configs:
- config_name: default
  data_files:
  - split: train
    path: "metadata.csv"