Loading mesh with io.obj.import_mesh returns torch tensor with device set to 'cpu' by default

[karthikey97]

So I tried loading a .obj mesh file using kaolin.io.obj.import_mesh function. By default the tensors are created in the cpu. I cannot change the device to gpu in place because the function returns a namedtuple which is immutable. I went through the obj loading code and there does not seem to be any flexibility to specify which device the tensors get loaded to. Am I doing something wrong with my setup that tensors are not created in gpu by default? I am working off 0.13 installed using pip.

[shumash]

Please try with the latest version. The obj importer now returns a SurfaceMesh object, and you can call cuda() on it.

See tutorial docs here