README.md

Switch (Swift LibTorch Wrapper)

This is some experimental attempt to wrap LibTorch, PyTorch's C++ API, with Swift.

Why

It's one of my random Sunday afternoon project, in which I am learning something new, so not much reason or discussion of pros/cons.

Goal

The goal of this project is to reproduce

the C++ end-to-end example

#include <torch/torch.h>

// Define a new Module.
struct Net : torch::nn::Module {
  Net() {
    // Construct and register two Linear submodules.
    fc1 = register_module("fc1", torch::nn::Linear(784, 64));
    fc2 = register_module("fc2", torch::nn::Linear(64, 32));
    fc3 = register_module("fc3", torch::nn::Linear(32, 10));
  }

  // Implement the Net's algorithm.
  torch::Tensor forward(torch::Tensor x) {
    // Use one of many tensor manipulation functions.
    x = torch::relu(fc1->forward(x.reshape({x.size(0), 784})));
    x = torch::dropout(x, /*p=*/0.5, /*train=*/is_training());
    x = torch::relu(fc2->forward(x));
    x = torch::log_softmax(fc3->forward(x), /*dim=*/1);
    return x;
  }

  // Use one of many "standard library" modules.
  torch::nn::Linear fc1{nullptr}, fc2{nullptr}, fc3{nullptr};
};

int main() {
  // Create a new Net.
  auto net = std::make_shared<Net>();

  // Create a multi-threaded data loader for the MNIST dataset.
  auto data_loader = torch::data::make_data_loader(
      torch::data::datasets::MNIST("./data").map(
          torch::data::transforms::Stack<>()),
      /*batch_size=*/64);

  // Instantiate an SGD optimization algorithm to update our Net's parameters.
  torch::optim::SGD optimizer(net->parameters(), /*lr=*/0.01);

  for (size_t epoch = 1; epoch <= 10; ++epoch) {
    size_t batch_index = 0;
    // Iterate the data loader to yield batches from the dataset.
    for (auto& batch : *data_loader) {
      // Reset gradients.
      optimizer.zero_grad();
      // Execute the model on the input data.
      torch::Tensor prediction = net->forward(batch.data);
      // Compute a loss value to judge the prediction of our model.
      torch::Tensor loss = torch::nll_loss(prediction, batch.target);
      // Compute gradients of the loss w.r.t. the parameters of our model.
      loss.backward();
      // Update the parameters based on the calculated gradients.
      optimizer.step();
      // Output the loss and checkpoint every 100 batches.
      if (++batch_index % 100 == 0) {
        std::cout << "Epoch: " << epoch << " | Batch: " << batch_index
                  << " | Loss: " << loss.item<float>() << std::endl;
        // Serialize your model periodically as a checkpoint.
        torch::save(net, "net.pt");
      }
    }
  }
}

Planning and Progress

• Set up a workspace for wrapping LibTorch with C and Swift.

• WIP: Port basic Tensor creation

  • Scalar
  • Tensor
    • Accessor
    • External Data
  • TensorOptions
    • dtype
    • device
  • Factory functions
    • variable size
    • arange
    • empty
    • eye
    • full
    • linspace
    • logspace
    • ones
    • rand
    • randn
    • randint
    • randperm
    • zeros

• Port basic Tensor manipulation.

  • resize
  • subscript

• Build computational graph with AutoGrad module.
• Wrap NN module.
• Build Module.

How to use

1. Build a Docker image for dev.

./docker/build.sh

This will build two images.

1. An image which only contains LibTorch built from source.
2. Development image with Swift and LibTorch.

2. Start a container

./start_container.sh

3. Run unit test

./run_test.sh