fix examples

Signed-off-by: Hui Zhang <zhtclz@foxmail.com>

Author: zh794390558

gcompiler/example/c++/basic/perf.cc

@@ -140,7 +140,7 @@ Status InitializeSession(int num_threads, const string& graph,
  LOG(INFO) << "Loading TensorFlow.";
 
  tensorflow::SessionOptions options;
- // set thread pool
+ //// set thread pool
  tensorflow::ConfigProto& config = options.config;
  config.mutable_gpu_options()->set_allow_growth(true);
  // config.mutable_gpu_options()->set_per_process_gpu_memory_fraction(0.5);
@@ -158,9 +158,9 @@ Status InitializeSession(int num_threads, const string& graph,
  graph_optimizer->set_name("SubGraphFusionOptimizer");*/
 
  LOG(INFO) << "Got config, " << config.device_count_size() << " devices";
-
  session->reset(tensorflow::NewSession(options));
  graph_def->reset(new GraphDef());
+ LOG(INFO) << "new session ok!";
  // tensorflow::GraphDef tensorflow_graph;
  //Status s = ReadBinaryProto(Env::Default(), graph, graph_def->get());
  //if (!s.ok()) {

gcompiler/example/c++/benchmark.cc

@@ -14,16 +14,20 @@ namespace tensorflow {
 
 void perf_of_inference() {
  Perf::Config cfg;
- cfg.graph =
- "/tmp-data/test/self/delta_infer/delta_infer/example/python/result.pb";
-
- cfg.input_layer = "Placeholder:0,Placeholder_1:0";
-
- cfg.input_layer_shape = "1,100,768:1,100,100";
- cfg.input_layer_type = "float,int32";
- cfg.input_layer_values = "1,1";
- cfg.output_layer =
- "import/e2e_model/Model_1/attn_decoder/LASDecoder/decoding_output:0";
+ //cfg.graph = "/tmp-data/test/self/delta_infer/delta_infer/example/python/result.pb";
+ cfg.graph = "/nfs/project/gaoyonghu/tf-inference-lib/unit_test/crash/models/frozen_graph_crash.pb";
+
+ //cfg.input_layer = "Placeholder:0,Placeholder_1:0";
+ cfg.input_layer = "inputs:0";
+
+ //cfg.input_layer_shape = "1,100,768:1,100,100";
+ cfg.input_layer_shape = "32,1000,40,1";
+ //cfg.input_layer_type = "float,int32";
+ cfg.input_layer_type = "float";
+ //cfg.input_layer_values = "1,1";
+ cfg.input_layer_values = "";
+ //cfg.output_layer = "import/e2e_model/Model_1/attn_decoder/LASDecoder/decoding_output:0";
+ cfg.output_layer = "softmax_output:0";
  cfg.target_layer = "";
  cfg.num_threads = 1;
  Perf infer(&cfg);

gcompiler/example/c++/gemm_perf.cc

@@ -29,7 +29,7 @@ int main(int argc, const char** argv) {
  if (argc < 5) {
  fprintf(
  stderr,
- "ERROR: need paramter batch_size, seq_len, head_num, size_per_head");
+ "ERROR: need paramter batch_size, seq_len, head_num, size_per_head\n");
  exit(1);
  }
 

gcompiler/example/c++/perf_test.cc

@@ -64,6 +64,10 @@ int main(int argc, const char** argv) {
  // tensorflow::test(fd);
  std::vector<int> batchs{4, 8, 16, 32, 64, 128};
  std::vector<int> seqs{50, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000};
+ if (argc != 3){
+ LOG(INFO) << "usage: bin <graph-path> <batch-size>";
+ exit(1);
+ }
  const char* graph = argv[1];
  int batch = atoi(argv[2]);
  // for(auto btid : batchs) {