diff --git a/paddle/fluid/inference/lite/test_engine.cc b/paddle/fluid/inference/lite/test_engine.cc
index e505af19d5389c074c5777d0235dfa055d1395a7..8e65fa2fbe36d80c3848f282e0fede45a0cf52f2 100644
--- a/paddle/fluid/inference/lite/test_engine.cc
+++ b/paddle/fluid/inference/lite/test_engine.cc
@@ -29,6 +29,7 @@ namespace inference {
 namespace lite {
 
 using inference::lite::AddTensorToBlockDesc;
+using paddle::inference::lite::AddFetchListToBlockDesc;
 using inference::lite::CreateTensor;
 using inference::lite::serialize_params;
 
@@ -65,7 +66,7 @@ void make_fake_model(std::string* model, std::string* param) {
   AddTensorToBlockDesc(block_, "x", std::vector<int64_t>({2, 4}), true);
   AddTensorToBlockDesc(block_, "y", std::vector<int64_t>({2, 4}), true);
   AddTensorToBlockDesc(block_, "z", std::vector<int64_t>({2, 4}), false);
-  AddTensorToBlockDesc(block_, "out", std::vector<int64_t>({2, 4}), false);
+  AddFetchListToBlockDesc(block_, "out");
 
   *block_->add_ops() = *feed0->Proto();
   *block_->add_ops() = *feed1->Proto();
diff --git a/paddle/fluid/operators/lite/lite_engine_op_test.cc b/paddle/fluid/operators/lite/lite_engine_op_test.cc
index 76c963ac652687cb0f65a0497b5c994f82d0d7aa..14088351cc8959f3a5168573f80555a2272ac1f9 100644
--- a/paddle/fluid/operators/lite/lite_engine_op_test.cc
+++ b/paddle/fluid/operators/lite/lite_engine_op_test.cc
@@ -25,6 +25,7 @@
 USE_NO_KERNEL_OP(lite_engine)
 
 using paddle::inference::lite::AddTensorToBlockDesc;
+using paddle::inference::lite::AddFetchListToBlockDesc;
 using paddle::inference::lite::CreateTensor;
 using paddle::inference::lite::serialize_params;
 namespace paddle {
@@ -60,7 +61,7 @@ TEST(LiteEngineOp, engine_op) {
   AddTensorToBlockDesc(block_, "x", std::vector<int64_t>({2, 4}), true);
   AddTensorToBlockDesc(block_, "y", std::vector<int64_t>({2, 4}), true);
   AddTensorToBlockDesc(block_, "z", std::vector<int64_t>({2, 4}), false);
-  AddTensorToBlockDesc(block_, "out", std::vector<int64_t>({2, 4}), false);
+  AddFetchListToBlockDesc(block_, "out");
   *block_->add_ops() = *feed1->Proto();
   *block_->add_ops() = *feed0->Proto();
   *block_->add_ops() = *elt_add->Proto();
diff --git a/paddle/fluid/operators/lite/ut_helper.h b/paddle/fluid/operators/lite/ut_helper.h
index f83b2a1a85c4fbda3383c5723fd00fb5ef0f1fc7..bc049dae77df651c06e50e1722ff81e7b0739aa8 100644
--- a/paddle/fluid/operators/lite/ut_helper.h
+++ b/paddle/fluid/operators/lite/ut_helper.h
@@ -42,6 +42,16 @@ void AddTensorToBlockDesc(framework::proto::BlockDesc* block,
   desc.SetPersistable(persistable);
   *var = *desc.Proto();
 }
+
+void AddFetchListToBlockDesc(framework::proto::BlockDesc* block,
+                             const std::string& name) {
+  using framework::proto::VarType;
+  auto* var = block->add_vars();
+  framework::VarDesc desc(name);
+  desc.SetType(VarType::FETCH_LIST);
+  *var = *desc.Proto();
+}
+
 void serialize_params(std::string* str, framework::Scope* scope,
                       const std::vector<std::string>& params) {
   std::ostringstream os;