BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20211207T055407Z
LOCATION:Online
DTSTART;TZID=America/Chicago:20211115T162500
DTEND;TZID=America/Chicago:20211115T165000
UID:submissions.supercomputing.org_SC21_sess423_ws_mlhpce104@linklings.com
SUMMARY:Production Deployment of Machine-Learned Rotorcraft Surrogate Mode
 ls on HPC
DESCRIPTION:Workshop\n\nProduction Deployment of Machine-Learned Rotorcraf
 t Surrogate Models on HPC\n\nBrewer\n\nWe explore how to optimally deploy 
 different types of machine-learned surrogate models used in rotorcraft aer
 odynamics on HPC. We first developed three different rotorcraft models at 
 three different orders of magnitude (2M, 44M, and 212M trainable parameter
 s) to use as test models. We tested three different types of inference ser
 ver deployments: (1) a Flask-based HTTP inference server, (2) TensorFlow S
 erving with gRPC protocol, and (3) RedisAI server with RESP protocol. We i
 nvestigated deployments on both DoD HPCMP's SCOUT and DoE OLCF's Summit PO
 WER9 supercomputers, demonstrated the ability to inference a million sampl
 es per second using 192 GPUs, and studied multiple scenarios on both Nvidi
 a T4 and V100 GPUs. We studied a range of concurrency levels both on the c
 lient-side and the server-side, and provide optimal configuration advice b
 ased on the type of deployment. Finally, we provide a simple Python-based 
 framework for benchmarking machine-learned surrogate models using the vari
 ous inference servers.\n\nTag: Online Only, Machine Learning and Artificia
 l Intelligence\n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR