mirror/darknet
1
0
Fork 0
mirror of https://github.com/pjreddie/darknet.git synced 2023-08-10 21:13:14 +03:00
Code Issues Projects Releases Wiki Activity

Optimized resizing of network for random=1

Browse Source
This commit is contained in:
AlexeyAB
2018-02-21 15:06:11 +03:00
parent 068090c3a4
commit 4b0be8c701
2 changed files with 22 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/convolutional_layer.c
View File

@ -359,6 +359,8 @@ void test_convolutional_layer()
void resize_convolutional_layer(convolutional_layer *l, int w, int h) void resize_convolutional_layer(convolutional_layer *l, int w, int h)
{ {
int old_w = l->w;
int old_h = l->h;
l->w = w; l->w = w;
l->h = h; l->h = h;
int out_w = convolutional_out_width(*l); int out_w = convolutional_out_width(*l);
@ -378,6 +380,7 @@ void resize_convolutional_layer(convolutional_layer *l, int w, int h)
} }
#ifdef GPU #ifdef GPU
if (old_w < w || old_h < h) {
cuda_free(l->delta_gpu); cuda_free(l->delta_gpu);
cuda_free(l->output_gpu); cuda_free(l->output_gpu);
@ -391,6 +394,7 @@ void resize_convolutional_layer(convolutional_layer *l, int w, int h)
l->x_gpu = cuda_make_array(l->output, l->batch*l->outputs); l->x_gpu = cuda_make_array(l->output, l->batch*l->outputs);
l->x_norm_gpu = cuda_make_array(l->output, l->batch*l->outputs); l->x_norm_gpu = cuda_make_array(l->output, l->batch*l->outputs);
} }
}
#ifdef CUDNN #ifdef CUDNN
cudnn_convolutional_setup(l); cudnn_convolutional_setup(l);
#endif #endif

14
src/network.c
View File

@ -328,6 +328,12 @@ int resize_network(network *net, int w, int h)
cuda_set_device(net->gpu_index); cuda_set_device(net->gpu_index);
if(gpu_index >= 0){ if(gpu_index >= 0){
cuda_free(net->workspace); cuda_free(net->workspace);
if (net->input_gpu) {
cuda_free(*net->input_gpu);
*net->input_gpu = 0;
cuda_free(*net->truth_gpu);
*net->truth_gpu = 0;
}
} }
#endif #endif
int i; int i;
@ -340,6 +346,7 @@ int resize_network(network *net, int w, int h)
//fflush(stderr); //fflush(stderr);
for (i = 0; i < net->n; ++i){ for (i = 0; i < net->n; ++i){
layer l = net->layers[i]; layer l = net->layers[i];
printf(" %d: layer = %d,", i, l.type);
if(l.type == CONVOLUTIONAL){ if(l.type == CONVOLUTIONAL){
resize_convolutional_layer(&l, w, h); resize_convolutional_layer(&l, w, h);
}else if(l.type == CROP){ }else if(l.type == CROP){
@ -371,13 +378,8 @@ int resize_network(network *net, int w, int h)
} }
#ifdef GPU #ifdef GPU
if(gpu_index >= 0){ if(gpu_index >= 0){
if(net->input_gpu) {
cuda_free(*net->input_gpu);
*net->input_gpu = 0;
cuda_free(*net->truth_gpu);
*net->truth_gpu = 0;
}
net->workspace = cuda_make_array(0, (workspace_size-1)/sizeof(float)+1); net->workspace = cuda_make_array(0, (workspace_size-1)/sizeof(float)+1);
printf(" CUDA allocate done! \n");
}else { }else {
free(net->workspace); free(net->workspace);
net->workspace = calloc(1, workspace_size); net->workspace = calloc(1, workspace_size);