tarsur909/summarize_sft
收藏Hugging Face2024-11-11 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/tarsur909/summarize_sft
下载链接
链接失效反馈官方服务：
资源简介：
该数据集来源于OpenAI的Summarize from Feedback任务，主要用于文本摘要任务。数据集包含多个特征，如id、subreddit、title、post、summary等，这些特征直接来源于原始数据集。此外，还包含一些通过预处理脚本添加的特征，如query、query_input_ids、reference_response_input_ids等。数据集分为训练集、验证集和测试集，分别包含116722、6447和6553个样本。
The dataset is a TL;DR SFT dataset designed for OpenAIs [Summarize from Feedback](https://openai.com/blog/summarization/) task. The dataset contains multiple fields such as id, subreddit, title, post, summary, reference_response, etc., which are directly extracted from the aforementioned dataset. Additionally, there are fields added by the preprocessing script, such as query, query_input_ids, reference_response_input_ids, etc. The dataset is divided into training, validation, and test sets, containing 116722, 6447, and 6553 samples respectively.
提供机构：
tarsur909
