制作自己的ChatGPT

Feb 11, 20235 min read

推荐：使用 NSDT场景设计器快速搭建 3D场景。

众所周知，ChatGPT 目前能够取得令人印象深刻的壮举。很可能许多人都有在他们自己的项目中使用该技术的想法。不过需要注意的是，ChatGPT 目前并没有官方的 API。使用非官方 API 可能会导致困难。

目前需要手动获取access token和cloudflare token才能使用API。此外，这些令牌必须每两小时手动更改一次。

1、ChatGPT

ChatGPT 在其设计中使用了 GPT-3 模型，并在此基础上开发了一个新模型。因此，新模型的输出结果往往与 GPT-3 相似。在撰写本文时，新模型的 ChatGPT 中使用了 text-davinci-002-render 模型，但目前尚未向公众开放。

虽然 ChatGPT 可能不是开创性的，但它提供了一个利用现有技术的新界面。通过利用强大的提示和高效的记忆窗口。因此，我们可以使用 LLM Chain 方法复制其功能，而不是破解 ChatGPT 的非官方 API。

2、LangChain

Langchain 是一个新的 python 包，它为LLM链提供了一个标准接口，与其他工具的大量集成，以及用于常见应用程序的端到端链。

LangChain 旨在协助四个主要领域，此处按复杂性递增的顺序列出：

LLM和提示

代理

记忆

在此处的官方文档中了解有关 langchain 的更多信息。

3、安装

要使用 langchain 包，你可以从 pypi 安装它。

pip install langchain

要从 langchain 获取最新更新，你可以使用这种安装方法。

pip install "git+https://github.com/hwchase17/langchain.git"

更多安装选项请阅读此处。

4、示例项目

你可以使用 ChatGPT 做很多事情，其中一个有趣的事情是为学生作业建立问答。所以这次我们将创建 AI 版的 Brainly。

下面是我们将从 ChatGPT 中得到的。

下面是 langchain 的提示：

from langchain.llms import OpenAI
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from langchain.chains import SimpleSequentialChainllm = OpenAI(temperature=.7)
template = """You are a teacher in physics for High School student. Given the text of question, it is your job to write a answer that question with example.
Question: {text}
Answer:
"""
prompt_template = PromptTemplate(input_variables=["text"], template=template)
answer_chain = LLMChain(llm=llm, prompt=prompt_template)
answer = answer_chain.run("What is the formula for Gravitational Potential Energy (GPE)?")
print(answer)

这是我们使用 langchain 从 GPT-3 获得的结果：

The formula for Gravitational Potential Energy (GPE) is GPE = mgh, where m is the mass of an object, g is the acceleration due to gravity, and h is the height of the object. For example, if an object with a mass of 10 kg is at a height of 5 meters, then the GPE would be GPE = 10 x 9.8 x 5 = 490 Joules.

5、聊天机器人

如果你需要创建像 AI 这样的聊天机器人，你可以使用 langchain 的内存。这是如何操作的示例。

from langchain.chains.conversation.memory import ConversationBufferMemory
from langchain import OpenAI, LLMChain, PromptTemplatetemplate = """You are a teacher in physics for High School student. Given the text of question, it is your job to write a answer that question with example.
{chat_history}
Human: {question}
AI:
"""
prompt_template = PromptTemplate(input_variables=["chat_history","question"], template=template)
memory = ConversationBufferMemory(memory_key="chat_history")llm_chain = LLMChain(llm=OpenAI(),prompt=prompt_template,verbose=True,memory=memory,
)llm_chain.predict(question="What is the formula for Gravitational Potential Energy (GPE)?")result = llm_chain.predict(question="What is Joules?")
print(result)

结果会是这样的：

$ python3 memory.py> Entering new LLMChain chain...
Prompt after formatting:
You are a teacher in physics for High School student. Given the text of question, it is your job to write a answer that question with example.Human: What is the formula for Gravitational Potential Energy (GPE)?
AI:> Finished LLMChain chain.> Entering new LLMChain chain...
Prompt after formatting:
You are a teacher in physics for High School student. Given the text of question, it is your job to write a answer that question with example.Human: What is the formula for Gravitational Potential Energy (GPE)?
AI: 
The formula for Gravitational Potential Energy (GPE) is GPE = mgh, where m is the mass of the object, g is the acceleration due to gravity, and h is the height of the object.For example, if an object has a mass of 10 kg and is at a height of 5 meters, then the gravitational potential energy of the object is GPE = 10 kg x 9.8 m/s2 x 5 m = 490 Joules.
Human: What is Joules?
AI:> Finished LLMChain chain.
Joules (J) is the SI unit of energy. It is defined as the amount of energy required to move an object of one kilogram at a speed of one meter per second. It is also equal to the work done when a force of one Newton is applied to an object and moved one meter in the direction of the force.