Python, CUDA, Driver Installation - Using Pip and Conda

I have a laptop with an Nvidia graphics chip, which is running Fedora 39, but i am using the Intel GPU (Mesa Intel® HD Graphics 5500). I was attempting to use Python, Numba, with CUDA and the Nvidia chip, and have installed via Pip and Conda the relevant libraries and toolkits.

Conda : cudatoolkit
Pip : cuda-python, nvidia-cuda-nvrtc-cu12

When using Spyder i receive the message :

CUDA driver library cannot be found. If you are sure that a CUDA driver is installed, try setting environment variable NUMBA_CUDA_DRIVER with the file path of the CUDA driver shared library.

Questions i have are :

  1. Do i need to install the Nvidia driver in the Fedora 39 OS as the OS graphics driver and will therefore have to use the laptop Nvidia IC ?
  2. Is it possible to use the Nvidia IC for CUDA, not install the Nvidia driver as part of the OS and continue to use the Intel graphics GPU ?
  3. For a desktop PC, can you use multiple graphics cards, with one providing your graphics output and the other for CUDA ?

For the laptop, being able to use the Nvidia chip for CUDA and the Intel GPU for graphics output is my preferred option.

What i am unsure about is if the Nvidia chip is used for graphics output, will CUDA be restricted in terms of the use of the chip, or is it possible to disrupt the graphics output etc ?

Any guidance gratefully received.