From 160eddddc4e265d5ee59a38797c30720bf46cd7c Mon Sep 17 00:00:00 2001
From: AlexeyAB <alexeyab84@gmail.com>
Date: Sun, 27 May 2018 13:53:42 +0000
Subject: [PATCH] Minor fix
---
src/network_kernels.cu | 41 +++++++++++++++++++++++++++++++++++++++--
1 files changed, 39 insertions(+), 2 deletions(-)
diff --git a/src/network_kernels.cu b/src/network_kernels.cu
index e319068..a11d61f 100644
--- a/src/network_kernels.cu
+++ b/src/network_kernels.cu
@@ -36,6 +36,10 @@
#include "blas.h"
}
+#ifdef OPENCV
+#include "opencv2/highgui/highgui_c.h"
+#endif
+
float * get_network_output_gpu_layer(network net, int i);
float * get_network_delta_gpu_layer(network net, int i);
float * get_network_output_gpu(network net);
@@ -51,7 +55,24 @@
fill_ongpu(l.outputs * l.batch, 0, l.delta_gpu, 1);
}
l.forward_gpu(l, state);
+ if(net.wait_stream)
+ cudaStreamSynchronize(get_cuda_stream());
state.input = l.output_gpu;
+/*
+ cuda_pull_array(l.output_gpu, l.output, l.batch*l.outputs);
+ if (l.out_w >= 0 && l.out_h >= 1 && l.c >= 3) {
+ int j;
+ for (j = 0; j < l.out_c; ++j) {
+ image img = make_image(l.out_w, l.out_h, 3);
+ memcpy(img.data, l.output+ l.out_w*l.out_h*j, l.out_w*l.out_h * 1 * sizeof(float));
+ char buff[256];
+ sprintf(buff, "layer-%d slice-%d", i, j);
+ show_image(img, buff);
+ }
+ cvWaitKey(0); // wait press-key in console
+ cvDestroyAllWindows();
+ }
+*/
}
}
@@ -64,6 +85,7 @@
for(i = net.n-1; i >= 0; --i){
state.index = i;
layer l = net.layers[i];
+ if (l.stopbackward) break;
if(i == 0){
state.input = original_input;
state.delta = original_delta;
@@ -78,11 +100,13 @@
void update_network_gpu(network net)
{
+ cuda_set_device(net.gpu_index);
int i;
int update_batch = net.batch*net.subdivisions;
float rate = get_current_rate(net);
for(i = 0; i < net.n; ++i){
layer l = net.layers[i];
+ l.t = get_current_batch(net);
if(l.update_gpu){
l.update_gpu(l, update_batch, rate, net.momentum, net.decay);
}
@@ -108,7 +132,15 @@
state.delta = 0;
state.truth = *net.truth_gpu;
state.train = 1;
+#ifdef CUDNN_HALF
+ int i;
+ for (i = 0; i < net.n; ++i) {
+ layer l = net.layers[i];
+ cuda_convert_f32_to_f16(l.weights_gpu, l.c*l.n*l.size*l.size, l.weights_gpu16);
+ }
+#endif
forward_network_gpu(net, state);
+ //cudaStreamSynchronize(get_cuda_stream());
backward_network_gpu(net, state);
}
@@ -176,6 +208,7 @@
{
int update_batch = net.batch*net.subdivisions;
float rate = get_current_rate(net);
+ l.t = get_current_batch(net);
if(l.update_gpu){
l.update_gpu(l, update_batch, rate, net.momentum, net.decay);
}
@@ -359,11 +392,14 @@
//printf("%f\n", errors[i]);
sum += errors[i];
}
+ //cudaDeviceSynchronize();
if (get_current_batch(nets[0]) % interval == 0) {
printf("Syncing... ");
+ fflush(stdout);
sync_nets(nets, n, interval);
printf("Done!\n");
}
+ //cudaDeviceSynchronize();
free(threads);
free(errors);
return (float)sum/(n);
@@ -372,7 +408,7 @@
float *get_network_output_layer_gpu(network net, int i)
{
layer l = net.layers[i];
- cuda_pull_array(l.output_gpu, l.output, l.outputs*l.batch);
+ if(l.type != REGION) cuda_pull_array(l.output_gpu, l.output, l.outputs*l.batch);
return l.output;
}
@@ -385,7 +421,8 @@
float *network_predict_gpu(network net, float *input)
{
- cuda_set_device(net.gpu_index);
+ if (net.gpu_index != cuda_get_device())
+ cuda_set_device(net.gpu_index);
int size = get_network_input_size(net) * net.batch;
network_state state;
state.index = 0;
--
Gitblit v1.10.0