[Revert] nvidia-cutlass-dsl[cu13] 4.5.1 -> 4.5.0#25938
Merged
Kangyan-Zhou merged 2 commits intoMay 21, 2026
Merged
Conversation
Contributor
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
Collaborator
Author
|
/tag-and-rerun-ci |
Kangyan-Zhou
force-pushed
the
fix_cutlass_dsl_variant
branch
from
dc34579 to
dc02d67
Compare
Kangyan-Zhou
changed the title
Contributor
|
Should force reinstall |
PR sgl-project#25576 bumped nvidia-cutlass-dsl[cu13] from 4.5.0 to 4.5.1. The bump exposed a latent file-level conflict between -libs-base and -libs-cu13 (both written by the additive [cu13] extra) as a hard GPUModuleOp TypeError on H100: -libs-cu13's pybind11 binding changed to the new MLIR-style ((operation: object)) without a matching bump to the Python wrapper in nvidia-cutlass-dsl, so loading -libs-cu13's .so makes the wrapper's old-style super().__init__() call fail. Two changes: 1. Revert the version bump (4.5.1 -> 4.5.0). At 4.5.0 both .so files expose a compatible binding, so the same coexistence no longer crashes. This removes the active TypeError on H100 and on the CUDA-13 Docker image for non-Blackwell users. 2. Add fix_cutlass_dsl_libs() to ci_install_dependency.sh, called from main() after download_flashinfer_cache. The function picks the right libs package per GPU family even at 4.5.0 to avoid two independent regressions that the silent conflict could still hit: Blackwell (IS_BLACKWELL=1, CU13): Purge -libs-base, force-reinstall -libs-cu13 so its files take precedence. -libs-base is CUDA-12.9-built and lacks the sm_110 arch alias that GB300/B200 need at cutlass import time. Non-Blackwell CU13 (H100, H200): Purge -libs-cu13, force-reinstall -libs-base. -libs-cu13 carries a CUDBG_EXCEPTION_WARP_ILLEGAL_ADDRESS regression in LoRA CUDA- graph capture on sm_90 (sgl-project#25743 / reverted by sgl-project#25756). Non-CU13: no-op (only -libs-base ever installed).
Kangyan-Zhou
force-pushed
the
fix_cutlass_dsl_variant
branch
from
e45a5d3 to
b04b737
Compare
Kangyan-Zhou
requested review from
Fridge003,
ispobock and
merrymercy
as code owners
github-actions
Bot
added
the
dependencies
Revert the version bump from PR sgl-project#25576. At 4.5.1, -libs-cu13's pybind11 binding changed to new MLIR-style ((operation: object)) without a matching bump to the Python wrapper in nvidia-cutlass-dsl, exposing the latent file-level conflict between -libs-base and -libs-cu13 (both written by the additive [cu13] extra) as a hard GPUModuleOp TypeError at kernel-compile time on CU13 runners. At 4.5.0 both .so files expose a compatible binding, so the same coexistence is silent and CI was empirically green on H100 and Blackwell during the post-sgl-project#25756, pre-sgl-project#25576 window. Going back to 4.5.0 restores that state. Supersedes sgl-project#25935 (which proposed the same revert but was closed).
Kangyan-Zhou
force-pushed
the
fix_cutlass_dsl_variant
branch
from
0e366cc to
da60371
Compare
Kangyan-Zhou
changed the title
Collaborator
Author
Let's revert this change first and then try your suggestions, this error is very confusing |
Kangyan-Zhou
added a commit
to Kangyan-Zhou/sglang
that referenced
this pull request
May 21, 2026
- Use "${REPO_ROOT}/python/pyproject.toml" instead of relative path so the
version probe doesn't depend on the working directory the script is
launched from (per gemini-code-assist review).
- Bump nvidia-cutlass-dsl[cu13] 4.5.0 -> 4.5.1 now that the wheel-mix
TypeError is mitigated by force_reinstall_cutlass_dsl_libs_cu13. This
re-applies sgl-project#25576 which was rolled back in sgl-project#25938 only because of the
install-order bug.
alisonshao
pushed a commit
that referenced
this pull request
May 21, 2026
- Use "${REPO_ROOT}/python/pyproject.toml" instead of relative path so the
version probe doesn't depend on the working directory the script is
launched from (per gemini-code-assist review).
- Bump nvidia-cutlass-dsl[cu13] 4.5.0 -> 4.5.1 now that the wheel-mix
TypeError is mitigated by force_reinstall_cutlass_dsl_libs_cu13. This
re-applies #25576 which was rolled back in #25938 only because of the
install-order bug.
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
3 participants
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.
Problem
nvidia-cutlass-dsl[cu13]has additive extras on PyPI: both-libs-baseAND-libs-cu13are installed together when[cu13]is requested. They write to the samesite-packagespaths with different content, causing aGPUModuleOp TypeErrorat kernel-compile time (vllm-project/vllm#40082).The correct libs package to keep depends on GPU family:
-libs-cu13wins-libs-base-libs-basewins-libs-cu13causesCUDBG_EXCEPTION_WARP_ILLEGAL_ADDRESSin LoRA CUDA-graph capture (#25743)-libs-baseinstalled, no conflictHistory
purge_cutlass_libs_base()(remove-libs-base, reinstall-libs-cu13) — fixed GB300 but caused H100 LoRA regressionGPUModuleOpTypeErrorChanges
Add
fix_cutlass_dsl_libs()called frommain()afterdownload_flashinfer_cache:-libs-base, force-reinstall-libs-cu13→ sm_110 support-libs-cu13, force-reinstall-libs-base→ no TypeError, no LoRA regressionTest Plan
base-b-test-*-gpu-large): eagle tests should pass (no moreGPUModuleOpTypeError), LoRA tests should remain greenbase-b-test-*-gpu-b200): sm_110 alias supported, cutlass import works🤖 Generated with Claude Code
CI States
Latest PR Test (Base): ⏳ Run #26208799937
Latest PR Test (Extra): ❌ Run #26208799771