IX2-0477
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Optimizing Inference with Wide EP + Disaggregated Serving - Line hint:
684
Context Before
Wide EP and disaggregated prefill are separate techniques that are often used together to achieve Pareto optimal performance. In this section, we walk through real results from InferenceX to build intuition for which combinations of parallelism strategy, wide EP, and disaggregated prefill are appropriate at different interactivity levels.
It helps to first understand what parallelism strategies fall on what parts of the Pareto frontier for single-node configurations. Take the example of DeepSeek R1 FP4 8k/1k on a single 8-GPU B200 node with TRT-LLM. The optimal strategy shifts as you move along the frontier, driven primarily by batch size and its effect on expert activation density.
Evidence
We collected expert activation ratio versus batch size data while profiling DeepSeek R1, which confirms that at batch sizes 16 and below, expert activation per layer is very low
Context After
Source: SemiAnalysis
② Atomic Claim
InferenceX 在 profiling DeepSeek R1 時收集 expert activation ratio vs batch size，結果確認 batch 16 以下每層 expert activation 都很低。
- Epistemic Mode:
ASSERTED - Mapping Status:
PARTIAL
③ Semantic Frame
{
"comparison_expression": "InferenceX 在 profiling DeepSeek R1 時收集 expert activation ratio vs batch size，結果確認 batch 16 以下每層 expert activation 都很低。",
"entities": [
{
"id": "02_companies/DeepSeek",
"label": "DeepSeek"
},
{
"id": "04_knowledge_base/Batch size",
"label": "batch size"
}
],
"frame_type": "COMPARISON",
"metric": "UNSPECIFIED_METRIC",
"operator": "COMPARES_WITH",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [
"16"
],
"temporal_mentions": []
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| comparison_entity_0 | DeepSeek | DeepSeek |
| comparison_entity_1 | 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。