nemotron-code-oracle-filtered-qwen3.5-122b-131k-opencode-sft-serveparity
收藏资源简介:
该数据集是一个专门为工具调用(函数调用)任务设计的对话数据集。数据集包含5,441个训练样本,总大小约259MB。每个样本包含以下核心字段:1)messages:多轮对话消息列表,每条消息包含角色(role)、内容(content)和可选的工具调用记录(tool_calls);2)tools:工具定义信息;3)task:任务类型标识;4)num_turns:对话轮数;5)num_tool_calls:工具调用次数。其中tool_calls字段详细记录了函数调用的类型(type)和具体函数信息(包括函数名name和参数arguments)。数据集适用于训练和评估支持工具调用的对话系统,特别是需要理解用户指令、选择适当工具并执行函数调用的场景。
This is a dialogue dataset specifically tailored for tool calling (function calling) tasks. It contains 5,441 training samples with a total size of approximately 259 MB. Each sample includes the following core fields: 1) messages: a list of multi-turn dialogue messages, where each message contains a role, content, and an optional tool call record (tool_calls); 2) tools: tool definition information; 3) task: task type identifier; 4) num_turns: number of dialogue turns; 5) num_tool_calls: number of tool calls. The tool_calls field specifically records the type of function call (type) and specific function information including the function name (name) and parameters (arguments). This dataset is applicable for training and evaluating dialogue systems that support tool calling, particularly scenarios requiring the understanding of user instructions, selection of appropriate tools, and execution of function calls.




