I made a HF Space which turns video into various types of frames (rgb and depth so far), and saves them as a tar.gz for download. Can I just send that tar.gz to a dataset that I create, and allow the dataset to grow with each submission of tar.gz file?
Any words of wisdom could save me hours of research, and would be super great.
Hi! Yes, this should work. The only issue is that our built-in loaders (CSV, JSON, …) don’t support tar.gz files, so you’ll have to create a loading script for the dataset to parse them.
Dataset loading script, alright. Is that some lines of code that I can add in my Huggingface Space?
I haven’t found any examples that have worked in Huggingface Spaces for use in creating Huggingface Datasets, or Huggingface Models, or Finetuning Models.
Where do I run the dataset loading script?