From f7a17f82eb43de864a4f980f235055da9685eef8 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Wed, 29 Jan 2014 00:28:42 +0000
Subject: [PATCH] Convolutional layers working w/ matrices
---
src/mini_blas.c | 93 +++++
Makefile | 2
src/convolutional_layer.h | 24
src/image.c | 68 ++--
src/mini_blas.h | 24 +
src/tests.c | 129 +++-----
src/image.h | 20
src/maxpool_layer.c | 28 +-
src/utils.c | 44 +-
nist_basic.cfg | 2
src/network.c | 70 ++--
src/matrix.c | 20
src/maxpool_layer.h | 8
src/matrix.h | 6
src/softmax_layer.h | 8
src/utils.h | 20
src/network.h | 24
src/connected_layer.c | 58 ++--
src/connected_layer.h | 24
src/data.c | 20
src/softmax_layer.c | 28 +
src/option_list.h | 2
src/convolutional_layer.c | 90 ++++--
src/option_list.c | 2
src/activations.h | 4
src/activations.c | 4
26 files changed, 459 insertions(+), 363 deletions(-)
diff --git a/Makefile b/Makefile
index dc08b46..fda7d88 100644
--- a/Makefile
+++ b/Makefile
@@ -1,6 +1,6 @@
CC=gcc
COMMON=-Wall `pkg-config --cflags opencv`
-CFLAGS= $(COMMON) -Ofast -ffast-math -flto
+CFLAGS= $(COMMON) -O3 -ffast-math -flto
UNAME = $(shell uname)
ifeq ($(UNAME), Darwin)
COMMON += -isystem /usr/local/Cellar/opencv/2.4.6.1/include/opencv -isystem /usr/local/Cellar/opencv/2.4.6.1/include
diff --git a/nist_basic.cfg b/nist_basic.cfg
index f5ea0a3..7142735 100644
--- a/nist_basic.cfg
+++ b/nist_basic.cfg
@@ -3,7 +3,7 @@
height=28
channels=1
filters=20
-size=5
+size=11
stride=1
activation=linear
diff --git a/src/activations.c b/src/activations.c
index b8bb79d..cc923d0 100644
--- a/src/activations.c
+++ b/src/activations.c
@@ -15,7 +15,7 @@
return RELU;
}
-double activate(double x, ACTIVATION a){
+float activate(float x, ACTIVATION a){
switch(a){
case LINEAR:
return x;
@@ -30,7 +30,7 @@
}
return 0;
}
-double gradient(double x, ACTIVATION a){
+float gradient(float x, ACTIVATION a){
switch(a){
case LINEAR:
return 1;
diff --git a/src/activations.h b/src/activations.h
index 889453f..fb2c54f 100644
--- a/src/activations.h
+++ b/src/activations.h
@@ -7,8 +7,8 @@
ACTIVATION get_activation(char *s);
-double activate(double x, ACTIVATION a);
-double gradient(double x, ACTIVATION a);
+float activate(float x, ACTIVATION a);
+float gradient(float x, ACTIVATION a);
#endif
diff --git a/src/connected_layer.c b/src/connected_layer.c
index 6871b2e..5f6631c 100644
--- a/src/connected_layer.c
+++ b/src/connected_layer.c
@@ -15,19 +15,19 @@
layer->inputs = inputs;
layer->outputs = outputs;
- layer->output = calloc(outputs, sizeof(double*));
- layer->delta = calloc(outputs, sizeof(double*));
+ layer->output = calloc(outputs, sizeof(float*));
+ layer->delta = calloc(outputs, sizeof(float*));
- layer->weight_updates = calloc(inputs*outputs, sizeof(double));
- layer->weight_momentum = calloc(inputs*outputs, sizeof(double));
- layer->weights = calloc(inputs*outputs, sizeof(double));
- double scale = 2./inputs;
+ layer->weight_updates = calloc(inputs*outputs, sizeof(float));
+ layer->weight_momentum = calloc(inputs*outputs, sizeof(float));
+ layer->weights = calloc(inputs*outputs, sizeof(float));
+ float scale = 2./inputs;
for(i = 0; i < inputs*outputs; ++i)
layer->weights[i] = rand_normal()*scale;
- layer->bias_updates = calloc(outputs, sizeof(double));
- layer->bias_momentum = calloc(outputs, sizeof(double));
- layer->biases = calloc(outputs, sizeof(double));
+ layer->bias_updates = calloc(outputs, sizeof(float));
+ layer->bias_momentum = calloc(outputs, sizeof(float));
+ layer->biases = calloc(outputs, sizeof(float));
for(i = 0; i < outputs; ++i)
//layer->biases[i] = rand_normal()*scale + scale;
layer->biases[i] = 0;
@@ -36,7 +36,7 @@
return layer;
}
-void update_connected_layer(connected_layer layer, double step, double momentum, double decay)
+void update_connected_layer(connected_layer layer, float step, float momentum, float decay)
{
int i;
for(i = 0; i < layer.outputs; ++i){
@@ -47,27 +47,27 @@
layer.weight_momentum[i] = step*(layer.weight_updates[i] - decay*layer.weights[i]) + momentum*layer.weight_momentum[i];
layer.weights[i] += layer.weight_momentum[i];
}
- memset(layer.bias_updates, 0, layer.outputs*sizeof(double));
- memset(layer.weight_updates, 0, layer.outputs*layer.inputs*sizeof(double));
+ memset(layer.bias_updates, 0, layer.outputs*sizeof(float));
+ memset(layer.weight_updates, 0, layer.outputs*layer.inputs*sizeof(float));
}
-void forward_connected_layer(connected_layer layer, double *input)
+void forward_connected_layer(connected_layer layer, float *input)
{
int i;
- memcpy(layer.output, layer.biases, layer.outputs*sizeof(double));
+ memcpy(layer.output, layer.biases, layer.outputs*sizeof(float));
int m = 1;
int k = layer.inputs;
int n = layer.outputs;
- double *a = input;
- double *b = layer.weights;
- double *c = layer.output;
+ float *a = input;
+ float *b = layer.weights;
+ float *c = layer.output;
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
for(i = 0; i < layer.outputs; ++i){
layer.output[i] = activate(layer.output[i], layer.activation);
}
}
-void learn_connected_layer(connected_layer layer, double *input)
+void learn_connected_layer(connected_layer layer, float *input)
{
int i;
for(i = 0; i < layer.outputs; ++i){
@@ -77,28 +77,28 @@
int m = layer.inputs;
int k = 1;
int n = layer.outputs;
- double *a = input;
- double *b = layer.delta;
- double *c = layer.weight_updates;
+ float *a = input;
+ float *b = layer.delta;
+ float *c = layer.weight_updates;
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
}
-void backward_connected_layer(connected_layer layer, double *input, double *delta)
+void backward_connected_layer(connected_layer layer, float *input, float *delta)
{
- memset(delta, 0, layer.inputs*sizeof(double));
+ memset(delta, 0, layer.inputs*sizeof(float));
int m = layer.inputs;
int k = layer.outputs;
int n = 1;
- double *a = layer.weights;
- double *b = layer.delta;
- double *c = delta;
+ float *a = layer.weights;
+ float *b = layer.delta;
+ float *c = delta;
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
}
/*
- void forward_connected_layer(connected_layer layer, double *input)
+ void forward_connected_layer(connected_layer layer, float *input)
{
int i, j;
for(i = 0; i < layer.outputs; ++i){
@@ -109,7 +109,7 @@
layer.output[i] = activate(layer.output[i], layer.activation);
}
}
- void learn_connected_layer(connected_layer layer, double *input)
+ void learn_connected_layer(connected_layer layer, float *input)
{
int i, j;
for(i = 0; i < layer.outputs; ++i){
@@ -120,7 +120,7 @@
}
}
}
- void backward_connected_layer(connected_layer layer, double *input, double *delta)
+ void backward_connected_layer(connected_layer layer, float *input, float *delta)
{
int i, j;
diff --git a/src/connected_layer.h b/src/connected_layer.h
index 05fb261..ce0181d 100644
--- a/src/connected_layer.h
+++ b/src/connected_layer.h
@@ -6,17 +6,17 @@
typedef struct{
int inputs;
int outputs;
- double *weights;
- double *biases;
+ float *weights;
+ float *biases;
- double *weight_updates;
- double *bias_updates;
+ float *weight_updates;
+ float *bias_updates;
- double *weight_momentum;
- double *bias_momentum;
+ float *weight_momentum;
+ float *bias_momentum;
- double *output;
- double *delta;
+ float *output;
+ float *delta;
ACTIVATION activation;
@@ -24,10 +24,10 @@
connected_layer *make_connected_layer(int inputs, int outputs, ACTIVATION activation);
-void forward_connected_layer(connected_layer layer, double *input);
-void backward_connected_layer(connected_layer layer, double *input, double *delta);
-void learn_connected_layer(connected_layer layer, double *input);
-void update_connected_layer(connected_layer layer, double step, double momentum, double decay);
+void forward_connected_layer(connected_layer layer, float *input);
