CareGPT (关怀GPT)是一个医疗大语言模型，同时它集合了数十个公开可用的医疗微调数据集和开放可用的医疗大语言.zip

<p align="left"> &nbsp中文&nbsp ｜ <a href="README_en.md">English</a> </p> <br> <div align="center"> <a href="https://github.com/WangRongsheng/CareGPT"> <img src="./assets/images/caregpt-logo.png" alt="Logo" height="280"> </a> </div> <div align="center"> <p align="center"> <h3> CareGPT (关怀GPT)：医疗LLM，开源驱动，共创健康未来 </h3> <p align="center"> <em>资源整合 / 开源模型 / 丰富数据 / 高效部署 </em> </p> <p align="center"> <a href='https://github.com/WangRongsheng/CareLlama'> <img src='https://img.shields.io/badge/Project-Page-Green'> </a> <a href='https://github.com/WangRongsheng/CareLlama'> <img src='https://img.shields.io/badge/Paper-Arxiv-red'> </a> <a href="#"> <img alt="GitHub Contributors" src="https://colab.research.google.com/assets/colab-badge.svg" /> </a> <a href='https://huggingface.co/wangrongsheng'> <img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue'> </a> </br> <a href="https://github.com/WangRongsheng/CareLlama/graphs/contributors"> <img alt="GitHub Contributors" src="https://img.shields.io/github/contributors/WangRongsheng/CareLlama" /> </a> <a href="https://github.com/WangRongsheng/CareLlama/issues"> <img alt="Issues" src="https://img.shields.io/github/issues/WangRongsheng/CareLlama?color=0088ff" /> </a> <a href="https://github.com/WangRongsheng/CareLlama/pulls"> <img alt="GitHub pull requests" src="https://img.shields.io/github/issues-pr/WangRongsheng/CareLlama?color=0088ff" /> </a> <a href=href="https://github.com/WangRongsheng/CareLlama/stargazers"> <img src="https://img.shields.io/github/stars/WangRongsheng/CareLlama?color=ccf"> </a> <a href=href="https://github.com/WangRongsheng/CareLlama"> <img src="https://img.shields.io/github/repo-size/WangRongsheng/CareLlama.svg?style=flat-square"> </a> </br> <!-- <a href=href="https://github.com/WangRongsheng/CareLlama"> <img src="https://visitor-badge.laobi.icu/badge?page_id=https://github.com/WangRongsheng/CareLlama"> </a> --> <a href=href="https://github.com/WangRongsheng/CareLlama"> <img src="https://img.shields.io/github/last-commit/WangRongsheng/CareLlama"> </a> <a href="https://github.com/WangRongsheng/CareLlama/blob/main/LICENSE"> <img alt="GitHub Contributors" src="https://img.shields.io/badge/License-CC%20BY--NC--SA%204.0-lightgrey.svg" /> </a> </p> </div> <!--center><kbd><img src="./docs/images/usage.png" height="550px"/></kbd></center--> <p align="center"> <a href="https://www.youtube.com/watch?v=ADVsEPbL9CQ"><strong>视频教程</strong></a> <a href="https://github.com/WangRongsheng/CareLlama/tree/main#5gradio%E9%83%A8%E7%BD%B2"><strong>安装部署</strong></a> <a href="https://huggingface.co/spaces/wangrongsheng/CareLlama"><strong>在线体验</strong></a> </p>  ⚡特性： 1. 添加[ChatGPT fine-tuning](https://github.com/WangRongsheng/CareLlama/tree/main/ChatGPT)实现，推荐有额度的朋友在ChatGPT上进行微调实验； 2. 支持[ChatGPT-Next-Web](https://github.com/WangRongsheng/CareLlama/tree/main#6chatgpt-next-web%E9%83%A8%E7%BD%B2)部署微调的模型； 3. 支持[Gradio](https://github.com/WangRongsheng/CareLlama/tree/main#5gradio%E9%83%A8%E7%BD%B2)部署微调的模型； 4. 支持LLaMA、LLaMA-2全系列模型训练； 5. 支持LoRA、QLoRA，包括后续PPO、DPO强化学习训练； 6. 支持模型与知识库结合问答； 7. 开源了超过[60个医院科室的导诊材料信息](https://github.com/WangRongsheng/CareLlama/tree/main/data/Hospital%20Guide)； 8. 开发了支持[GPT-4/ChatGPT模型蒸馏医学数据](https://huggingface.co/spaces/wangrongsheng/DataMaker)的工具，能够批量生成各种用于构建知识库和微调的数据； 9. 聚合了丰富的开源医学LLM、LLM训练的医学数据、LLM部署资料、LLM测评以及相关LLM的资源整理； 10. 我们参与了医学LLM的[CMB榜单评测-IvyGPT](https://cmedbenchmark.llmzoo.com/static/leaderboard.html)，在测试中，我们领先ChatGPT及一众开源医学LLM； 11. 