Add files via upload
allow user to load embedding & rerank models from cache
This commit is contained in:
zealot52099
GitHub
parent
382d338ab3
commit
0aa58372bb
@ -1,305 +1,329 @@
|
|||||||
import json
|
import json
|
||||||
import pickle
|
import pickle
|
||||||
import faiss
|
import faiss
|
||||||
import pickle
|
import pickle
|
||||||
import os
|
import os
|
||||||
|
|
||||||
from loguru import logger
|
from loguru import logger
|
||||||
from sentence_transformers import SentenceTransformer
|
from sentence_transformers import SentenceTransformer
|
||||||
from langchain_community.vectorstores import FAISS
|
from langchain_community.vectorstores import FAISS
|
||||||
from config.config import embedding_path, doc_dir, qa_dir, knowledge_pkl_path, data_dir, base_dir, vector_db_dir
|
from config.config import embedding_path, doc_dir, qa_dir, knowledge_pkl_path, data_dir, vector_db_dir, rerank_path
|
||||||
from langchain.embeddings import HuggingFaceBgeEmbeddings
|
from langchain.embeddings import HuggingFaceBgeEmbeddings
|
||||||
from langchain_community.document_loaders import DirectoryLoader, TextLoader, JSONLoader
|
from langchain_community.document_loaders import DirectoryLoader, TextLoader, JSONLoader
|
||||||
from langchain_text_splitters import CharacterTextSplitter, RecursiveCharacterTextSplitter, RecursiveJsonSplitter
|
from langchain_text_splitters import CharacterTextSplitter, RecursiveCharacterTextSplitter, RecursiveJsonSplitter
|
||||||
from BCEmbedding import EmbeddingModel, RerankerModel
|
from BCEmbedding import EmbeddingModel, RerankerModel
|
||||||
# from util.pipeline import EmoLLMRAG
|
# from util.pipeline import EmoLLMRAG
|
||||||
from transformers import AutoTokenizer, AutoModelForCausalLM
|
from transformers import AutoTokenizer, AutoModelForCausalLM
|
||||||
from langchain.document_loaders.pdf import PyPDFDirectoryLoader
|
from langchain.document_loaders.pdf import PyPDFDirectoryLoader
|
||||||
from langchain.document_loaders import UnstructuredFileLoader,DirectoryLoader
|
from langchain.document_loaders import UnstructuredFileLoader,DirectoryLoader
|
||||||
from langchain_community.llms import Cohere
|
from langchain_community.llms import Cohere
|
||||||
from langchain.retrievers import ContextualCompressionRetriever
|
from langchain.retrievers import ContextualCompressionRetriever
|
||||||
from langchain.retrievers.document_compressors import FlashrankRerank
|
from langchain.retrievers.document_compressors import FlashrankRerank
|
||||||
from langchain_core.documents.base import Document
|
from langchain_core.documents.base import Document
|
||||||
from FlagEmbedding import FlagReranker
|
from FlagEmbedding import FlagReranker
|
||||||
|
|
||||||
class Data_process():
|
class Data_process():
|
||||||
def __init__(self):
|
def __init__(self):
|
||||||
self.vector_db_dir = vector_db_dir
|
self.chunk_size: int=1000
|
||||||
self.doc_dir = doc_dir
|
self.chunk_overlap: int=100
|
||||||
self.qa_dir = qa_dir
|
|
||||||
self.knowledge_pkl_path = knowledge_pkl_path
|
def load_embedding_model(self, model_name='BAAI/bge-small-zh-v1.5', device='cpu', normalize_embeddings=True):
|
||||||
self.chunk_size: int=1000
|
"""
|
||||||
self.chunk_overlap: int=100
|
加载嵌入模型。
|
||||||
|
|
||||||
def load_embedding_model(self, model_name="BAAI/bge-small-zh-v1.5", device='cpu', normalize_embeddings=True):
|
参数:
|
||||||
"""
|
- model_name: 模型名称，字符串类型，默认为"BAAI/bge-small-zh-v1.5"。
|
||||||
加载嵌入模型。
|
- device: 指定模型加载的设备，'cpu' 或 'cuda'，默认为'cpu'。
|
||||||
|
- normalize_embeddings: 是否标准化嵌入向量，布尔类型，默认为 True。
|
||||||
参数:
|
"""
|
||||||
- model_name: 模型名称，字符串类型，默认为"BAAI/bge-small-zh-v1.5"。
|
if not os.path.exists(embedding_path):
|
||||||
- device: 指定模型加载的设备，'cpu' 或 'cuda'，默认为'cpu'。
|
os.makedirs(embedding_path, exist_ok=True)
|
||||||
- normalize_embeddings: 是否标准化嵌入向量，布尔类型，默认为 True。
|
embedding_model_path = os.path.join(embedding_path,model_name.split('/')[1] + '.pkl')
|
||||||
"""
|
logger.info('Loading embedding model...')