+void backward_connected_layer(connected_layer layer, float *input, float *delta);
+void learn_connected_layer(connected_layer layer, float *input);
+void update_connected_layer(connected_layer layer, float step, float momentum, float decay);
#endif
diff --git a/src/convolutional_layer.c b/src/convolutional_layer.c
index 53eb7bf..cdfe9e1 100644
--- a/src/convolutional_layer.c
+++ b/src/convolutional_layer.c
@@ -9,7 +9,7 @@
h = layer.out_h;
w = layer.out_w;
c = layer.n;
- return double_to_image(h,w,c,layer.output);
+ return float_to_image(h,w,c,layer.output);
}
image get_convolutional_delta(convolutional_layer layer)
@@ -18,7 +18,7 @@
h = layer.out_h;
w = layer.out_w;
c = layer.n;
- return double_to_image(h,w,c,layer.delta);
+ return float_to_image(h,w,c,layer.delta);
}
convolutional_layer *make_convolutional_layer(int h, int w, int c, int n, int size, int stride, ACTIVATION activation)
@@ -34,14 +34,14 @@
layer->stride = stride;
layer->size = size;
- layer->filters = calloc(c*n*size*size, sizeof(double));
- layer->filter_updates = calloc(c*n*size*size, sizeof(double));
- layer->filter_momentum = calloc(c*n*size*size, sizeof(double));
+ layer->filters = calloc(c*n*size*size, sizeof(float));
+ layer->filter_updates = calloc(c*n*size*size, sizeof(float));
+ layer->filter_momentum = calloc(c*n*size*size, sizeof(float));
- layer->biases = calloc(n, sizeof(double));
- layer->bias_updates = calloc(n, sizeof(double));
- layer->bias_momentum = calloc(n, sizeof(double));
- double scale = 2./(size*size);
+ layer->biases = calloc(n, sizeof(float));
+ layer->bias_updates = calloc(n, sizeof(float));
+ layer->bias_momentum = calloc(n, sizeof(float));
+ float scale = 2./(size*size);
for(i = 0; i < c*n*size*size; ++i) layer->filters[i] = rand_normal()*scale;
for(i = 0; i < n; ++i){
//layer->biases[i] = rand_normal()*scale + scale;
@@ -50,9 +50,9 @@
out_h = (h-size)/stride + 1;
out_w = (w-size)/stride + 1;
- layer->col_image = calloc(out_h*out_w*size*size*c, sizeof(double));
- layer->output = calloc(out_h * out_w * n, sizeof(double));
- layer->delta = calloc(out_h * out_w * n, sizeof(double));
+ layer->col_image = calloc(out_h*out_w*size*size*c, sizeof(float));
+ layer->output = calloc(out_h * out_w * n, sizeof(float));
+ layer->delta = calloc(out_h * out_w * n, sizeof(float));
layer->activation = activation;
layer->out_h = out_h;
layer->out_w = out_w;
@@ -63,18 +63,18 @@
return layer;
}
-void forward_convolutional_layer(const convolutional_layer layer, double *in)
+void forward_convolutional_layer(const convolutional_layer layer, float *in)
{
int m = layer.n;
int k = layer.size*layer.size*layer.c;
int n = ((layer.h-layer.size)/layer.stride + 1)*
((layer.w-layer.size)/layer.stride + 1);
- memset(layer.output, 0, m*n*sizeof(double));
+ memset(layer.output, 0, m*n*sizeof(float));
- double *a = layer.filters;
- double *b = layer.col_image;
- double *c = layer.output;
+ float *a = layer.filters;
+ float *b = layer.col_image;
+ float *c = layer.output;
im2col_cpu(in, layer.c, layer.h, layer.w, layer.size, layer.stride, b);
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
@@ -94,7 +94,7 @@
int i,j;
int size = layer.out_h*layer.out_w;
for(i = 0; i < layer.n; ++i){
- double sum = 0;
+ float sum = 0;
for(j = 0; j < size; ++j){
sum += layer.delta[j+i*size];
}
@@ -111,14 +111,33 @@
int k = ((layer.h-layer.size)/layer.stride + 1)*
((layer.w-layer.size)/layer.stride + 1);
- double *a = layer.delta;
- double *b = layer.col_image;
- double *c = layer.filter_updates;
+ float *a = layer.delta;
+ float *b = layer.col_image;
+ float *c = layer.filter_updates;
gemm(0,1,m,n,k,1,a,k,b,k,1,c,n);
}
-void update_convolutional_layer(convolutional_layer layer, double step, double momentum, double decay)
+void backward_convolutional_layer(convolutional_layer layer, float *delta)
+{
+ int m = layer.size*layer.size*layer.c;
+ int k = layer.n;
+ int n = ((layer.h-layer.size)/layer.stride + 1)*
+ ((layer.w-layer.size)/layer.stride + 1);
+
+ float *a = layer.filters;
+ float *b = layer.delta;
+ float *c = layer.col_image;
+
+
+ memset(c, 0, m*n*sizeof(float));
+ gemm(1,0,m,n,k,1,a,m,b,n,1,c,n);
+
+ memset(delta, 0, layer.h*layer.w*layer.c*sizeof(float));
+ col2im_cpu(c, layer.c, layer.h, layer.w, layer.size, layer.stride, delta);
+}
+
+void update_convolutional_layer(convolutional_layer layer, float step, float momentum, float decay)
{
int i;
int size = layer.size*layer.size*layer.c*layer.n;
@@ -133,9 +152,9 @@
}
/*
-void backward_convolutional_layer2(convolutional_layer layer, double *input, double *delta)
+void backward_convolutional_layer2(convolutional_layer layer, float *input, float *delta)
{
- image in_delta = double_to_image(layer.h, layer.w, layer.c, delta);
+ image in_delta = float_to_image(layer.h, layer.w, layer.c, delta);
image out_delta = get_convolutional_delta(layer);
int i,j;
for(i = 0; i < layer.n; ++i){
@@ -156,10 +175,10 @@
}
-void learn_convolutional_layer(convolutional_layer layer, double *input)
+void learn_convolutional_layer(convolutional_layer layer, float *input)
{
int i;
- image in_image = double_to_image(layer.h, layer.w, layer.c, input);
+ image in_image = float_to_image(layer.h, layer.w, layer.c, input);
image out_delta = get_convolutional_delta(layer);
gradient_delta_convolutional_layer(layer);
for(i = 0; i < layer.n; ++i){
@@ -168,7 +187,7 @@
}
}
-void update_convolutional_layer(convolutional_layer layer, double step, double momentum, double decay)
+void update_convolutional_layer(convolutional_layer layer, float step, float momentum, float decay)
{
int i,j;
for(i = 0; i < layer.n; ++i){
@@ -190,21 +209,28 @@
void test_convolutional_layer()
{
convolutional_layer l = *make_convolutional_layer(4,4,1,1,3,1,LINEAR);
- double input[] = {1,2,3,4,
+ float input[] = {1,2,3,4,
5,6,7,8,
9,10,11,12,
13,14,15,16};
- double filter[] = {.5, 0, .3,
+ float filter[] = {.5, 0, .3,
0 , 1, 0,
.2 , 0, 1};
- double delta[] = {1, 2,
+ float delta[] = {1, 2,
3, 4};
+ float in_delta[] = {.5,1,.3,.6,
+ 5,6,7,8,
+ 9,10,11,12,
+ 13,14,15,16};
l.filters = filter;
forward_convolutional_layer(l, input);
l.delta = delta;
learn_convolutional_layer(l);
- image filter_updates = double_to_image(3,3,1,l.filter_updates);
+ image filter_updates = float_to_image(3,3,1,l.filter_updates);
print_image(filter_updates);
+ printf("Delta:\n");
+ backward_convolutional_layer(l, in_delta);
+ pm(4,4,in_delta);
}
image get_convolutional_filter(convolutional_layer layer, int i)
@@ -212,7 +238,7 @@
int h = layer.size;
int w = layer.size;
int c = layer.c;
- return double_to_image(h,w,c,layer.filters+i*h*w*c);
+ return float_to_image(h,w,c,layer.filters+i*h*w*c);
}
void visualize_convolutional_layer(convolutional_layer layer, char *window)
diff --git a/src/convolutional_layer.h b/src/convolutional_layer.h
index e2e6cdc..c4de24e 100644
--- a/src/convolutional_layer.h
+++ b/src/convolutional_layer.h
@@ -10,28 +10,28 @@
int n;
int size;
int stride;
- double *filters;
- double *filter_updates;
- double *filter_momentum;
+ float *filters;
+ float *filter_updates;
+ float *filter_momentum;
- double *biases;
- double *bias_updates;
- double *bias_momentum;
+ float *biases;
+ float *bias_updates;
+ float *bias_momentum;
- double *col_image;
- double *delta;
- double *output;
+ float *col_image;
+ float *delta;
+ float *output;
ACTIVATION activation;
} convolutional_layer;
convolutional_layer *make_convolutional_layer(int h, int w, int c, int n, int size, int stride, ACTIVATION activation);
-void forward_convolutional_layer(const convolutional_layer layer, double *in);
+void forward_convolutional_layer(const convolutional_layer layer, float *in);
void learn_convolutional_layer(convolutional_layer layer);
