Part of the TPC Seminar Series
Speaker: Koichi Yamada, Sr. Principal Engineer in the Data Center and AI Group (DCAI) at Intel
Date: Wednesday, August 7, 2024
Time: 11:00 A.M. to 12:15 P.M. (Central U.S. Time)
Location: Virtual
Abstract:
The Aurora Supercomputer is an exascale supercomputer designed by Intel and HPE for Argonne National Laboratory. Aurora will enable breakthrough science, innovation, and discovery to address some of the world’s biggest challenges. Intel team has been collaborating with Argonne Nation Laboratory and HPE for creating a state-of-the-art generative AI model for science by leveraging the full potential of the Aurora supercomputer. This talk will present the optimized AI software stack and various distributed training strategy solutions Intel team is providing for Aurora, that can enable efficient training of various LLM model sizes from billions to trillion parameters.
Biography:
Koichi Yamada is a Sr. Principal Engineer in the Data Center and AI Group (DCAI) at Intel where he leads the development of AI software systems and optimization methods. In his technical career at Intel, he has worked on and led numerous technology developments including computer architecture, operating system design, binary translation system and computer security. Koichi holds more than 40 U.S. patents.