Skip to content

perf(amd): port MI355X Qwen3.5 FP8 disagg sweep to srt-slurm - #2628

Draft
cquil11 wants to merge 4 commits into
agent/srt-slurm-amd-integrationfrom
agent/port-qwen35-fp8-mi355x-srt-slurm
Draft

perf(amd): port MI355X Qwen3.5 FP8 disagg sweep to srt-slurm#2628
cquil11 wants to merge 4 commits into
agent/srt-slurm-amd-integrationfrom
agent/port-qwen35-fp8-mi355x-srt-slurm

Conversation

@cquil11

@cquil11 cquil11 commented Aug 17, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • replace the active MI355X Qwen3.5 FP8 legacy AMD disaggregated launcher with the existing native srt-slurm SGLang Router + MoRI recipe
  • preserve the production 1P1D TP8 topology and complete c8/c16/c32/c64/c128 8k1k search space
  • run every point inside one eight-hour Slurm allocation through the repository's unchanged benchmark_serving.py
  • remove the model-specific legacy wrapper that only translated matrix variables into amd_utils/submit.sh
  • consolidate the temporary -srt-disagg validation key back into the original production config key

Validation

  • srt-slurm dry-run passes against feat(runtime): add AMD accelerator support srt-slurm#1 at c609754b5622f96d5c12a93149e245308d4f1e9b
  • generated matrix contains exactly one topology job; the recipe owns all five concurrency points
  • 231 matrix-generation and validation tests pass
  • changed YAML parses and git diff --check passes

Stack

Based on #2542. This remains draft until the exact-head MI355X full recipe sweep completes.

@github-actions

Copy link
Copy Markdown
Contributor

Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase As a PR reviewer and CODEOWNER, I have reviewed this and have.

For PR verification, add the full-sweep-fail-fast label (strongly recommended) to this PR — the benchmark sweep only runs on labeled PRs. Use full-sweep-enabled only if you need matrix jobs to keep running past a failure.

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. See GitHub's docs on re-running failed jobs


感谢你的贡献!请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md,然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人,你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板,包括保留英文语句 As a PR reviewer and CODEOWNER, I have reviewed this and have

如需进行 PR 验证,请为此 PR 添加 full-sweep-fail-fast 标签(强烈推荐)— 基准测试 sweep 仅在带有标签的 PR 上运行。仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动(flake),重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档

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

Labels

None yet

Projects

Status: No status

Development

Successfully merging this pull request may close these issues.

1 participant