yuzaboto/VioletSalvia
收藏Hugging Face2024-03-30 更新2024-03-04 收录
下载链接：
https://hf-mirror.com/datasets/yuzaboto/VioletSalvia
下载链接
链接失效反馈官方服务：
资源简介：
Edit (30/03/2024):
This repo has been included in a masters application for university. This note is here for confirmation, since the usernames are different
-----------------------------------------------------
In-Development dataset to fix some core knowledge/ability tasks in language models.
Data follows the alpaca format (Instruction-Input-Response), and formated as a json file.
The json file holds a single array of dictionaries. Each dictionary has 3 keys, "Instruction","Input","Response".
Data has been generated using conventional code (no self-instruct), and pulls from subsets of pile_v1 for when raw text data is needed.
Questions aim to be un-ambiguous and mostly aimed at foundational knowledge. (rather that long instruction-chains or trivia)
Project could be dropped at any point.
---
license: cc-by-nc-4.0
---
提供机构：
yuzaboto
原始信息汇总
数据集概述
数据集格式
- 遵循alpaca格式，包含三个部分：Instruction（指令）、Input（输入）、Response（响应）。
- 数据以JSON文件形式存储，文件中包含一个字典数组，每个字典包含"Instruction"、"Input"、"Response"三个键。
数据来源
- 数据生成未使用自指导代码，而是从pile_v1数据集的子集中提取原始文本数据。
数据内容
- 问题设计旨在清晰无歧义，主要针对基础知识，避免复杂的指令链或琐碎问题。
使用许可
- 数据集遵循CC-BY-NC-4.0许可协议。
项目状态
- 该项目可能随时被终止。
