188 lines
6.9 KiB
C++
188 lines
6.9 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/tensor_utils.h"
|
|
#include "paddle/phi/kernels/cast_kernel.h"
|
|
#include "paddle/phi/kernels/full_kernel.h"
|
|
#include "paddle/phi/kernels/funcs/eigen/common.h"
|
|
#include "paddle/phi/kernels/funcs/eigen/eigen_function.h"
|
|
#include "paddle/phi/kernels/impl/expand_kernel_impl.h"
|
|
|
|
namespace phi {
|
|
template <typename Context, typename T, int Dims>
|
|
void ExpandBackward(const Context& dev_ctx,
|
|
const DenseTensor& out_grad,
|
|
const std::vector<int>& reshape_dims_vec,
|
|
const std::vector<int>& reduce_dims_vec,
|
|
DenseTensor* in_grad) {
|
|
size_t reshape_size = reshape_dims_vec.size();
|
|
size_t reduce_size = reduce_dims_vec.size();
|
|
dev_ctx.template Alloc<T>(in_grad);
|
|
in_grad->data<T>();
|
|
|
|
if constexpr (std::is_same_v<T, dtype::float16> ||
|
|
std::is_same_v<T, dtype::bfloat16>) {
|
|
const DenseTensor out_grad_fp32 =
|
|
Cast<T, Context>(dev_ctx, out_grad, DataType::FLOAT32);
|
|
DenseTensor in_grad_fp32;
|
|
in_grad_fp32.Resize(in_grad->dims());
|
|
dev_ctx.template Alloc<float>(&in_grad_fp32);
|
|
|
|
auto x_grad = EigenVector<float>::Flatten(in_grad_fp32);
|
|
Eigen::DSizes<int64_t, Dims * 2> reshape_dims;
|
|
for (size_t i = 0; i < reshape_size; ++i) {
|
|
reshape_dims[i] = reshape_dims_vec[i];
|
|
}
|
|
Eigen::DSizes<int64_t, Dims> reduce_dims;
|
|
for (size_t i = 0; i < reduce_size; ++i) {
|
|
reduce_dims[i] = reduce_dims_vec[i];
|
|
}
|
|
const auto out_grad0 = EigenVector<float>::Flatten(out_grad_fp32);
|
|
auto& place = *dev_ctx.eigen_device();
|
|
funcs::EigenBroadcastGrad<std::decay_t<decltype(place)>, float, Dims>::Eval(
|
|
place, x_grad, out_grad0, reduce_dims, reshape_dims);
|
|
|
|
if constexpr (std::is_same_v<T, dtype::float16>) {
|
|
CastKernel<float, Context>(
|
|
dev_ctx, in_grad_fp32, DataType::FLOAT16, in_grad);
|
|
} else {
|
|
CastKernel<float, Context>(
|
|
dev_ctx, in_grad_fp32, DataType::BFLOAT16, in_grad);
|
|
}
|
|
} else {
|
|
auto x_grad = EigenVector<T>::Flatten(*in_grad);
|
|
Eigen::DSizes<int64_t, Dims * 2> reshape_dims;
|
|
for (size_t i = 0; i < reshape_size; ++i) {
|
|
reshape_dims[i] = reshape_dims_vec[i];
|
|
}
|
|
Eigen::DSizes<int64_t, Dims> reduce_dims;
|
|
for (size_t i = 0; i < reduce_size; ++i) {
|
|
reduce_dims[i] = reduce_dims_vec[i];
|
|
}
|
|
auto out_grad0 = EigenVector<T>::Flatten(out_grad);
|
|
auto& place = *dev_ctx.eigen_device();
|
|
funcs::EigenBroadcastGrad<std::decay_t<decltype(place)>, T, Dims>::Eval(
|
|
place, x_grad, out_grad0, reduce_dims, reshape_dims);
|
|
}
|
|
}
|
|
|
|
template <typename T, typename Context>
|
|
void ExpandGradKernel(const Context& dev_ctx,
|
|
const DenseTensor& x,
|
|
const DenseTensor& out_grad,
|
|
const IntArray& shape,
|
|
DenseTensor* in_grad) {
|
|
auto x_dims = x.dims();
|
|
auto out_grad_dims = out_grad.dims();
|
|
std::vector<int64_t> expand_shape = vectorize<int64_t>(out_grad_dims);
