骆驼(Luotuo): 开源中文大语言模型

骆驼(Luotuo)项目是由冷子昂 @ 商汤科技, 陈启源 @ 华中师范大学以及李鲁鲁 @ 商汤科技发起的中文大语言模型开源项目，包含了一系列大语言模型、数据、管线和应用。

骆驼项目不是商汤科技的官方产品。

我们将项目命名为骆驼 Luotuo (Camel) 主要是因为，Meta之前的项目LLaMA（驼马）和斯坦福之前的项目alpaca(羊驼)都属于偶蹄目-骆驼科（Artiodactyla-Camelidae）。而且骆驼科只有三个属，再不起这名字就来不及了。

项目重要更新 [ ... ]

[2023-07-12] 骆驼嵌入更新中模型 <a href="https://colab.research.google.com/github/LC1332/Luotuo-Text-Embedding/blob/main/notebook/Luotuo_Embedding_Visualization_Medium.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> 。我们后面将准备再训一个英语的嵌入模型。

[2023-06-07] 最近很多精力都在做 Chat凉宫春日, 可以点这个体验 <a href="https://colab.research.google.com/github/LC1332/Prophet-Andrew-Ng/blob/main/prophet-code/haruhiLangChain.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> ,这个项目还在持续招人, 欢迎联系

子项目一览

<table style="table-layout:fixed; width: 100%;"> <tr> <td style="text-align:center; width:50%;" width="50%"> Chat凉宫春日 <a href="https://github.com/LC1332/Chat-Haruhi-Suzumiya"> <img src="https://github.com/LC1332/Luotuo-Chinese-LLM/blob/main/image/ProjHaruhi.jpg" alt="Contributor 2" height="250" width="100%"> Chat凉宫春日是模仿凉宫春日等一系列动漫人物，使用近似语气、个性和剧情聊天的语言模型方案 </a> </td> <td style="text-align:center; width:50%;" width="50%"> 骆驼嵌入 <a href="https://github.com/LC1332/Luotuo-Text-Embedding"> <img src="https://github.com/LC1332/Luotuo-Chinese-LLM/blob/main/image/ProjEmbedding.png" alt="Contributor 2" height="250" width="100%"> 骆驼嵌入: Generative Text Embedding Model distilled from OpenAI API </a> </td> </tr> <tr> <td style="text-align:center; width:50%;" width="50%"> 骆驼QA <a href="https://github.com/LC1332/Luotuo-QA"> <img src="https://github.com/LC1332/Luotuo-Chinese-LLM/blob/main/image/ProjQA.png" alt="Contributor 1" height="250" width="100%"> 骆驼QA: Better Conversational Question Answering Model with Answer Completion </a> </td> <td style="text-align:center; width:50%;"> 迷你骆驼 <a href="https://github.com/LC1332/Mini-Luotuo"> <img src="https://github.com/LC1332/Mini-Luotuo/blob/main/images/ProjectIcon.png" alt="Contributor 2" height="250"> 迷你骆驼:一系列蒸馏指令数据得到的中文语言模型 </a> </td> </tr> <tr> <td style="text-align:center; width:50%;"> 丝绸之路 <a href="https://github.com/LC1332/Luotuo-Silk-Road"> <img src="https://github.com/LC1332/Luotuo-Chinese-LLM/blob/main/image/icon_silk_road.png" alt="Contributor 2" height="250"> 丝绸之路: 构建中文大语言模型的数据基础 </a> </td> <td style="text-align:center; width:50%;"> Vanilla 骆驼 <a href="https://github.com/LC1332/Chinese-alpaca-lora"> <img src="https://github.com/LC1332/Luotuo-Chinese-LLM/blob/main/image/icon_luotuo.png" alt="Contributor 2" height="250"> 骆驼: An Instruction-following Chinese Language model, LoRA tuning on LLaMA </a> </td> </tr> <tr> <td style="text-align:center; width:50%;"> 骆驼先知 <a href="https://github.com/LC1332/Prophet-Andrew-Ng"> <img src="https://github.com/LC1332/Prophet-Andrew-Ng/blob/main/figures/icon.png" alt="Contributor 2" height="250"> 骆驼先知是模仿纪伯伦的《先知》进行哲学讨论。项目包含了Andrew Ng吴恩达Prompt工程的笔记和LangChain的笔记 </a> </td> <td style="text-align:center; width:50%;"> 丝绸魔法书 <a href="https://github.com/LC1332/Luotuo-Silk-Magic-Book"> <img src="https://github.com/LC1332/Luotuo-Chinese-LLM/blob/main/image/icon_silk_scroll.png" alt="Contributor 2" height="250"> 丝绸魔法书记录了大语言模型的一些魔法提示词(prompt)。 </a> </td> </tr> <tr> <td style="text-align:center;"> 骆驼RPG <a href="https://github.com/LC1332/Chinese-generative-agents"> LuotuoRPG是斯坦福Generative Agents的中文版本尝试。 </a> </td> <td style="text-align:center;"> 丫丫-MOSS <a href="https://github.com/qiyuan-chen/Yaya-Moss-Alpaca-LoRA"> 丫丫是基于复旦MOSS的LoRA训练代码 </a> </td> </tr> <tr> <td style="text-align:center;"> 骆驼大乱斗 <a href="https://github.com/LC1332/Luotuo-Chinese-LLM"> 骆驼大乱斗正在构建之中... 骆驼大乱斗: Generating Massive Content for a Text-based Fighting Game </a> </td> <td style="text-align:center;"> 骆驼CLIP <a href="https://github.com/LC1332/Luotuo-Chinese-LLM"> 骆驼CLIP正在构建之中... 骆驼CLIP: Aligning Existed CLIP model with Multiple Prior Leveraged </a> </td> </tr> </table>

