[https://nvbugs/6669206][test] Unwaive test_kimi_k3_gen_dep[1] fixed by #18226 - #18371
Conversation
NVIDIA#18226 The Kimi K3 MLA decode backend-selection defect behind this failure was fixed in NVIDIA#18226 (keep CuTe-DSL MLA decode for H=96 speculative-verify batches). Remove the now-stale waiver so unittest/tools/test_layer_wise_benchmarks.py::test_kimi_k3_gen_dep[1] runs again on B200. Verified locally on B200 with the fix present: the test passes (1 passed, ~105s). Signed-off-by: Brian Nguyen <brnguyen@nvidia.com>
|
/bot run --stage-list "DGX_B200-PyTorch-*" |
|
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)
💤 Files with no reviewable changes (1)
Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review. WalkthroughThe change removes the waiver entry for Changes
Estimated code review effort: 1 (Trivial) | ~2 minutes Merge Risk: ⚪ Minimal · up to This localized change re-enables an existing test after its underlying fix was merged; no actionable merge-blocking risk remains beyond normal checks and review. 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
Full details: Docstring CoverageExplanation No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0 files. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
|
PR_Github #69983 [ run ] triggered by Bot. Commit: |
|
/bot skip --comment "Waiver-removal only (one line); the underlying fix landed in #18226. The unwaived test unittest/tools/test_layer_wise_benchmarks.py::test_kimi_k3_gen_dep[1] passed in the targeted DGX_B200-PyTorch CI run (L0_MergeRequest_PR #57263, single-GPU test job 63/63 stages green)." |
|
PR_Github #69983 [ run ] completed with state |
|
PR_Github #70032 [ skip ] triggered by Bot. Commit: |
|
PR_Github #70032 [ skip ] completed with state |
Description
The Kimi K3 MLA decode backend-selection defect behind
test_kimi_k3_gen_dep[1](nvbugs/6669206) was fixed in #18226, which keeps CuTe-DSL MLA decode for H=96 speculative-verify batches instead of downgrading them to trtllm-gen (whose MLA decode rejects64 < num_heads_q < 128). That fix is already merged, but the test's waiver inwaives.txtwas left behind, so the test remains skipped in CI. This PR removes the now-stale waiver to re-enable it.Test Coverage
unittest/tools/test_layer_wise_benchmarks.py::test_kimi_k3_gen_dep[1](single-GPU, DGX_B200-PyTorch). Verified passing locally on B200 with the merged fix (1 passed, ~105s).PR Checklist
Dev Engineer Review
unittest/tools/test_layer_wise_benchmarks.py::test_kimi_k3_gen_dep[1].waives.txtformat.#18226and local B200 verification.QA Engineer Review
test-db/orqa/files were modified.tests/integration/test_lists/waives.txt.