diff --git a/paddle/fluid/inference/tensorrt/engine.cc b/paddle/fluid/inference/tensorrt/engine.cc
index ef58057dadc0ed38ca80a1ad1e47c8a0f0c8dfa2..e151ef124a7510d3ea9bc16606bdbffe5c69091a 100644
--- a/paddle/fluid/inference/tensorrt/engine.cc
+++ b/paddle/fluid/inference/tensorrt/engine.cc
@@ -238,6 +238,7 @@ void TensorRTEngine::FreezeNetwork() {
   infer_engine_.reset(infer_builder_->buildEngineWithConfig(
       *network(), *infer_builder_config_));
 #else
+  infer_builder_config_->setFlag(nvinfer1::BuilderFlag::kSPARSE_WEIGHTS);
   infer_ptr<nvinfer1::IHostMemory> plan(infer_builder_->buildSerializedNetwork(
       *network(), *infer_builder_config_));
   infer_ptr<nvinfer1::IRuntime> runtime(createInferRuntime(&logger_));