Added Linear Regression. This closes #1

Author: prabhuomkar

CMakeLists.txt

@@ -7,10 +7,10 @@ find_package(Torch REQUIRED)
 add_executable(pytorch-cpp main.cpp)
 target_link_libraries(pytorch-cpp "${TORCH_LIBRARIES}")
 
-add_executable(pytorch_basics tutorials/basics/pytorch_basics/main.cpp)
-target_link_libraries(pytorch_basics "${TORCH_LIBRARIES}")
-# add_executable(linear_regression tutorials/basics/linear_regression/main.cpp)
-# target_link_libraries(linear_regression "${TORCH_LIBRARIES}")
+# add_executable(pytorch_basics tutorials/basics/pytorch_basics/main.cpp)
+# target_link_libraries(pytorch_basics "${TORCH_LIBRARIES}")
+add_executable(linear_regression tutorials/basics/linear_regression/main.cpp)
+target_link_libraries(linear_regression "${TORCH_LIBRARIES}")
 # add_executable(logistic_regression tutorials/basics/logistic_regression/main.cpp)
 # target_link_libraries(logistic_regression "${TORCH_LIBRARIES}")
 # add_executable(feedforward_neural_network tutorials/basics/feedforward_neural_network/main.cpp)

tutorials/basics/linear_regression/main.cpp

@@ -1,7 +1,46 @@
 // Copyright 2019 Omkar Prabhu
 #include <torch/torch.h>
+#include <ATen/ATen.h>
 #include <iostream>
+#include <vector>
 
 int main() {
   std::cout << "Linear Regression" << std::endl;
+
+  // Hyper parameters
+  int input_size = 1;
+  int output_size = 1;
+  int num_epochs = 60;
+  double learning_rate = 0.001;
+
+  // Sample dataset
+  auto x_train = torch::randint(0, 10, {15, 1});
+  auto y_train = torch::randint(0, 10, {15, 1});
+
+  // Linear regression model
+  auto model = torch::nn::Linear(input_size, output_size);
+
+  // Loss and optimizer
+  auto criterion = torch::nn::L1Loss();
+  auto optimizer = torch::optim::SGD(model->parameters(), torch::optim::SGDOptions(learning_rate));
+
+  for (int epoch = 0; epoch < num_epochs; epoch++) {
+    // Array to tensors
+    auto inputs = x_train;
+    auto targets = y_train;
+
+    // Forward pass
+    auto outputs = model(inputs);
+    auto loss = criterion(outputs, targets);
+
+    // Backward and optimize
+    optimizer.zero_grad();
+    loss.backward();
+    optimizer.step();
+
+    if ((epoch+1) % 5 == 0) {
+      std::cout << "Epoch [" << (epoch+1) << "/" << num_epochs << "], Loss: " << loss.item().toFloat() << std::endl;
+    }
+  } 
+
 }