#import gradio as gr #import cv2 #def to_black(image): # output = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) # return output #interface = gr.Interface(fn=to_black, inputs="image", outputs="image") #print('here') #interface.launch() #print(share_url) #print(local_url) #print(app) #interface.launch(inbrowser =True, share=True, port=8888) #url = interface.share() #print(url) from langchain.chains import RetrievalQA from langchain.document_loaders import UnstructuredFileLoader, TextLoader, CSVLoader from langchain.document_loaders import CSVLoader from langchain.document_loaders import TextLoader from langchain.vectorstores import DocArrayInMemorySearch from langchain.indexes import VectorstoreIndexCreator from langchain.prompts import PromptTemplate from langchain.embeddings.huggingface import HuggingFaceEmbeddings from langchain import HuggingFacePipeline import torch from langchain.vectorstores import FAISS from langchain.text_splitter import RecursiveCharacterTextSplitter from langchain.chains.base import Chain from langchain.chains import ConversationalRetrievalChain from langchain.chains.summarize import load_summarize_chain import gradio as gr from typing import List from tqdm import tqdm import logging import argparse import os import string CHUNK_SIZE=600 CHUNK_OVERLAP = 100 SEARCH_TOP_K = 5 logger = logging.getLogger("bio_LLM_logger") def tree(filepath, ignore_dir_names=None, ignore_file_names=None): """返回两个列表，第一个列表为 filepath 下全部文件的完整路径, 第二个为对应的文件名""" if ignore_dir_names is None: ignore_dir_names = [] if ignore_file_names is None: ignore_file_names = [] ret_list = [] if isinstance(filepath, str): if not os.path.exists(filepath): print("路径不存在") return None, None elif os.path.isfile(filepath) and os.path.basename(filepath) not in ignore_file_names: return [filepath], [os.path.basename(filepath)] elif os.path.isdir(filepath) and os.path.basename(filepath) not in ignore_dir_names: for file in os.listdir(filepath): fullfilepath = os.path.join(filepath, file) if os.path.isfile(fullfilepath) and os.path.basename(fullfilepath) not in ignore_file_names: ret_list.append(fullfilepath) if os.path.isdir(fullfilepath) and os.path.basename(fullfilepath) not in ignore_dir_names: ret_list.extend(tree(fullfilepath, ignore_dir_names, ignore_file_names)[0]) return ret_list, [os.path.basename(p) for p in ret_list] def load_file(file_path, chunk_size=CHUNK_SIZE, chunk_overlap=CHUNK_OVERLAP): if file_path.lower().endswith(".pdf"): loader = UnstructuredFileLoader(file_path, mode="elements") text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap= chunk_overlap) docs = loader.load_and_split(text_splitter=text_splitter) elif file_path.lower().endswith(".txt"): loader = TextLoader(file_path, autodetect_encoding=True) text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap= chunk_overlap) docs = loader.load_and_split(text_splitter=text_splitter) elif file_path.lower().endswith(".csv"): loader = CSVLoader(file_path) text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap= chunk_overlap) docs = loader.load_and_split(text_splitter=text_splitter) else: print("unsupported the file format") return docs #class summary_chain: # def init_cfg(self, # llm_model: Chain, def summary(model, chain_type, PROMPT, REFINE_PROMPT,docs): if chain_type == "stuff": chain = load_summarize_chain(model, chain_type="stuff", prompt=PROMPT) elif chain_type == "refine": chain = load_summarize_chain(model, chain_type="refine", question_prompt=PROMPT, refine_prompt=REFINE_PROMPT) print(chain.run(docs)) class QA_Localdb: llm_model_chain: Chain = None embeddings: object = None top_k: int = SEARCH_TOP_K chunk_size: int = CHUNK_SIZE def init_cfg(self, llm_model: Chain, embedding_model: str, #embedding_device: str, top_k = SEARCH_TOP_K, ): self.llm_model_chain = llm_model self.embeddings = HuggingFaceEmbeddings(model_name = embedding_model) self.top_k = top_k def init_knowledge_vector_store(self, file_path: str or List[str], vectorstore_path: str or os.PathLike = None, ): loaded_files = [] failed_files = [] if isinstance(file_path, str): if not os.path.exists(file_path): print("unknown path") return None elif os.path.isfile(file_path): file = os.path.split(file_path)[-1] try: docs = load_file(file_path) logger.info(f"{file} sucessful loaded") loaded_files.append(file_path) except Exception as e: logger.error(e) logger.info(f"{file} unsucessful loaded") return None elif os.path.isdir(file_path): docs=[] for fullfilepath, file in tqdm(zip(*tree(file_path, ignore_dir_names=['tmp_files'])), desc="load