seongil-dn/korean_retrieval_subset

该数据集包含三个主要特征：正面样本（positive）、困难负面样本（hard_negative）和查询（query）。数据集仅包含一个训练集（train），共有287,413个样本，总大小为963,545,224字节，下载大小为563,283,601字节。数据集的默认配置文件指定了数据文件路径为data/train-*。The dataset contains three main features: positive, hard_negative, and query. The dataset includes only a training set (train) with 287,413 examples, totaling 963,545,224 bytes, and a download size of 563,283,601 bytes. The default configuration file specifies the data file path as data/train-*.