IEEE Northern Virginia Communications Society (ComSoc) and George Mason University Department of Computer Science invite you to attend the following CS Lecture Series Seminar:
Title: Deep Learning Compilers and Kernel Generation
When: April 17, 2024 from 2:00 PM to 3:00 PM
Speaker: Vinod Grover, NVIDIA
Location: Nguyen Engineering Building, Conference Room 4201
Zoom Link: (https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgmu.zoom.us%2Fj%2F99133375939%3Fpwd%3DK003K3JuYmR3L2FWYldPWWZrRUVFQT09&data=05%7C02%7Ckhassan1%40gmu.edu%7C7f98fa3c7c194c32f75f08dc58cd04dd%7C9e857255df574c47a0c00546460380cb%7C0%7C0%7C638482884783417855%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=AdtnvDXUbK6PsccONuQhjAY1O%2BY%2B0flZNjMLGaTZISs%3D&reserved=0)
Abstract:
Deep Learning Compilers have made great progress in recent years and started to be used in training and inference of models. Yet many challenges still remain. High performance model graphs are still written as kernels by experts and these are also deployed manually using custom operation hooks in frameworks. In this talk we will describe some of the challenges in compiling dynamic shapes and designing intermediate representations that enable automatic kernel generation.
Bio:
Vinod Grover is a Sr. Distinguished Engineer at NVIDIA. He has been with NVIDIA since 2007 where he led a team that developed the CUDA C++ language and compiler for programming GPUs. CUDA C++ has been a key catalyst in driving productivity, performance and innovation in many fields. Since 2017, Vinod has been focused on bringing the same productivity and performance to the problem of accelerating Deep Learning models by using important ideas from language design and compiler technology. He is currently leading a small group that is focused on problems of performance and productivity to the development of ML models. Previously Vinod held engineering, research and management positions at Sun Microsystems and Microsoft.
Vinod holds a Bachelor's degree in Physics from IIT Delhi and a Master's degree in Computer Science from Syracuse University.
Co-sponsored by: George Mason University – Department of Computer Science
Bldg: Nguyen Engineering Building, Conference Room 4201, 4400 University Dr, Fairfax, Virginia, United States, 22030, Virtual: https://events.vtools.ieee.org/m/417186