mlfoundations-dev/OH_DCFT_v1_wo_unnatural_instructions

该数据集包含两个主要特征：sample和source。sample特征是一个列表，包含from和value两个子特征，它们的类型都是字符串。source特征的类型也是字符串。数据集有一个训练集（train）分割，包含1,028,698个样本，总大小为1,741,581,655.205821字节。数据集的下载大小为991,069,872字节。配置部分指定了默认配置下的数据文件路径。

The dataset contains two main features: sample and source. The sample feature is a list containing two sub-features: from and value, both of which are of string type. The source feature is also of string type. The dataset has one training split (train) with 1,028,698 samples and a total size of 1,741,581,655.205821 bytes. The download size of the dataset is 991,069,872 bytes. The configuration section specifies the data file paths under the default configuration.