-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
381 lines (320 loc) · 12.6 KB
/
utils.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
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
# -*- coding:utf-8 -*-
# @Author: Wei Yi
import os
import random
import pickle
import BE
import shutil
from config import Config
from chinese_word_segmentation.sampler import GibbsSampler
from chinese_word_segmentation.FreqDict import FreqDict
from chinese_word_segmentation.DataLoader import DataLoader
from produce_tf import produce_tfrecord
from to_tfdata import produce_tfrecord as pt
from BERT_PRED import BERT_PRED
def read_from_pkl(filename):
pkl = open(filename, 'rb')
data = pickle.load(pkl)
pkl.close()
return data
def save_as_pkl(filename, data):
output_pkl = open(filename, 'wb')
pickle.dump(data, output_pkl)
output_pkl.close()
def check_dir(dir_name):
if not os.path.exists(dir_name):
os.mkdir(dir_name)
if os.path.exists("data/train_step.pkl"):
os.remove("data/train_step.pkl")
if os.path.exists("data/fq.pkl"):
os.remove("data/fq.pkl")
check_dir("data")
check_dir("temp")
check_dir("tfdata")
check_dir("output")
INPUT_FILE = "data/data.txt"
TEMP_FILE = "temp/temp.txt"
OUTPUT_FILE = "data/train.txt"
REM_FILE = "data/data.txt"
TRAIN_FILE = "data/msit.txt"
if os.path.exists(OUTPUT_FILE):
os.remove(OUTPUT_FILE)
if os.path.exists(TRAIN_FILE):
os.remove(TRAIN_FILE)
CUR_STAGE = 0
def read_txt_file(filename):
with open(filename, 'r', encoding="utf-8") as file:
return file.read().split('\n')
def save_txt_file(filename, sents):
with open(filename, 'w', encoding="utf-8") as file:
for idx, sent in enumerate(sents):
if idx != len(sents) - 1:
file.write(sent + '\n')
else:
file.write(sent)
def random_cutting(sents, thres=6):
result = []
if thres in set(["BE", "nVBE", "jieba"]):
s2s = read_from_pkl("seg_" + thres + ".pkl")
for sent in sents:
if sent not in s2s.keys():
result.append(sent)
continue
result.append(s2s[sent])
return result
def sampling(input_filename, output_filename):
gs = GibbsSampler()
fq = FreqDict()
if os.path.exists("data/fq.pkl"):
fq = FreqDict(load=True, pkl_name="data/fq.pkl")
gs.sample(freq_dict=fq, textfile=input_filename, max_iters=100, save=True, save_name=output_filename)
def produce_sampling_file(output_cnt):
global INPUT_FILE, TEMP_FILE
seg_type = ["BE", "nVBE", "jieba"]
for cnt in range(output_cnt):
sampling_filename = "temp/sample_" + str(cnt) + ".txt"
sents = read_txt_file(INPUT_FILE)
sents = random_cutting(sents, thres=seg_type[cnt])
save_txt_file(TEMP_FILE, sents=sents)
sampling(input_filename=TEMP_FILE, output_filename=sampling_filename)
def check_word_len(sent):
for w in sent.split(' '):
if len(w) > 1:
return False
return True
def append_pred():
global OUTPUT_FILE, TRAIN_FILE
has_output = os.path.exists(TRAIN_FILE)
selected_cnt = 0
if has_output:
with open(TRAIN_FILE, 'r', encoding="utf-8") as file:
selected_cnt += len(file.read().split("\n"))
with open("temp/predict.txt", 'r', encoding="utf-8") as file:
with open(TRAIN_FILE, 'a', encoding="utf-8") as afile:
lines = [line for line in file.read().split("\n") if len(line) > 0]
for i, line in enumerate(lines):
afile.write(line)
if i != len(lines) - 1:
afile.write("\n")
with open(TRAIN_FILE, 'r', encoding="utf-8") as file:
with open(OUTPUT_FILE, 'w', encoding="utf-8") as rfile:
lines = list()
for line in file.read().split("\n"):
if len(line) > 0:
lines.append(line)
for i, line in enumerate(lines):
rfile.write(line)
if i != len(lines) - 1:
rfile.write("\n")
def tri_select_sentences(filename_list, is_init=False):
global OUTPUT_FILE, REM_FILE, TRAIN_FILE
sents_one = read_txt_file(filename_list[0])
sents_two = read_txt_file(filename_list[1])
sents_thr = read_txt_file(filename_list[2])
selected_cnt = 0
new_sent_cnt = 0
has_output = os.path.exists(TRAIN_FILE)
if has_output:
with open(TRAIN_FILE, 'r', encoding="utf-8") as file:
selected_cnt += len(file.read().split("\n"))
with open(TRAIN_FILE, 'a', encoding="utf-8") as file:
with open(REM_FILE, 'w', encoding="utf-8") as rfile:
for idx in range(len(sents_one)):
is_last = True if idx == len(sents_one) - 1 else False
if sents_one[idx] == "":
continue
agree = 0
if sents_one[idx] == sents_two[idx]:
agree += 1
if sents_one[idx] == sents_thr[idx]:
agree += 1
if sents_two[idx] == sents_thr[idx]:
agree += 1
is_all_one = check_word_len(sents_one[idx])
if is_all_one:
is_all_one = check_word_len(sents_two[idx])
if is_all_one:
is_all_one = check_word_len(sents_thr[idx])
if is_init:
is_all_one = False
if agree == 3:
# if agree == 3 and not is_all_one:
selected_cnt += 1
new_sent_cnt += 1
file.write(sents_one[idx] + '\n')
else:
if is_last:
rfile.write("".join(sents_one[idx].split(' ')) + '\n')
else:
rfile.write("".join(sents_one[idx].split(' ')) + '\n')
