BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160548Z
LOCATION:Track 3
DTSTART;TZID=America/New_York:20201112T122500
DTEND;TZID=America/New_York:20201112T125500
UID:submissions.supercomputing.org_SC20_sess209_ws_mlhpce104@linklings.com
SUMMARY:Accelerating GPU-Based Machine Learning in Python Using MPI Librar
 y: A Case Study with MVAPICH2-GDR
DESCRIPTION:Workshop\n\nAccelerating GPU-Based Machine Learning in Python 
 Using MPI Library: A Case Study with MVAPICH2-GDR\n\nGhazimirsaeed, Anthon
 y, Shafi, Subramoni, Panda\n\nThe growth of big data applications during t
 he last decade has led to a surge in the deployment and popularity of mach
 ine learning (ML) libraries. On the other hand, the high performance offer
 ed by GPUs makes them well suited for ML problems. To take advantage of GP
 U performance for ML, NVIDIA has recently developed the cuML library. cuML
  is the GPU counterpart of Scikit-learn, and provides similar Pythonic int
 erfaces to Scikit-learn while hiding the complexities of writing GPU compu
 te kernels directly using CUDA. To support execution of ML workloads on Mu
 lti-Node Multi-GPU (MNMG) systems, the cuML library exploits the NVIDIA Co
 llective Communications Library (NCCL) as a backend for collective communi
 cations between processes. On the other hand, MPI is a de facto standard f
 or communication in HPC systems. Among various MPI libraries, MVAPICH2-GDR
  is the pioneer in optimizing GPU communication. \n\nThis paper explores v
 arious aspects and challenges of providing MPI-based communication support
  for GPU-accelerated cuML applications. More specifically, it proposes a P
 ython API to take advantage of MPI-based communications for cuML applicati
 ons. It also gives an in-depth analysis, characterization, and benchmarkin
 g of the cuML algorithms such as K-Means, Nearest Neighbors, Random Forest
 , and tSVD. Moreover, it provides a comprehensive performance evaluation a
 nd profiling study for MPI-based versus NCCL-based communication for these
  algorithms. The evaluation results show that the proposed MPI-based commu
 nication approach achieves up to 1.6x, 1.25x, 1.25x, and 1.36x speedup for
  K-Means, Nearest Neighbors, Linear Regression, and tSVD, respectively on 
 up to 32 GPUs.\n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR