wehub-resource-sync
81 lines
3.4 KiB
C++
81 lines
3.4 KiB
C++
//
|
|
// BroadCastAdd.cpp
|
|
// MNNConverter
|
|
//
|
|
// Created by MNN on 2019/01/31.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include <stdio.h>
|
|
#include "liteOpConverter.hpp"
|
|
|
|
DECLARE_OP_COVERTER(AddTflite);
|
|
|
|
MNN::OpType AddTflite::opType(int quantizedModel) {
|
|
if (quantizedModel == 1)
|
|
return MNN::OpType_QuantizedAdd;
|
|
return MNN::OpType_Extra;
|
|
}
|
|
|
|
MNN::OpParameter AddTflite::type(int quantizedModel) {
|
|
if (quantizedModel == 1)
|
|
return MNN::OpParameter_QuantizedAdd;
|
|
return MNN::OpParameter_Extra;
|
|
}
|
|
|
|
void AddTflite::run(MNN::OpT* dstOp, const std::unique_ptr<tflite::OperatorT>& tfliteOp,
|
|
const std::vector<std::unique_ptr<tflite::TensorT>>& tfliteTensors,
|
|
const std::vector<std::unique_ptr<tflite::BufferT>>& tfliteModelBuffer,
|
|
const std::vector<std::unique_ptr<tflite::OperatorCodeT>>& tfliteOpSet, int quantizedModel) {
|
|
const auto& addOption = tfliteOp->builtin_options.AsAddOptions();
|
|
if (quantizedModel == 1) {
|
|
auto AddParam = new MNN::QuantizedAddT;
|
|
|
|
DCHECK(tfliteOp->inputs.size() == 2) << "tflite Reshape input ERROR";
|
|
|
|
// input1
|
|
const int input1Index = tfliteOp->inputs[0];
|
|
const auto& input1Tensor = tfliteTensors[input1Index];
|
|
AddParam->input1QuantizedParam = std::unique_ptr<MNN::QuantizedParamT>(new MNN::QuantizedParamT);
|
|
AddParam->input1QuantizedParam->zeroPoint = input1Tensor->quantization->zero_point[0];
|
|
AddParam->input1QuantizedParam->scale = input1Tensor->quantization->scale[0];
|
|
|
|
// input1
|
|
const int input2Index = tfliteOp->inputs[1];
|
|
const auto& input2Tensor = tfliteTensors[input2Index];
|
|
AddParam->input2QuantizedParam = std::unique_ptr<MNN::QuantizedParamT>(new MNN::QuantizedParamT);
|
|
AddParam->input2QuantizedParam->zeroPoint = input2Tensor->quantization->zero_point[0];
|
|
AddParam->input2QuantizedParam->scale = input2Tensor->quantization->scale[0];
|
|
|
|
// output
|
|
const int outputIndex = tfliteOp->outputs[0];
|
|
const auto& outputTensor = tfliteTensors[outputIndex];
|
|
AddParam->outputQuantizedParam = std::unique_ptr<MNN::QuantizedParamT>(new MNN::QuantizedParamT);
|
|
AddParam->outputQuantizedParam->zeroPoint = outputTensor->quantization->zero_point[0];
|
|
AddParam->outputQuantizedParam->scale = outputTensor->quantization->scale[0];
|
|
|
|
AddParam->activationType = static_cast<MNN::FusedActivation>(addOption->fused_activation_function);
|
|
|
|
dstOp->main.value = AddParam;
|
|
} else {
|
|
auto extraOpParam = new MNN::ExtraT;
|
|
extraOpParam->engine = "Tflite";
|
|
extraOpParam->type = "BinaryActivation";
|
|
extraOpParam->attr.resize(2);
|
|
extraOpParam->attr[0].reset(new MNN::AttributeT);
|
|
extraOpParam->attr[1].reset(new MNN::AttributeT);
|
|
extraOpParam->attr[0]->key = "opType";
|
|
extraOpParam->attr[0]->i = tflite::BuiltinOperator_ADD;
|
|
extraOpParam->attr[1]->key = "activationType";
|
|
if (nullptr != addOption) {
|
|
extraOpParam->attr[1]->i = addOption->fused_activation_function;
|
|
} else {
|
|
extraOpParam->attr[1]->i = 0;
|
|
}
|
|
dstOp->main.value = extraOpParam;
|
|
}
|
|
}
|
|
|
|
using namespace tflite;
|
|
REGISTER_CONVERTER(AddTflite, BuiltinOperator_ADD);
|