第四期书生大模型实战营 - 基础岛闯关作业4 - InternLM + LlamaIndex RAG 实践RAG RAG

项目地址

算力平台 HF-Mirror

RAG

RAG工作原理

向量数据库`Vector-DB`

发展进程

优化方法

RAG VS Fine-tuning

LlamaIndex

LlamaIndex + RAG

作业

两个任务的脚本仅model配置部分存在少量区别

1. 浦语 API

1.1 准备环境

安装环境

conda create -n llamaindex python=3.10
conda activate llamaindex

# 安装python依赖
pip install einops==0.7.0 protobuf==5.26.1

# 安装 Llamaindex和相关依赖
conda activate llamaindex
pip install llama-index==0.11.20
pip install llama-index-llms-replicate==0.3.0
pip install llama-index-llms-openai-like==0.2.0
pip install llama-index-embeddings-huggingface==0.3.1
pip install llama-index-embeddings-instructor==0.2.1
pip install torch==2.5.0 torchvision==0.20.0 torchaudio==2.5.0 --index-url https://download.pytorch.org/whl/cu121

下载Sentence Transformer模型

huggingface.co
modelscope

相对轻量、支持中文且效果较好的源词向量模型

cd ~
mkdir llamaindex_demo
mkdir model
cd ~/llamaindex_demo
touch download_hf.py

import os

# 设置环境变量
os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'

# 下载模型
os.system('huggingface-cli download --resume-download sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 --local-dir /root/model/sentence-transformer')

cd /root/llamaindex_demo
conda activate llamaindex
python download_hf.py

modelscope下载:

apt-get install git-lfs
git lfs install
git clone https://www.modelscope.cn/Ceceliachenen/paraphrase-multilingual-MiniLM-L12-v2.git

下载NLTK相关资源

# 这里通过码云下载
cd /root
git clone https://gitee.com/yzy0612/nltk_data.git  --branch gh-pages
cd nltk_data
mv packages/*  ./
cd tokenizers
unzip punkt.zip
cd ../taggers
unzip averaged_perceptron_tagger.zip

2. 实操

文档链接

浦语官网和硅基流动都提供了InternLM的类OpenAI接口格式的免费的 API，可以访问以下两个了解两个 API 的使用方法和 Key。

浦语官方 API：internlm.intern-ai.org.cn/api/documen… 硅基流动：cloud.siliconflow.cn/models?mfs=…

运行以下指令，新建一个python文件

cd ~/llamaindex_demo
touch test_internlm.py

打开test_internlm.py 贴入以下代码

from openai import OpenAI
import os
base_url = "https://internlm-chat.intern-ai.org.cn/puyu/api/v1/"
# api_key = "sk-请填写准确的 token！"
api_key = os.getenv('api_key')
model="internlm2.5-latest"

# base_url = "https://api.siliconflow.cn/v1"
# api_key = "sk-请填写准确的 token！"
# model="internlm/internlm2_5-7b-chat"

client = OpenAI(
    api_key=api_key , 
    base_url=base_url,
)

chat_rsp = client.chat.completions.create(
    model=model,
    messages=[{"role": "user", "content": "xtuner是什么？"}],
)

for choice in chat_rsp.choices:
    print(choice.message.content)

之后运行

conda activate llamaindex
cd ~/llamaindex_demo/
python test_internlm.py

运行如下

获取知识库

cd ~/llamaindex_demo
mkdir data
cd data
git clone https://github.com/InternLM/xtuner.git
mv xtuner/README_zh-CN.md ./

运行以下指令，新建一个python文件

cd ~/llamaindex_demo
touch llamaindex_RAG.py

打开llamaindex_RAG.py贴入以下代码

import os 
os.environ['NLTK_DATA'] = '/root/nltk_data'

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llama_index.core.settings import Settings
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.legacy.callbacks import CallbackManager
from llama_index.llms.openai_like import OpenAILike

# Create an instance of CallbackManager
callback_manager = CallbackManager()

api_base_url =  "https://internlm-chat.intern-ai.org.cn/puyu/api/v1/"
model = "internlm2.5-latest"
api_key = os.getenv('api_key')