-void update_convolutional_layer(convolutional_layer layer, double step, double momentum, double decay);
+void update_convolutional_layer(convolutional_layer layer, float step, float momentum, float decay);
void visualize_convolutional_layer(convolutional_layer layer, char *window);
-//void backward_convolutional_layer(convolutional_layer layer, double *input, double *delta);
+void backward_convolutional_layer(convolutional_layer layer, float *delta);
//void backpropagate_convolutional_layer_convolve(image input, convolutional_layer layer);
//void visualize_convolutional_filters(convolutional_layer layer, char *window);
diff --git a/src/data.c b/src/data.c
index 0b396d7..2c5932b 100644
--- a/src/data.c
+++ b/src/data.c
@@ -19,10 +19,10 @@
return lines;
}
-void fill_truth(char *path, char **labels, int k, double *truth)
+void fill_truth(char *path, char **labels, int k, float *truth)
{
int i;
- memset(truth, 0, k*sizeof(double));
+ memset(truth, 0, k*sizeof(float));
for(i = 0; i < k; ++i){
if(strstr(path, labels[i])){
truth[i] = 1;
@@ -36,7 +36,7 @@
data d;
d.shallow = 0;
d.X.rows = n;
- d.X.vals = calloc(d.X.rows, sizeof(double*));
+ d.X.vals = calloc(d.X.rows, sizeof(float*));
d.y = make_matrix(n, k);
for(i = 0; i < n; ++i){
@@ -106,8 +106,8 @@
data d;
d.shallow = 0;
matrix X = csv_to_matrix(filename);
- double *truth_1d = pop_column(&X, target);
- double **truth = one_hot_encode(truth_1d, X.rows, k);
+ float *truth_1d = pop_column(&X, target);
+ float **truth = one_hot_encode(truth_1d, X.rows, k);
matrix y;
y.rows = X.rows;
y.cols = k;
@@ -123,7 +123,7 @@
int i;
for(i = d.X.rows-1; i > 0; --i){
int index = rand()%i;
- double *swap = d.X.vals[index];
+ float *swap = d.X.vals[index];
d.X.vals[index] = d.X.vals[i];
d.X.vals[i] = swap;
@@ -156,10 +156,10 @@
train.X.cols = test.X.cols = d.X.cols;
train.y.cols = test.y.cols = d.y.cols;
- train.X.vals = calloc(train.X.rows, sizeof(double*));
- test.X.vals = calloc(test.X.rows, sizeof(double*));
- train.y.vals = calloc(train.y.rows, sizeof(double*));
- test.y.vals = calloc(test.y.rows, sizeof(double*));
+ train.X.vals = calloc(train.X.rows, sizeof(float*));
+ test.X.vals = calloc(test.X.rows, sizeof(float*));
+ train.y.vals = calloc(train.y.rows, sizeof(float*));
+ test.y.vals = calloc(test.y.rows, sizeof(float*));
for(i = 0; i < start; ++i){
train.X.vals[i] = d.X.vals[i];
diff --git a/src/image.c b/src/image.c
index df8e1b8..62ee5f7 100644
--- a/src/image.c
+++ b/src/image.c
@@ -16,7 +16,7 @@
for(k = 0; k < source.c; ++k){
for(i = 0; i < source.h; ++i){
for(j = 0; j < source.w; ++j){
- double val = get_pixel(source, i,j,k);
+ float val = get_pixel(source, i,j,k);
set_pixel(dest, h+i, w+j, k, val);
}
}
@@ -45,14 +45,14 @@
void normalize_image(image p)
{
- double *min = calloc(p.c, sizeof(double));
- double *max = calloc(p.c, sizeof(double));
+ float *min = calloc(p.c, sizeof(float));
+ float *max = calloc(p.c, sizeof(float));
int i,j;
for(i = 0; i < p.c; ++i) min[i] = max[i] = p.data[i*p.h*p.w];
for(j = 0; j < p.c; ++j){
for(i = 0; i < p.h*p.w; ++i){
- double v = p.data[i+j*p.h*p.w];
+ float v = p.data[i+j*p.h*p.w];
if(v < min[j]) min[j] = v;
if(v > max[j]) max[j] = v;
}
@@ -72,17 +72,17 @@
free(max);
}
-double avg_image_layer(image m, int l)
+float avg_image_layer(image m, int l)
{
int i;
- double sum = 0;
+ float sum = 0;
for(i = 0; i < m.h*m.w; ++i){
sum += m.data[l*m.h*m.w + i];
}
return sum/(m.h*m.w);
}
-void threshold_image(image p, double t)
+void threshold_image(image p, float t)
{
int i;
for(i = 0; i < p.w*p.h*p.c; ++i){
@@ -93,8 +93,8 @@
image copy_image(image p)
{
image copy = p;
- copy.data = calloc(p.h*p.w*p.c, sizeof(double));
- memcpy(copy.data, p.data, p.h*p.w*p.c*sizeof(double));
+ copy.data = calloc(p.h*p.w*p.c, sizeof(float));
+ memcpy(copy.data, p.data, p.h*p.w*p.c*sizeof(float));
return copy;
}
@@ -168,11 +168,11 @@
image make_image(int h, int w, int c)
{
image out = make_empty_image(h,w,c);
- out.data = calloc(h*w*c, sizeof(double));
+ out.data = calloc(h*w*c, sizeof(float));
return out;
}
-image double_to_image(int h, int w, int c, double *data)
+image float_to_image(int h, int w, int c, float *data)
{
image out = make_empty_image(h,w,c);
out.data = data;
@@ -181,12 +181,12 @@
void zero_image(image m)
{
- memset(m.data, 0, m.h*m.w*m.c*sizeof(double));
+ memset(m.data, 0, m.h*m.w*m.c*sizeof(float));
}
void zero_channel(image m, int c)
{
- memset(&(m.data[c*m.h*m.w]), 0, m.h*m.w*sizeof(double));
+ memset(&(m.data[c*m.h*m.w]), 0, m.h*m.w*sizeof(float));
}
void rotate_image(image m)
@@ -194,7 +194,7 @@
int i,j;
for(j = 0; j < m.c; ++j){
for(i = 0; i < m.h*m.w/2; ++i){
- double swap = m.data[j*m.h*m.w + i];
+ float swap = m.data[j*m.h*m.w + i];
m.data[j*m.h*m.w + i] = m.data[j*m.h*m.w + (m.h*m.w-1 - i)];
m.data[j*m.h*m.w + (m.h*m.w-1 - i)] = swap;
}
@@ -212,19 +212,19 @@
return out;
}
-void add_scalar_image(image m, double s)
+void add_scalar_image(image m, float s)
{
int i;
for(i = 0; i < m.h*m.w*m.c; ++i) m.data[i] += s;
}
-void scale_image(image m, double s)
+void scale_image(image m, float s)
{
int i;
for(i = 0; i < m.h*m.w*m.c; ++i) m.data[i] *= s;
}
-image make_random_kernel(int size, int c, double scale)
+image make_random_kernel(int size, int c, float scale)
{
int pad;
if((pad=(size%2==0))) ++size;
@@ -280,34 +280,34 @@
return out;
}
-double get_pixel(image m, int x, int y, int c)
+float get_pixel(image m, int x, int y, int c)
{
assert(x < m.h && y < m.w && c < m.c);
return m.data[c*m.h*m.w + x*m.w + y];
}
-double get_pixel_extend(image m, int x, int y, int c)
+float get_pixel_extend(image m, int x, int y, int c)
{
if(x < 0 || x >= m.h || y < 0 || y >= m.w || c < 0 || c >= m.c) return 0;
return get_pixel(m, x, y, c);
}
-void set_pixel(image m, int x, int y, int c, double val)
+void set_pixel(image m, int x, int y, int c, float val)
{
assert(x < m.h && y < m.w && c < m.c);
m.data[c*m.h*m.w + x*m.w + y] = val;
}
-void set_pixel_extend(image m, int x, int y, int c, double val)
+void set_pixel_extend(image m, int x, int y, int c, float val)
{
if(x < 0 || x >= m.h || y < 0 || y >= m.w || c < 0 || c >= m.c) return;
set_pixel(m, x, y, c, val);
}
-void add_pixel(image m, int x, int y, int c, double val)
+void add_pixel(image m, int x, int y, int c, float val)
{
assert(x < m.h && y < m.w && c < m.c);
m.data[c*m.h*m.w + x*m.w + y] += val;
}
-void add_pixel_extend(image m, int x, int y, int c, double val)
+void add_pixel_extend(image m, int x, int y, int c, float val)
{
if(x < 0 || x >= m.h || y < 0 || y >= m.w || c < 0 || c >= m.c) return;
add_pixel(m, x, y, c, val);
@@ -329,7 +329,7 @@
}
for(x = xstart; x < xend; x += stride){
for(y = ystart; y < yend; y += stride){
- double sum = 0;
+ float sum = 0;
for(i = 0; i < kernel.h; ++i){
for(j = 0; j < kernel.w; ++j){
sum += get_pixel(kernel, i, j, kc)*get_pixel_extend(m, x+i-kernel.h/2, y+j-kernel.w/2, mc);
@@ -340,9 +340,9 @@
}
}
-double single_convolve(image m, image kernel, int x, int y)
+float single_convolve(image m, image kernel, int x, int y)
{
- double sum = 0;
+ float sum = 0;
int i, j, k;
for(i = 0; i < kernel.h; ++i){
for(j = 0; j < kernel.w; ++j){
@@ -366,7 +366,7 @@
int j;
for(i = 0; i < m.h; i += stride){
for(j = 0; j < m.w; j += stride){
- double val = single_convolve(m, kernel, i, j);
+ float val = single_convolve(m, kernel, i, j);
set_pixel(out, i/stride, j/stride, channel, val);
}
}
@@ -380,20 +380,20 @@
for(k = 0; k < m.c; ++k){
for(i = 0; i < m.h; ++i){
for(j = 0; j< m.w; ++j){
- double val = get_pixel(m, i, j, k);
+ float val = get_pixel(m, i, j, k);
set_pixel(out, i*stride, j*stride, k, val);
}
}
}
}
-void single_update(image m, image update, int x, int y, double error)
+void single_update(image m, image update, int x, int y, float error)
