Conversation
elvircrn
force-pushed
the
topk-ids-optional-do-expand
branch
6 times, most recently
from
June 10, 2026 21:53
0195a68 to
66c31f0
Compare
elvircrn
force-pushed
the
topk-ids-optional-do-expand
branch
from
June 11, 2026 08:41
66c31f0 to
9caa69c
Compare
elvircrn
added a commit
to elvircrn/vllm
that referenced
this pull request
Jun 11, 2026
…=True Adds do_expand as independent constructor param. When do_expand=True + use_cudagraph=True, routes via expert_tokens_meta without topk_ids. Keeps use_nvfp4_dispatch support.
elvircrn
added a commit
to elvircrn/vllm
that referenced
this pull request
Jun 11, 2026
elvircrn
added a commit
to elvircrn/vllm
that referenced
this pull request
Jun 11, 2026
- deepep_v2: build topk_ids from expert counts for do_expand+cudagraph instead of returning None (which _prepare treats as "keep original") - flashinfer_cutedsl_moe: use topk_ids.size(1) as effective top_k so expanded layout (top_k=1 per row) works with FlashInfer CuteDSL - fused_moe_modular_method: remove shared_experts/inplace args removed in PR vllm-project#45193
Contributor
|
This pull request has merge conflicts that must be resolved before it can be |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
With do_expand=True, DeepEP v2 returns tokens contiguous per expert and expert_tokens_meta provides routing via GPU cumsum. topk_ids are not needed.
moe_problem_sizeaccepttopk_ids=None, default topk=1_receiverwhenrecv_topk_idx is None_finalizewhentopk_ids is None