kadirnar/Emilia-DE-SmolLM2

该数据集包含以下字段：持续时间（浮点型）、说话者（字符串型）、语言（字符串型）、dnsmos（语音质量评分，浮点型）、phone_count（电话数量，整型）、唯一标识符（字符串型）、源文件夹路径（字符串型）、输入ID序列（整型序列）、标签序列（长整型序列）和注意力掩码序列（字整型序列）。数据集分为训练集，包含约200万4966个示例，总大小约为21.68GB。

The dataset includes fields such as duration (float64), speaker (string), language (string), dnsmos (voice quality score, float64), phone_count (number of phones, int64), unique identifier (_id, string), source folder path (string), input ID sequence (int32 sequence), label sequence (int64 sequence), and attention mask sequence (int8 sequence). The dataset is split into a training set, containing approximately 2,004,966 examples, with a total size of about 21.68GB.