IX2-0327
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Multi Token Prediction (MTP) - Line hint:
538
Context Before
Source: SemiAnalysis InferenceX ↗
Evidence
At large batch sizes, the inference regime is less memory-bandwidth bound compared to for low batch sizes
Context After
In terms of cost, MTP can drive huge cost savings, in the below table, we see that DeepSeek-R1-0528 run on FP4 using Dynamo TRT costs 0.057 per million total tokens.
② Atomic Claim
在大 batch size 下，推論較不受記憶體頻寬限制。
- Epistemic Mode:
ASSERTED - Mapping Status:
COMPLETE
③ Semantic Frame
{
"attribute": "BANDWIDTH",
"context_nodes": [],
"entity": {
"id": "04_knowledge_base/Batch size",
"label": "batch size"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [],
"value_text": "在大 batch size 下，推論較不受記憶體頻寬限制。"
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| entity | batch size | 04_knowledge_base/Batch size |
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。