我们基于自有数据集在不同基座LLM上训练开源了多个医疗LLM，您可以直接下载体验； <div align="center"> <a href="https://github.com/WangRongsheng/CareGPT"> <img src="./assets/images/caregpt.jpg" alt="Logo" height="280"> </a> </div> # 🎁数据集 #### 预训练数据 - [LLM-Pretrain-FineTune/data_pretrain](https://github.com/X-jun-0130/LLM-Pretrain-FineTune/tree/main/data_pretrain) - [MedicalGPT/pretrain](https://github.com/shibing624/MedicalGPT/tree/main/data/pretrain) - [zysj](https://www.zysj.com.cn/) - [TCM-Ancient-Books](https://github.com/xiaopangxia/TCM-Ancient-Books) (近700项中医药古籍文本) - [epfl-llm/guidelines](https://huggingface.co/datasets/epfl-llm/guidelines) #### 监督训练数据 - [icliniq-10k(en)](https://drive.google.com/file/d/1ZKbqgYqWc7DJHs3N9TQYQVPdDQmZaClA/view?usp=sharing) - [HealthCareMagic-100k(en)](https://drive.google.com/file/d/1lyfqIwlLSClhgrCutWuEe_IACNq6XNUt/view?usp=sharing) - [ShenNong_TCM_Dataset](https://huggingface.co/datasets/michaelwzhu/ShenNong_TCM_Dataset) - ✅[ChatMed_Consult_Dataset](https://huggingface.co/datasets/michaelwzhu/ChatMed_Consult_Dataset) - [Chinese-medical-dialogue-data](https://github.com/Toyhom/Chinese-medical-dialogue-data) - [cMedQA2](https://github.com/zhangsheng93/cMedQA2) - ✅[Huatuo-26M](https://github.com/FreedomIntelligence/Huatuo-26M) - [cMedQA2](https://github.com/zhangsheng93/cMedQA2) - [webMedQA](https://github.com/hejunqing/webMedQA) - [PubMedQA](https://pubmedqa.github.io/) - [CMCQA](https://github.com/WENGSYX/CMCQA) - ✅[QiZhenGPT](https://github.com/CMKRG/QiZhenGPT/tree/main/data) - ✅[LLM-Pretrain-FineTune/data_sft](https://github.com/X-jun-0130/LLM-Pretrain-FineTune/tree/main/data_sft) - [Medical-Dialogue-System](https://github.com/UCSD-AI4H/Medical-Dialogue-System) - [IMCS-V2](https://github.com/lemuria-wchen/imcs21) - [CHIP-MDCFNPC](https://tianchi.aliyun.com/dataset/95414) - [MedDG](https://tianchi.aliyun.com/dataset/95414) - ✅[HuatuoGPT-sft-data-v1](https://huggingface.co/datasets/FreedomIntelligence/HuatuoGPT-sft-data-v1) - [MedicalGPT/finetune](https://github.com/shibing624/MedicalGPT/tree/main/data/finetune) - ✅[shibing624/medical](https://huggingface.co/datasets/shibing624/medical) - [medAlpaca/data](https://github.com/kbressem/medAlpaca#data-overview) - ✅[Zhongjing/sft](https://github.com/SupritYoung/Zhongjing) - [medical_dialog](https://huggingface.co/datasets/medical_dialog) - [huatuo_encyclopedia_qa](https://huggingface.co/datasets/FreedomIntelligence/huatuo_encyclopedia_qa) - [Med-ChatGLM/data](https://github.com/SCIR-HI/Med-ChatGLM/tree/main/data) - [CMB](https://github.com/FreedomIntelligence/CMB) - [GenMedGPT-5k(en)](https://drive.google.com/file/d/1nDTKZ3wZbZWTkFMBkxlamrzbNz0frugg/view?usp=sharing) - [Alpaca-CoT(general)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT) - ✅[DISC-Med-SFT](https://huggingface.co/datasets/Flmc/DISC-Med-SFT) - ✅[HuatuoGPT2_sft_instruct](https://huggingface.co/datasets/FreedomIntelligence/HuatuoGPT2_sft_instruct_GPT4_50K) - [FreedomIntelligence/Medbase_data](https://huggingface.co/datasets/FreedomIntelligence/Medbase_data) #### 奖励训练数据 - [MedicalGPT/reward](https://github.com/shibing624/MedicalGPT/tree/main/data/reward) - [Zhongjing/rw](https://github.com/SupritYoung/Zhongjing/tree/main/data) - [comparison_gpt4_data](https://huggingface.co/datasets/wangrongsheng/comparison_gpt4_data) - [HH-RLHF](https://huggingface.co/datasets/Anthropic/hh-rlhf) - [UltraFeedback](https://github.com/OpenBMB/UltraFeedback) # 🗜️全流程训练 ## 1.安装依赖 ```python conda create -n llm python=3.1