Skip to content

Using Dataset and Model Card Templates

We provide Dataset and Model Card templates specific to the course field site at which data is colected. These templates follow the Imageomics and ABC templates, which include guidance and examples for the various metadata sections, reference information for Hugging Face's particular flavor of markdown, and the appropriate NSF & NSERC grant acknowledgments. We also include acknowledgments for the specified field sites.

To use a template for a new dataset or model repository on Hugging Face (HF), simply copy and paste the contents of the appropriate template (Dataset Card or Model Card) into your README.md file.1 Then, follow the descriptions under each section to fill in the appropriate information. This is meant to be an iterative process throughout the life of your project, so do not worry if you cannot answer all parts at the beginning—that's to be expected!

Practice makes perfect!

If you have never filled out a dataset card before, or are unsure of how to find the answers to fill in the sections, we ran a workshop to help familiarize our members with this process. In particular, the portion where we walked through filling out part of a dataset card as we did exploratory data analysis (EDA) was recorded and is available on the Imageomics YouTube Channel. Read the story of the workshop and clone the repo to follow along with the 1 hour and 15 minute lesson!

  1. The templates can also be added to your repository thorugh the website user interface (UI): Navigate to the "Model/Dataset Card" tab on your repo, select "Create Model/Dataset Card", copy and paste the template contents into the README.md file, and add your content. ↩