First why do we need to install tensorflow with cuda does it mean we dont need to install cuda globally. I installed cuda(i.e for os) globally using nvidia documentation and tensorflow cuda using pip install tensorflow[and -cuda].

Created virtual environment using venv in vscode. Pip freeze gave me cuda 12.5

I checked version of cuda usage it is showing 12.3 whereas global cuda version is 12.5 . I am confused and is it something to worry.