/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve. 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/framework/eigen.h" #include "paddle/framework/op_registry.h" namespace paddle { namespace operators { template class DecayedAdagradOpKernel : public framework::OpKernel { public: void Compute(const framework::ExecutionContext& ctx) const override { auto param_out_tensor = ctx.Output("ParamOut"); auto moment_out_tensor = ctx.Output("MomentOut"); param_out_tensor->mutable_data(ctx.GetPlace()); moment_out_tensor->mutable_data(ctx.GetPlace()); float decay = ctx.Attr("decay"); float epsilon = ctx.Attr("epsilon"); auto param = framework::EigenVector::Flatten( *ctx.Input("Param")); auto grad = framework::EigenVector::Flatten( *ctx.Input("Grad")); auto moment = framework::EigenVector::Flatten( *ctx.Input("Moment")); auto lr = framework::EigenVector::Flatten( *ctx.Input("LearningRate")); auto param_out = framework::EigenVector::Flatten(*param_out_tensor); auto moment_out = framework::EigenVector::Flatten(*moment_out_tensor); auto& place = *ctx.template device_context().eigen_device(); moment_out.device(place) = decay * moment + (1 - decay) * grad * grad; Eigen::DSizes m_dsize(moment_out_tensor->numel()); param_out.device(place) = param - lr.broadcast(m_dsize) * grad / (moment_out.sqrt() + epsilon); } }; } // namespace operators } // namespace paddle