About Glitchdata

This site distributes datasets and trained models, publishes news on every release, and hosts courses on putting them to work.

How it works

• Datasets are documented data with a card, links to the files and, for tables, a preview of the first rows. We don't store the data: every download comes straight from the publisher that hosts it.
• Models are trained models with the framework, task and parameter count you need to run them.
• Anyone can share a dataset or model under their own name. Every submission, and every change to one, is checked by an admin before it goes public.
• News covers new releases, updates to existing downloads, and what is coming next.
• Courses are made up of ordered lessons. Enrol in one and your progress is saved as you go.
• Preview lessons are open to everyone, so you can see what a course is like before signing up.
• Guides are standalone articles — shorter reading that sits alongside the courses.
• Topics group everything by subject, so related courses and guides stay together.

Getting started

Create a free account, browse the catalogue, and enrol in whatever looks useful. Your dashboard keeps everything you are part-way through in one list.