wehub-resource-sync
107 lines
3.6 KiB
C++
107 lines
3.6 KiB
C++
//
|
|
// ReduceGrad.cpp
|
|
// MNN
|
|
//
|
|
// Created by MNN on 2019/05/24.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include "OpGrad.hpp"
|
|
using namespace std;
|
|
namespace MNN {
|
|
using namespace MNN::Express;
|
|
|
|
class ReduceGrad : public OpGrad {
|
|
public:
|
|
virtual std::vector<Express::VARP> onGrad(Express::EXPRP expr,
|
|
const std::vector<Express::VARP>& backwardOutput) override {
|
|
std::vector<Express::VARP> result;
|
|
auto inputs = expr->inputs();
|
|
result.resize(inputs.size());
|
|
std::unique_ptr<OpT> forwardOp(expr->get()->UnPack());
|
|
std::vector<int> reductionDims = forwardOp->main.AsReductionParam()->dim;
|
|
auto keepDim = forwardOp->main.AsReductionParam()->keepDims;
|
|
if (inputs.size() > 1) {
|
|
reductionDims.clear();
|
|
auto ptr = inputs[1]->readMap<int32_t>();
|
|
auto shape = inputs[1]->getInfo();
|
|
for (int i = 0; i < shape->size; ++i) {
|
|
reductionDims.emplace_back(ptr[i]);
|
|
}
|
|
}
|
|
if (reductionDims.empty()) {
|
|
auto shape = inputs[0]->getInfo();
|
|
for (int i = 0; i < shape->dim.size(); ++i) {
|
|
reductionDims.emplace_back(i);
|
|
}
|
|
}
|
|
|
|
VARP mask = _ZerosLike(inputs[0]) + _Scalar<float>(1.0f);
|
|
auto outputDiff = backwardOutput[0];
|
|
|
|
// implement other reduction op's grad below
|
|
if (forwardOp->main.AsReductionParam()->operation == ReductionType_SUM) {
|
|
// do not need to modify grads, just copy them, so, pass
|
|
}
|
|
|
|
if (forwardOp->main.AsReductionParam()->operation == ReductionType_MEAN) {
|
|
auto gradCount = _Size(outputDiff);
|
|
auto inputCount = _Size(inputs[0]);
|
|
outputDiff = _Multiply(outputDiff, _Cast<float>(gradCount) / _Cast<float>(inputCount));
|
|
}
|
|
|
|
if (forwardOp->main.AsReductionParam()->operation == ReductionType_MAXIMUM) {
|
|
auto output = Variable::create(expr);
|
|
if (!keepDim) {
|
|
output = _Unsqueeze(output, reductionDims);
|
|
}
|
|
mask = _Sign(inputs[0] - output) + _Scalar<float>(1.0f);
|
|
mask = mask / _ReduceSum(mask);
|
|
}
|
|
|
|
if (forwardOp->main.AsReductionParam()->operation == ReductionType_MINIMUM) {
|
|
auto output = Variable::create(expr);
|
|
if (!keepDim) {
|
|
output = _Unsqueeze(output, reductionDims);
|
|
}
|
|
mask = _Sign(output - inputs[0]) + _Scalar<float>(1.0f);
|
|
mask = mask / _ReduceSum(mask);
|
|
}
|
|
|
|
if (forwardOp->main.AsReductionParam()->operation == ReductionType_PROD) {
|
|
auto output = Variable::create(expr);
|
|
if (!keepDim) {
|
|
output = _Unsqueeze(output, reductionDims);
|
|
}
|
|
mask = output / inputs[0];
|
|
}
|
|
|
|
// this should be common operations, to expand grads to inputs shape
|
|
if (!keepDim) {
|
|
outputDiff = _Unsqueeze(outputDiff, reductionDims);
|
|
}
|
|
|
|
result[0] = mask * outputDiff;
|
|
return result;
|
|
}
|
|
};
|
|
class FillGrad : public OpGrad {
|
|
public:
|
|
virtual std::vector<Express::VARP> onGrad(Express::EXPRP expr,
|
|
const std::vector<Express::VARP>& backwardOutput) override {
|
|
return {backwardOutput[0].sum({})};
|
|
}
|
|
};
|
|
|
|
static void _create() {
|
|
static ReduceGrad _c;
|
|
OpGrad::insert(OpType_Reduction, &_c);
|
|
static FillGrad _d;
|
|
OpGrad::insert(OpType_Fill, &_d);
|
|
|
|
}
|
|
|
|
REGISTER_GRAD(ReduceGrad_cpp, _create);
|
|
};
|
|
|