torch-points-kernels
torch-points-kernels copied to clipboard
Pytorch kernels for spatial operations on point clouds
At the moment we have two variables to control the maximun number of threads started on the GPU. There should be only one.
ImportError: /opt/conda/lib/python3.8/site-packages/torch_points_kernels-0.6.10-py3.8-linux-x86_64.egg/torch_points_kernels/points_cpu.so: undefined symbol: _ZN3c106detail19maybe_wrap_dim_slowEllb
import torch_points_kernels.points_cuda as tpcuda ImportError: DLL load failed while importing points_cuda