LangChain-Parsl: Connect Large Language Model Agents to High Performance Computing Resource

Abstract

Large Language Models (LLMs) can improve performance in answering questions beyond their contextual understanding by running external tools, such as a calculator for arithmetics, an online query for real-time weather, et al. For scientific applications, this enables the LLM to perform and analyze simulation runs for more accurate answers. However, the increasing scale of scientific computing requires high-performance computers (HPCs), which are managed by job schedulers. In this work, we implemented Parsl to the LangChain tool calling to bridge the gap between the LLM agent and the HPC resource. Two implementations were set up and tested on a local Nvidia GPU workstation and the Polaris/ALCF HPC system. The first setup was implemented by modifying the LangChain tool calling, which converts the LangChain tool calls to Parsl functions and queues them to the Parsl workers for parallel execution. The second approach was achieved by designing a Parsl ensemble function as an LLM tool, which performed parallel tasks. With these implementations, the LLM agent workflow was prompted to run molecular dynamics simulations, with different protein structures and simulation conditions. The results show that our Parsl implementations enable parallel execution of scientific tools that invoked by LLM agents on both local GPU workstations and HPC platforms.

Ähnliche Arbeiten