-
Notifications
You must be signed in to change notification settings - Fork 2
/
base.py
193 lines (152 loc) · 5.53 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
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
from os.path import join, dirname
import tensorflow as tf
from gensim.models.word2vec import Word2Vec
from core.processing.lemmatization.mystem import MystemWrapper
from core.processing.pos.mystem_wrap import POSMystemWrapper
from core.source.embeddings.rusvectores import RusvectoresEmbedding
from networks.context.debug import DebugKeys
class LabelCalculationMode:
FIRST_APPEARED = u'take_first_appeared'
AVERAGE = u'average'
class CommonModelSettings(object):
GPUMemoryFraction = 0.25
# private settings
_test_on_epoch = range(0, 30000, 50)
_use_class_weights = True
_dropout = 0.5
_classes_count = 3
_keep_tokens = True
_default_stemmer = MystemWrapper()
_default_pos_tagger = POSMystemWrapper(_default_stemmer.mystem)
_terms_per_context = 50
_bags_per_minibatch = 6
_bag_size = 3
_word_embedding = None
_optimiser = tf.train.AdadeltaOptimizer(
learning_rate=0.5,
epsilon=10e-6,
rho=0.95)
_word_embedding_path = "../../../data/w2v/news_rusvectores2.bin.gz" \
if not DebugKeys.UseDebugEmbeddingPlaceholder \
else "../../../data/w2v/banki_ru_300_e5.txt.gz"
_word_embedding_is_binary = True \
if not DebugKeys.UseDebugEmbeddingPlaceholder \
else False
_term_embedding_matrix = None # Includes embeddings of: words, entities, tokens.
_class_weights = None
_use_pos_emb = True
_pos_emb_size = 5
_dist_emb_size = 5
_relations_label_calc_mode = LabelCalculationMode.AVERAGE
def __init__(self, load_embedding=True):
if DebugKeys.LoadWordEmbedding:
print "Loading embedding: {}".format(self._word_embedding_path)
if load_embedding:
self._word_embedding = RusvectoresEmbedding.from_file(
filepath=join(dirname(__file__), self._word_embedding_path),
binary=self._word_embedding_is_binary,
stemmer=self.Stemmer,
pos_tagger=self.PosTagger)
if DebugKeys.LoadWordEmbedding:
print "Embedding has been loaded."
@property
def DistanceEmbeddingSize(self):
return self._dist_emb_size
@property
def RelationLabelCalculationMode(self):
return self._relations_label_calc_mode
@property
def TermEmbeddingMatrix(self):
return self._term_embedding_matrix
@property
def TermEmbeddingShape(self):
return self._term_embedding_matrix.shape
@property
def TotalAmountOfTermsInEmbedding(self):
"""
Returns vocabulary size -- total amount of words/terms,
for which embedding has been provided
"""
return self.TermEmbeddingShape(0)
def set_term_embedding(self, embedding_matrix):
self._term_embedding_matrix = embedding_matrix
def set_class_weights(self, class_weights):
assert(isinstance(class_weights, list))
assert(len(class_weights) == self._classes_count)
self._class_weights = class_weights
def update_terms_per_context(self, min_possible_value):
assert(isinstance(min_possible_value, int) and min_possible_value > 0)
self._terms_per_context = min(min_possible_value, self._terms_per_context)
@property
def ClassesCount(self):
return self._classes_count
@property
def Stemmer(self):
return self._default_stemmer
@property
def PosTagger(self):
return self._default_pos_tagger
@property
def ClassWeights(self):
return self._class_weights
@property
def Optimiser(self):
return self._optimiser
@property
def TestOnEpochs(self):
return self._test_on_epoch
@property
def BatchSize(self):
return self.BagSize * self.BagsPerMinibatch
@property
def BagSize(self):
return self._bag_size
@property
def BagsPerMinibatch(self):
return self._bags_per_minibatch
@property
def Dropout(self):
return self._dropout
@property
def KeepTokens(self):
return self._keep_tokens
@property
def TermsPerContext(self):
return self._terms_per_context
@property
def UseClassWeights(self):
return self._use_class_weights
@property
def WordEmbedding(self):
return self._word_embedding
@property
def UsePOSEmbedding(self):
return self._use_pos_emb
@property
def PosEmbeddingSize(self):
return self._pos_emb_size
@property
def Epochs(self):
return max(self.TestOnEpochs) + 1
def _internal_get_parameters(self):
return [
("use_class_weights", self.UseClassWeights),
("dropout", self.Dropout),
("classes_count", self.ClassesCount),
("keep_tokens", self.KeepTokens),
("class_weights", self.ClassWeights),
("default_stemmer", self.Stemmer),
("default_pos_tagger", self.PosTagger),
("terms_per_context", self.TermsPerContext),
("bags_per_minibatch", self.BagsPerMinibatch),
("bag_size", self.BagSize),
("batch_size", self.BatchSize),
("word_embedding_path", self._word_embedding_path),
("use_pos_emb", self.UsePOSEmbedding),
("pos_emb_size", self.PosEmbeddingSize),
("dist_embedding_size", self.DistanceEmbeddingSize),
("relations_label_calc_mode", self.RelationLabelCalculationMode),
("optimizer", self.Optimiser)
]
def get_parameters(self):
return [list(p) for p in zip(*self._internal_get_parameters())]