In this project I finally switched to using pipenv instead of conda. I also tried virtualenv :|

First you need to install linuxbrew or macbrew then use to install pipenv

Install pipenv

brew install pipenv
pipenv lock

Add env to kernel

pipenv run python -m ipykernel install --user --name=line-reader

Install dependencies

pipenv install --dev

Update dependencies

pipenv sync --dev

Install new dependency

pipenv install some_dependency

Remove the environment

pipenv --rm

To avoid switching between tensorflow gpu and cpu versions when moving my code between my local and the cloud I installed relevant version using pip instead

pipenv run pip install tensorflow==2.0.0-beta1
# Or
pipenv run pip install tensorflow-gpu==2.0.0-beta1