wehub-resource-sync
109 lines
4.1 KiB
C++
109 lines
4.1 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.
|
|
|
|
#include "paddle/phi/kernels/fake_quantize_grad_kernel.h"
|
|
#include "paddle/phi/core/kernel_registry.h"
|
|
#include "paddle/phi/core/tensor_utils.h"
|
|
|
|
namespace phi {
|
|
|
|
template <typename T, typename Context>
|
|
void QuantizeGradFunc(const Context& dev_ctx,
|
|
const DenseTensor& dout,
|
|
DenseTensor* dx) {
|
|
PADDLE_ENFORCE_NOT_NULL(dx,
|
|
common::errors::PreconditionNotMet(
|
|
"The QuantizeGradFunc output dx is nullptr"));
|
|
// Initialize dx as same as d_out
|
|
dev_ctx.template Alloc<T>(dx);
|
|
phi::Copy(dev_ctx, dout, dev_ctx.GetPlace(), false, dx);
|
|
}
|
|
|
|
template <typename T, typename Context>
|
|
void FakeChannelWiseQuantizeDequantizeAbsMaxGradKernel(const Context& dev_ctx,
|
|
const DenseTensor& dout,
|
|
int bit_length,
|
|
int round_type,
|
|
int quant_axis,
|
|
DenseTensor* dx) {
|
|
QuantizeGradFunc<T, Context>(dev_ctx, dout, dx);
|
|
}
|
|
|
|
template <typename T, typename Context>
|
|
void FakeQuantizeDequantizeAbsMaxGradKernel(const Context& dev_ctx,
|
|
const DenseTensor& dout,
|
|
int bit_length,
|
|
int round_type,
|
|
DenseTensor* dx) {
|
|
QuantizeGradFunc<T, Context>(dev_ctx, dout, dx);
|
|
}
|
|
|
|
template <typename T, typename Context>
|
|
void FakeQuantizeDequantizeMovingAverageAbsMaxGradKernel(
|
|
const Context& dev_ctx,
|
|
const DenseTensor& dout,
|
|
float moving_rate,
|
|
int bit_length,
|
|
bool is_test,
|
|
int round_type,
|
|
DenseTensor* dx) {
|
|
QuantizeGradFunc<T, Context>(dev_ctx, dout, dx);
|
|
}
|
|
|
|
} // namespace phi
|
|
|
|
PD_REGISTER_KERNEL(fake_channel_wise_quantize_dequantize_abs_max_grad,
|
|
CPU,
|
|
ALL_LAYOUT,
|
|
phi::FakeChannelWiseQuantizeDequantizeAbsMaxGradKernel,
|
|
float) {
|
|
kernel->InputAt(0).SetBackend(phi::Backend::ALL_BACKEND);
|
|
}
|
|
|
|
PD_REGISTER_KERNEL(fake_quantize_dequantize_abs_max_grad,
|
|
CPU,
|
|
ALL_LAYOUT,
|
|
phi::FakeQuantizeDequantizeAbsMaxGradKernel,
|
|
float) {
|
|
kernel->InputAt(0).SetBackend(phi::Backend::ALL_BACKEND);
|
|
}
|
|
|
|
PD_REGISTER_KERNEL(fake_quantize_dequantize_moving_average_abs_max_grad,
|
|
CPU,
|
|
ALL_LAYOUT,
|
|
phi::FakeQuantizeDequantizeMovingAverageAbsMaxGradKernel,
|
|
float) {
|
|
kernel->InputAt(0).SetBackend(phi::Backend::ALL_BACKEND);
|
|
}
|
|
|
|
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
|
|
PD_REGISTER_KERNEL(fake_channel_wise_quantize_dequantize_abs_max_grad,
|
|
GPU,
|
|
ALL_LAYOUT,
|
|
phi::FakeChannelWiseQuantizeDequantizeAbsMaxGradKernel,
|
|
float) {}
|
|
PD_REGISTER_KERNEL(fake_quantize_dequantize_abs_max_grad,
|
|
GPU,
|
|
ALL_LAYOUT,
|
|
phi::FakeQuantizeDequantizeAbsMaxGradKernel,
|
|
float,
|
|
phi::float16) {}
|
|
PD_REGISTER_KERNEL(fake_quantize_dequantize_moving_average_abs_max_grad,
|
|
GPU,
|
|
ALL_LAYOUT,
|
|
phi::FakeQuantizeDequantizeMovingAverageAbsMaxGradKernel,
|
|
float,
|
|
phi::float16) {}
|
|
#endif
|