Description: In this seminar, we discuss the paper that demonstrates the first climate-scale, numerical ocean simulations improved through distributed, online inference of Deep Neural Networks (DNN) using SmartSim.

SmartSim is a library dedicated to enabling online analysis and Machine Learning (ML) for traditional HPC simulations. We detail the SmartSim architecture, benchmarks on heterogenous HPC systems, and a few use cases. This research demonstrates the capability of SmartSim by using it to run a 12-member ensemble of global-scale, high-resolution ocean simulations, each spanning 19 compute nodes, all communicating with the same ML architecture at each simulation timestep. In total, 970 billion inferences are collectively served by running the ensemble for a total of 120 simulated years. Finally, we show our solution is stable over the full duration of the model integrations, and that the inclusion of machine learning has minimal impact on the simulation runtimes. We conclude with a discussion of recent research into optimization of inference server and simulation deployment.

Presenter(s): Sam Partee, HPE/Cray
Location: Webcast
Date & Time: February 22, 2022, 2:00p - 3:30p ET

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 22 February 2022.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.