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
brew install pipenv
pipenv lock
pipenv run python -m ipykernel install --user --name=line-reader
pipenv install --dev
pipenv sync --dev
pipenv install some_dependency
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