Fix Nightly NV CI - #33564
Merged
Merged
Fix Nightly NV CI#33564
Conversation
b8zhong
requested review from
BBuf,
Edwardf0t1,
Fridge003,
HaiShaw,
Ying1123,
ch-wan,
ispobock and
merrymercy
as code owners
August 4, 2026 15:10
Contributor
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
Collaborator
Author
|
/rerun-test test/registered/8-gpu-models/test_longcat_flash_lite_fp8.py |
Contributor
|
Results for 🚀 🚀 |
b8zhong
force-pushed
the
brayden-fix-longcat-ngram-oe-vocab-size
branch
from
August 4, 2026 15:23
7f2b9d9 to
4ee81f1
Compare
Collaborator
Author
|
/rerun-test test/registered/8-gpu-models/test_glm52_fp8.py |
Contributor
|
Results for 🚀 🚀 |
b8zhong
changed the title
LongcatFlashConfig normalizes architectures to LongcatFlashForCausalLM, which
sits in _DEEPSEEK_FAMILY_ARCHS, so _deepseek_moe_quant_resolution forced
moe_runner_backend=flashinfer_trtllm on SM100. LongCat picks its top-12 from
n_routed_experts + zero_expert_num (256 + 128) logits and splits the identity
experts out afterwards, which trtllm-gen's internal routing cannot express --
it only sees the 256 routed logits. The nightly B200 run hit the guard:
assert TopKOutputChecker.format_is_bypassed(topk_output)
Exclude the LongCat archs from that override so they fall back to auto
(deep_gemm/triton), the path that scored gsm8k 0.935 on H200.
Also skip the TP8+EP8+deepep variant: DeepEP's low-latency dispatch asserts
num_topk <= kNumMaxTopK (11 in internode_ll.cu) and LongCat moe_topk is 12,
so the variant has never been runnable since #30975 made --moe-a2a-backend
take effect for LongCat.
Collaborator
Author
|
/rerun-test test/registered/8-gpu-models/test_longcat_flash_lite_fp8.py |
Contributor
|
Results for 🚀 🚀 |
Fridge003
approved these changes
Aug 6, 2026
sagearc
pushed a commit
to sagearc/sglang
that referenced
this pull request
Aug 13, 2026
sgl-project#33779) Co-authored-by: Brayden Zhong <b8zhong@uwaterloo.ca> Signed-off-by: Sage Ahrac <sagiahrak@gmail.com>
saturn-acc
pushed a commit
to saturn-acc/sglang
that referenced
this pull request
Aug 16, 2026
Co-authored-by: Brayden Zhong <brayden@radixark.ai> Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
Atituiset
pushed a commit
to Atituiset/sglang
that referenced
this pull request
Sep 10, 2026
Co-authored-by: Brayden Zhong <brayden@radixark.ai> Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
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
2 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.
NgramEmbeddingpassedexclusive_oe_embedder_size_sums[-1](a 0-dim CUDA tensor) asnum_embeddings, so the fused Triton embedding kernel got a tensor where it wants atl.constexprint and failed to compile. Wrap it inint().Fixes nightly
test_longcat_flash_lite_fp8.py:Verified with a 2-rank repro at the real LongCat-Flash-Lite config: crashes before, passes after, output bit-identical to the non-Triton masked path.
CI States
Latest PR Test (Base): 🚫 Run #31061386671
Latest PR Test (Extra): ❌ Run #31061386535