You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
66 lines
2.3 KiB
66 lines
2.3 KiB
/* 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. */
|
|
|
|
#include "gather.cu.h"
|
|
#include "paddle/framework/eigen.h"
|
|
#include "paddle/operators/gather_op.h"
|
|
#include "scatter.cu.h"
|
|
|
|
namespace paddle {
|
|
namespace operators {
|
|
|
|
template <typename T>
|
|
class GatherOpCUDAKernel : public framework::OpKernel<T> {
|
|
public:
|
|
void Compute(const framework::ExecutionContext &ctx) const override {
|
|
PADDLE_ENFORCE(platform::is_gpu_place(ctx.GetPlace()),
|
|
"This kernel only runs on GPU device.");
|
|
auto *x = ctx.Input<Tensor>("X");
|
|
auto *index = ctx.Input<Tensor>("Index");
|
|
auto *output = ctx.Output<Tensor>("Out");
|
|
|
|
output->mutable_data<T>(ctx.GetPlace());
|
|
|
|
GPUGather<T>(ctx.device_context(), *x, *index, output);
|
|
}
|
|
};
|
|
|
|
template <typename T>
|
|
class GatherGradOpCUDAKernel : public framework::OpKernel<T> {
|
|
public:
|
|
void Compute(const framework::ExecutionContext &ctx) const override {
|
|
PADDLE_ENFORCE(platform::is_gpu_place(ctx.GetPlace()),
|
|
"This kernel only runs on GPU device.");
|
|
auto *Index = ctx.Input<Tensor>("Index");
|
|
auto *dX = ctx.Output<Tensor>(framework::GradVarName("X"));
|
|
auto *dO = ctx.Input<Tensor>(framework::GradVarName("Out"));
|
|
auto *x = ctx.Input<Tensor>("X");
|
|
|
|
dX->mutable_data<T>(ctx.GetPlace());
|
|
auto dxt = framework::EigenVector<T>::Flatten(*dX);
|
|
auto &place = *ctx.template device_context<platform::CUDADeviceContext>()
|
|
.eigen_device();
|
|
dxt.device(place) = dxt.constant(static_cast<T>(0));
|
|
|
|
GPUScatterAssign<T>(ctx.device_context(), *dO, *Index, dX);
|
|
}
|
|
};
|
|
|
|
} // namespace operators
|
|
} // namespace paddle
|
|
|
|
namespace ops = paddle::operators;
|
|
REGISTER_OP_CUDA_KERNEL(gather, ops::GatherOpCUDAKernel<float>);
|
|
REGISTER_OP_CUDA_KERNEL(gather_grad, ops::GatherGradOpCUDAKernel<float>);
|