nixiesearch/querygen-data-v4
收藏Hugging Face2024-10-21 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/nixiesearch/querygen-data-v4
下载链接
链接失效反馈官方服务：
资源简介：
该数据集用于训练Nixiesearch的querygen-v4模型，包含了多个公开的查询-文档数据集，格式适用于因果语言模型训练。数据集的具体来源包括MSMARCO、HotpotQA、NQ、MIRACL en、SQUAD和TriviaQA，训练集和测试集的行数分别为900000和8491。数据集可能包含重复的查询。
The querygen-v4 training dataset is used to train the not-yet-published querygen-v4 model from Nixiesearch. It is a combination of multiple open query-document datasets suitable for Causal LLM training. The dataset includes train data from MSMARCO, HotpotQA, NQ, MIRACL en, SQUAD, and TriviaQA, totaling 900,000 rows for training and 8,491 rows for testing. The dataset is in ZSTD-compressed JSONL format with two string columns: query and doc. The dataset is licensed under the Apache License 2.0.
提供机构：
nixiesearch
