maxseats/aihub-464-preprocessed-680GB-set-8

该数据集包含音频数据、转录文本、输入特征和标签。音频数据的采样率为16000Hz，转录文本为字符串类型，输入特征为浮点数序列，标签为整数序列。数据集分为训练集、测试集和验证集，训练集包含41997个样本，测试集和验证集各包含5250个样本。数据集的下载大小为21244860674字节，总大小为60773101067.0字节。

The dataset includes audio data, transcripts, input features, and labels. The audio data has a sampling rate of 16000Hz, transcripts are of string type, input features are sequences of float32, and labels are sequences of int64. The dataset is divided into training, test, and validation sets, with the training set containing 41997 samples, and the test and validation sets each containing 5250 samples. The download size of the dataset is 21244860674 bytes, and the total size is 60773101067.0 bytes.

原始信息汇总

数据集概述

数据特征

• audio:
  • 采样率: 16000
• transcripts:
  • 数据类型: string
• input_features:
  • 序列类型: float32
• labels:
  • 序列类型: int64

数据集划分

• train:
  • 字节数: 48617786264.18269
  • 样本数: 41997
• test:
  • 字节数: 6077657401.408652
  • 样本数: 5250
• valid:
  • 字节数: 6077657401.408652
  • 样本数: 5250

数据集大小

• 下载大小: 21244860674
• 总大小: 60773101067.0

配置

• config_name: default
  • 数据文件路径:
    • train: data/train-*
    • test: data/test-*
    • valid: data/valid-*