# api_base_url =  "https://api.siliconflow.cn/v1"
# model = "internlm/internlm2_5-7b-chat"
# api_key = "请填写 API Key"

llm =OpenAILike(model=model, api_base=api_base_url, api_key=api_key, is_chat_model=True,callback_manager=callback_manager)

#初始化一个HuggingFaceEmbedding对象，用于将文本转换为向量表示
embed_model = HuggingFaceEmbedding(
#指定了一个预训练的sentence-transformer模型的路径
    model_name="/root/model/sentence-transformer"
)
#将创建的嵌入模型赋值给全局设置的embed_model属性，
#这样在后续的索引构建过程中就会使用这个模型。
Settings.embed_model = embed_model

#初始化llm
Settings.llm = llm

#从指定目录读取所有文档，并加载数据到内存中
documents = SimpleDirectoryReader("/root/llamaindex_demo/data").load_data()
#创建一个VectorStoreIndex，并使用之前加载的文档来构建索引。
# 此索引将文档转换为向量，并存储这些向量以便于快速检索。
index = VectorStoreIndex.from_documents(documents)
# 创建一个查询引擎，这个引擎可以接收查询并返回相关文档的响应。
query_engine = index.as_query_engine()
response = query_engine.query("xtuner是什么?")

print(response)

使用新的知识库

conda activate llamaindex
cd ~/llamaindex_demo/
python llamaindex_RAG.py

运行结果(使用LlamaIndex RAG（仅API）之后)如下:

使用LlamaIndex web

# 安装依赖
pip install streamlit==1.39.0 
cd ~/llamaindex_demo
touch app.py

# app.py
import streamlit as st
import os
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.legacy.callbacks import CallbackManager
from llama_index.llms.openai_like import OpenAILike

# Create an instance of CallbackManager
callback_manager = CallbackManager()

api_base_url =  "https://internlm-chat.intern-ai.org.cn/puyu/api/v1/"
model = "internlm2.5-latest"
api_key = os.getenv('api_key')

# api_base_url =  "https://api.siliconflow.cn/v1"
# model = "internlm/internlm2_5-7b-chat"
# api_key = "请填写 API Key"

llm =OpenAILike(model=model, api_base=api_base_url, api_key=api_key, is_chat_model=True,callback_manager=callback_manager)

st.set_page_config(page_title="llama_index_demo", page_icon="🦜🔗")
st.title("llama_index_demo")

# 初始化模型
@st.cache_resource
def init_models():
    embed_model = HuggingFaceEmbedding(
        model_name="/root/model/paraphrase-multilingual-MiniLM-L12-v2"
    )
    Settings.embed_model = embed_model

    #用初始化llm
    Settings.llm = llm

    documents = SimpleDirectoryReader("/root/llamaindex_demo/data").load_data()
    index = VectorStoreIndex.from_documents(documents)
    query_engine = index.as_query_engine()

    return query_engine

# 检查是否需要初始化模型
if 'query_engine' not in st.session_state:
    st.session_state['query_engine'] = init_models()

def greet2(question):
    response = st.session_state['query_engine'].query(question)
    return response

      
# Store LLM generated responses
if "messages" not in st.session_state.keys():
    st.session_state.messages = [{"role": "assistant", "content": "你好，我是你的助手，有什么我可以帮助你的吗？"}]    

    # Display or clear chat messages
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.write(message["content"])

def clear_chat_history():
    st.session_state.messages = [{"role": "assistant", "content": "你好，我是你的助手，有什么我可以帮助你的吗？"}]

st.sidebar.button('Clear Chat History', on_click=clear_chat_history)

# Function for generating LLaMA2 response
def generate_llama_index_response(prompt_input):
    return greet2(prompt_input)

# User-provided prompt
if prompt := st.chat_input():
    st.session_state.messages.append({"role": "user", "content": prompt})
    with st.chat_message("user"):
        st.write(prompt)

