---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: messages
    list:
    - name: question
      dtype: string
    - name: answer
      dtype: string
    - name: thought
      dtype: string
    - name: bbox
      dtype: string
  - name: metadata
    struct:
    - name: website
      dtype: string
    - name: url
      dtype: string
    - name: image_w
      dtype: int32
    - name: image_h
      dtype: int32
  splits:
  - name: train
    num_bytes: 1495075856
    num_examples: 5000
  download_size: 1495371868
  dataset_size: 1495075856
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---