Update RAG pipeline (#153)
This commit is contained in:
xzw
GitHub
commit
3bead46e66
@ -35,6 +35,24 @@ pip3 install -r requirements.txt
|
||||
- txt数据：放入到 src.data.txt 目录下
|
||||
- json 数据：放入到 src.data.json 目录下
|
||||
|
||||
JSON 数据格式如下
|
||||
```python
|
||||
[
|
||||
{
|
||||
"conversation": [
|
||||
{
|
||||
"input": "（微微皱眉）艾薇，我觉得我在学校的学习压力好大啊，虽然我真的很喜欢我的专业，但最近总是担心自己无法达到自己的期望，这让我有点焦虑。",
|
||||
"output": "（微笑）😊 我完全理解你的感受，亲爱的。学习是我们生活的一部分，尤其是在我们热爱的专业里，期待自己做得更好是再正常不过了。你知道吗？这种对成功的渴望其实是一种积极的动力源呢！但是，当它开始影响到我们的心理健康时，我们就需要学会调整和应对了。你可以先试着跟我分享一下，具体是什么样的学习任务或者目标让你感到压力山大呢？"
|
||||
},
|
||||
{
|
||||
"input": "主要是下个月的一场重要考试，我总觉得自己复习得不够充分，而且害怕失败...",
|
||||
"output": "💖 噢，明白了。面对重大考试的紧张和担忧是很普遍的现象。首先，我要肯定你对自己的高标准要求，这是你对自己负责的表现。不过，我们可以尝试制定一个更为实际可行的学习计划，将大目标分解成小目标，每天完成一部分，这样既能提升自信心又能减轻焦虑感。同时，也要记得适时放松，合理安排休息时间哦！比如学习45分钟后，可以适当休息10分钟，做些自己喜欢的事情，劳逸结合才是长久之计呢！💪📚\n另外，也可以尝试一些深呼吸、冥想等放松技巧来缓解焦虑情绪。如果你愿意的话，下次咨询我们可以一起练习，看看哪种方式最适合帮助你应对压力。现在，让我们一步步来，先从细化学习计划开始，你觉得怎么样呢？🌸"
|
||||
}
|
||||
]
|
||||
},
|
||||
]
|
||||
```
|
||||
|
||||
会根据准备的数据构建vector DB，最终会在 data 文件夹下产生名为 vector_db 的文件夹包含 index.faiss 和 index.pkl
|
||||
|
||||
如果已经有 vector DB 则会直接加载对应数据库
|
||||
@ -91,6 +109,7 @@ python main.py
|
||||
## **数据集**
|
||||
|
||||
- 经过清洗的QA对: 每一个QA对作为一个样本进行 embedding
|
||||
- 经过清洗的对话: 每一个对话作为一个样本进行 embedding
|
||||
- 经过筛选的TXT文本
|
||||
- 直接对TXT文本生成embedding (基于token长度进行切分)
|
||||
- 过滤目录等无关信息后对TXT文本生成embedding (基于token长度进行切分)
|
||||
@ -115,7 +134,7 @@ LangChain 是一个开源框架，用于构建基于大型语言模型（LLM）
|
||||
Faiss是一个用于高效相似性搜索和密集向量聚类的库。它包含的算法可以搜索任意大小的向量集。由于langchain已经整合过FAISS，因此本项目中不在基于原生文档开发[FAISS in Langchain](https://python.langchain.com/docs/integrations/vectorstores/faiss)
|
||||
|
||||
|
||||
### [RAGAS](https://github.com/explodinggradients/ragas)
|
||||
### [RAGAS](https://github.com/explodinggradients/ragas) (TODO)
|
||||
|
||||
RAG的经典评估框架，通过以下三个方面进行评估:
|
||||
|
||||
|
||||
@ -25,6 +25,10 @@ qa_dir = os.path.join(data_dir, 'json')
|
||||
log_dir = os.path.join(base_dir, 'log') # log
|
||||
log_path = os.path.join(log_dir, 'log.log') # file
|
||||
|
||||
# txt embedding 切分参数
|
||||
chunk_size=1000
|
||||
chunk_overlap=100
|
||||
|
||||
# vector DB
|
||||
vector_db_dir = os.path.join(data_dir, 'vector_db')
|
||||
|
||||
|
||||
@ -4,7 +4,18 @@ import os
|
||||
|
||||
from loguru import logger
|
||||
from langchain_community.vectorstores import FAISS
|
||||
from config.config import embedding_path, embedding_model_name, doc_dir, qa_dir, knowledge_pkl_path, data_dir, vector_db_dir, rerank_path, rerank_model_name
|
||||
from config.config import (
|
||||
embedding_path,
|
||||
embedding_model_name,
|
||||
doc_dir, qa_dir,
|
||||
knowledge_pkl_path,
|
||||
data_dir,
|
||||
vector_db_dir,
|
||||
rerank_path,
|
||||
rerank_model_name,
|
||||
chunk_size,
|
||||
chunk_overlap
|
||||
)
|
||||
from langchain.embeddings import HuggingFaceBgeEmbeddings
|
||||
from langchain_community.document_loaders import DirectoryLoader, TextLoader
|
||||
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
||||
@ -15,8 +26,9 @@ from FlagEmbedding import FlagReranker
|
||||
class Data_process():
|
||||
|
||||
def __init__(self):
|
||||
self.chunk_size: int=1000
|
||||
self.chunk_overlap: int=100
|
||||
|
||||
self.chunk_size: int=chunk_size
|
