sonic-moe: Add sonic-moe kernels #531
Open
adarshxs wants to merge 12 commits intohuggingface:mainfrom
Open
Conversation
danieldk
reviewed
Apr 8, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds SonicMoE as a Python-only kernel for accelerated Mixture-of-Experts on Hopper and Blackwell GPUs. SonicMoE uses CuTe-DSL grouped GEMMs and Triton routing kernels to deliver state-of-the-art MoE throughput with IO-aware tile scheduling - see arXiv:2512.14080. The kernel vendors QuACK (v0.2.5) for its GEMM infrastructure and declares nvidia-cutlass-dsl as a CUDA python-depends. Built and verified with
nix run .#build-and-copy, and tested end-to-end on H100 viaget_kernel("kernels-community/sonic-moe", version=1)with forward/backward correctness checks against the torch baseline.