Workaround for this is to install Tensorflow 1 (e.g., `pip install tensorflow==1.15.2`) instead of the default, which is now Tensorflow 2.