cornfieldrm/pair-preference-dataset-700K_subset-4-of-4_gemma-2b_1of4_iter1_bs128_lr1e-5_conf-0.8

---
dataset_info:
features:
- name: rejected
list:
- name: content
dtype: string
- name: role
dtype: string
- name: rejected_score
dtype: float64
- name: chosen_score
dtype: float64
- name: chosen
list:
- name: content
dtype: string
- name: role
dtype: string
- name: length
dtype: int64
- name: chosen_prob
dtype: float64
- name: messages
list:
- name: content
dtype: string
- name: role
dtype: string
splits:
- name: train
num_bytes: 301022781.5413197
num_examples: 44392
download_size: 196056771
dataset_size: 301022781.5413197
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
---

This dataset is designed for analyzing and comparing the quality of different dialogue contents. It includes multiple features such as rejected and chosen, each containing content and role fields, along with related scores like rejected_score and chosen_score. Additionally, there are features like length and chosen_prob. The dataset is divided into a training set, containing 44392 samples.

原始信息汇总

数据集概述

数据集特征

• rejected

  • content: 数据类型为字符串
  • role: 数据类型为字符串

• rejected_score: 数据类型为浮点数

• chosen_score: 数据类型为浮点数

• chosen

  • content: 数据类型为字符串
  • role: 数据类型为字符串

• length: 数据类型为整数

• chosen_prob: 数据类型为浮点数

• messages

  • content: 数据类型为字符串
  • role: 数据类型为字符串

数据集分割

• train
  • num_bytes: 301022781.5413197 字节
  • num_examples: 44392 个样本

数据集大小

• download_size: 196056771 字节
• dataset_size: 301022781.5413197 字节