diff --git a/paddle/fluid/inference/analysis/ir_passes/tensorrt_subgraph_pass.cc b/paddle/fluid/inference/analysis/ir_passes/tensorrt_subgraph_pass.cc
index 4ffe5f575c232ccfc0089cb86e28737e56b32f94..9c42b83e7add348433635b1899087324e4e370d4 100644
--- a/paddle/fluid/inference/analysis/ir_passes/tensorrt_subgraph_pass.cc
+++ b/paddle/fluid/inference/analysis/ir_passes/tensorrt_subgraph_pass.cc
@@ -63,7 +63,6 @@ std::unique_ptr<framework::ir::Graph> analysis::TensorRtSubgraphPass::ApplyImpl(
 void TensorRtSubgraphPass::CreateTensorRTOp(framework::ir::Node *node,
                                             Graph *graph) const {
   auto *op_desc = node->Op();
-  static int counter{0};
   auto &subgraph = *Agent(node).subgraph();
   PADDLE_ENFORCE(!subgraph.empty());
 
@@ -192,8 +191,6 @@ void TensorRtSubgraphPass::CreateTensorRTOp(framework::ir::Node *node,
           block_desc.Proto()->SerializeAsString());
   SetAttr(op_desc->Proto(), "max_batch_size", Get<int>("max_batch_size"));
   SetAttr(op_desc->Proto(), "workspace_size", Get<int>("workspace_size"));
-  SetAttr(op_desc->Proto(), "engine_uniq_key",
-          "trt-" + std::to_string(counter++));
   SetAttr(op_desc->Proto(), "parameters", ExtractParameters(graph->Nodes()));
   SetAttr(op_desc->Proto(), "output_name_mapping", output_mapping);
 }
diff --git a/paddle/fluid/operators/tensorrt/tensorrt_engine_op.cc b/paddle/fluid/operators/tensorrt/tensorrt_engine_op.cc
index f1ab59e3972efe598472b2d18077b15ce88716c4..b993c55fad13e892efd51648b78704bec83bf2b4 100644
--- a/paddle/fluid/operators/tensorrt/tensorrt_engine_op.cc
+++ b/paddle/fluid/operators/tensorrt/tensorrt_engine_op.cc
@@ -29,7 +29,6 @@ class TensorRTEngineOpMaker : public framework::OpProtoAndCheckerMaker {
     AddInput("Xs", "A list of inputs.").AsDuplicable();
     AddOutput("Ys", "A list of outputs").AsDuplicable();
     AddAttr<std::string>("subgraph", "the subgraph.");
-    AddAttr<std::string>("engine_uniq_key", "unique key for the TRT engine.");
     AddAttr<int>("max_batch_size", "the maximum batch size.");
     AddAttr<int>("workspace_size", "the workspace size.");
     AddComment("TensorRT engine operator.");
diff --git a/paddle/fluid/operators/tensorrt/tensorrt_engine_op.h b/paddle/fluid/operators/tensorrt/tensorrt_engine_op.h
index c19c315f798316d78d36d5fce788faf4abd0686d..88c4f508474e66953b79fb92ff1eb0b53a539f07 100644
--- a/paddle/fluid/operators/tensorrt/tensorrt_engine_op.h
+++ b/paddle/fluid/operators/tensorrt/tensorrt_engine_op.h
@@ -65,7 +65,6 @@ using inference::tensorrt::TensorRTEngine;
 
 class TensorRTEngineOp : public framework::OperatorBase {
  private:
-  std::string engine_name_;
   std::vector<std::string> input_names_;
   std::unordered_set<std::string> param_names_;
   mutable std::unique_ptr<TensorRTEngine> trt_engine_;
@@ -78,7 +77,6 @@ class TensorRTEngineOp : public framework::OperatorBase {
                    const framework::VariableNameMap &outputs,
                    const framework::AttributeMap &attrs)
       : framework::OperatorBase(type, inputs, outputs, attrs) {
-    engine_name_ = Attr<std::string>("engine_uniq_key");
     input_names_ = Inputs("Xs");
     max_batch_size_ = Attr<int>("max_batch_size");
     workspace_size_ = Attr<int>("workspace_size");