fn
max_pool3d_backward_f64
void max_pool3d_backward_f64(const double * g, const int * argmax, double * dx, int B, int C, int D, int H, int W, int OD, int OH, int OW)Double-precision counterpart to max_pool3d_backward_f32.
Parameters
gconst double*Upstream gradient of shape
(B, C, OD, OH, OW).argmaxconst int32_t*Per-output winning flat 3-D index.
dxdouble*Output gradient of shape
(B, C, D, H, W); accumulated.B, C, D, H, W, OD, OH, OWintLayout extents.