This repository was archived by the owner on Jul 1, 2026. It is now read-only.
perf(tests): share canonical humaneval-multiturn rollout + widen secret gitignore - #8
Merged
Merged
Conversation
…secret gitignore Audit P-1: tests/test_code_humaneval_multiturn.py ran FIVE full 3-turn rollouts, each spawning a sandboxed pytest subprocess per turn (~7 s per rollout, 22.5 s across the three that exercise the SAME canonical run). The three tests asserting different properties of one [bad, gold, gold] rollout (canonical dict shape, history growth, per-turn components) now consume a single module-scoped fixture. Measured: 22.5 s -> 7.7 s for those three (-14.8 s, -60%); file wall 40 s -> 24.8 s. Test count unchanged (22 in file; suite stays 1917 passed / 7 skipped / 2 deselected / 0 failed). Kept separate on purpose: the turn-penalty test (needs base_reward exactly 1.0 -> gold x3 script) and the max_turns-override test (different rollout params). Audit L-1: the Secrets block of .gitignore covered only .env, .env.local, *.key. Widened to the standard credential-shaped set (.env.*, secrets.*, credentials.*, .netrc, *.pem, *.p12, *.pfx, id_rsa*/ed25519*/ecdsa*, .aws/.gcp/.azure dirs). No tracked file matches any new pattern (verified), so this is purely preventive.
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Two approved items from the repo audit (P-1 perf + L-1 hygiene). No audit documents are in this PR — only the fixes.
P-1 — shared rollout fixture (measured)
tests/test_code_humaneval_multiturn.pyran five full 3-turn rollouts; each turn spawns a sandboxed pytest subprocess (~7 s/rollout). Three of the five assert different properties of the SAME canonical[bad, gold, gold]run — they now share one module-scoped fixture.Kept separate by design: turn-penalty (needs gold×3 for
base_reward == 1.0) and max_turns-override (different rollout params).L-1 — .gitignore secret-pattern widening
Secrets block grew from
{.env, .env.local, *.key}to the standard credential-shaped set (.env.*,secrets.*,credentials.*,.netrc,*.pem,*.p12,*.pfx, SSH key names, cloud-CLI cred dirs). Verified zero tracked files match any new pattern — purely preventive.Audit corrections for the record
git ls-filesshows 0 tracked cache files; nothing to untrack.