Skip to content

fix: drop orphaned sharded_moe - #31139

Closed
GaetanLepage wants to merge 1 commit into
microsoft:mainfrom
GaetanLepage:drop-orphaned-sharded-moe
Closed

fix: drop orphaned sharded_moe#31139
GaetanLepage wants to merge 1 commit into
microsoft:mainfrom
GaetanLepage:drop-orphaned-sharded-moe

Conversation

@GaetanLepage

@GaetanLepage GaetanLepage commented Jul 29, 2026

Copy link
Copy Markdown

Description

Drop the orphaned ShardedMoE CUDA contrib op.

Motivation and Context

ShardedMoE's backend, contrib_ops/cuda/moe/ft_moe/moe_kernel.h and the ort_fastertransformer::CutlassMoeFCRunner it wraps, was removed in the MoE GEMM refactor (#28467) which moved everything to contrib_ops/cuda/llm/moe_gemm/.
moe.cc/moe_base.h were ported to the new onnxruntime::llm::kernels::cutlass_kernels API, but sharded_moe.{h,cc} were not, so the file still includes the deleted header and fails to compile whenever onnxruntime_USE_NCCL is enabled:

contrib_ops/cuda/collective/sharded_moe.h:6:10: fatal error:
contrib_ops/cuda/moe/ft_moe/moe_kernel.h: No such file or directory

The op has been dead (uncompilable) since that refactor.
Remove it along with its build/registration entries so NCCL builds again; all other NCCL collective ops (AllReduce, AllGather, AllToAll, DistributedMatMul, ...) are unaffected.

Related:
- #28467 (comment) (cc @SuperSandro2000)
- NixOS/nixpkgs#546053

@azure-pipelines

Copy link
Copy Markdown
Azure Pipelines:
There may be pipelines that require an authorized user to comment /azp run to run.

@tianleiwu

Copy link
Copy Markdown
Contributor

See another PR #31191 for this issue.

@tianleiwu tianleiwu closed this Jul 30, 2026
@GaetanLepage
GaetanLepage deleted the drop-orphaned-sharded-moe branch July 30, 2026 23:09
@GaetanLepage

Copy link
Copy Markdown
Author

Indeed, sorry for the noise.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants