wehub-resource-sync
112 lines
4.6 KiB
C++
112 lines
4.6 KiB
C++
// Copyright (c) 2024 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 <math.h>
|
|
|
|
#include "paddle/phi/kernels/funcs/eigen/common.h"
|
|
#include "paddle/phi/kernels/funcs/eigen/eigen_function.h"
|
|
#include "paddle/phi/kernels/nadam_kernel.h"
|
|
|
|
namespace phi {
|
|
|
|
template <typename T, typename Context>
|
|
void NAdamKernel(const Context& dev_ctx,
|
|
const DenseTensor& param,
|
|
const DenseTensor& grad,
|
|
const DenseTensor& learning_rate,
|
|
const DenseTensor& momentum_decay_pow,
|
|
const DenseTensor& beta2_pow,
|
|
const DenseTensor& mu_product,
|
|
const DenseTensor& moment1,
|
|
const DenseTensor& moment2,
|
|
const optional<DenseTensor>& master_param UNUSED,
|
|
float beta1,
|
|
float beta2,
|
|
float epsilon,
|
|
float momentum_decay,
|
|
bool multi_precision UNUSED,
|
|
DenseTensor* param_out,
|
|
DenseTensor* momentum_decay_pow_out,
|
|
DenseTensor* beta2_pow_out,
|
|
DenseTensor* mu_product_out,
|
|
DenseTensor* moment1_out,
|
|
DenseTensor* moment2_out,
|
|
DenseTensor* master_param_out UNUSED) {
|
|
dev_ctx.template Alloc<T>(param_out);
|
|
dev_ctx.template Alloc<T>(momentum_decay_pow_out);
|
|
dev_ctx.template Alloc<T>(beta2_pow_out);
|
|
dev_ctx.template Alloc<T>(mu_product_out);
|
|
dev_ctx.template Alloc<T>(moment1_out);
|
|
dev_ctx.template Alloc<T>(moment2_out);
|
|
|
|
T beta1_ = static_cast<T>(beta1);
|
|
T beta2_ = static_cast<T>(beta2);
|
|
T epsilon_ = static_cast<T>(epsilon);
|
|
T momentum_decay_ = static_cast<T>(momentum_decay);
|
|
|
|
auto eigen_param = EigenVector<T>::Flatten(param);
|
|
auto eigen_grad = EigenVector<T>::Flatten(grad);
|
|
auto eigen_lr = EigenVector<T>::Flatten(learning_rate);
|
|
auto eigen_momentum_decay_pow = EigenVector<T>::Flatten(momentum_decay_pow);
|
|
auto eigen_beta2_pow = EigenVector<T>::Flatten(beta2_pow);
|
|
auto eigen_mu_product = EigenVector<T>::Flatten(mu_product);
|
|
auto eigen_moment1 = EigenVector<T>::Flatten(moment1);
|
|
auto eigen_moment2 = EigenVector<T>::Flatten(moment2);
|
|
|
|
auto eigen_param_out = EigenVector<T>::Flatten(*param_out);
|
|
auto eigen_momentum_decay_pow_out =
|
|
EigenVector<T>::Flatten(*momentum_decay_pow_out);
|
|
auto eigen_beta2_pow_out = EigenVector<T>::Flatten(*beta2_pow_out);
|
|
auto eigen_mu_product_out = EigenVector<T>::Flatten(*mu_product_out);
|
|
auto eigen_moment1_out = EigenVector<T>::Flatten(*moment1_out);
|
|
auto eigen_moment2_out = EigenVector<T>::Flatten(*moment2_out);
|
|
|
|
eigen_momentum_decay_pow_out =
|
|
eigen_momentum_decay_pow * static_cast<T>(0.96);
|
|
eigen_beta2_pow_out = eigen_beta2_pow * beta2_;
|
|
|
|
auto eigen_mu_t =
|
|
beta1_ *
|
|
(static_cast<T>(1) -
|
|
static_cast<T>(0.5) * eigen_momentum_decay_pow_out.pow(momentum_decay_));
|
|
auto eigen_mu_t_1 =
|
|
beta1_ *
|
|
(static_cast<T>(1) -
|
|
static_cast<T>(0.5) * eigen_momentum_decay_pow_out.pow(momentum_decay_) *
|
|
std::pow(static_cast<T>(0.96), momentum_decay_));
|
|
|
|
eigen_mu_product_out = eigen_mu_product * eigen_mu_t;
|
|
auto eigen_mu_product_t_1 = eigen_mu_product_out * eigen_mu_t_1;
|
|
|
|
eigen_moment1_out =
|
|
beta1_ * eigen_moment1 + (static_cast<T>(1) - beta1_) * eigen_grad;
|
|
eigen_moment2_out = beta2_ * eigen_moment2 +
|
|
(static_cast<T>(1) - beta2_) * eigen_grad * eigen_grad;
|
|
|
|
Eigen::DSizes<int, 1> p_dsize(param_out->numel());
|
|
auto eigen_moment1_hat = eigen_mu_t_1 * eigen_moment1_out /
|
|
(static_cast<T>(1) - eigen_mu_product_t_1) +
|
|
(static_cast<T>(1) - eigen_mu_t) * eigen_grad /
|
|
(static_cast<T>(1) - eigen_mu_product_out);
|
|
auto eigen_moment2_hat =
|
|
eigen_moment2_out / (static_cast<T>(1) - eigen_beta2_pow_out);
|
|
|
|
eigen_param_out = eigen_param - eigen_lr.broadcast(p_dsize) *
|
|
eigen_moment1_hat /
|
|
(eigen_moment2_hat.sqrt() + epsilon_);
|
|
}
|
|
} // namespace phi
|