blob: 826b077f6e28577e4b06c0014fdee521af3e5e83 (
plain)
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
|
//
// Copyright © 2017 Arm Ltd. All rights reserved.
// SPDX-License-Identifier: MIT
//
#pragma once
#include <armnn/INetwork.hpp>
#include <armnn/Types.hpp>
#include <armnn/Tensor.hpp>
namespace armnn
{
struct QuantizerOptions
{
QuantizerOptions() : QuantizerOptions(DataType::QuantisedAsymm8, false) {}
QuantizerOptions(DataType activationFormat) : QuantizerOptions(activationFormat, false) {}
QuantizerOptions(DataType activationFormat, bool preserveType)
: m_ActivationFormat(activationFormat)
, m_PreserveType(preserveType) {}
DataType m_ActivationFormat;
bool m_PreserveType;
};
using INetworkQuantizerPtr = std::unique_ptr<class INetworkQuantizer, void(*)(INetworkQuantizer* quantizer)>;
/// Quantizer class Quantizes a float32 InputNetwork
class INetworkQuantizer
{
public:
/// Create Quantizer object and return raw pointer
static INetworkQuantizer* CreateRaw(INetwork* inputNetwork, const QuantizerOptions& options = QuantizerOptions());
/// Create Quantizer object wrapped in unique_ptr
static INetworkQuantizerPtr Create(INetwork* inputNetwork, const QuantizerOptions& options = QuantizerOptions());
/// Destroy Quantizer object
static void Destroy(INetworkQuantizer* quantizer);
/// Overrides the default quantization values for the input layer with the given id
virtual void OverrideInputRange(LayerBindingId layerId, float min, float max) = 0;
/// Refine input network with a set of refinement data for specified LayerBindingId
virtual void Refine(const InputTensors& inputTensors) = 0;
/// Extract final quantized network
virtual INetworkPtr ExportNetwork() = 0;
protected:
virtual ~INetworkQuantizer() {}
};
} //namespace armnn
|
