167-PFlops Deep Learning for Electron Microscopy: From Learning Physics to Atomic Manipulation

Patton, Robert M.; Johnston, J. Travis; Young, Steven R.; Schuman, Catherine D.; March, Don D.; Potok, Thomas E.; Rose, Derek C.; Lim, Seung-Hwan; Karnowski, Thomas P.; Ziatdinov, Maxim A.; Kalinin, Sergei V.

doi:10.1109/SC.2018.00053

Title: 167-PFlops Deep Learning for Electron Microscopy: From Learning Physics to Atomic Manipulation

Abstract

An artificial intelligence system called MENNDL, which used 25,200 Nvidia Volta GPUs on Oak Ridge National Laboratory’s Summit machine, automatically designed an optimal deep learning network in order to extract structural information from raw atomic-resolution microscopy data. In a few hours, MENNDL creates and evaluates millions of networks using a scalable, parallel, asynchronous genetic algorithm augmented with a support vector machine to automatically find a superior deep learning network topology and hyper-parameter set than a human expert can find in months. For the application of electron microscopy, the system furthers the goal of improving our understanding of the electron-beam-matter interactions and real-time image-based feedback, which enables a huge step beyond human capacity toward nanofabricating materials automatically. Finally, MENNDL has been scaled to the 4,200 available nodes of Summit achieving a measured 152.5 PFlops, with an estimated sustained performance of 167 PFlops when the entire machine is available.

Authors:

Patton, Robert M. ^[1]; Johnston, J. Travis ^[1]; Young, Steven R. ^[1]; Schuman, Catherine D. ^[1]; March, Don D. ^[1]; Potok, Thomas E. ^[1]; Rose, Derek C. ^[1]; Lim, Seung-Hwan ^[1]; Karnowski, Thomas P. ^[1]; Ziatdinov, Maxim A. ^[1]; Kalinin, Sergei V. ^[1]

Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)

Publication Date:: Fri Nov 16 00:00:00 EST 2018

Research Org.:: Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)

Sponsoring Org.:: USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR)

OSTI Identifier:: 1485808

Grant/Contract Number:: AC05-00OR22725

Resource Type:: Accepted Manuscript

Journal Name:: International Conference for High Performance Computing, Networking, Storage and Analysis

Additional Journal Information:: Conference: SC’18: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, Dallas, TX (2018), Dallas, TX (United States), 11-16 Nov. 2018

Publisher:: ACM

Country of Publication:: United States

Language:: English

Subject:: 97 MATHEMATICS AND COMPUTING

Citation Formats


                    Patton, Robert M., Johnston, J. Travis, Young, Steven R., Schuman, Catherine D., March, Don D., Potok, Thomas E., Rose, Derek C., Lim, Seung-Hwan, Karnowski, Thomas P., Ziatdinov, Maxim A., and Kalinin, Sergei V. 167-PFlops Deep Learning for Electron Microscopy: From Learning Physics to Atomic Manipulation.  United States: N. p., 2018. 
Web.  doi:10.1109/SC.2018.00053.

Copy to clipboard


                    Patton, Robert M., Johnston, J. Travis, Young, Steven R., Schuman, Catherine D., March, Don D., Potok, Thomas E., Rose, Derek C., Lim, Seung-Hwan, Karnowski, Thomas P., Ziatdinov, Maxim A., & Kalinin, Sergei V. 167-PFlops Deep Learning for Electron Microscopy: From Learning Physics to Atomic Manipulation.  United States.  https://doi.org/10.1109/SC.2018.00053

Copy to clipboard


                    Patton, Robert M., Johnston, J. Travis, Young, Steven R., Schuman, Catherine D., March, Don D., Potok, Thomas E., Rose, Derek C., Lim, Seung-Hwan, Karnowski, Thomas P., Ziatdinov, Maxim A., and Kalinin, Sergei V. Fri .  
"167-PFlops Deep Learning for Electron Microscopy: From Learning Physics to Atomic Manipulation".  United States.  https://doi.org/10.1109/SC.2018.00053.  https://www.osti.gov/servlets/purl/1485808.

Copy to clipboard


                    
@article{osti_1485808,

  title        = {167-PFlops Deep Learning for Electron Microscopy: From Learning Physics to Atomic Manipulation},

  author       = {Patton, Robert M. and Johnston, J. Travis and Young, Steven R. and Schuman, Catherine D. and March, Don D. and Potok, Thomas E. and Rose, Derek C. and Lim, Seung-Hwan and Karnowski, Thomas P. and Ziatdinov, Maxim A. and Kalinin, Sergei V.},

  abstractNote = {An artificial intelligence system called MENNDL, which used 25,200 Nvidia Volta GPUs on Oak Ridge National Laboratory’s Summit machine, automatically designed an optimal deep learning network in order to extract structural information from raw atomic-resolution microscopy data. In a few hours, MENNDL creates and evaluates millions of networks using a scalable, parallel, asynchronous genetic algorithm augmented with a support vector machine to automatically find a superior deep learning network topology and hyper-parameter set than a human expert can find in months. For the application of electron microscopy, the system furthers the goal of improving our understanding of the electron-beam-matter interactions and real-time image-based feedback, which enables a huge step beyond human capacity toward nanofabricating materials automatically. Finally, MENNDL has been scaled to the 4,200 available nodes of Summit achieving a measured 152.5 PFlops, with an estimated sustained performance of 167 PFlops when the entire machine is available.},

  doi          = {10.1109/SC.2018.00053},

  journal      = {International Conference for High Performance Computing, Networking, Storage and Analysis},

  number       = ,

  volume       = ,

  place        = {United States},

  year         = {Fri Nov 16 00:00:00 EST 2018},

  month        = {Fri Nov 16 00:00:00 EST 2018}

}

Copy to clipboard

Journal Article:

