[AMD] perf: enable FlyDSL w4a16 MoE for Kimi INT4#1777
Open
amd-asalykov wants to merge 3 commits into
Open
Conversation
seungrokj
reviewed
Jun 15, 2026
seungrokj
reviewed
Jun 15, 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Replace default triton w4a16 MoE kernel with more performant FlyDSL implementation for Kimi INT4 MI355X
Note
Low Risk
Benchmark and serving-flag changes only; no application auth or data paths. Risk is limited to reproducibility and CI cost from expanded sweeps and a nightly container pin.
Overview
Updates the Kimi K2.5 INT4 vLLM MI355X benchmark to use FlyDSL for w4a16 MoE instead of the default Triton path, and pins a digest-suffixed ROCm nightly image (
vllm-openai-rocm:nightly-b8336c3…).The runner script
kimik2.5_int4_mi355x.shadds--moe-backend flydsland a compilation pass that setsfuse_allreduce_rmsto false. CI config expands the fixed-seq-len sweep: concurrency up to 128 (from 64) and an additional TP=4 row for both 1k/1k and 8k/1k scenarios.perf-changelog.yamlrecords the config-key change for PR #1777.Reviewed by Cursor Bugbot for commit be23347. Bugbot is set up for automated code reviews on this repo. Configure here.