TorchGT: A Holistic System for Large-scale Graph Transformer Training

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Meng, Sun, Jie, Hu, Qinghao, Sun, Peng, Wang, Zeke, Wen, Yonggang, Zhang, Tianwei
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914877077979136
author Zhang, Meng
Sun, Jie
Hu, Qinghao
Sun, Peng
Wang, Zeke
Wen, Yonggang
Zhang, Tianwei
author_facet Zhang, Meng
Sun, Jie
Hu, Qinghao
Sun, Peng
Wang, Zeke
Wen, Yonggang
Zhang, Tianwei
contents Graph Transformer is a new architecture that surpasses GNNs in graph learning. While there emerge inspiring algorithm advancements, their practical adoption is still limited, particularly on real-world graphs involving up to millions of nodes. We observe existing graph transformers fail on large-scale graphs mainly due to heavy computation, limited scalability and inferior model quality. Motivated by these observations, we propose TorchGT, the first efficient, scalable, and accurate graph transformer training system. TorchGT optimizes training at different levels. At algorithm level, by harnessing the graph sparsity, TorchGT introduces a Dual-interleaved Attention which is computation-efficient and accuracy-maintained. At runtime level, TorchGT scales training across workers with a communication-light Cluster-aware Graph Parallelism. At kernel level, an Elastic Computation Reformation further optimizes the computation by reducing memory access latency in a dynamic way. Extensive experiments demonstrate that TorchGT boosts training by up to 62.7x and supports graph sequence lengths of up to 1M.
format Preprint
id arxiv_https___arxiv_org_abs_2407_14106
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle TorchGT: A Holistic System for Large-scale Graph Transformer Training
Zhang, Meng
Sun, Jie
Hu, Qinghao
Sun, Peng
Wang, Zeke
Wen, Yonggang
Zhang, Tianwei
Distributed, Parallel, and Cluster Computing
Artificial Intelligence
Machine Learning
Graph Transformer is a new architecture that surpasses GNNs in graph learning. While there emerge inspiring algorithm advancements, their practical adoption is still limited, particularly on real-world graphs involving up to millions of nodes. We observe existing graph transformers fail on large-scale graphs mainly due to heavy computation, limited scalability and inferior model quality. Motivated by these observations, we propose TorchGT, the first efficient, scalable, and accurate graph transformer training system. TorchGT optimizes training at different levels. At algorithm level, by harnessing the graph sparsity, TorchGT introduces a Dual-interleaved Attention which is computation-efficient and accuracy-maintained. At runtime level, TorchGT scales training across workers with a communication-light Cluster-aware Graph Parallelism. At kernel level, an Elastic Computation Reformation further optimizes the computation by reducing memory access latency in a dynamic way. Extensive experiments demonstrate that TorchGT boosts training by up to 62.7x and supports graph sequence lengths of up to 1M.
title TorchGT: A Holistic System for Large-scale Graph Transformer Training
topic Distributed, Parallel, and Cluster Computing
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2407.14106