* Copyright (c) 2024 Huawei Technologies Co., Ltd.
* This program is free software, you can redistribute it and/or modify it under the terms and conditions of
* CANN Open Software License Agreement Version 2.0 (the "License").
* Please refer to the License for details. You may not use this file except in compliance with the License.
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
* See LICENSE in the root of the software repository for the full text of the License.
*/
#include <mki/base/kernel_base.h>
#include <mki_loader/op_register.h>
#include <mki/utils/log/log.h>
#include <mki/utils/const/op_const.h>
#include "tiling/unpad_tiling.h"
#include "tiling/tiling_data.h"
#include "mixkernels/utils/common.h"
static constexpr uint32_t TENSOR_INPUT_NUM = 4;
static constexpr uint32_t TENSOR_OUTPUT_NUM = 3;
namespace AtbOps {
using namespace Mki;
class UnpadKernel : public KernelBase {
public:
explicit UnpadKernel(const std::string &kernelName, const BinHandle *handle) noexcept
: KernelBase(kernelName, handle)
{
}
bool DimsCheck(const LaunchParam &launchParam) const
{
auto inTensor0 = launchParam.GetInTensor(0);
auto inTensor1 = launchParam.GetInTensor(DIM_1);
auto inTensor2 = launchParam.GetInTensor(DIM_2);
auto inTensor3 = launchParam.GetInTensor(DIM_3);
MKI_CHECK(inTensor0.desc.dims[0] > 0 && inTensor0.desc.dims[1] > 0,
"in tensor0 dims[0] or dims[1] is invalid", return false);
uint32_t batch = static_cast<uint32_t>(inTensor0.desc.dims[0]);
MKI_CHECK(inTensor1.desc.dims[0] == batch && inTensor1.desc.dims[1] == 1,
"in tensor1 dims[0] or dims[1] is invalid", return false);
MKI_CHECK(inTensor2.desc.dims[0] == 1 && inTensor2.desc.dims[1] == 1,
"in tensor2 dims[0] or dims[1] is invalid", return false);
MKI_CHECK(inTensor3.desc.dims[0] == batch && inTensor3.desc.dims[1] == 1,
"in tensor3 dims[0] or dims[1] is invalid", return false);
uint32_t maxSeqlen = static_cast<uint32_t>(inTensor0.desc.dims[1]);
MKI_CHECK(batch <= UINT32_MAX / maxSeqlen, "batch * maxSeqlen exceeds uint32_t maximum (overflow)", return false);
uint32_t totalLen = batch * maxSeqlen;
auto outTensor0 = launchParam.GetOutTensor(0);
auto outTensor1 = launchParam.GetOutTensor(1);
auto outTensor2 = launchParam.GetOutTensor(2);
MKI_CHECK(outTensor0.desc.dims[0] == 1 && outTensor0.desc.dims[1] == totalLen,
"out tensor0 dims[0] or dims[1] is invalid", return false);
MKI_CHECK(outTensor1.desc.dims[0] == batch && outTensor1.desc.dims[1] == 1,
"out tensor1 dims[0] or dims[1] is invalid", return false);
MKI_CHECK(outTensor2.desc.dims[0] == 1 && outTensor2.desc.dims[1] == totalLen,
"out tensor2 dims[0] or dims[1] is invalid", return false);
return true;
}
bool CanSupport(const LaunchParam &launchParam) const override
{
MKI_CHECK(launchParam.GetInTensorCount() == TENSOR_INPUT_NUM,
"in tensor num invalid", return false);
MKI_CHECK(launchParam.GetOutTensorCount() == TENSOR_OUTPUT_NUM,
"out tensor num invalid", return false);
auto inTensor0 = launchParam.GetInTensor(DIM_0);
MKI_CHECK(inTensor0.desc.dtype == TENSOR_DTYPE_INT64,
"in tensor 0 dtype invalid", return false);
auto inTensor1 = launchParam.GetInTensor(DIM_1);
MKI_CHECK(inTensor1.desc.dtype == TENSOR_DTYPE_INT32,
"in tensor 1 dtype invalid", return false);
auto inTensor2 = launchParam.GetInTensor(DIM_2);
MKI_CHECK(inTensor2.desc.dtype == TENSOR_DTYPE_INT64,
"in tensor 2 dtype invalid", return false);
auto inTensor3 = launchParam.GetInTensor(DIM_3);
MKI_CHECK(inTensor3.desc.dtype == TENSOR_DTYPE_INT32,
"in tensor 3 dtype invalid", return false);
for (size_t i = 0; i < TENSOR_INPUT_NUM; i++) {
auto inTensor = launchParam.GetInTensor(i);
MKI_CHECK(inTensor.desc.format == TENSOR_FORMAT_ND,
"in tensor " << i << " format invalid", return false);
MKI_CHECK(inTensor.desc.dims.size() == DIM_2,
"in tensor " << i << " dim num invalid", return false);
}
for (size_t i = 0; i < TENSOR_OUTPUT_NUM; i++) {
auto outTensor = launchParam.GetOutTensor(i);
MKI_CHECK(outTensor.desc.format == TENSOR_FORMAT_ND,
"out tensor " << i << " format invalid", return false);
MKI_CHECK(outTensor.desc.dims.size() == DIM_2,
"out tensor " << i << " dim num invalid", return false);
}
auto outTensor0 = launchParam.GetOutTensor(0);
MKI_CHECK(outTensor0.desc.dtype == TENSOR_DTYPE_INT64,
"outTensor0 dtype invalid", return false);
bool outputCheck = DimsCheck(launchParam);
return outputCheck;
}
uint64_t GetTilingSize(const LaunchParam &launchParam) const override
{
(void)launchParam;
return sizeof(UnpadTilingData);
}
Status InitImpl(const LaunchParam &launchParam) override
{
return UnpadTiling(launchParam, kernelInfo_);
}
};
REG_KERNEL_BASE(UnpadKernel);
}