IX2-0468
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Optimizing Inference with Wide EP + Disaggregated Serving - Line hint:
680
Context Before
Optimizing Inference with Wide EP + Disaggregated Serving
Evidence
disaggregated prefill are appropriate at different interactivity levels
Context After
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.
At the highest interactivity levels (batch 1-16), pure TP outperforms any configuration involving EP. At low batch sizes, only a small fraction of experts activate per step. With EP, these activations are distributed unevenly across GPUs: at batch 4, only 32 of 256 experts fire, and any given GPU has roughly a low double digit percent chance of receiving zero routed tokens in a given layer. TP avoids this by sharding every expert across all GPUs, so all 8 GPUs participate equally in every expert computation regardless of which experts the router selects. 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.
② Atomic Claim
並進一步判斷何時適合使用 disaggregated prefill。
- Epistemic Mode:
ASSERTED - Mapping Status:
PARTIAL
③ Semantic Frame
{
"attribute": "UNSPECIFIED_ATTRIBUTE",
"context_nodes": [],
"entity": {
"id": "04_knowledge_base/Disaggregated prefill",
"label": "disaggregated prefill"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [],
"value_text": "並進一步判斷何時適合使用 disaggregated prefill。"
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| entity | disaggregated prefill | 04_knowledge_base/Disaggregated prefill |
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。