Files
paddlepaddle--paddle/paddle/phi/kernels/onednn/dequantize_kernel.cc
T
2026-07-13 12:40:42 +08:00

110 lines
4.0 KiB
C++

// Copyright (c) 2023 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/dequantize_kernel.h"
#include "paddle/phi/backends/onednn/onednn_context.h"
#include "paddle/phi/backends/onednn/onednn_helper.h"
#include "paddle/phi/backends/onednn/onednn_reuse.h"
#include "paddle/phi/core/enforce.h"
#include "paddle/phi/core/kernel_registry.h"
namespace phi {
template <typename T, typename Context>
void DeQuantKernel(const Context& dev_ctx,
const DenseTensor& x,
const float quantization_scale,
const float quantization_shift,
DenseTensor* out) {
PADDLE_ENFORCE(quantization_scale != 0.0f,
common::errors::InvalidArgument(
"Dequantization scale must be different than 0.0f"));
const auto q_shift = static_cast<int32_t>(quantization_shift);
PADDLE_ENFORCE_GE(q_shift,
0,
common::errors::InvalidArgument(
"Dequantization shift must be greater or equal to 0"));
PADDLE_ENFORCE_LE(q_shift,
255,
common::errors::InvalidArgument(
"Dequantization shift must be lower or equal to 255"));
const bool with_shift = q_shift != 0;
auto x_tz = vectorize<int64_t>(x.dims());
auto x_type = funcs::ToOneDNNDataType(x.dtype());
auto out_type = funcs::ToOneDNNDataType(out->dtype());
dnnl::primitive_attr attrs;
static constexpr int32_t mask = 0; // same shift and scale for whole tensor
attrs.set_scales_mask(DNNL_ARG_DST, mask);
if (with_shift) {
attrs.set_zero_points_mask(DNNL_ARG_SRC, mask);
}
funcs::ReorderOneDNNHandler reorder_handler(
x_tz, x.dtype(), x_type, out->dtype(), out_type, dev_ctx.GetEngine());
auto reorder_src_memory_p = reorder_handler.AcquireSrcMemory(
x.mem_desc(), funcs::to_void_cast(x.data<T>()));
auto reorder_dst_memory_p =
reorder_handler.AcquireDstMemory(out, x.mem_desc(), dev_ctx.GetPlace());
auto reorder_p = reorder_handler.AcquireReorder(
reorder_dst_memory_p, reorder_src_memory_p, attrs);
auto& astream = OneDNNContext::tls().get_stream();
auto scales_md = dnnl::memory::desc(
{1}, dnnl::memory::data_type::f32, dnnl::memory::format_tag::x);
auto scales_mem =
dnnl::memory(scales_md,
dev_ctx.GetEngine(),
funcs::to_void_cast<float>(&quantization_scale));
auto zero_points_md = dnnl::memory::desc(
{1}, dnnl::memory::data_type::s32, dnnl::memory::format_tag::x);
auto zero_points_mem = dnnl::memory(zero_points_md,
dev_ctx.GetEngine(),
funcs::to_void_cast<int32_t>(&q_shift));
std::unordered_map<int, dnnl::memory> reorder_args;
reorder_args.insert({DNNL_ARG_SRC, *reorder_src_memory_p});
reorder_args.insert({DNNL_ARG_DST, *reorder_dst_memory_p});
reorder_args.insert({DNNL_ARG_ATTR_SCALES | DNNL_ARG_DST, scales_mem});
if (with_shift) {
reorder_args.insert(
{DNNL_ARG_ATTR_ZERO_POINTS | DNNL_ARG_SRC, zero_points_mem});
}
reorder_p->execute(astream, reorder_args);
astream.wait();
out->set_mem_desc(reorder_dst_memory_p->get_desc());
}
} // namespace phi
PD_REGISTER_KERNEL(dequantize,
OneDNN,
ONEDNN,
phi::DeQuantKernel,
uint8_t,
int8_t,
phi::bfloat16) {
kernel->OutputAt(0).SetDataType(phi::DataType::FLOAT32);
}