fix for CUDA streams

src/convolutional_layer.c

@@ -523,6 +523,7 @@ convolutional_layer make_convolutional_layer(int batch, int steps, int h, int w,
         stride_x = stride_y = l.stride = l.stride_x = l.stride_y = 1; // use stride=1 in host-layer
     }
 
+    l.wait_stream_id = -1;
     l.deform = deform;
     l.assisted_excitation = assisted_excitation;
     l.share_layer = share_layer;

src/route_layer.c

@@ -15,6 +15,7 @@ route_layer make_route_layer(int batch, int n, int *input_layers, int *input_siz
     l.input_sizes = input_sizes;
     l.groups = groups;
     l.group_id = group_id;
+    l.wait_stream_id = -1;
     int i;
     int outputs = 0;
     for(i = 0; i < n; ++i){