file"): try: docs += load_file(fullfilepath) loaded_files.append(fullfilepath) except Exception as e: logger.error(e) failed_files.append(file) if len(failed_files) > 0: logger.info('unloaded files are as follows') for file in failed_files: logger.info(f"{file}\n") else: docs = [] for file in file_path: try: docs += load_file(file) logger.info(f"{file} sucessful loaded") loaded_files.append(file) except Exception as e: logger.error(e) logger.info(f"{file} unsucessful loaded") if len(docs) > 0: logger.info("sucessful loaded, generating vector store") if vectorstore_path and os.path.isdir(vectorstore_path) and "index.faiss" in os.listdir(vectorstore_path): print("temp") # vector_store = load_vector_store(vectorstore_path, self.embeddings) # vector_store.add_documents(docs) # torch_gc() else: if not vectorstore_path: vectorstore_path = "" vector_store = FAISS.from_documents(docs, self.embeddings) #vector_store.save_local(vectorstore_path) return vector_store, loaded_files else: logger.info("file load failed") ''' def delete_file_from_vector_store(self, filepath: str or List[str], vs_path): vector_store = load_vector_store(vs_path, self.embeddings) status = vector_store.delete_doc(filepath) return status def update_file_from_vector_store(self, filepath: str or List[str], vs_path, docs: List[Document], ): vector_store = load_vector_store(vs_path, self.embeddings) status = vector_store.update_doc(filepath, docs) return status def list_file_from_vector_store(self, vs_path, fullpath=False): vector_store = load_vector_store(vs_path, self.embeddings) docs = vector_store.list_docs() if fullpath: return docs else: return [os.path.split(doc)[-1] for doc in docs] ''' def QA_model(): # file_path = "/mnt/petrelfs/lvying/LLM/BoMA/data/test/OPUS-DSD.pdf" file_path = "doc1.txt" # file_path = "/mnt/petrelfs/lvying/LLM/BoMA/data/test/Interageting-Prior-into-DA.pdf" # file_path = "/mnt/petrelfs/lvying/LLM/BoMA/data/test/" model_path = "/mnt/petrelfs/lvying/LLM/BoMA/models/LLM/Llama-2-13b-chat-hf" embedding_path = "/mnt/petrelfs/lvying/LLM/BoMA/text2vec/instructor-xl/" model = HuggingFacePipeline.from_model_id(model_id="daryl149/llama-2-7b-chat-hf", task="text-generation", model_kwargs={ "torch_dtype" : torch.float32, "low_cpu_mem_usage" :True, "temperature": 0.2, "max_length": 2048, # "device_map": "auto", "repetition_penalty":1.1} ) print(model.model_id) QA = QA_Localdb() QA.init_cfg(llm_model=model, embedding_model = "sentence-transformers/paraphrase-MiniLM-L6-v2") vector_store, _ =QA.init_knowledge_vector_store(file_path) retriever = vector_store.as_retriever(search_kwargs={"k": 3}) print("loading LLM...") prompt_template = ("Below is an instruction that describes a task. " "Write a response that appropriately completes the request.\n\n" "### Instruction:\n{context}\n{question}\n\n### Response: ") PROMPT = PromptTemplate( template=prompt_template, input_variables=["context", "question"] ) chain_type_kwargs = {"prompt": PROMPT} #print(chain_type_kwargs) ''' qa_stuff = RetrievalQA.from_chain_type( llm = model, chain_type="stuff", retriever = retriever, chain_type_kwargs = chain_type_kwargs, # verbose = True ) while True: print("Input Qusetion:") query = input() if len(query.strip())==0: break print(qa_stuff.run(query)) ''' ''' qa = ConversationalRetrievalChain.from_llm( llm = QA.llm_model_chain, chain_type="stuff", retriever = retriever, combine_docs_chain_kwargs = chain_type_kwargs, # verbose = True ) ''' qa = RetrievalQA.from_chain_type( llm = QA.llm_model_chain, chain_type="stuff", retriever = retriever, chain_type_kwargs = chain_type_kwargs, # verbose = True ) return qa qa_temp = QA_model() def temp(query): return qa_temp.run(query) def answer_question(query): print(query) chat_history = [] threshold_history = 10 # Remembered historical conversations i = 0 if i>threshold_history: chat_history = [] print("Send a Message:") #query = context #if len(query.strip())==0: # break result = qa_temp({"question":query, "chat_history": chat_history}) print(type(result["answer"])) chat_history.append((query, result["answer"])) i = i + 1 resp = result["answer"] return str(resp) iface = gr.Interface( fn = temp, inputs="text", outputs="text",) #title="问答界面", #description="输入问题和相关文本，得到问题的答案。", #article="这里是相关的文本。可以输入一些段落或者问题的背景。", #examples=[ # ["Gradio是什么？", "Gradio是一个用于构建和部署机器学习模型的开源库。"], # ["Python的创始人是谁？", "Python的创始人是Guido van Rossum。"] #]) #print(iface.launch(share=True)) #print("======Finish======") #share_url = iface.share() #print(share_url) iface.launch()