theojiang/MINI_CIVETv2_key_idea_retrieval_dataset_v3.1_gtebase_msmarco

该数据集包含两个配置项：data和metadata。data配置项包含段落输入ID、段落注意力掩码和问题嵌入三个特征，数据集分为训练集和验证集，训练集包含507990个样本，验证集包含500个样本。metadata配置项包含索引特征，训练集包含917个样本，验证集包含50个样本。数据集的总下载大小为2610513560字节，总大小为11647750820.00231字节。

The dataset contains two configurations: data and metadata. The data configuration includes three features: passage_input_ids, passage_attention_mask, and question_embeddings, representing passage input IDs, passage attention masks, and question embeddings, respectively. The dataset is divided into a training set with 507,990 samples and a validation set with 500 samples. The metadata configuration includes a feature: indices, representing indices. The training set contains 917 samples, and the validation set contains 50 samples. The total download size of the dataset is 2,610,513,560 bytes, and the total size is 11,647,750,820.00231 bytes.