论文外挂插件引用论文

langchain+6b 外挂知识库
https://blog.csdn.net/hugo_lei/article/details/135327051
https://zhuanlan.zhihu.com/p/643531454
cpu langchain https://zhuanlan.zhihu.com/p/645386589
https://github.com/chatchat-space/Langchain-Chatchat
已更名
langchainchatchat
https://blog.csdn.net/weixin_44729138/article/details/136042471
pwd问题https://blog.csdn.net/morecccc/article/details/136574595#:~:text=ModuleNotFoundError%3A%20No%20module%20named%20%27pwd%27%20%E6%89%A7%E8%A1%8Cinit_database%E7%9A%84%E6%97%B6%E5%80%99%E6%8A%A5%E9%94%99%EF%BC%8C%E8%A7%A3%E5%86%B3%E6%96%B9%E6%B3%95%E5%A6%82%E4%B8%8B%EF%BC%9A%201%E3%80%81%E5%88%9B%E5%BB%BApwd.py%20%E4%BB%A3%E7%A0%81%E5%A6%82%E4%B8%8B,%2A%20def%20get_username%20%28%29%3A%20return%20getpwuid%28getuid%28%29%29%5B%200%5D%202%E3%80%81%E6%8A%8Apwd.py%E6%94%BE%E5%85%A5%E5%88%B0python%E7%8E%AF%E5%A2%83%E7%9A%84Lib%E7%9B%AE%E5%BD%95%E4%B8%8B%EF%BC%8C%E6%88%91%E7%94%A8%E7%9A%84conda%E7%9A%84%E8%99%9A%E6%8B%9F%E7%8E%AF%E5%A2%83%EF%BC%8C%E7%9B%AE%E5%BD%95%E6%98%AF%E5%9C%A8
conda虚拟环境需要安装`conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch`
server_config 改chatglm-6b为cpu

https://github.com/chatchat-space/Langchain-Chatchat/wiki/%E5%BC%80%E5%8F%91%E7%8E%AF%E5%A2%83%E9%83%A8%E7%BD%B2/bf96a2450eb72052236f577c2a2db73264d9a5da

加-i https://pypi.tuna.tsinghua.edu.cn/simple 临时使用国内源

git lfs clone https://www.modelscope.cn/Jerry0/m3e-base.git
git lfs clone https://www.modelscope.cn/ZhipuAI/chatglm3-6b.git

问题：libGL.so.1:cannot open shared 
`ip install opencv-python-headless`

---
外挂知识库2
wenda
https://blog.csdn.net/engchina/article/details/130592913
  https://blog.csdn.net/qq_45841569/article/details/131487887
 缺少docx 安装python-docx
 预训练不要在plugins目录里
 langchain 0.0.228
 
 ---
 
p-tuningv2 推荐
地址
https://github.com/THUDM/ChatGLM-6B/tree/main/ptuning
进入`../ChatGLM3/finetune_demo`
下载示例`wget https://cloud.tsinghua.edu.cn/f/b3f119a008264b1cabd1/?dl=1`
j解压
`tar -xvf index.html?dl=1`



LLaMA --面板
https://zhuanlan.zhihu.com/p/655008197


---
`conda env list`

选择需要切换到的环境，用下面命令切换，如切换到chat

`conda activate chat`

python3 startup.py -a


chatglm-6b训练

git clone https://github.com/THUDM/ChatGLM3.git
(gitcode.net)
 pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple/
安装git-lfs
然后 git lfs install
https://modelscope.cn/models/ZhipuAI/chatglm3-6b/summary
新目录model
git clone https://www.modelscope.cn/ZhipuAI/chatglm3-6b.git
配置 glm3
进入glm3/basic_demo
修改web_demo_streamlit.py

MODEL_PATH = os.environ.get('MODEL_PATH', '../../chatglm3-6b')
Windows要改为正斜杠
若要cpu计算则
	model = AutoModel.from_pretrained(MODEL_PATH, trust_remote_code=True, device_map="auto").float()
        
`streamlit run web_demo_streamlit.py`

---
理论基础：
https://modelscope.cn/models/ZhipuAI/ChatGLM-6B/summary
模型：
https://github.com/THUDM/ChatGLM3

chatGLM-6B

windows安装python后没有pip

FastAPI基本路由

根路径路由
```
from fastapi import FastAPI

app = FastAPI()

@app.get("/")
def read_root():
    return {"Hello": "World"}
```


路径参数
```
@app.get("/items/{item_id}")
def read_item(item_id: int, q: str = None):
    return {"item_id": item_id, "q": q}
```

fastapi请求数据
定义/item/，接受skip,limit参数（整型）,默认值0，10
```
@app.get("/items/")
def read_item(skip:int =0 ,limit int =10):
return {"skip": skip,"limit": limit}