diff --git a/paddle/fluid/inference/api/details/zero_copy_tensor.cc b/paddle/fluid/inference/api/details/zero_copy_tensor.cc
index 66dec0157d98e776b38ec8af81a0c006bc732bf4..77ab6bd590e552642387c1f400cd07a7f31bec8d 100644
--- a/paddle/fluid/inference/api/details/zero_copy_tensor.cc
+++ b/paddle/fluid/inference/api/details/zero_copy_tensor.cc
@@ -23,7 +23,8 @@
 #include "paddle/fluid/platform/float16.h"
 #include "paddle/phi/core/allocator.h"
 #ifdef PADDLE_WITH_ONNXRUNTIME
-#include "paddle/fluid/inference/api/onnxruntime_predictor.h"
+#include "onnxruntime_c_api.h"    // NOLINT
+#include "onnxruntime_cxx_api.h"  // NOLINT
 #endif
 
 namespace paddle_infer {
diff --git a/paddle/fluid/inference/api/onnxruntime_predictor.cc b/paddle/fluid/inference/api/onnxruntime_predictor.cc
index bd9de252a0962bc27a23b949b428d8f18f96190f..eb561667fe1f3d1f624dd928fc53780bb5e63cf2 100644
--- a/paddle/fluid/inference/api/onnxruntime_predictor.cc
+++ b/paddle/fluid/inference/api/onnxruntime_predictor.cc
@@ -81,7 +81,7 @@ bool CheckConvertToONNX(const AnalysisConfig &config) {
 bool ONNXRuntimePredictor::Init() {
   VLOG(3) << "ONNXRuntime Predictor::init()";
 
-  // Now ONNXRuntime only suuport CPU
+  // Now ONNXRuntime only support CPU
   const char *device_name = config_.use_gpu() ? "Cuda" : "Cpu";
   if (config_.use_gpu()) {
     place_ = paddle::platform::CUDAPlace(config_.gpu_device_id());
diff --git a/paddle/fluid/inference/api/onnxruntime_predictor_tester.cc b/paddle/fluid/inference/api/onnxruntime_predictor_tester.cc
index 2be2de9c60bb1c3fdedf13212d50a6f4e155d4df..4a702edacc903cb9bea53d71b365481f14847c0e 100644
--- a/paddle/fluid/inference/api/onnxruntime_predictor_tester.cc
+++ b/paddle/fluid/inference/api/onnxruntime_predictor_tester.cc
@@ -49,10 +49,6 @@ TEST(ONNXRuntimePredictor, onnxruntime_on) {
 
   ASSERT_TRUE(predictor);
   ASSERT_TRUE(!predictor->Clone());
-  ASSERT_TRUE(predictor->scope_);
-  ASSERT_TRUE(predictor->sub_scope_);
-  ASSERT_EQ(predictor->scope_->parent(), nullptr);
-  ASSERT_EQ(predictor->sub_scope_->parent(), predictor->scope_.get());
   // Dummy Input Data
   std::vector<int64_t> input_shape = {-1, 3, 224, 224};
   std::vector<float> input_data(1 * 3 * 224 * 224, 1.0);