AMR-Transformer: Enabling Efficient Long-range Interaction for Complex Neural Fluid Simulation

Zeyi Xu; Jinfan Liu; Kuangxu Chen; Ye Chen; Zhangli Hu; Bingbing Ni

doi:10.1109/cvpr52734.2025.00545

CVPR2025顶会

AMR-Transformer: Enabling Efficient Long-range Interaction for Complex Neural Fluid Simulation

Zeyi Xu, Jinfan Liu, Kuangxu Chen, Ye Chen, Zhangli Hu, Bingbing Ni

arXiv DOI 出版方

2025年份

1顶会引用

摘要

Figure 1. AMR tokenizer results in high-resolution simulation. Visualization of the AMR tokenizer applied to 1024 × 1024 shock wave and explosion simulations. The AMR tokenizer captures fine-scale structures while reducing token count. Right panels compare the regular 512 × 512 grid with the AMR partitioning, each right panel displays the total cell count and partitioning scheme (bottom-right), along with the mean squared error (MSE) relative to the 1024 × 1024 ground truth (top-left).