clm_670bbe900445425380fdfc31382bf230_r1

NIEK2-0112

① SA Source

• Source: 開啟完整 SA 文章
• Section: GPU and LPU Integration: Attention FFN Disaggregation (AFD)
• Line hint: 101

Context Before

Source: MegaScale-Infer, SemiAnalysis

Evidence

If a GPU only performs attention operations, its HBM capacity can be fully allocated to KV cache, increasing the total number of tokens it can process, which then increases the tokens each expert processes on average

Context After

Source: SemiAnalysis

② Atomic Claim

若一顆 GPU 只執行 attention，其 HBM 容量可全部配置給 KV cache，提高可處理的總 token 數，進而提高每個 expert 平均處理的 token 數。

• Epistemic Mode: HYPOTHETICAL
• Mapping Status: PARTIAL

③ Semantic Frame

{
  "frame_type": "NARY_RELATION",
  "participants": [
    {
      "node": {
        "id": "04_knowledge_base/GPU",
        "label": "GPU"
      },
      "role": "subject"
    },
    {
      "node": {
        "id": "04_knowledge_base/HBM",
        "label": "HBM"
      },
      "role": "participant"
    },
    {
      "node": {
        "id": "04_knowledge_base/KV cache",
        "label": "KV cache"
      },
      "role": "participant"
    }
  ],
  "qualifiers": {
    "condition_text": null,
    "numeric_mentions": [],
    "temporal_mentions": []
  },
  "relation_type": "RUNS"
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
subject	GPU	GPU
participant	HBM	HBM
participant	KV cache	04_knowledge_base/KV cache

⑤ Human Review

請在 Properties 逐項確認：

• 原文 → Atomic Claim 是否忠實
• Atomic Claim → Semantic Frame 是否忠實
• Canonical Entity mapping 是否正確
• Epistemic mode 是否保留原文語氣
• 最後選擇 review_action

Review state

Markdown 內文不是正式 approval。只有 Apply bridge 寫入的 Decision Ledger event 才是正式決策。