Conv5x5on8Relu< INP_H, INP_W, OUT_W, OUT_W_PAD, STEP_H, STEP_W, B, C, M, KH, KW, GROUP, IS_RELU > Class Template Reference

Vector implementation for 5x5 BCHW, stores weights and biases, requires KH==KW==5, INP_W%4==0, OUT_W_PAD%8=0, STEP_H==STEP_W==1, GROUP==1, assumes weights are padded to MxCx5x8, Conv5x5on8Relu<28,28,24,24,1,1,1,2,4,5,5,1,1> total = 21401.

#include <conv.h>

Public Member Functions
	Conv5x5on8Relu (float(&w)[M *C *KH *8], float(&b)[M])
void	filter (input_window< float > *in, output_window< float > *out)

Static Public Member Functions
static void	registerKernelClass ()

---

The documentation for this class was generated from the following files:

• design/aie_src/conv.h
• design/aie_src/conv.cc