Files
paddlepaddle--paddle/paddle/phi/kernels/impl/gumbel_softmax_grad_kernel_impl.h
T
2026-07-13 12:40:42 +08:00

62 lines
2.0 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 "paddle/phi/core/dense_tensor.h"
#include "paddle/phi/kernels/funcs/axis_utils.h"
#include "paddle/phi/kernels/funcs/math_function.h"
#include "paddle/phi/kernels/funcs/softmax.h"
#include "paddle/phi/kernels/funcs/softmax_impl.h"
namespace phi {
template <typename T, typename Context>
void GumbelSoftmaxGradKernel(const Context& dev_ctx,
const DenseTensor& out,
const DenseTensor& dout,
int axis,
DenseTensor* dx) {
const int rank = dx->dims().size();
axis = funcs::CanonicalAxis(axis, rank);
int64_t axis_dim = dx->dims()[axis];
// allocate memory on device.
dev_ctx.template Alloc<T>(dx);
if (dx->numel() == 0) {
return;
}
// For 0D Tensor
if (rank == 0) {
funcs::set_constant(dev_ctx, dx, static_cast<T>(0.0));
return;
}
// TODO(large-tensor): SoftmaxGradFunctor not support int64
PADDLE_ENFORCE_LE_INT_MAX(axis_dim, "axis_dim");
const int size_to_axis = funcs::SizeToAxis(axis, dx->dims());
const int size_from_axis = funcs::SizeFromAxis(axis, dx->dims());
DenseTensor dx_2d(*dx), out_2d(out), dout_2d(dout);
dx_2d.Resize({size_to_axis, size_from_axis});
out_2d.Resize({size_to_axis, size_from_axis});
dout_2d.Resize({size_to_axis, size_from_axis});
funcs::SoftmaxGradFunctor<Context, T>()(
dev_ctx, axis_dim, &out_2d, &dout_2d, &dx_2d);
}
} // namespace phi