wehub-resource-sync
84 lines
3.5 KiB
C++
84 lines
3.5 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/common/amp_type_traits.h"
|
|
#include "paddle/phi/kernels/adadelta_kernel.h"
|
|
#include "paddle/phi/kernels/funcs/eigen/common.h"
|
|
#include "paddle/phi/kernels/funcs/math_function.h"
|
|
|
|
namespace phi {
|
|
|
|
template <typename T, typename Context>
|
|
void AdadeltaKernel(const Context& dev_ctx,
|
|
const DenseTensor& param,
|
|
const DenseTensor& grad,
|
|
const DenseTensor& avg_squared_grad,
|
|
const DenseTensor& avg_squared_update,
|
|
const DenseTensor& learning_rate,
|
|
const optional<DenseTensor>& master_param,
|
|
float rho,
|
|
float epsilon,
|
|
bool multi_precision,
|
|
DenseTensor* param_out,
|
|
DenseTensor* avg_squared_grad_out,
|
|
DenseTensor* avg_squared_update_out,
|
|
DenseTensor* master_param_outs) {
|
|
using MT = typename dtype::template MPTypeTrait<T>::Type;
|
|
dev_ctx.template Alloc<T>(param_out);
|
|
dev_ctx.template Alloc<MT>(avg_squared_grad_out);
|
|
dev_ctx.template Alloc<MT>(avg_squared_update_out);
|
|
|
|
MT rho_ = static_cast<MT>(rho);
|
|
MT epsilon_ = static_cast<MT>(epsilon);
|
|
|
|
auto eigen_param = EigenVector<T>::Flatten(param);
|
|
auto eigen_grad = EigenVector<T>::Flatten(grad);
|
|
// Squared gradient accumulator
|
|
auto eigen_avg_squared_grad = EigenVector<MT>::Flatten(avg_squared_grad);
|
|
// Squared updates accumulator
|
|
auto eigen_avg_squared_update = EigenVector<MT>::Flatten(avg_squared_update);
|
|
auto eigen_param_out = EigenVector<T>::Flatten(*param_out);
|
|
auto eigen_avg_squared_grad_out =
|
|
EigenVector<MT>::Flatten(*avg_squared_grad_out);
|
|
auto eigen_avg_squared_update_out =
|
|
EigenVector<MT>::Flatten(*avg_squared_update_out);
|
|
auto& place = *dev_ctx.eigen_device();
|
|
auto eigen_grad_cast = eigen_grad.template cast<MT>();
|
|
eigen_avg_squared_grad_out.device(place) =
|
|
rho_ * eigen_avg_squared_grad + (1 - rho_) * eigen_grad_cast.square();
|
|
auto update =
|
|
-(((eigen_avg_squared_update + epsilon_).sqrt()) /
|
|
((eigen_avg_squared_grad_out + epsilon_).sqrt()) * eigen_grad_cast);
|
|
Eigen::DSizes<int, 1> m_dsize(avg_squared_update_out->numel());
|
|
auto lr = EigenVector<MT>::Flatten(learning_rate);
|
|
if (multi_precision) {
|
|
auto eigen_master_param_out = EigenVector<MT>::Flatten(*master_param_outs);
|
|
auto eigen_master_param = EigenVector<MT>::Flatten(*master_param);
|
|
|
|
eigen_master_param_out.device(place) =
|
|
eigen_master_param + lr.broadcast(m_dsize) * update;
|
|
eigen_param_out.device(place) =
|
|
(eigen_param.template cast<MT>() + lr.broadcast(m_dsize) * update)
|
|
.template cast<T>();
|
|
} else {
|
|
eigen_param_out.device(place) =
|
|
eigen_param + (lr.broadcast(m_dsize) * update).template cast<T>();
|
|
}
|
|
eigen_avg_squared_update_out.device(place) =
|
|
rho_ * eigen_avg_squared_update + (1 - rho_) * update.square();
|
|
}
|
|
|
|
} // namespace phi
|