IX2-0076
① SA Source
- Source: 開啟完整 SA 文章
- Section:
Prefill and Decode Phases - Line hint:
105
Context Before
Prefill and Decode Phases
Inference contains two main phases: prefill and decode. Prefill occurs during the first forward pass of a request’s lifetime. It is computationally intensive since all tokens in the request are processed in parallel. This phase is responsible for “filling up” the KV cache for a sequence. After prefill, responses are generated (or decoded) one token at a time. Each forward pass loads the entire KV cache for a sequence from HBM, while only performing the computation for a single token, making decode memory (bandwidth) intensive.
Evidence
Context After
Disaggregated Prefill
Disaggregated prefill (aka PD disaggregation or simply “disagg”) is the practice of separating the prefill and decode phases across separate pools of GPUs or clusters. These separate prefill and decode pools can be tuned independently and scaled to match the needs of workloads.
② Atomic Claim
當 prefill 與 decode 在同一個 engine 上執行時，prefill 會持續打斷 decode batches，造成整體效能變差。
- Epistemic Mode:
ASSERTED - Mapping Status:
COMPLETE
③ Semantic Frame
{
"attribute": "COUNT",
"context_nodes": [
{
"id": "04_knowledge_base/Decode",
"label": "decode"
}
],
"entity": {
"id": "04_knowledge_base/Prefill",
"label": "prefill"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [],
"value_text": "當 prefill 與 decode 在同一個 engine 上執行時，prefill 會持續打斷 decode batches，造成整體效能變差。"
}
}④ Canonical Entity Mapping
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。