Detection is accelerated by 7 percent (fused conv and batch_norm layers)

AlexeyAB

2018-04-03 1b2c70f82a9ec88c60bb277f9f5f0073c96691e7

 src/cpu_gemm.c

@@ -1,4 +1,4 @@
#include "mini_blas.h"
//#include "mini_blas.h"

void cpu_gemm_nn(int TA, int TB, int M, int N, int K, float ALPHA, 
        float *A, int lda,