metadata
dataset_info:
features:
- name: prompt
dtype: string
- name: chosen
dtype: string
- name: rejected
dtype: string
- name: pred_chosen
dtype: float32
- name: pred_rejected
dtype: float32
splits:
- name: train
num_bytes: 37788099
num_examples: 19999
download_size: 23251925
dataset_size: 37788099
configs:
- config_name: default
data_files:
- split: train
path: data/train-*