unfold_grad_kernel.cc 3.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94
// 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.

#include "paddle/phi/kernels/unfold_grad_kernel.h"

#include "paddle/phi/backends/xpu/enforce_xpu.h"
#include "paddle/phi/core/kernel_registry.h"
#include "paddle/phi/kernels/funcs/unfold_functor.h"

namespace phi {

template <typename T, typename Context>
void UnfoldGradKernel(const Context& ctx,
                      const DenseTensor& x,
                      const DenseTensor& out_grad,
                      const std::vector<int>& kernel_sizes,
                      const std::vector<int>& strides,
                      const std::vector<int>& paddings,
                      const std::vector<int>& dilations,
                      DenseTensor* x_grad) {
  using XPUType = typename XPUTypeTrait<T>::Type;
  ctx.template Alloc<T>(x_grad);
  const std::string data_format = phi::DataLayoutToString(x.layout());
  bool is_nchw = data_format == "NCHW";
  PADDLE_ENFORCE_EQ(is_nchw,
                    true,
                    phi::errors::PreconditionNotMet(
                        "Unfold grad op only supports datalayout == NCHW"));

  auto x_dims = x_grad->dims();
  int n = static_cast<int>(x_dims[0]);
  int c = static_cast<int>(x_dims[1]);
  int h = static_cast<int>(x_dims[2]);
  int w = static_cast<int>(x_dims[3]);

  int out_height = phi::funcs::CalcOutputSize(x_dims[2],
                                              kernel_sizes[0],
                                              dilations[0],
                                              paddings[0],
                                              paddings[2],
                                              strides[0]);
  int out_width = phi::funcs::CalcOutputSize(x_dims[3],
                                             kernel_sizes[1],
                                             dilations[1],
                                             paddings[1],
                                             paddings[3],
                                             strides[1]);

  xpu::ctx_guard RAII_GUARD(ctx.x_context());
  XPUType* out_grad_trans =
      RAII_GUARD.alloc_l3_or_gm<XPUType>(out_grad.numel());

  int r = xpu::transpose(
      ctx.x_context(),
      reinterpret_cast<const XPUType*>(out_grad.data<T>()),
      out_grad_trans,
      {n, c, kernel_sizes[0], kernel_sizes[1], out_height, out_width},
      {0, 4, 5, 1, 2, 3});
  PADDLE_ENFORCE_XDNN_SUCCESS(r, "transpose");

  r = xpu::col2im(ctx.x_context(),
                  out_grad_trans,
                  reinterpret_cast<XPUType*>(x_grad->data<T>()),
                  n,
                  c,
                  h,
                  w,
                  kernel_sizes,
                  strides,
                  paddings,
                  dilations,
                  is_nchw);
  PADDLE_ENFORCE_XDNN_SUCCESS(r, "col2im");
}

}  // namespace phi

PD_REGISTER_KERNEL(unfold_grad,
                   XPU,
                   ALL_LAYOUT,
                   phi::UnfoldGradKernel,
                   float,
                   phi::dtype::float16) {}