Can this work be run under Cuda 9.0? about kpconv-pytorch HOT 2 CLOSED

I think it is a bad idea to use an older version of Pytorch, because I assume get_worker_info is not the only module that will be missing. You could have unexpected errors anywhere. If you really don't have any other solution, then you will have to modify the implementation everywhere there is a bug. For example here it's simple because the get_worker_info is just a debug function you can delete it with no problem. But I cannot ensure that everything will work for the rest of the code.

If you work on a shared server, the best solution in my opinion is to use docker containers so that you can have your own version of nvidia drivers, CUDA, and Pytorch.

Otherwise try to search online if its possible to install Pytorch with older CUDA version, but I don't think that will be easy.

Good lucj with your issue.
Hugues

from kpconv-pytorch.

Thanks, switched to Cuda 10.2 instead~

from kpconv-pytorch.