diff --git a/src/avgpool_layer.c b/src/avgpool_layer.c
index 1306fd38..2b595aac 100644
--- a/src/avgpool_layer.c
+++ b/src/avgpool_layer.c
@@ -1,5 +1,6 @@
 #include "avgpool_layer.h"
 #include "dark_cuda.h"
+#include "utils.h"
 #include <stdio.h>
 
 avgpool_layer make_avgpool_layer(int batch, int w, int h, int c)
diff --git a/src/batchnorm_layer.c b/src/batchnorm_layer.c
index 018294df..253ce288 100644
--- a/src/batchnorm_layer.c
+++ b/src/batchnorm_layer.c
@@ -1,5 +1,6 @@
 #include "batchnorm_layer.h"
 #include "blas.h"
+#include "utils.h"
 #include <stdio.h>
 
 layer make_batchnorm_layer(int batch, int w, int h, int c)
diff --git a/src/blas.c b/src/blas.c
index 96fc067f..212fccdb 100644
--- a/src/blas.c
+++ b/src/blas.c
@@ -1,4 +1,5 @@
 #include "blas.h"
+#include "utils.h"
 
 #include <math.h>
 #include <assert.h>
@@ -358,4 +359,4 @@ void fix_nan_and_inf_cpu(float *input, size_t size)
         if (isnan(val) || isinf(val))
             input[i] = 1.0f / i;  // pseudo random value
     }
-}
\ No newline at end of file
+}
diff --git a/src/gru_layer.c b/src/gru_layer.c
index b24bee94..787f46ea 100644
--- a/src/gru_layer.c
+++ b/src/gru_layer.c
@@ -36,36 +36,36 @@ layer make_gru_layer(int batch, int inputs, int outputs, int steps, int batch_no
     l.steps = steps;
     l.inputs = inputs;
 
-    l.input_z_layer = (layer*)malloc(sizeof(layer));
+    l.input_z_layer = (layer*)xcalloc(1,sizeof(layer));
     fprintf(stderr, "\t\t");
     *(l.input_z_layer) = make_connected_layer(batch, steps, inputs, outputs, LINEAR, batch_normalize);
     l.input_z_layer->batch = batch;
 
-    l.state_z_layer = (layer*)malloc(sizeof(layer));
+    l.state_z_layer = (layer*)xcalloc(1,sizeof(layer));
     fprintf(stderr, "\t\t");
     *(l.state_z_layer) = make_connected_layer(batch, steps, outputs, outputs, LINEAR, batch_normalize);
     l.state_z_layer->batch = batch;
 
 
 
-    l.input_r_layer = (layer*)malloc(sizeof(layer));
+    l.input_r_layer = (layer*)xcalloc(1,sizeof(layer));
     fprintf(stderr, "\t\t");
     *(l.input_r_layer) = make_connected_layer(batch, steps, inputs, outputs, LINEAR, batch_normalize);
     l.input_r_layer->batch = batch;
 
-    l.state_r_layer = (layer*)malloc(sizeof(layer));
+    l.state_r_layer = (layer*)xcalloc(1,sizeof(layer));
     fprintf(stderr, "\t\t");
     *(l.state_r_layer) = make_connected_layer(batch, steps, outputs, outputs, LINEAR, batch_normalize);
     l.state_r_layer->batch = batch;
 
 
 
-    l.input_h_layer = (layer*)malloc(sizeof(layer));
+    l.input_h_layer = (layer*)xcalloc(1,sizeof(layer));
     fprintf(stderr, "\t\t");
     *(l.input_h_layer) = make_connected_layer(batch, steps, inputs, outputs, LINEAR, batch_normalize);
     l.input_h_layer->batch = batch;
 
-    l.state_h_layer = (layer*)malloc(sizeof(layer));
+    l.state_h_layer = (layer*)xcalloc(1,sizeof(layer));
     fprintf(stderr, "\t\t");
     *(l.state_h_layer) = make_connected_layer(batch, steps, outputs, outputs, LINEAR, batch_normalize);
     l.state_h_layer->batch = batch;