[bugfix] Fix dummy-run and multi-node issues in MoE routing and MTP#4947
Merged
wangxiyuan merged 1 commit intovllm-project:mainfrom Dec 15, 2025
Merged
[bugfix] Fix dummy-run and multi-node issues in MoE routing and MTP#4947wangxiyuan merged 1 commit intovllm-project:mainfrom
wangxiyuan merged 1 commit intovllm-project:mainfrom
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request introduces fixes for Mixture-of-Experts (MoE) functionality, specifically targeting dummy-run and multi-node scenarios. The main changes involve updating the MoE communication method for FUSED_ALLTOALL to use MC2 components, which likely resolves issues in multi-node setups. Additionally, a guard on the expert parallelism size has been removed, enabling this path for larger configurations. A minor cleanup in a C++ kernel is also included. The changes appear to correctly address the intended fixes. However, I've pointed out that a docstring in moe_comm_method.py is now outdated due to these changes, which could impact future maintainability.
Contributor
|
👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:
If CI fails, you can run linting and testing checks locally according Contributing and Testing. |
kiscad
force-pushed
the
fix-smoking
branch
2 times, most recently
from
4020532 to
972aee2
Compare
kiscad
changed the title
kiscad
changed the title
kiscad
force-pushed
the
fix-smoking
branch
2 times, most recently
from
b7e217f to
7b3c477
Compare
Contributor
|
This pull request has conflicts, please resolve those before we can evaluate the pull request. |
kiscad
force-pushed
the
fix-smoking
branch
3 times, most recently
from
84c68f8 to
c8a182f
Compare
Signed-off-by: mojave2 <chenchen145@huawei.com>
weijinqian0
added
ready
weijinqian0
approved these changes
Dec 15, 2025
wangxiyuan
approved these changes
Dec 15, 2025
chenaoxuan
pushed a commit
to chenaoxuan/vllm-ascend
that referenced
this pull request
Dec 20, 2025
…llm-project#4947) ### What this PR does / why we need it? - Fix a premature `return` in `moe_init_routing_quant_v2.cpp` so the routing kernel completes correctly instead of exiting early in certain paths. - Switch `FusedAlltoAllCommImpl` to use the MC2-based token dispatcher and prepare/finalize routines, aligning MoE communication with the MC2 algorithm optimized for Ascend devices. - Add a temporary override in `MtpProposer` to map `FUSED_ALLTOALL` back to `ALLTOALL` until the MoE communication type selection logic is fully finalized, avoiding incorrect behavior in dummy-run flows. - Simplify the MoE communication selection for Ascend 910-93 in `NPUModelRunner` by removing the EP-size guard on `FUSED_ALLTOALL`, which fixes failures in multi-node / larger-EP configurations while keeping MC2 routing under the configured token capacity. - vLLM version: v0.12.0 - vLLM main: vllm-project/vllm@ad32e3e Signed-off-by: mojave2 <chenchen145@huawei.com>
ZRJ026
pushed a commit
to ZRJ026/vllm-ascend
that referenced
this pull request
Feb 28, 2026
…llm-project#4947) ### What this PR does / why we need it? - Fix a premature `return` in `moe_init_routing_quant_v2.cpp` so the routing kernel completes correctly instead of exiting early in certain paths. - Switch `FusedAlltoAllCommImpl` to use the MC2-based token dispatcher and prepare/finalize routines, aligning MoE communication with the MC2 algorithm optimized for Ascend devices. - Add a temporary override in `MtpProposer` to map `FUSED_ALLTOALL` back to `ALLTOALL` until the MoE communication type selection logic is fully finalized, avoiding incorrect behavior in dummy-run flows. - Simplify the MoE communication selection for Ascend 910-93 in `NPUModelRunner` by removing the EP-size guard on `FUSED_ALLTOALL`, which fixes failures in multi-node / larger-EP configurations while keeping MC2 routing under the configured token capacity. - vLLM version: v0.12.0 - vLLM main: vllm-project/vllm@ad32e3e Signed-off-by: mojave2 <chenchen145@huawei.com> Signed-off-by: zrj026 <zhangrunjiang026@gmail.com>
ZRJ026
pushed a commit
to ZRJ026/vllm-ascend
that referenced
this pull request
Mar 4, 2026
…llm-project#4947) ### What this PR does / why we need it? - Fix a premature `return` in `moe_init_routing_quant_v2.cpp` so the routing kernel completes correctly instead of exiting early in certain paths. - Switch `FusedAlltoAllCommImpl` to use the MC2-based token dispatcher and prepare/finalize routines, aligning MoE communication with the MC2 algorithm optimized for Ascend devices. - Add a temporary override in `MtpProposer` to map `FUSED_ALLTOALL` back to `ALLTOALL` until the MoE communication type selection logic is fully finalized, avoiding incorrect behavior in dummy-run flows. - Simplify the MoE communication selection for Ascend 910-93 in `NPUModelRunner` by removing the EP-size guard on `FUSED_ALLTOALL`, which fixes failures in multi-node / larger-EP configurations while keeping MC2 routing under the configured token capacity. - vLLM version: v0.12.0 - vLLM main: vllm-project/vllm@ad32e3e Signed-off-by: mojave2 <chenchen145@huawei.com> Signed-off-by: zrj026 <zhangrunjiang026@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
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.
What this PR does / why we need it?
returninmoe_init_routing_quant_v2.cppso the routing kernel completes correctly instead of exiting early in certain paths.FusedAlltoAllCommImplto use the MC2-based token dispatcher and prepare/finalize routines, aligning MoE communication with the MC2 algorithm optimized for Ascend devices.MtpProposerto mapFUSED_ALLTOALLback toALLTOALLuntil the MoE communication type selection logic is fully finalized, avoiding incorrect behavior in dummy-run flows.NPUModelRunnerby removing the EP-size guard onFUSED_ALLTOALL, which fixes failures in multi-node / larger-EP configurations while keeping MC2 routing under the configured token capacity.Does this PR introduce any user-facing change?
How was this patch tested?