-
Notifications
You must be signed in to change notification settings - Fork 0
/
prepare_predictions.py
206 lines (166 loc) · 6.99 KB
/
prepare_predictions.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
from tweets import Helpers, tweets_preprocessor
from models import Sklearn
from tensorflow.keras.models import load_model
from utils import save_to_file, load_classifier
from data import train_data, test_data
from sklearn.model_selection import train_test_split
from tensorflow.keras.preprocessing.text import Tokenizer
from configs import get_preprocessing_algorithm
from pathlib import Path
import numpy as np
from bert_tokenization import FullTokenizer
import tensorflow_hub as hub
saved_models_pathes = [str(lst) for lst in list(Path("./data-saved-models/").rglob("*.h5"))] + \
[str(lst) for lst in list(Path("./data-saved-models/").rglob("*.pickle"))]
SEED = 7
NETWORKS_KEYS = ['LSTM', 'LSTM_DROPOUT', 'BI_LSTM', 'LSTM_CNN', 'FASTTEXT', 'RCNN', 'CNN', 'RNN', 'GRU']
BERT_KEY = ['BERT']
BERT_MODEL = {
'BERT': 'bert_en_uncased_L-24_H-1024_A-16',
'BERT_VERSION': 1,
}
OMIT_BERT = True
CLASSIFIERS_KEYS = ['RIDGE', 'SVC', 'LOGISTIC_REGRESSION', 'SGD', 'DECISION_TREE', 'RANDOM_FOREST']
PREPROCESSING_ALGORITHM_IDS = [
'1258a9d2',
'60314ef9',
'4c2e484d',
'8b7db91c',
'7bc816a1',
'a85c8435',
'b054e509',
'2e359f0b',
'71bd09db',
'd3cc3c6e',
]
PREDICTIONS = {
'bert': {},
'glove-true': {},
'glove-false': {},
'classifiers': {},
}
STATICS = {}
def flatten(lst):
return [sub_sub_l for sub_l in lst for sub_sub_l in sub_l]
def get_probs(lst):
return [np.exp(d) / (1 + np.exp(d)) for d in lst]
correct_targets_saved = False
for folder in PREDICTIONS.keys():
is_classifier = folder == 'classifiers'
is_bert = folder == 'bert'
keys = None
if is_bert and OMIT_BERT:
continue
if is_classifier:
keys = CLASSIFIERS_KEYS
elif is_bert:
keys = BERT_KEY
else:
keys = NETWORKS_KEYS
for key in keys:
try:
model_path = [x for x in saved_models_pathes if f'{folder}/{key}/' in x and f'SEED-{SEED}' in x][0]
except:
continue
data = model_path.split('/')[-1].split('-')
preprocessing_algorithm_id = data[1]
preprocessing_algorithm = get_preprocessing_algorithm(
preprocessing_algorithm_id,
join=(is_classifier or is_bert)
)
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
)
y_train = np.asarray(train_targets)
y_val = np.asarray(val_targets)
y_test = np.asarray(test_data.target.values)
if not correct_targets_saved:
STATICS['y_train'] = y_train.tolist()
STATICS['y_val'] = y_val.tolist()
STATICS['y_test'] = y_test.tolist()
correct_targets_saved = True
if is_bert:
x_train = np.asarray(train_inputs)
x_val = np.asarray(val_inputs)
x_test = np.asarray(test_data_preprocessed)
bert_layer = hub.KerasLayer(
f'https://tfhub.dev/tensorflow/{BERT_MODEL["BERT"]}/{BERT_MODEL["BERT_VERSION"]}',
trainable=True
)
vocab_file = bert_layer.resolved_object.vocab_file.asset_path.numpy()
do_lower_case = bert_layer.resolved_object.do_lower_case.numpy()
tokenizer = FullTokenizer(vocab_file, do_lower_case)
x_train, INPUT_LENGTH = Helpers.get_bert_input(x_train, tokenizer)
x_val = Helpers.get_bert_input(x_val, tokenizer, input_length=INPUT_LENGTH)
x_test = Helpers.get_bert_input(x_test, tokenizer, input_length=INPUT_LENGTH)
model = load_model(model_path, custom_objects={'KerasLayer': hub.KerasLayer})
print(key)
# print(f'a - {model.evaluate(x_train, y_train, verbose=0)}')
# print(f'va - {model.evaluate(x_val, y_val, verbose=0)}')
# print(f'ta - {model.evaluate(x_test, y_test, verbose=1)}')
# print('----------------------------')
x_train_predictions = model.predict(x_train)
x_val_predictions = model.predict(x_val)
x_test_predictions = model.predict(x_test)
PREDICTIONS[folder][f'{key}-{preprocessing_algorithm_id}'] = {
'x_train': flatten(x_train_predictions.tolist()),
'x_val': flatten(x_val_predictions.tolist()),
'x_test': flatten(x_test_predictions.tolist()),
}
save_to_file('./predictions_bert_backup.json', {**PREDICTIONS, **STATICS})
elif not is_classifier:
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,
)
model = load_model(model_path)
print(key)
print(model_path)
# print(f'a - {model.evaluate(x_train, y_train, verbose=0)}')
# print(f'va - {model.evaluate(x_val, y_val, verbose=0)}')
# print(f'ta - {model.evaluate(x_test, y_test, verbose=0)}')
print('----------------------------')
x_train_predictions = model.predict(x_train)
x_val_predictions = model.predict(x_val)
x_test_predictions = model.predict(x_test)
PREDICTIONS[folder][f'{key}-{preprocessing_algorithm_id}'] = {
'x_train': flatten(x_train_predictions.tolist()),
'x_val': flatten(x_val_predictions.tolist()),
'x_test': flatten(x_test_predictions.tolist()),
}
else:
vectorizer = Sklearn.VECTORIZERS[data[2]](**{
'binary': True,
'ngram_range': (int(data[3]), int(data[4]))
})
x_train = vectorizer.fit_transform(train_inputs).todense()
x_val = vectorizer.transform(val_inputs).todense()
x_test = vectorizer.transform(test_data_preprocessed).todense()
cls = load_classifier(model_path)
print(key)
# print(f'a - {cls.score(x_train, y_train)}')
# print(f'va - {cls.score(x_val, y_val)}')
# print(f'ta - {cls.score(x_test, y_test)}')
# print('----------------------------')
PREDICTIONS[folder][f'{key}-{preprocessing_algorithm_id}'] = {
'x_train': cls.predict(x_train).tolist(),
'x_val': cls.predict(x_val).tolist(),
'x_test': cls.predict(x_test).tolist(),
}
save_to_file('./predictions_v_6.json', {**PREDICTIONS, **STATICS})