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
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
|
//
// Copyright © 2020 Arm Ltd and Contributors. All rights reserved.
// SPDX-License-Identifier: MIT
//
#pragma once
#include "Optimization.hpp"
#include <armnnUtils/DataLayoutIndexed.hpp>
#include <ResolveType.hpp>
namespace armnn
{
namespace optimizations
{
template <typename ConvLayer, armnn::DataType ArmnnType,
typename T = armnn::ResolveType<ArmnnType>>
class FuseBatchNorm
{
public:
/// Run for every exclusive connection between any base Convolution layer and a child BatchNorm layer for not
/// quantized layers.
/// The child will be removed, the base will be removed if it's left unconnected. A new Convolution layer will
/// be added, its weights and bias will be calculated using the weights and bias of the base Convolution layer
/// combined with the parameters of the child BatchNorm layer.
void Run(Graph& graph, InputSlot& connection) const
{
Layer& base = connection.GetConnectedOutputSlot()->GetOwningLayer();
Layer& child = connection.GetOwningLayer();
bool depthwise = (base.GetType() == LayerType::DepthwiseConvolution2d);
ARMNN_ASSERT(base.GetType() == LayerType::Convolution2d || depthwise);
ARMNN_ASSERT(child.GetType() == LayerType::BatchNormalization);
if (base.GetDataType() == ArmnnType && child.GetDataType() == ArmnnType)
{
OutputSlot* parentOut = base.GetInputSlot(0).GetConnectedOutputSlot();
auto convLayer = PolymorphicDowncast<ConvLayer*>(&base);
auto batchNormLayer = PolymorphicDowncast<BatchNormalizationLayer*>(&child);
// Read convolution and batch norm parameters
BatchNormalizationDescriptor batchNormDescriptor = batchNormLayer->GetParameters();
auto epsilon = batchNormDescriptor.m_Eps;
IgnoreUnused(epsilon);
ConstTensor betaTensor(batchNormLayer->m_Beta->GetTensorInfo(), batchNormLayer->m_Beta->Map(true));
ConstTensor gammaTensor(batchNormLayer->m_Gamma->GetTensorInfo(), batchNormLayer->m_Gamma->Map(true));
ConstTensor meanTensor(batchNormLayer->m_Mean->GetTensorInfo(), batchNormLayer->m_Mean->Map(true));
ConstTensor varTensor(batchNormLayer->m_Variance->GetTensorInfo(), batchNormLayer->m_Variance->Map(true));
auto convDescriptor = convLayer->GetParameters();
ConstTensor weightsTensor;
if (convLayer->GetNumInputSlots() > 1)
{
ARMNN_ASSERT_MSG(convLayer->GetInputSlots()[1].GetConnection() != nullptr,
"FuseBatchNorm: Weight data should not be null.");
InputSlot & oldSlotWeights = const_cast<InputSlot&>(convLayer->GetInputSlots()[1]);
OutputSlot & constantSlotWeights = const_cast<OutputSlot&>(*oldSlotWeights.GetConnectedOutputSlot());
ConstantLayer* weightLayer = PolymorphicDowncast<ConstantLayer*>(
&constantSlotWeights.GetOwningLayer());
weightsTensor = ConstTensor(weightLayer->m_LayerOutput->GetTensorInfo(),
weightLayer->m_LayerOutput->Map(true));
}
else
{
ARMNN_ASSERT_MSG(convLayer->m_Weight != nullptr,
"FuseBatchNorm: Bias data should not be null if bias is enabled.");
weightsTensor = ConstTensor(convLayer->m_Weight->GetTensorInfo(), convLayer->m_Weight->Map(true));
}
armnnUtils::DataLayoutIndexed dataLayout(convDescriptor.m_DataLayout);
auto weightsShape = weightsTensor.GetInfo().GetShape();
const unsigned int inputChannels = parentOut->GetTensorInfo().GetShape()[dataLayout.GetChannelsIndex()];
const unsigned int depthMultiplier = depthwise ? weightsShape[3] / inputChannels : 1;
const unsigned int outputChannels = depthwise ? weightsShape[3] : weightsShape[0];
const unsigned int weightsHeight = depthwise ? weightsShape[1] :
