Skip to content

Conversation

ntjohnson1
Copy link
Collaborator

@ntjohnson1 ntjohnson1 commented Sep 2, 2025

Closes #451
Reiterating some discussion on the ticket. Basically things were flaky because I used an adhoc method to generate bad steps to reject. If we just negate the function handle then as we approach the true solution we are approaching the maximum rather than the minimum. I ran this like 50 times and didn't get a flake.


📚 Documentation preview 📚: https://pyttb--452.org.readthedocs.build/en/452/

@ntjohnson1
Copy link
Collaborator Author

CC: @jeremy-myers in case I misdiagnosed or it helps clarify

@dmdunla
Copy link
Collaborator

dmdunla commented Sep 2, 2025

@ntjohnson I like the current fix for the bad step rejection, and I will merge this PR shortly. However, I do think that addressing #441 will help with reproducibility and stability of the tests in the long run. A robust RNG solution for pyttb may not have helped in this situation, but while tracking down this issue, I noticed that testing could use a consistent approach for using RNGs for reproducibility.

@dmdunla dmdunla merged commit d0254cd into sandialabs:main Sep 2, 2025
11 checks passed
@ntjohnson1
Copy link
Collaborator Author

Agreed. #441 still makes sense, but I expect it to be a moderate amount of plumbing. Pending how long the documentation overhaul is planned for it might make sense to start after to avoid conflicts.

@ntjohnson1 ntjohnson1 deleted the nick/fix_gcp_test branch September 2, 2025 17:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Potential issue with RNG in gcp_opt testing

2 participants