RedHat, CUDA 12.4 - torch GPU not avaiable

Torch version 2.4.0+cu124 is installted,
The nvidia driver 550.90.07 is also installed.
nvcc (release 12.4, V12.4.131) is installed.

But torch.cuda.is_available() returns False.

Does someone has an idea to fix this issue.
Why does torch not work on Red Hat Enterprise Linux 8.10 (Ootpa), Linux 4.18.0-553.8.1.el8_10.x86_64 with CUDA?

You’ve enabled MIG so make sure the right MIG setup is used and the desired slice is passed to CUDA_VISIBLE_DEVICES.