Publishing and limits

What happens when you publish a model, and what you can change afterwards.

Publishing

  1. Log in with GitHub and open Publish.
  2. Drop in your .onnx file and, if you have one, your peekml.yaml.
  3. PeekML checks that the YAML matches the model and does a test run. Any problems are listed so you can fix them.
  4. If everything passes, you're taken to the new page at /your-username/model-name.

Limits

LimitValue
One model fileUp to 100 MB
All your models togetherUp to 500 MB (shown on the Publish page)

Every visitor downloads the model file the first time they open its page, so smaller is better: under 25 MB loads in seconds, even on phones. Saving a model with int8 quantization usually makes it about 4× smaller.

Changing or removing a model

  • Improve the page: choose View on your model's page, then click the pencil icon to edit its metadata. Choose Save to update the page and prediction fields, or Cancel to discard your edits. The model's web address stays the same.
  • Delete: use Delete on your model's page. This frees its storage and can't be undone.