Skip to content

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

rl-prefill-consistency-dump

Verified

训推一致性(prefill)采集侧:为训练/推理两侧采集可比的 prefill dump(输入对齐 PROMPTS_ONLY → 参数对齐校验(dump 前前置:校验两侧 config 超参/结构/权重,不一致输出参数不一致点并对齐)→ msprobe 接入采集 → 采集验证 → 恢复代码)。按用户需求判断用法:① 仅采集 dump——用户只要 dump 数据,走采集主流程(0→7→9),直接产出两侧可比 dump;② 配合分析——用户要求找所有独立差异点/差异点隔离验证/配合 rl-consistency-dump-analysis 分析时,才按分析侧打桩指令用 msprobe `load` 覆盖边界二次采集(阶段 8)再恢复代码。差异分析、首差异定位、打桩指令、整体结论由 rl-consistency-dump-analysis 承担。触发词:训推一致、训推一致性、prefill 比对、msprobe dump/采集、精度比对数据采集、打桩二次采集、参数对齐校验、参数不一致点、mHC 配置不一致、权重不同步、清卡验环境后采数据;已采 dump 的分析转 rl-consistency-dump-analysis。

scripts/param_alignment_mapping.example.json

View on GitHub
{
  "model": "DeepSeek-V4-Flash-Vision (起点示例; 新模型按现场重建映射)",
  "items": [
    {
      "semantic": "Hyper-Connections enable 开关",
      "train": "enable_mhc_connections",
      "infer": null,
      "infer_expected": true,
      "note": "推理侧无独立开关(结构上恒启用), 训练侧必须为 True"
    },
    {
      "semantic": "Hyper-Connections 流数 (hc_mult)",
      "train": "mhc_num_residual_streams",
      "infer": "hc_mult",
      "note": ""
    },
    {
      "semantic": "Hyper-Connections Sinkhorn 迭代",
      "train": "mhc_sinkhorn_iterations",
      "infer": "hc_sinkhorn_iters",
      "note": ""
    },
    {
      "semantic": "Hyper-Connections eps",
      "train": null,
      "infer": "hc_eps",
      "note": "训练侧无对应字段时标 MISSING_SIDE, 查训练代码是否硬编码同值"
    },
    {
      "semantic": "CSA 窗口",
      "train": "csa_window_size",
      "infer": "sliding_window",
      "note": ""
    },
    {
      "semantic": "CSA 压缩率",
      "train": "csa_compress_ratios",
      "infer": "compress_ratios",
      "note": ""
    },
    {
      "semantic": "DSA indexer heads",
      "train": "dsa_indexer_n_heads",
      "infer": "index_n_heads",
      "note": ""
    },
    {
      "semantic": "DSA indexer head_dim",
      "train": "dsa_indexer_head_dim",
      "infer": "index_head_dim",
      "note": ""
    },
    {
      "semantic": "DSA indexer topk",
      "train": "dsa_indexer_topk",
      "infer": "index_topk",
      "note": ""
    },
    {
      "semantic": "MTP 层数",
      "train": "mtp_num_layers",
      "infer": "num_nextn_predict_layers",
      "note": ""
    },
    {
      "semantic": "Hash 层数",
      "train": "moe_n_hash_layers",
      "infer": "num_hash_layers",
      "note": ""
    },
    {
      "semantic": "隐藏维度",
      "train": "hidden_size",
      "infer": "hidden_size",
      "note": ""
    },
    {
      "semantic": "层数",
      "train": "num_layers",
      "infer": "num_hidden_layers",
      "note": ""
    },
    {
      "semantic": "注意力头",
      "train": "num_attention_heads",
      "infer": "num_attention_heads",
      "note": ""
    },
    {
      "semantic": "词表大小",
      "train": "actual_vocab_size",
      "infer": "vocab_size",
      "note": ""
    }
  ]
}
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