# Gegenerate_llama_index_response last message is not from assistant
if st.session_state.messages[-1]["role"] != "assistant":
    with st.chat_message("assistant"):
        with st.spinner("Thinking..."):
            response = generate_llama_index_response(prompt)
            placeholder = st.empty()
            placeholder.markdown(response)
    message = {"role": "assistant", "content": response}
    st.session_state.messages.append(message)

# 运行该文件,即可启动网页版
streamlit run app.py

2. 本地部署InternLM

2.1 准备环境

python基础依赖包

pip install einops==0.7.0 protobuf==5.26.1

安装 Llamaindex 与 Pytorch

# 安装 Llamaindex 和相关的包：
conda activate llamaindex
pip install llama-index==0.10.38 llama-index-llms-huggingface==0.2.0 "transformers[torch]==4.41.1" "huggingface_hub[inference]==0.23.1" huggingface_hub==0.23.1 sentence-transformers==2.7.0 sentencepiece==0.2.0

# 安装 LlamaIndex 词嵌入向量依赖：
pip install llama-index-embeddings-huggingface==0.2.0 llama-index-embeddings-instructor==0.1.3
# 在这一步请确定llama-index-embeddings-huggingface安装成功
# 如果存在not found错误，请重新安装
# pip install llama-index-embeddings-huggingface==0.2.0
# 确保 huggingface_hub==0.23.1

# 安装 Pytorch
conda install pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.7 -c pytorch -c nvidia

# 验证pytorch和cuda版本
import torch
print(torch.__version__)        # 应输出类似 '2.0.1'
print(torch.version.cuda)       # 应输出 '11.7'
print(torch.cuda.is_available())# 应输出 True

关于 LlamaIndex 与 Pytorch 安装顺序的说明

关于本次实验的运行环境，我们建议的是如上 先安装 Llamaindex ，再安装 Pytorch。

“先安装 Pytorch 再安装 Llamaindex”存在的问题是：匹配CUDA 11.7的torch安装在前，但是其后安装 LLamaIndex 相关的指令会稳定触发torch的升级到最新版本，而新版本的PyTorch (2.5.1) 默认使用CUDA 12.4，导致 Pytorch 版本与 CUDA版本不匹配。

这样，当进行到模型推理的步骤时，就会报错：

RuntimeError: CUDA error: CUBLAS_STATUS_NOT_INITIALIZED when calling cublasCreate(handle) 这时候就需要再次重新安装正确的torch：

conda install pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.7 -c pytorch -c nvidia 导致操作步骤的繁杂与冗余。

下载Sentence Transformer模型
下载NLTK相关资源
LlamaIndex HuggingFaceLLM

5.1. 运行以下指令，将share文件夹中已经存在的 InternLM2 1.8B 软连接出来

cd ~/model
ln -s /root/share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b/ ./

5.2. 运行以下指令，新建一个 python 文件

cd ~/llamaindex_demo
touch llamaindex_internlm.py

5.3. 打开 llamaindex_internlm.py 贴入以下代码

from llama_index.llms.huggingface import HuggingFaceLLM
from llama_index.core.llms import ChatMessage
llm = HuggingFaceLLM(
    model_name="/root/model/internlm2-chat-1_8b",
    tokenizer_name="/root/model/internlm2-chat-1_8b",
    model_kwargs={"trust_remote_code":True},
    tokenizer_kwargs={"trust_remote_code":True}
)

rsp = llm.chat(messages=[ChatMessage(content="xtuner是什么？")])
print(rsp)

5.4. 之后运行

conda activate llamaindex
cd ~/llamaindex_demo/
python llamaindex_internlm.py

结果如下:

LlamaIndex RAG

6.1. 运行以下命令，获取知识库

cd ~/llamaindex_demo
mkdir data
cd data
git clone https://github.com/InternLM/xtuner.git
mv xtuner/README_zh-CN.md ./

