diff --git a/deeplearning4j/deeplearning4j-cuda/src/test/java/org/deeplearning4j/ValidateCuDNN.java b/deeplearning4j/deeplearning4j-cuda/src/test/java/org/deeplearning4j/ValidateCuDNN.java
index a9e4d185f..1d393aaf4 100644
--- a/deeplearning4j/deeplearning4j-cuda/src/test/java/org/deeplearning4j/ValidateCuDNN.java
+++ b/deeplearning4j/deeplearning4j-cuda/src/test/java/org/deeplearning4j/ValidateCuDNN.java
@@ -248,7 +248,7 @@ public class ValidateCuDNN extends BaseDL4JTest {
             Nd4j.getRandom().setSeed(12345);
             INDArray features = Nd4j.rand(fShape);
             INDArray labels = Nd4j.rand(lShape);
-            Nd4j.getExecutioner().exec(new IsMax(labels, 1));
+            labels = Nd4j.exec(new IsMax(labels, 1));
 
             List<CuDNNValidationUtil.TestCase> testCaseList = new ArrayList<>();
 
@@ -256,7 +256,7 @@ public class ValidateCuDNN extends BaseDL4JTest {
             for (int i = 0; i < 6; i++) {
                 INDArray f = Nd4j.rand(fShape);
                 INDArray l = Nd4j.rand(lShape);
-                Nd4j.getExecutioner().exec(new IsMax(l, 1));
+                Nd4j.exec(new IsMax(l, 1))[0];
                 dataSets.add(new DataSet(f, l));
             }
             DataSetIterator iter = new ExistingDataSetIterator(dataSets);
diff --git a/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/objdetect/Yolo2OutputLayer.java b/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/objdetect/Yolo2OutputLayer.java
index 2610603bc..bb551cd3f 100644
--- a/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/objdetect/Yolo2OutputLayer.java
+++ b/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/objdetect/Yolo2OutputLayer.java
@@ -191,8 +191,8 @@ public class Yolo2OutputLayer extends AbstractLayer<org.deeplearning4j.nn.conf.l
         //We also need 1_ij^noobj, which is (a) no object, or (b) object present in grid cell, but this box doesn't
         // have the highest IOU
         INDArray mask1_ij_obj = Nd4j.create(DataType.BOOL, iou.shape(), 'c');
-        Nd4j.getExecutioner().execAndReturn(new IsMax(iou, mask1_ij_obj, 1));
-        Nd4j.getExecutioner().execAndReturn(new BroadcastMulOp(mask1_ij_obj, maskObjectPresentBool, mask1_ij_obj, 0,2,3));
+        Nd4j.exec(new IsMax(iou, mask1_ij_obj, 1));
+        Nd4j.exec(new BroadcastMulOp(mask1_ij_obj, maskObjectPresentBool, mask1_ij_obj, 0,2,3));
         INDArray mask1_ij_noobj = Transforms.not(mask1_ij_obj);
         mask1_ij_obj = mask1_ij_obj.castTo(input.dataType());
 
diff --git a/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/pooling/GlobalPoolingLayer.java b/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/pooling/GlobalPoolingLayer.java
index 413fa0983..fc3b155c2 100644
--- a/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/pooling/GlobalPoolingLayer.java
+++ b/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/pooling/GlobalPoolingLayer.java
@@ -296,7 +296,7 @@ public class GlobalPoolingLayer extends AbstractLayer<org.deeplearning4j.nn.conf
 
         switch (poolingType) {
             case MAX:
-                INDArray isMax = Nd4j.getExecutioner().exec(new IsMax(inputArray.dup(), poolDim));
+                INDArray isMax = Nd4j.exec(new IsMax(inputArray.dup(), poolDim))[0];
                 return Nd4j.getExecutioner().exec(new BroadcastMulOp(isMax, epsilon, isMax, broadcastDims));
             case AVG:
                 //if out = avg(in,dims) then dL/dIn = 1/N * dL/dOut
diff --git a/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/util/MaskedReductionUtil.java b/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/util/MaskedReductionUtil.java
index 3e48b6dbc..5ea4fc8c3 100644
--- a/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/util/MaskedReductionUtil.java
+++ b/deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/util/MaskedReductionUtil.java
@@ -136,7 +136,7 @@ public class MaskedReductionUtil {
                 Nd4j.getExecutioner().exec(new BroadcastAddOp(input, negInfMask, withInf, 0, 2));
                 //At this point: all the masked out steps have value -inf, hence can't be the output of the MAX op
 
-                INDArray isMax = Nd4j.getExecutioner().exec(new IsMax(withInf, 2));
+                INDArray isMax = Nd4j.exec(new IsMax(withInf, 2))[0];
 
                 return Nd4j.getExecutioner().exec(new BroadcastMulOp(isMax, epsilon2d, isMax, 0, 1));
             case AVG:
@@ -296,7 +296,7 @@ public class MaskedReductionUtil {
                 Nd4j.getExecutioner().exec(new BroadcastAddOp(input, negInfMask, withInf, dimensions));
                 //At this point: all the masked out steps have value -inf, hence can't be the output of the MAX op
 
-                INDArray isMax = Nd4j.getExecutioner().exec(new IsMax(withInf, 2, 3));
+                INDArray isMax = Nd4j.exec(new IsMax(withInf, 2, 3))[0];
 
                 return Nd4j.getExecutioner().exec(new BroadcastMulOp(isMax, epsilon2d, isMax, 0, 1));
             case AVG: