IParameterSurfaceLifecycle

Interfaces · AiDotNet.Models.Parameters

Implemented by parameter sources whose structure or storage becomes ready in lifecycle stages.

Known implementations: ComponentAccessorParameterSource<T>, ComponentCollectionParameterSource<T>, DiffusionModelBase<T>, LayerBase<T>, NeuralNetworkBase<T>, NoisePredictorBase<T>, VAEModelBase<T>

How It Works

This is a capability contract, not a trainability inference. A source may contain trainable, frozen, fitted, buffered, or external slots; their explicit manifest roles remain authoritative.

Implementations must be idempotent. Describe must not allocate parameter storage, while concrete intents may materialize only slots whose shapes are already resolved. Truly data-dependent shapes remain deferred and produce the normal readiness diagnostic rather than being guessed from a dummy forward.

Methods

Method Summary
void PrepareParameterSurface(ParameterSurfaceIntent intent) Advances this source only as far as the requested operation requires.