README.md

metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: example_id
      dtype: string
    - name: description
      dtype: string
    - name: qa
      list:
        - name: answer
          dtype: string
        - name: question
          dtype: string
  splits:
    - name: train
      num_bytes: 7060345226.106
      num_examples: 13647
    - name: test
      num_bytes: 254553007
      num_examples: 500
    - name: val
      num_bytes: 261171671
      num_examples: 500
  download_size: 7544082762
  dataset_size: 7576069904.106
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
      - split: val
        path: data/val-*
license: apache-2.0

The dataset consists of a collection of general questions and orders/instructions added to google/docci dataset. 4000 test example moved into train dataset.