cong hao

Speaker: Cong Hao

Date: Thursday, February 16th, 2023

Time: 3:00 - 4:00 pm

Location: HH 1010

Host: Shiyu Chang 

Title: Smart Reconfigurable Computing for GNN and Transformer

Abstract:

Reconfigurable computing, especially using FPGAs (field-programmable gate array), usually can deliver high performance and low energy for domain-specific applications. However, it is always challenging to develop good FPGA designs, given the algorithm complexity and required domain expertise. In this talk, we introduce two popular applications, transformer and graph neural network (GNN), and their smart acceleration techniques on FPGA. First, we discuss M3ViT, a multi-task learning algorithm using vision transformer and mixture-of-expert, co-designed with its FPGA optimization. Second, we introduce FlowGNN, a generic architecture for a wide range of GNN models, which can deliver real-time performance with up to 400x times faster than CPU and GPU. Finally, we briefly introduce GNNBuilder, an end-to-end automated GNN accelerator generator that digests standard PyTorch and generates its FPGA implementation.

Bio:

Dr. Cong (Callie) Hao is an assistant professor in ECE at Georgia Tech. She was a postdoctoral fellow at Georgia Tech from 2020-2021 and at UIUC from 2018-2020. She received the Ph.D. degree in Electrical Engineering from Waseda University in 2017, and the M.S. and B.S. degrees in Computer Science and Engineering from Shanghai Jiao Tong University. Her primary research interests lie in the joint area of efficient hardware design and machine learning algorithms, including software/hardware co-design for reconfigurable and high-efficiency computing and agile electronic design automation tools.