I have a computer with a decent GPU, and I tried to get Tensorflow GPU working. I will start by saying that I am a long time Arch Linux user, and I am very new to Ubuntu, but I wanted to give it a shot in this instance.
I created a virtualenv where I installed tensorflow-gpu and all the required stuff. My code worked, but it was not using the GPU, so after checking stuff for a while, I thought the problem might be with CUDA. Based on Tensorflow documentation, Tensorflow 2.0 (the one that installed with pip) need Cuda 10.0, so I went to install that.
I followed this tutorial: https://towardsdatascience.com/how-to-install-tensorflow-gpu-on-ubuntu-18-04-1c1d2d6d6fd2
In particular, I did step 2, because the Nvidia drivers were already working. I went to the Nvidia website, downloaded Cuda 10.0, and I executed the 4 instructions in that tutorial, which were the same in the Nvidia website.
wget https://developer.nvidia.com/compute/cuda/10.0/Prod/local_installers/cuda-repo-ubuntu1804-10-0-local-10.0.130-410.48_1.0-1_amd64 sudo dpkg -i cuda-repo-ubuntu1804-10-0-local-10.0.130-410.48_1.0-1_amd64 sudo apt-key add /var/cuda-repo-10-0-local-10.0.130-410.48/7fa2af80.pub sudo apt-get update
Once I did this I realized that somehow it decided to downgraded my nvidia drivers to 410.48 and that the computer required a reboot.
I typed "nvidia-smi" and it gave me "Failed to initialize NVML: Driver/library version mismatch".
This is when I panicked, I searched for this on Google, and someone had solved something similar by removing cuda, so thought that if I uninstall Cuda, it will revert to the original configuration, and I did:
"sudo apt-get remove cuda"
So then I did "nvidia-smi" again, and I got the same error as before. And this is when I decided to reboot the computer with "sudo reboot" and the computer is not booting again.