|
||||||
logger.info('Loading embedding model...')
|
if os.path.exists(embedding_model_path):
|
||||||
try:
|
try:
|
||||||
embeddings = HuggingFaceBgeEmbeddings(
|
with open(embedding_model_path , 'rb') as f:
|
||||||
model_name=model_name,
|
embeddings = pickle.load(f)
|
||||||
model_kwargs={'device': device},
|
logger.info('Embedding model loaded.')
|
||||||
encode_kwargs={'normalize_embeddings': normalize_embeddings}
|
return embeddings
|
||||||
)
|
except Exception as e:
|
||||||
except Exception as e:
|
logger.error(f'Failed to load embedding model from {embedding_model_path}')
|
||||||
logger.error(f'Failed to load embedding model: {e}')
|
try:
|
||||||
return None
|
embeddings = HuggingFaceBgeEmbeddings(
|
||||||
|
model_name=model_name,
|
||||||
logger.info('Embedding model loaded.')
|
model_kwargs={'device': device},
|
||||||
return embeddings
|
encode_kwargs={'normalize_embeddings': normalize_embeddings})
|
||||||
|
logger.info('Embedding model loaded.')
|
||||||
def load_rerank_model(self, model_name='BAAI/bge-reranker-large'):
|
with open(embedding_model_path, 'wb') as file:
|
||||||
"""
|
pickle.dump(embeddings, file)
|
||||||
加载重排名模型。
|
except Exception as e:
|
||||||
|
logger.error(f'Failed to load embedding model: {e}')
|
||||||
参数:
|
return None
|
||||||
- model_name (str): 模型的名称。默认为 'BAAI/bge-reranker-large'。
|
return embeddings
|
||||||
|
|
||||||
返回:
|
def load_rerank_model(self, model_name='BAAI/bge-reranker-large'):
|
||||||
- FlagReranker 实例。
|
"""
|
||||||
|
加载重排名模型。
|
||||||
异常:
|
|
||||||
- ValueError: 如果模型名称不在批准的模型列表中。
|
参数:
|
||||||
- Exception: 如果模型加载过程中发生任何其他错误。
|
- model_name (str): 模型的名称。默认为 'BAAI/bge-reranker-large'。
|
||||||
"""
|
|
||||||
try:
|
返回:
|
||||||
reranker_model = FlagReranker(model_name, use_fp16=True)
|
- FlagReranker 实例。
|
||||||
except Exception as e:
|
|
||||||
logger.error(f'Failed to load rerank model: {e}')
|
异常:
|
||||||
raise
|
- ValueError: 如果模型名称不在批准的模型列表中。
|
||||||
|
- Exception: 如果模型加载过程中发生任何其他错误。
|
||||||
return reranker_model
|
|
||||||
|
"""
|
||||||
|
if not os.path.exists(rerank_path):
|
||||||
def extract_text_from_json(self, obj, content=None):
|
os.makedirs(rerank_path, exist_ok=True)
|
||||||
"""
|
rerank_model_path = os.path.join(rerank_path, model_name.split('/')[1] + '.pkl')
|
||||||
抽取json中的文本，用于向量库构建
|
logger.info('Loading rerank model...')
