Tensorflow doesn't detect GPU when running jupiter notebook

Everything was ok until i executed a cell with "!pip install.." in it. Up until that point, I was running my notebook without any problems, and my GPU was being detected.

Currently, I have Python 3.10.13 installed along with the following:

• cudatoolkit 11.2.2 h933977f_10
• cudnn 8.1.0.77 h3e0f4f4_0
• tensorflow 2.10.1 pypi_0

to make it work on windows native.

Now if i run a .py with the same conda enviroment everything works as it should.

import tensorflow as tfprint(tf.config.list_physical_devices('GPU'))

Result:[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

However, if i run the same code in a notebook with the same enviroment, no GPU will be found.

I've tried creating a new conda enviroment, a new notebook, and cloned the repository in another folder, but the issue persists.