My specs are
Ubuntu 20.04 via WSL2
Windows 10 21H2
Nvidia latest driver, 512.95 compatible with GeForce RTX 3060 (GPU-enabled)
Cuda Driver Version/ Runtime Version 11.6/11.4
Tensorflow 2.9.1
Python 3.9.7.
I have downloaded the local Cuda installer, cudnn-local-repo-ubuntu2004-8.4.1.50_1.0-1_amd64 (1) but have not installed it yet.
My question is, should I be aiming to have Cuda v 11.2 and cuDNN 8.1 (not 8.4) according to Tested Build Configurations for GPU at
Build from source on Windows | TensorFlow?
how can I downgrade my Cuda and cuDNN if thats whats required for it to work?
Thank you v much for any advice you can give me - I am a student trying to make it work for days!