mirror of https://github.com/alibaba/MNN.git
111 lines
4.0 KiB
C++
111 lines
4.0 KiB
C++
//
|
|
// ReduceGradTest.cpp
|
|
// MNNTests
|
|
//
|
|
// Created by MNN on 2022/07/12.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include <MNN/expr/Expr.hpp>
|
|
#include <MNN/expr/ExprCreator.hpp>
|
|
#include "MNNTestSuite.h"
|
|
#include "TestUtils.h"
|
|
#include "../tools/train/source/grad/OpGrad.hpp"
|
|
|
|
using namespace MNN;
|
|
using namespace MNN::Express;
|
|
|
|
class ReduceGradTest : public MNNTestCase {
|
|
public:
|
|
char name[20] = "Reduce";
|
|
virtual ~ReduceGradTest() = default;
|
|
|
|
bool checkResult(VARP output, VARP outputDiff, std::vector<float> expectedOutput, const char* subname) {
|
|
const int len = expectedOutput.size();
|
|
auto opExpr = output->expr().first;
|
|
auto grad = OpGrad::get(opExpr->get()->type());
|
|
if (grad == nullptr) {
|
|
MNN_ERROR("no grad defined for: %s %s\n", name, subname);
|
|
}
|
|
auto inputGrad = grad->onGrad(opExpr, {outputDiff});
|
|
auto gotOutput = inputGrad[0]->readMap<float>();
|
|
|
|
for (int i = 0; i < len; ++i) {
|
|
auto diff = ::fabsf(gotOutput[i] - expectedOutput[i]);
|
|
if (diff > 0.001) {
|
|
MNN_ERROR("%s %s grad test failed, expected: %f, but got: %f!\n", name, subname, expectedOutput[i], gotOutput[i]);
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
virtual bool run(int precision) {
|
|
const int len = 5;
|
|
auto input = _Input({len}, NCHW);
|
|
const float inpudata[] = {-1.0, -2.0, 0.0, 4.0, -5.0};
|
|
auto inputPtr = input->writeMap<float>();
|
|
memcpy(inputPtr, inpudata, len * sizeof(float));
|
|
std::vector<float> outputDiffVec = {0.1};
|
|
|
|
{
|
|
auto output = _ReduceSum(input);
|
|
auto ptr = output->readMap<float>();
|
|
const std::vector<float> expectedOutput = {0.1, 0.1, 0.1, 0.1, 0.1};
|
|
auto outputDiff = _Const(outputDiffVec.data(), {});
|
|
if (!checkResult(output, outputDiff, expectedOutput, "ReduceSum")) {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
{
|
|
auto output = _ReduceMean(input);
|
|
const std::vector<float> expectedOutput = {0.0200, 0.0200, 0.0200, 0.0200, 0.0200};
|
|
auto outputDiff = _Const(outputDiffVec.data(), {});
|
|
if (!checkResult(output, outputDiff, expectedOutput, "ReduceMean")) {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
{
|
|
const float inpudata[] = {-1.0, -2.0, 0.0, 4.0, 4.0};
|
|
auto inputPtr = input->writeMap<float>();
|
|
memcpy(inputPtr, inpudata, len * sizeof(float));
|
|
auto output = _ReduceMax(input);
|
|
const std::vector<float> expectedOutput = {0.0, 0.0, 0.0, 0.05, 0.05};
|
|
auto outputDiff = _Const(outputDiffVec.data(), {});
|
|
if (!checkResult(output, outputDiff, expectedOutput, "ReduceMax")) {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
{
|
|
const float inpudata[] = {-2.0, -2.0, 0.0, 4.0, 4.0};
|
|
auto inputPtr = input->writeMap<float>();
|
|
memcpy(inputPtr, inpudata, len * sizeof(float));
|
|
auto output = _ReduceMin(input);
|
|
const std::vector<float> expectedOutput = {0.05, 0.05, 0.0, 0.0, 0.0};
|
|
auto outputDiff = _Const(outputDiffVec.data(), {});
|
|
if (!checkResult(output, outputDiff, expectedOutput, "ReduceMin")) {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
{
|
|
const float inpudata[] = {-1.0, -2.0, 1.0, 4.0, -5.0};
|
|
auto inputPtr = input->writeMap<float>();
|
|
memcpy(inputPtr, inpudata, len * sizeof(float));
|
|
auto output = _ReduceProd(input);
|
|
const std::vector<float> expectedOutput = {4.0000, 2.0000, -4.0000, -1.0000, 0.8000};
|
|
auto outputDiff = _Const(outputDiffVec.data(), {});
|
|
if (!checkResult(output, outputDiff, expectedOutput, "ReduceProd")) {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
return true;
|
|
}
|
|
};
|
|
|
|
MNNTestSuiteRegister(ReduceGradTest, "grad/reduce");
|