arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.01872cs.DCcs.LG
收藏

Grappa:仅梯度通信的可扩展图神经网络训练

Grappa: Gradient-Only Communication for Scalable Graph Neural Network Training

Max Planck Institute for Software Systems (MPI-SWS)(马克斯·普朗克软件系统研究所);Vienna University of Economics and Business (WU)(维也纳经济与商业大学)

Chongyang Xu, Christoph Siebenbrunner, Laurent Bindschaedler

AI总结:

Grappa通过仅梯度通信实现可扩展图神经网络训练,提升训练速度和精度,适用于大规模图数据。

摘要:

跨分区边主导了分布式GNN训练的成本:每次迭代中获取远程特征和激活会超出网络的承受能力,随着图的深入和分区数量的增长。Grappa是一种分布式GNN训练框架,强制仅进行梯度通信:在每次迭代中,分区独立训练并仅交换梯度以进行全局更新。为了弥补隔离造成的精度损失,Grappa(i)定期重新分区以暴露新的邻居,(ii)应用一种轻量级的覆盖校正梯度聚合,灵感来自重要性采样。我们提出了一个渐近无偏的梯度校正估计器,用于开发一种最小距离的批处理变体,该变体与常见的深度学习包兼容。我们还引入了一种收缩版本,以提高实际中的稳定性。在真实和合成图上的实验证明,Grappa在平均上比最先进的系统快4倍(最高13倍),在更深层模型中实现了更好的精度,并在商用硬件上实现了万亿边规模的训练。Grappa是模型无关的,支持完整图和小批量训练,并不依赖高速互连或缓存。

英文摘要:

Cross-partition edges dominate the cost of distributed GNN training: fetching remote features and activations per iteration overwhelms the network as graphs deepen and partition counts grow. Grappa is a distributed GNN training framework that enforces gradient-only communication: during each iteration, partitions train in isolation and exchange only gradients for the global update. To recover accuracy lost to isolation, Grappa (i) periodically repartitions to expose new neighborhoods and (ii) applies a lightweight coverage-corrected gradient aggregation inspired by importance sampling. We present an asymptotically unbiased estimator for gradient correction, which we use to develop a minimum-distance batch-level variant that is compatible with common deep-learning packages. We also introduce a shrinkage version that improves stability in practice. Empirical results on real and synthetic graphs show that Grappa trains GNNs 4x faster on average (up to 13x) than state-of-the-art systems, achieves better accuracy especially for deeper models, and sustains training at the trillion-edge scale on commodity hardware. Grappa is model-agnostic, supports full-graph and mini-batch training, and does not rely on high-bandwidth interconnects or caching.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