使用Ollama

推荐

1. Ollama 本地运行大模型(LLM)完全指南
2. Ollama中文学习

应用

查看可支持的模型：https://ollama.com/library

1. 查看运行中的模型

ollama ps

2. 停止模型

方法1: kill -9 端口号

方法2: 关闭ollama的窗口

3. API调用模型

• 启动模型
• 单次调用模型

from ollama import Client
import time

def llm(model, message):
  client = Client(host='http://localhost:11434')
  response = client.chat(model=model, messages=[
    {
      'role': 'user',
      'content': message,
    },
  ])
  return response['message']['content']

start = time.time()
res = llm('qwen2', '您好，请介绍一下自己')
end = time.time()
print(res)
print(f'time: {end-start} s')

• 多次调用模型

import json
import requests

model = "qwen2"
def chat(messages):
    r = requests.post(
        "http://0.0.0.0:11434/api/chat",
        json={"model": model, "messages": messages, "stream": True},
    )
    r.raise_for_status()
    output = ""
    for line in r.iter_lines():
        body = json.loads(line)
        if "error" in body:
            raise Exception(body["error"])
        if body.get("done") is False:
            message = body.get("message", "")
            content = message.get("content", "")
            output += content
            print(content, end="", flush=True)

        if body.get("done", False):
            message["content"] = output
            return message
def main():
    messages = []
    while True:
        user_input = input("Enter a prompt: ")
        if not user_input:
            exit()
        print()
        messages.append({"role": "user", "content": user_input})
        message = chat(messages)
        messages.append(message)
        print("\n\n")

if __name__ == "__main__":
    main()

3. 对接langchain

• 安装langchain

pip install langchain

pip install -U langchain-community

• langchain启动模型

from langchain.llms import Ollama
ollama = Ollama(base_url='http://localhost:11434',model="qwen2")
print(ollama("你好，请介绍自己"))

4. 文档问答【RAG】

• 加载文档

LangChain 的 WebBaseLoader 来从任何网页加载文本。

from langchain.document_loaders import WebBaseLoader

loader = WebBaseLoader("https://www.gutenberg.org/files/1727/1727-h/1727-h.htm")

data = loader.load()

• 选择向量库和嵌入模型

向量库：pip install chromadb

嵌入模型：ollama run nomic-embed-text

from langchain.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import OllamaEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import RetrievalQA
from langchain.llms import Ollama

# 链接模型
ollama = Ollama(base_url='http://localhost:11434',model="qwen2")

# 加载文档
loader = WebBaseLoader("https://www.ruanyifeng.com/blog/2024/07/weekly-issue-308.html")
data = loader.load()
# print(data)

# 文档分割
text_splitter=RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
all_splits = text_splitter.split_documents(data)

# 文档存到向量库
oembed = OllamaEmbeddings(base_url="http://localhost:11434", model="nomic-embed-text")
vectorstore = Chroma.from_documents(documents=all_splits, embedding=oembed)

• 通过相似度搜索匹配

from langchain.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import OllamaEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import RetrievalQA
from langchain.llms import Ollama

# 链接模型
ollama = Ollama(base_url='http://localhost:11434',model="qwen2")

# 加载文档
loader = WebBaseLoader("https://www.ruanyifeng.com/blog/2024/07/weekly-issue-308.html")
data = loader.load()
# print(data)

# 文档分割
text_splitter=RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
all_splits = text_splitter.split_documents(data)

# 文档存到向量库
oembed = OllamaEmbeddings(base_url="http://localhost:11434", model="nomic-embed-text")
vectorstore = Chroma.from_documents(documents=all_splits, embedding=oembed)

# 相似度匹配
question="请问找不到工作，还能读博么？"
docs = vectorstore.similarity_search(question)
print(f'匹配长度：{len(docs)}')
print(f'匹配内容：{docs}')

• 将问题和文档的相关部分拼接在一起，叫做链，发送给模型

from langchain.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import OllamaEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import RetrievalQA
from langchain.llms import Ollama

# 链接模型
ollama = Ollama(base_url='http://localhost:11434',model="qwen2")

# 加载文档
loader = WebBaseLoader("https://www.ruanyifeng.com/blog/2024/07/weekly-issue-308.html")
data = loader.load()
# print(data)

# 文档分割
text_splitter=RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
all_splits = text_splitter.split_documents(data)

# 文档存到向量库
oembed = OllamaEmbeddings(base_url="http://localhost:11434", model="nomic-embed-text")
vectorstore = Chroma.from_documents(documents=all_splits, embedding=oembed)

# 相似度匹配
question="请问找不到工作，还能读博么？"
docs = vectorstore.similarity_search(question)
print(f'匹配长度：{len(docs)}')
print(f'匹配内容：{docs}')

# 将搜索结果与问题拼接
qachain=RetrievalQA.from_chain_type(ollama, retriever=vectorstore.as_retriever())
res = qachain.invoke({"query": question})

print(res['result'])