From 054e2b1954aafb15b0e983180dda309cfd5d831f Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Thu, 12 May 2016 20:36:11 +0000
Subject: [PATCH] not sure
---
src/convolutional_layer.c | 132 ++++++++++++++++++++++++++++++++++++++------
1 files changed, 114 insertions(+), 18 deletions(-)
diff --git a/src/convolutional_layer.c b/src/convolutional_layer.c
index b9fd3c9..d76dfcd 100644
--- a/src/convolutional_layer.c
+++ b/src/convolutional_layer.c
@@ -1,5 +1,6 @@
#include "convolutional_layer.h"
#include "utils.h"
+#include "batchnorm_layer.h"
#include "im2col.h"
#include "col2im.h"
#include "blas.h"
@@ -7,6 +8,52 @@
#include <stdio.h>
#include <time.h>
+void swap_binary(convolutional_layer *l)
+{
+ float *swap = l->filters;
+ l->filters = l->binary_filters;
+ l->binary_filters = swap;
+
+ #ifdef GPU
+ swap = l->filters_gpu;
+ l->filters_gpu = l->binary_filters_gpu;
+ l->binary_filters_gpu = swap;
+ #endif
+}
+
+void binarize_filters2(float *filters, int n, int size, char *binary, float *scales)
+{
+ int i, k, f;
+ for(f = 0; f < n; ++f){
+ float mean = 0;
+ for(i = 0; i < size; ++i){
+ mean += fabs(filters[f*size + i]);
+ }
+ mean = mean / size;
+ scales[f] = mean;
+ for(i = 0; i < size/8; ++i){
+ binary[f*size + i] = (filters[f*size + i] > 0) ? 1 : 0;
+ for(k = 0; k < 8; ++k){
+ }
+ }
+ }
+}
+
+void binarize_filters(float *filters, int n, int size, float *binary)
+{
+ int i, f;
+ for(f = 0; f < n; ++f){
+ float mean = 0;
+ for(i = 0; i < size; ++i){
+ mean += fabs(filters[f*size + i]);
+ }
+ mean = mean / size;
+ for(i = 0; i < size; ++i){
+ binary[f*size + i] = (filters[f*size + i] > 0) ? mean : -mean;
+ }
+ }
+}
+
int convolutional_out_height(convolutional_layer l)
{
int h = l.h;
@@ -41,7 +88,7 @@
return float_to_image(w,h,c,l.delta);
}
-convolutional_layer make_convolutional_layer(int batch, int h, int w, int c, int n, int size, int stride, int pad, ACTIVATION activation, int batch_normalize)
+convolutional_layer make_convolutional_layer(int batch, int h, int w, int c, int n, int size, int stride, int pad, ACTIVATION activation, int batch_normalize, int binary, int xnor)
{
int i;
convolutional_layer l = {0};
@@ -51,6 +98,7 @@
l.w = w;
l.c = c;
l.n = n;
+ l.binary = binary;
l.batch = batch;
l.stride = stride;
l.size = size;
@@ -65,7 +113,7 @@
// float scale = 1./sqrt(size*size*c);
float scale = sqrt(2./(size*size*c));
- for(i = 0; i < c*n*size*size; ++i) l.filters[i] = 2*scale*rand_uniform() - scale;
+ for(i = 0; i < c*n*size*size; ++i) l.filters[i] = scale*rand_uniform(-1, 1);
int out_h = convolutional_out_height(l);
int out_w = convolutional_out_width(l);
l.out_h = out_h;
@@ -78,6 +126,12 @@
l.output = calloc(l.batch*out_h * out_w * n, sizeof(float));
l.delta = calloc(l.batch*out_h * out_w * n, sizeof(float));
+ if(binary){
+ l.binary_filters = calloc(c*n*size*size, sizeof(float));
+ l.cfilters = calloc(c*n*size*size, sizeof(char));
+ l.scales = calloc(n, sizeof(float));
+ }
+
if(batch_normalize){
l.scales = calloc(n, sizeof(float));
l.scale_updates = calloc(n, sizeof(float));
@@ -86,9 +140,8 @@
}
l.mean = calloc(n, sizeof(float));
- l.spatial_mean = calloc(n*l.batch, sizeof(float));
-
l.variance = calloc(n, sizeof(float));
+
l.rolling_mean = calloc(n, sizeof(float));
l.rolling_variance = calloc(n, sizeof(float));
