异步使用langchain

一.先利用langchain官方文档的AI功能问问

• 然后看他给的 Verified Sources
  
• 这个页面里面虽然有些函数是异步函数，但是并非专门讲解异步的

二.langchain async api

还不如直接谷歌搜 一下搜到， 上面那个AI文档问答没给出这个链接

• 官方示例

  import asyncio
  import time
  
  from langchain.llms import OpenAI
  from langchain.prompts import PromptTemplate
  from langchain.chains import LLMChain
  
  
  def generate_serially():
      llm = OpenAI(temperature=0.9)
      prompt = PromptTemplate(
          input_variables=["product"],
          template="What is a good name for a company that makes {product}?",
      )
      chain = LLMChain(llm=llm, prompt=prompt)
      for _ in range(5):
          resp = chain.run(product="toothpaste")
          print(resp)
  
  
  async def async_generate(chain):
      resp = await chain.arun(product="toothpaste")
      print(resp)
  
  
  async def generate_concurrently():
      llm = OpenAI(temperature=0.9)
      prompt = PromptTemplate(
          input_variables=["product"],
          template="What is a good name for a company that makes {product}?",
      )
      chain = LLMChain(llm=llm, prompt=prompt)
      tasks = [async_generate(chain) for _ in range(5)]
      await asyncio.gather(*tasks)
  
  
  s = time.perf_counter()
  # If running this outside of Jupyter, use asyncio.run(generate_concurrently())
  await generate_concurrently()
  elapsed = time.perf_counter() - s
  print("\033[1m" + f"Concurrent executed in {elapsed:0.2f} seconds." + "\033[0m")
  
  s = time.perf_counter()
  generate_serially()
  elapsed = time.perf_counter() - s
  print("\033[1m" + f"Serial executed in {elapsed:0.2f} seconds." + "\033[0m")
  

• 不过官方代码报错了
  

• 我让copilot修改一下，能跑了

  import time
  import asyncio
  from langchain.llms import OpenAI
  from langchain.prompts import PromptTemplate
  from langchain.chains import LLMChain
  
  
  def generate_serially():
      llm = OpenAI(temperature=0.9)
      prompt = PromptTemplate(
          input_variables=["product"],
          template="What is a good name for a company that makes {product}?",
      )
      chain = LLMChain(llm=llm, prompt=prompt)
      for _ in range(5):
          resp = chain.run(product="toothpaste")
          print(resp)
  
  
  async def async_generate(chain):
      resp = await chain.arun(product="toothpaste")
      print(resp)
  
  
  async def generate_concurrently():
      llm = OpenAI(temperature=0.9)
      prompt = PromptTemplate(
          input_variables=["product"],
          template="What is a good name for a company that makes {product}?",
      )
      chain = LLMChain(llm=llm, prompt=prompt)
      tasks = [async_generate(chain) for _ in range(5)]
      await asyncio.gather(*tasks)
  
  
  async def main():
      s = time.perf_counter()
      await generate_concurrently()
      elapsed = time.perf_counter() - s
      print("\033[1m" + f"Concurrent executed in {elapsed:0.2f} seconds." + "\033[0m")
  
      s = time.perf_counter()
      generate_serially()
      elapsed = time.perf_counter() - s
      print("\033[1m" + f"Serial executed in {elapsed:0.2f} seconds." + "\033[0m")
  
  
  asyncio.run(main())
  	
  

  

• 这还有一篇官方blog
  
  

三.串行，异步速度比较

• 先学习一下掘金上看到的一篇：https://juejin.cn/post/7231907374688436284
• 为了更方便的看到异步效果，我在原博主的基础上，print里面加了一个提示
  
  

# 引入time和asyncio模块
import time
import asyncio
# 引入OpenAI类
from langchain.llms import OpenAI


# 定义异步函数async_generate，该函数接收一个llm参数和一个name参数
async def async_generate(llm, name):
    # 调用OpenAI类的agenerate方法，传入字符串列表["Hello, how are you?"]并等待响应
    resp = await llm.agenerate(["Hello, how are you?"])
    # 打印响应结果的生成文本和函数名
    print(f"{name}: {resp.generations[0][0].text}")


# 定义异步函数generate_concurrently
async def generate_concurrently():
    # 创建OpenAI实例，并设置temperature参数为0.9
    llm = OpenAI(temperature=0.9)
    # 创建包含10个async_generate任务的列表
    tasks = [async_generate(llm, f"Function {i}") for i in range(10)]
    # 并发执行任务
    await asyncio.gather(*tasks)


# 主函数
# 如果在Jupyter Notebook环境运行该代码，则无需手动调用await generate_concurrently()，直接在下方执行单元格即可执行该函数
# 如果在命令行或其他环境下运行该代码，则需要手动调用asyncio.run(generate_concurrently())来执行该函数
asyncio.run(generate_concurrently())

免费用户一分钟只能3次，实在是有点难蚌

• 整合一下博主的代码，对两个速度进行比较，但是这个调用限制真的很搞人啊啊啊

  import time
  import asyncio
  from langchain.llms import OpenAI
  
  
  async def async_generate(llm, name):
      resp = await llm.agenerate(["Hello, how are you?"])
      # print(f"{name}: {resp.generations[0][0].text}")
  
  
  async def generate_concurrently():
      llm = OpenAI(temperature=0.9)
      tasks = [async_generate(llm, f"Function {i}") for i in range(3)]
      await asyncio.gather(*tasks)
  
  
  def generate_serially():
      llm = OpenAI(temperature=0.9)
      for _ in range(3):
          resp = llm.generate(["Hello, how are you?"])
          # print(resp.generations[0][0].text)
  
  
  async def main():
      s = time.perf_counter()
      await generate_concurrently()
      elapsed = time.perf_counter() - s
      print("\033[1m" + f"Concurrent executed in {elapsed:0.2f} seconds." + "\033[0m")
  
      s = time.perf_counter()
      generate_serially()
      elapsed = time.perf_counter() - s
      print("\033[1m" + f"Serial executed in {elapsed:0.2f} seconds." + "\033[0m")
  
  
  asyncio.run(main())
  

  
  

• 再看一篇blog
  • 作者将代码开源在这里了：https://github.com/gabrielcassimiro17/async-langchain
  • 测试一下它的async_chain.py文件
    
• 读取csv的时候路径一直报错，还好不久前总结了一篇blog：Python中如何获取各种目录路径

  • 直接获取当前脚本路径了

    import os
    import pandas as pd
    
    # Get the directory where the script is located
    script_directory = os.path.dirname(os.path.abspath(__file__))
    
    # Construct the path to the CSV file
    csv_path = os.path.join(script_directory, 'wine_subset.csv')
    
    # Read the CSV file
    df = pd.read_csv(csv_path)
    

    • sequential_run.py 就不跑了… 一天200次调用都快没了
• 主要是看看两者区别