wehub-resource-sync
85 lines
3.4 KiB
C++
85 lines
3.4 KiB
C++
//
|
|
// PackTflite.cpp
|
|
// MNNConverter
|
|
//
|
|
// Created by MNN on 2019/11/25.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include <stdio.h>
|
|
#include "liteOpConverter.hpp"
|
|
|
|
DECLARE_OP_COVERTER(PackTflite);
|
|
MNN::OpType PackTflite::opType(int quantizedModel) {
|
|
return MNN::OpType_Pack;
|
|
}
|
|
MNN::OpParameter PackTflite::type(int quantizedModel) {
|
|
return MNN::OpParameter_PackParam;
|
|
}
|
|
|
|
void PackTflite::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 param = new MNN::PackParamT;
|
|
auto opt=tfliteOp->builtin_options.AsPackOptions();
|
|
param->axis=opt->axis;
|
|
dstOp->main.value = param;
|
|
}
|
|
|
|
DECLARE_OP_COVERTER(SplitTflite);
|
|
MNN::OpType SplitTflite::opType(int quantizedModel) {
|
|
return MNN::OpType_Slice;
|
|
}
|
|
MNN::OpParameter SplitTflite::type(int quantizedModel) {
|
|
return MNN::OpParameter_Slice;
|
|
}
|
|
|
|
void SplitTflite::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 param = new MNN::SliceT;
|
|
param->sourceType = MNN::NetSource_TENSORFLOW;
|
|
auto opt=tfliteOp->builtin_options.AsSplitOptions();
|
|
param->slicePoints.resize(1);
|
|
param->slicePoints[0] = opt->num_splits;
|
|
dstOp->main.value = param;
|
|
auto originInput = dstOp->inputIndexes[1];
|
|
auto axisInput = dstOp->inputIndexes[0];
|
|
const auto& axisTensor = tfliteTensors[axisInput];
|
|
auto axisBuffer = tfliteModelBuffer[axisTensor->buffer].get();
|
|
param->axis = ((int32_t*)(axisBuffer->data.data()))[0];
|
|
|
|
dstOp->inputIndexes.resize(1);
|
|
dstOp->inputIndexes[0] = originInput;
|
|
}
|
|
|
|
DECLARE_OP_COVERTER(StridedSliceTflite);
|
|
MNN::OpType StridedSliceTflite::opType(int quantizedModel) {
|
|
return MNN::OpType_StridedSlice;
|
|
}
|
|
MNN::OpParameter StridedSliceTflite::type(int quantizedModel) {
|
|
return MNN::OpParameter_StridedSliceParam;
|
|
}
|
|
|
|
void StridedSliceTflite::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 param = new MNN::StridedSliceParamT;
|
|
param->T = MNN::DataType_DT_FLOAT;
|
|
auto opt=tfliteOp->builtin_options.AsStridedSliceOptions();
|
|
param->beginMask = opt->begin_mask;
|
|
param->endMask = opt->end_mask;
|
|
param->ellipsisMask = opt->ellipsis_mask;
|
|
param->newAxisMask = opt->new_axis_mask;
|
|
param->shrinkAxisMask = opt->shrink_axis_mask;
|
|
dstOp->main.value = param;
|
|
}
|
|
|
|
using namespace tflite;
|
|
REGISTER_CONVERTER(PackTflite, BuiltinOperator_PACK);
|
|
REGISTER_CONVERTER(SplitTflite, BuiltinOperator_SPLIT);
|
|
REGISTER_CONVERTER(StridedSliceTflite, BuiltinOperator_STRIDED_SLICE);
|