satan9394/dsh-llm-finetuning

Name: satan9394/dsh-llm-finetuning
Author: satan9394

satan9394★ 0JavaScript最后同步: 2026-08-19

DSH skill: LLM 微调路由与配方, SFT/DPO/GRPO 选型（受 wshobson/agents 38k★ 启发）

README 摘要

dsh-llm-finetuning DSH（DeepSeek Harness）技能插件： LLM 微调路由与配方。先判断是否该微调（RAG/提示词出口）、方法路由（SFT/DPO/ORPO/KTO/GRPO+RLVR/CPT）、LoRA/QLoRA 配方、评估先行（eval harness）、量化导出。受 wshobson/agents（38k★ MIT）启发的中文原创精简版。安装触发方式描述中包含"微调 / fine-tune / LoRA / QLoRA / DPO / GRPO / RLVR / 偏好优化 / 强化训练 / eval harness"等关键词时自动触发。能力 - 出口优先路由（易变事实→RAG、未定行为→提示词、稳定知识→CPT+SFT） - 方法选型（SFT/DPO/KTO/GRPO 按数据形态） - LoRA/QLoRA 配方要点与评估先行纪律 - 量化导出（FP8/NVFP4/AWQ/GGUF）与质量验证许可 MIT

在 GitHub 查看完整 README →

Agent/智能体dsh-pluginagent

分类

ruvnet/ruflo

🌊 The original agent meta-harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, RAG integration, and native Claude Code / Codex / Hermes and many more Integrated

★ 68,716

volcengine/OpenViking

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

★ 31,761

titanwings/colleague-skill

将冰冷的离别化为温暖的 Skill，欢迎加入数字生命1.0！Transforming cold farewells into warm skills? It's giving rebirth era. Welcome to Digital Life 1.0. 🫶

★ 23,757