profile_picture
Jiangfei Duan
Ph.D. student, The Chinese University of Hong Kong
dj021@ie.cuhk.edu.hk

I am a first-year Ph.D. student at MMLab, CUHK. Before Joining CUHK, I received my B.S. degree from University of Chinese Academy of Sciences. During my undergraduate period, I conducted some preliminary computer vision related research supervised by Prof. Jie Zhang and Prof. Shiguang Shan. Afterwards, I turned to research on distributed DNN training advised by Prof. Dahua Lin. My research interests lie in the area of MLSys. Currently, I am working on topics related to automatic parallel training of DNN.

Interests

  • Machine Learning System
  • Large Scale Distributed Training

Education

The Chinese University of Hong Kong
2021 - present
Ph.D. Department of Information Engineering
University of Chinese Academy of Sciences
2016 - 2020
B.S. Computer Science and Technology
Outstanding graduate, supervised by Prof. Shiguang Shan

Projects

Large Scale Distributed Training, at Sensetime Research
Optmizing large scale data parallel training performance. With sparse communication and system optimization, We train alexnet in 1 minute on a 1000 V100 cluster with parrots.
Auto-Parallel Training, at MMLab, CUHK
Exploring a general method to automatically parallel DNN training on a given cluster. Currently, I am working on an evaluation system to simualte the performance of different parallel strategy, which we believe can advance the research on optimization algorithms of parallel strategy.