layer_norm_backward_f64

void layer_norm_backward_f64(const double * x, const double * gamma, const double * saved_mean, const double * saved_rstd, const double * g, double * dx, double * dgamma, double * dbeta, int outer, int N)

source edit

Double-precision counterpart to layer_norm_backward_f32.

Parameters

xconst double*

Original forward input of shape (outer, N).

gammaconst double*

Affine weight of shape (N,).

saved_mean, saved_rstdconst double*

Per-row statistics from the matching forward call.

gconst double*

Upstream gradient of shape (outer, N).

dxdouble*

Output gradient w.r.t. x; overwritten.

dgamma, dbetadouble*

Output gradients of shape (N,); accumulated — caller must zero them before the first invocation.

outer, Nstd::size_t

Layout extents.