BeastyZ/ConRetriever

该数据集包含多个配置，每个配置都包含查询、正面示例和负面示例。这些数据用于训练e5-R-mistral-7b模型。具体配置包括eli5、fever、fiqa、hotpot_qa、msmarco_passage、nq、quora_duplicate、squad、synthetic和triviaqa。每个配置的数据文件路径也已提供。

原始信息汇总

数据集概述

数据集配置

配置名称	特征	训练集信息
eli5	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 25298207<br>num_examples: 25000
fever	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 461884567<br>num_examples: 71257
fiqa	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 74970232<br>num_examples: 5500
hotpot_qa	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 466370386<br>num_examples: 68657
msmarco_passage	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 1469115771<br>num_examples: 400782
nq	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 508591032<br>num_examples: 76923
quora_duplicate	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 46431262<br>num_examples: 60000
squad	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 1394419158<br>num_examples: 74377
synthetic	query: string<br>positive: sequence<br>negative: sequence<br>instruction: string	num_bytes: 450333569<br>num_examples: 224791
triviaqa	query: string<br>positive: sequence<br>negative: sequence	num_bytes: 375282874<br>num_examples: 52738

数据集文件路径

配置名称	训练数据文件路径
eli5	eli5/train-*
fever	fever/train-*
fiqa	fiqa/train-*
hotpot_qa	hotpot_qa/train-*
msmarco_passage	msmarco_passage/train-*
nq	nq/train-*
quora_duplicate	quora_duplicate/train-*
squad	squad/train-*
synthetic	synthetic/train-*
triviaqa	triviaqa/train-*

搜集汇总

背景与挑战

背景概述

该数据集是一个用于信息检索或对比学习任务的英语文本数据集，包含超过100万行数据，涵盖多个子集如eli5、fever和msmarco_passage等。数据以parquet格式存储，总大小为3.04 GB，主要用于训练e5-R-mistral-7b模型，支持文本嵌入和检索任务。