TensorFlow Extended, version control and .gitignore

When working on a dataset with TensorFlow, generating the examples, statistics, schema, transform graph, etc. and using git as a version control for the Jupyter Notebook, what is the best practice regarding all the files created by TensorFlow?
Should everything also be stored in the version control system with git, or should part of it be kept away using .gitignore?
What is the best practice?
Thank you.

If you want to save the status of a run for sharing with colleagues or to use in another project, check them into github. It’s common not to check generated artifacts into github since the notebook should be able to create them on a new environment.