wehub-resource-sync
112 lines
3.7 KiB
C++
112 lines
3.7 KiB
C++
/* Copyright (c) 2021 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. */
|
|
|
|
#include "paddle/phi/kernels/scale_kernel.h"
|
|
|
|
#include "paddle/phi/backends/cpu/cpu_context.h"
|
|
#include "paddle/phi/common/scalar.h"
|
|
#include "paddle/phi/core/dense_tensor.h"
|
|
#include "paddle/phi/core/kernel_registry.h"
|
|
#include "paddle/phi/kernels/funcs/eigen/common.h"
|
|
|
|
#include "paddle/phi/kernels/funcs/eigen/eigen_function.h"
|
|
namespace phi {
|
|
|
|
template <typename T, typename Context>
|
|
void ScaleKernel(const Context& dev_ctx,
|
|
const DenseTensor& x,
|
|
const Scalar& scale,
|
|
const Scalar& bias,
|
|
bool bias_after_scale,
|
|
DenseTensor* out) {
|
|
// calc
|
|
dev_ctx.template Alloc<T>(out);
|
|
auto eigen_out = EigenVector<T>::Flatten(*out);
|
|
auto eigen_x = EigenVector<T>::Flatten(x);
|
|
auto& dev = *dev_ctx.eigen_device();
|
|
// TODO(chenweihang): now the eigen function here need the dtype of scale,
|
|
// eigen_x, bias should be same, so here need cast for two scalar arg,
|
|
// maybe we declare that the type of scale and bias is T?
|
|
if (x.numel() <= 0 || (!x.IsInitialized())) {
|
|
return;
|
|
}
|
|
funcs::EigenScale<std::decay_t<decltype(dev)>, T>::Eval(
|
|
dev, eigen_out, eigen_x, scale.to<T>(), bias.to<T>(), bias_after_scale);
|
|
}
|
|
|
|
template <typename T, typename Context>
|
|
void DivScaleKernel(const Context& dev_ctx,
|
|
const DenseTensor& x,
|
|
const Scalar& scale,
|
|
DenseTensor* out) {
|
|
dev_ctx.template Alloc<T>(out);
|
|
auto eigen_out = EigenVector<T>::Flatten(*out);
|
|
auto eigen_x = EigenVector<T>::Flatten(x);
|
|
auto& dev = *dev_ctx.eigen_device();
|
|
if (x.numel() <= 0 || (!x.IsInitialized())) {
|
|
return;
|
|
}
|
|
funcs::EigenDiv<std::decay_t<decltype(dev)>, T>::Eval(
|
|
dev, eigen_out, eigen_x, scale.to<T>());
|
|
}
|
|
|
|
#ifdef _WIN32
|
|
INSTANCE_SCALAR_KERNEL(int, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(int64_t, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(float, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(double, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(bfloat16, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(float16, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(uint8_t, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(int8_t, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(int16_t, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(complex64, CPUContext)
|
|
INSTANCE_SCALAR_KERNEL(complex128, CPUContext)
|
|
#endif
|
|
} // namespace phi
|
|
|
|
PD_REGISTER_KERNEL(scale,
|
|
CPU,
|
|
ALL_LAYOUT,
|
|
phi::ScaleKernel,
|
|
bool,
|
|
float,
|
|
double,
|
|
phi::bfloat16,
|
|
phi::float16,
|
|
uint8_t,
|
|
int8_t,
|
|
int16_t,
|
|
int,
|
|
int64_t,
|
|
phi::complex64,
|
|
phi::complex128) {}
|
|
|
|
PD_REGISTER_KERNEL(div_scale,
|
|
CPU,
|
|
ALL_LAYOUT,
|
|
phi::DivScaleKernel,
|
|
bool,
|
|
float,
|
|
double,
|
|
phi::bfloat16,
|
|
phi::float16,
|
|
uint8_t,
|
|
int8_t,
|
|
int16_t,
|
|
int,
|
|
int64_t,
|
|
phi::complex64,
|
|
phi::complex128) {}
|