I did not find a tutorial on how to set the preview for my dataset. Is there any guidance?

I did not find a tutorial on how to set the preview for my dataset. Is there any guidance? Thanks

what is your repository? Please feel free to open a discussion on the repo and we will give you support there.
Also: see the docs Data files Configuration, hopefully you will find a guidance (see also the datasets examples at datasets-examples (Datasets examples))

Hi @severo , here is my repo: jinggu/layout-understanding · Datasets at Hugging Face. I am trying to create a preview that show some images and corresponding text caption. So I created a test folder and a data.csv. I hope that csv file will guide the dataset preview content. However, the preview still show three tags: train, validation and test. How can I just show the preview with only one tag, and with some images and corresponding text caption?

I think you have to:

  • delete data.csv from the root
  • rename test_set_real_life.json to metadata.jsonl
  • change its format to JSONL (ie: remove the enclosing array: every line is a JSON object)

If you want more control on the splits, read File names and splits

Thanks! Could you share an example of jsonl? I am not sure what key and value should be included so to include the image and corresponding caption.