{
int i, j, k;
for(i = 0; i < update.h; ++i){
for(j = 0; j < update.w; ++j){
for(k = 0; k < update.c; ++k){
- double val = get_pixel_extend(m, x+i-update.h/2, y+j-update.w/2, k);
+ float val = get_pixel_extend(m, x+i-update.h/2, y+j-update.w/2, k);
add_pixel(update, i, j, k, val*error);
}
}
@@ -417,7 +417,7 @@
}
for(i = istart; i < iend; i += stride){
for(j = jstart; j < jend; j += stride){
- double error = get_pixel(out, (i-istart)/stride, (j-jstart)/stride, channel);
+ float error = get_pixel(out, (i-istart)/stride, (j-jstart)/stride, channel);
single_update(m, update, i, j, error);
}
}
@@ -428,13 +428,13 @@
*/
}
-void single_back_convolve(image m, image kernel, int x, int y, double val)
+void single_back_convolve(image m, image kernel, int x, int y, float val)
{
int i, j, k;
for(i = 0; i < kernel.h; ++i){
for(j = 0; j < kernel.w; ++j){
for(k = 0; k < kernel.c; ++k){
- double pval = get_pixel(kernel, i, j, k) * val;
+ float pval = get_pixel(kernel, i, j, k) * val;
add_pixel_extend(m, x+i-kernel.h/2, y+j-kernel.w/2, k, pval);
}
}
@@ -457,7 +457,7 @@
}
for(i = istart; i < iend; i += stride){
for(j = jstart; j < jend; j += stride){
- double val = get_pixel(out, (i-istart)/stride, (j-jstart)/stride, channel);
+ float val = get_pixel(out, (i-istart)/stride, (j-jstart)/stride, channel);
single_back_convolve(m, kernel, i, j, val);
}
}
diff --git a/src/image.h b/src/image.h
index 1865857..72c4b2c 100644
--- a/src/image.h
+++ b/src/image.h
@@ -7,18 +7,18 @@
int h;
int w;
int c;
- double *data;
+ float *data;
} image;
-void scale_image(image m, double s);
-void add_scalar_image(image m, double s);
+void scale_image(image m, float s);
+void add_scalar_image(image m, float s);
void normalize_image(image p);
void z_normalize_image(image p);
-void threshold_image(image p, double t);
+void threshold_image(image p, float t);
void zero_image(image m);
void rotate_image(image m);
void subtract_image(image a, image b);
-double avg_image_layer(image m, int l);
+float avg_image_layer(image m, int l);
void embed_image(image source, image dest, int h, int w);
image collapse_image_layers(image source, int border);
@@ -30,14 +30,14 @@
image make_image(int h, int w, int c);
image make_empty_image(int h, int w, int c);
image make_random_image(int h, int w, int c);
-image make_random_kernel(int size, int c, double scale);
-image double_to_image(int h, int w, int c, double *data);
+image make_random_kernel(int size, int c, float scale);
+image float_to_image(int h, int w, int c, float *data);
image copy_image(image p);
image load_image(char *filename);
-double get_pixel(image m, int x, int y, int c);
-double get_pixel_extend(image m, int x, int y, int c);
-void set_pixel(image m, int x, int y, int c, double val);
+float get_pixel(image m, int x, int y, int c);
+float get_pixel_extend(image m, int x, int y, int c);
+void set_pixel(image m, int x, int y, int c, float val);
image get_image_layer(image m, int l);
diff --git a/src/matrix.c b/src/matrix.c
index 68e6f8d..96bd332 100644
--- a/src/matrix.c
+++ b/src/matrix.c
@@ -13,7 +13,7 @@
free(m.vals);
}
-double matrix_accuracy(matrix truth, matrix guess)
+float matrix_accuracy(matrix truth, matrix guess)
{
int k = truth.cols;
int i;
@@ -22,7 +22,7 @@
int class = max_index(guess.vals[i], k);
if(truth.vals[i][class]) ++count;
}
- return (double)count/truth.rows;
+ return (float)count/truth.rows;
}
void matrix_add_matrix(matrix from, matrix to)
@@ -42,9 +42,9 @@
matrix m;
m.rows = rows;
m.cols = cols;
- m.vals = calloc(m.rows, sizeof(double *));
+ m.vals = calloc(m.rows, sizeof(float *));
for(i = 0; i < m.rows; ++i){
- m.vals[i] = calloc(m.cols, sizeof(double));
+ m.vals[i] = calloc(m.cols, sizeof(float));
}
return m;
}
@@ -55,7 +55,7 @@
matrix h;
h.rows = n;
h.cols = m->cols;
- h.vals = calloc(h.rows, sizeof(double *));
+ h.vals = calloc(h.rows, sizeof(float *));
for(i = 0; i < n; ++i){
int index = rand()%m->rows;
h.vals[i] = m->vals[index];
@@ -64,9 +64,9 @@
return h;
}
-double *pop_column(matrix *m, int c)
+float *pop_column(matrix *m, int c)
{
- double *col = calloc(m->rows, sizeof(double));
+ float *col = calloc(m->rows, sizeof(float));
int i, j;
for(i = 0; i < m->rows; ++i){
col[i] = m->vals[i][c];
@@ -90,18 +90,18 @@
int n = 0;
int size = 1024;
- m.vals = calloc(size, sizeof(double*));
+ m.vals = calloc(size, sizeof(float*));
while((line = fgetl(fp))){
if(m.cols == -1) m.cols = count_fields(line);
if(n == size){
size *= 2;
- m.vals = realloc(m.vals, size*sizeof(double*));
+ m.vals = realloc(m.vals, size*sizeof(float*));
}
m.vals[n] = parse_fields(line, m.cols);
free(line);
++n;
}
- m.vals = realloc(m.vals, n*sizeof(double*));
+ m.vals = realloc(m.vals, n*sizeof(float*));
m.rows = n;
return m;
}
diff --git a/src/matrix.h b/src/matrix.h
index 098eb9e..01d825d 100644
--- a/src/matrix.h
+++ b/src/matrix.h
@@ -2,7 +2,7 @@
#define MATRIX_H
typedef struct matrix{
int rows, cols;
- double **vals;
+ float **vals;
} matrix;
matrix make_matrix(int rows, int cols);
@@ -11,9 +11,9 @@
matrix csv_to_matrix(char *filename);
matrix hold_out_matrix(matrix *m, int n);
-double matrix_accuracy(matrix truth, matrix guess);
+float matrix_accuracy(matrix truth, matrix guess);
void matrix_add_matrix(matrix from, matrix to);
-double *pop_column(matrix *m, int c);
+float *pop_column(matrix *m, int c);
#endif
diff --git a/src/maxpool_layer.c b/src/maxpool_layer.c
index ccf9bee..8c409b9 100644
--- a/src/maxpool_layer.c
+++ b/src/maxpool_layer.c
@@ -6,7 +6,7 @@
int h = (layer.h-1)/layer.stride + 1;
int w = (layer.w-1)/layer.stride + 1;
int c = layer.c;
- return double_to_image(h,w,c,layer.output);
+ return float_to_image(h,w,c,layer.output);
}
image get_maxpool_delta(maxpool_layer layer)
@@ -14,7 +14,7 @@
int h = (layer.h-1)/layer.stride + 1;
int w = (layer.w-1)/layer.stride + 1;
int c = layer.c;
- return double_to_image(h,w,c,layer.delta);
+ return float_to_image(h,w,c,layer.delta);
}
maxpool_layer *make_maxpool_layer(int h, int w, int c, int stride)
@@ -25,41 +25,41 @@
layer->w = w;
layer->c = c;
layer->stride = stride;
- layer->output = calloc(((h-1)/stride+1) * ((w-1)/stride+1) * c, sizeof(double));
- layer->delta = calloc(((h-1)/stride+1) * ((w-1)/stride+1) * c, sizeof(double));
+ layer->output = calloc(((h-1)/stride+1) * ((w-1)/stride+1) * c, sizeof(float));
+ layer->delta = calloc(((h-1)/stride+1) * ((w-1)/stride+1) * c, sizeof(float));
return layer;
}
-void forward_maxpool_layer(const maxpool_layer layer, double *in)
+void forward_maxpool_layer(const maxpool_layer layer, float *in)
{
- image input = double_to_image(layer.h, layer.w, layer.c, in);
+ image input = float_to_image(layer.h, layer.w, layer.c, in);
image output = get_maxpool_image(layer);
int i,j,k;
for(i = 0; i < output.h*output.w*output.c; ++i) output.data[i] = -DBL_MAX;
for(k = 0; k < input.c; ++k){
for(i = 0; i < input.h; ++i){
for(j = 0; j < input.w; ++j){
- double val = get_pixel(input, i, j, k);
- double cur = get_pixel(output, i/layer.stride, j/layer.stride, k);
+ float val = get_pixel(input, i, j, k);
+ float cur = get_pixel(output, i/layer.stride, j/layer.stride, k);
if(val > cur) set_pixel(output, i/layer.stride, j/layer.stride, k, val);
}
}
}
}
-void backward_maxpool_layer(const maxpool_layer layer, double *in, double *delta)
+void backward_maxpool_layer(const maxpool_layer layer, float *in, float *delta)
{
- image input = double_to_image(layer.h, layer.w, layer.c, in);
- image input_delta = double_to_image(layer.h, layer.w, layer.c, delta);
+ image input = float_to_image(layer.h, layer.w, layer.c, in);
+ image input_delta = float_to_image(layer.h, layer.w, layer.c, delta);
image output_delta = get_maxpool_delta(layer);
image output = get_maxpool_image(layer);
