wehub-resource-sync
79 lines
3.1 KiB
C++
79 lines
3.1 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.
|
|
|
|
#pragma once
|
|
|
|
#include "paddle/phi/common/scalar.h"
|
|
#include "paddle/phi/core/dense_tensor.h"
|
|
#include "paddle/phi/core/device_context.h"
|
|
|
|
namespace phi {
|
|
|
|
template <typename T, typename Context>
|
|
void FlashAttnUnpaddedGradKernel(const Context& dev_ctx,
|
|
const DenseTensor& q,
|
|
const DenseTensor& k,
|
|
const DenseTensor& v,
|
|
const DenseTensor& cu_seqlens_q,
|
|
const DenseTensor& cu_seqlens_k,
|
|
const DenseTensor& out,
|
|
const DenseTensor& softmax_lse,
|
|
const DenseTensor& seed_offset,
|
|
const optional<DenseTensor>& attn_mask,
|
|
const DenseTensor& dout,
|
|
const Scalar& max_seqlen_q,
|
|
const Scalar& max_seqlen_k,
|
|
float scale,
|
|
float dropout,
|
|
bool causal,
|
|
DenseTensor* dq,
|
|
DenseTensor* dk,
|
|
DenseTensor* dv);
|
|
|
|
template <typename T, typename Context>
|
|
void FlashAttnGradKernel(const Context& dev_ctx,
|
|
const DenseTensor& q,
|
|
const DenseTensor& k,
|
|
const DenseTensor& v,
|
|
const DenseTensor& out,
|
|
const DenseTensor& softmax_lse,
|
|
const DenseTensor& seed_offset,
|
|
const optional<DenseTensor>& attn_mask,
|
|
const DenseTensor& dout,
|
|
float dropout,
|
|
bool causal,
|
|
DenseTensor* dq,
|
|
DenseTensor* dk,
|
|
DenseTensor* dv);
|
|
|
|
template <typename T, typename Context>
|
|
void FlashAttnWithSparseMaskGradKernel(
|
|
const Context& dev_ctx,
|
|
const DenseTensor& q,
|
|
const DenseTensor& k,
|
|
const DenseTensor& v,
|
|
const DenseTensor& attn_mask_start_row_indices,
|
|
const DenseTensor& out,
|
|
const DenseTensor& softmax_lse,
|
|
const DenseTensor& seed_offset,
|
|
const DenseTensor& dout,
|
|
float dropout,
|
|
bool causal,
|
|
int attn_mask_start_row,
|
|
DenseTensor* dq,
|
|
DenseTensor* dk,
|
|
DenseTensor* dv);
|
|
|
|
} // namespace phi
|