metadata
dataset_info:
features:
- name: question
dtype: string
- name: docId
dtype: int64
- name: answers
sequence: string
- name: data_split
dtype: string
- name: bounding_boxes
sequence:
sequence: int64
- name: word_list
sequence: string
- name: image_raw
dtype: image
- name: ground_truth
dtype: string
splits:
- name: train
num_bytes: 21886657466.22558
num_examples: 30000
- name: distill
num_bytes: 6903780998.354423
num_examples: 9463
- name: test
num_bytes: 1823888087.1854649
num_examples: 2500
- name: valid
num_bytes: 2078502810.0255556
num_examples: 2849
download_size: 32064774574
dataset_size: 32692829361.79102
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: distill
path: data/distill-*
- split: test
path: data/test-*
- split: valid
path: data/valid-*