TorchLstmCellLayer<T>

Layers ยท AiDotNet.NeuralNetworks.Layers

A PyTorch-compatible LSTM step retaining independent input and recurrent biases.

Inheritance: LayerBase<T> โ†’ TorchLstmCellLayer<T>

Implements: IShapeContract

How It Works

Both biases contribute to the gates and remain independently trainable. Folding them into one bias preserves inference but changes optimizer updates.

Constructors

Constructor Summary
TorchLstmCellLayer(int inputSize, int hiddenSize) Creates a cell with i, f, g, o gate ordering and two distinct biases.

Properties

Property Summary
bool RequiresMultipleInputs { get; }
bool SupportsTraining { get; }

Methods

Method Summary
IReadOnlyList<ILayer<T>> GetSubLayers()
IReadOnlyList<OutputAxisContract> OutputAxesFor(int inputRank)
Tensor<T> Forward(Tensor<T> input, Tensor<T> state)
ValueTuple<Tensor<T>, Tensor<T>> SplitState(Tensor<T> state)
void ResetState()