int i,j,k;
for(k = 0; k < input.c; ++k){
for(i = 0; i < input.h; ++i){
for(j = 0; j < input.w; ++j){
- double val = get_pixel(input, i, j, k);
- double cur = get_pixel(output, i/layer.stride, j/layer.stride, k);
- double d = get_pixel(output_delta, i/layer.stride, j/layer.stride, k);
+ float val = get_pixel(input, i, j, k);
+ float cur = get_pixel(output, i/layer.stride, j/layer.stride, k);
+ float d = get_pixel(output_delta, i/layer.stride, j/layer.stride, k);
if(val == cur) {
set_pixel(input_delta, i, j, k, d);
}
diff --git a/src/maxpool_layer.h b/src/maxpool_layer.h
index 0afe68a..27d6f55 100644
--- a/src/maxpool_layer.h
+++ b/src/maxpool_layer.h
@@ -6,14 +6,14 @@
typedef struct {
int h,w,c;
int stride;
- double *delta;
- double *output;
+ float *delta;
+ float *output;
} maxpool_layer;
image get_maxpool_image(maxpool_layer layer);
maxpool_layer *make_maxpool_layer(int h, int w, int c, int stride);
-void forward_maxpool_layer(const maxpool_layer layer, double *in);
-void backward_maxpool_layer(const maxpool_layer layer, double *in, double *delta);
+void forward_maxpool_layer(const maxpool_layer layer, float *in);
+void backward_maxpool_layer(const maxpool_layer layer, float *in, float *delta);
#endif
diff --git a/src/mini_blas.c b/src/mini_blas.c
index 3af36e5..b9a4304 100644
--- a/src/mini_blas.c
+++ b/src/mini_blas.c
@@ -1,8 +1,10 @@
#include <stdlib.h>
+#include <stdio.h>
#include <math.h>
+#include <time.h>
-void pm(int M, int N, double *A)
+void pm(int M, int N, float *A)
{
int i,j;
for(i =0 ; i < M; ++i){
@@ -14,28 +16,37 @@
printf("\n");
}
-void gemm(int TA, int TB, int M, int N, int K, double ALPHA,
- double *A, int lda,
- double *B, int ldb,
- double BETA,
- double *C, int ldc)
+void gemm(int TA, int TB, int M, int N, int K, float ALPHA,
+ float *A, int lda,
+ float *B, int ldb,
+ float BETA,
+ float *C, int ldc)
{
- // Assume TA = 0, beta = 1 LULZ
+ // Assume beta = 1 LULZ
int i,j,k;
if(TB && !TA){
for(i = 0; i < M; ++i){
for(j = 0; j < N; ++j){
- register double sum = 0;
+ register float sum = 0;
for(k = 0; k < K; ++k){
sum += ALPHA*A[i*lda+k]*B[k+j*ldb];
}
C[i*ldc+j] += sum;
}
}
+ }else if(TA && !TB){
+ for(i = 0; i < M; ++i){
+ for(k = 0; k < K; ++k){
+ register float A_PART = ALPHA*A[k*lda+i];
+ for(j = 0; j < N; ++j){
+ C[i*ldc+j] += A_PART*B[k*ldb+j];
+ }
+ }
+ }
}else{
for(i = 0; i < M; ++i){
for(k = 0; k < K; ++k){
- register double A_PART = ALPHA*A[i*lda+k];
+ register float A_PART = ALPHA*A[i*lda+k];
for(j = 0; j < N; ++j){
C[i*ldc+j] += A_PART*B[k*ldb+j];
}
@@ -44,7 +55,7 @@
}
}
-void im2row(double *image, int h, int w, int c, int size, int stride, double *matrix)
+void im2row(float *image, int h, int w, int c, int size, int stride, float *matrix)
{
int i;
int mc = c;
@@ -64,7 +75,7 @@
matrix[i] = image[pc*h*w+ph*w+pw];
}
}
-void im2col(double *image, int h, int w, int c, int size, int stride, double *matrix)
+void im2col(float *image, int h, int w, int c, int size, int stride, float *matrix)
{
int b,p;
int blocks = ((h-size)/stride+1)*((w-size)/stride+1);
@@ -84,9 +95,9 @@
}
//From Berkeley Vision's Caffe!
-void im2col_cpu(double* data_im, const int channels,
+void im2col_cpu(float* data_im, const int channels,
const int height, const int width, const int ksize, const int stride,
- double* data_col)
+ float* data_col)
{
int c,h,w;
int height_col = (height - ksize) / stride + 1;
@@ -106,3 +117,59 @@
}
}
+void col2im_cpu(float* data_col, const int channels,
+ const int height, const int width, const int ksize, const int stride,
+ float* data_im)
+{
+ int c,h,w;
+ int height_col = (height - ksize) / stride + 1;
+ int width_col = (width - ksize) / stride + 1;
+ int channels_col = channels * ksize * ksize;
+ for ( c = 0; c < channels_col; ++c) {
+ int w_offset = c % ksize;
+ int h_offset = (c / ksize) % ksize;
+ int c_im = c / ksize / ksize;
+ for ( h = 0; h < height_col; ++h) {
+ for ( w = 0; w < width_col; ++w) {
+ data_im[(c_im * height + h * stride + h_offset) * width
+ + w * stride + w_offset]+= data_col[(c * height_col + h) * width_col + w];
+ }
+ }
+ }
+}
+
+float *random_matrix(int rows, int cols)
+{
+ int i;
+ float *m = calloc(rows*cols, sizeof(float));
+ for(i = 0; i < rows*cols; ++i){
+ m[i] = (float)rand()/RAND_MAX;
+ }
+ return m;
+}
+
+void time_random_matrix(int TA, int TB, int m, int k, int n)
+{
+ float *a = random_matrix(m,k);
+ float *b = random_matrix(k,n);
+ float *c = random_matrix(m,n);
+ int i;
+ clock_t start = clock(), end;
+ for(i = 0; i<1000; ++i){
+ gemm(TA,TB,m,n,k,1,a,k,b,n,1,c,n);
+ }
+ end = clock();
+ printf("Matrix Multiplication %dx%d * %dx%d, TA=%d, TB=%d: %lf ms\n",m,k,k,n, TA, TB, (double)(end-start)/CLOCKS_PER_SEC);
+}
+
+void test_blas()
+{
+ time_random_matrix(0,0,100,100,100);
+ time_random_matrix(1,0,100,100,100);
+ time_random_matrix(0,1,100,100,100);
+
+ time_random_matrix(0,1,1000,100,100);
+ time_random_matrix(1,0,1000,100,100);
+
+}
+
diff --git a/src/mini_blas.h b/src/mini_blas.h
index 46a37d3..ff82a60 100644
--- a/src/mini_blas.h
+++ b/src/mini_blas.h
@@ -1,11 +1,15 @@
-void pm(int M, int N, double *A);
-void gemm(int TA, int TB, int M, int N, int K, double ALPHA,
- double *A, int lda,
- double *B, int ldb,
- double BETA,
- double *C, int ldc);
-void im2row(double *image, int h, int w, int c, int size, int stride, double *matrix);
-void im2col(double *image, int h, int w, int c, int size, int stride, double *matrix);
-void im2col_cpu(double* data_im, const int channels,
+void pm(int M, int N, float *A);
+void gemm(int TA, int TB, int M, int N, int K, float ALPHA,
+ float *A, int lda,
+ float *B, int ldb,
+ float BETA,
+ float *C, int ldc);
+void im2row(float *image, int h, int w, int c, int size, int stride, float *matrix);
+void im2col(float *image, int h, int w, int c, int size, int stride, float *matrix);
+void im2col_cpu(float* data_im, const int channels,
const int height, const int width, const int ksize, const int stride,
- double* data_col);
+ float* data_col);
+void col2im_cpu(float* data_col, const int channels,
+ const int height, const int width, const int ksize, const int stride,
+ float* data_im);
+void test_blas();
diff --git a/src/network.c b/src/network.c
index 2ce13d8..29e22e4 100644
--- a/src/network.c
+++ b/src/network.c
@@ -21,7 +21,7 @@
return net;
}
-void forward_network(network net, double *input)
+void forward_network(network net, float *input)
{
int i;
for(i = 0; i < net.n; ++i){
@@ -48,7 +48,7 @@
}
}
-void update_network(network net, double step, double momentum, double decay)
+void update_network(network net, float step, float momentum, float decay)
{
int i;
for(i = 0; i < net.n; ++i){
@@ -69,7 +69,7 @@
}
}
-double *get_network_output_layer(network net, int i)
+float *get_network_output_layer(network net, int i)
{
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
@@ -86,12 +86,12 @@
}
return 0;
}
-double *get_network_output(network net)
+float *get_network_output(network net)
{
return get_network_output_layer(net, net.n-1);
}
-double *get_network_delta_layer(network net, int i)
+float *get_network_delta_layer(network net, int i)
{
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
@@ -109,16 +109,16 @@
return 0;
}
-double *get_network_delta(network net)
+float *get_network_delta(network net)
{
return get_network_delta_layer(net, net.n-1);
}
-double calculate_error_network(network net, double *truth)
+float calculate_error_network(network net, float *truth)
{
- double sum = 0;
- double *delta = get_network_delta(net);
- double *out = get_network_output(net);
+ float sum = 0;
+ float *delta = get_network_delta(net);
+ float *out = get_network_output(net);
int i, k = get_network_output_size(net);
for(i = 0; i < k; ++i){
