LayerPass

classin packageklyn.math.nn.layers

public class LayerPass<T extends AbstractFloat = Float>

LayerPass

Explicit per-batch forward context, released by the trainer after backward.

Example

context = layer.forwardTraining(batch)

Properties

Modifier and Type	Member	Description
public readonly property	inputs `inputs as NDArray<T>`	Original input tensor.
public readonly property	outputs `outputs as NDArray<T>`	Output tensor, including any configured activation.
public readonly property	saved `saved as NDArray<T>`	Optional layer-specific intermediate, such as affine pre-activations.

Constructors

Modifier and Type	Member	Description
public	LayerPass `LayerPass(inputs as NDArray<T>, outputs as NDArray<T>, saved as NDArray<T> = null)`	Retains tensors without copying their storage.