-
Notifications
You must be signed in to change notification settings - Fork 1
/
main.py
324 lines (286 loc) · 14.8 KB
/
main.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
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
import os
import argparse
import data_model
from data_model import Config, Saver, Progress
from empty_training import EmptyTraining
from serializer import JsonClassSerialize
from tf_word2vec import *
import utilities
import random
# os.environ["CUDA_VISIBLE_DEVICES"] = "0"
parser = argparse.ArgumentParser(description='Word2Vec training tool')
parser.add_argument('-train-model', action='store',
default="skipgram",
dest='train_model',
help='Use for creating config, possible model is: cbow, skipgram')
parser.add_argument('-train-mode', action='store',
default="word2vec",
dest='train_mode',
help='Use for creating config, possible mode is: word2vec, doc2vec')
parser.add_argument('-train-type', action='store',
default=None,
dest='train_type',
help='Use for training, possible type: normal, empty')
parser.add_argument('-create-embedding', action='store_true',
default=False,
dest='is_create_embedding',
help='Create embedding from training model')
parser.add_argument('-create-doc-embedding', action='store_true',
default=False,
dest='is_create_doc_embedding',
help='Create doc embedding from training model')
parser.add_argument('-create-doc-mapper', action='store_true',
default=False,
dest='is_create_doc_mapper',
help='Create doc mapper')
parser.add_argument('-create-category-mapper', action='store_true',
default=False,
dest='is_create_category_mapper',
help='Create category mapper')
parser.add_argument('-eval-doc-embedding', action='store_true',
default=False,
dest='is_eval_doc_embedding',
help='Evaluate doc embedding result, require doc_embedding.json and doc_mapper.json')
parser.add_argument('-eval-new-doc-embedding', action='store_true',
default=False,
dest='is_eval_new_doc_embedding',
help='Evaluate new doc embedding result, require whole network and eval query')
parser.add_argument('-eval-doc-rele-embedding', action='store_true',
default=False,
dest='is_eval_doc_rele_embedding',
help='Evaluate doc relevant embedding result, require input query')
parser.add_argument('-eval-doc-rele-prediction', action='store_true',
default=False,
dest='is_eval_doc_rele_prediction',
help='Evaluate query prediction result')
parser.add_argument('-eval-query', action='store',
default=None,
dest='eval_query',
help='Query for evaluate doc relevant')
parser.add_argument('-create-word-mapper', action='store_true',
default=False,
dest='is_create_word_mapper',
help='Create word_mapper from list of csv')
parser.add_argument('-create-word-count', action='store_true',
default=False,
dest='is_create_word_count',
help='Create word_count from list of csv')
parser.add_argument('-create-config', action='store_true',
default=False,
dest='is_create_config',
help='Create config file from list of csv folder')
parser.add_argument('-create-cnn-config', action='store_true',
default=False,
dest='is_create_cnn_config',
help='Create cnn config file from list of csv folder')
parser.add_argument('-csv-folder-path', action='store',
dest='csv_folder_path',
default=None,
help='Path to csv folder. Eg: ./data/*csv')
parser.add_argument('-word-count-path', action='store',
dest='word_count_path',
default=None,
help='Path to word_count.json')
parser.add_argument('-min-word-count', action='store',
dest='min_word_count',
default=None,
help='Set minimum of count for building vocabulary, if set, use this instead vocabulary size')
parser.add_argument('-vocabulary-size', action='store',
dest='vocabulary_size',
default=10000,
help='Set vocabulary size for building vocabulary')
parser.add_argument('-save-path', action='store',
default=None,
dest='save_folder_path',
help="Set save path for creating mapper/config or loading training config")
parser.add_argument('-doc-embedding-path', action='store',
default=None,
dest='doc_embedding_path',
help="Set doc embedding path for evaluating")
parser.add_argument('-word-mapper-path', action='store',
dest='mapper_path',
default=None,
help='Set word_mapper path for training!')
parser.add_argument('-doc-mapper-path', action='store',
dest='doc_mapper_path',
default=None,
help='Set doc_mapper path for training!')
parser.add_argument('-category-mapper-path', action='store',
dest='category_mapper_path',
default=None,
help='Set category_mapper_path path for training!')