delta[i] = truth[i] - out[i];
@@ -129,17 +129,17 @@
int get_predicted_class_network(network net)
{
- double *out = get_network_output(net);
+ float *out = get_network_output(net);
int k = get_network_output_size(net);
return max_index(out, k);
}
-double backward_network(network net, double *input, double *truth)
+float backward_network(network net, float *input, float *truth)
{
- double error = calculate_error_network(net, truth);
+ float error = calculate_error_network(net, truth);
int i;
- double *prev_input;
- double *prev_delta;
+ float *prev_input;
+ float *prev_delta;
for(i = net.n-1; i >= 0; --i){
if(i == 0){
prev_input = input;
@@ -152,7 +152,7 @@
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
learn_convolutional_layer(layer);
//learn_convolutional_layer(layer);
- //if(i != 0) backward_convolutional_layer(layer, prev_input, prev_delta);
+ if(i != 0) backward_convolutional_layer(layer, prev_delta);
}
else if(net.types[i] == MAXPOOL){
maxpool_layer layer = *(maxpool_layer *)net.layers[i];
@@ -171,49 +171,49 @@
return error;
}
-double train_network_datum(network net, double *x, double *y, double step, double momentum, double decay)
+float train_network_datum(network net, float *x, float *y, float step, float momentum, float decay)
{
forward_network(net, x);
int class = get_predicted_class_network(net);
- double error = backward_network(net, x, y);
+ float error = backward_network(net, x, y);
update_network(net, step, momentum, decay);
//return (y[class]?1:0);
return error;
}
-double train_network_sgd(network net, data d, int n, double step, double momentum,double decay)
+float train_network_sgd(network net, data d, int n, float step, float momentum,float decay)
{
int i;
- double error = 0;
+ float error = 0;
for(i = 0; i < n; ++i){
int index = rand()%d.X.rows;
error += train_network_datum(net, d.X.vals[index], d.y.vals[index], step, momentum, decay);
//if((i+1)%10 == 0){
- // printf("%d: %f\n", (i+1), (double)correct/(i+1));
+ // printf("%d: %f\n", (i+1), (float)correct/(i+1));
//}
}
return error/n;
}
-double train_network_batch(network net, data d, int n, double step, double momentum,double decay)
+float train_network_batch(network net, data d, int n, float step, float momentum,float decay)
{
int i;
int correct = 0;
for(i = 0; i < n; ++i){
int index = rand()%d.X.rows;
- double *x = d.X.vals[index];
- double *y = d.y.vals[index];
+ float *x = d.X.vals[index];
+ float *y = d.y.vals[index];
forward_network(net, x);
int class = get_predicted_class_network(net);
backward_network(net, x, y);
correct += (y[class]?1:0);
}
update_network(net, step, momentum, decay);
- return (double)correct/n;
+ return (float)correct/n;
}
-void train_network(network net, data d, double step, double momentum, double decay)
+void train_network(network net, data d, float step, float momentum, float decay)
{
int i;
int correct = 0;
@@ -226,7 +226,7 @@
}
visualize_network(net);
cvWaitKey(100);
- printf("Accuracy: %f\n", (double)correct/d.X.rows);
+ printf("Accuracy: %f\n", (float)correct/d.X.rows);
}
int get_network_output_size_layer(network net, int i)
@@ -294,10 +294,10 @@
}
}
-double *network_predict(network net, double *input)
+float *network_predict(network net, float *input)
{
forward_network(net, input);
- double *out = get_network_output(net);
+ float *out = get_network_output(net);
return out;
}
@@ -307,7 +307,7 @@
int k = get_network_output_size(net);
matrix pred = make_matrix(test.X.rows, k);
for(i = 0; i < test.X.rows; ++i){
- double *out = network_predict(net, test.X.vals[i]);
+ float *out = network_predict(net, test.X.vals[i]);
for(j = 0; j < k; ++j){
pred.vals[i][j] = out[j];
}
@@ -319,7 +319,7 @@
{
int i,j;
for(i = 0; i < net.n; ++i){
- double *output = 0;
+ float *output = 0;
int n = 0;
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
@@ -343,8 +343,8 @@
output = layer.output;
n = layer.inputs;
}
- double mean = mean_array(output, n);
- double vari = variance_array(output, n);
+ float mean = mean_array(output, n);
+ float vari = variance_array(output, n);
fprintf(stderr, "Layer %d - Mean: %f, Variance: %f\n",i,mean, vari);
if(n > 100) n = 100;
for(j = 0; j < n; ++j) fprintf(stderr, "%f, ", output[j]);
@@ -353,10 +353,10 @@
}
}
-double network_accuracy(network net, data d)
+float network_accuracy(network net, data d)
{
matrix guess = network_predict_data(net, d);
- double acc = matrix_accuracy(d.y, guess);
+ float acc = matrix_accuracy(d.y, guess);
free_matrix(guess);
return acc;
}
diff --git a/src/network.h b/src/network.h
index fa109dd..17cc10b 100644
--- a/src/network.h
+++ b/src/network.h
@@ -17,22 +17,22 @@
void **layers;
LAYER_TYPE *types;
int outputs;
- double *output;
+ float *output;
} network;
network make_network(int n);
-void forward_network(network net, double *input);
-double backward_network(network net, double *input, double *truth);
-void update_network(network net, double step, double momentum, double decay);
-double train_network_sgd(network net, data d, int n, double step, double momentum,double decay);
-double train_network_batch(network net, data d, int n, double step, double momentum,double decay);
-void train_network(network net, data d, double step, double momentum, double decay);
+void forward_network(network net, float *input);
+float backward_network(network net, float *input, float *truth);
+void update_network(network net, float step, float momentum, float decay);
+float train_network_sgd(network net, data d, int n, float step, float momentum,float decay);
+float train_network_batch(network net, data d, int n, float step, float momentum,float decay);
+void train_network(network net, data d, float step, float momentum, float decay);
matrix network_predict_data(network net, data test);
-double network_accuracy(network net, data d);
-double *get_network_output(network net);
-double *get_network_output_layer(network net, int i);
-double *get_network_delta_layer(network net, int i);
-double *get_network_delta(network net);
+float network_accuracy(network net, data d);
+float *get_network_output(network net);
+float *get_network_output_layer(network net, int i);
+float *get_network_delta_layer(network net, int i);
+float *get_network_delta(network net);
int get_network_output_size_layer(network net, int i);
int get_network_output_size(network net);
image get_network_image(network net);
diff --git a/src/option_list.c b/src/option_list.c
index 1b32ebb..7902cd9 100644
--- a/src/option_list.c
+++ b/src/option_list.c
@@ -59,7 +59,7 @@
return def;
}
-double option_find_double(list *l, char *key, double def)
+float option_find_float(list *l, char *key, float def)
{
char *v = option_find(l, key);
if(v) return atof(v);
diff --git a/src/option_list.h b/src/option_list.h
index 0270465..60e37fe 100644
--- a/src/option_list.h
+++ b/src/option_list.h
@@ -6,7 +6,7 @@
char *option_find(list *l, char *key);
char *option_find_str(list *l, char *key, char *def);
int option_find_int(list *l, char *key, int def);
-double option_find_double(list *l, char *key, double def);
+float option_find_float(list *l, char *key, float def);
void option_unused(list *l);
#endif
diff --git a/src/softmax_layer.c b/src/softmax_layer.c
index b213e5b..1e01bd2 100644
--- a/src/softmax_layer.c
+++ b/src/softmax_layer.c
@@ -8,15 +8,16 @@
fprintf(stderr, "Softmax Layer: %d inputs\n", inputs);
softmax_layer *layer = calloc(1, sizeof(softmax_layer));
layer->inputs = inputs;
- layer->output = calloc(inputs, sizeof(double));
- layer->delta = calloc(inputs, sizeof(double));
+ layer->output = calloc(inputs, sizeof(float));
+ layer->delta = calloc(inputs, sizeof(float));
return layer;
}
-void forward_softmax_layer(const softmax_layer layer, double *input)
+/* UNSTABLE!