}
@@ -107,6 +160,15 @@
l.delta_gpu = cuda_make_array(l.delta, l.batch*out_h*out_w*n);
l.output_gpu = cuda_make_array(l.output, l.batch*out_h*out_w*n);
+ if(binary){
+ l.binary_filters_gpu = cuda_make_array(l.filters, c*n*size*size);
+ }
+ if(xnor){
+ l.binary_filters_gpu = cuda_make_array(l.filters, c*n*size*size);
+ l.binary_input_gpu = cuda_make_array(0, l.inputs*l.batch);
+ }
+ l.xnor = xnor;
+
if(batch_normalize){
l.mean_gpu = cuda_make_array(l.mean, n);
l.variance_gpu = cuda_make_array(l.variance, n);
@@ -114,12 +176,6 @@
l.rolling_mean_gpu = cuda_make_array(l.mean, n);
l.rolling_variance_gpu = cuda_make_array(l.variance, n);
- l.spatial_mean_gpu = cuda_make_array(l.spatial_mean, n*l.batch);
- l.spatial_variance_gpu = cuda_make_array(l.spatial_mean, n*l.batch);
-
- l.spatial_mean_delta_gpu = cuda_make_array(l.spatial_mean, n*l.batch);
- l.spatial_variance_delta_gpu = cuda_make_array(l.spatial_mean, n*l.batch);
-
l.mean_delta_gpu = cuda_make_array(l.mean, n);
l.variance_delta_gpu = cuda_make_array(l.variance, n);
@@ -148,7 +204,7 @@
void test_convolutional_layer()
{
- convolutional_layer l = make_convolutional_layer(1, 5, 5, 3, 2, 5, 2, 1, LEAKY, 1);
+ convolutional_layer l = make_convolutional_layer(1, 5, 5, 3, 2, 5, 2, 1, LEAKY, 1, 0, 0);
l.batch_normalize = 1;
float data[] = {1,1,1,1,1,
1,1,1,1,1,
@@ -201,13 +257,25 @@
#endif
}
-void bias_output(float *output, float *biases, int batch, int n, int size)
+void add_bias(float *output, float *biases, int batch, int n, int size)
{
int i,j,b;
for(b = 0; b < batch; ++b){
for(i = 0; i < n; ++i){
for(j = 0; j < size; ++j){
- output[(b*n + i)*size + j] = biases[i];
+ output[(b*n + i)*size + j] += biases[i];
+ }
+ }
+ }
+}
+
+void scale_bias(float *output, float *scales, int batch, int n, int size)
+{
+ int i,j,b;
+ for(b = 0; b < batch; ++b){
+ for(i = 0; i < n; ++i){
+ for(j = 0; j < size; ++j){
+ output[(b*n + i)*size + j] *= scales[i];
}
}
}
@@ -223,13 +291,42 @@
}
}
-void forward_convolutional_layer(const convolutional_layer l, network_state state)
+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;
- bias_output(l.output, l.biases, l.batch, l.n, out_h*out_w);
+ fill_cpu(l.outputs*l.batch, 0, l.output, 1);
+ /*
+ if(l.binary){
+ binarize_filters(l.filters, l.n, l.c*l.size*l.size, l.binary_filters);
+ binarize_filters2(l.filters, l.n, l.c*l.size*l.size, l.cfilters, l.scales);
+ swap_binary(&l);
+ }
+ */
+
+ if(l.binary){
+ int m = l.n;
+ int k = l.size*l.size*l.c;
+ int n = out_h*out_w;
+
+ char *a = l.cfilters;
+ float *b = l.col_image;
+ 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_bin(m,n,k,1,a,k,b,n,c,n);
+ c += n*m;
+ state.input += l.c*l.h*l.w;
+ }
+ scale_bias(l.output, l.scales, l.batch, l.n, out_h*out_w);
+ add_bias(l.output, l.biases, l.batch, l.n, out_h*out_w);
+ activate_array(l.output, m*n*l.batch, l.activation);
+ return;
+ }
int m = l.n;
int k = l.size*l.size*l.c;
@@ -248,10 +345,9 @@
}
if(l.batch_normalize){
- mean_cpu(l.output, l.batch, l.n, l.out_h*l.out_w, l.mean);
- variance_cpu(l.output, l.mean, l.batch, l.n, l.out_h*l.out_w, l.variance);
- normalize_cpu(l.output, l.mean, l.variance, l.batch, l.n, l.out_h*l.out_w);
+ 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);
}
--
Gitblit v1.10.0