weightsShape[dataLayout.GetHeightIndex()];
const unsigned int weightsWidth = depthwise ? weightsShape[2] :
weightsShape[dataLayout.GetWidthIndex()];
const auto* weightsBuffer = static_cast<const T*>(weightsTensor.GetMemoryArea());
const auto* betaBuffer = static_cast<const T*>(betaTensor.GetMemoryArea());
const auto* gammaBuffer = static_cast<const T*>(gammaTensor.GetMemoryArea());
const auto* meanBuffer = static_cast<const T*>(meanTensor.GetMemoryArea());
const auto* varBuffer = static_cast<const T*>(varTensor.GetMemoryArea());
std::vector<T> weightsVector (weightsBuffer, weightsBuffer + weightsTensor.GetNumElements());
std::vector<T> betaVector (betaBuffer, betaBuffer + betaTensor.GetNumElements());
std::vector<T> gammaVector (gammaBuffer, gammaBuffer + gammaTensor.GetNumElements());
std::vector<T> meanVector (meanBuffer, meanBuffer + meanTensor.GetNumElements());
std::vector<T> varianceVector(varBuffer, varBuffer + varTensor.GetNumElements());
// fusedWeights = ( gamma * weights ) / ( std - epsilon);
std::vector<T> fusedWeightsVector(weightsVector.size());
for (unsigned int cInput = 0; cInput < inputChannels; ++cInput)
{
for (unsigned int cOut = 0; cOut < outputChannels; ++cOut)
{
T mult = gammaVector[cOut] / static_cast<T>(sqrtf (varianceVector[cOut] + epsilon));
for (unsigned int h = 0; h < weightsHeight; ++h)
{
for (unsigned int w = 0; w < weightsWidth; ++w)
{
unsigned int weightsIdx = 0;
if (depthwise)
{
cInput = cOut / depthMultiplier;
weightsIdx = w * outputChannels + cOut +
h * weightsWidth * outputChannels;
}
else if (convDescriptor.m_DataLayout == DataLayout::NHWC)
{
weightsIdx = cOut * weightsHeight * weightsWidth * inputChannels +
h * weightsWidth * inputChannels +
w * inputChannels +
cInput;
}
else
{
weightsIdx = cOut * weightsWidth * weightsHeight * inputChannels +
cInput * weightsWidth * weightsHeight +
h * weightsWidth +
w;
}
fusedWeightsVector[weightsIdx] = mult * weightsVector[weightsIdx];
}
}
}
}
ConstTensor fusedWeightsTensor(weightsTensor.GetInfo(), fusedWeightsVector);
// fusedBias = (gamma * (bias - mean)) / (variance - epsilon) + beta;
std::vector<T> fusedBiasVector(outputChannels);
bool biasWasEnabledBeforeOpt = convDescriptor.m_BiasEnabled;
if (biasWasEnabledBeforeOpt)
{
ConstTensor biasTensor;
if (convLayer->GetNumInputSlots() > 1)
{
ARMNN_ASSERT_MSG(convLayer->GetInputSlots()[2].GetConnection() != nullptr,
"FuseBatchNorm: Bias data should not be null if bias is enabled.");
InputSlot & oldSlotBias = const_cast<InputSlot&>(convLayer->GetInputSlots()[2]);
OutputSlot & constantSlotBias = const_cast<OutputSlot&>(*oldSlotBias.GetConnectedOutputSlot());
ConstantLayer* biasLayer = PolymorphicDowncast<ConstantLayer*>(
&constantSlotBias.GetOwningLayer());
biasTensor = ConstTensor(biasLayer->m_LayerOutput->GetTensorInfo(),
biasLayer->m_LayerOutput->Map(true));
}
else
{
ARMNN_ASSERT_MSG(convLayer->m_Bias != nullptr,
"FuseBatchNorm: Bias data should not be null if bias is enabled.");
biasTensor = ConstTensor(convLayer->m_Bias->GetTensorInfo(), convLayer->m_Bias->Map(true));
}
const auto* biasBuffer = static_cast<const T*>(biasTensor.GetMemoryArea());
std::vector<T> biasVector(biasBuffer, biasBuffer + biasTensor.GetNumElements());
for (unsigned int cOut = 0; cOut < outputChannels; ++cOut)
{
fusedBiasVector[cOut] = ((gammaVector[cOut] * (biasVector[cOut] - meanVector[cOut])) /
sqrtf(varianceVector[cOut] + epsilon)) + betaVector[cOut];
}
}
else
{
convDescriptor.m_BiasEnabled = true;
