[Recipe][Docs] Add MXFP4 intermediate variable and scheduler delay advice for Kimi-K2 - #1566
Merged
Conversation
Added environment variable for MXFP4 intermediate processing and recommended scheduler delay factor for large concs.
Contributor
🏷️ CI GuideRuns automatically on every eligible PR before approval:
Heavy model tests:
|
valarLip
approved these changes
Jul 12, 2026
wanzhenchn
added a commit
that referenced
this pull request
Jul 13, 2026
gyohuangxin
pushed a commit
that referenced
this pull request
Jul 14, 2026
…rting (#1580) * [ci][mesh] isolate ATOMesh worker compile caches * [ci][mesh] add kimi-k25-2p1d-dpa-tp4 case adjust pd_worker_layout for kimi-2p1d-dpa * Add ATOMesh case-level benchmark selection * remove kimi-2.5 decode dpa and tbo conifg * Revert "remove kimi-2.5 decode dpa and tbo conifg" This reverts commit d04f3f8. * Revert "[ci][mesh] add kimi-k25-2p1d-dpa-tp4 case" This reverts commit 9a3c6a8. * update Kimi-K2 env. config from PR ##1566 * fix Image source for dashboard * fix dashboard rerun selection by throughput * split atomesh dashboard charts by model * fix ATOMesh pd rank mapping for Spur cluster * fix code style --------- Co-authored-by: yuechguo <yuechguo@amd.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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Added environment variable for MXFP4 intermediate processing and recommended scheduler delay factor for large concs.
AITER_MXFP4_INTERMEDIATE=1added in [RadeonFlow] MXFP4 (a4w4) MoE backend for gfx950 aiter#3832 can makes moe gemm2 output directly out fp4 output, which can be then consumed directly byAITER_QUICK_REDUCE_QUANTIZATION=INT4Motivation
Technical Details
Test Plan
Test Result
Submission Checklist