mirror of https://github.com/alibaba/MNN.git
80 lines
3.1 KiB
C++
80 lines
3.1 KiB
C++
|
//
|
||
|
|
// ShapeSlice.cpp
|
||
|
|
// MNN
|
||
|
|
//
|
||
|
|
// Created by MNN on 2019/01/10.
|
||
|
|
// Copyright © 2018, Alibaba Group Holding Limited
|
||
|
|
//
|
||
|
|
|
||
|
|
#include "Macro.h"
|
||
|
|
#include "SizeComputer.hpp"
|
||
|
|
|
||
|
|
namespace MNN {
|
||
|
|
|
||
|
|
class SliceComputer : public SizeComputer {
|
||
|
|
virtual bool onComputeSize(const MNN::Op* op, const std::vector<Tensor*>& inputs,
|
||
|
|
const std::vector<Tensor*>& outputs) const override {
|
||
|
|
MNN_ASSERT(1 == inputs.size());
|
||
|
|
MNN_ASSERT(2 <= outputs.size());
|
||
|
|
auto slice = op->main_as_Slice();
|
||
|
|
|
||
|
|
auto& input = inputs[0]->buffer();
|
||
|
|
|
||
|
|
int axis = slice->axis();
|
||
|
|
if (axis == -1) {
|
||
|
|
axis += input.dimensions;
|
||
|
|
}
|
||
|
|
|
||
|
|
if (MNN::NetSource_CAFFE == slice->sourceType()) {
|
||
|
|
// caffe Slice
|
||
|
|
int previous = 0;
|
||
|
|
for (int i = 0; i < slice->slicePoints()->size(); ++i) {
|
||
|
|
int sliceIndex = slice->slicePoints()->data()[i];
|
||
|
|
auto& output = outputs[i]->buffer();
|
||
|
|
output.dimensions = input.dimensions;
|
||
|
|
::memcpy(output.dim, input.dim, input.dimensions * sizeof(halide_dimension_t));
|
||
|
|
output.type = input.type;
|
||
|
|
output.dim[axis].extent = sliceIndex - previous;
|
||
|
|
previous = sliceIndex;
|
||
|
|
}
|
||
|
|
|
||
|
|
// Compute Last
|
||
|
|
auto& output = outputs[outputs.size() - 1]->buffer();
|
||
|
|
::memcpy(output.dim, input.dim, input.dimensions * sizeof(halide_dimension_t));
|
||
|
|
|
||
|
|
output.dim[axis].extent = input.dim[axis].extent - previous;
|
||
|
|
} else {
|
||
|
|
// tensorflow Split
|
||
|
|
if (1 == slice->slicePoints()->size()) {
|
||
|
|
// scalar
|
||
|
|
const int numSplits = slice->slicePoints()->data()[0];
|
||
|
|
MNN_ASSERT(numSplits == outputs.size());
|
||
|
|
MNN_ASSERT(0 == input.dim[axis].extent % numSplits);
|
||
|
|
const int splitDim = input.dim[axis].extent / numSplits;
|
||
|
|
for (int i = 0; i < numSplits; i++) {
|
||
|
|
auto& output = outputs[i]->buffer();
|
||
|
|
output.dimensions = input.dimensions;
|
||
|
|
output.type = input.type;
|
||
|
|
::memcpy(output.dim, input.dim, input.dimensions * sizeof(halide_dimension_t));
|
||
|
|
output.dim[axis].extent = splitDim;
|
||
|
|
}
|
||
|
|
} else {
|
||
|
|
// one dimension tensor, ex: [5,30]=>[5,4]+[5,15]+[5,11], slicePoints is [4, 15, 11]
|
||
|
|
MNN_ASSERT(slice->slicePoints()->size() == outputs.size());
|
||
|
|
for (int i = 0; i < slice->slicePoints()->size(); i++) {
|
||
|
|
auto& output = outputs[i]->buffer();
|
||
|
|
output.type = input.type;
|
||
|
|
output.dimensions = input.dimensions;
|
||
|
|
::memcpy(output.dim, input.dim, input.dimensions * sizeof(halide_dimension_t));
|
||
|
|
output.dim[axis].extent = slice->slicePoints()->data()[i];
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
return true;
|
||
|
|
}
|
||
|
|
};
|
||
|
|
|
||
|
|
REGISTER_SHAPE(SliceComputer, OpType_Slice);
|
||
|
|
} // namespace MNN
|