87 lines
2.7 KiB
C++
87 lines
2.7 KiB
C++
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
// you may not use this file except in compliance with the License.
|
|
// You may obtain a copy of the License at
|
|
//
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
//
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
// See the License for the specific language governing permissions and
|
|
// limitations under the License.
|
|
|
|
#pragma once
|
|
#include <string>
|
|
|
|
#include "paddle/common/hostdevice.h"
|
|
#include "paddle/phi/core/dense_tensor.h"
|
|
#include "paddle/phi/kernels/funcs/eigen/common.h"
|
|
|
|
namespace phi {
|
|
using Array1 = Eigen::DSizes<int64_t, 1>;
|
|
template <typename T>
|
|
struct KLDivLossBackward {
|
|
bool log_target = false;
|
|
|
|
HOSTDEVICE KLDivLossBackward(bool logTarget) : log_target(logTarget) {}
|
|
|
|
HOSTDEVICE T operator()(const T& target, const T& grad) const {
|
|
if (log_target) {
|
|
return static_cast<T>(-1.) * std::exp(target) * grad;
|
|
} else {
|
|
if (target <= 0) {
|
|
return 0;
|
|
} else {
|
|
return static_cast<T>(-1.) * target * grad;
|
|
}
|
|
}
|
|
}
|
|
};
|
|
|
|
template <typename T, typename Context>
|
|
void KLDivLossGradKernel(const Context& dev_ctx,
|
|
const DenseTensor& x UNUSED,
|
|
const DenseTensor& label,
|
|
const DenseTensor& d_out,
|
|
const std::string& reduction,
|
|
bool log_target,
|
|
DenseTensor* d_x) {
|
|
if (d_x->numel() == 0) {
|
|
dev_ctx.template Alloc<T>(d_x);
|
|
return;
|
|
}
|
|
|
|
auto& place = *dev_ctx.eigen_device();
|
|
auto* target = &label;
|
|
auto* input_grad = d_x;
|
|
auto* loss_grad = &d_out;
|
|
|
|
// TODO(large-tensor): downstream functors may still use int; guard until
|
|
// upgraded.
|
|
int64_t n = input_grad->dims()[0];
|
|
|
|
const int64_t numel = input_grad->numel();
|
|
const int64_t expand = numel / loss_grad->numel();
|
|
|
|
dev_ctx.template Alloc<T>(input_grad);
|
|
|
|
auto target_t = EigenVector<T>::Flatten(*target);
|
|
|
|
auto input_grad_t = EigenVector<T>::Flatten(*input_grad);
|
|
auto loss_grad_t = EigenVector<T>::Flatten(*loss_grad);
|
|
|
|
auto loss_grad_expand = loss_grad_t.broadcast(Array1(expand));
|
|
auto grad_t = loss_grad_expand;
|
|
input_grad_t.device(place) =
|
|
target_t.binaryExpr(grad_t, KLDivLossBackward<T>(log_target));
|
|
|
|
if ("mean" == reduction) {
|
|
input_grad_t.device(place) = input_grad_t / static_cast<T>(numel);
|
|
} else if ("batchmean" == reduction) {
|
|
input_grad_t.device(place) = input_grad_t / static_cast<T>(n);
|
|
}
|
|
}
|
|
} // namespace phi
|