+void forward_softmax_layer(const softmax_layer layer, float *input)
{
int i;
- double sum = 0;
+ float sum = 0;
for(i = 0; i < layer.inputs; ++i){
sum += exp(input[i]);
}
@@ -24,8 +25,25 @@
layer.output[i] = exp(input[i])/sum;
}
}
+*/
+void forward_softmax_layer(const softmax_layer layer, float *input)
+{
+ int i;
+ float sum = 0;
+ float largest = 0;
+ for(i = 0; i < layer.inputs; ++i){
+ if(input[i] > largest) largest = input[i];
+ }
+ for(i = 0; i < layer.inputs; ++i){
+ sum += exp(input[i]-largest);
+ }
+ sum = largest+log(sum);
+ for(i = 0; i < layer.inputs; ++i){
+ layer.output[i] = exp(input[i]-sum);
+ }
+}
-void backward_softmax_layer(const softmax_layer layer, double *input, double *delta)
+void backward_softmax_layer(const softmax_layer layer, float *input, float *delta)
{
int i;
for(i = 0; i < layer.inputs; ++i){
diff --git a/src/softmax_layer.h b/src/softmax_layer.h
index 1a0d760..bfcd390 100644
--- a/src/softmax_layer.h
+++ b/src/softmax_layer.h
@@ -3,12 +3,12 @@
typedef struct {
int inputs;
- double *delta;
- double *output;
+ float *delta;
+ float *output;
} softmax_layer;
softmax_layer *make_softmax_layer(int inputs);
-void forward_softmax_layer(const softmax_layer layer, double *input);
-void backward_softmax_layer(const softmax_layer layer, double *input, double *delta);
+void forward_softmax_layer(const softmax_layer layer, float *input);
+void backward_softmax_layer(const softmax_layer layer, float *input, float *delta);
#endif
diff --git a/src/tests.c b/src/tests.c
index af22ddb..00cd1a1 100644
--- a/src/tests.c
+++ b/src/tests.c
@@ -14,6 +14,9 @@
#include <stdlib.h>
#include <stdio.h>
+#define _GNU_SOURCE
+#include <fenv.h>
+
void test_convolve()
{
image dog = load_image("dog.jpg");
@@ -26,7 +29,7 @@
convolve(dog, kernel, 1, 0, edge, 1);
}
end = clock();
- printf("Convolutions: %lf seconds\n", (double)(end-start)/CLOCKS_PER_SEC);
+ printf("Convolutions: %lf seconds\n", (float)(end-start)/CLOCKS_PER_SEC);
show_image_layers(edge, "Test Convolve");
}
@@ -38,11 +41,11 @@
int size = 11;
int stride = 4;
int n = 40;
- double *filters = make_random_image(size, size, dog.c*n).data;
+ float *filters = make_random_image(size, size, dog.c*n).data;
int mw = ((dog.h-size)/stride+1)*((dog.w-size)/stride+1);
int mh = (size*size*dog.c);
- double *matrix = calloc(mh*mw, sizeof(double));
+ float *matrix = calloc(mh*mw, sizeof(float));
image edge = make_image((dog.h-size)/stride+1, (dog.w-size)/stride+1, n);
@@ -54,7 +57,7 @@
gemm(0,0,n,mw,mh,1,filters,mh,matrix,mw,1,edge.data,mw);
}
end = clock();
- printf("Convolutions: %lf seconds\n", (double)(end-start)/CLOCKS_PER_SEC);
+ printf("Convolutions: %lf seconds\n", (float)(end-start)/CLOCKS_PER_SEC);
show_image_layers(edge, "Test Convolve");
cvWaitKey(0);
}
@@ -72,11 +75,11 @@
int n = 1;
int stride = 1;
int size = 3;
- double eps = .00000001;
+ float eps = .00000001;
image test = make_random_image(5,5, 1);
convolutional_layer layer = *make_convolutional_layer(test.h,test.w,test.c, n, size, stride, RELU);
image out = get_convolutional_image(layer);
- double **jacobian = calloc(test.h*test.w*test.c, sizeof(double));
+ float **jacobian = calloc(test.h*test.w*test.c, sizeof(float));
forward_convolutional_layer(layer, test.data);
image base = copy_image(out);
@@ -90,19 +93,19 @@
jacobian[i] = partial.data;
test.data[i] -= eps;
}
- double **jacobian2 = calloc(out.h*out.w*out.c, sizeof(double));
+ float **jacobian2 = calloc(out.h*out.w*out.c, sizeof(float));
image in_delta = make_image(test.h, test.w, test.c);
image out_delta = get_convolutional_delta(layer);
for(i = 0; i < out.h*out.w*out.c; ++i){
out_delta.data[i] = 1;
- //backward_convolutional_layer(layer, test.data, in_delta.data);
+ backward_convolutional_layer(layer, in_delta.data);
image partial = copy_image(in_delta);
jacobian2[i] = partial.data;
out_delta.data[i] = 0;
}
int j;
- double *j1 = calloc(test.h*test.w*test.c*out.h*out.w*out.c, sizeof(double));
- double *j2 = calloc(test.h*test.w*test.c*out.h*out.w*out.c, sizeof(double));
+ float *j1 = calloc(test.h*test.w*test.c*out.h*out.w*out.c, sizeof(float));
+ float *j2 = calloc(test.h*test.w*test.c*out.h*out.w*out.c, sizeof(float));
for(i = 0; i < test.h*test.w*test.c; ++i){
for(j =0 ; j < out.h*out.w*out.c; ++j){
j1[i*out.h*out.w*out.c + j] = jacobian[i][j];
@@ -112,12 +115,11 @@
}
- image mj1 = double_to_image(test.w*test.h*test.c, out.w*out.h*out.c, 1, j1);
- image mj2 = double_to_image(test.w*test.h*test.c, out.w*out.h*out.c, 1, j2);
+ image mj1 = float_to_image(test.w*test.h*test.c, out.w*out.h*out.c, 1, j1);
+ image mj2 = float_to_image(test.w*test.h*test.c, out.w*out.h*out.c, 1, j2);
printf("%f %f\n", avg_image_layer(mj1,0), avg_image_layer(mj2,0));
show_image(mj1, "forward jacobian");
show_image(mj2, "backward jacobian");
-
}
void test_load()
@@ -145,7 +147,7 @@
rotate_image(dog);
}
end = clock();
- printf("Rotations: %lf seconds\n", (double)(end-start)/CLOCKS_PER_SEC);
+ printf("Rotations: %lf seconds\n", (float)(end-start)/CLOCKS_PER_SEC);
show_image(dog, "Test Rotate");
image random = make_random_image(3,3,3);
@@ -159,18 +161,18 @@
void test_parser()
{
network net = parse_network_cfg("test_parser.cfg");
- double input[1];
+ float input[1];
int count = 0;
- double avgerr = 0;
+ float avgerr = 0;
while(++count < 100000000){
- double v = ((double)rand()/RAND_MAX);
- double truth = v*v;
+ float v = ((float)rand()/RAND_MAX);
+ float truth = v*v;
input[0] = v;
forward_network(net, input);
- double *out = get_network_output(net);
- double *delta = get_network_delta(net);
- double err = pow((out[0]-truth),2.);
+ float *out = get_network_output(net);
+ float *delta = get_network_delta(net);
+ float err = pow((out[0]-truth),2.);
avgerr = .99 * avgerr + .01 * err;
if(count % 1000000 == 0) printf("%f %f :%f AVG %f \n", truth, out[0], err, avgerr);
delta[0] = truth - out[0];
@@ -192,9 +194,9 @@
srand(0);
int i = 0;
char *labels[] = {"cat","dog"};
- double lr = .00001;
- double momentum = .9;
- double decay = 0.01;
+ float lr = .00001;
+ float momentum = .9;
+ float decay = 0.01;
while(i++ < 1000 || 1){
data train = load_data_image_pathfile_random("train_paths.txt", 1000, labels, 2);
train_network(net, train, lr, momentum, decay);
@@ -207,32 +209,33 @@
{
srand(444444);
srand(888888);
- network net = parse_network_cfg("nist_basic.cfg");
+ network net = parse_network_cfg("nist.cfg");
data train = load_categorical_data_csv("mnist/mnist_train.csv", 0, 10);
data test = load_categorical_data_csv("mnist/mnist_test.csv",0,10);
normalize_data_rows(train);
normalize_data_rows(test);
//randomize_data(train);
int count = 0;
- double lr = .0005;
- double momentum = .9;
- double decay = 0.01;
+ float lr = .0005;
+ float momentum = .9;
+ float decay = 0.01;
clock_t start = clock(), end;
while(++count <= 100){
- visualize_network(net);
- double loss = train_network_sgd(net, train, 10000, lr, momentum, decay);
+ //visualize_network(net);
+ float loss = train_network_sgd(net, train, 1000, lr, momentum, decay);
printf("%5d Training Loss: %lf, Params: %f %f %f, ",count*100, loss, lr, momentum, decay);
end = clock();
- printf("Time: %lf seconds\n", (double)(end-start)/CLOCKS_PER_SEC);
+ printf("Time: %lf seconds\n", (float)(end-start)/CLOCKS_PER_SEC);
start=end;
cvWaitKey(100);
//lr /= 2;
if(count%5 == 0){
- double train_acc = network_accuracy(net, train);
+ float train_acc = network_accuracy(net, train);
fprintf(stderr, "\nTRAIN: %f\n", train_acc);
- double test_acc = network_accuracy(net, test);
+ float test_acc = network_accuracy(net, test);
fprintf(stderr, "TEST: %f\n\n", test_acc);
