metadata
dataset_info:
features:
- name: chosen
list:
- name: content
dtype: string
- name: role
dtype: string
- name: rejected
list:
- name: content
dtype: string
- name: role
dtype: string
- name: task_id
dtype: int32
- name: step_id
dtype: int32
- name: source_name
dtype: string
splits:
- name: train
num_bytes: 588633697
num_examples: 30960
download_size: 60644221
dataset_size: 588633697
configs:
- config_name: default
data_files:
- split: train
path: data/train-*