* Copyright (c) 2025 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 "onnx_common.h"
#include "op_cv_proto_extend.h"
using namespace std;
using namespace ge;
using ge::Operator;
namespace {
const int INPUT_NUM = 2;
const int OUTPUT_NUM = 1;
const int PADS_SIZE = 8;
}
namespace domi {
using NodeProto = ge::onnx::NodeProto;
static Status ParseParamsUpsample(const Message* op_src, ge::Operator& op_dest)
{
const NodeProto* node = dynamic_cast<const NodeProto*>(op_src);
if (node == nullptr) {
return FAILED;
}
op_dest.SetAttr("original_type", ge::AscendString("ai.onnx::11::Upsample"));
std::string mode_value = "nearest";
for (const auto& attr : node->attribute()) {
if (attr.name() == "mode" && attr.type() == ge::onnx::AttributeProto::STRING) {
mode_value = attr.s();
}
}
if (mode_value != "nearest" && mode_value != "linear") {
OP_LOGE(
GetOpName(op_dest).c_str(), "Mode attr of Upsample only supports nearest and linear, current is %s .",
mode_value.c_str());
return FAILED;
}
op_dest.SetAttr("name", node->name());
op_dest.SetAttr("mode", mode_value);
op_dest.DynamicInputRegister("x", INPUT_NUM);
op_dest.DynamicOutputRegister("y", OUTPUT_NUM);
return SUCCESS;
}
static Status ParseOpToGraphUpsample(const ge::Operator& op, Graph& graph)
{
std::string ori_name;
if (op.GetAttr("name", ori_name) != SUCCESS) {
OP_LOGE(GetOpName(op).c_str(), "get name from op failed.");
return FAILED;
}
auto data0 = op::Data((ori_name + "_data0").c_str()).set_attr_index(0);
auto data1 = op::Data((ori_name + "_data1").c_str()).set_attr_index(1);
std::vector<float> pads_vector(PADS_SIZE, 0);
int64_t len = pads_vector.size();
const vector<int64_t> dims = {len};
ge::Tensor const_value = Vec2Tensor(pads_vector, dims, DT_FLOAT, ge::FORMAT_ND);
auto const_op = op::Const((ori_name + "_const_data").c_str()).set_attr_value(const_value);
auto ret_x = ChangeFormatFromOnnx(data0, 0, ge::FORMAT_NCHW, false);
if (ret_x != ge::GRAPH_SUCCESS) {
OP_LOGE(GetOpName(op).c_str(), "update upsample_x format failed.");
return FAILED;
}
std::string mode_value1;
if (op.GetAttr("mode", mode_value1) != SUCCESS) {
OP_LOGE(GetOpName(op).c_str(), "get value of mode from op failed.");
return FAILED;
}
auto resize_op = op::Resize((ori_name + "_resize").c_str())
.set_input_x(data0)
.set_input_roi(const_op)
.set_input_scales(data1)
.set_attr_mode(mode_value1);
ChangeFormatFromOnnx(resize_op, 0, ge::FORMAT_NCHW, true);
ChangeFormatFromOnnx(resize_op, 0, ge::FORMAT_NCHW, false);
std::vector<ge::Operator> inputs = {data0, const_op, data1};
std::vector<std::pair<ge::Operator, std::vector<size_t>>> output_indexs;
output_indexs.emplace_back(resize_op, std::vector<size_t>{0});
graph.SetInputs(inputs).SetOutputs(output_indexs);
return SUCCESS;
}
REGISTER_CUSTOM_OP("PartitionedCall")
.FrameworkType(ONNX)
.OriginOpType({ge::AscendString("ai.onnx::8::Upsample"),
ge::AscendString("ai.onnx::9::Upsample"),
ge::AscendString("ai.onnx::11::Upsample"),
ge::AscendString("ai.onnx::12::Upsample"),
ge::AscendString("ai.onnx::13::Upsample"),
ge::AscendString("ai.onnx::14::Upsample"),
ge::AscendString("ai.onnx::15::Upsample"),
ge::AscendString("ai.onnx::16::Upsample")})
.ParseParamsFn(ParseParamsUpsample)
.ParseOpToGraphFn(ParseOpToGraphUpsample)
.ImplyType(ImplyType::TVM);
}