SA260324R-0094_r1

SA260324R-0094

① SA Source

• Source: 開啟完整 SA 文章
• Section: GPU and LPU Integration: Attention FFN Disaggregation (AFD)
• Line hint: 103

Context Before

Now with an understanding of what LPUs are good for we can understand how they fit into inference setups. NVIDIA introduced LPUs to improve the performance of high interactivity scenarios.

Evidence

In those scenarios, LPUs can leverage their low-latency capabilities to improve the decode phase latencies.

Context After

One way LPUs can improve decode phase latencies is by applying the Attention FFN Disaggregation (AFD) technique, introduced in MegaScale-Infer and Step-3.

② Atomic Claim

在這類情境下，LPU 可利用低延遲能力，改善 decode 階段的延遲。

• Epistemic Mode: ASSERTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "attribute": "LATENCY_PERFORMANCE",
  "context_nodes": [
    {
      "id": "04_knowledge_base/Decode",
      "label": "decode"
    }
  ],
  "entity": {
    "id": "04_knowledge_base/Groq LPU",
    "label": "LPU"
  },
  "frame_type": "ATTRIBUTE",
  "qualifiers": {
    "condition_text": "高互動情境",
    "numeric_mentions": [],
    "temporal_mentions": []
  },
  "value": {
    "numeric_mentions": [],
    "value_text": "可利用低延遲能力改善 decode 階段的延遲。"
  }
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
entity	LPU	04_knowledge_base/Groq LPU
context_node	decode	Decode

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。