|
||||||
|
if os.path.exists(rerank_model_path):
|
||||||
参数:
|
try:
|
||||||
- obj: dict,list,str
|
with open(rerank_model_path , 'rb') as f:
|
||||||
- content: str
|
reranker_model = pickle.load(f)
|
||||||
|
logger.info('Rerank model loaded.')
|
||||||
返回:
|
return reranker_model
|
||||||
- content: str
|
except Exception as e:
|
||||||
"""
|
logger.error(f'Failed to load embedding model from {rerank_model_path}')
|
||||||
if isinstance(obj, dict):
|
try:
|
||||||
for key, value in obj.items():
|
reranker_model = FlagReranker(model_name, use_fp16=True)
|
||||||
try:
|
logger.info('Rerank model loaded.')
|
||||||
content = self.extract_text_from_json(value, content)
|
with open(rerank_model_path, 'wb') as file:
|
||||||
except Exception as e:
|
pickle.dump(reranker_model, file)
|
||||||
print(f"Error processing value: {e}")
|
except Exception as e:
|
||||||
elif isinstance(obj, list):
|
logger.error(f'Failed to load rerank model: {e}')
|
||||||
for index, item in enumerate(obj):
|
raise
|
||||||
try:
|
|
||||||
content = self.extract_text_from_json(item, content)
|
return reranker_model
|
||||||
except Exception as e:
|
|
||||||
print(f"Error processing item: {e}")
|
|
||||||
elif isinstance(obj, str):
|
def extract_text_from_json(self, obj, content=None):
|
||||||
content += obj
|
"""
|
||||||
return content
|
抽取json中的文本，用于向量库构建
|
||||||
|
|
||||||
|
参数:
|
||||||
def split_document(self, data_path, chunk_size=500, chunk_overlap=100):
|
- obj: dict,list,str
|
||||||
"""
|
- content: str
|
||||||
切分data_path文件夹下的所有txt文件
|
|
||||||
|
返回:
|
||||||
参数:
|
- content: str
|
||||||
- data_path: str
|
"""
|
||||||
- chunk_size: int
|
if isinstance(obj, dict):
|
||||||
- chunk_overlap: int
|
for key, value in obj.items():
|
||||||
|
try:
|
||||||
返回：
|
content = self.extract_text_from_json(value, content)
|
||||||
- split_docs: list
|
except Exception as e:
|
||||||
"""
|
print(f"Error processing value: {e}")
|
||||||
|
elif isinstance(obj, list):
|
||||||
|
for index, item in enumerate(obj):
|
||||||
# text_spliter = CharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
|
try:
|
||||||
text_spliter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
|
content = self.extract_text_from_json(item, content)
|
||||||
split_docs = []
|
except Exception as e:
|
||||||
logger.info(f'Loading txt files from {data_path}')
|
print(f"Error processing item: {e}")
|
||||||
if os.path.isdir(data_path):
|
elif isinstance(obj, str):
|
||||||
loader = DirectoryLoader(data_path, glob="**/*.txt",show_progress=True)
|
content += obj
|
||||||
docs = loader.load()
|
return content
|
||||||
split_docs = text_spliter.split_documents(docs)
|
|
||||||
elif data_path.endswith('.txt'):
|
|
||||||
file_path = data_path
|
def split_document(self, data_path, chunk_size=500, chunk_overlap=100):
|
||||||
logger.info(f'splitting file {file_path}')
|
"""
|
||||||
text_loader = TextLoader(file_path, encoding='utf-8')
|
切分data_path文件夹下的所有txt文件
|
||||||
text = text_loader.load()
|
|
||||||
splits = text_spliter.split_documents(text)
|
参数:
|
||||||
split_docs = splits
|
- data_path: str
|
||||||
logger.info(f'split_docs size {len(split_docs)}')
|
- chunk_size: int
|
||||||
return split_docs
|
- chunk_overlap: int
|
||||||
|
|
||||||
|
返回：
|
||||||
def split_conversation(self, path):
|
- split_docs: list
|
||||||
"""
|
"""
|
||||||
按conversation块切分path文件夹下的所有json文件
|
|
||||||
##TODO 限制序列长度
|
|
||||||
"""
|
# text_spliter = CharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
|
||||||
