orionweller/contrastive-pretraining
收藏Hugging Face2026-04-24 更新2025-10-25 收录
下载链接:
https://hf-mirror.com/datasets/orionweller/contrastive-pretraining
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含了多种语言的查询和文档对,用于对比预训练任务。每个语言配置都有对应的查询和文档数据,数据按照 LightOn 风格的模式组织。数据集的每个条目包括一个索引、一个查询、一个文档、一个表示是否丢弃该条目的布尔值和一个表示重复次数的整数。
The dataset contains query/document pairs in multiple languages for contrastive pretraining tasks. Each language configuration has corresponding query and document data, organized in a LightOn-style schema. Each entry in the dataset includes an index, a query, a document, a boolean indicating whether to drop the entry, and an integer representing the number of duplicates.
提供机构:
orionweller



