聊聊ShareGPT格式的微调数据集

小半 • 2024年3月23日下午5:13 • 技术漫谈 • 阅读 549

概述

ShareGPT格式的数据集中，一般是如下格式：

[
  {
    "conversations": [
      {
        "from": "human",
        "value": "I saw a dress that I liked. It was originally priced at $200 but it's on sale for 20% off. Can you tell me how much it will cost after the discount?"
      },
      {
        "from": "function_call",
        "value": "{"name": "calculate_discount", "arguments": {"original_price": 200, "discount_percentage": 20}}"
      },
      {
        "from": "observation",
        "value": "{"discounted_price": 160}"
      },
      {
        "from": "gpt",
        "value": "The dress will cost you $160 after the 20% discount."
      }
    ],
    "system": "系统提示词（选填）",
    "tools": "[{"name": "calculate_discount", "description": "Calculate the discounted price", "parameters": {"type": "object", "properties": {"original_price": {"type": "number", "description": "The original price of the item"}, "discount_percentage": {"type": "number", "description": "The percentage of discount"}}, "required": ["original_price", "discount_percentage"]}}]"
  }
]

function_call表示函数调用，什么是函数调用？其作用是什么？由于大模型的数据一般都是截止于某个时间点之前的数据，不具备实时性。比如，我要问今天的天气，正常来说，由于模型参数的局限性，是不会知道的。但基于函数调用的功能，就解决了这个问题。所谓的 function_call，在某个程度来说，可以理解为API调用，这个API就是一个function，提供了某种功能。

observation表示观测结果，即 function_call的执行结果。

tools表示工具，即对 function_call的总结描述。