6.2. 运行以下指令，新建一个 python 文件

cd ~/llamaindex_demo
touch llamaindex_RAG.py

6.3. 打开llamaindex_RAG.py贴入以下代码 - 实现将6.1的readme文件完成向量化,导入模型

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.llms.huggingface import HuggingFaceLLM

#初始化一个HuggingFaceEmbedding对象，用于将文本转换为向量表示
embed_model = HuggingFaceEmbedding(
#指定了一个预训练的sentence-transformer模型的路径
    model_name="/root/model/sentence-transformer"
)
#将创建的嵌入模型赋值给全局设置的embed_model属性，
#这样在后续的索引构建过程中就会使用这个模型。
Settings.embed_model = embed_model

llm = HuggingFaceLLM(
    model_name="/root/model/internlm2-chat-1_8b",
    tokenizer_name="/root/model/internlm2-chat-1_8b",
    model_kwargs={"trust_remote_code":True},
    tokenizer_kwargs={"trust_remote_code":True}
)
#设置全局的llm属性，这样在索引查询时会使用这个模型。
Settings.llm = llm

#从指定目录读取所有文档，并加载数据到内存中
documents = SimpleDirectoryReader("/root/llamaindex_demo/data").load_data()
#创建一个VectorStoreIndex，并使用之前加载的文档来构建索引。
# 此索引将文档转换为向量，并存储这些向量以便于快速检索。
index = VectorStoreIndex.from_documents(documents)
# 创建一个查询引擎，这个引擎可以接收查询并返回相关文档的响应。
query_engine = index.as_query_engine()
response = query_engine.query("xtuner是什么?")

print(response)

重新执行1.7的指令即可发现知识库(README_zh-CN.md)已导入,从而获取到想要的内容

使用LlamaIndex web

# 安装依赖
pip install streamlit==1.39.0 
cd ~/llamaindex_demo
touch app.py

# app.py
import streamlit as st
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.llms.huggingface import HuggingFaceLLM

st.set_page_config(page_title="llama_index_demo", page_icon="🦜🔗")
st.title("llama_index_demo")

# 初始化模型
@st.cache_resource
def init_models():
    embed_model = HuggingFaceEmbedding(
        model_name="/root/model/sentence-transformer"
    )
    Settings.embed_model = embed_model

    llm = HuggingFaceLLM(
        model_name="/root/model/internlm2-chat-1_8b",
        tokenizer_name="/root/model/internlm2-chat-1_8b",
        model_kwargs={"trust_remote_code": True},
        tokenizer_kwargs={"trust_remote_code": True}
    )
    Settings.llm = llm

    documents = SimpleDirectoryReader("/root/llamaindex_demo/data").load_data()
    index = VectorStoreIndex.from_documents(documents)
    query_engine = index.as_query_engine()

    return query_engine

# 检查是否需要初始化模型
if 'query_engine' not in st.session_state:
    st.session_state['query_engine'] = init_models()

def greet2(question):
    response = st.session_state['query_engine'].query(question)
    return response


# Store LLM generated responses
if "messages" not in st.session_state.keys():
    st.session_state.messages = [{"role": "assistant", "content": "你好，我是你的助手，有什么我可以帮助你的吗？"}]

    # Display or clear chat messages
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.write(message["content"])

def clear_chat_history():
    st.session_state.messages = [{"role": "assistant", "content": "你好，我是你的助手，有什么我可以帮助你的吗？"}]

st.sidebar.button('Clear Chat History', on_click=clear_chat_history)

# Function for generating LLaMA2 response
def generate_llama_index_response(prompt_input):
    return greet2(prompt_input)

# User-provided prompt
if prompt := st.chat_input():
    st.session_state.messages.append({"role": "user", "content": prompt})
    with st.chat_message("user"):
        st.write(prompt)

# Gegenerate_llama_index_response last message is not from assistant
if st.session_state.messages[-1]["role"] != "assistant":
    with st.chat_message("assistant"):
        with st.spinner("Thinking..."):
            response = generate_llama_index_response(prompt)
            placeholder = st.empty()
            placeholder.markdown(response)
    message = {"role": "assistant", "content": response}
    st.session_state.messages.append(message)

