Transolver++: An Accurate Neural Solver for PDEs on Million-Scale Geometries
Huakun Luo, Haixu Wu, Hang Zhou, Lanxiang Xing, Yichen Di, Jianmin Wang, Mingsheng Long · 2025
Read Transolver++: An Accurate Neural Solver for PDEs on Million-Scale Geometries by Huakun Luo, Haixu Wu, Hang Zhou, Lanxiang Xing, Yichen Di, Jianmin Wang, Mingsheng Long (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-02414.txt.
Description
Huakun Luo and co-authors scale a neural solver for flow rules to meshes with a million points, in a paper from 2025. Plant grade sims need that scale, yet past neural solvers topped out at tens of thousands of points. The new solver learns the state of the whole field at once. It adds two boosts for scale. One splits work leanly across chips, and one tunes locally to hold fine state marks.
One chip can now take a million points, a first for this class, and more chips stretch the load in a straight line. Scores lift by a wide edge on six test rules and by more on plant grade car and craft shapes a hundred times past past test size. The authors aim the tool at real plant design loops.
Catalog record
| CREATOR | Huakun Luo, Haixu Wu, Hang Zhou, Lanxiang Xing, Yichen Di, Jianmin Wang, Mingsheng Long |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-02414