For step profiling.

ge/executor/ge_executor.cc

@@ -813,6 +813,7 @@ Status GeExecutor::LoadModelWithQ(uint32_t &model_id, const ModelData &model_dat
 */
 Status GeExecutor::ExecModel(uint32_t model_id, void *stream, const ge::RunModelData &run_input_data,
                              ge::RunModelData &run_output_data, bool async_mode) {
+
   std::vector<GeTensorDesc> input_desc = {};
   std::vector<GeTensorDesc> output_desc = {};
   return ExecModel(model_id, stream, run_input_data, input_desc, run_output_data, output_desc, async_mode);