ChaiML/DPO_experiment_triplets_by_elo_v1

该数据集包含对话、选择的完成、拒绝的完成和elo评分等特征。对话特征包括内容和角色，选择的完成和拒绝的完成也包含内容和角色。elo评分是一个浮点数序列。数据集分为训练集，包含9261个示例，大小为64163058字节。The dataset includes features such as conversations, chosen completion, rejected completion, and elo scores. The conversation feature includes content and role, and the chosen and rejected completions also include content and role. The elo score is a sequence of float64. The dataset is divided into a training set, containing 9261 examples, with a size of 64163058 bytes.

原始信息汇总

数据集概述

数据集信息

特征

• conversations:
  • content: 字符串类型
  • role: 字符串类型
• chosen_completion:
  • content: 字符串类型
  • role: 字符串类型
• rejected_completion:
  • content: 字符串类型
  • role: 字符串类型
• elo_scores: 浮点数序列

数据分割

• train:
  • num_bytes: 64163058
  • num_examples: 9261

数据集大小

• download_size: 26160970
• dataset_size: 64163058

配置

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