Part of the TPC Seminar Series

---

Abstract:

Large language models (LLMs) have the potential to serve as an assistant to aid scientists with scientific discovery. One important area needed for this assistance is that of parallel, scientific code generation. Existing Large language models, such as GPT-x, StarCoder, and Code Llama, etc. are being used to generate parallel, scientific code with some limitations. This talk will discuss some of the limitations and present our framework, called LASSI, to address some of the limitations. LASSI, is an LLM-based Automated Self-correcting pipeline for generating parallel ScIentific codes. LASSI features a pipeline that automatically generates code, tests for compilation, and checks execution. LASSI has been extended to refactor code to generate energy-efficient parallel code for a given target platform.

Biography:

Valerie Taylor is the Director of the Mathematics and Computer Science Division and a Distinguished Fellow at Argonne National Laboratory. Her research is in high-performance computing, with a focus on performance analysis, modeling and tuning of parallel, scientific applications. Her current work is on energy efficient methods and the use of Large Language Models for parallel, scientific code generation. Prior to joining Argonne, she was the Senior Associate Dean of Academic Affairs in the College of Engineering and a Regents Professor and the Royce E. Wisenbaker Professor in the Department of Computer Science and Engineering at Texas A&M University. She is also the CEO and President of the Center for Minorities and People with Disabilities in IT (CMD-IT). Valerie Taylor is an IEEE Fellow, ACM Fellow, and AAAS Fellow.