parser.add_argument('-config-path', action='store',
dest='config_path',
default=None,
help='Set config path for training!')
parser.add_argument('-use-preprocessor', action='store_true',
dest='use_preprocessor',
default=True,
help='Should use preprocessor when extract word for building word_count. When training, use config!')
parser.add_argument('-CUDA_VISIBLE_DEVICES', action='store',
dest='CUDA_VISIBLE_DEVICES',
default="0",
help='Set cuda visible device')
parser.add_argument('-use-cpu', action='store_true',
dest='is_use_cpu',
default=False,
help='Set use CPU instead of GPU')
results = parser.parse_args()
os.environ["CUDA_VISIBLE_DEVICES"] = results.CUDA_VISIBLE_DEVICES
seri = JsonClassSerialize()
def build_word_count(save_folder_path, csv_folder_path, use_preprocessor):
assert csv_folder_path is not None
word_count = data_model.build_word_count(csv_folder_path, use_preprocessor)
seri.save(word_count, os.path.join(save_folder_path, "word_count.json"))
return word_count
def build_doc_mapper(save_path, csv_folder_path):
doc_mapper = data_model.DocMapper()
doc_mapper.build_mapper(csv_folder_path)
seri.save(doc_mapper, os.path.join(save_path, "doc_mapper.json"))
def build_category_mapper(save_path, csv_folder_path):
category_mapper = data_model.CategoryMapper()
category_mapper.build_mapper(csv_folder_path)
seri.save(category_mapper, os.path.join(save_path, "category_mapper.json"))
def main():
train_data_saver = Saver()
if results.is_create_cnn_config:
config = data_model.ConfigFactory.generate_cnn_config(results.save_folder_path, results.csv_folder_path)
seri.save(config, os.path.join(results.save_folder_path, "cnn_config.json"))
return
if results.is_create_category_mapper:
build_category_mapper(results.save_folder_path, results.csv_folder_path)
return
if results.is_create_word_count:
build_word_count(results.save_folder_path, results.csv_folder_path, results.use_preprocessor)
return
if results.is_create_doc_mapper:
build_doc_mapper(results.save_folder_path, results.csv_folder_path)
return
if results.is_eval_doc_embedding:
assert (results.doc_embedding_path)
assert (results.doc_mapper_path)
doc_mapper = seri.load(results.doc_mapper_path)
doc_embedding = train_data_saver.load_doc_embedding(doc_mapper, results.doc_embedding_path)
top_eval = 10
for reversed_info in random.sample(list(doc_mapper.reversed_doc_mapper.values()), top_eval):
print(doc_embedding.similar_by(reversed_info[0]))
return
if results.is_create_word_mapper:
print("Creating mapper!")
if results.csv_folder_path is not None:
print("Creating word_count.json from csv folder {}".format(results.csv_folder_path))
word_count = build_word_count(results.save_folder_path, results.csv_folder_path, results.use_preprocessor)
else:
assert results.word_count_path is not None
print("Loading word_count.json from {}".format(results.word_count_path))
word_count = seri.load(results.word_count_path)
if results.min_word_count is not None:
word_mapper = word_count.get_vocab_by_min_count(int(results.min_word_count))
print("Successfully create word_mapper length {} with min_word_count {}".format(
word_mapper.get_vocabulary_size(),
results.min_word_count))
else:
word_mapper = word_count.get_vocab_by_size(int(results.vocabulary_size))
print("Successfully create word_mapper length {} with vocabulary_size {}".format(
word_mapper.get_vocabulary_size(),
results.vocabulary_size))
seri.save(word_mapper, os.path.join(results.save_folder_path, "word_mapper.json"))
return
if results.is_create_config:
print("Creating config!")
