一句话：提出 PULSE，以统一最小单元原则消除时序图网络训练中的计算、存储与数据加载冗余，训练吞吐最高提升 6.67 倍。
时序图网络（TGN）训练在计算、存储与数据加载中存在普遍冗余，造成效率下降、内存压力增大以及 CPU-GPU 数据搬运过多。为此，作者提出端到端训练框架 PULSE，以统一最小单元原则系统消除冗余，并设计三个协同单元：最小输入单元 MIU 做组件级去重和算子级计算重构；最小存储单元 MSU 进行依赖引导的消息重构，仅保留不可复现条目并支持按需恢复；最小复用单元 MRU 管理 GPU 内存，结合 BlockPool 缓冲分配器与二分时序复用策略，缓解碎片并利用批次间局部性。在代表性基准上的实验表明，PULSE 相比当前最优基线将训练吞吐最高提升 6.67 倍。
Temporal Graph Network (TGN) is increasingly adopted to model evolving relationships in dynamic graphs. However, the training pipeline is plagued by pervasive redundancy in computation, storage, and data loading. These redundancies harm computational efficiency, exacerbate memory pressure, and induce excessive CPU-GPU data transfers. We present PULSE, an end-to-end TGN training framework that systematically eliminates redundancies guided by a unified minimal-unit principle. To realize such principle, PULSE defines three synergetic units: 1) the Minimal Input Unit (MIU) for component-wise deduplication and operator-level reconstruction of redundant computations, 2) the Minimal Storage Unit (MSU) for dependency-guided message reconstruction, only preserving irreproducible entries while enabling on-demand recovery of others, and 3) the Minimal Reuse Unit (MRU) for GPU memory management, combining a BlockPool-based buffer allocator with a bipartite temporal reuse strategy to mitigate fragmentation and exploit inter-batch locality. Experimental results on representative benchmarks demonstrate that PULSE improves training throughput by up to 6.67× over the state-of-the-art baselines.
@inproceedings{Wang0MYW0S0L0026,
title = {Efficient Temporal Graph Network Training via Unified Redundancy Elimination},
author = {Yiqing Wang and Hailong Yang and Kejie Ma and Enze Yu and Pengbo Wang and Xin You and Qingxiao Sun and Chenhao Xie and Zhongzhi Luan and Yi Liu and Depei Qian},
booktitle = {ASPLOS},
year = {2026},
doi = {10.1145/3779212.3790157}
}