Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Rocauc refactor #533
There is currently a bug in
It seems like the bug occurs because the default params for
This PR attempts to address the bug by:
Let me know what you think and if you have any suggestions!
left a comment
I think this is an excellent solution (it took me a few minutes to wrap my head around it) but it makes sense to me, and I think that it clearly communicates to the user what needs to be done in the case you discovered! Thank you also for adding the three tests for each of the three possibilities for methods having these scoring functions!
I had a few comments for you to review, but I think once we get the CI tests passing we can go ahead and merge this in.
We should probably also create an issue to document this situation in the ROCAUC docs, and follow up on that while its fresh in our heads.
Thank you so much @rebeccabilbro for taking on this challenging bug and sticking with it until you found a solution!
@bbengfort - ok, just pushed all those updates. One question on your suggestion to add a new issue for us to capture the binary decision case in the YB docs; #252 does have a bullet for docs, and this current PR won't close the issue because there are still a couple of todos associated with the issue. I think instead of opening a new issue, once this is merged in, I'll just open a new PR to wrap up the remaining stuff from the issue. Sound good?