perf(dpa1): optimize graph CUDA inference and deployment - #5758

Merged

OutisLi merged 13 commits into

deepmodeling:masterfrom

OutisLi:pr/dpa1

Jul 15, 2026

Commits on Jul 11, 2026

perf(pt/dpa1): optimize graph CUDA inference and deployment
OutisLi
committed
bcf3c54
perf(pt/dpa1): add compact canonical graph deployment
OutisLi
committed
a89ac04
perf(dpa1): specialize compact descriptor mask handling
OutisLi
committed
3f13664

Commits on Jul 12, 2026

Commits on Jul 13, 2026

refactor(graph): align dual-CSR field ordering
OutisLi
committed
2540154
fix(lmp): expose device graph inference through C API
OutisLi
committed
45bf817
fix(cuda): restore full architecture set on CUDA 13
OutisLi
committed
e5d0dbe
Merge upstream/master into pr/dpa1
OutisLi
committed
c2538f0
Merge remote-tracking branch 'upstream/master' into pr/dpa1
OutisLi
committed
e96287c

Commits on Jul 14, 2026

Merge branch 'master' into pr/dpa1
OutisLi
authored
2d34f9e