IX2-0482
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Optimizing Inference with Wide EP + Disaggregated Serving - Line hint:
690
Context Before
Source: SemiAnalysis
Evidence
In the highest throughput, lowest interactivity region of the frontier, batch sizes are large (128+) and configurations shift to full DEP: attention weights are fully replicated across all GPUs as independent data-parallel ranks, experts are distributed via EP
Context After
Source: SemiAnalysis InferenceX ↗
② Atomic Claim
在 frontier 中 throughput 最高、interactivity 最低的區域，batch size 很大（128+），configuration 會轉成 full DEP：attention weights 在所有 GPUs 上完整複製成獨立 DP ranks，而 experts 則透過 EP 分散。
- Epistemic Mode:
ASSERTED - Mapping Status:
COMPLETE
③ Semantic Frame
{
"attribute": "THROUGHPUT",
"context_nodes": [],
"entity": {
"id": "04_knowledge_base/GPU",
"label": "GPUs"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [
"128"
],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [
"128"
],
"value_text": "在 frontier 中 throughput 最高、interactivity 最低的區域，batch size 很大（128+），configuration 會轉成 full DEP：attention weights 在所有 GPUs 上完整複製成獨立 DP ranks，而 experts 則透過 EP 分散。"
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| entity | GPUs | GPU |
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。