Skip to content
View rui08984-dot's full-sized avatar

Block or report rui08984-dot

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
rui08984-dot/README.md

你好,我是 rui08984-dot 👋

我在 8GB 显卡上折腾大模型推理优化,也给 AI 编程代理搭基础设施。 所有东西都实测过、有数据、能复现 —— 不写没验证过的结论。


🔧 8GB 显存跑大模型(实测调优,含完整深度曲线)

把 27B–35B 级别的模型塞进 8GB 笔记本显卡,附生产配置与实测数据:

项目 说明
bonsai2-27b-8g-tuning 三值量化 Bonsai-2-27B · PrismML 版本 A/B 对比(+15% PP / +24% TG)
ornith-9b-kvmem-8g-tuning KV 缓存放内存 → 128K 上下文 @ 42.7 tok/s
ornith-1.5-35b-8g-tuning 35B MoE 到 118K 上下文的深度曲线 + turbo4 KV
qwen3.6-35b-8g-tuning Qwen3.6-35B-A3B 基线生产配置
kat-coder-35b-8g-tuning 编码模型 · 工具循环修复 · agentic 基准
zhrp-gemma4-26b-8g-tuning 中文角色扮演调优 · 三级 ubatch 配方

🤖 AI 代理基础设施

项目 说明
SessionRelay 本地优先的 AI 会话记忆层 · 五源捕获 · 中文+语义检索 · MCP 16 工具
pi-bridge 把本地/云端模型变成 MCP 工具 · 命令行验收兜底 · 证据链 · 失败自动升档
agent-state-stack 跨代理共享状态栈 · 开场简报注入 · 看板与依赖图

🎲 其他

项目 说明
jujian 局鉴 · 社交推理游戏复盘台 —— 抓出互相矛盾的声称,每条结论可回查
agent-skills Agent Skills 技能包 · AI 音乐制作 SOP + 中文网文写作管线

💡 如果哪个项目帮到你了,点个 ⭐ 就是最好的支持 —— 也能让更多人找到它。

Popular repositories Loading

  1. agent-skills agent-skills Public

    中文 Agent Skills 技能包:工程纪律四件套(会话交接/进度锚/项目地图/不可逆变更) + AI 音乐生产 + 中文网文流水线 | Chinese agent skills: long-project discipline + creative pipelines

    1

  2. qwen3.6-35b-8g-tuning qwen3.6-35b-8g-tuning Public

    Qwen3.6-35B-A3B on 8GB - baseline production config and measured depth curves (turbo4 KV)

    Python 1

  3. ornith-1.5-35b-8g-tuning ornith-1.5-35b-8g-tuning Public

    Ornith-1.5-35B-A3B on RTX 4060 8GB - measured depth curves (PP/TG to 118K) + turbo4 KV production config + tool-loop fix

    Python

  4. kat-coder-35b-8g-tuning kat-coder-35b-8g-tuning Public

    KAT-Coder-V2.5-Dev 35B on 8GB - depth curves to 124K (PP 850), thinking governance flags, multi-round tool-loop fix, agentic bench

    Python

  5. bonsai2-27b-8g-tuning bonsai2-27b-8g-tuning Public

    Ternary-Bonsai-2-27B (PTQ1_0 QAT) on 8GB - PrismML b10709 A/B (+15% PP/+24% TG), quality gate, production config

    Python

  6. zhrp-gemma4-26b-8g-tuning zhrp-gemma4-26b-8g-tuning Public

    Chinese-calibrated Gemma-4-26B RP on 8GB - three-tier ubatch recipe (64K/118K/128K), NO_PINNED, measured PP/TG

    Python