[None][CI] Waive flaky test_vbench_dimension_score_wan (nvbugs/6357628)#15529
Open
chang-l wants to merge 1 commit into
Open
[None][CI] Waive flaky test_vbench_dimension_score_wan (nvbugs/6357628)#15529chang-l wants to merge 1 commit into
chang-l wants to merge 1 commit into
Conversation
VBench per-dimension scores (subject_consistency, aesthetic_quality, dynamic_degree) vary far more than the test's 0.05 band across its non-deterministic generation seed (dynamic_degree is binary 0/1), so the test fails intermittently on different dimensions. Reproduced FAIL/PASS/FAIL across three identical 4-GPU runs at upstream/main HEAD; not a code regression. Tracked by nvbugs/6357628. Signed-off-by: Chang Liu <liuc@nvidia.com>
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Enterprise Run ID: 📒 Files selected for processing (1)
📝 WalkthroughWalkthroughA single SKIP waiver entry for ChangesTest Waiver Addition
Estimated code review effort🎯 1 (Trivial) | ⏱️ ~2 minutes 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Waive the flaky test
examples/visual_gen/test_visual_gen.py::test_vbench_dimension_score_wan(SKIP), tracked by nvbugs/6357628.Why
The test is flaky by design: it generates a video with a non-deterministic seed (
VisualGenParams.seeddefaults toNone) and compares absolute per-dimension VBench scores against a single fixed-seed HF-reference golden with a uniform 0.05 band. Several dimensions vary far more than 0.05 across seeds —dynamic_degreeis effectively binary (0/1), andaesthetic_quality/imaging_qualityswing 0.15–0.25.Reproduced on
upstream/mainHEAD (B200, exact CI config: 4 GPU,cfg_size=2,ulysses_size=1, full 6-dimension VBench), three identical-config runs → FAIL / PASS / FAIL, each failing on a different dimension. Single-GPU and 2-GPU both pass at a fixed seed, so this is not a code regression (suspect commits #13978 / #14961 / #14687 ruled out).This mirrors the already-waived sibling tests
test_vbench_dimension_score_wan22_a14b_fp8/_nvfp4(nvbugs/6310230).Recommended real fix (for visual_gen owners @NVIDIA/trt-llm-torch-visual-gen-devs)
Pin a fixed generation seed (verified deterministic), re-baseline all 6 goldens from pinned-seed runs, and special-case the binary
dynamic_degree(exact match, not a 0.05 band). Tracked in nvbugs/6357628.🤖 Generated with Claude Code
Summary by CodeRabbit