How to run tensorflow benchmark_model through ngraph cpu

[zhaify]

HI ALL,
I want to run TF benchmark_model through ngraph cpu, is there some doc to guide me do these?

I have try to load libngraph_bridge.so manually in benchmark_model.cc like below:

  TF_LoadLibrary(
      "libngraph_bridge.so",
      status);

and add customize_config into SessionOptions refer UT

  tensorflow::SessionOptions options;
    options.config.mutable_graph_options()
        ->mutable_optimizer_options()
        ->set_opt_level(tensorflow::OptimizerOptions_Level_L0);
    options.config.mutable_graph_options()
        ->mutable_rewrite_options()
        ->set_constant_folding(tensorflow::RewriterConfig::OFF);
    tensorflow::ConfigProto& config = options.config;
    auto* custom_config = options.config.mutable_graph_options()
                              ->mutable_rewrite_options()
                              ->add_custom_optimizers();
    custom_config->set_name("ngraph-optimizer");
    (*custom_config->mutable_parameter_map())["ngraph_backend"].set_s("CPU");
    (*custom_config->mutable_parameter_map())["device_id"].set_s("0");   
    options.config.mutable_graph_options()
        ->mutable_rewrite_options()
        ->set_min_graph_nodes(-1);

    options.config.mutable_graph_options()
        ->mutable_rewrite_options()
        ->set_meta_optimizer_iterations(tensorflow::RewriterConfig::ONE);

  config.set_allow_soft_placement(true);


  LOG(INFO) << "Got config, " << config.device_count_size() << " devices";

  session->reset(tensorflow::NewSession(options));
  graph_def->reset(new GraphDef());
  tensorflow::GraphDef tensorflow_graph;
  Status s = ReadBinaryProto(Env::Default(), graph, graph_def->get());

but these does not work!