Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Unifying Framework to Enable Artificial Intelligence in High Performance Computing Workflows

Jens Domke, Mohamed Wahib, Anshu Dubey, Tal Ben-Nun, Erik W. Draeger

TL;DR

The paper addresses the fragmentation of HPC software ecosystems and the lack of interoperable HPC-AI tooling necessary to exploit future hardware. It proposes a living, phased framework design with features such as data transformers, locality expressiveness, and semantics-constrained language subsets to enable plug-and-play integration of HPC and AI/ML components. An implementation plan combines pilot studies, incremental refactoring of existing libraries, and co-design with AI-assisted tooling to accelerate development and evaluation. The authors argue that a single scalable HPC-AI framework will accelerate scientific discovery by reducing rework across hardware generations and enabling efficient, sustained optimization of complex workflows.

Abstract

Current trends point to a future where large-scale scientific applications are tightly-coupled HPC/AI hybrids. Hence, we urgently need to invest in creating a seamless, scalable framework where HPC and AI/ML can efficiently work together and adapt to novel hardware and vendor libraries without starting from scratch every few years. The current ecosystem and sparsely-connected community are not sufficient to tackle these challenges, and we require a breakthrough catalyst for science similar to what PyTorch enabled for AI.

A Unifying Framework to Enable Artificial Intelligence in High Performance Computing Workflows

TL;DR

The paper addresses the fragmentation of HPC software ecosystems and the lack of interoperable HPC-AI tooling necessary to exploit future hardware. It proposes a living, phased framework design with features such as data transformers, locality expressiveness, and semantics-constrained language subsets to enable plug-and-play integration of HPC and AI/ML components. An implementation plan combines pilot studies, incremental refactoring of existing libraries, and co-design with AI-assisted tooling to accelerate development and evaluation. The authors argue that a single scalable HPC-AI framework will accelerate scientific discovery by reducing rework across hardware generations and enabling efficient, sustained optimization of complex workflows.

Abstract

Current trends point to a future where large-scale scientific applications are tightly-coupled HPC/AI hybrids. Hence, we urgently need to invest in creating a seamless, scalable framework where HPC and AI/ML can efficiently work together and adapt to novel hardware and vendor libraries without starting from scratch every few years. The current ecosystem and sparsely-connected community are not sufficient to tackle these challenges, and we require a breakthrough catalyst for science similar to what PyTorch enabled for AI.
Paper Structure (7 sections, 3 figures)

This paper contains 7 sections, 3 figures.

Table of Contents

  1. LATEST IN SCIENTIFIC PROGRAMMING
  2. OBJECTIVES AND FEATURES FOR THE FRAMEWORK
  3. Desired Features:
  4. DEVELOPMENT APPROACH
  5. IMPLEMENTATION PROPOSAL
  6. CONCLUDING REMARKS
  7. ACKNOWLEDGMENTS

Figures (3)

  • Figure 1: The design space that a multiphysics HPC framework must take into account already in the absence of AI/ML integration or offloading.
  • Figure 2: Various ways in which we anticipate use of AI/ML in High-Performance Computing software.
  • Figure 3: Workflow of the proposed HPC framework. Phases enable gradual application integration --- the more information provided, the more capabilities can be unlocked to reduce the overhead of utilizing modern hardware.