# json_spliter = RecursiveJsonSplitter(max_chunk_size=500)
|
text_spliter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
|
||||||
logger.info(f'Loading json files from {path}')
|
split_docs = []
|
||||||
split_qa = []
|
logger.info(f'Loading txt files from {data_path}')
|
||||||
if os.path.isdir(path):
|
if os.path.isdir(data_path):
|
||||||
# loader = DirectoryLoader(path, glob="**/*.json",show_progress=True)
|
loader = DirectoryLoader(data_path, glob="**/*.txt",show_progress=True)
|
||||||
# jsons = loader.load()
|
docs = loader.load()
|
||||||
|
split_docs = text_spliter.split_documents(docs)
|
||||||
for root, dirs, files in os.walk(path):
|
elif data_path.endswith('.txt'):
|
||||||
for file in files:
|
file_path = data_path
|
||||||
if file.endswith('.json'):
|
logger.info(f'splitting file {file_path}')
|
||||||
file_path = os.path.join(root, file)
|
text_loader = TextLoader(file_path, encoding='utf-8')
|
||||||
logger.info(f'splitting file {file_path}')
|
text = text_loader.load()
|
||||||
with open(file_path, 'r', encoding='utf-8') as f:
|
splits = text_spliter.split_documents(text)
|
||||||
data = json.load(f)
|
split_docs = splits
|
||||||
# print(data)
|
logger.info(f'split_docs size {len(split_docs)}')
|
||||||
for conversation in data:
|
return split_docs
|
||||||
# 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,'')
|
def split_conversation(self, path):
|
||||||
# split_qa.append(Document(page_content = content))
|
"""
|
||||||
#按conversation块切分
|
按conversation块切分path文件夹下的所有json文件
|
||||||
content = self.extract_text_from_json(conversation['conversation'], '')
|
##TODO 限制序列长度
|
||||||
logger.info(f'content====={content}')
|
"""
|
||||||
split_qa.append(Document(page_content = content))
|
# json_spliter = RecursiveJsonSplitter(max_chunk_size=500)
|
||||||
# logger.info(f'split_qa size====={len(split_qa)}')
|
logger.info(f'Loading json files from {path}')
|
||||||
return split_qa
|
split_qa = []
|
||||||
|
if os.path.isdir(path):
|
||||||
|
# loader = DirectoryLoader(path, glob="**/*.json",show_progress=True)
|
||||||
def load_knowledge(self, knowledge_pkl_path):
|
# jsons = loader.load()
|
||||||
'''
|
|
||||||
读取或创建知识.pkl
|
for root, dirs, files in os.walk(path):
|
||||||
'''
|
for file in files:
|
||||||
if not os.path.exists(knowledge_pkl_path):
|
if file.endswith('.json'):
|
||||||
split_doc = self.split_document(doc_dir)
|
file_path = os.path.join(root, file)
|
||||||
split_qa = self.split_conversation(qa_dir)
|
logger.info(f'splitting file {file_path}')
|
||||||
knowledge_chunks = split_doc + split_qa
|
with open(file_path, 'r', encoding='utf-8') as f:
|
||||||
with open(knowledge_pkl_path, 'wb') as file:
|
data = json.load(f)
|
||||||
pickle.dump(knowledge_chunks, file)
|
# print(data)
|
||||||
else:
|
for conversation in data:
|
||||||
with open(knowledge_pkl_path , 'rb') as f:
|
# for dialog in conversation['conversation']:
|
||||||
knowledge_chunks = pickle.load(f)
|
##按qa对切分,将每一轮qa转换为langchain_core.documents.base.Document
|
||||||
return knowledge_chunks
|
# content = self.extract_text_from_json(dialog,'')
|
||||||
|
# split_qa.append(Document(page_content = content))
|
||||||
|
#按conversation块切分
|
||||||
def create_vector_db(self, emb_model):
|
content = self.extract_text_from_json(conversation['conversation'], '')
|
||||||
'''
|
logger.info(f'content====={content}')
|
||||||
创建并保存向量库
|
split_qa.append(Document(page_content = content))
|
||||||
'''
|
# logger.info(f'split_qa size====={len(split_qa)}')
|
||||||
logger.info(f'Creating index...')
