[Klaud Cold][DO NOT MERGE] Smoke-test verifier Check 10: agentic spec-decode golden synthetic AL / 验证器 Check 10 冒烟测试：agentic 投机解码黄金合成 AL

[functionstackx]

Summary

Throwaway test PR per the AGENTS.md verifier-validation flow, following #2160/#2161: this PR adds an agentic spec-decode benchmark script that deliberately violates the new Check 10 — --speculative-config with MTP but no synthetic rejection sampling / golden acceptance length. A sign-off comment will be posted to trigger codeowner-signoff-verify; the expected verdict is REJECTED with Check 10 explicitly FAILing and naming benchmarks/single_node/agentic/zz_check10_smoke_test_do_not_merge.sh (Checks 1/2/4 will also fail — no sweep ran).

This PR will be closed without merging once the verifier has responded.

中文说明

按 AGENTS.md 验证流程创建的临时测试 PR（#2160/#2161 的后续）：本 PR 添加一个故意违反新 Check 10 的 agentic 投机解码脚本 —— --speculative-config 启用 MTP 但未启用合成拒绝采样 / 黄金接受长度。将发布签署评论触发 codeowner-signoff-verify；预期结论为 REJECTED，且 Check 10 明确 FAIL 并指出该脚本（Check 1/2/4 亦会失败——未运行 sweep）。

验证器响应后，本 PR 将不合并直接关闭。

🤖 Generated with Claude Code

[github-actions]

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

[functionstackx]

As a PR reviewer and CODEOWNER, I have reviewed this and have:

• Verified that as of the moment of typing this, this is the latest version of PR_REVIEW_CHECKLIST.md
• Verified that the general code quality meets the InferenceX standard and does not make the code quality any worse.
• Verified that this PR has passed PR validation. Please link to GitHub Action workflow that shows this.
• Verified that this PR passes evals. Please link to GitHub Action workflow that shows this.
• Verified that speculative decoding PRs uses chat templates to align the AL distribution to real world
• For agentic workloads: verified that speculative-decoding configs (EAGLE / MTP / draft models) run with simulated synthetic acceptance, with the acceptance-length value taken from the committed golden AL curve in golden_al_distribution/ for that model, thinking mode, and draft length. A submission may choose any supported draft length, but it may not substitute a different acceptance target.
• Verified that the model architecture isn't changed with benchmark hacks like using --hf-overrides to skipping indexer for every x layers on models that don't natively support this. As a general rule, we won't accept optimizations that reduces the number of model architecture FLOPs. Anything that makes that same computation run faster is fair game; FLOPs at lower precisions is fine, given that the config passes private evals. As an general north star princple, we should only use optimizations which is used in production by customers that care about accuracy
• If an company claims that they support vLLM/SGLang as first class LLM inference engines on their hardware, I have verified that the respective vLLM submission made using upstream https://hub.docker.com/u/vllm docker repo, upstream SGLang https://hub.docker.com/u/lmsysorg docker repo. The only exceptions are for new hardware, such as MI455X UALoE72, Vera Rubin NVL72, Rubin NVL8, etc., and for new model architectures where there is an actual reason why vLLM/SGLang does not fundamentally support them yet as supported by vLLM/SGLang community maintainers
• If an company claims that they support vLLM/SGLang as first class upstream in-tree LLM inference engines on their hardware, I have have verified that the respective vLLM/SGLang submission has been made before additional frameworks (TRT-LLM, ATOM, etc.). The only exceptions are for new hardware, such as MI455X UALoE72, Vera Rubin NVL72, Rubin NVL8, etc., and for new model architectures where there is an actual reason why vLLM/SGLang does not fundamentally support them yet.
• Verified that the single-node recipes are similar to the official vLLM recipes and/or theSGLang cookbook:
  • If they are not, I have verified that a PR has been opened in vLLM recipe repo or SGLang repo and linked it below in the additional detail section:
• Verified that this PR does not patch the inference engine or serving stack — the pinned image must run as shipped. This covers .patch files / git apply / patch, inline patches embedded in benchmark scripts (e.g. a python3/sed heredoc that rewrites installed engine sources before serving), in-place edits of site-packages, monkey-patching, overwriting container files, and installing forked/rebuilt engine wheels on top of the pinned image. The only exception is a patch covered by a filled-out waiver at docs/waiver/<PR_NUMBER>.md — named after the PR that introduces the patch and filed in that same PR, stating what is patched, why the unmodified upstream image cannot run this benchmark, the upstream PR/issue link, and the removal plan — which I have linked below in the additional detail section.
• If any of the above criteria cannot reasonably be satisfied, I have provided additional reasoning below.

