search

facebookresearch / pytorch3d

PyTorch3D is FAIR's library of reusable components for deep learning with 3D data
https://pytorch3d.org/
Other
8.7k stars 1.3k forks source link

Why the knn_points only support float32? #1700

Closed poodarchu closed 9 months ago

poodarchu commented 9 months ago

❓ Questions on how to use PyTorch3D

. when calculating the derivatives https://github.com/facebookresearch/pytorch3d/blob/main/pytorch3d/csrc/knn/knn_cpu.cpp#L118 why need to convert data as float? Can this be done in double directly?
bottler commented 9 months ago

PyTorch3D generally uses float32 for locations. You could compile a version of that function taking doubles and using them directly if you want.