Free Publicly Available Full Text

Accepted Manuscript (DOE)

Publisher's Version of Record

https://doi.org/10.1109/SC.2018.00053

Other availability

Search WorldCat to find libraries that may hold this journal

Save / Share:

Export Metadata

Save to My Library

Works referencing / citing this record:

Deep materials informatics: Applications of deep learning in materials science
journal, June 2019

Agrawal, Ankit; Choudhary, Alok
MRS Communications, Vol. 9, Issue 3
DOI: 10.1557/mrc.2019.73

Similar Records in DOE PAGES and OSTI.GOV collections:

167-PFlops deep learning for electron microscopy: from learning physics to atomic manipulation

Conference Patton, Robert ; Johnston, Travis ; Young, Steven ; ...

An artificial intelligence system called MENNDL, which used 25,200 NVIDIA Volta GPUs on Oak Ridge National Laboratory's Summit machine, automatically designed an optimal deep learning network in order to extract structural information from raw atomic-resolution microscopy data. In a few hours, MENNDL creates and evaluates millions of networks using a scalable, parallel, asynchronous genetic algorithm augmented with a support vector machine to automatically find a superior deep learning network topology and hyper-parameter set than a human expert can find in months. For the application of electron microscopy, the system furthers the goal of improving our understanding of the electron-beam-matter interactionsmore »« less
https://doi.org/10.1109/SC.2018.00053

Full Text Available
Fine-Grained Exploitation of Mixed Precision for Faster CNN Training

Conference Johnston, Travis ; Young, Steven ; Schuman, Catherine ; ...

As deep convolutional neural networks (CNNs) have become increasingly popular and successful at an ever-widening number of machine learning tasks specialized hardware has become increasingly available for training and deploying them. NVIDIA's recent Volta architecture includes tensor cores which perform a fused operation reduced and mixed precision (16-bit multiply, 32-bit accumulate). Recent research indicates that, typically, very little is lost (in terms of training accuracy) when half precision is used in place of single precision, and performance gains can be made by doing arithmetic in reduced precision. In this work we demonstrate that making layer-by-layer choices as to the arithmetic/datamore »« less
https://doi.org/10.1109/MLHPC49564.2019.00007

Full Text Available
Fine-Grained Exploitation of Mixed Precision for Faster CNN Training

Conference Johnston, Travis ; Young, Steven ; Schuman, Catherine ; ...

As deep convolutional neural networks (CNNs) have become increasingly popular and successful at an ever-widening number of machine learning tasks specialized hardware has become increasingly available for training and deploying them. NVIDIA's recent Volta architecture includes tensor cores which perform a fused operation reduced and mixed precision (16-bit multiply, 32-bit accumulate). Recent research indicates that, typically, very little is lost (in terms of training accuracy) when half precision is used in place of single precision, and performance gains can be made by doing arithmetic in reduced precision. In this work we demonstrate that making layer-by-layer choices as to the arithmetic/datamore »« less
https://doi.org/10.1109/MLHPC49564.2019.00007

Full Text Available
Performance Analysis of PIConGPU: Particle-in-Cell on GPUs using NVIDIA’s NSight Systems and NSight Compute

Technical Report Leinhauser, Matthew ; Young, Jeffrey ; Bastrakov, Sergei ; ...

PIConGPU, Particle In Cell on GPUs, is an open source simulations framework for plasma and laser-plasma physics used to develop advanced particle accelerators for radiation therapy of cancer, high energy physics and photon science. While PIConGPU has been optimized for at least 5 years to run well on NVIDIA GPU-based clusters, there has been limited exploration by the development team of potential scalability bottlenecks using recently updated and new tools including NVIDIA’s NVProf tool and the brand-new NVIDIA NSight Suite (Systems and Compute) tools. PIConGPU is a highly optimized application that runs production jobs at scale on a system Oakmore »« less
https://doi.org/10.2172/1761619

Full Text Available
Scaling Deep Learning on GPU and Knights Landing clusters

Journal Article You, Yang ; Buluc, Aydin ; Demmel, James - International Conference for High Performance Computing, Networking, Storage and Analysis (Online)

The speed of deep neural networks training has become a big bottleneck of deep learning research and development. For example, training GoogleNet by ImageNet dataset on one Nvidia K20 GPU needs 21 days. To speed up the training process, the current deep learning systems heavily rely on the hardware accelerators. However, these accelerators have limited on-chip memory compared with CPUs. To handle large datasets, they need to fetch data from either CPU memory or remote processors. We use both self-hosted Intel Knights Landing (KNL) clusters and multi-GPU clusters as our target platforms. From an algorithm aspect, current distributed machine learningmore »« less
Cited by 39
https://doi.org/10.1145/3126908.3126912

Full Text Available

Similar Records

Title: 167-PFlops Deep Learning for Electron Microscopy: From Learning Physics to Atomic Manipulation

Abstract

Citation Formats

Deep materials informatics: Applications of deep learning in materials science journal, June 2019

Deep materials informatics: Applications of deep learning in materials science
journal, June 2019