To anyone who wants to use install cuda and cudnn for anything. Please get a free redhat developer account for free and install rhel for free for personal, non commercial or non production use or use ubuntu or centos and then install cuda and cudnn. It might save you your sanity.
Reason :-
Cuda and cudnn require gcc and g++ lower than 9. And guess what you might need to compile that compiler version from source. Have fun with the not so good docs for compiling. Seriously compiling the linux kernel on my machine seemed easier than compiling gcc.

  • 1
    Or use the Nvidia Docker image, which is literally built to handle stuff like this.

    Also gcc 9 is recent enough to have prebuilt packages for most distros?
  • 1
    @RememberMe I didnt rmember about the docker image at the time. I will use that
  • 1
    I think I will not use the docker image anyways. There is no official image for some of the stuff I use. The community is likely not updated that much and I am not gonna maintain a docker image.
Add Comment