# 运行该文件,即可启动网页版
streamlit run app.py

结果如下:

3. Streamlit+LlamaIndex+浦语API的 Space 部署到 Hugging Face

space SDK选中 Streamlit
git clone选ssh方式,https方式需科学上网
用于模型启动的app.py代码如下:

import os
import streamlit as st
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.legacy.callbacks import CallbackManager
from llama_index.llms.openai_like import OpenAILike

# 配置 API
api_base_url = "https://internlm-chat.intern-ai.org.cn/puyu/api/v1/"
model = "internlm2.5-latest"
api_key = os.getenv("api_key")

# 检查 API 密钥
if not api_key:
    st.error("API key is not set. Please configure the environment variable 'api_key'.")
    st.stop()

# Streamlit 页面配置
st.set_page_config(page_title="LlamaIndex Demo", page_icon="🦜🔗")
st.title("LlamaIndex Demo")

# 初始化查询引擎
@st.cache_resource
def init_query_engine():
    try:
        # 确保数据路径存在
        data_path = "./data"
        if not os.path.exists(data_path) or not os.listdir(data_path):
            st.error(f"Data directory '{data_path}' is missing or empty.")
            st.stop()
    
        # 使用 HuggingFaceEmbedding 替代 SentenceTransformer
        embed_model = HuggingFaceEmbedding(
            model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
        )

        # 创建CallbackManager实例
        callback_manager = CallbackManager()

        # 初始化 OpenAILike 模型
        llm = OpenAILike(
            model=model,
            api_base=api_base_url,
            api_key=api_key,
            is_chat_model=True,
            callback_manager=callback_manager
        )
        
        Settings.embed_model = embed_model
    
        #用初始化llm
        Settings.llm = llm

        # 加载文档
        documents = SimpleDirectoryReader(data_path).load_data()

        # 构建索引
        index = VectorStoreIndex.from_documents(documents)
        # index = VectorStoreIndex.from_documents(documents, embed_model=embed_model, llm=llm)

        # 构建查询引擎
        query_engine = index.as_query_engine()

        return query_engine
    except Exception as e:
        st.error(f"Error initializing query engine: {e}")
        st.stop()

# 初始化查询引擎
if 'query_engine' not in st.session_state:
    st.session_state['query_engine'] = init_query_engine()

# 聊天历史管理
if "messages" not in st.session_state:
    st.session_state.messages = [
        {"role": "assistant", "content": "你好，我是你的助手，有什么我可以帮助你的吗？"}
    ]

# 清除聊天历史
def clear_chat_history():
    st.session_state.messages = [
        {"role": "assistant", "content": "你好，我是你的助手，有什么我可以帮助你的吗？"}
    ]

st.sidebar.button("Clear Chat History", on_click=clear_chat_history)

# 生成助手响应
def generate_response(prompt):
    try:
        response = st.session_state['query_engine'].query(prompt)
        return response.response if response else "无法生成有效响应。"
    except Exception as e:
        return f"Error: {e}"

# 展示聊天历史
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.write(message["content"])

# 用户输入
if prompt := st.chat_input():
    # 添加用户消息到会话历史
    st.session_state.messages.append({"role": "user", "content": prompt})
    with st.chat_message("user"):
        st.write(prompt)

    # 生成助手的响应
    with st.chat_message("assistant"):
        with st.spinner("Thinking..."):
            response = generate_response(prompt)
            st.write(response)
        st.session_state.messages.append({"role": "assistant", "content": response})

requirements.txt内容如下:

sentence-transformers>=2.6.1
llama-index==0.11.20
llama-index-llms-replicate==0.3.0
llama-index-llms-openai-like==0.2.0
llama-index-embeddings-huggingface==0.3.1
llama-index-embeddings-instructor==0.2.1
torch==2.5.0
torchvision==0.20.0
torchaudio==2.5.0

部署站点

运行结果如下:

第四期书生大模型实战营 - 基础岛闯关作业4 - InternLM + LlamaIndex RAG 实践

RAG