Graph Neural Network-based Tracking as a Service

TL;DR

This paper introduces a GNN-based tracking as a service platform using NVIDIA Triton to address computational and deployment challenges, demonstrating its performance on the Perlmutter supercomputer.

Contribution

It presents a novel GNN tracking service with a custom backend for Triton, enabling efficient deployment and inference in particle physics applications.

Findings

Achieved improved inference performance on supercomputing hardware.

Addressed deployment challenges by integrating GNN into a scalable service.

Demonstrated feasibility of GNN-based tracking in production environments.

Abstract

Recent studies have shown promising results for track finding in dense environments using Graph Neural Network (GNN)-based algorithms. However, GNN-based track finding is computationally slow on CPUs, necessitating the use of coprocessors to accelerate the inference time. Additionally, the large input graph size demands a large device memory for efficient computation, a requirement not met by all computing facilities used for particle physics experiments, particularly those lacking advanced GPUs. Furthermore, deploying the GNN-based track-finding algorithm in a production environment requires the installation of all dependent software packages, exclusively utilized by this algorithm. These computing challenges must be addressed for the successful implementation of GNN-based track-finding algorithm into production settings. In response, we introduce a ``GNN-based tracking as a service'' approach, incorporating a custom backend within the NVIDIA Triton inference server to facilitate GNN-based tracking. This paper presents the performance of this approach using the Perlmutter supercomputer at NERSC.