nthakur/bge-full-data
收藏官方服务：
资源简介：
这是一个转换为Tevatron格式的数据集，基于原始的`cfli/bge-full-data`。它包含了用于训练的15个检索划分，如sts、msmarco_passage、hotpotqa等。在数据集的后处理中，原始文档的全文被推送到`text`字段，而`title`字段则为空。由于原始文档或查询ID不可用，因此通过计算文本的MD5哈希来创建唯一的查询和文档ID。
This is a ported version of the original `cfli/bge-full-data` into Tevatron format containing 15 retrieval splits for training, such as sts, msmarco_passage, hotpotqa, etc. In the postprocessing of the dataset, the full text of the original document is pushed into the `text` field, while the `title` field is left empty. As the original document or query IDs were unavailable, unique query and document IDs are created by computing the MD5 hash of the text.
提供机构：
nthakur