printf("%d, %f, %f\n", count, train_acc, test_acc);
+ lr *= .5;
}
}
}
@@ -253,24 +256,24 @@
int n = 30;
for(i = 0; i < n; ++i){
int count = 0;
- double lr = .0005;
- double momentum = .9;
- double decay = .01;
+ float lr = .0005;
+ float momentum = .9;
+ float decay = .01;
network net = parse_network_cfg("nist.cfg");
while(++count <= 15){
- double acc = train_network_sgd(net, train, train.X.rows, lr, momentum, decay);
+ float acc = train_network_sgd(net, train, train.X.rows, lr, momentum, decay);
printf("Training Accuracy: %lf Learning Rate: %f Momentum: %f Decay: %f\n", acc, lr, momentum, decay );
lr /= 2;
}
matrix partial = network_predict_data(net, test);
- double acc = matrix_accuracy(test.y, partial);
+ float acc = matrix_accuracy(test.y, partial);
printf("Model Accuracy: %lf\n", acc);
matrix_add_matrix(partial, prediction);
acc = matrix_accuracy(test.y, prediction);
printf("Current Ensemble Accuracy: %lf\n", acc);
free_matrix(partial);
}
- double acc = matrix_accuracy(test.y, prediction);
+ float acc = matrix_accuracy(test.y, prediction);
printf("Full Ensemble Accuracy: %lf\n", acc);
}
@@ -279,19 +282,19 @@
network net = parse_network_cfg("connected.cfg");
matrix m = csv_to_matrix("train.csv");
//matrix ho = hold_out_matrix(&m, 2500);
- double *truth = pop_column(&m, 0);
- //double *ho_truth = pop_column(&ho, 0);
+ float *truth = pop_column(&m, 0);
+ //float *ho_truth = pop_column(&ho, 0);
int i;
clock_t start = clock(), end;
int count = 0;
while(++count <= 300){
for(i = 0; i < m.rows; ++i){
int index = rand()%m.rows;
- //image p = double_to_image(1690,1,1,m.vals[index]);
+ //image p = float_to_image(1690,1,1,m.vals[index]);
//normalize_image(p);
forward_network(net, m.vals[index]);
- double *out = get_network_output(net);
- double *delta = get_network_delta(net);
+ float *out = get_network_output(net);
+ float *delta = get_network_delta(net);
//printf("%f\n", out[0]);
delta[0] = truth[index] - out[0];
// printf("%f\n", delta[0]);
@@ -299,8 +302,8 @@
//backward_network(net, m.vals[index], );
update_network(net, .00001, 0,0);
}
- //double test_acc = error_network(net, m, truth);
- //double valid_acc = error_network(net, ho, ho_truth);
+ //float test_acc = error_network(net, m, truth);
+ //float valid_acc = error_network(net, ho, ho_truth);
//printf("%f, %f\n", test_acc, valid_acc);
//fprintf(stderr, "%5d: %f Valid: %f\n",count, test_acc, valid_acc);
//if(valid_acc > .70) break;
@@ -311,12 +314,12 @@
truth = pop_column(&test, 0);
for(i = 0; i < test.rows; ++i){
forward_network(net, test.vals[i]);
- double *out = get_network_output(net);
+ float *out = get_network_output(net);
if(fabs(out[0]) < .5) fprintf(fp, "0\n");
else fprintf(fp, "1\n");
}
fclose(fp);
- printf("Neural Net Learning: %lf seconds\n", (double)(end-start)/CLOCKS_PER_SEC);
+ printf("Neural Net Learning: %lf seconds\n", (float)(end-start)/CLOCKS_PER_SEC);
}
void test_split()
@@ -326,30 +329,6 @@
printf("%d, %d, %d\n", train.X.rows, split[0].X.rows, split[1].X.rows);
}
-double *random_matrix(int rows, int cols)
-{
- int i, j;
- double *m = calloc(rows*cols, sizeof(double));
- for(i = 0; i < rows; ++i){
- for(j = 0; j < cols; ++j){
- m[i*cols+j] = (double)rand()/RAND_MAX;
- }
- }
- return m;
-}
-
-void test_blas()
-{
- int m = 1000, n = 1000, k = 1000;
- double *a = random_matrix(m,k);
- double *b = random_matrix(k,n);
- double *c = random_matrix(m,n);
- int i;
- for(i = 0; i<1000; ++i){
- gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
- }
-}
-
void test_im2row()
{
int h = 20;
@@ -362,16 +341,18 @@
int mw = ((h-size)/stride+1)*((w-size)/stride+1);
int mh = (size*size*c);
int msize = mc*mw*mh;
- double *matrix = calloc(msize, sizeof(double));
+ float *matrix = calloc(msize, sizeof(float));
int i;
for(i = 0; i < 1000; ++i){
im2col_cpu(test.data, c, h, w, size, stride, matrix);
- image render = double_to_image(mh, mw, mc, matrix);
+ image render = float_to_image(mh, mw, mc, matrix);
}
}
int main()
{
+ //feenableexcept(FE_DIVBYZERO | FE_INVALID | FE_OVERFLOW);
+
//test_blas();
//test_convolve_matrix();
// test_im2row();
diff --git a/src/utils.c b/src/utils.c
index 5180fe6..41ee768 100644
--- a/src/utils.c
+++ b/src/utils.c
@@ -123,9 +123,9 @@
return count;
}
-double *parse_fields(char *line, int n)
+float *parse_fields(char *line, int n)
{
- double *field = calloc(n, sizeof(double));
+ float *field = calloc(n, sizeof(float));
char *c, *p, *end;
int count = 0;
int done = 0;
@@ -143,36 +143,36 @@
return field;
}
-double mean_array(double *a, int n)
+float mean_array(float *a, int n)
{
int i;
- double sum = 0;
+ float sum = 0;
for(i = 0; i < n; ++i) sum += a[i];
return sum/n;
}
-double variance_array(double *a, int n)
+float variance_array(float *a, int n)
{
int i;
- double sum = 0;
- double mean = mean_array(a, n);
+ float sum = 0;
+ float mean = mean_array(a, n);
for(i = 0; i < n; ++i) sum += (a[i] - mean)*(a[i]-mean);
- double variance = sum/n;
+ float variance = sum/n;
return variance;
}
-double constrain(double a, double max)
+float constrain(float a, float max)
{
if(a > abs(max)) return abs(max);
if(a < -abs(max)) return -abs(max);
return a;
}
-void normalize_array(double *a, int n)
+void normalize_array(float *a, int n)
{
int i;
- double mu = mean_array(a,n);
- double sigma = sqrt(variance_array(a,n));
+ float mu = mean_array(a,n);
+ float sigma = sqrt(variance_array(a,n));
for(i = 0; i < n; ++i){
a[i] = (a[i] - mu)/sigma;
}
@@ -180,7 +180,7 @@
sigma = sqrt(variance_array(a,n));
}
-void translate_array(double *a, int n, double s)
+void translate_array(float *a, int n, float s)
{
int i;
for(i = 0; i < n; ++i){
@@ -188,18 +188,18 @@
}
}
-void scale_array(double *a, int n, double s)
+void scale_array(float *a, int n, float s)
{
int i;
for(i = 0; i < n; ++i){
a[i] *= s;
}
}
-int max_index(double *a, int n)
+int max_index(float *a, int n)
{
if(n <= 0) return -1;
int i, max_i = 0;
- double max = a[0];
+ float max = a[0];
for(i = 1; i < n; ++i){
if(a[i] > max){
max = a[i];
@@ -209,20 +209,20 @@
return max_i;
}
-double rand_normal()
+float rand_normal()
{
int i;
- double sum= 0;
- for(i = 0; i < 12; ++i) sum += (double)rand()/RAND_MAX;
+ float sum= 0;
+ for(i = 0; i < 12; ++i) sum += (float)rand()/RAND_MAX;
return sum-6.;
}
-double **one_hot_encode(double *a, int n, int k)
+float **one_hot_encode(float *a, int n, int k)
{
int i;
- double **t = calloc(n, sizeof(double*));
+ float **t = calloc(n, sizeof(float*));
for(i = 0; i < n; ++i){
- t[i] = calloc(k, sizeof(double));
+ t[i] = calloc(k, sizeof(float));
int index = (int)a[i];
t[i][index] = 1;
}
diff --git a/src/utils.h b/src/utils.h
index cf38016..8185107 100644
--- a/src/utils.h
+++ b/src/utils.h
@@ -13,15 +13,15 @@
list *parse_csv_line(char *line);
char *copy_string(char *s);
int count_fields(char *line);
-double *parse_fields(char *line, int n);
-void normalize_array(double *a, int n);
-void scale_array(double *a, int n, double s);
-void translate_array(double *a, int n, double s);
-int max_index(double *a, int n);
-double constrain(double a, double max);
-double rand_normal();
-double mean_array(double *a, int n);
-double variance_array(double *a, int n);
-double **one_hot_encode(double *a, int n, int k);
+float *parse_fields(char *line, int n);
+void normalize_array(float *a, int n);
+void scale_array(float *a, int n, float s);
+void translate_array(float *a, int n, float s);
+int max_index(float *a, int n);
+float constrain(float a, float max);
+float rand_normal();
+float mean_array(float *a, int n);
+float variance_array(float *a, int n);
+float **one_hot_encode(float *a, int n, int k);
#endif
--
Gitblit v1.10.0