There are instructions, like:
https://linuxconfig.org/how-to-install-nvidia-cuda-toolkit-on-fedora-29-linux
(However, links in them seem old.)
NVidia's site shows essentially the same old instructions for RPM-based ("local") install as the above link.
1. The dnf can both download and install. There is no need for 'wget'. There definitely is no justification for using 'rpm'; the "install a file" part yum/dnf has been able to do for ages.
2. Without the "-y" option the dnf will not auto-install; you can see the estimated download size and can abort.
However, I'd use the RPM(network):
Code:
sudo dnf config-manager --add-repo http://developer.download.nvidia.com/compute/cuda/repos/fedora29/x86_64/cuda-fedora29.repo
sudo dnf install cuda-10-2
What version of Fedora do you have? Only 30 and 31 are currently supported. NVidia's latest dnf-repo being for 29 either means that NVidia does not care or that those packages install to 30 (or 31).
Did you try an RPM-based install or the dreaded
runfile?
I don't think that NVidia's blob differs much, no matter which system it is for.