TeoGchx/BEAT_HML3D_chinese

该数据集包含多种数据类型，包括语音（speech）、动作（motion）、文本（text）和元数据（meta_data）。语音数据以float32序列形式存储，动作数据是float32的序列的序列，文本数据是字符串序列。元数据包括持续时间（duration）、名称（name）、帧数（num_frames）、单词数（num_words）和说话者（speaker）等信息。数据集分为训练集（train）、测试集（test）、验证集（val）和附加集（additional），每个部分都有特定的字节数和示例数。

The dataset includes various types of data such as speech, motion, text, and metadata. Speech data is stored as sequences of float32, motion data as sequences of sequences of float32, and text data as sequences of strings. Metadata includes information such as duration, name, number of frames, number of words, and speaker. The dataset is divided into training (train), testing (test), validation (val), and additional (additional) sets, each with specific byte sizes and number of examples.