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
|
// Copyright 2022 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#ifndef COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_EXECUTION_PROCESSING_CUSTOM_INPUT_PROCESSOR_H_
#define COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_EXECUTION_PROCESSING_CUSTOM_INPUT_PROCESSOR_H_
#include <vector>
#include "base/containers/flat_map.h"
#include "base/time/time.h"
#include "components/segmentation_platform/internal/execution/processing/query_processor.h"
#include "components/segmentation_platform/public/proto/model_metadata.pb.h"
namespace segmentation_platform::processing {
class FeatureProcessorState;
class InputDelegateHolder;
// CustomInputProcessor adds support to a larger variety of data type
// (timestamps, strings mapped to enums, etc), transforming them into valid
// input tensor to use when executing the ML model.
class CustomInputProcessor : public QueryProcessor {
public:
CustomInputProcessor(const base::Time prediction_time,
InputDelegateHolder* input_delegate_holder);
CustomInputProcessor(base::flat_map<FeatureIndex, Data>&& data,
const base::Time prediction_time,
InputDelegateHolder* input_delegate_holder);
~CustomInputProcessor() override;
using FeatureListQueryProcessorCallback =
base::OnceCallback<void(std::unique_ptr<FeatureProcessorState>)>;
// QueryProcessor implementation.
void Process(std::unique_ptr<FeatureProcessorState> feature_processor_state,
QueryProcessorCallback callback) override;
template <typename IndexType>
using TemplateCallback =
base::OnceCallback<void(std::unique_ptr<FeatureProcessorState>,
base::flat_map<IndexType, Tensor>)>;
// Process a data mapping with a customized index type and return the tensor
// values in |callback|. Appends the input to the provided `result` and
// returns it.
template <typename IndexType>
void ProcessIndexType(
base::flat_map<IndexType, proto::CustomInput> custom_inputs,
std::unique_ptr<FeatureProcessorState> feature_processor_state,
std::unique_ptr<base::flat_map<IndexType, Tensor>> result,
TemplateCallback<IndexType> callback);
private:
// Helper method to handle async custom inputs for `ProcessIndexType()`
template <typename IndexType>
void OnGotProcessedValue(
base::flat_map<IndexType, proto::CustomInput> custom_inputs,
std::unique_ptr<FeatureProcessorState> feature_processor_state,
std::unique_ptr<base::flat_map<IndexType, Tensor>> result,
TemplateCallback<IndexType> callback,
IndexType current_index,
size_t current_tensor_length,
bool error,
Tensor current_value);
// Helper function for parsing a single sync custom input and insert the
// result along with the corresponding feature index.
QueryProcessor::Tensor ProcessSingleCustomInput(
const proto::CustomInput& custom_input,
FeatureProcessorState* feature_processor_state);
// Add a tensor value for CustomInput::FILL_PREDICTION_TIME type and return
// whether it succeeded.
bool AddPredictionTime(const proto::CustomInput& custom_input,
std::vector<ProcessedValue>& out_tensor);
// Add a tensor value for CustomInput::FILL_DEVICE_RAM type and return
// whether it succeeded.
bool AddDeviceRAMInMB(const proto::CustomInput& custom_input,
std::vector<ProcessedValue>& out_tensor);
// Add a tensor value for CustomInput::FILL_DEVICE_OS type and return
// whether it succeeded.
bool AddDeviceOSVersionNumber(const proto::CustomInput& custom_input,
std::vector<ProcessedValue>& out_tensor);
// Add a tensor value for CustomInput::FILL_DEVICE_PPI type and return
// whether it succeeded.
bool AddDevicePPI(const proto::CustomInput& custom_input,
std::vector<ProcessedValue>& out_tensor);
// Add a tensor value for CustomInput::TIME_RANGE_BEFORE_PREDICTION type and
// return whether it succeeded.
bool AddTimeRangeBeforePrediction(const proto::CustomInput& custom_input,
std::vector<ProcessedValue>& out_tensor);
// Add a tensor value for CustomInput::FILL_FROM_INPUT_CONTEXT and return
// whether it succeeded.
bool AddFromInputContext(const proto::CustomInput& custom_input,
FeatureProcessorState* feature_processor_state,
std::vector<ProcessedValue>& out_tensor);
// Add a random number for CustomInput::FILL_RANDOM and return whether it
// succeeded.
bool AddRandom(const proto::CustomInput& custom_input,
std::vector<ProcessedValue>& out_tensor);
const raw_ptr<InputDelegateHolder, AcrossTasksDanglingUntriaged>
input_delegate_holder_;
// List of custom inputs to process into input tensors.
base::flat_map<FeatureIndex, proto::CustomInput> custom_inputs_;
// Time at which we expect the model execution to run.
base::Time prediction_time_;
base::WeakPtrFactory<CustomInputProcessor> weak_ptr_factory_{this};
};
} // namespace segmentation_platform::processing
#endif // COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_EXECUTION_PROCESSING_CUSTOM_INPUT_PROCESSOR_H_
|
