openAI大会（openAI API简易使用教程）

准备

创建openAI 账号（https://platform.openai.com/overview）  
  
  
  
  
  ，右上角personal 
  
  
  
  
  
，创建API key           。

2. 安装包

pip install openai pip install --upgrade tiktoken

tiktoken 是用来计算每次查询时的token数


 


 


 


 


 

，因为openAI是根据token数计费 
 
 
 
 
 ，不是必须安装 
 
  
 
  
 
  
 
  
 
  
 。

API调用

api key 可以直接明文写在代码中  
   
   
   
   
   
，也可以通过环境变量方式获取

import os import openai # OPENAI_API_KEY是自己设定的环境变量名 openai.api_key = os.getenv("OPENAI_API_KEY") # 明文 openai.api_key = *************

openAI提供了几种不同场景的模型

 

 

 

 

 
，主要有text completion  
  
  
  
  
  、code completion 
  
  
  
  
  
、chat completion


 


 


 


 


 

、image completion
 

 

 

 

 

 ，例如chat completion           ，则调用方式为

openai.ChatCompletion.create( model="gpt-3.5-turbo", messages=[ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Who won the world series in 2020?"}, {"role": "assistant", "content": "The Los Angeles Dodgers won the World Series in 2020."}, {"role": "user", "content": "Where was it played?"} ] )

其中

model 是具体的模型
 

 
 

 
 

 
 

 
 

 
 
，gpt-3.5-turbo是openAI最先进的语言模型










，当然也可以用其他模型










。

role                 ，三种固定值
 
 
 
 
 
 
 
 
 
 
 
，

system：类似一种前提















，表示后续的对话以此情景为基础；

user：提问者；

assistant：当对话需要结合上下文时            ，通过它让模型知道之前的对话内容
  
  
  
  
  
 。

content是具体的对话内容

发起一次请求到响应会存在几秒钟的延迟
 
  
 
  
 
  
 
  
 
  
 
，response 格式如下所示

{ id: chatcmpl-6p9XYPYSTTRi0xEviKjjilqrWU2Ve, object: chat.completion, created: 1677649420, model: gpt-3.5-turbo, usage: {prompt_tokens: 56, completion_tokens: 31, total_tokens: 87}, choices: [ { message: { role: assistant, content: The 2020 World Series was played in Arlington, Texas at the Globe Life Field, which was the new home stadium for the Texas Rangers.}, finish_reason: stop, index: 0 } ] }

提取回复内容response[choices][0][message][content]

每次response中不同finish-reason值代表不同状态：

stop：API返回完整内容 length：由于max_token限制










，回答不完整 
  
  
  
  
  。 content_filter：回复被过滤 null：API还在思考答案

计算token数量

openAI的gpt-3.5-turbo-0301模型token最多4096  
  
  
  
  
  ，超过限制只能缩短请求内容 


 


 


 


 


 

。

而且请求的token和回复的token数会被加一起计费 
  
  
  
  
  
，例如说输入了10个token


 


 


 


 


 

，openAI回复了20个token 
 
 
 
 
 ，那么最终收费是按照30个token进行收费
 
 
 
 
 
。 import tiktoken encoding = tiktoken.encoding_for_model("gpt-3.5-turbo") def num_tokens_from_string(string: str, encoding_name: str) -> int: """Returns the number of tokens in a text string.""" encoding = tiktoken.get_encoding(encoding_name) num_tokens = len(encoding.encode(string)) return num_tokens num_tokens_from_string("tiktoken is great!", "cl100k_base")

how to count tokens with tiktoken

示例

中文请求 # 中文 Reponse = openai.ChatCompletion.create( model="gpt-3.5-turbo", messages=[ {"role": "user", "content": "上海在哪里"}, ] )

回复

{ "choices": [ { "finish_reason": "stop", "index": 0, "message": { "content": "\n\n\u4e0a\u6d77\u4f4d\u4e8e\u4e2d\u56fd\u4e1c\u90e8\u6cbf\u6d77\u5730\u5e26\uff0c\u6bd7\u90bb\u6c5f\u82cf\u548c\u6d59\u6c5f\u4e24\u7701\uff0c\u5904\u4e8e\u957f\u6c5f\u53e3\u548c\u676d\u5dde\u6e7e\u4e4b\u95f4\uff0c\u5730\u7406\u5750\u6807\u4e3a31.23\u00b0N, 121.47\u00b0E\u3002", "role": "assistant" } } ], "created": 1678794854, "id": "chatcmpl-6txWIqBPu6GIbaN7fwnosvKzfkLEE", "model": "gpt-3.5-turbo-0301", "object": "chat.completion", "usage": { "completion_tokens": 63, "prompt_tokens": 13, "total_tokens": 76 } } 翻译 reponse = openai.ChatCompletion.create( model="gpt-3.5-turbo", messages=[ {"role": "system", "content": "You are a helpful assistant that translates English to French."}, {"role": "user", "content": Translate the following English text to French: "{text}"} ] )

回复

{ "choices": [ { "finish_reason": "stop", "index": 0, "message": { "content": "Je suis d\u00e9sol\u00e9, je ne peux pas traduire cette demande car il ny a pas de texte fourni entre les accolades. Veuillez ajouter du texte \u00e0 traduire.", "role": "assistant" } } ], "created": 1678798886, "id": "chatcmpl-6tyZKMOnfUUeRIszSmgovSKvzsNfA", "model": "gpt-3.5-turbo-0301", "object": "chat.completion", "usage": { "completion_tokens": 41, "prompt_tokens": 34, "total_tokens": 75 } }

也可以不加system

reponse = openai.ChatCompletion.create( model="gpt-3.5-turbo", messages=[ {"role": "user", "content": Translate the following English text to French: "{text}"} ] )

回复：

{ "choices": [ { "finish_reason": "stop", "index": 0, "message": { "content": "\n\n\"{text}\" is already in English and does not need to be translated.", "role": "assistant" } } ], "created": 1678799017, "id": "chatcmpl-6tybRxkYc92IE1mXELPuWnw2OFmgE", "model": "gpt-3.5-turbo-0301", "object": "chat.completion", "usage": { "completion_tokens": 17, "prompt_tokens": 18, "total_tokens": 35 } }