Giters

CVMI-Lab / PAConv

(CVPR 2021) PAConv: Position Adaptive Convolution with Dynamic Kernel Assembling on Point Clouds

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Cuda Problem

LeslauD opened this issue · comments

commented

My program gets stuck when executing the following line:

gpu = load('gpconv_cuda', [
os.path.join(gpu_path, 'operator.cpp'),
os.path.join(gpu_path, 'assign_score_withk_gpu.cu'),
os.path.join(gpu_path, 'assign_score_withk_halfkernel_gpu.cu'),
], build_directory=gpu_path, verbose=False)

commented

I once had the same problem. I set --load('gpconv_cuda', ... verbose=False) --verbose to True . And then run load function will get a path . CD the {path}/gpconv_cuda. If you see the lock file, delete it and you can run it.
It should be that the last compilation was not completed but the lock file was not deleted, so it has been deadlocked.