-
Notifications
You must be signed in to change notification settings - Fork 0
/
app_han.py
291 lines (234 loc) · 9.01 KB
/
app_han.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
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
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
# -*- coding: utf-8 -*-
# Implementation of Hierarchical Attentional Networks for Document Classification
# http://www.cs.cmu.edu/~./hovy/papers/16HLT-hierarchical-attention-networks.pdf
# https://github.com/arunarn2/HierarchicalAttentionNetworks
# requirements: python3.6, tensorflow 1.4.0, keras 2.0.8 !!!
from keras.callbacks import Callback, ModelCheckpoint
from keras import backend as K
from keras.models import Model
from keras import initializers
from keras.engine.topology import Layer
from keras.optimizers import Adam, RMSprop
from keras.layers import Dense, Input
from keras.layers import Embedding, GRU, Bidirectional, TimeDistributed
from utils import get_glove_embeddings, log_model, ensure_path_exists
from data import train_data, test_data
from tweets import Helpers, tweets_preprocessor
from sklearn.model_selection import train_test_split
from keras.preprocessing.text import Tokenizer
from configs import get_preprocessing_algorithm
import numpy as np
import uuid
PREDICTIONS = {
'x_train': [],
'x_val': [],
'x_test': [],
}
class TestDataCallback(Callback):
def __init__(self, x_test, y_test, x_train, x_val):
super().__init__()
self.accuracy = []
self.loss = []
self.x_train = x_train
self.x_val = x_val
self.x_test = x_test
self.y_test = y_test
def on_epoch_end(self, epoch, logs=None):
score = self.model.evaluate(self.x_test, self.y_test, verbose=1)
self.loss.append(score[0])
PREDICTIONS['x_train'].append([v[0] for v in self.model.predict(self.x_train).tolist()])
PREDICTIONS['x_val'].append([v[0] for v in self.model.predict(self.x_val).tolist()])
PREDICTIONS['x_test'].append([v[0] for v in self.model.predict(self.x_test).tolist()])
self.accuracy.append(score[1])
# class defining the custom attention layer
class HierarchicalAttentionNetwork(Layer):
def __init__(self, attention_dim):
self.init = initializers.get('normal')
self.supports_masking = True
self.attention_dim = attention_dim
super(HierarchicalAttentionNetwork, self).__init__()
def build(self, input_shape):
assert len(input_shape) == 3
self.W = K.variable(self.init((input_shape[-1], self.attention_dim)))
self.b = K.variable(self.init((self.attention_dim,)))
self.u = K.variable(self.init((self.attention_dim, 1)))
self.trainable_weights = [self.W, self.b, self.u]
super(HierarchicalAttentionNetwork, self).build(input_shape)
def compute_mask(self, inputs, mask=None):
return mask
def call(self, x, mask=None):
# size of x :[batch_size, sel_len, attention_dim]
# size of u :[batch_size, attention_dim]
# uit = tanh(xW+b)
uit = K.tanh(K.bias_add(K.dot(x, self.W), self.b))
ait = K.exp(K.squeeze(K.dot(uit, self.u), -1))
if mask is not None:
# Cast the mask to floatX to avoid float64 upcasting
ait *= K.cast(mask, K.floatx())
ait /= K.cast(K.sum(ait, axis=1, keepdims=True) + K.epsilon(), K.floatx())
weighted_input = x * K.expand_dims(ait)
output = K.sum(weighted_input, axis=1)
return output
def compute_output_shape(self, input_shape):
return input_shape[0], input_shape[-1]
OPTIMIZERS = {
'adam': Adam,
'rmsprop': RMSprop
}
DEFAULTS = {
'ACTIVATION': 'sigmoid',
'OPTIMIZER': 'adam',
'LEARNING_RATE': 1e-4,
}
def get_han_model(config):
embedding_layer = Embedding(**config['EMBEDDING_OPTIONS'], mask_zero=True)
sentence_input = Input(shape=(config['EMBEDDING_OPTIONS']['input_length'],), dtype='int32')
embedded_sequences = embedding_layer(sentence_input)
lstm_word = Bidirectional(GRU(config['GRU_UNITS'], return_sequences=True))(embedded_sequences)
attn_word = HierarchicalAttentionNetwork(config['ATTN_UNITS'])(lstm_word)
sentenceEncoder = Model(sentence_input, attn_word)
review_input = Input(shape=(1, config['EMBEDDING_OPTIONS']['input_length']), dtype='int32')
review_encoder = TimeDistributed(sentenceEncoder)(review_input)
lstm_sentence = Bidirectional(GRU(config['GRU_UNITS'], return_sequences=True))(review_encoder)
attn_sentence = HierarchicalAttentionNetwork(config['ATTN_UNITS'])(lstm_sentence)
preds = Dense(
