Aratako/Self-Instruct-Qwen2.5-72B-Instruct-60k
收藏Hugging Face2024-12-16 更新2024-12-21 收录
下载链接：
https://hf-mirror.com/datasets/Aratako/Self-Instruct-Qwen2.5-72B-Instruct-60k
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是通过使用Magpie和Evol-Instruct技术创建的约6万条日语合成instruction数据。首先，使用Qwen/Qwen2.5-72B-Instruct-GPTQ-Int8模型对已有的合成instruction数据进行类别标注，然后利用Self-Instruct方法生成新的instruction。数据集包含多个特征，如primary_tag、seed_task、generated_instruction等，主要用于文本生成任务。
This dataset consists of approximately 60,000 Japanese synthetic instruction data created using Magpie and Evol-Instruct techniques. Initially, the Qwen/Qwen2.5-72B-Instruct-GPTQ-Int8 model was used to categorize existing synthetic instruction data, and then new instructions were generated using the Self-Instruct method. The dataset includes various features such as primary_tag, seed_task, generated_instruction, etc., and is primarily used for text generation tasks.
提供机构：
Aratako
