OnDevice Classifier Support
For help, include the app build, iPhone or iPad model, iOS version, selected classifier, and the non-sensitive error text. Do not send API keys, Hugging Face tokens, or private input examples.
Support
Common Topics
Use the current app build and check the model’s device and OS requirements. For Julia-1 or Laya GGUF models, select a supported variant from the built-in library. If memory is tight, unload the current model and choose a smaller or 8-bit variant.
Downloads pause when the app goes to the background. Reopen the library and resume the transfer. Check network access and free storage; use Retry if the server interrupted it.
Add a read-access token from your own Hugging Face account after obtaining access to the repository. Public catalog downloads do not require a token.
Check the provider, endpoint, model name and API key in the connection settings. Provider availability, quotas and charges are managed by that provider.
Image classifiers use their built-in class list. Switch to a text decision model for editable categories, scoring or yes/no tasks.