nicher92/magpie_llama70b_260k_filtered_swedish
收藏Hugging Face2025-02-13 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/nicher92/magpie_llama70b_260k_filtered_swedish
下载链接
链接失效反馈官方服务：
资源简介：
这是一个包含约260k个瑞典语指令-响应对的数据集，从大约650k个原始对中过滤而来。数据集包含了正常的问答、数学和编程问答以及多项选择题和答案。过滤过程包括去除重复项、评分低于良好或优秀的指令、响应评分低于-10的项以及长度小于10或大于2048的指令和响应。
This dataset consists of approximately 260k Swedish instruction-response pairs, filtered from about 650k original pairs. It includes normal QA, math and coding QA, and multiple-choice questions and answers. The filtering process involves removing duplicates, instructions scored less than good or excellent, responses scored below -10 by ArmoRM-Llama3-8B-v0.1, and instructions and responses that are less than 10 or more than 2048 in length.
提供机构：
nicher92
