metadata
dataset_info:
features:
- name: image
dtype: image
- name: elements
sequence:
- name: role
dtype: string
- name: text
dtype: string
splits:
- name: train
num_bytes: 188454566523
num_examples: 2000000
download_size: 188179589916
dataset_size: 188454566523
configs:
- config_name: default
data_files:
- split: train
path: data/train-*