GriNNder uses structured storage offloading with partition-wise caching and gradient regathering to train full-graph GNNs on limited memory, achieving up to 9.78x speedup over baselines.
ICLR Workshop on Representation Learning on Graphs and Manifolds (ICLRW) , year=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
GriNNder: Breaking the Memory Capacity Wall in Full-Graph GNN Training with Storage Offloading
GriNNder uses structured storage offloading with partition-wise caching and gradient regathering to train full-graph GNNs on limited memory, achieving up to 9.78x speedup over baselines.