Synthyra/uniref50

该数据集包含文本序列作为特征，有三个数据划分：训练集包含68340540个示例，大小为19850004390.27字节；验证集和测试集各包含10000个示例，大小均为2904572.37字节。整个数据集的大小为19855813535.0字节，下载大小为19881625108字节。

The dataset includes text sequences as features, and is divided into three splits: the training set contains 68340540 examples, with a size of 19850004390.27 bytes; the validation set and test set each contain 10000 examples, with a size of 2904572.37 bytes respectively. The total size of the dataset is 19855813535.0 bytes, and the download size is 19881625108 bytes.

搜集汇总

背景与挑战

背景概述

Synthyra/uniref50是一个大型文本数据集，采用parquet格式，数据规模在10M到100M之间，包含超过6800万行数据，主要分为train、valid和test三个子集。数据集由Hugging Face托管，支持多种数据处理库。