metadata
dataset_info:
features:
- name: instruction
dtype: string
- name: prefer
dtype: string
- name: rejected
dtype: string
- name: images
list:
- name: bytes
dtype: binary
- name: path
dtype: 'null'
splits:
- name: train
num_bytes: 2112198428
num_examples: 49972
download_size: 2089954601
dataset_size: 2112198428
configs:
- config_name: default
data_files:
- split: train
path: data/train-*