diff --git a/fastdeploy/runtime/backends/paddle/paddle_backend.cc b/fastdeploy/runtime/backends/paddle/paddle_backend.cc
index e817f272b..6df92d279 100644
--- a/fastdeploy/runtime/backends/paddle/paddle_backend.cc
+++ b/fastdeploy/runtime/backends/paddle/paddle_backend.cc
@@ -88,13 +88,6 @@ void PaddleBackend::BuildOption(const PaddleBackendOption& option) {
   if (!option.enable_log_info) {
     config_.DisableGlogInfo();
   }
-  if (!option.delete_pass_names.empty()) {
-    auto pass_builder = config_.pass_builder();
-    for (int i = 0; i < option.delete_pass_names.size(); i++) {
-      FDINFO << "Delete pass : " << option.delete_pass_names[i] << std::endl;
-      pass_builder->DeletePass(option.delete_pass_names[i]);
-    }
-  }
   if (option.cpu_thread_num <= 0) {
     config_.SetCpuMathLibraryNumThreads(8);
   } else {
@@ -239,6 +232,15 @@ bool PaddleBackend::InitFromPaddle(const std::string& model_buffer,
            << " to set TensorRT dynamic shape." << std::endl;
     config_.EnableTunedTensorRtDynamicShape(shape_range_info, false);
   }
+  // Note(zhoushunjie): The pass deletion should be executed just before
+  // creating predictor.
+  if (!option.delete_pass_names.empty()) {
+    auto pass_builder = config_.pass_builder();
+    for (int i = 0; i < option.delete_pass_names.size(); i++) {
+      FDINFO << "Delete pass : " << option.delete_pass_names[i] << std::endl;
+      pass_builder->DeletePass(option.delete_pass_names[i]);
+    }
+  }
   predictor_ = paddle_infer::CreatePredictor(config_);
   initialized_ = true;
   return true;