diff --git a/libnd4j/include/ops/declarable/generic/nn/fusedBatchNorm.cpp b/libnd4j/include/ops/declarable/generic/nn/fusedBatchNorm.cpp
index ccdf60f40..d9e48d1c1 100644
--- a/libnd4j/include/ops/declarable/generic/nn/fusedBatchNorm.cpp
+++ b/libnd4j/include/ops/declarable/generic/nn/fusedBatchNorm.cpp
@@ -89,8 +89,8 @@ namespace sd {
             else {
                 //REQUIRE_TRUE(block.width() == 3, 0, "CUSTOM_OP fused_batch_norm: when isTraining=true then number of input arrays must be equal to 3, but got %i instead !", block.width());
                 std::vector<Nd4jLong> shape = {iD};
-                mean = NDArrayFactory::create_(scale->ordering(), shape, sd::DataType::FLOAT32, block.launchContext());
-                variance = NDArrayFactory::create_(scale->ordering(), shape, sd::DataType::FLOAT32, block.launchContext());
+                mean = NDArrayFactory::create_(scale->ordering(), shape, scale->dataType(), block.launchContext());
+                variance = NDArrayFactory::create_(scale->ordering(), shape, scale->dataType(), block.launchContext());
             }
 
 
@@ -104,7 +104,7 @@ namespace sd {
 
             const int restSize = x->lengthOf() / iD;
 
-            auto xAffected = NDArrayFactory::create(x->ordering(), {restSize, iD}, sd::DataType::FLOAT32, block.launchContext());
+            auto xAffected = NDArrayFactory::create(x->ordering(), {restSize, iD}, mean->dataType(), block.launchContext());
             xAffected.assign(xCast);
 
             const int restSizeMinusOne = (restSize > 1) ? (restSize - 1) : 1;