std::vector<T> biasVector(outputChannels, T(0));
for (unsigned int cOut = 0; cOut < outputChannels; ++cOut)
{
fusedBiasVector[cOut] = ((gammaVector[cOut] * (biasVector[cOut] - meanVector[cOut])) /
sqrtf(varianceVector[cOut] + epsilon)) + betaVector[cOut];
}
}
ConstTensor fusedBiasTensor(TensorInfo({outputChannels}, ArmnnType, 0.0f, 0, true), fusedBiasVector);
// Insert the new convolution layer that has batch norm parameters fused into
const std::string name = std::string("fused-") + child.GetName() + std::string("-into-") + base.GetName();
auto& newConv2dLayer = *graph.InsertNewLayer<ConvLayer>(base.GetInputSlot(0),
convDescriptor,
name.c_str());
newConv2dLayer.m_Weight = std::make_unique<ScopedTensorHandle>(fusedWeightsTensor);
newConv2dLayer.m_Bias = std::make_unique<ScopedTensorHandle>(ConstTensor(fusedBiasTensor));
// Connect weights and bias from old to new Conv2d layer
// This optimization will always have 3 input slots on the Conv2d base layer
if (newConv2dLayer.GetNumInputSlots() > 1)
{
ConstantLayer* weightLayer = PolymorphicDowncast<ConstantLayer*>(
&base.GetInputSlot(1).GetConnectedOutputSlot()->GetOwningLayer());
// Remove old connection and connect to new layer2d
weightLayer->GetOutputSlot(0).Disconnect(base.GetInputSlot(1));
weightLayer->GetOutputSlot(0).Connect(newConv2dLayer.GetInputSlot(1));
weightLayer->m_LayerOutput = newConv2dLayer.m_Weight;
// Move bias const layers as normal if it was enabled before the optimisation
ConstantLayer* biasLayer;
if (biasWasEnabledBeforeOpt)
{
biasLayer = PolymorphicDowncast<ConstantLayer*>(
&base.GetInputSlot(2).GetConnectedOutputSlot()->GetOwningLayer());
// Remove old connection and connect to new layer2d
biasLayer->GetOutputSlot(0).Disconnect(base.GetInputSlot(2));
biasLayer->GetOutputSlot(0).Connect(newConv2dLayer.GetInputSlot(2));
}
// Otherwise create a new bias layer and add to the new convolution2d
else
{
// Add in bias constant layer
biasLayer = graph.AddLayer<ConstantLayer>("Bias");
biasLayer->GetOutputSlot(0).SetTensorInfo(fusedBiasTensor.GetInfo());
biasLayer->GetOutputSlot(0).Connect(newConv2dLayer.GetInputSlot(2));
}
biasLayer->m_LayerOutput = newConv2dLayer.m_Bias;
}
// Reconnects with original parent.
newConv2dLayer.GetOutputSlot().MoveAllConnections(*parentOut);
// Parent is now the new convolution2d layer.
parentOut = &newConv2dLayer.GetOutputSlot();
// Moves connections in child output to parent layer.
// Child layer will be removed as it's left unconnected.
// Base layer will be removed if left unconnected.
child.GetOutputSlot().MoveAllConnections(*parentOut);
}
}
protected:
FuseBatchNorm() = default;
~FuseBatchNorm() = default;
};
using FuseBatchNormIntoConvolution2DFloat32 =
OptimizeForExclusiveConnection<Convolution2dLayer,
BatchNormalizationLayer,
FuseBatchNorm<Convolution2dLayer, armnn::DataType::Float32>>;
using FuseBatchNormIntoConvolution2DFloat16 =
OptimizeForExclusiveConnection<Convolution2dLayer,
BatchNormalizationLayer,
FuseBatchNorm<Convolution2dLayer, armnn::DataType::Float16>>;
using FuseBatchNormIntoDepthwiseConvolution2DFloat32 =
OptimizeForExclusiveConnection<DepthwiseConvolution2dLayer,
BatchNormalizationLayer,
FuseBatchNorm<DepthwiseConvolution2dLayer, armnn::DataType::Float32>>;
using FuseBatchNormIntoDepthwiseConvolution2DFloat16 =
OptimizeForExclusiveConnection<DepthwiseConvolution2dLayer,
BatchNormalizationLayer,
FuseBatchNorm<DepthwiseConvolution2dLayer, armnn::DataType::Float16>>;
} // namespace optimizations
} // namespace armnn
|
