col2im_3d_f32

void col2im_3d_f32(const float * cols, float * dx, int C, int D, int H, int W, int KD, int KH, int KW, int OD, int OH, int OW, int stride_d, int stride_h, int stride_w, int pad_d, int pad_h, int pad_w, int dilation_d, int dilation_h, int dilation_w)

source edit

Single-precision 3-D col2im fold (gradient transform).

Adjoint of im2col_3d_f32: scatter-accumulates the column-major gradient back into the volumetric input gradient $dx$ . Used in the conv3d backward pass.

Parameters

colsconst float*

Column matrix, shape

(C \cdot K_D \cdot K_H \cdot K_W, O_D \cdot O_H \cdot O_W)

dxfloat*

Output gradient tensor, shape

(C, D, H, W)

; must be zeroed by the caller before this call.

C, D, H, W, KD, KH, KW, OD, OH, OWint

Shape parameters.

stride_d, stride_h, stride_w, pad_d, pad_h, pad_w, dilation_d, dilation_h, dilation_wint

Match the forward convolution.

Notes

Overlapping receptive fields accumulate by sum.

Loading…

Lucid3.0