metadata
dataset_info:
features:
- name: prompt
dtype: string
- name: chosen
list:
- name: content
dtype: string
- name: role
dtype: string
- name: rejected
list:
- name: content
dtype: string
- name: role
dtype: string
- name: all_generated_responses
sequence: string
- name: all_rm_scores
sequence: 'null'
splits:
- name: train
num_bytes: 150350
num_examples: 50
- name: test
num_bytes: 33559
num_examples: 10
download_size: 124077
dataset_size: 183909
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
tags:
- dpo