IX2-0111
① SA Source
- Source: 開啟完整 SA 文章
- Section:
DeepSeek R1 - Line hint:
163
Context Before
Source: SemiAnalysis InferenceX ↗
Evidence
For MI355X Disaggregated inference serving, AMD recommends using SGLang with MoRI. MoRI is AMD’s MoE dispatch/combine collective and KV Cache transfer library ↗ built from first principles by AMD’s cracked 10x China-based engineering team. Although MoRI needs much more open CI and testing, we are strong supporters of the direction that MoRI is taking. This is because instead of taking AMD’s historical approach, which was to fork NVIDIA’s NCCL into RCCL, MoRI is built from scratch by taking the lessons from RCCL/NCCL and building an entirely new package from first principles. The use of MoRI has also delivered good speedups in the span of more than a month, with throughput per GPU increasing by more than 20% in the 20-45 tok/s/user interactivity range.
Context After
Source: SemiAnalysis InferenceX ↗
② Atomic Claim
MoRI 是 AMD 從零打造的 MoE dispatch/combine collective 與 KV Cache transfer library，主要由 AMD 位於中國的核心工程團隊開發。
- Epistemic Mode:
ASSERTED - Mapping Status:
PARTIAL
③ Semantic Frame
{
"attribute": "UNSPECIFIED_ATTRIBUTE",
"context_nodes": [
{
"id": "02_companies/AMD",
"label": "AMD"
},
{
"id": "04_knowledge_base/Mixture of Experts",
"label": "MoE"
},
{
"id": "04_knowledge_base/KV cache",
"label": "KV Cache"
}
],
"entity": {
"id": "04_knowledge_base/AMD MoRI backend",
"label": "MoRI"
},
"frame_type": "ATTRIBUTE",
"qualifiers": {
"condition_text": null,
"numeric_mentions": [],
"temporal_mentions": []
},
"value": {
"numeric_mentions": [],
"value_text": "MoRI 是 AMD 從零打造的 MoE dispatch/combine collective 與 KV Cache transfer library，主要由 AMD 位於中國的核心工程團隊開發。"
}
}④ Canonical Entity Mapping
| Role | Surface Label | Canonical Target |
|---|---|---|
| entity | MoRI | 04_knowledge_base/AMD MoRI backend |
| context_0 | AMD | AMD |
| context_1 | MoE | 04_knowledge_base/Mixture of Experts |
| context_2 | KV Cache | 04_knowledge_base/KV cache |
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。