dclm_100BT-shuffled
收藏资源简介:
该数据集包含89,269,902个训练样本,总大小为522GB。数据文件存储在'train'分割路径下(data/train-*)。每个样本包含7个字段:文本内容(text)、唯一标识符(id)、来源URL(url)、语言标签(language)、语言置信度分数(language_score)、fasttext模型分数(fasttext_score)以及来源数据集标识(dataset)。所有字段均明确标注了数据类型,其中文本和标识类字段为字符串类型,分数值为浮点数。数据集下载大小约为316GB。
This dataset comprises 89,269,902 training samples with a total size of 522 GB. All data files are stored under the 'train' split, following the path pattern data/train-*. Each sample includes 7 fields: text content (text), unique identifier (id), source URL (url), language label (language), language confidence score (language_score), fasttext model score (fasttext_score), and source dataset identifier (dataset). All fields are explicitly annotated with their data types, where text and identifier fields are of string type, while score values are floating-point numbers. The approximate download size of the dataset is 316 GB.



