pszemraj/flan-comp-add_data

该数据集是一个用于文本生成任务的数据集，包含两个配置：deduped和default。每个配置都包含输入、目标和任务名称的特征，并且都有训练集的分割信息。数据集还包括一些补充的指令数据，这些数据被添加到FLAN中以创建FLAN-compressed ++。

This dataset is an English dataset for text-to-text generation tasks, licensed under odc-by. The dataset size is between 1M and 10M, containing two configurations: deduped and default. Each configuration includes input and target features, as well as a task name. The dataset is used for training, with detailed training set information such as the number of bytes and samples. Additionally, the dataset includes supplemental data sources for FLAN++, such as tasksource-instruct-v0, deepmind/code-contests, etc.