1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
|
//
// Copyright © 2017,2019-2024 Arm Ltd and Contributors. All rights reserved.
// SPDX-License-Identifier: MIT
//
#include "GatherLayer.hpp"
#include "LayerCloneBase.hpp"
#include <armnn/TypesUtils.hpp>
#include <armnn/backends/WorkloadData.hpp>
#include <armnn/backends/WorkloadFactory.hpp>
namespace armnn
{
GatherLayer::GatherLayer(const GatherDescriptor& param, const char* name)
: LayerWithParameters(2, 1, LayerType::Gather, param, name)
{
}
std::unique_ptr<IWorkload> GatherLayer::CreateWorkload(const armnn::IWorkloadFactory& factory) const
{
GatherQueueDescriptor descriptor;
SetAdditionalInfo(descriptor);
return factory.CreateWorkload(LayerType::Gather, descriptor, PrepInfoAndDesc(descriptor));
}
GatherLayer* GatherLayer::Clone(Graph& graph) const
{
return CloneBase<GatherLayer>(graph, m_Param, GetName());
}
std::vector<TensorShape> GatherLayer::InferOutputShapes(const std::vector<TensorShape>& inputShapes) const
{
if (inputShapes.size() != 2)
{
throw armnn::Exception("inputShapes' size is \"" + std::to_string(inputShapes.size()) +
"\" - should be \"2\".");
}
const TensorShape& params = inputShapes[0];
const TensorShape& indices = inputShapes[1];
if (indices.GetDimensionality() == Dimensionality::Scalar && indices.GetNumDimensions() == 1)
{
return std::vector<TensorShape>({ TensorShape(Dimensionality::Scalar)});
}
const unsigned int paramsDim = params.GetNumDimensions();
const unsigned int indicesDim = indices.GetNumDimensions();
const unsigned int outputDim = paramsDim - 1 + indicesDim;
std::vector<unsigned int> dimSizes;
unsigned int axis = static_cast<unsigned int>(m_Param.m_Axis);
if (m_Param.m_Axis < 0)
{
int32_t axis_aux = static_cast<int32_t>(paramsDim) + m_Param.m_Axis;
axis = static_cast<unsigned int> (axis_aux);
}
for (unsigned int i = 0; i < axis; ++i)
{
dimSizes.push_back(params[i]);
}
for (unsigned int i = axis; i < indicesDim + axis; ++i)
{
dimSizes.push_back(indices[i - axis]);
}
for (unsigned int i = 1 + axis; i < paramsDim; ++i)
{
dimSizes.push_back(params[i]);
}
return std::vector<TensorShape>({ TensorShape({outputDim, dimSizes.data()})});
}
void GatherLayer::ValidateTensorShapesFromInputs()
{
VerifyLayerConnections(2, CHECK_LOCATION());
const TensorShape& outputShape = GetOutputSlot(0).GetTensorInfo().GetShape();
VerifyShapeInferenceType(outputShape, m_ShapeInferenceMethod);
std::vector<TensorShape> inferredShapes = InferOutputShapes(
{GetInputSlot(0).GetTensorInfo().GetShape(),
GetInputSlot(1).GetTensorInfo().GetShape()});
if (inferredShapes.size() != 1)
{
throw armnn::LayerValidationException("inferredShapes has "
+ std::to_string(inferredShapes.size()) +
" elements - should only have 1.");
}
if (inferredShapes[0].GetDimensionality() != Dimensionality::Specified &&
inferredShapes[0].GetDimensionality() != Dimensionality::Scalar)
{
throw armnn::LayerValidationException("inferredShapes' dimensionality is neither specified nor scalar.");
}
ValidateAndCopyShape(outputShape, inferredShapes[0], m_ShapeInferenceMethod, "GatherLayer");
}
void GatherLayer::ExecuteStrategy(IStrategy& strategy) const
{
strategy.ExecuteStrategy(this, GetParameters(), {}, GetName());
}
} // namespace armnn
|
