mirror of https://github.com/alibaba/MNN.git
50 lines
1.5 KiB
C++
50 lines
1.5 KiB
C++
//
|
|
// ShapeReduceJoin.cpp
|
|
// MNN
|
|
//
|
|
// Created by MNN on 2019/01/10.
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
|
//
|
|
|
|
#include "shape/SizeComputer.hpp"
|
|
#include "core/Macro.h"
|
|
|
|
namespace MNN {
|
|
class ReduceJoinComputer : public SizeComputer {
|
|
public:
|
|
virtual bool onComputeSize(const MNN::Op* op, const std::vector<Tensor*>& inputs,
|
|
const std::vector<Tensor*>& outputs) const override {
|
|
MNN_ASSERT(2 == inputs.size());
|
|
MNN_ASSERT(1 == outputs.size());
|
|
|
|
auto output = outputs[0];
|
|
auto input = inputs[0];
|
|
auto axis = inputs[1];
|
|
|
|
// support reduce 1 dimension, only
|
|
MNN_ASSERT(axis->size() == axis->buffer().type.bytes());
|
|
|
|
MNN_ASSERT(axis->host<int32_t>()[0] >= 0);
|
|
std::vector<int> shape;
|
|
for (int i = 0; i < input->buffer().dimensions; i++) {
|
|
if (i != axis->host<int32_t>()[0]) {
|
|
shape.push_back(input->buffer().dim[i].extent);
|
|
} else {
|
|
if (op->main_as_ReduceJoin()->keepDims()) {
|
|
shape.push_back(1);
|
|
}
|
|
}
|
|
}
|
|
output->buffer().dimensions = (int)shape.size();
|
|
for (int i = 0; i < shape.size(); i++) {
|
|
output->buffer().dim[i].extent = shape[i];
|
|
}
|
|
output->setType(DataType_DT_STRING);
|
|
TensorUtils::getDescribe(outputs[0])->dimensionFormat = MNN_DATA_FORMAT_NHWC;
|
|
return true;
|
|
}
|
|
};
|
|
|
|
REGISTER_SHAPE_INPUTS(ReduceJoinComputer, OpType_ReduceJoin, {1});
|
|
} // namespace MNN
|