GSR-GNN: Training Acceleration and Memory-Saving Framework of Deep GNNs on Circuit Graph
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911550039654400 |
|---|---|
| author | Luo, Yuebo Li, Shiyang Feng, Yifei Kancharla, Vishal Huang, Shaoyi Ding, Caiwen |
| author_facet | Luo, Yuebo Li, Shiyang Feng, Yifei Kancharla, Vishal Huang, Shaoyi Ding, Caiwen |
| contents | Graph Neural Networks (GNNs) show strong promise for circuit analysis, but scaling to modern large-scale circuit graphs is limited by GPU memory and training cost, especially for deep models. We revisit deep GNNs for circuit graphs and show that, when trainable, they significantly outperform shallow architectures, motivating an efficient, domain-specific training framework. We propose Grouped-Sparse-Reversible GNN (GSR-GNN), which enables training GNNs with up to hundreds of layers while reducing both compute and memory overhead. GSR-GNN integrates reversible residual modules with a group-wise sparse nonlinear operator that compresses node embeddings without sacrificing task-relevant information, and employs an optimized execution pipeline to eliminate fragmented activation storage and reduce data movement. On sampled circuit graphs, GSR-GNN achieves up to 87.2\% peak memory reduction and over 30$\times$ training speedup with negligible degradation in correlation-based quality metrics, making deep GNNs practical for large-scale EDA workloads. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_27156 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | GSR-GNN: Training Acceleration and Memory-Saving Framework of Deep GNNs on Circuit Graph Luo, Yuebo Li, Shiyang Feng, Yifei Kancharla, Vishal Huang, Shaoyi Ding, Caiwen Machine Learning Artificial Intelligence Graph Neural Networks (GNNs) show strong promise for circuit analysis, but scaling to modern large-scale circuit graphs is limited by GPU memory and training cost, especially for deep models. We revisit deep GNNs for circuit graphs and show that, when trainable, they significantly outperform shallow architectures, motivating an efficient, domain-specific training framework. We propose Grouped-Sparse-Reversible GNN (GSR-GNN), which enables training GNNs with up to hundreds of layers while reducing both compute and memory overhead. GSR-GNN integrates reversible residual modules with a group-wise sparse nonlinear operator that compresses node embeddings without sacrificing task-relevant information, and employs an optimized execution pipeline to eliminate fragmented activation storage and reduce data movement. On sampled circuit graphs, GSR-GNN achieves up to 87.2\% peak memory reduction and over 30$\times$ training speedup with negligible degradation in correlation-based quality metrics, making deep GNNs practical for large-scale EDA workloads. |
| title | GSR-GNN: Training Acceleration and Memory-Saving Framework of Deep GNNs on Circuit Graph |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2603.27156 |