RandLA-Net-pytorch icon indicating copy to clipboard operation
RandLA-Net-pytorch copied to clipboard

fixed bug: features and indexes on diffent devices when using cuda.

Open zergzzlun opened this issue 3 years ago • 0 comments

Hi there.

Using CUDA version of local spatial encoding in the 'gather' branch (#17) will get the following error:

RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking arugment for argument index in method wrapper_gather)

This can be fixed by moving indexes (expanded_idx) to the same device of features.

zergzzlun avatar Aug 26 '21 18:08 zergzzlun