distilabel-internal-testing/fine-preferences-magpie-v4

该数据集是通过distilabel工具生成的，主要用于对话生成任务。数据集中包含多个配置（如chat_generation_0、chat_generation_1、chat_generation_2和default），每个配置都有不同的特征和结构。数据集的特征包括文本、ID、URL、文件路径、语言、语言评分、对话内容、模型名称等。数据集的结构以JSON格式呈现，包含对话内容、系统提示、生成内容等。数据集的标签包括synthetic、distilabel和rlaif，表明数据集是合成的，并且与强化学习和人工智能反馈相关。

This dataset was generated using the distilabel tool and is primarily used for dialogue generation tasks. The dataset contains multiple configurations (e.g., chat_generation_0, chat_generation_1, chat_generation_2, and default), each with different features and structures. The features of the dataset include text, ID, URL, file path, language, language score, conversation content, model names, etc. The dataset structure is presented in JSON format, containing conversation content, system prompts, generated content, etc. The datasets tags include synthetic, distilabel, and rlaif, indicating that the dataset is synthetic and related to reinforcement learning and AI feedback.

原始信息汇总

数据集概述

数据集结构

配置信息

• config_name: chat_generation_0

  • features:
    • text: string
    • id: string
    • dump: string
    • url: string
    • file_path: string
    • language: string
    • language_score: float64
    • token_count: int64
    • score: float64
    • int_score: int64
    • system_prompt: string
    • conversation: list
      • content: string
      • role: string
    • model_name-0: string
    • generation-0: string
    • distilabel_metadata: struct
      • raw_output_chat_generation_0: string
  • splits:
    • train
      • num_bytes: 1706637
      • num_examples: 100
  • download_size: 800699
  • dataset_size: 1706637

• config_name: chat_generation_1

  • features:
    • text: string
    • id: string
    • dump: string
    • url: string
    • file_path: string
    • language: string
    • language_score: float64
    • token_count: int64
    • score: float64
    • int_score: int64
    • system_prompt: string
    • conversation: list
      • content: string
      • role: string
    • model_name-1: string
    • generation-1: string
    • distilabel_metadata: struct
      • raw_output_chat_generation_1: string
  • splits:
    • train
      • num_bytes: 1694451
      • num_examples: 100
  • download_size: 805594
  • dataset_size: 1694451

• config_name: chat_generation_2

  • features:
    • text: string
    • id: string
    • dump: string
    • url: string
    • file_path: string
    • language: string
    • language_score: float64
    • token_count: int64
    • score: float64
    • int_score: int64
    • system_prompt: string
    • conversation: list
      • content: string
      • role: string
    • model_name-2: string
    • generation-2: string
    • distilabel_metadata: struct
      • raw_output_chat_generation_2: string
  • splits:
    • train
      • num_bytes: 1692913
      • num_examples: 100
  • download_size: 796622
  • dataset_size: 1692913

• config_name: default

  • features:
    • text: string
    • id: string
    • dump: string
    • url: string
    • file_path: string
    • language: string
    • language_score: float64
    • token_count: int64
    • score: float64
    • int_score: int64
    • system_prompt: string
    • conversation: list
      • content: string
      • role: string
    • gen_conv_model_name: string
    • generations: sequence: string
    • distilabel_metadata: struct
      • raw_output_chat_generation_2: string
    • generations_model_names: sequence: string
  • splits:
    • train
      • num_bytes: 1976083
      • num_examples: 100
  • download_size: 917721
  • dataset_size: 1976083

数据文件路径

• config_name: chat_generation_0

  • data_files:
    • split: train
      • path: chat_generation_0/train-*

• config_name: chat_generation_1

  • data_files:
    • split: train
      • path: chat_generation_1/train-*

• config_name: chat_generation_2

  • data_files:
    • split: train
      • path: chat_generation_2/train-*

• config_name: default

  • data_files:
    • split: train
      • path: data/train-*

标签

• synthetic
• distilabel
• rlaif