Skip to content

kali20gakki/msAgent / skills/accuracy/rl-consistency-dump-analysis

rl-consistency-dump-analysis

Verified

通用训推一致性(prefill)的 dump 数据分析 skill。分析**已采集**的训推两侧 dump:探查 dump 结构 → 语义边界映射(不要求算子同名)→ 逐边界精确门禁比对 → 定位首差异 → ULP 量化区分舍入/逻辑级 → 追放大链 → 差异点隔离验证(打桩迭代:输出打桩指令给采集侧,等它用 msprobe `load` 覆盖边界再采后比下游,枚举所有独立差异点)→ 排除根因假设 → 输出整体结论 → 产出分析报告。与具体框架/硬件解耦,适用 msprobe PrecisionDebugger 或等价工具采的任何 dump。触发词:训推一致性分析、dump 数据分析、dump 已采好帮我分析、首差异定位、所有差异点、差异点隔离、打桩指令、对齐报告、训推不一致结论、比较两侧 tensor。采集与打桩再采由 rl-prefill-consistency-dump 负责,本 skill 只消费已落盘 dump、输出打桩指令驱动再采。

references/task_card.json

View on GitHub
{
  "model": null,
  "scope": "prefill 训推一致性 dump 分析",
  "train_dump_dir": null,
  "infer_dump_dir": null,
  "step": null,
  "rank": null,
  "request_id": null,
  "prompt_token_length": null,
  "train_input_token_length": null,
  "infer_prefill_token_length": null,
  "padding_handling": null,
  "train_framework": null,
  "inference_engine": null,
  "inference_engine_version": null,
  "hardware": null,
  "topology": { "tp": null, "pp": null, "ep": null },
  "evidence_tool": "msprobe PrecisionDebugger(或等价)",
  "target": "首差异位置 + 根因 + 证据链 + 是否可解(默认;可改为仅首差异)",
  "stop_condition": "首差异有确定性根因并完成 ULP 定性;产出分析结论记录与分析报告",

  "stage1_probe": {
    "dump_metadata": null,
    "construct_op_counts": null,
    "tensor_file_counts": null,
    "common_ops": [],
    "prefix_patterns": null,
    "fused_candidates": []
  },
  "stage2_mapping": {
    "mapping_file": null,
    "boundary_count": null,
    "evidence_missing_boundaries": []
  },
  "stage3_compare": {
    "results_file": null,
    "total": null,
    "exact": null,
    "not_exact": null,
    "evidence_missing": null
  },
  "stage4_first_divergence": {
    "name": null,
    "semantic_boundary": null,
    "max_abs_diff": null,
    "nz_count": null
  },
  "stage5_ulp": {
    "ulp_multiplier_dist": null,
    "classification": null
  },
  "stage6_excluded_hypotheses": [],
  "stage7_conclusion": null,
  "stage8_reports": {
    "report_source": null,
    "output_dir": null
  },

  "notes": "只记事实,不知道的字段留 null / [],不编默认值。"
}
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall