coffee-barista-dataset
收藏资源简介:
Coffee Barista Training Dataset(咖啡师训练数据集)是一个专门用于监督微调(SFT)的中文指令数据集,旨在训练语言模型(如 Qwen2.5-3B-Instruct)成为专业的“金牌咖啡师”。该数据集聚焦于咖啡相关领域,涵盖了文本生成和问答任务。数据规模在1千到1万条之间,当前最新版本(v21)包含1392条高质量样本。数据以JSONL格式存储,每条样本包含四个字段:instruction(指令)、input(输入,通常为空字符串)、output(输出)和 system(系统提示)。该版本已移除“think”标签并包含工具调用(tool call)内容。数据集适用于训练或微调模型以处理咖啡制作、知识问答、顾客服务等专业对话场景。
The Coffee Barista Training Dataset is a specialized Chinese instruction dataset for supervised fine-tuning (SFT), designed to train language models (such as Qwen2.5-3B-Instruct) to become professional gold medal baristas. It focuses on the coffee-related domain and covers text generation and question-answering tasks. The dataset size ranges from 1,000 to 10,000 entries, with the latest version (v21) containing 1,392 high-quality samples. Data is stored in JSONL format, with each sample including four fields: instruction, input (usually an empty string), output, and system prompt. This version has removed the think label and includes tool call content. The dataset is suitable for training or fine-tuning models to handle professional dialogue scenarios such as coffee making, knowledge Q&A, and customer service.
数据集概述
名称:Coffee Barista Training Dataset(金牌咖啡师训练数据集)
许可证:MIT
语言:中文(zh)
规模:1,000 至 10,000 条数据(1K < n < 10K)
任务类别:文本生成(text-generation)、问答(question-answering)
标签:咖啡(coffee)、监督微调(sft)、Qwen、中文(chinese)、咖啡师(barista)
模型基准:Qwen2.5-3B-Instruct
数据集内容
- 用途:专为 SFT(监督微调)设计的金牌咖啡师问答数据集。
- 数据格式:JSONL 格式,每行包含四个字段:
instruction:指令input:输入(当前为空字符串)output:输出(回答)system:系统提示
- 版本:
- 当前最新版本 v21,共 1,392 条数据。
- v21 版本特点:去除了
think标签,包含工具调用(tool call)。
仓库结构
coffee-barista-dataset/ ├── coffee_latest.jsonl ← 最新版 (当前 v21, 1392条) ├── archive/ │ └── coffee_v21.jsonl ← v21 历史快照 │ └── ... ← 未来迭代 └── README.md
快速加载
python from datasets import load_dataset ds = load_dataset("ynanxiu/coffee-barista-dataset", split="train")
默认加载 coffee_latest.jsonl





