Struct std::AdaLN

#include <std/ml/norm.h> source

struct AdaLN {
    &FloatTensor W;
    &FloatTensor b;
    unsigned long dim;
    unsigned long condDim;
    int gated;
    float eps;
}

Fields

§&FloatTensor Wsource

[condDim, chunks * dim], init 0

§&FloatTensor bsource

[chunks * dim], init 0

§unsigned long dimsource

§unsigned long condDimsource

§int gatedsource

1: a third chunk, the residual gate

§float epssource

§struct AdaLNOut forward(const &FloatTensor x, const &FloatTensor cond) constsource · definition

x [m, dim], cond [m, condDim] (one conditioning row per sample).

§void collect_params(&Vec<&FloatTensor> out) constsource · definition

Parameterized (module.h): the trained tensors, and what a copy or a file carries.

§void collect_state(&Vec<&FloatTensor> out) constsource · definition

§void free() consumesource · definition

= drop()

§static struct AdaLN new(unsigned long dim, unsigned long condDim, int gated)source · definition