---
dataset_info:
  features:
  - name: image_ref
    dtype: image
  - name: image_win
    dtype: image
  - name: image_lose
    dtype: image
  - name: caption
    dtype: string
  splits:
  - name: train
    num_bytes: 34095891715.694
    num_examples: 30406
  download_size: 34178832827
  dataset_size: 34095891715.694
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---