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

arXiv: https://arxiv.org/abs/2602.01872 · HTML: https://mlsanity.com/paper/2602.01872/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 58 references that are in the map.
- First posted on arXiv 2 Feb 2026.

## Builds on (most cited first, 23 of 23)

- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Semi-Supervised Classification with Graph Convolutional Networks](https://mlsanity.com/paper/1609.02907/) (2016)
- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [Communication-Efficient Learning of Deep Networks from Decentralized Data](https://mlsanity.com/paper/1602.05629/) (2016)
- [Inductive Representation Learning on Large Graphs](https://mlsanity.com/paper/1706.02216/) (2017)
- [Trust Region Policy Optimization](https://mlsanity.com/paper/1502.05477/) (2015)
- [Open Graph Benchmark: Datasets for Machine Learning on Graphs](https://mlsanity.com/paper/2005.00687/) (2020)
- [GraphSAINT: Graph Sampling Based Inductive Learning Method](https://mlsanity.com/paper/1907.04931/) (2019)
- [Local SGD Converges Fast and Communicates Little](https://mlsanity.com/paper/1805.09767/) (2018)
- [FastGCN: Fast Learning with Graph Convolutional Networks via Importance Sampling](https://mlsanity.com/paper/1801.10247/) (2018)
- [Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks](https://mlsanity.com/paper/1905.07953/) (2019)
- [Can Decentralized Algorithms Outperform Centralized Algorithms? A Case Study for Decentralized Parallel Stochastic Gradient Descent](https://mlsanity.com/paper/1705.09056/) (2017)
- [Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks](https://mlsanity.com/paper/1909.01315/) (2019)
- [AliGraph: A Comprehensive Graph Neural Network Platform](https://mlsanity.com/paper/1902.08730/) (2019)
- [DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphs](https://mlsanity.com/paper/2010.05337/) (2020)
- [Community Interaction and Conflict on the Web](https://mlsanity.com/paper/1803.03697/) (2018)
- [Dorylus: Affordable, Scalable, and Accurate GNN Training with Distributed CPU Servers and Serverless Threads](https://mlsanity.com/paper/2105.11118/) (2021)
- [BNS-GCN: Efficient Full-Graph Training of Graph Convolutional Networks with Partition-Parallelism and Random Boundary Node Sampling](https://mlsanity.com/paper/2203.10983/) (2022)
- [IGB: Addressing The Gaps In Labeling, Features, Heterogeneity, and Size of Public Graph Datasets for Deep Learning Research](https://mlsanity.com/paper/2302.13522/) (2023)
- [Accelerating Graph Sampling for Graph Machine Learning using GPUs](https://mlsanity.com/paper/2009.06693/) (2020)
- [Knowing your FATE: Friendship, Action and Temporal Explanations for User Engagement Prediction on Social Apps](https://mlsanity.com/paper/2006.06427/) (2020)
- [MGG: Accelerating Graph Neural Networks with Fine-grained intra-kernel Communication-Computation Pipelining on Multi-GPU Platforms](https://mlsanity.com/paper/2209.06800/) (2022)
- [Legion: Automatically Pushing the Envelope of Multi-GPU System for Billion-Scale GNN Training](https://mlsanity.com/paper/2305.16588/) (2023)

## Abstract (arXiv)

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.