[CI] fix bench_jit_per_token_group_quant_8bit sglang import cascade - #453
Open
arathi-hlab wants to merge 3 commits into
Open
arathi-hlab wants to merge 3 commits into
arathi-hlab wants to merge 3 commits into
Conversation
arathi-hlab
force-pushed
the
ci/fix-jit-8bit-bench-sglang-import
branch
from
August 27, 2026 17:35
8aabfc4 to
a1e16f3
Compare
PR sgl-project#450 skipped this bench because it raised `ModuleNotFoundError: No module named 'orjson'`. Root cause is that sgl_kernel.gemm.sgl_per_token_group_quant_8bit only reads `SGLANG_PER_TOKEN_GROUP_QUANT_8BIT_V2` via `sglang.srt.utils.get_bool_env_var` when `enable_v2 is None`; that path imports the full sglang package, whose common.py needs orjson, psutil, pybase64, etc. — none installed in the bench container. Passing enable_v2=False from the bench short-circuits the sglang lookup entirely. v2 is already covered by bench_jit_per_token_group_quant_8bit_v2.py, so this bench legitimately wants the v1 kernel. - bench: pin enable_v2=False in aot_per_token_group_quant_8bit. - CI: run bench_jit_per_token_group_quant_8bit.py again instead of the skip echo added in sgl-project#450.
arathi-hlab
force-pushed
the
ci/fix-jit-8bit-bench-sglang-import
branch
from
August 27, 2026 17:35
a1e16f3 to
fd910ae
Compare
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Follow-up to #443. The
Run Sglang Kernel Benchmarksstep is now failing onbench_jit_per_token_group_quant_8bit.pywith:Root cause
sgl_kernel.gemm.sgl_per_token_group_quant_8bitonly importssglang.srt.utils.get_bool_env_varwhenenable_v2 is None, to read theSGLANG_PER_TOKEN_GROUP_QUANT_8BIT_V2env var. The bench was leavingenable_v2unset, soPYTHONPATH=/root/sglang/python(added in #443 to unblock this) made sglang findable — butsglang/__init__.pycascades intosrt/utils/common.pywhich pullsorjson,psutil,pybase64,requests,starlette,torchvision, … none installed in the CI container. Chasing those withpip installs is fragile.Fix
enable_v2=Falsein the AOT calls inbench_jit_per_token_group_quant_8bit.pysosgl_kernelnever reaches the sglang-lookup branch. v2 is already covered bybench_jit_per_token_group_quant_8bit_v2.py.export PYTHONPATH=/root/sglang/pythonfrom the bench step; nothing else in that step needs it.