[PowerX B300] enable strict measured power for Kimi-K3 AgentX / [PowerX B300] 为 Kimi-K3 AgentX 启用严格实测功耗

[edwingao28]

Description

Enable strict measured power for kimik3-fp4-b300-vllm-agentic-dspark (Tracks #3592: Kimi-K3 × B300 AgentX 0/11). Closed draft #3619 was changelog-only and did not enable require-power; this PR enables it and wires AgentX forwarding.

Config

• kimik3-fp4-b300-vllm-agentic-dspark → agentic-coding arm: require-power: true next to dram-utilization: 0.75
• Topology, concurrency [1,2,4,8,16,24,32,40,48,56,70], image, and recipe unchanged
• No recipe REQUIRE_POWER churn: single-node AgentX already collects via ENABLE_AGENTX_POWER; CI forwards strictness from the matrix flag

Plumbing (required on current main for AgentX require-power to take effect)

• Accept require-power on AgenticCodingConfig / agentic matrix entries
• Forward the flag from scenario arms into generated AgentX matrix rows
• Pass require-power into sweep-agentic / sweep-multi-node-agentic in run-sweep.yml
• Same wiring shape as open [PowerX] MiniMax-M3 B200 AgentX measured-power coverage / 补齐 MiniMax-M3 B200 AgentX 功耗覆盖 #3628 (MiniMax B200); expect a clean conflict resolution if that lands first

Planner / validation

• Base: origin/main @ 1c5e6b9f5e7b53042c6ac831621ac72568fd8405
• Head: 043738b6123602bbeea160d2c4ccb243d06e6b24
• python -m infx.matrix.generate test-config ... --config-keys kimik3-fp4-b300-vllm-agentic-dspark --scenario-type agentic-coding --no-evals → 11/11 rows, all require-power: true, concs [1,2,4,8,16,24,32,40,48,56,70]
• python -m infx.matrix.plan --base-ref origin/main --head-ref HEAD → single_node.agentic: 11, agentic_evals: 11 (evals strip require-power)
• python -m infx.workflows.validate_perf_changelog ... → passed

No sweep/dispatch labels applied on open.

中文

为 kimik3-fp4-b300-vllm-agentic-dspark 启用严格实测功耗（关联 #3592：Kimi-K3 × B300 AgentX 0/11）。已关闭的草稿 #3619 仅改 changelog，未启用 require-power。本 PR 在 agentic-coding 场景臂设置 require-power: true，并接通 AgentX 矩阵/schema/run-sweep 转发；拓扑与 11 点并发不变，未改 recipe。接线形态与开放中的 #3628 相同，若其先合入需处理冲突。本地规划结果为 11 吞吐点 + 11 评估；changelog 校验通过。打开 PR 时未添加任何 sweep/dispatch 标签。

AI model disclosure

• Model/version: gpt-5.2-codex-high
• Role: Implementation, local planner/changelog validation, and PR preparation

Related Issue

Tracks #3592.

Type of Change

• Bug fix
• New feature
• Configuration change
• Documentation update
• Other (please describe): AgentX require-power matrix/workflow wiring needed for the flag to take effect on current main

Checklist

• I have completed the AI model disclosure and kept it current
• I have tested my changes locally
• I have updated documentation if necessary
• For every change that can affect benchmark performance and every recipe addition or modification, I have appended a new entry to the physical end of inferencex-e2e/perf-changelog.yaml and have not edited historical entries
• Before merging via reuse, an authorized maintainer (OWNER/MEMBER/COLLABORATOR) has commented /use <run_id> (or the legacy /reuse-sweep-run) on this PR. Do this only once there is a final full sweep that is all green with evals passing, since after this comment the sweep label will no longer automatically kick off new sweeps. Remove and re-add the label to force one.

中文

AI：gpt-5.2-codex-high；实现、本地规划/changelog 校验与 PR 准备。关联 #3592；配置变更 + AgentX require-power 接线；GPU 验收与合并前复用仍待完成。打开时未添加任何 sweep 标签。

  

[github-actions]

Thanks for the contribution!

• Review: If this PR changes files owned by someone other than a repository admin or @SemiAnalysisAI/core, ask one eligible CODEOWNER to complete the latest PR_REVIEW_CHECKLIST.md before contacting a core maintainer on Slack. Follow the template exactly, including As a PR reviewer and CODEOWNER, I have reviewed this and have, so sign-off verification triggers.
• PR verification: Sweeps only run on labeled PRs. Add full-sweep-fail-fast (strongly recommended); use full-sweep-enabled only when matrix jobs should continue after a failure.
• After merging: PR authors must ensure all GitHub Actions jobs pass. Transient failures often pass on rerun; see how to rerun failed jobs.

中文

感谢你的贡献！

• **审阅：**如果 PR 修改的文件归属于仓库管理员及 @SemiAnalysisAI/core 之外的 CODEOWNER，请先联系一位有资格的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，再通过 Slack 联系核心维护者。必须严格遵循模板，并保留 As a PR reviewer and CODEOWNER, I have reviewed this and have，才能触发签核验证。
• **PR 验证：**扫描仅在带有标签的 PR 上运行。强烈建议添加 full-sweep-fail-fast；仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。
• **合并后：**PR 作者必须确保所有 GitHub Actions 任务通过。临时性失败通常可以通过重新运行恢复；参见重新运行失败任务的说明。