vuron/adept: Платформа для обучения и вывода нейронных сетей, с API приближенным к PyTorch.

src/dispatch/batchnorm.hpp

37 строк2 KB

kolkir

Batchnorm2d backward implementation

23 фев 2025, 01:20

23 фев 2025, 01:20d9615da

#pragma once

#include <adept/dispatch/dispatcher.hpp>
#include <adept/tensor.hpp>

namespace adept {

inline std::tuple<Tensor, Tensor, Tensor> batchnorm2d_fwd(const Tensor& input,
                                                          const Tensor& weight,
                                                          const Tensor& bias,
                                                          Tensor& running_mean,
                                                          Tensor& running_var,
                                                          float32_t eps,
                                                          float32_t momentum,
                                                          bool is_train) {
  static auto& func = Dispatcher::instance().find("batchnorm2d_fwd");
  return Dispatcher::instance()
      .call<std::tuple<Tensor, Tensor, Tensor>, const Tensor&, const Tensor&, const Tensor&,
            Tensor&, Tensor&, float32_t, float32_t, bool>(func, input, weight, bias, running_mean,
                                                          running_var, eps, momentum, is_train);
}

inline std::tuple<Tensor, Tensor> batchnorm2d_bwd(Tensor& in_grad,
                                                  const Tensor& out_grad,
                                                  const Tensor& input,
                                                  const Tensor& weight,
                                                  const Tensor& save_mean,
                                                  const Tensor& save_var,
                                                  float32_t eps) {
  static auto& func = Dispatcher::instance().find("batchnorm2d_bwd");
  return Dispatcher::instance()
      .call<std::tuple<Tensor, Tensor>, Tensor&, const Tensor&, const Tensor&, const Tensor&,
            const Tensor&, const Tensor&, float32_t>(func, in_grad, out_grad, input, weight,
                                                     save_mean, save_var, eps);
}

}  // namespace adept