README.md

metadata

dataset_info:
  features:
    - name: caption
      dtype: string
    - name: path
      dtype: string
    - name: source
      dtype: string
    - name: md5
      dtype: string
  splits:
    - name: train
      num_bytes: 2247590531
      num_examples: 2271892
  download_size: 1144751977
  dataset_size: 2247590531
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

This dataset is presented in Improving Long-Text Alignment for Text-to-Image Diffusion Models.