(s):: OSTI ID: 1811041

Journal Information:: Journal of Parallel and Distributed Computing, Journal Name: Journal of Parallel and Distributed Computing Vol. 147; ISSN 0743-7315

Publisher:: ElsevierCopyright Statement

Country of Publication:: United States

Language:: English

References (21)

The Materials Data Facility: Data Services to Advance Materials Science Research Blaiszik, B.; Chard, K.; Pruyne, J. JOM, Vol. 68, Issue 8 https://doi.org/10.1007/s11837-016-2001-3	journal	July 2016
OCPMDM: Online computation platform for materials data mining Zhang, Qing; Chang, Dongping; Zhai, Xiuyun Chemometrics and Intelligent Laboratory Systems, Vol. 177 https://doi.org/10.1016/j.chemolab.2018.04.004	journal	June 2018
Python Materials Genomics (pymatgen): A robust, open-source python library for materials analysis Ong, Shyue Ping; Richards, William Davidson; Jain, Anubhav Computational Materials Science, Vol. 68 https://doi.org/10.1016/j.commatsci.2012.10.028	journal	February 2013
AFLOW-ML: A RESTful API for machine-learning predictions of materials properties Gossett, Eric; Toher, Cormac; Oses, Corey Computational Materials Science, Vol. 152 https://doi.org/10.1016/j.commatsci.2018.03.075	journal	September 2018
Matminer: An open source toolkit for materials data mining Ward, Logan; Dunn, Alexander; Faghaninia, Alireza Computational Materials Science, Vol. 152 https://doi.org/10.1016/j.commatsci.2018.05.018	journal	September 2018
Machine learning applications in cancer prognosis and prediction Kourou, Konstantina; Exarchos, Themis P.; Exarchos, Konstantinos P. Computational and Structural Biotechnology Journal, Vol. 13 https://doi.org/10.1016/j.csbj.2014.11.005	journal	January 2015
An online tool for predicting fatigue strength of steel alloys based on ensemble data mining Agrawal, Ankit; Choudhary, Alok International Journal of Fatigue, Vol. 113 https://doi.org/10.1016/j.ijfatigue.2018.04.017	journal	August 2018
“Memo” Functions and Machine Learning Michie, Donald Nature, Vol. 218, Issue 5136 https://doi.org/10.1038/218019a0	journal	April 1968
1,500 scientists lift the lid on reproducibility Baker, Monya Nature, Vol. 533, Issue 7604 https://doi.org/10.1038/533452a	journal	May 2016
Deep learning improves prediction of CRISPR–Cpf1 guide RNA activity Kim, Hui Kwon; Min, Seonwoo; Song, Myungjae Nature Biotechnology, Vol. 36, Issue 3 https://doi.org/10.1038/nbt.4061	journal	January 2018
The Open Quantum Materials Database (OQMD): assessing the accuracy of DFT formation energies Kirklin, Scott; Saal, James E.; Meredig, Bryce npj Computational Materials, Vol. 1, Issue 1 https://doi.org/10.1038/npjcompumats.2015.10	journal	December 2015
A general-purpose machine learning framework for predicting properties of inorganic materials Ward, Logan; Agrawal, Ankit; Choudhary, Alok npj Computational Materials, Vol. 2, Issue 1 https://doi.org/10.1038/npjcompumats.2016.28	journal	August 2016
Commentary: The Materials Project: A materials genome approach to accelerating materials innovation Jain, Anubhav; Ong, Shyue Ping; Hautier, Geoffroy APL Materials, Vol. 1, Issue 1 https://doi.org/10.1063/1.4812323	journal	July 2013
Adaptive enhanced sampling by force-biasing using neural networks Guo, Ashley Z.; Sevgen, Emre; Sidky, Hythem The Journal of Chemical Physics, Vol. 148, Issue 13 https://doi.org/10.1063/1.5020733	journal	April 2018
Deep learning to represent subgrid processes in climate models Rasp, Stephan; Pritchard, Michael S.; Gentine, Pierre Proceedings of the National Academy of Sciences, Vol. 115, Issue 39 https://doi.org/10.1073/pnas.1810286115	journal	September 2018
XSEDE: Accelerating Scientific Discovery Towns, John; Cockerill, Timothy; Dahan, Maytal Computing in Science & Engineering, Vol. 16, Issue 5 https://doi.org/10.1109/MCSE.2014.80	journal	September 2014
Shining Light into Black Boxes Morin, A.; Urban, J.; Adams, P. D. Science, Vol. 336, Issue 6078 https://doi.org/10.1126/science.1218263	journal	April 2012
Enhancing reproducibility for computational methods Stodden, V.; McNutt, M.; Bailey, D. H. Science, Vol. 354, Issue 6317 https://doi.org/10.1126/science.aah6168	journal	December 2016
Reproducible big data science: A case study in continuous FAIRness Madduri, Ravi; Chard, Kyle; D’Arcy, Mike PLOS ONE, Vol. 14, Issue 4 https://doi.org/10.1371/journal.pone.0213013	journal	April 2019
A data ecosystem to support machine learning in materials science Blaiszik, Ben; Ward, Logan; Schwarting, Marcus MRS Communications, Vol. 9, Issue 4 https://doi.org/10.1557/mrc.2019.118	journal	October 2019
Applying Artificial Intelligence to Address the Knowledge Gaps in Cancer Care Simon, George; DiNardo, Courtney D.; Takahashi, Koichi The Oncologist, Vol. 24, Issue 6 https://doi.org/10.1634/theoncologist.2018-0257	journal	November 2018

Similar Records

A data ecosystem to support machine learning in materials science

Journal Article · 2019 · MRS Communications · OSTI ID:1607645

ScienceSearch: Enabling Search through Automatic Metadata Generation

Conference · 2018 · OSTI ID:1602828

Related Subjects

97 MATHEMATICS AND COMPUTING
DLHub
Learning systems
Machine learning
Model serving

DLHub: Simplifying publication, discovery, and use of machine learning models in science

Citation Formats

References (21)

Similar Records

Related Subjects