corniclr25/stack-mined-java-v1

该数据集包含四个主要特征：查询（query）、文档（document）、负面样本（negatives）和元数据（metadata）。元数据中进一步包含目标（objective）结构，该结构下又分为配对（paired）、自我（self）和三元组（triplet）。数据集分为一个训练集（train），包含16000000个样本，总大小为142451950689字节。下载大小为50443007078字节。

The dataset contains four main features: query, document, negatives, and metadata. Query and document are string types, negatives is a sequence of strings, and metadata is a structured data containing the objective field, which in turn contains three subfields: paired, self, and triplet. The dataset is divided into a training set (train) with 16,000,000 samples, the total size of the dataset is 142,451,950,689 bytes, and the download size is 50,443,007,078 bytes.