wehub-resource-sync
67 lines
2.6 KiB
C++
67 lines
2.6 KiB
C++
//
|
|
// QuantizeTflite.cpp
|
|
// MNNConverter
|
|
//
|
|
// Created by MNN on 2023/04/07.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include <stdio.h>
|
|
#include "liteOpConverter.hpp"
|
|
#include "TfliteUtils.hpp"
|
|
|
|
DECLARE_OP_COVERTER(QuantizeTflite);
|
|
MNN::OpType QuantizeTflite::opType(int quantizedModel) {
|
|
return MNN::OpType_Cast;
|
|
}
|
|
MNN::OpParameter QuantizeTflite::type(int quantizedModel) {
|
|
return MNN::OpParameter_CastParam;
|
|
}
|
|
|
|
void QuantizeTflite::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){
|
|
auto inputIndex = tfliteOp->inputs[0];
|
|
auto outputIndex = tfliteOp->outputs[0];
|
|
const auto& inputTensor = tfliteTensors[inputIndex];
|
|
const auto& outputTensor = tfliteTensors[outputIndex];
|
|
if (inputTensor->type == tflite::TensorType_INT8 || inputTensor->type == tflite::TensorType_FLOAT16 || inputTensor->type == tflite::TensorType_FLOAT32) {
|
|
if (outputTensor->type == tflite::TensorType_INT8 || outputTensor->type == tflite::TensorType_FLOAT16 || outputTensor->type == tflite::TensorType_FLOAT32) {
|
|
dstOp->type = MNN::OpType_Identity;
|
|
dstOp->main.type = MNN::OpParameter_NONE;
|
|
return;
|
|
}
|
|
}
|
|
#if 0
|
|
auto extraOpParam = new MNN::ExtraT;
|
|
extraOpParam->engine = "Tflite";
|
|
extraOpParam->type = "Quantize";
|
|
extraOpParam->attr.resize(2);
|
|
extraOpParam->attr[0].reset(new MNN::AttributeT);
|
|
extraOpParam->attr[1].reset(new MNN::AttributeT);
|
|
extraOpParam->attr[0]->key = "scale";
|
|
extraOpParam->attr[0]->f = outputTensor->quantization->scale[0];
|
|
extraOpParam->attr[1]->key = "zero_point";
|
|
extraOpParam->attr[1]->i = (int)outputTensor->quantization->zero_point[0];
|
|
dstOp->main.value = extraOpParam;
|
|
#else
|
|
// just add cast, quant/dequant is runtime insert
|
|
auto param = new MNN::CastParamT;
|
|
// other type to int8 => cast + quant
|
|
if (outputTensor->type == tflite::TensorType_INT8) {
|
|
param->srcT = TfliteDataTypeToMNN(inputTensor->type);
|
|
param->dstT = MNN::DataType_DT_FLOAT;
|
|
}
|
|
// int8 to other type => dequant + cast
|
|
if (inputTensor->type == tflite::TensorType_INT8) {
|
|
param->srcT = MNN::DataType_DT_FLOAT;
|
|
param->dstT = TfliteDataTypeToMNN(outputTensor->type);
|
|
}
|
|
dstOp->main.value = param;
|
|
#endif
|
|
}
|
|
|
|
using namespace tflite;
|
|
REGISTER_CONVERTER(QuantizeTflite, BuiltinOperator_QUANTIZE);
|