TensorFlow GPU Nightmare!

[muhendis80]

Hi, I can't get TensorFlow to use the GPU. My computer has an RTX 4070 graphics card. I've tried installing every CUDA and cuDNN version, but TensorFlow still can't detect them and keeps using the CPU. On the other hand, when I use the same code and run PyTorch with the command pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124, PyTorch immediately accesses the GPU without any issues. Please provide a similarly simple solution for TensorFlow, as I can't figure this out.

[ALOK158]

I am taking into account that you have a latest version of all mentioned above

Well, first you should verify whether your GPU is available or not via the given code
print("Is GPU available:", tf.config.list_physical_devices('GPU'))

If it fails, then you should ensure that path environment variable includes CuSA and CUDNN

[AbstractEyes]

I had this same exact problem. I'm just not getting anything through tensorflow, but the gpu is most definitely available through the environment.
I've been here for hours trying to debug it. I've run through the ringer. PATH set, reinstalled drivers, correct versions of cuda, WSL2, none of it worked.