Refactor JIT kernel CI to use run_suite.py registration system#21239
Merged
merrymercy merged 3 commits intomainfrom Mar 24, 2026
Merged
Refactor JIT kernel CI to use run_suite.py registration system#21239merrymercy merged 3 commits intomainfrom
merrymercy merged 3 commits intomainfrom
Conversation
Migrate pr-test-jit-kernel.yml from raw pytest/shell invocation to the centralized run_suite.py + register_cuda_ci system, introducing three new suites: - stage-b-kernel-unit-1-gpu-large (per-commit unit tests) - stage-b-kernel-benchmark-1-gpu-large (per-commit benchmarks) - nightly-kernel-1-gpu (nightly full tests) Changes: - Add register_cuda_ci calls to all 35 test files and 23 benchmark files - Extend run_suite.py glob to discover jit_kernel tests/benchmarks - Register new suites in PER_COMMIT_SUITES and NIGHTLY_SUITES - Fix missing __main__ guards in test_nvfp4_*.py files - Fix test_custom_all_reduce.py __main__ to handle both torchrun worker and direct pytest invocation - Disable multi-GPU and self-skipping tests/benchmarks with reasons Made-with: Cursor
merrymercy
requested review from
BBuf,
DarkSharpness,
Fridge003,
HydraQYH,
Kangyan-Zhou,
bingxche,
celve,
ispobock and
yuan-luo
as code owners
Contributor
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
github-actions
bot
added
quant
Contributor
Author
|
/tag-and-rerun-ci |
… logs - Remove jit-kernel-unit-test-nightly from pr-test-jit-kernel.yml; add nightly-test-kernel-1-gpu-h100 to nightly-test-nvidia.yml with matching env (SGLANG_JIT_KERNEL_RUN_FULL_TESTS, SGLANG_JIT_DEEPGEMM_FAST_WARMUP, SGLANG_PR_TEST_BYPASS_MAINTENANCE_ON_MAIN) and workflow_dispatch filter. - Refresh register_cuda_ci est_time values using observed wall times from PR #21239 run 23465359984: jit-kernel-unit-test (68277933098) and jit-kernel-benchmark-test (68277933115). Per-commit times use ~1.15x elapsed; nightly uses 4x stage-b (min 120, cap 1200) except FA4 (120/900; FA4 elapsed is a skip on H100). Made-with: Cursor
1 task
adityavaid
pushed a commit
to adityavaid/sglang
that referenced
this pull request
Mar 24, 2026
adityavaid
pushed a commit
to adityavaid/sglang
that referenced
this pull request
Mar 24, 2026
0-693
pushed a commit
to 0-693/sglang
that referenced
this pull request
Mar 25, 2026
johnnycxm
pushed a commit
to johnnycxm/sglang
that referenced
this pull request
Mar 25, 2026
johnnycxm
pushed a commit
to johnnycxm/sglang
that referenced
this pull request
Mar 25, 2026
5 tasks
JustinTong0323
pushed a commit
to JustinTong0323/sglang
that referenced
this pull request
Apr 7, 2026
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
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.
Summary
pr-test-jit-kernel.ymlfrom rawpytest/shell invocation to the centralizedrun_suite.py+register_cuda_ciregistration systemstage-b-kernel-unit-1-gpu-large,stage-b-kernel-benchmark-1-gpu-large,nightly-kernel-1-gpuregister_cuda_cicalls to all 35 test files and 23 benchmark files underpython/sglang/jit_kernel/run_suite.pyglob to discover JIT kernel tests and benchmarks alongsidetest/registered/__main__guards intest_nvfp4_*.pyfiles, fixtest_custom_all_reduce.pyto dispatch between torchrun worker and pytestTest plan
jit-kernel-unit-testjob discovers and runs all unit tests viarun_suite.py --hw cuda --suite stage-b-kernel-unit-1-gpu-largejit-kernel-benchmark-testjob discovers and runs all benchmarks viarun_suite.py --hw cuda --suite stage-b-kernel-benchmark-1-gpu-largejit-kernel-unit-test-nightlyjob runs full tests viarun_suite.py --hw cuda --suite nightly-kernel-1-gpu --nightlytest_custom_all_reduce,bench_custom_all_reduce,bench_norm_impls) are correctly skippedMade with Cursor