|
|
|
|
if (x.numel() == 0 || out_grad.numel() == 0 ||
|
|
(in_grad && in_grad->numel() == 0)) {
|
|
dev_ctx.template Alloc<T>(in_grad);
|
|
if (in_grad->numel() != 0) {
|
|
Full<T, Context>(dev_ctx, in_grad->dims(), 0, in_grad);
|
|
}
|
|
return;
|
|
}
|
|
|
|
if (in_grad->dims() == out_grad_dims) {
|
|
Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, in_grad);
|
|
return;
|
|
}
|
|
auto vec_in_dims = vectorize<int64_t>(x_dims);
|
|
auto diff = expand_shape.size() - vec_in_dims.size();
|
|
vec_in_dims.insert(vec_in_dims.begin(), diff, 1);
|
|
// 1. reshape_dims_vec is the broadcast parameter.
|
|
// 2. reduce_dims_vec is the dimension parameter to compute gradients. For
|
|
// each dimension expanded, the gradients should be summed to original
|
|
// size.
|
|
std::vector<int> repeat_times(vec_in_dims.size());
|
|
for (size_t i = 0; i < vec_in_dims.size(); ++i) {
|
|
repeat_times[i] = expand_shape[i] / vec_in_dims[i];
|
|
}
|
|
std::vector<int> reshape_dims_vec;
|
|
std::vector<int> reduce_dims_vec;
|
|
for (size_t i = 0; i < repeat_times.size(); ++i) {
|
|
reduce_dims_vec.push_back(reshape_dims_vec.size());
|
|
reshape_dims_vec.push_back(repeat_times[i]);
|
|
reshape_dims_vec.push_back(vec_in_dims[i]);
|
|
}
|
|
|
|
int dims = reduce_dims_vec.size();
|
|
|
|
PADDLE_ENFORCE_GE(dims,
|
|
0,
|
|
common::errors::InvalidArgument(
|
|
"The rank of the input 'Out@GRAD' for "
|
|
"expand_v2_grad op must be greater than or "
|
|
"equal to 0, but the value received is %d.",
|
|
dims));
|
|
PADDLE_ENFORCE_LE(dims,
|
|
MAX_RANK_SUPPORTED,
|
|
common::errors::InvalidArgument(
|
|
"The rank of the input 'Out@GRAD' for "
|
|
"expand_v2_grad op must be less than or equal "
|
|
"to %d, but the value received is %d.",
|
|
MAX_RANK_SUPPORTED,
|
|
dims));
|
|
switch (dims) {
|
|
case 0:
|
|
ExpandBackward<Context, T, 1>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
case 1:
|
|
ExpandBackward<Context, T, 1>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
case 2:
|
|
ExpandBackward<Context, T, 2>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
case 3:
|
|
ExpandBackward<Context, T, 3>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
case 4:
|
|
ExpandBackward<Context, T, 4>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
case 5:
|
|
ExpandBackward<Context, T, 5>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
case 6:
|
|
ExpandBackward<Context, T, 6>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
case 7:
|
|
ExpandBackward<Context, T, 7>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
case 8:
|
|
ExpandBackward<Context, T, 8>(
|
|
dev_ctx, out_grad, reshape_dims_vec, reduce_dims_vec, in_grad);
|
|
break;
|
|
default:
|
|
PADDLE_THROW(common::errors::InvalidArgument(
|
|
"Only support tensor with rank being between 1 and %d. But "
|
|
"received tensor's rank = %d.",
|
|
MAX_RANK_SUPPORTED,
|
|
dims));
|
|
}
|
|
}
|
|
|
|
} // namespace phi
|