1,
activation=config.get('ACTIVATION', DEFAULTS['ACTIVATION'])
)(attn_sentence)
model = Model(review_input, preds)
model.compile(
optimizer=OPTIMIZERS[
config.get('OPTIMIZER', DEFAULTS['OPTIMIZER'])
](
lr=config.get(
'LEARNING_RATE',
DEFAULTS['LEARNING_RATE']
)
),
loss='binary_crossentropy',
metrics=['accuracy']
)
return model
# model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['acc'])
# print("model fitting - Hierachical attention network")
# model.fit(x_train, y_train, validation_data=(x_val, y_val), nb_epoch=10, batch_size=100)
def fit(model, data, config):
is_with_test_data = len(data) == 6
if is_with_test_data:
x_train, y_train, x_val, y_val, x_test, y_test = data
else:
x_train, y_train, x_val, y_val = data
callbacks = []
if is_with_test_data:
test_data_callback = TestDataCallback(
x_test=x_test,
y_test=y_test,
x_train=x_train,
x_val=x_val
)
callbacks.append(test_data_callback)
if config.get('DIR') is not None and config.get('PREFIX') is not None:
suffix = '-e{epoch:03d}-a{acc:03f}-va{val_acc:03f}-ta.hdf5'
callbacks.append(ModelCheckpoint(
config['DIR'] + config['PREFIX'] + suffix,
verbose=1,
monitor='val_loss',
save_best_only=True,
mode='auto'
))
history = model.fit(
x=x_train, y=y_train,
batch_size=config['BATCH_SIZE'],
epochs=config['EPOCHS'],
verbose=1,
validation_data=(
x_val,
y_val
),
callbacks=callbacks
)
model_history = history.history.copy()
model_history['test_loss'] = test_data_callback.loss
model_history['test_accuracy'] = test_data_callback.accuracy
model_history = {
k: [round(float(v), 6) for v in data] for k, data in model_history.items()
}
return model_history
HAN_CONFIG = {
'TYPE': 'HAN',
'BATCH_SIZE': 32,
'EPOCHS': 5,
'OPTIMIZER': 'adam',
'LEARNING_RATE': 1e-4,
'EMBEDDING_OPTIONS': {
'output_dim': 256,
},
'GRU_UNITS': 128,
'ATTN_UNITS': 128,
}
TRAIN_UUID = str(uuid.uuid4())
SEED = 7
USE_GLOVE = False
NETWORK_KEY = 'HAN'
PREPROCESSING_ALGORITHM_ID = 'a85c8435'
PREPROCESSING_ALGORITHM = get_preprocessing_algorithm(PREPROCESSING_ALGORITHM_ID, join=True)
CONFIG = HAN_CONFIG.copy()
if USE_GLOVE:
CONFIG['GLOVE'] = {
'SIZE': 200
}
GLOVE = f'glove.twitter.27B.{CONFIG["GLOVE"]["SIZE"]}d.txt'
GLOVE_FILE_PATH = f'./data/glove/{GLOVE}'
GLOVE_EMBEDDINGS = get_glove_embeddings(GLOVE_FILE_PATH)
CONFIG['TRAIN_UUID'] = TRAIN_UUID
CONFIG['UUID'] = str(uuid.uuid4())
CONFIG['PREPROCESSING_ALGORITHM'] = PREPROCESSING_ALGORITHM
CONFIG['PREPROCESSING_ALGORITHM_UUID'] = PREPROCESSING_ALGORITHM_ID
CONFIG['HISTORY'] = None
# CONFIG['DIR'] = f'./data-saved-models/glove-true/{NETWORK_KEY}/'
# ensure_path_exists(CONFIG['DIR'])
# CONFIG['PREFIX'] = f'{NETWORK_KEY}-{PREPROCESSING_ALGORITHM_ID}-SEED-{SEED}'
train_data['preprocessed'] = tweets_preprocessor.preprocess(
train_data.text,
PREPROCESSING_ALGORITHM,
keywords=train_data.keyword,
locations=train_data.location
)
test_data['preprocessed'] = tweets_preprocessor.preprocess(
test_data.text,
PREPROCESSING_ALGORITHM,
keywords=test_data.keyword,
locations=test_data.location
)
train_inputs, val_inputs, train_targets, val_targets = train_test_split(
train_data['preprocessed'],
train_data['target'],
test_size=0.3,
random_state=SEED
)
keras_tokenizer = Tokenizer()
(x_train, x_val, x_test), input_dim, input_len = Helpers.get_model_inputs(
(train_inputs, val_inputs, test_data.preprocessed),
keras_tokenizer
)
y_train = train_targets
y_val = val_targets
y_test = test_data.target.values
x_train = np.array([[v] for v in x_train])
x_val = np.array([[v] for v in x_val])
x_test = np.array([[v] for v in x_test])
CONFIG['EMBEDDING_OPTIONS']['input_dim'] = input_dim
CONFIG['EMBEDDING_OPTIONS']['input_length'] = input_len
if USE_GLOVE:
Helpers.with_glove_embedding_options(CONFIG, keras_tokenizer, GLOVE_EMBEDDINGS)
model = get_han_model(CONFIG)
history = fit(model, (x_train, y_train, x_val, y_val, x_test, y_test), CONFIG)
CONFIG['HISTORY'] = history
loss_index = history['val_loss'].index(min(history['val_loss']))
CONFIG['PREDICTIONS'] = {
'x_train': PREDICTIONS['x_train'][loss_index],
'x_val': PREDICTIONS['x_val'][loss_index],
'x_test': PREDICTIONS['x_test'][loss_index],
}
log_model(CONFIG)
print('VAL_ACC: {} ; TEST ACC: {}'.format(history['val_acc'][loss_index], history['test_accuracy'][loss_index]))