This paper represents a massive amount of work that will have significant impact on research practices in ML. Reviewers are strongly positive, with the only concern being that this is about ML practice and not ML itself. I do not share this concern.