用分块注意力让物理模拟模型在单张显卡上跑百万级点
MSPT: Efficient Large-Scale Physical Modeling via Parallelized Multi-Scale Attention
- 分层处理：局部点注意力+全局块级注意力，兼顾细节与整体
- 单卡支持百万级点，内存和计算开销大幅降低
- 适用于工业级流体、弹性等复杂物理仿真任务
工业级物理模拟的神经求解器面临关键可扩展性挑战：如何高效捕捉数百万空间单元间的精细局部交互与长程全局依赖。我们提出多尺度补丁变换器（MSPT），通过补丁内的局部点注意力与粗粒度补丁级表示的全局注意力相结合实现双尺度建模。为高效划分不规则几何区域，采用球树进行空间一致的补丁分割。该设计使MSPT可在单张GPU上扩展至百万级点。我们在标准偏微分方程基准（弹性、塑性、流体动力学、多孔介质流动）及大规模气动数据集（ShapeNet-Car、Ahmed-ML）上验证方法，达到当前最优精度，同时显著降低内存占用与计算成本。
原文摘要 · Abstract (English)
A key scalability challenge in neural solvers for industrial-scale physics simulations is efficiently capturing both fine-grained local interactions and long-range global dependencies across millions of spatial elements. We introduce the Multi-Scale Patch Transformer (MSPT), an architecture that combines local point attention within patches with global attention to coarse patch-level representations. To partition the input domain into spatially-coherent patches, we employ ball trees, which handle irregular geometries efficiently. This dual-scale design enables MSPT to scale to millions of points on a single GPU. We validate our method on standard PDE benchmarks (elasticity, plasticity, fluid dynamics, porous flow) and large-scale aerodynamic datasets (ShapeNet-Car, Ahmed-ML), achieving state-of-the-art accuracy with substantially lower memory footprint and computational cost.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。