39 lines
1.1 KiB
C++
39 lines
1.1 KiB
C++
//
|
|
// Normalize.cpp
|
|
// MNNConverter
|
|
//
|
|
// Created by MNN on 2019/01/31.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include "OpConverter.hpp"
|
|
|
|
class Normalize : public OpConverter {
|
|
public:
|
|
virtual void run(MNN::OpT* dstOp, const caffe::LayerParameter& parameters, const caffe::LayerParameter& weight);
|
|
Normalize() {
|
|
}
|
|
virtual ~Normalize() {
|
|
}
|
|
virtual MNN::OpType opType() {
|
|
return MNN::OpType_Normalize;
|
|
}
|
|
virtual MNN::OpParameter type() {
|
|
return MNN::OpParameter_Normalize;
|
|
}
|
|
};
|
|
|
|
void Normalize::run(MNN::OpT* dstOp, const caffe::LayerParameter& parameters, const caffe::LayerParameter& weight) {
|
|
auto normizeT = new MNN::NormalizeT;
|
|
dstOp->main.value = normizeT;
|
|
auto& l = parameters.norm_param();
|
|
normizeT->channelShared = l.channel_shared();
|
|
normizeT->acrossSpatial = l.across_spatial();
|
|
normizeT->eps = l.eps();
|
|
auto& scaleBlob = weight.blobs(0);
|
|
for (int i = 0; i < scaleBlob.data_size(); ++i) {
|
|
normizeT->scale.push_back(scaleBlob.data(i));
|
|
}
|
|
}
|
|
static OpConverterRegister<Normalize> a("Normalize");
|