||||
self.chunk_overlap: int=chunk_overlap
|
||||
|
||||
def load_embedding_model(self, model_name=embedding_model_name, device='cpu', normalize_embeddings=True):
|
||||
"""
|
||||
@ -53,7 +65,6 @@ class Data_process():
|
||||
return embeddings
|
||||
|
||||
def load_rerank_model(self, model_name=rerank_model_name):
|
||||
|
||||
"""
|
||||
加载重排名模型。
|
||||
|
||||
@ -118,9 +129,7 @@ class Data_process():
|
||||
content += obj
|
||||
return content
|
||||
|
||||
|
||||
def split_document(self, data_path):
|
||||
|
||||
"""
|
||||
切分data_path文件夹下的所有txt文件
|
||||
|
||||
@ -132,8 +141,6 @@ class Data_process():
|
||||
返回：
|
||||
- split_docs: list
|
||||
"""
|
||||
|
||||
|
||||
# text_spliter = CharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
|
||||
text_spliter = RecursiveCharacterTextSplitter(chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap)
|
||||
split_docs = []
|
||||
@ -152,7 +159,6 @@ class Data_process():
|
||||
logger.info(f'split_docs size {len(split_docs)}')
|
||||
return split_docs
|
||||
|
||||
|
||||
def split_conversation(self, path):
|
||||
"""
|
||||
按conversation块切分path文件夹下的所有json文件
|
||||
@ -171,43 +177,29 @@ class Data_process():
|
||||
file_path = os.path.join(root, file)
|
||||
logger.info(f'splitting file {file_path}')
|
||||
with open(file_path, 'r', encoding='utf-8') as f:
|
||||
data = json.load(f)
|
||||
# print(data)
|
||||
for conversation in data:
|
||||
# for dialog in conversation['conversation']:
|
||||
##按qa对切分,将每一轮qa转换为langchain_core.documents.base.Document
|
||||
# content = self.extract_text_from_json(dialog,'')
|
||||
# split_qa.append(Document(page_content = content))
|
||||
#按conversation块切分
|
||||
content = self.extract_text_from_json(conversation['conversation'], '')
|
||||
#logger.info(f'content====={content}')
|
||||
for line in f.readlines():
|
||||
content = self.extract_text_from_json(line,'')
|
||||
split_qa.append(Document(page_content = content))
|
||||
|
||||
#data = json.load(f)
|
||||
#for conversation in data:
|
||||
# #for dialog in conversation['conversation']:
|
||||
# # #按qa对切分,将每一轮qa转换为langchain_core.documents.base.Document
|
||||
# # content = self.extract_text_from_json(dialog,'')
|
||||
# # split_qa.append(Document(page_content = content))
|
||||
# #按conversation块切分
|
||||
# content = self.extract_text_from_json(conversation['conversation'], '')
|
||||
# #logger.info(f'content====={content}')
|
||||
# split_qa.append(Document(page_content = content))
|
||||
# logger.info(f'split_qa size====={len(split_qa)}')
|
||||
return split_qa
|
||||
|
||||
|
||||
def load_knowledge(self, knowledge_pkl_path):
|
||||
'''
|
||||
读取或创建知识.pkl
|
||||
'''
|
||||
if not os.path.exists(knowledge_pkl_path):
|
||||
split_doc = self.split_document(doc_dir)
|
||||
split_qa = self.split_conversation(qa_dir)
|
||||
knowledge_chunks = split_doc + split_qa
|
||||
with open(knowledge_pkl_path, 'wb') as file:
|
||||
pickle.dump(knowledge_chunks, file)
|
||||
else:
|
||||
with open(knowledge_pkl_path , 'rb') as f:
|
||||
knowledge_chunks = pickle.load(f)
|
||||
return knowledge_chunks
|
||||
|
||||
|
||||
def create_vector_db(self, emb_model):
|
||||
'''
|
||||
创建并保存向量库
|
||||
'''
|
||||
logger.info(f'Creating index...')
