void save_convolutional_weights_binary(layer l, FILE *fp)
{
#ifdef GPU
if(gpu_index >= 0){
pull_convolutional_layer(l);
}
#endif
binarize_weights(l.weights, l.n, l.c*l.size*l.size, l.binary_weights);
int size = l.c*l.size*l.size;
int i, j, k;
fwrite(l.biases, sizeof(float), l.n, fp);
if (l.batch_normalize){
fwrite(l.scales, sizeof(float), l.n, fp);
fwrite(l.rolling_mean, sizeof(float), l.n, fp);
fwrite(l.rolling_variance, sizeof(float), l.n, fp);
}
for(i = 0; i < l.n; ++i){
float mean = l.binary_weights[i*size];
if(mean < 0) mean = -mean;
fwrite(&mean, sizeof(float), 1, fp);
for(j = 0; j < size/8; ++j){
int index = i*size + j*8;
unsigned char c = 0;
for(k = 0; k < 8; ++k){
if (j*8 + k >= size) break;
if (l.binary_weights[index + k] > 0) c = (c | 1<<k);
}
fwrite(&c, sizeof(char), 1, fp);
}
}
}
void forward_convolutional_layer(convolutional_layer l, network_state state)
{
int out_h = convolutional_out_height(l);
int out_w = convolutional_out_width(l);
int i;
fill_cpu(l.outputs*l.batch, 0, l.output, 1);
if(l.xnor){
binarize_weights(l.weights, l.n, l.c*l.size*l.size, l.binary_weights);
swap_binary(&l);
binarize_cpu(state.input, l.c*l.h*l.w*l.batch, l.binary_input);
state.input = l.binary_input;
}
int m = l.n;
int k = l.size*l.size*l.c;
int n = out_h*out_w;
float *a = l.weights;
float *b = state.workspace;
float *c = l.output;
for(i = 0; i < l.batch; ++i){
im2col_cpu(state.input, l.c, l.h, l.w,
l.size, l.stride, l.pad, b);
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
c += n*m;
state.input += l.c*l.h*l.w;
}
if(l.batch_normalize){
forward_batchnorm_layer(l, state);
}
add_bias(l.output, l.biases, l.batch, l.n, out_h*out_w);
activate_array(l.output, m*n*l.batch, l.activation);
if(l.binary || l.xnor) swap_binary(&l);
}
void forward_convolutional_layer(convolutional_layer l, network net)
{
int i, j;
fill_cpu(l.outputs*l.batch, 0, l.output, 1);
if(l.xnor){
binarize_weights(l.weights, l.n, l.c/l.groups*l.size*l.size, l.binary_weights);
swap_binary(&l);
binarize_cpu(net.input, l.c*l.h*l.w*l.batch, l.binary_input);
net.input = l.binary_input;
}
int m = l.n/l.groups;
int k = l.size*l.size*l.c/l.groups;
int n = l.out_w*l.out_h;
for(i = 0; i < l.batch; ++i){
for(j = 0; j < l.groups; ++j){
float *a = l.weights + j*l.nweights/l.groups;
float *b = net.workspace;
float *c = l.output + (i*l.groups + j)*n*m;
im2col_cpu(net.input + (i*l.groups + j)*l.c/l.groups*l.h*l.w,
l.c/l.groups, l.h, l.w, l.size, l.stride, l.pad, b);
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
}
}
if(l.batch_normalize){
forward_batchnorm_layer(l, net);
} else {
add_bias(l.output, l.biases, l.batch, l.n, l.out_h*l.out_w);
}
activate_array(l.output, l.outputs*l.batch, l.activation);
if(l.binary || l.xnor) swap_binary(&l);
}
