AffineCouplingFlowLayer<T>
Layers · AiDotNet.NeuralNetworks.Layers
An affine coupling layer with a WaveNet-like transform network (WaveGlow; Glow-TTS §3.3): the first half of the channels passes through and parameterizes a shift and log-scale applied to the second half.
Inheritance: LayerBase<T> → AffineCouplingFlowLayer<T>
Implements: IInvertibleFlowStep<T>, IShapeContract
For Beginners
Half of the data decides how to stretch and shift the other half, which keeps the step exactly reversible however complicated the deciding network is.
How It Works
Reference implementation attentions.CouplingBlock and modules.WN: x₀, x₁ = split(x);
h = start(x₀) (weight-normalized 1×1 convolution); layers gated layers
a = tanh(u) ⊙ σ(v) with [u; v] = Dropout(conv_i(h)) (weight-normalized, kernel
kernelSize, dilation rate^i), each feeding a residual and a skip output through a weight-
normalized 1×1 convolution (the last only a skip); [m; s] = end(Σ skips), an ordinary 1×1 convolution started
at zero so the layer starts as the identity; z₁ = m + e^s ⊙ x₁, log-determinant Σ s.
Constructors
| Constructor | Summary |
|---|---|
AffineCouplingFlowLayer(int channels, int hidden, int kernelSize, int dilationRate, int layers, double dropoutRate) |
Creates the coupling layer. |
Properties
| Property | Summary |
|---|---|
bool SupportsTraining { get; } |
Methods
| Method | Summary |
|---|---|
Dictionary<string, string> GetParameterRoles() |
Returns parameter roles for per-role optimizer configuration (e.g., weight decay exemption for biases). |
IReadOnlyList<ILayer<T>> GetSubLayers() |
|
IReadOnlyList<OutputAxisContract> OutputAxesFor(int inputRank) |
Derived from this type’s [TensorLayout] declarations. |
IReadOnlyList<Tensor<T>> GetTrainableParameters() |
Returns all trainable parameter tensors marked with [TrainableParameter]. |
ValueTuple<Tensor<T>, Tensor<T>> Transform(Tensor<T> input, bool reverse) |
|
void ResetState() |
|
void SetTrainableParameters(IReadOnlyList<Tensor<T>> parameters) |
Replaces trainable parameter tensors (e.g., with ParameterBuffer views). |
void ZeroGrad() |
Clears all gradient fields discovered by convention ({paramName}Gradient). |