项目重要更新 [ ... ]

[2023-05-20] 发布项目迷你骆驼:一系列蒸馏指令数据得到的中文语言模型, 3.5B小模型和测试代码 <a href="https://colab.research.google.com/github/LC1332/Mini-Luotuo/blob/main/3.5B_minimal.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>

[2023-05-06] 升级了数据集批量翻译的脚本 <a href="https://colab.research.google.com/github/LC1332/Luotuo-Chinese-LLM/blob/main/notebook/improvedTranslation.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>

[2023-04-27] 尝试翻译了一下斯坦福25个Agents生成的Generative Agents的工作。代码仓库，colab链接 <a href="https://colab.research.google.com/github/LC1332/Chinese-generative-agents/blob/main/notebook/Chinese_story_turbo.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>

[2023-4-16] 骆驼嵌入代码已经发布，可以用这个colab链接体验 <a href="https://colab.research.google.com/github/LC1332/Luotuo-Text-Embedding/blob/main/notebook/Luotuo_Embedding_Visualization.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> 可支持文本可视化，聚类，模糊搜索等应用

快速上手

	Colab链接	细节
Chat凉宫春日(图文)	<a href="https://colab.research.google.com/github/LC1332/Chat-Haruhi-Suzumiya/blob/main/notebook/gradio_with_Image.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>	能够根据台词搜索图片的gradio版本
Chat凉宫春日	<a href="https://colab.research.google.com/github/LC1332/Prophet-Andrew-Ng/blob/main/prophet-code/haruhiLangChain.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>	李鲁鲁最早开发的gradio Chat凉宫春日
骆驼先知	<a href="https://colab.research.google.com/github/LC1332/Prophet-Andrew-Ng/blob/main/prophet-code/prophet-gradio.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>	骆驼先知的Gradio交互版本
骆驼QA	<a href="https://colab.research.google.com/github/LC1332/Luotuo-QA/blob/main/colab/LuotuoQA_Gradio.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>	骆驼QA的0.1模型，基于一段给定文本做问答
骆驼嵌入(小)	<a href="https://colab.research.google.com/github/LC1332/Luotuo-Text-Embedding/blob/main/notebook/Luotuo_Embedding_Visualization.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>	LuotuoBERT 文本可视化聚类搜索
驼铃C 文本摘要	<a href="https://colab.research.google.com/github/LC1332/Luotuo-Chinese-LLM/blob/main/notebook/TuoLingC_evaluation_code.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>	基于GLM-6B的文本摘要模型
批量翻译	<a href="https://colab.research.google.com/github/LC1332/Luotuo-Chinese-LLM/blob/main/notebook/improvedTranslation.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>	利用GPT接口对数据集进行批量翻译
骆驼RPG	<a href="https://colab.research.google.com/github/LC1332/Chinese-generative-agents/blob/main/notebook/Chinese_story_turbo.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>	一个斯坦福35个Agents生成的中文版
驼铃B Chat哈利波特	-	Chat哈利波特的初步汇报
驼铃A	<a href="https://colab.research.google.com/github/LC1332/Luotuo-Chinese-LLM/blob/main/notebook/TuoLing_evaluation_code.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In