Ownership & move-only Graph

app/Accuracy/accuracy_check.cpp

@@ -34,7 +34,7 @@ int main() {
   }
   Shape sh({static_cast<size_t>(count_pic), 227, 227, 3});
   Tensor t = make_tensor<float>(res, sh);
-  Graph graph(6);
+  Graph graph;
   Shape sh1({1, 5, 5, 3});
   std::vector<float> vec;
   vec.reserve(75);
@@ -43,23 +43,29 @@ int main() {
   }
   Tensor input = t;
   Tensor output = make_tensor(vec, sh1);
-  auto a1 = std::make_shared<InputLayer>(kNchw, kNchw, 1, 2);
+  auto a1 = std::make_unique<InputLayer>(kNchw, kNchw, 1, 2);
+  Layer* a1_ptr = a1.get();
   std::vector<float> kernelvec = {1, 1, 1, 1, 1, 1, 1, 1, 1};
   Shape sh2({3, 3});
   Tensor kernel = make_tensor(kernelvec, sh2);
-  auto a2 = std::make_shared<ConvolutionalLayer>(1, 0, 0, kernel);
+  auto a2 = std::make_unique<ConvolutionalLayer>(1, 0, 0, kernel);
+  Layer* a2_ptr = a2.get();
   Shape poolshape = {2, 2};
-  auto a3 = std::make_shared<EWLayer>("linear", 2.0F, 3.0F);
-  auto a4 = std::make_shared<PoolingLayer>(poolshape, "average");
-  auto a5 = std::make_shared<OutputLayer>();
-  auto a6 = std::make_shared<FCLayer>();
-  graph.setInput(a1, input);
-  graph.makeConnection(a1, a2);
-  graph.makeConnection(a2, a3);
-  graph.makeConnection(a3, a4);
-  graph.makeConnection(a4, a5);
-  graph.makeConnection(a5, a6);
-  graph.setOutput(a5, output);
+  auto a3 = std::make_unique<EWLayer>("linear", 2.0F, 3.0F);
+  Layer* a3_ptr = a3.get();
+  auto a4 = std::make_unique<PoolingLayer>(poolshape, "average");
+  Layer* a4_ptr = a4.get();
+  auto a5 = std::make_unique<OutputLayer>();
+  Layer* a5_ptr = a5.get();
+  auto a6 = std::make_unique<FCLayer>();
+  Layer* a6_ptr = a6.get();
+  graph.setInput(a1_ptr, input);
+  graph.makeConnection(a1_ptr, a2_ptr);
+  graph.makeConnection(a2_ptr, a3_ptr);
+  graph.makeConnection(a3_ptr, a4_ptr);
+  graph.makeConnection(a4_ptr, a5_ptr);
+  graph.makeConnection(a5_ptr, a6_ptr);
+  graph.setOutput(a5_ptr, output);
   graph.inference();
   std::vector<float> tmp = *output.as<float>();
   std::vector<float> tmp_output = softmax<float>(*output.as<float>());