* Copyright (c) 2026 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.
*/
* \file pad_v3_grad.cc
* \brief
*/
#include "log/log.h"
#include "register/op_impl_registry.h"
#include "util/shape_util.h"
#include "op_api/op_util.h"
using namespace ge;
namespace {
constexpr size_t INDEX_X = 0;
constexpr size_t INDEX_PADDINGS = 1;
constexpr size_t INDEX_Y = 0;
constexpr size_t INDEX_PADDINGS_CONTIGUOUS = 1;
constexpr size_t PAIR = 2;
static constexpr int64_t UNKNOWN_DIM_VALUE_ = -1L;
}
namespace ops {
template <typename T>
static ge::graphStatus PadV3GradInfershape(
const gert::InferShapeContext* context, const gert::Shape* x_shape, const gert::Tensor* paddings_tensor,
gert::Shape* y_shape)
{
const T* paddings_value = paddings_tensor->GetData<T>();
const size_t paddings_num = static_cast<size_t>(paddings_tensor->GetShapeSize());
auto attrs = context->GetAttrs();
OP_CHECK_NULL_WITH_CONTEXT(context, attrs);
const bool* paddings_contiguous = attrs->GetAttrPointer<bool>(INDEX_PADDINGS_CONTIGUOUS);
OP_CHECK_NULL_WITH_CONTEXT(context, paddings_contiguous);
OP_LOGD(context->GetNodeName(), "Begin to do PadV3GradInfershape");
OP_LOGD(context->GetNodeName(), "input x = %s", Ops::Base::ToString(*x_shape).c_str());
size_t input_dim_size = x_shape->GetDimNum();
OP_CHECK_IF(
input_dim_size == 0,
OP_LOGE_FOR_INVALID_SHAPE_WITH_REASON(
context->GetNodeName(), "x", Ops::Base::ToString(*x_shape).c_str(), "x cannot be an empty tensor"),
return ge::GRAPH_FAILED);
if (input_dim_size * PAIR != paddings_num) {
OP_LOGE_FOR_INVALID_SHAPESIZE_WITH_REASON(
context->GetNodeName(), "paddings", std::to_string(paddings_num).c_str(),
"The shape size of paddings should be equal to twice of the input x rank");
return ge::GRAPH_FAILED;
}
y_shape->SetDimNum(input_dim_size);
int64_t index_cof = 1;
int64_t index_offset = input_dim_size;
if (*paddings_contiguous) {
index_cof = PAIR;
index_offset = 1;
}
for (size_t i = 0; i < input_dim_size; ++i) {
auto pad_front = paddings_value[index_cof * i];
auto pad_end = paddings_value[index_cof * i + index_offset];
int64_t dim_value =
x_shape->GetDim(i) == UNKNOWN_DIM_VALUE_ ? UNKNOWN_DIM_VALUE_ : (x_shape->GetDim(i) - pad_front - pad_end);
if (x_shape->GetDim(i) != UNKNOWN_DIM_VALUE_ && dim_value < 0) {
OP_LOGE_FOR_INVALID_SHAPEDIM_WITH_REASON(
context->GetNodeName(), "y", std::to_string(dim_value).c_str(),
"The shape dim of y must be greater than or equal to 0");
return ge::GRAPH_FAILED;
}
y_shape->SetDim(i, dim_value);
}
OP_LOGD(context->GetNodeName(), "output y = %s", Ops::Base::ToString(*y_shape).c_str());
OP_LOGD(context->GetNodeName(), "End to do PadV3GradInfershape");
return ge::GRAPH_SUCCESS;
}
static ge::graphStatus SetAllUnknownDim(const int64_t rank, gert::Shape* output_shape)
{
output_shape->SetDimNum(rank);
for (int64_t i = 0; i < rank; ++i) {
output_shape->SetDim(i, UNKNOWN_DIM_VALUE_);
}
OP_LOGD("SetAllUnknownDim", "set all dim = -1, output = %s", Ops::Base::ToString(*output_shape).c_str());
return ge::GRAPH_SUCCESS;
}
static ge::graphStatus InferShape4PadV3Grad(gert::InferShapeContext* context)
{
const gert::Shape* x_shape = context->GetInputShape(INDEX_X);
OP_CHECK_NULL_WITH_CONTEXT(context, x_shape);
gert::Shape* y_shape = context->GetOutputShape(INDEX_Y);
OP_CHECK_NULL_WITH_CONTEXT(context, y_shape);
if (Ops::Base::IsUnknownRank(*x_shape)) {
Ops::Base::SetUnknownRank(*y_shape);
return GRAPH_SUCCESS;
}
const gert::Tensor* paddings_tensor = context->GetInputTensor(INDEX_PADDINGS);
OP_CHECK_NULL_WITH_CONTEXT(context, paddings_tensor);
if (!IsConstTensor(paddings_tensor)) {
return SetAllUnknownDim(x_shape->GetDimNum(), y_shape);
}
ge::DataType paddings_dtype = paddings_tensor->GetDataType();
switch (paddings_dtype) {
case ge::DT_INT32: {
return PadV3GradInfershape<int32_t>(context, x_shape, paddings_tensor, y_shape);
}
case ge::DT_INT64: {
return PadV3GradInfershape<int64_t>(context, x_shape, paddings_tensor, y_shape);
}
default:
OP_LOGE_FOR_INVALID_DTYPE(
context->GetNodeName(), "paddings", Ops::Base::ToString(paddings_dtype).c_str(), "int32 or int64");
return ge::GRAPH_FAILED;
}
return ge::GRAPH_FAILED;
}
IMPL_OP_INFERSHAPE(PadV3Grad).InferShape(InferShape4PadV3Grad).InputsDataDependency({INDEX_PADDINGS});
}