[AMD] dsv4 atom-disagg eval sweep — validate reduced ATOM logging

[Oseltamivir]

Description

Eval-only / all-evals sweep to validate the reduced ATOM multinode logging on hardware. Based on current main (includes merged #1879: 8k1k-only --all-evals, per-conc score labels, --expected-concs), seeded from #1818's dsv4 disagg scaffolding, with the perf-changelog.yaml entry pointed at dsv4-fp4-mi355x-atom-disagg so the sweep exercises the ATOM engine path.

Reduced ATOM logging (the thing under test)

The atom engine (atom.entrypoints.openai_server) is vLLM-based and reads VLLM_LOGGING_LEVEL, but env_atom.sh only set VLLM_LOG_LEVEL (a no-op for vLLM) — so the engine logged its per-request KV-connector flood ([CONSUMER]/[PRODUCER]/[PD-*]/Engine Core, ~6.5k lines in run 27924353733) at INFO.

• env_atom.sh: set VLLM_LOGGING_LEVEL=WARNING (the real lever) + ATOM_UVICORN_ACCESS_LOG=0, ATOM_UVICORN_LOG_LEVEL=warning, ATOMESH_LOG_LEVEL=warn. All overridable.
• server_atom.sh: atomesh router --log-level info → ${ATOMESH_LOG_LEVEL}.

Labels

all-evals, full-sweep-enabled, evals-only (all-evals also disables artifact reuse, so the run executes fresh).

Validation target

Compare the atom-disagg eval job logs from this sweep against run 27924353733: the [CONSUMER]/[PRODUCER] INFO flood and uvicorn access lines should be gone.

Note: test/validation PR; not intended to supersede #1818.

[github-actions]

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

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

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

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

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.

---

感谢你的贡献！对于 vLLM 与 SGLang，请确保你的 recipe 与官方 vLLM recipes 和/或 SGLang cookbook 保持一致

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

如果不一致，请先创建一个 PR，之后我们才能将你的单节点 PR 合并到 master 分支。让我们确保文档保持一流水准，使整个 ML 社区都能从你的辛勤工作中受益！谢谢

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

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。如果选择重新运行失败的任务，PR 作者有责任确保其最终通过。参见 GitHub 关于重新运行失败任务的文档：https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

一般而言，PR 作者应先向相应公司的 CODEOWNERS 请求审阅并获得 PR 批准，然后再请求核心维护者审阅。

如需更多帮助，PR 作者可通过 Slack 联系核心维护者。

[github-actions]

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

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

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

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

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.

---

感谢你的贡献！对于 vLLM 与 SGLang，请确保你的 recipe 与官方 vLLM recipes 和/或 SGLang cookbook 保持一致

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

如果不一致，请先创建一个 PR，之后我们才能将你的单节点 PR 合并到 master 分支。让我们确保文档保持一流水准，使整个 ML 社区都能从你的辛勤工作中受益！谢谢

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

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。如果选择重新运行失败的任务，PR 作者有责任确保其最终通过。参见 GitHub 关于重新运行失败任务的文档：https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

一般而言，PR 作者应先向相应公司的 CODEOWNERS 请求审阅并获得 PR 批准，然后再请求核心维护者审阅。

如需更多帮助，PR 作者可通过 Slack 联系核心维护者。

[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 9e44be0. Configure here.}

[cursor]

MoE input token scaling removed

Medium Severity

The DP+EP and MTP branches no longer recompute MORI_MOE_MAX_INPUT_TOKENS_PREFILL / MORI_MOE_MAX_INPUT_TOKENS_DECODE, so every non–DeepSeek-V4-Pro disagg job keeps the static defaults from env.sh even when dispatch tokens and bench concurrency are scaled. That desyncs MoE limits from the tuned DP+EP path.

Additional Locations (2)

• benchmarks/multi_node/amd_utils/server_sglang.sh#L220-L221
• benchmarks/multi_node/amd_utils/server_sglang.sh#L256-L257

 

^{Reviewed by Cursor Bugbot for commit 9e44be0. Configure here.}

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27941462066
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27941462066

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27942496140
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27942496140