|
return split_qa
|
||||||
split_doc = self.split_document(self.doc_dir)
|
|
||||||
split_qa = self.split_conversation(self.qa_dir)
|
|
||||||
# logger.info(f'split_doc == {len(split_doc)}')
|
def load_knowledge(self, knowledge_pkl_path):
|
||||||
# logger.info(f'split_qa == {len(split_qa)}')
|
'''
|
||||||
# logger.info(f'split_doc type == {type(split_doc[0])}')
|
读取或创建知识.pkl
|
||||||
# logger.info(f'split_qa type== {type(split_qa[0])}')
|
'''
|
||||||
db = FAISS.from_documents(split_doc + split_qa, emb_model)
|
if not os.path.exists(knowledge_pkl_path):
|
||||||
db.save_local(vector_db_dir)
|
split_doc = self.split_document(doc_dir)
|
||||||
return db
|
split_qa = self.split_conversation(qa_dir)
|
||||||
|
knowledge_chunks = split_doc + split_qa
|
||||||
|
with open(knowledge_pkl_path, 'wb') as file:
|
||||||
def load_vector_db(self, knowledge_pkl_path=knowledge_pkl_path, doc_dir=doc_dir, qa_dir=qa_dir):
|
pickle.dump(knowledge_chunks, file)
|
||||||
'''
|
else:
|
||||||
读取向量库
|
with open(knowledge_pkl_path , 'rb') as f:
|
||||||
'''
|
knowledge_chunks = pickle.load(f)
|
||||||
# current_os = platform.system()
|
return knowledge_chunks
|
||||||
emb_model = self.load_embedding_model()
|
|
||||||
if not os.path.exists(vector_db_dir) or not os.listdir(vector_db_dir):
|
|
||||||
db = self.create_vector_db(emb_model)
|
def create_vector_db(self, emb_model):
|
||||||
else:
|
'''
|
||||||
db = FAISS.load_local(vector_db_dir, emb_model, allow_dangerous_deserialization=True)
|
创建并保存向量库
|
||||||
return db
|
'''
|
||||||
|
logger.info(f'Creating index...')
|
||||||
|
split_doc = self.split_document(doc_dir)
|
||||||
def retrieve(self, query, vector_db, k=5):
|
split_qa = self.split_conversation(qa_dir)
|
||||||
'''
|
# logger.info(f'split_doc == {len(split_doc)}')
|
||||||
基于query对向量库进行检索
|
# logger.info(f'split_qa == {len(split_qa)}')
|
||||||
'''
|
# logger.info(f'split_doc type == {type(split_doc[0])}')
|
||||||
retriever = vector_db.as_retriever(search_kwargs={"k": k})
|
# logger.info(f'split_qa type== {type(split_qa[0])}')
|
||||||
docs = retriever.invoke(query)
|
db = FAISS.from_documents(split_doc + split_qa, emb_model)
|
||||||
return docs, retriever
|
db.save_local(vector_db_dir)
|
||||||
|
return db
|
||||||
##FlashrankRerank效果一般
|
|
||||||
# def rerank(self, query, retriever):
|
|
||||||
# compressor = FlashrankRerank()
|
def load_vector_db(self, knowledge_pkl_path=knowledge_pkl_path, doc_dir=doc_dir, qa_dir=qa_dir):
|
||||||
# compression_retriever = ContextualCompressionRetriever(base_compressor=compressor, base_retriever=retriever)
|
'''
|
||||||
# compressed_docs = compression_retriever.get_relevant_documents(query)
|
读取向量库
|
||||||
# return compressed_docs
|
'''
|
||||||
|
# current_os = platform.system()
|
||||||
def rerank(self, query, docs):
|
emb_model = self.load_embedding_model()
|
||||||
reranker = self.load_rerank_model()
|
if not os.path.exists(vector_db_dir) or not os.listdir(vector_db_dir):
|
