Installation fails on Cuda12
Closed this issue · 3 comments
Hi,
I was wondering if there is a new wheel for cuda12 installation? I tried this on a debian 12 system with cuda12, however the setup fails to build the wheel. I also tried with cuda11.3, where it fails because debian12 has a g++ version 12 and it requires a g++ version <=10.
Thanks
Unfortunately, the strict CUDA version requirement is a result of the OpenFold dependency. There may be a way to work around it with a modified OpenFold fork, but at the moment I would suggest trying to install a compatible g++ with cuda11.3.
In theory, you should also be able to install a g++ version compatible with your cuda inside the conda environment. Although we also had cuda11.6 installed, and when I encountered the same problem (g++ version mismatch), I just did the installation with cuda11.6 and it was done without a problem
That is what I ended up doing, installing a g++10 on my debian12 setup within a venv and LMOD setup. For now please feel free to close the issue. Thank you for the help