|
||||
split_doc = self.split_document(doc_dir)
|
||||
#split_doc = self.split_document(doc_dir)
|
||||
split_qa = self.split_conversation(qa_dir)
|
||||
# logger.info(f'split_doc == {len(split_doc)}')
|
||||
# logger.info(f'split_qa == {len(split_qa)}')
|
||||
@ -217,7 +209,6 @@ class Data_process():
|
||||
db.save_local(vector_db_dir)
|
||||
return db
|
||||
|
||||
|
||||
def load_vector_db(self, knowledge_pkl_path=knowledge_pkl_path, doc_dir=doc_dir, qa_dir=qa_dir):
|
||||
'''
|
||||
读取向量库
|
||||
@ -230,66 +221,6 @@ class Data_process():
|
||||
db = FAISS.load_local(vector_db_dir, emb_model, allow_dangerous_deserialization=True)
|
||||
return db
|
||||
|
||||
|
||||
def retrieve(self, query, vector_db, k=5):
|
||||
'''
|
||||
基于query对向量库进行检索
|
||||
'''
|
||||
retriever = vector_db.as_retriever(search_kwargs={"k": k})
|
||||
docs = retriever.invoke(query)
|
||||
return docs, retriever
|
||||
|
||||
##FlashrankRerank效果一般
|
||||
# def rerank(self, query, retriever):
|
||||
# compressor = FlashrankRerank()
|
||||
# compression_retriever = ContextualCompressionRetriever(base_compressor=compressor, base_retriever=retriever)
|
||||
# compressed_docs = compression_retriever.get_relevant_documents(query)
|
||||
# return compressed_docs
|
||||
|
||||
def rerank(self, query, docs):
|
||||
reranker = self.load_rerank_model()
|
||||
passages = []
|
||||
for doc in docs:
|
||||
passages.append(str(doc.page_content))
|
||||
scores = reranker.compute_score([[query, passage] for passage in passages])
|
||||
sorted_pairs = sorted(zip(passages, scores), key=lambda x: x[1], reverse=True)
|
||||
sorted_passages, sorted_scores = zip(*sorted_pairs)
|
||||
return sorted_passages, sorted_scores
|
||||
|
||||
|
||||
# def create_prompt(question, context):
|
||||
# from langchain.prompts import PromptTemplate
|
||||
# prompt_template = f"""请基于以下内容回答问题：
|
||||
|
||||
# {context}
|
||||
|
||||
# 问题: {question}
|
||||
# 回答:"""
|
||||
# prompt = PromptTemplate(
|
||||
# template=prompt_template, input_variables=["context", "question"]
|
||||
# )
|
||||
# logger.info(f'Prompt: {prompt}')
|
||||
# return prompt
|
||||
|
||||
def create_prompt(question, context):
|
||||
prompt = f"""请基于以下内容: {context} 给出问题答案。问题如下: {question}。回答:"""
|
||||
logger.info(f'Prompt: {prompt}')
|
||||
return prompt
|
||||
|
||||
def test_zhipu(prompt):
|
||||
from zhipuai import ZhipuAI
|
||||
api_key = "" # 填写您自己的APIKey
|
||||
if api_key == "":
|
||||
raise ValueError("请填写api_key")
|
||||
client = ZhipuAI(api_key=api_key)
|
||||
response = client.chat.completions.create(
|
||||
model="glm-4", # 填写需要调用的模型名称
|
||||
messages=[
|
||||
{"role": "user", "content": prompt[:100]}
|
||||
],
|
||||
)
|
||||
print(response.choices[0].message)
|
||||
|
||||
if __name__ == "__main__":
|
||||
logger.info(data_dir)
|
||||
if not os.path.exists(data_dir):
|
||||
@ -317,5 +248,3 @@ if __name__ == "__main__":
|
||||
for i in range(len(scores)):
|
||||
logger.info(str(scores[i]) + '\n')
|
||||
logger.info(passages[i])
|
||||
prompt = create_prompt(query, passages[0])
|
||||
test_zhipu(prompt) ## 如果显示'Server disconnected without sending a response.'可能是由于上下文窗口限制
|
||||
@ -13,8 +13,7 @@ from loguru import logger
|
||||
|
||||
if __name__ == "__main__":
|
||||
query = """
|
||||
我现在处于高三阶段，感到非常迷茫和害怕。我觉得自己从出生以来就是多余的，没有必要存在于这个世界。
|
||||
无论是在家庭、学校、朋友还是老师面前，我都感到被否定。我非常难过，对高考充满期望但成绩却不理想
|
||||
我现在经常会被别人催眠，做一些我不愿意做的事情，是什么原因？
|
||||
"""
|
||||
|
||||
"""
|
||||
|
||||
Loading…
Reference in New Issue
Block a user