||||||
passages = []
|
db = self.create_vector_db(emb_model)
|
||||||
for doc in docs:
|
else:
|
||||||
passages.append(str(doc.page_content))
|
db = FAISS.load_local(vector_db_dir, emb_model, allow_dangerous_deserialization=True)
|
||||||
scores = reranker.compute_score([[query, passage] for passage in passages])
|
return db
|
||||||
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 retrieve(self, query, vector_db, k=5):
|
||||||
|
'''
|
||||||
|
基于query对向量库进行检索
|
||||||
# def create_prompt(question, context):
|
'''
|
||||||
# from langchain.prompts import PromptTemplate
|
retriever = vector_db.as_retriever(search_kwargs={"k": k})
|
||||||
# prompt_template = f"""请基于以下内容回答问题：
|
docs = retriever.invoke(query)
|
||||||
|
return docs, retriever
|
||||||
# {context}
|
|
||||||
|
##FlashrankRerank效果一般
|
||||||
# 问题: {question}
|
# def rerank(self, query, retriever):
|
||||||
# 回答:"""
|
# compressor = FlashrankRerank()
|
||||||
# prompt = PromptTemplate(
|
# compression_retriever = ContextualCompressionRetriever(base_compressor=compressor, base_retriever=retriever)
|
||||||
# template=prompt_template, input_variables=["context", "question"]
|
# compressed_docs = compression_retriever.get_relevant_documents(query)
|
||||||
# )
|
# return compressed_docs
|
||||||
# logger.info(f'Prompt: {prompt}')
|
|
||||||
# return prompt
|
def rerank(self, query, docs):
|
||||||
|
reranker = self.load_rerank_model()
|
||||||
def create_prompt(question, context):
|
passages = []
|
||||||
prompt = f"""请基于以下内容: {context} 给出问题答案。问题如下: {question}。回答:"""
|
for doc in docs:
|
||||||
logger.info(f'Prompt: {prompt}')
|
passages.append(str(doc.page_content))
|
||||||
return prompt
|
scores = reranker.compute_score([[query, passage] for passage in passages])
|
||||||
|
sorted_pairs = sorted(zip(passages, scores), key=lambda x: x[1], reverse=True)
|
||||||
def test_zhipu(prompt):
|
sorted_passages, sorted_scores = zip(*sorted_pairs)
|
||||||
from zhipuai import ZhipuAI
|
return sorted_passages, sorted_scores
|
||||||
api_key = "" # 填写您自己的APIKey
|
|
||||||
if api_key == "":
|
|
||||||
raise ValueError("请填写api_key")
|
# def create_prompt(question, context):
|
||||||
client = ZhipuAI(api_key=api_key)
|
# from langchain.prompts import PromptTemplate
|
||||||
response = client.chat.completions.create(
|
# prompt_template = f"""请基于以下内容回答问题：
|
||||||
model="glm-4", # 填写需要调用的模型名称
|
|
||||||
messages=[
|
# {context}
|
||||||
{"role": "user", "content": prompt[:100]}
|
|
||||||
],
|
# 问题: {question}
|
||||||
)
|
# 回答:"""
|
||||||
print(response.choices[0].message)
|
# prompt = PromptTemplate(
|
||||||
|
# template=prompt_template, input_variables=["context", "question"]
|
||||||
if __name__ == "__main__":
|
# )
|
||||||
logger.info(data_dir)
|
# logger.info(f'Prompt: {prompt}')
|
||||||
if not os.path.exists(data_dir):
|
# return prompt
|
||||||
os.mkdir(data_dir)
|
|
||||||
dp = Data_process()
|
def create_prompt(question, context):
|
||||||
