five

yimingzhang/hh-rlhf-sft

收藏
Hugging Face2024-05-30 更新2024-06-12 收录
下载链接:
https://hf-mirror.com/datasets/yimingzhang/hh-rlhf-sft
下载链接
链接失效反馈
官方服务:
资源简介:
--- dataset_info: features: - name: chosen dtype: string - name: rejected dtype: string - name: chosen-chat list: - name: content dtype: string - name: role dtype: string - name: rejected-chat list: - name: content dtype: string - name: role dtype: string - name: chosen-safety dtype: string - name: rejected-safety dtype: string - name: messages list: - name: content dtype: string - name: role dtype: string splits: - name: train_sft num_bytes: 528401375 num_examples: 160800 - name: test_sft num_bytes: 28357007 num_examples: 8552 download_size: 314367654 dataset_size: 556758382 configs: - config_name: default data_files: - split: train_sft path: data/train_sft-* - split: test_sft path: data/test_sft-* ---

The dataset includes multiple features such as chosen and rejected, both of which are string types. Additionally, there are chosen-chat and rejected-chat, which are lists containing content and role, also string types. There are also chosen-safety and rejected-safety, as well as messages, the latter being a list containing content and role. The dataset is divided into two parts: train_sft and test_sft, containing 160800 and 8552 examples respectively. The configuration name of the dataset is default, and the data file paths are in the data/ directory.
提供机构:
yimingzhang
原始信息汇总

数据集概述

数据集特征

  • chosen:字符串类型
  • rejected:字符串类型
  • chosen-chat:列表类型,包含以下字段:
    • content:字符串类型
    • role:字符串类型
  • rejected-chat:列表类型,包含以下字段:
    • content:字符串类型
    • role:字符串类型
  • chosen-safety:字符串类型
  • rejected-safety:字符串类型
  • messages:列表类型,包含以下字段:
    • content:字符串类型
    • role:字符串类型

数据集分割

  • train_sft
    • 数据大小:528401375字节
    • 示例数量:160800
  • test_sft
    • 数据大小:28357007字节
    • 示例数量:8552

数据集大小

  • 下载大小:314367654字节
  • 数据集大小:556758382字节

数据文件配置

  • default 配置:
    • train_sft:路径为 data/train_sft-*
    • test_sft:路径为 data/test_sft-*
5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作