Open MatthCorn opened 2 weeks ago
I was not able to compile pointops and chamfer distance with torch 2.0. But there is a pure PyTorch implementation of pointops functions. Actually, I only used functions to index knn points and fps. This may be an solution if you are not able to use torch with a lower version.
Hello !
I was trying to compile pointops and i got an error about my cuda version :
RuntimeError: The detected CUDA version (12.0) mismatches the version that was used to compile PyTorch (11.7). Please make sure to use the same CUDA versions.
i am using as you see cuda 11.7 with pytorch 2.0.1 and i have seen some of you succeding compilation with cuda 11.6.
Has anyone encounter this issue and can help ?
Thank you, Matthieu