diff --git a/paddle/fluid/inference/analysis/analyzer.cc b/paddle/fluid/inference/analysis/analyzer.cc
index 1249c404db53d14055b6301a2f8ef5b325a264ac..e51d6cfeb931d50a9a573df29c916ebd3da403d1 100644
--- a/paddle/fluid/inference/analysis/analyzer.cc
+++ b/paddle/fluid/inference/analysis/analyzer.cc
@@ -24,7 +24,7 @@
 
 namespace paddle {
 
-DEFINE_bool(inference_analysis_enable_tensorrt_subgraph_engine, true,
+DEFINE_bool(inference_analysis_enable_tensorrt_subgraph_engine, false,
             "Enable subgraph to TensorRT engine for acceleration");
 
 DEFINE_string(inference_analysis_graphviz_log_root, "./",
diff --git a/paddle/fluid/inference/api/api_tensorrt_subgraph_engine.cc b/paddle/fluid/inference/api/api_tensorrt_subgraph_engine.cc
index 93de7a5209e7dc289b4b02e73ef3bb20bfc8c774..48704ba235210c848c66bc4f28a4e780f8a32e84 100644
--- a/paddle/fluid/inference/api/api_tensorrt_subgraph_engine.cc
+++ b/paddle/fluid/inference/api/api_tensorrt_subgraph_engine.cc
@@ -32,7 +32,7 @@ class TensorRTSubgraphPredictor : public NativePaddlePredictor {
       : NativePaddlePredictor(config), config_(config) {}
 
   bool Init(const std::shared_ptr<framework::Scope>& parent_scope) {
-    FLAGS_IA_enable_tensorrt_subgraph_engine = true;
+    FLAGS_inference_analysis_enable_tensorrt_subgraph_engine = true;
     VLOG(3) << "Predictor::init()";
     FLAGS_tensorrt_max_batch_size = config_.max_batch_size;
     FLAGS_tensorrt_workspace_size = config_.workspace_size;