erdi28/honaz_instruction_dataset

该数据集是关于土耳其小镇Honaz的独特数据集，基于三篇土耳其学术文章构建，并翻译成英文。数据集的目的是测试模型是否能够从这种特定的小众数据中学习，并用于模型微调和DPO对齐。数据集包含训练集，共有1010个样本，总大小为1661261字节。

The Honaz dataset is a unique dataset about the town of Honaz, known for its rich history and natural beauty. The dataset is built upon three academic articles in Turkish, covering various aspects of Honaz, including historical migrations, developments from the 19th to the 20th century, and the vegetation cover around Mount Honaz. The goal of the dataset is to translate this detailed localized information into English and test if a model can learn from such niche data. Additionally, the dataset includes a script for creating your own instruction and alignment dataset from local files.