# faiss_index, knowledge_chunks = dp.load_index_and_knowledge(knowledge_pkl_path='')
|
prompt = f"""请基于以下内容: {context} 给出问题答案。问题如下: {question}。回答:"""
|
||||||
vector_db = dp.load_vector_db()
|
logger.info(f'Prompt: {prompt}')
|
||||||
# 按照query进行查询
|
return prompt
|
||||||
# query = "儿童心理学说明-内容提要-目录 《儿童心理学》1993年修订版说明 《儿童心理学》是1961年初全国高等学校文科教材会议指定朱智贤教授编 写的。1962年初版，1979年再版。"
|
|
||||||
# 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，我现在感到非常孤独、累和迷茫。您能给我提供一些建议吗？"
|
def test_zhipu(prompt):
|
||||||
# query = "这在一定程度上限制了其思维能力，特别是辩证 逻辑思维能力的发展。随着年龄的增长，初中三年级学生逐步克服了依赖性"
|
from zhipuai import ZhipuAI
|
||||||
# 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"
|
api_key = "" # 填写您自己的APIKey
|
||||||
query = "我现在心情非常差，有什么解决办法吗？"
|
if api_key == "":
|
||||||
docs, retriever = dp.retrieve(query, vector_db, k=10)
|
raise ValueError("请填写api_key")
|
||||||
logger.info(f'Query: {query}')
|
client = ZhipuAI(api_key=api_key)
|
||||||
logger.info("Retrieve results:")
|
response = client.chat.completions.create(
|
||||||
for i, doc in enumerate(docs):
|
model="glm-4", # 填写需要调用的模型名称
|
||||||
logger.info(str(i) + '\n')
|
messages=[
|
||||||
logger.info(doc)
|
{"role": "user", "content": prompt[:100]}
|
||||||
# print(f'get num of docs:{len(docs)}')
|
],
|
||||||
# print(docs)
|
)
|
||||||
passages,scores = dp.rerank(query, docs)
|
print(response.choices[0].message)
|
||||||
logger.info("After reranking...")
|
|
||||||
for i in range(len(scores)):
|
if __name__ == "__main__":
|
||||||
logger.info(str(scores[i]) + '\n')
|
logger.info(data_dir)
|
||||||
logger.info(passages[i])
|
if not os.path.exists(data_dir):
|
||||||
prompt = create_prompt(query, passages[0])
|
os.mkdir(data_dir)
|
||||||
|
dp = Data_process()
|
||||||
|
# faiss_index, knowledge_chunks = dp.load_index_and_knowledge(knowledge_pkl_path='')
|
||||||
|
vector_db = dp.load_vector_db()
|
||||||
|
# 按照query进行查询
|
||||||
|
# query = "儿童心理学说明-内容提要-目录 《儿童心理学》1993年修订版说明 《儿童心理学》是1961年初全国高等学校文科教材会议指定朱智贤教授编 写的。1962年初版，1979年再版。"
|
||||||
|
# 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，我现在感到非常孤独、累和迷茫。您能给我提供一些建议吗？"
|
||||||
|
# query = "这在一定程度上限制了其思维能力，特别是辩证 逻辑思维能力的发展。随着年龄的增长，初中三年级学生逐步克服了依赖性"
|
||||||
|
# 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"
|
||||||
|
# query = "我现在心情非常差，有什么解决办法吗？"
|
||||||
|
query = "我最近总感觉胸口很闷，但医生检查过说身体没问题。可我就是觉得喘不过气来，尤其是看到那些旧照片，想起过去的日子"
|
||||||
|
docs, retriever = dp.retrieve(query, vector_db, k=10)
|
||||||
|
logger.info(f'Query: {query}')
|
||||||
|
logger.info("Retrieve results:")
|
||||||
|
for i, doc in enumerate(docs):
|
||||||
|
logger.info(str(i) + '\n')
|
||||||
|
logger.info(doc)
|
||||||
|
# print(f'get num of docs:{len(docs)}')
|
||||||
|
# print(docs)
|
||||||
|
passages,scores = dp.rerank(query, docs)
|
||||||
|
logger.info("After reranking...")
|
||||||
|
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
|
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
|
||||||
Loading…
Reference in New Issue
Block a user