decagon-pytorch/src/torch_stablesort/torch_stablesort.cpp

#include <torch/extension.h>

#include <iostream>
#include <vector>
#include <algorithm>


template<typename fun, typename... Ts>
void dispatch(torch::Tensor input, Ts&& ... args) {
  switch(input.type().scalarType()) {
    case torch::ScalarType::Double:
      return fun<double>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::Float:
      return fun<float>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::Half:
      throw std::runtime_error("Half-precision float not supported");
    case torch::ScalarType::ComplexHalf:
      throw std::runtime_error("Half-precision complex float not supported");
    case torch::ScalarType::ComplexFloat:
      return fun<float64_t>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::ComplexDouble:
      return fun<float128_t>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::Long:
      return fun<int64_t>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::Int:
      return fun<int32_t>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::Short:
      return fun<int16_t>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::Char:
      return fun<int8_t>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::Byte:
      return fun<uint8_t>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::Bool:
      return fun<bool>(input, std::forward<Ts>(args)...);
    case torch::ScalarType::QInt32:
      throw std::runtime_error("QInt32 not supported");
    case torch::ScalarType::QInt16:
      throw std::runtime_error("QInt16 not supported");
    case torch::ScalarType::QInt8:
      throw std::runtime_error("QInt8 not supported");
    case torch::ScalarType::BFloat16:
      throw std::runtime_error("BFloat16 not supported");
    default:
      throw std::runtime_error("Unknown scalar type");
  }
}


std::vector<at::Tensor> stable_sort_forward(
  torch::Tensor input,
  int dim,
  bool descending,
  torch::optional<torch::Tensor> out = nullptr) {



  auto X = torch::cat({old_h, input}, /*dim=*/1);

  auto gate_weights = torch::addmm(bias, X, weights.transpose(0, 1));
  auto gates = gate_weights.chunk(3, /*dim=*/1);

  auto input_gate = torch::sigmoid(gates[0]);
  auto output_gate = torch::sigmoid(gates[1]);
  auto candidate_cell = torch::elu(gates[2], /*alpha=*/1.0);

  auto new_cell = old_cell + candidate_cell * input_gate;
  auto new_h = torch::tanh(new_cell) * output_gate;

  return {new_h,
          new_cell,
          input_gate,
          output_gate,
          candidate_cell,
          X,
          gate_weights};
}

/ tanh'(z) = 1 - tanh^2(z)
torch::Tensor d_tanh(torch::Tensor z) {
  return 1 - z.tanh().pow(2);
}

// elu'(z) = relu'(z) + { alpha * exp(z) if (alpha * (exp(z) - 1)) < 0, else 0}
torch::Tensor d_elu(torch::Tensor z, torch::Scalar alpha = 1.0) {
  auto e = z.exp();
  auto mask = (alpha * (e - 1)) < 0;
  return (z > 0).type_as(z) + mask.type_as(z) * (alpha * e);
}

std::vector<torch::Tensor> stable_sort_backward(
    torch::Tensor grad_h,
    torch::Tensor grad_cell,
    torch::Tensor new_cell,
    torch::Tensor input_gate,
    torch::Tensor output_gate,
    torch::Tensor candidate_cell,
    torch::Tensor X,
    torch::Tensor gate_weights,
    torch::Tensor weights) {
  auto d_output_gate = torch::tanh(new_cell) * grad_h;
  auto d_tanh_new_cell = output_gate * grad_h;
  auto d_new_cell = d_tanh(new_cell) * d_tanh_new_cell + grad_cell;

  auto d_old_cell = d_new_cell;
  auto d_candidate_cell = input_gate * d_new_cell;
  auto d_input_gate = candidate_cell * d_new_cell;

  auto gates = gate_weights.chunk(3, /*dim=*/1);
  d_input_gate *= d_sigmoid(gates[0]);
  d_output_gate *= d_sigmoid(gates[1]);
  d_candidate_cell *= d_elu(gates[2]);

  auto d_gates =
      torch::cat({d_input_gate, d_output_gate, d_candidate_cell}, /*dim=*/1);

  auto d_weights = d_gates.t().mm(X);
  auto d_bias = d_gates.sum(/*dim=*/0, /*keepdim=*/true);

  auto d_X = d_gates.mm(weights);
  const auto state_size = grad_h.size(1);
  auto d_old_h = d_X.slice(/*dim=*/1, 0, state_size);
  auto d_input = d_X.slice(/*dim=*/1, state_size);

  return {d_old_h, d_input, d_weights, d_bias, d_old_cell};
}

std::vector<torch::Tensor> stable_argsort_forward() {

}

std::vector<torch::Tensor> stable_argsort_backward() {

}

PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
  m.def("stable_sort_forward", &stable_sort_forward, "Stable sort forward");
  m.def("stable_sort_backward", &stable_sort_backward, "Stable sort backward");
  m.def("stable_argsort_forward", &stable_argsort_forward, "Stable argsort forward");
  m.def("stable_argsort_backward", &stable_argsort_backward, "Stable argsort backward");
}