前言

赶上ai的大潮,尝试做个问答式文档库

项目

使用OpenAI训练一个简单的文档库,这里用王者荣耀早期公开的数据进行测试.

安装环境

!pip install gpt_index
!pip install langchain

引入包

from gpt_index import SimpleDirectoryReader, GPTListIndex, GPTSimpleVectorIndex, LLMPredictor, PromptHelper
from langchain import OpenAI
import sys
#from google.colab import drive
import os

设置环境变量(key)

os.environ["OPENAI_API_KEY"] = 'Your OpenAI Key 你的OpenAI Key'

声明方法

将文件夹内的文件加入训练数据集合并输出训练好的数据

def construct_index(directory_path):
  # set maximum input size
  max_input_size = 4096
  # set number of output tokens
  num_outputs = 256
  # set maximum chunk overlap
  max_chunk_overlap = 20
  # set chunk size limit
  chunk_size_limit = 600

  prompt_helper = PromptHelper(max_input_size, num_outputs, max_chunk_overlap, chunk_size_limit=chunk_size_limit)

  # define LLM  text-davinci-003
  llm_predictor = LLMPredictor(llm=OpenAI(temperature=0, model_name="text-ada-001", max_tokens=num_outputs))
  
  documents = SimpleDirectoryReader(directory_path).load_data()
  
  index = GPTSimpleVectorIndex(documents, llm_predictor=llm_predictor, prompt_helper=prompt_helper)
  
  index.save_to_disk('index.json')
  
  return index

带训练数据集合的提问

def ask_bot(input_index = 'index.json'):
  index = GPTSimpleVectorIndex.load_from_disk(input_index)
  while True:
    query = input('What do you want to ask the bot?   \n')
    response = index.query(query, response_mode="compact")
    print ("\nBot says: \n\n" + response.response + "\n\n\n")

生成训练数据

调用方法,将content文件夹内的所有txt都加入训练

index = construct_index("/content/")

使用训练好的数据集进行提问

ask_bot('index.json')

运行效果

鸣谢

1littlecoder