yguooo/summarize_from_feedback_tldr_3_filtered_oai_preprocessing_llama3_scene2
收藏Hugging Face2024-11-25 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/yguooo/summarize_from_feedback_tldr_3_filtered_oai_preprocessing_llama3_scene2
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是用于OpenAI的Summarize from Feedback任务的TL;DR SFT数据集。数据集直接从https://github.com/openai/summarize-from-feedback/tree/700967448d10004279f138666442bf1497d0e705#reddit-tldr-dataset获取。数据集包含多个特征，如id、subreddit、title、post、summary等，这些特征直接从原始数据集中提取。此外，预处理脚本还添加了一些新的特征，如query、query_token、reference_response_token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 directly taken from OpenAIs Summarize from Feedback task, containing the Reddit TL;DR dataset. The dataset features include post ID, subreddit, post title, post content, post summary, reference response, etc. Additionally, some features added by the preprocessing script, such as query, tokenized version of the query, tokenized version of the reference response, etc. The dataset is divided into training, validation, and test parts, each with corresponding sample numbers and byte counts.
提供机构：
yguooo
