This repository has been archived by the owner on Feb 22, 2020. It is now read-only.
/
base.py
144 lines (108 loc) · 4.6 KB
/
base.py
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
# Tencent is pleased to support the open source community by making GNES available.
#
# Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# pylint: disable=low-comment-ratio
from typing import List, Any, Union, Dict, Callable
import numpy as np
from ..base import TrainableBase
class BaseEncoder(TrainableBase):
def encode(self, data: Any, *args, **kwargs) -> Any:
pass
def _copy_from(self, x: 'BaseEncoder') -> None:
pass
class BaseImageEncoder(BaseEncoder):
def encode(self, img: List['np.ndarray'], *args, **kwargs) -> np.ndarray:
pass
class BaseVideoEncoder(BaseEncoder):
def encode(self, img: List['np.ndarray'], *args, **kwargs) -> np.ndarray:
pass
class BaseTextEncoder(BaseEncoder):
def encode(self, text: List[str], *args, **kwargs) -> np.ndarray:
pass
class BaseNumericEncoder(BaseEncoder):
def encode(self, text: np.ndarray, *args, **kwargs) -> np.ndarray:
pass
class BaseBinaryEncoder(BaseEncoder):
def encode(self, data: np.ndarray, *args, **kwargs) -> bytes:
if data.dtype != np.uint8:
raise ValueError('data must be np.uint8 but received %s' % data.dtype)
return data.tobytes()
class CompositionalEncoder(BaseEncoder):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self._component = None # type: List['BaseEncoder']
@property
def component(self) -> Union[List['BaseEncoder'], Dict[str, 'BaseEncoder']]:
return self._component
@property
def is_pipeline(self):
return isinstance(self.component, list)
@component.setter
def component(self, comps: Callable[[], Union[list, dict]]):
if not callable(comps):
raise TypeError('component must be a callable function that returns '
'a List[BaseEncoder]')
if not getattr(self, 'init_from_yaml', False):
self._component = comps()
else:
self.logger.info('component is omitted from construction, '
'as it is initialized from yaml config')
def close(self):
super().close()
# pipeline
if isinstance(self.component, list):
for be in self.component:
be.close()
# no typology
elif isinstance(self.component, dict):
for be in self.component.values():
be.close()
elif self.component is None:
pass
else:
raise TypeError('component must be dict or list, received %s' % type(self.component))
def _copy_from(self, x: 'CompositionalEncoder'):
if isinstance(self.component, list):
for be1, be2 in zip(self.component, x.component):
be1._copy_from(be2)
elif isinstance(self.component, dict):
for k, v in self.component.items():
v._copy_from(x.component[k])
else:
raise TypeError('component must be dict or list, received %s' % type(self.component))
@classmethod
def to_yaml(cls, representer, data):
tmp = super()._dump_instance_to_yaml(data)
tmp['component'] = data.component
return representer.represent_mapping('!' + cls.__name__, tmp)
@classmethod
def from_yaml(cls, constructor, node):
obj, data, from_dump = super()._get_instance_from_yaml(constructor, node)
if not from_dump and 'component' in data:
obj.component = lambda: data['component']
return obj
class PipelineEncoder(CompositionalEncoder):
def encode(self, data: Any, *args, **kwargs) -> Any:
if not self.component:
raise NotImplementedError
for be in self.component:
data = be.encode(data, *args, **kwargs)
return data
def train(self, data, *args, **kwargs):
if not self.component:
raise NotImplementedError
for idx, be in enumerate(self.component):
be.train(data, *args, **kwargs)
if idx + 1 < len(self.component):
data = be.encode(data, *args, **kwargs)