Fix few-shot sampler isolation - #1334
Open
atirna wants to merge 1 commit into
Open
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
I fixed the few-shot sampler so sequential variance seeds rotate independent copies of the task pool. I also fixed balanced sampling to retain falsy labels and use a local seeded RNG instead of mutating process-global random state.
The existing sequential assertion now compares the sampled offset with the unchanged pool. Previously, its expected value was read from the pool after sampling had mutated it.
Fixes #1307
Fixes #1309
Why
Sequential variance runs should choose offsets from the original few-shot pool for each seed, without changing later sampling or other consumers of the task pool. Balanced sampling should preserve valid empty labels and not affect unrelated random consumers.
Verification
main: the candidate sampler regression tests produced 4 failures and 2 passesuv run --no-sync pytest tests/unit/prompt/test_prompt_manager.py -q(6 passed)uv run --no-sync ruff format --check src/lighteval/tasks/prompt_manager.py tests/unit/prompt/test_prompt_manager.pyuv run --no-sync ruff check src/lighteval/tasks/prompt_manager.py tests/unit/prompt/test_prompt_manager.py