jaeyong2/ja-persona-cot-inst

该数据集包含两个特征：content和text，数据类型均为字符串。数据集分为一个训练集，包含645,000个样本，总大小为2,480,381,909字节。数据集的语言为日语，许可证为cc-by-nc-sa-4.0。使用方式是通过HuggingFace的`load_dataset`函数加载数据集。开发过程包括从另一个数据集加载问题，并使用Qwen/Qwen2-72B-Instruct模型生成带有COT的答案。研究得到了TPU Research Cloud program的支持。

This dataset includes two features: content and text, both of which are of string type. The dataset is divided into a training set containing 645,000 samples, with a total size of 2,480,381,909 bytes. The dataset is in Japanese and uses the CC-BY-NC-SA-4.0 license. The development process of the dataset involves loading the question dataset from jaeyong2/persona-inst and using the Qwen/Qwen2-72B-Instruct model to generate answers with COT (Chain of Thought).