diff --git a/paddle/fluid/inference/api/api_tensorrt_subgraph_engine_tester.cc b/paddle/fluid/inference/api/api_tensorrt_subgraph_engine_tester.cc
index 8f1a72316d6c146ebc9a86ced739ef088a3b4267..9e7425eddd2df07ffe897f908aad360abe42117a 100644
--- a/paddle/fluid/inference/api/api_tensorrt_subgraph_engine_tester.cc
+++ b/paddle/fluid/inference/api/api_tensorrt_subgraph_engine_tester.cc
@@ -37,6 +37,7 @@ void CompareTensorRTWithFluid(bool enable_tensorrt) {
   config1.use_gpu = true;
   config1.fraction_of_gpu_memory = 0.3;
   config1.device = 0;
+  config1.max_batch_size = 10;
 
   auto predictor0 =
       CreatePaddlePredictor<NativeConfig, PaddleEngineKind::kNative>(config0);