From 0f1a31648c5292fa49b35eac90a2ee676d6c13e6 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Sat, 31 Jan 2015 06:05:23 +0000
Subject: [PATCH] idk, probably something changed
---
src/convolutional_layer.c | 187 ++++------------------------------------------
1 files changed, 18 insertions(+), 169 deletions(-)
diff --git a/src/convolutional_layer.c b/src/convolutional_layer.c
index 4ca6104..6848511 100644
--- a/src/convolutional_layer.c
+++ b/src/convolutional_layer.c
@@ -1,6 +1,9 @@
#include "convolutional_layer.h"
#include "utils.h"
-#include "mini_blas.h"
+#include "im2col.h"
+#include "col2im.h"
+#include "blas.h"
+#include "gemm.h"
#include <stdio.h>
#include <time.h>
@@ -62,12 +65,12 @@
layer->biases = calloc(n, sizeof(float));
layer->bias_updates = calloc(n, sizeof(float));
- float scale = 1./(size*size*c);
- scale = .01;
- for(i = 0; i < c*n*size*size; ++i) layer->filters[i] = scale*2*(rand_uniform()-.5);
+ float scale = 1./sqrt(size*size*c);
+ //scale = .05;
+ for(i = 0; i < c*n*size*size; ++i) layer->filters[i] = scale*rand_normal();
for(i = 0; i < n; ++i){
//layer->biases[i] = rand_normal()*scale + scale;
- layer->biases[i] = .5;
+ layer->biases[i] = scale;
}
int out_h = convolutional_out_height(*layer);
int out_w = convolutional_out_width(*layer);
@@ -77,15 +80,15 @@
layer->delta = calloc(layer->batch*out_h * out_w * n, sizeof(float));
#ifdef GPU
- layer->filters_cl = cl_make_array(layer->filters, c*n*size*size);
- layer->filter_updates_cl = cl_make_array(layer->filter_updates, c*n*size*size);
+ layer->filters_gpu = cuda_make_array(layer->filters, c*n*size*size);
+ layer->filter_updates_gpu = cuda_make_array(layer->filter_updates, c*n*size*size);
- layer->biases_cl = cl_make_array(layer->biases, n);
- layer->bias_updates_cl = cl_make_array(layer->bias_updates, n);
+ layer->biases_gpu = cuda_make_array(layer->biases, n);
+ layer->bias_updates_gpu = cuda_make_array(layer->bias_updates, n);
- layer->col_image_cl = cl_make_array(layer->col_image, out_h*out_w*size*size*c);
- layer->delta_cl = cl_make_array(layer->delta, layer->batch*out_h*out_w*n);
- layer->output_cl = cl_make_array(layer->output, layer->batch*out_h*out_w*n);
+ layer->col_image_gpu = cuda_make_array(layer->col_image, out_h*out_w*size*size*c);
+ layer->delta_gpu = cuda_make_array(layer->delta, layer->batch*out_h*out_w*n);
+ layer->output_gpu = cuda_make_array(layer->output, layer->batch*out_h*out_w*n);
#endif
layer->activation = activation;
@@ -140,7 +143,6 @@
float *b = layer.col_image;
float *c = layer.output;
-
for(i = 0; i < layer.batch; ++i){
im2col_cpu(in, layer.c, layer.h, layer.w,
layer.size, layer.stride, layer.pad, b);
@@ -170,7 +172,9 @@
int n = layer.size*layer.size*layer.c;
int k = convolutional_out_height(layer)*
convolutional_out_width(layer);
+
gradient_array(layer.output, m*k*layer.batch, layer.activation, layer.delta);
+
learn_bias_convolutional_layer(layer);
if(delta) memset(delta, 0, layer.batch*layer.h*layer.w*layer.c*sizeof(float));
@@ -204,7 +208,7 @@
axpy_cpu(layer.n, layer.learning_rate, layer.bias_updates, 1, layer.biases, 1);
scal_cpu(layer.n, layer.momentum, layer.bias_updates, 1);
- scal_cpu(size, 1.-layer.learning_rate*layer.decay, layer.filters, 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);
}
@@ -263,158 +267,3 @@
return single_filters;
}
-#ifdef GPU
-
-cl_kernel get_convolutional_learn_bias_kernel()
-{
- static int init = 0;
- static cl_kernel kernel;
- if(!init){
- kernel = get_kernel("src/convolutional_layer.cl", "learn_bias", 0);
- init = 1;
- }
- return kernel;
-}
-
-void learn_bias_convolutional_layer_ongpu(convolutional_layer layer)
-{
- int size = convolutional_out_height(layer) * convolutional_out_width(layer);
-
- cl_setup();
- cl_kernel kernel = get_convolutional_learn_bias_kernel();
- cl_command_queue queue = cl.queue;
-
- cl_uint i = 0;
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.batch), (void*) &layer.batch);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.n), (void*) &layer.n);
- cl.error = clSetKernelArg(kernel, i++, sizeof(size), (void*) &size);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.delta_cl), (void*) &layer.delta_cl);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.bias_updates_cl), (void*) &layer.bias_updates_cl);
- check_error(cl);
-