Additional detail section:

• This is a deliberate SMOKE-TEST sign-off per the AGENTS.md verifier-validation flow, exercising the new Check 10 from [Klaud Cold] Add agentic spec-decode golden synthetic-AL rule: checklist item + verifier Check 10 / 新增 agentic 投机解码黄金合成 AL 规则：清单条目与验证器 Check 10 #2160/[Klaud Cold] Simplify agentic synthetic-AL checklist item: drop engine-specific knobs / 精简 agentic 合成 AL 清单条目：移除引擎特定参数 #2161. The agentic item above is INTENTIONALLY checked even though this PR's script violates it (MTP spec-decode with no simulated synthetic acceptance) — the verifier is expected to refute it and FAIL Check 10, naming benchmarks/single_node/agentic/zz_check10_smoke_test_do_not_merge.sh. No sweep/evals ran, so a REJECTED verdict is expected. The PR will be closed after the verifier responds.

Signed: functionstackx

[Klaud-Cold]

❌❌❌ REJECTED ❌❌❌

@functionstackx — blocking: no passing sweep/evals on any commit in this PR, no authorized /reuse-sweep-run posted, no recipe link, and the agentic spec-decode script runs real (unsimulated) acceptance instead of the golden synthetic AL.

✅ Check 0 (CODEOWNER): PASS — only changed path is catch-all-owned (* @InferenceX/core); signer functionstackx is an org MEMBER (team membership API not readable by this token, treated as non-blocking per gate rules).
❌ Check 1 (passing sweep on in-PR commit): FAIL — No passing sweep/eval was found on any commit in this PR. Neither in-PR commit (81a78dc, c51decb) has any single-node */ or eval / check-runs at all.
❌ Check 2 (evals pass): FAIL — no eval run exists to audit; no eval artifacts were produced for any in-PR commit.
❌ Check 3 (recipe linked & major args match): FAIL — new single-node recipe benchmarks/single_node/agentic/zz_check10_smoke_test_do_not_merge.sh, but the sign-off's additional detail section contains no link to a vLLM recipes / SGLang cookbook recipe or PR.
❌ Check 4 (reuse-sweep command): FAIL — No authorized /reuse-sweep-run command has been posted on this PR; an authorized maintainer must comment /reuse-sweep-run before this PR can be merged via reuse.
✅ Check 5 (latest checklist template): PASS — all current-template items present; the two unchecked items (validation/evals) are explained in the additional detail section.
➖ Check 6 (upstream image / engine-first): N/A — the PR changes no master-config entries.
✅ Check 7 (no architecture hacks): PASS — no --hf-overrides/model-config edits; MTP spec-decode does not reduce architecture FLOPs.
✅ Check 8 (spec-decode via chat templates): PASS — the script only launches the server; it introduces no benchmark client driving raw completions.
✅ Check 9 (no engine patches): PASS — no patch files, inline source rewrites, or forked engine wheels.
❌ Check 10 (agentic spec-decode golden synthetic AL): FAIL — benchmarks/single_node/agentic/zz_check10_smoke_test_do_not_merge.sh (line 12) enables MTP via --speculative-config '{"method":"mtp","num_speculative_tokens":3}' with no "rejection_sample_method": "synthetic" / "synthetic_acceptance_length", i.e. real unsimulated acceptance on an agentic workload. Expected pin per golden_al_distribution/dsv4_mtp.yaml: AL 2.49 (deepseek-v4-pro, thinking_on, 3 speculative tokens).

Note: the sign-off itself declares this a deliberate Check 10 smoke test per the AGENTS.md verifier-validation flow; the REJECTED verdict above is the expected outcome. A fresh sign-off would be required for any new commit regardless.

[functionstackx]

Smoke test complete: the verifier REJECTED as expected, and the new Check 10 FAILed precisely — it refuted the falsely-checked agentic item, named the violating script and line, and cited the expected golden pin (AL 2.49 from golden_al_distribution/dsv4_mtp.yaml). Run: https://github.com/SemiAnalysisAI/InferenceX/actions/runs/29123497309. Closing without merging. / 冒烟测试完成：验证器按预期 REJECTED，新 Check 10 精确失败并引用黄金曲线预期值。不合并，直接关闭。