[AMD] dsv4-fp4-mi355x-sglang: switch fixed-seq-len search space to TP4 by seungrokj · Pull Request #1768 · SemiAnalysisAI/InferenceX

seungrokj · 2026-06-15T03:59:13Z

Summary

Switch dsv4-fp4-mi355x-sglang fixed-seq-len search space from TP8 to TP4 for both isl=1024 and isl=8192 scenarios
Expand isl=8192 coverage: add TP4 dp-attn sweep (conc 32–2048) alongside TP4 TP-only sweep (conc 1–32)
Update perf-changelog.yaml with change description

Test plan

Verify dsv4-fp4-mi355x-sglang fixed-seq-len CI jobs launch with TP4 configs
Confirm isl=8192 dp-attn and TP-only rows both appear in sweep results

🤖 Generated with Claude Code

Note

Low Risk
YAML-only benchmark matrix changes; no serving or application code paths are modified, though CI will run different TP/concurrency combinations on MI355X.

Overview
Updates dsv4-fp4-mi355x-sglang in .github/configs/amd-master.yaml so fixed-seq-len jobs emphasize TP4 alongside the existing TP8 + dp-attn high-concurrency row for both isl=1024 and isl=8192.

For each scenario, the former TP8, dp-attn=false sweep (conc 1–32) is replaced by TP4 rows: dp-attn=true (conc 16–128) and dp-attn=false (conc 1–32). The TP8, dp-attn=true sweep (conc 64–2048) is unchanged.

Adds a perf-changelog.yaml entry for dsv4-fp4-mi355x-sglang describing the TP4 search-space shift and expanded isl=8192 coverage.

^{Reviewed by Cursor Bugbot for commit b519ea6. Bugbot is set up for automated code reviews on this repo. Configure here.}

…ter yaml Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…able isl=1024 and expand isl=8192 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

… space update Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

github-actions · 2026-06-15T03:59:20Z

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers.

If additional help is needed, PR authors can reach out to core maintainers over Slack.

functionstackx

@seungrokj can u try looking into sgl mtp too? would switching it help? +viz @1am9trash

…dp-attn conc=512 point Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…oint for dsv4-fp4-mi355x-sglang Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…n and TP4 TP-only rows for isl=1024 and isl=8192 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

github-actions · 2026-06-15T06:41:14Z

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27528680882
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27528680882

github-actions · 2026-06-15T09:37:35Z

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27528759942
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27528759942

github-actions · 2026-06-15T15:05:36Z

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27528759942
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27528759942

seungrokj · 2026-06-15T15:06:50Z

@functionstackx @cquil11 can you plz approve this ?

billishyahao

LGTM

seungrokj · 2026-06-16T01:53:33Z

/reuse-sweep-run

seungrokj · 2026-06-16T01:53:50Z

/merge-prs

Oseltamivir

lgtm

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit b519ea6. Configure here.}

cursor · 2026-06-16T01:57:56Z

      search-space:
      - { tp: 8, dp-attn: true, conc-start: 64, conc-end: 2048 }
-      - { tp: 8, dp-attn: false, conc-start: 1 , conc-end: 32 }
+      - { tp: 4, dp-attn: true, conc-start: 16, conc-end: 128 }


TP4 dp-attn conc mismatch

Medium Severity

The new TP4 dp-attn: true fixed-seq-len rows use conc-start 16 and conc-end 128, but the same PR’s perf-changelog.yaml entry documents that sweep as concurrency 32–2048. The CI sweep will not cover the high-concurrency TP4 dp-attn range the changelog and PR describe.

Additional Locations (2)

.github/configs/amd-master.yaml#L2157-L2158

perf-changelog.yaml#L3837-L3838

^{Reviewed by Cursor Bugbot for commit b519ea6. Configure here.}

seungrokj and others added 3 commits June 15, 2026 11:35

[AMD] update dsv4-fp4-mi355x-sglang fixed-seq-len search space in mas…

c8541ac

…ter yaml Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

[AMD] update dsv4-fp4-mi355x-sglang fixed-seq-len search space: re-en…

a1f0502

…able isl=1024 and expand isl=8192 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

[AMD] perf-changelog: dsv4-fp4-mi355x-sglang TP4 fixed-seq-len search…

9628d11

… space update Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

seungrokj requested a review from a team June 15, 2026 03:59

seungrokj requested review from 1am9trash, billishyahao, chunfangamd and yctseng0211 as code owners June 15, 2026 03:59

github-project-automation Bot added this to InferenceMAX Board Jun 15, 2026

seungrokj added the AMD label Jun 15, 2026

functionstackx reviewed Jun 15, 2026

View reviewed changes

cursor Bot reviewed Jun 15, 2026

View reviewed changes

Comment thread .github/configs/amd-master.yaml Outdated

[AMD] narrow dsv4-fp4-mi355x-sglang fixed-seq-len to single isl=8192 …

50b4a94

…dp-attn conc=512 point Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

cursor Bot reviewed Jun 15, 2026

View reviewed changes

Comment thread .github/configs/amd-master.yaml Outdated

seungrokj and others added 4 commits June 15, 2026 13:34

[AMD] add memory budget comments and set isl=1024 dp-attn conc=2048 p…

d7b8862

…oint for dsv4-fp4-mi355x-sglang Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

[AMD] clarify memory budget comments for dsv4-fp4-mi355x-sglang

4327aab

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

[AMD] expand dsv4-fp4-mi355x-sglang fixed-seq-len: add TP8/TP4 dp-att…

24bde2a

…n and TP4 TP-only rows for isl=1024 and isl=8192 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Merge branch 'main' into amd/dsv4_sgl_0616

c6daac8

seungrokj added the full-sweep-enabled label Jun 15, 2026

[AMD] remove stale memory budget comments from amd-master.yaml

64213b9

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

billishyahao approved these changes Jun 16, 2026

View reviewed changes

Oseltamivir approved these changes Jun 16, 2026

View reviewed changes

Merge branch 'main' into amd/dsv4_sgl_0616

b519ea6

seungrokj merged commit 12fa1d2 into main Jun 16, 2026
25 checks passed

seungrokj deleted the amd/dsv4_sgl_0616 branch June 16, 2026 01:57

github-project-automation Bot moved this to Done in InferenceMAX Board Jun 16, 2026

cursor Bot reviewed Jun 16, 2026

View reviewed changes

Oseltamivir mentioned this pull request Jun 16, 2026

[AMD] perf-changelog: duplicate dsv4-fp4-mi355x-sglang TP4 fixed-seq-len entry #1795

Open

1 task

Conversation

seungrokj commented Jun 15, 2026 • edited by cursor Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

github-actions Bot commented Jun 15, 2026

Uh oh!

functionstackx left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

github-actions Bot commented Jun 15, 2026

Uh oh!

github-actions Bot commented Jun 15, 2026

Uh oh!

github-actions Bot commented Jun 15, 2026

Uh oh!

seungrokj commented Jun 15, 2026

Uh oh!

billishyahao left a comment

Choose a reason for hiding this comment

Uh oh!

seungrokj commented Jun 16, 2026

Uh oh!

seungrokj commented Jun 16, 2026

Uh oh!

Oseltamivir left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

cursor Bot Jun 16, 2026

Choose a reason for hiding this comment

TP4 dp-attn conc mismatch

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

seungrokj commented Jun 15, 2026 •

edited by cursor Bot

Loading