build_config(results.save_folder_path, results.csv_folder_path, results.train_model, results.train_mode)
return
# save_folder_path = results.save_folder_path
config_path = results.config_path
word_mapper_path = results.mapper_path
assert utilities.exists(config_path)
config = train_data_saver.load_config(config_path)
train_data = data_model.DataModelFactory.generate_data_model(config)
assert utilities.exists(word_mapper_path)
train_data_saver.restore_word_mapper(train_data, word_mapper_path)
if utilities.exists(train_data_saver.get_progress_path()):
train_data_saver.restore_progress(train_data, train_data_saver.get_progress_path())
else:
train_data_saver.init_progress(train_data, train_data.config)
train_vec = NetworkFactory.generate_network(config)
if config.mode == "doc2vec":
doc_mapper = seri.load(results.doc_mapper_path)
train_data.set_doc_mapper_data(doc_mapper)
elif config.mode == "docrelevant":
category_mapper = seri.load(results.category_mapper_path)
train_data.set_category_mapper(category_mapper)
train_vec.use_cpu = results.is_use_cpu
train_vec.set_train_data(train_data, train_data_saver)
train_vec.restore_last_training_if_exists()
if results.is_eval_new_doc_embedding:
assert(results.eval_query is not None)
print(train_vec.predict([results.eval_query]))
return
if results.is_create_embedding:
assert (utilities.exists(train_data_saver.get_progress_path()))
print("Creating word embedding from {}".format(train_data.config.save_folder_path))
train_data_saver.save_word_embedding(train_vec.final_embeddings,
train_data.word_mapper.reversed_dictionary)
return
if results.is_create_doc_embedding:
assert (utilities.exists(train_data_saver.get_progress_path()))
print("Creating doc embedding from {}".format(train_data.config.save_folder_path))
doc_embedding = train_vec.get_doc_embedding()
train_data_saver.save_doc_embedding(doc_embedding.embedding,
doc_embedding.doc_mapper.reversed_doc_mapper)
train_data_saver.save_doc_mapper(doc_embedding.doc_mapper)
print(doc_embedding.similar_by(doc_embedding.doc_mapper.reversed_doc_mapper["0"][0]))
return
if results.is_eval_doc_rele_embedding:
assert (utilities.exists(train_data_saver.get_progress_path()))
doc_embedding = train_vec.get_doc_embedding()
if results.eval_query is not None:
print(train_vec.retrieve_by_query([results.eval_query]))
else:
top_eval = 10
query_list = []
for reversed_mapper in random.sample(list(doc_embedding.doc_mapper.reversed_doc_mapper.values()), top_eval):
org_idx = reversed_mapper[0]
csv_path = reversed_mapper[1]
line_number = reversed_mapper[2]
query = utilities.extract_query_from_csv(org_idx, csv_path, line_number)
query_list.append(query)
print(train_vec.retrieve_by_query(query_list))
return
if results.is_eval_doc_rele_prediction:
assert (utilities.exists(train_data_saver.get_progress_path()))
doc_embedding = train_vec.get_doc_embedding()
category_mapper = train_data.category_mapper
top_eval = 10
acc = 0
for reversed_mapper in random.sample(list(doc_embedding.doc_mapper.reversed_doc_mapper.values()), top_eval):
org_idx = reversed_mapper[0]
csv_path = reversed_mapper[1]
line_number = reversed_mapper[2]
post_idx, title, tags, content, catId = utilities.extract_info_from_csv(org_idx, csv_path, line_number)
train_word = preprocessor.get_train_word_from_title_and_tags(title, tags)
prediction = train_vec.get_query_prediction(train_word)
prediction_catId = category_mapper.reversed_dictionary[str(prediction)]
print("Query {}".format(" ".join(train_word)))
print("Prediction catId {} - true catId".format(prediction_catId, catId))
if prediction_catId == catId:
acc += 1
print("Total accuracy {}".format(acc / top_eval))
return
if results.train_type == "empty":
train_vec.empty_training()
elif results.train_type is not None:
train_vec.train()
def build_config(save_folder_path, csv_folder_path, train_model, train_mode):
config = data_model.ConfigFactory.generate_config(save_folder_path, csv_folder_path, train_model, train_mode)
seri.save(config, os.path.join(save_folder_path, "config.json"))
if __name__ == "__main__":
main()
# build_vocab("./temp/", "./data/longdata/*.csv", 10000)
# build_config( "./temp/shortdata/", "./data/shortdata/*.csv")
# word_count = seri.load("./temp/word_count.json")
# print(word_count.word_count["long_văn"])
# vocab = word_count.get_vocab(min_count=100)
# print(len(vocab.dictionary))
# word_count.draw_histogram()
# print(word_count.word_count["sửa_ti_vi_tại"])