NIEK2-0128
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Speculative Decoding - Line hint:
137
Context Before
Speculative Decoding
A different way LPUs could improve decode phase latencies is by accelerating a speculative decoding setup, where we deploy draft models or Multi-Token Prediction (MTP) layers onto LPUs.
Evidence
Speculative decoding usually boosts output token per decode step by 1.5 to 2 tokens, depending on the draft model / MTP accuracy
Context After
Source: SemiAnalysis
② Atomic Claim
Speculative decoding 通常可將每個 decode step 的輸出 token 數提高至 1.5～2 個，實際效果取決於 draft model／MTP 的準確率。
- Epistemic Mode:
ASSERTED - Mapping Status:
COMPLETE
③ Semantic Frame
{
"attribute": "COUNT",
"context_nodes": [
{
"id": "04_knowledge_base/Decode",
"label": "decode"
},
{
"id": "04_knowledge_base/Multi-Token Prediction",
"label": "MTP"
}
],
"entity": {
"id": "04_knowledge_base/Speculative Decoding",
"label": "Speculative decoding"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [
"1.5",
"2"
],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [
"1.5",
"2"
],
"value_text": "Speculative decoding 通常可將每個 decode step 的輸出 token 數提高至 1.5～2 個，實際效果取決於 draft model／MTP 的準確率。"
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| entity | Speculative decoding | 04_knowledge_base/Speculative Decoding |
| context_0 | decode | Decode |
| context_1 | MTP | 04_knowledge_base/Multi-Token Prediction |
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。