wehub-resource-sync
118 lines
4.3 KiB
Plaintext
118 lines
4.3 KiB
Plaintext
// 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 <algorithm>
|
|
|
|
#include "paddle/phi/backends/gpu/gpu_info.h"
|
|
#include "paddle/phi/backends/gpu/gpu_primitives.h"
|
|
#include "paddle/phi/core/kernel_registry.h"
|
|
#include "paddle/phi/kernels/funcs/blas/blas.h"
|
|
#include "paddle/phi/kernels/funcs/eigen/common.h"
|
|
#include "paddle/phi/kernels/funcs/rank_attention.cu.h"
|
|
|
|
namespace phi {
|
|
|
|
template <typename T, typename Context>
|
|
void RankAttentionGradOpCUDAKernel(const Context &dev_ctx,
|
|
const DenseTensor &x,
|
|
const DenseTensor &rank_offset,
|
|
const DenseTensor &rank_param,
|
|
const DenseTensor &input_help,
|
|
const DenseTensor &ins_rank,
|
|
const DenseTensor &out_grad,
|
|
int max_rank UNUSED,
|
|
int max_size,
|
|
DenseTensor *rank_param_grad) {
|
|
auto *dout = &out_grad;
|
|
auto *drank_para = rank_param_grad;
|
|
|
|
// get dim
|
|
auto x_dims = x.dims();
|
|
auto ins_num = x_dims[0];
|
|
auto x_fea_dim = x_dims[1];
|
|
auto para_dims = rank_param.dims();
|
|
auto para_row = para_dims[0];
|
|
auto para_col = para_dims[1];
|
|
auto rank_offset_dims = rank_offset.dims();
|
|
auto rank_offset_max_rank =
|
|
(rank_offset_dims[1] - 1) / 2; // Not use param max_rank
|
|
int64_t block_matrix_row =
|
|
static_cast<int64_t>(rank_offset_max_rank) * x_fea_dim;
|
|
auto &place = *dev_ctx.eigen_device();
|
|
|
|
int64_t max_ins = std::max(ins_num, static_cast<int64_t>(max_size));
|
|
// initialize out grad
|
|
dev_ctx.template Alloc<T>(drank_para);
|
|
auto drank_para_eigen = EigenVector<T>::Flatten(*drank_para);
|
|
drank_para_eigen.device(place) = drank_para_eigen.constant(static_cast<T>(0));
|
|
|
|
// copy data
|
|
DenseTensor param_grad;
|
|
param_grad.Resize({max_ins * block_matrix_row, para_col});
|
|
dev_ctx.template Alloc<T>(¶m_grad);
|
|
|
|
// initialize
|
|
auto param_grad_eigen = EigenVector<T>::Flatten(param_grad);
|
|
param_grad_eigen.device(place) = param_grad_eigen.constant(static_cast<T>(0));
|
|
// get data ptr
|
|
const T *input_help_data = input_help.data<T>();
|
|
const T *ins_rank_data = ins_rank.data<T>();
|
|
T *param_grad_data = param_grad.data<T>();
|
|
|
|
auto blas = funcs::GetBlas<GPUContext, T>(dev_ctx);
|
|
T alpha = 1;
|
|
T beta = 0;
|
|
|
|
// get param_grad
|
|
CBLAS_TRANSPOSE transA = CblasTrans;
|
|
CBLAS_TRANSPOSE transB = CblasNoTrans;
|
|
int64_t strideA = block_matrix_row;
|
|
int64_t strideB = para_col;
|
|
blas.BatchedGEMM(transA,
|
|
transB,
|
|
block_matrix_row,
|
|
para_col,
|
|
1,
|
|
alpha,
|
|
input_help_data,
|
|
dout->data<T>(),
|
|
beta,
|
|
param_grad_data,
|
|
ins_num,
|
|
strideA,
|
|
strideB);
|
|
// merge param_grad to get drank_para
|
|
int64_t expanded_grad_rows = ins_num * block_matrix_row;
|
|
|
|
merge_rank_attention_param_grad(dev_ctx.stream(),
|
|
param_grad_data,
|
|
expanded_grad_rows,
|
|
para_col,
|
|
drank_para->data<T>(),
|
|
para_row,
|
|
para_col,
|
|
ins_rank_data,
|
|
ins_num,
|
|
rank_offset_max_rank,
|
|
x_fea_dim);
|
|
}
|
|
} // namespace phi
|
|
|
|
PD_REGISTER_KERNEL(rank_attention_grad,
|
|
GPU,
|
|
ALL_LAYOUT,
|
|
phi::RankAttentionGradOpCUDAKernel,
|
|
float,
|
|
double) {}
|