Repository navigation
docs(model-architecture): 线性注意力机制精读与通俗拆解入库（250，I-242）——新开分组、KDA 原型与双 archify 图 - #1222
Merged
Merged
Conversation
…archify 图; - 新增 250 精读：腾讯技术工程《AI Infra进阶：3万字深入理解线性注意力机制》全文 + 10 个一手信源逐条对账（Kimi Linear/GDN/DeltaNet 并行化/vLLM v0.32.0 文档/FWP 通读，Devil 论文 ID 勘误为 2210.10340）；KV Cache 三重代价 → 门控演进链 → chunk-wise 两级并行 → hybrid 3:1 兜底与 vLLM 落地三件事；五规律三争议，字数门 6000 内 - 新开 docs/research/model-architecture/ 分组；纯标准库原型 linear_attention_lab.py（17 断言：KDA/DeltaNet/GDN 顺序 vs 分块偏差 4.4e-16；D1–D5 破坏实验） - 两张 archify 图四件套（kda--evolution-ladder、kda--chunkwise-two-level，finalize 四门回执齐） - 登记面：research readme 第十一节、knowledge-map、mermaid README、CHANGELOG、issue I-242 🤖 Generated with [Claude Code](https://github.com/claude), [CodeX](https://openai.com), [Gemini](https://github.com/apps/gemini-code-assist) Co-Authored-By: Aurelius Huang<threefish.ai@gmail.com>
ThreeFish-AI
added a commit
that referenced
this pull request
Oct 30, 2026
冲突处置：CHANGELOG 与 knowledge-map 取双侧并集；issue.md 双方撞号 I-242——对方线性注意力条目（#1222）已合主干保留 I-242，我方 Trustworthy RSI 条目重编为 I-243 并留映射注记。 🤖 Generated with [Claude Code](https://github.com/claude), [CodeX](https://openai.com), [Gemini](https://github.com/apps/gemini-code-assist) Co-Authored-By: Aurelius Huang<threefish.ai@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
概要
新开
docs/research/model-architecture/分组，入库首篇精读 250 线性注意力机制精读与通俗拆解：以腾讯技术工程《AI Infra进阶：3万字深入理解线性注意力机制》（2026-10-29）为主信源，补读 10 个一手信源逐条对账，配纯标准库原型与两张 archify 图（I-242）。内容主线
S_t=S_{t-1}T_t+H_t→ 块间 scan + 块内 WY 下三角系统一次求解，精确重写无近似信源与核验
交付物
docs/research/model-architecture/250-linear-attention-kda.md（+_category_.json，position 11）docs/research/model-architecture/assets/linear_attention_lab.py：17 断言全绿（KDA/DeltaNet/GDN 顺序 vs 分块最大偏差 4.4e-16）+ D1–D5 破坏实验kda--evolution-ladder（workflow）、kda--chunkwise-two-level（dataflow），.mmd / candidate JSON / HTML / 双主题 PNG / finalize 四门回执齐验收
python3 docs/research/model-architecture/assets/linear_attention_lab.py selftest→ 17/17… count docs/research/model-architecture/250-linear-attention-kda.md→ 6000（门禁 ≤6000，退出码 0）🤖 Generated with Claude Code