- const size_t global_size[] = {layer.n};
-
- cl.error = clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
- check_error(cl);
-}
-
-cl_kernel get_convolutional_bias_kernel()
-{
- static int init = 0;
- static cl_kernel kernel;
- if(!init){
- kernel = get_kernel("src/convolutional_layer.cl", "bias", 0);
- init = 1;
- }
- return kernel;
-}
-
-void bias_output_gpu(const convolutional_layer layer)
-{
- int out_h = convolutional_out_height(layer);
- int out_w = convolutional_out_width(layer);
- int size = out_h*out_w;
-
- cl_setup();
- cl_kernel kernel = get_convolutional_bias_kernel();
- cl_command_queue queue = cl.queue;
-
- cl_uint i = 0;
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.n), (void*) &layer.n);
- cl.error = clSetKernelArg(kernel, i++, sizeof(size), (void*) &size);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.biases_cl), (void*) &layer.biases_cl);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.output_cl), (void*) &layer.output_cl);
- check_error(cl);
-
- const size_t global_size[] = {layer.n*size, layer.batch};
-
- cl.error = clEnqueueNDRangeKernel(queue, kernel, 2, 0, global_size, 0, 0, 0, 0);
- check_error(cl);
-}
-
-//#define TIMEIT
-
-void forward_convolutional_layer_gpu(convolutional_layer layer, cl_mem in)
-{
- int i;
- int m = layer.n;
- int k = layer.size*layer.size*layer.c;
- int n = convolutional_out_height(layer)*
- convolutional_out_width(layer);
-
- bias_output_gpu(layer);
-
- for(i = 0; i < layer.batch; ++i){
- im2col_ongpu(in, i*layer.c*layer.h*layer.w, layer.c, layer.h, layer.w, layer.size, layer.stride, layer.pad, layer.col_image_cl);
- cl_mem a = layer.filters_cl;
- cl_mem b = layer.col_image_cl;
- cl_mem c = layer.output_cl;
- gemm_ongpu_offset(0,0,m,n,k,1.,a,0,k,b,0,n,1.,c,i*m*n,n);
- }
- activate_array_ongpu(layer.output_cl, m*n*layer.batch, layer.activation);
-}
-
-void backward_convolutional_layer_gpu(convolutional_layer layer, cl_mem in, cl_mem delta_cl)
-{
- int i;
- int m = layer.n;
- int n = layer.size*layer.size*layer.c;
- int k = convolutional_out_height(layer)*
- convolutional_out_width(layer);
- gradient_array_ongpu(layer.output_cl, m*k*layer.batch, layer.activation, layer.delta_cl);
- learn_bias_convolutional_layer_ongpu(layer);
-
- if(delta_cl) scal_ongpu(layer.batch*layer.h*layer.w*layer.c, 0, delta_cl, 1);
-
- for(i = 0; i < layer.batch; ++i){
- cl_mem a = layer.delta_cl;
- cl_mem b = layer.col_image_cl;
- cl_mem c = layer.filter_updates_cl;
-
- im2col_ongpu(in, i*layer.c*layer.h*layer.w, layer.c, layer.h, layer.w, layer.size, layer.stride, layer.pad, layer.col_image_cl);
- gemm_ongpu_offset(0,1,m,n,k,1,a,i*m*k,k,b,0,k,1,c,0,n);
-
- if(delta_cl){
-
- cl_mem a = layer.filters_cl;
- cl_mem b = layer.delta_cl;
- cl_mem c = layer.col_image_cl;
-
- gemm_ongpu_offset(1,0,n,k,m,1,a,0,n,b,i*k*m,k,0,c,0,k);
-
- col2im_ongpu(layer.col_image_cl, i*layer.c*layer.h*layer.w, layer.c, layer.h, layer.w, layer.size, layer.stride, layer.pad, delta_cl);
- }
- }
-}
-
-void pull_convolutional_layer(convolutional_layer layer)
-{
- cl_read_array(layer.filters_cl, layer.filters, layer.c*layer.n*layer.size*layer.size);
- cl_read_array(layer.biases_cl, layer.biases, layer.n);
- cl_read_array(layer.filter_updates_cl, layer.filter_updates, layer.c*layer.n*layer.size*layer.size);
- cl_read_array(layer.bias_updates_cl, layer.bias_updates, layer.n);
-}
-
-void push_convolutional_layer(convolutional_layer layer)
-{
- cl_write_array(layer.filters_cl, layer.filters, layer.c*layer.n*layer.size*layer.size);
- cl_write_array(layer.biases_cl, layer.biases, layer.n);
- cl_write_array(layer.filter_updates_cl, layer.filter_updates, layer.c*layer.n*layer.size*layer.size);
- cl_write_array(layer.bias_updates_cl, layer.bias_updates, layer.n);
-}
-
-void update_convolutional_layer_gpu(convolutional_layer layer)
-{
- int size = layer.size*layer.size*layer.c*layer.n;
- axpy_ongpu(layer.n, layer.learning_rate, layer.bias_updates_cl, 1, layer.biases_cl, 1);
- scal_ongpu(layer.n,layer.momentum, layer.bias_updates_cl, 1);
-
- scal_ongpu(size, 1.-layer.learning_rate*layer.decay, layer.filters_cl, 1);
- axpy_ongpu(size, layer.learning_rate, layer.filter_updates_cl, 1, layer.filters_cl, 1);
- scal_ongpu(size, layer.momentum, layer.filter_updates_cl, 1);
- pull_convolutional_layer(layer);
-}
-
-
-#endif
-
--
Gitblit v1.10.0