From 6553b3f0e3e55fc30a99c7d4b5798aa86d18a114 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Mon, 30 Mar 2015 02:31:47 +0000
Subject: [PATCH] no comment
---
src/convolutional_layer.c | 40 +++++++++++++++++-----------------------
1 files changed, 17 insertions(+), 23 deletions(-)
diff --git a/src/convolutional_layer.c b/src/convolutional_layer.c
index 7782e3d..e20a41c 100644
--- a/src/convolutional_layer.c
+++ b/src/convolutional_layer.c
@@ -41,15 +41,11 @@
return float_to_image(h,w,c,layer.delta);
}
-convolutional_layer *make_convolutional_layer(int batch, int h, int w, int c, int n, int size, int stride, int pad, ACTIVATION activation, float learning_rate, float momentum, float decay)
+convolutional_layer *make_convolutional_layer(int batch, int h, int w, int c, int n, int size, int stride, int pad, ACTIVATION activation)
{
int i;
convolutional_layer *layer = calloc(1, sizeof(convolutional_layer));
- layer->learning_rate = learning_rate;
- layer->momentum = momentum;
- layer->decay = decay;
-
layer->h = h;
layer->w = w;
layer->c = c;
@@ -133,17 +129,16 @@
void backward_bias(float *bias_updates, float *delta, int batch, int n, int size)
{
- float alpha = 1./batch;
int i,b;
for(b = 0; b < batch; ++b){
for(i = 0; i < n; ++i){
- bias_updates[i] += alpha * sum_array(delta+size*(i+b*n), size);
+ bias_updates[i] += sum_array(delta+size*(i+b*n), size);
}
}
}
-void forward_convolutional_layer(const convolutional_layer layer, float *in)
+void forward_convolutional_layer(const convolutional_layer layer, network_state state)
{
int out_h = convolutional_out_height(layer);
int out_w = convolutional_out_width(layer);
@@ -160,18 +155,17 @@
float *c = layer.output;
for(i = 0; i < layer.batch; ++i){
- im2col_cpu(in, layer.c, layer.h, layer.w,
+ im2col_cpu(state.input, layer.c, layer.h, layer.w,
layer.size, layer.stride, layer.pad, b);
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
c += n*m;
- in += layer.c*layer.h*layer.w;
+ state.input += layer.c*layer.h*layer.w;
}
activate_array(layer.output, m*n*layer.batch, layer.activation);
}
-void backward_convolutional_layer(convolutional_layer layer, float *in, float *delta)
+void backward_convolutional_layer(convolutional_layer layer, network_state state)
{
- float alpha = 1./layer.batch;
int i;
int m = layer.n;
int n = layer.size*layer.size*layer.c;
@@ -181,40 +175,40 @@
gradient_array(layer.output, m*k*layer.batch, layer.activation, layer.delta);
backward_bias(layer.bias_updates, layer.delta, layer.batch, layer.n, k);
- if(delta) memset(delta, 0, layer.batch*layer.h*layer.w*layer.c*sizeof(float));
+ if(state.delta) memset(state.delta, 0, layer.batch*layer.h*layer.w*layer.c*sizeof(float));
for(i = 0; i < layer.batch; ++i){
float *a = layer.delta + i*m*k;
float *b = layer.col_image;
float *c = layer.filter_updates;
- float *im = in+i*layer.c*layer.h*layer.w;
+ float *im = state.input+i*layer.c*layer.h*layer.w;
im2col_cpu(im, layer.c, layer.h, layer.w,
layer.size, layer.stride, layer.pad, b);
- gemm(0,1,m,n,k,alpha,a,k,b,k,1,c,n);
+ gemm(0,1,m,n,k,1,a,k,b,k,1,c,n);
- if(delta){
+ if(state.delta){
a = layer.filters;
b = layer.delta + i*m*k;
c = layer.col_image;
gemm(1,0,n,k,m,1,a,n,b,k,0,c,k);
- col2im_cpu(layer.col_image, layer.c, layer.h, layer.w, layer.size, layer.stride, layer.pad, delta+i*layer.c*layer.h*layer.w);
+ col2im_cpu(layer.col_image, layer.c, layer.h, layer.w, layer.size, layer.stride, layer.pad, state.delta+i*layer.c*layer.h*layer.w);
}
}
}
-void update_convolutional_layer(convolutional_layer layer)
+void update_convolutional_layer(convolutional_layer layer, int batch, float learning_rate, float momentum, float decay)
{
int size = layer.size*layer.size*layer.c*layer.n;
- axpy_cpu(layer.n, layer.learning_rate, layer.bias_updates, 1, layer.biases, 1);
- scal_cpu(layer.n, layer.momentum, layer.bias_updates, 1);
+ axpy_cpu(layer.n, learning_rate/batch, layer.bias_updates, 1, layer.biases, 1);
+ scal_cpu(layer.n, momentum, layer.bias_updates, 1);
- axpy_cpu(size, -layer.decay, layer.filters, 1, layer.filter_updates, 1);
- axpy_cpu(size, layer.learning_rate, layer.filter_updates, 1, layer.filters, 1);
- scal_cpu(size, layer.momentum, layer.filter_updates, 1);
+ axpy_cpu(size, -decay*batch, layer.filters, 1, layer.filter_updates, 1);
+ axpy_cpu(size, learning_rate/batch, layer.filter_updates, 1, layer.filters, 1);
+ scal_cpu(size, momentum, layer.filter_updates, 1);
}
--
Gitblit v1.10.0