I have been having a hard time connecting my jupyter notebook to my laptops GPU. I have tried creating a new environment on Anaconda, Installing tensorflow GPU, deleted and installed a new kernel. Downloaded CUDA and CUDNN. It still shows there are no GPU available. I used an Acer Nitro 5 with GTX 1650.

CodePudding user response：

This is most likely caused due to incorrect kernel. Run this code on a Python terminal. If cuda_visible_devices gives you the result you want that mean notebook and the root Python installations are different even if they have the same version. My suggestion would be to use Anaconda to install everything including Jupyter. That has helped me solve this issue on my system.