Skip to main content
A LoRA adapter specialises a base model without swapping it out. Apply one, generate, remove it.

The namespace

Scale

scale controls how strongly the adapter pulls the base model. Leave it unset for the adapter’s own default.
Lower values blend the adapter with the base behaviour; higher values commit to it. If output degrades after applying an adapter, scale is the first thing to turn down.

What is applied right now

Removing

Removal is keyed by adapter id.

Switching between adapters

Adapters stack, so removing first is what keeps two of them from fighting.

The base model has to support it

An adapter is trained against a specific base model. Applying one to a different base either fails or produces nonsense. Check the model’s supportsLora flag before offering adapters in your interface.
Kotlin exposes the short form only: apply, remove, removeAll, list. Catalog management (registering adapters, querying a catalog, checking compatibility) is available on the Swift and Electron SDKs.