BUG: restore RandomState multinomial stream compatibility - #32300
Draft
vetrovk wants to merge 1 commit into
Draft
Conversation
Copy link
Copy Markdown
Sorry, something went wrong.
Copy link
Copy Markdown
Member
|
I'm converting this to draft because the discussion in #32299 is leaning toward just documenting the behavior change. |
Sorry, something went wrong.
ngoldbaum
marked this pull request as draft
August 18, 2026 16:46
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PR summary
Fixes #32299.
This restores the pre-2.5 random stream behavior of RandomState.multinomial without reverting the Generator.multinomial fix from #31238.
The regression was caused by routing RandomState.multinomial through legacy_random_binomial, which changed both zero-probability RNG consumption and the inversion path. This PR adds a multinomial-specific legacy helper that preserves the previous behavior while retaining the corrected Generator behavior.
Regression tests cover the reported sequence, RNG stream consumption, the inversion path, and the existing BTPE behavior.
First time committer introduction
Hi, I'm Kirill. This is my first contribution to NumPy.
I came across #32299 while looking for a small bug I could help with. I was interested in it because the change affects reproducibility of the legacy RandomState API. I reproduced the issue on different NumPy versions and worked through the regression to understand what had changed.
I'm still getting familiar with the NumPy codebase, so feedback is very welcome.
AI Disclosure
I used OpenAI Codex as a development assistant for repository exploration, reproducing the issue, tracing the relevant code paths, implementing the fix, and running tests and validation.
I reviewed and understood the resulting changes and verified the behavior and test results myself. The pull request is being submitted and all project communication is being handled by me.