with open(TRAIN_FILE, 'r', encoding="utf-8") as file:
with open(OUTPUT_FILE, 'w', encoding="utf-8") as rfile:
lines = list()
for line in file.read().split("\n"):
if len(line) > 0:
lines.append(line)
for i, line in enumerate(lines):
rfile.write(line)
if i != len(lines) - 1:
rfile.write("\n")
fq = FreqDict()
if os.path.exists("data/fq.pkl"):
fq = FreqDict(load=True, pkl_name="data/fq.pkl")
with open("temp/fq.txt", 'w', encoding="utf-8") as file:
for idx in range(len(sents_one)):
file.write(sents_one[idx] + '\n')
data_loader = DataLoader()
fq = data_loader.update_freq_dict(filename="temp/fq.txt", freq_dict=fq, need_return=True)
fq.save("data/fq.pkl")
save_as_pkl("data/selected_sents.pkl", selected_cnt)
save_as_pkl("data/new_sents.pkl", new_sent_cnt)
def to_label(line):
words = line.split(' ')
label = []
for word in words:
label.append("[BOS]")
rem = len(word) - 1
for _ in range(rem):
label.append("[IOS]")
return "".join(words), " ".join(label)
def to_seg_data(OUTPUT_FILE, out_filename=None):
if out_filename is None:
out_filename = OUTPUT_FILE
seg_lines = []
with open(OUTPUT_FILE, 'r', encoding="utf-8") as file:
lines = file.read().split('\n')
for line in lines:
line, label = to_label(line)
if len(line) > 22:
continue
seg_lines.append(line + '|' + label)
with open(out_filename, 'w', encoding="utf-8") as file:
for seg in seg_lines:
file.write(seg + '\n')
def labeled_txt_to_tf():
global OUTPUT_FILE
to_seg_data(OUTPUT_FILE)
produce_tfrecord()
def del_checkpoint():
filename_list = os.listdir("output/")
for filename in filename_list:
need_del = True
if filename.find("model.ckpt") != -1 or filename == "checkpoint":
need_del = True
if need_del:
os.remove("output/" + filename)
def mv_checkpoint(outdir="output/"):
filename_list = os.listdir("output")
need_del_name_list = ["events.out", "model.ckpt-0", "graph.pbtxt"]
for filename in filename_list:
need_del = False
for name in need_del_name_list:
if filename.find(name) != -1:
need_del = True
break
if need_del:
os.remove("output/" + filename)
continue
if filename.find(".data") != -1:
new_name = "model.ckpt.data-00000-of-00001"
os.rename("output/" + filename, outdir + new_name)
if filename.find(".meta") != -1:
new_name = "model.ckpt.meta"
os.rename("output/" + filename, outdir + new_name)
if filename.find("index") != -1:
new_name = "model.ckpt.index"
os.rename("output/" + filename, outdir + new_name)
if filename.find("checkpoint") != -1:
new_name = "checkpoint"
os.rename("output/" + filename, outdir + new_name)
def save_ckpt():
global CUR_STAGE
save_ckpt_dir = "stage_" + str(CUR_STAGE) + "_ckpt/"
if not os.path.exists(save_ckpt_dir):
os.mkdir(save_ckpt_dir)
CUR_STAGE += 1
filename_list = os.listdir("output")
need_del_name_list = ["events.out", "model.ckpt-0", "graph.pbtxt"]
for filename in filename_list:
need_del = False
for name in need_del_name_list:
if filename.find(name) != -1:
need_del = True
break
if need_del:
os.remove("output/" + filename)
continue
if filename.find(".data") != -1:
new_name = "model.ckpt.data-00000-of-00001"
os.rename("output/" + filename, save_ckpt_dir + new_name)
if filename.find(".meta") != -1:
new_name = "model.ckpt.meta"
os.rename("output/" + filename, save_ckpt_dir + new_name)
if filename.find("index") != -1:
new_name = "model.ckpt.index"
os.rename("output/" + filename, save_ckpt_dir + new_name)
if filename.find("checkpoint") != -1:
new_name = "checkpoint"
os.rename("output/" + filename, save_ckpt_dir + new_name)
shutil.rmtree("output/")
shutil.copytree(save_ckpt_dir, "output/")
def copy_good_ckpt():
good_ckpt_dir = "stage_" + str(Config.last_ckpt) + "_ckpt/"
shutil.rmtree("output/")
shutil.copytree(good_ckpt_dir, "output/")
def label2text(text, label):
res = []
label = label.split(' ')[1:-1]
idx = 0
for i in range(len(label)):
if label[i] != "[BOS]" and label[i] != "[IOS]":
continue
if idx >= len(text):
break
if label[i] == "[BOS]" and res != []:
res.append(' ')
res.append(text[idx])
idx += 1
return "".join(res)
def from_label_to_text(text_file, label_file, output_file):
with open(text_file, 'r', encoding="utf-8") as file:
gt = file.read().split('\n')
with open(label_file, 'r', encoding="utf-8") as file:
pred = file.read().split('\n')
with open(output_file, 'w', encoding="utf-8") as file:
for i in range(len(gt)):
if gt[i] == "":
continue
text = "".join(gt[i].split(' '))
if len(text) == 0:
continue
pred_label = pred[i]
prediction = label2text(text, pred_label)
if i == len(gt) - 1:
file.write(prediction)
else:
file.write(prediction + '\n')
def predict(pred_filename, output_filename, output_dir, as_text):
to_seg_data(pred_filename, "temp/temp.txt")
save_as_pkl("pred_filename.pkl", "temp/temp.txt")
save_as_pkl("output_filename.pkl", output_filename)
save_as_pkl("output_dir.pkl", output_dir)
pt()
BERT_PRED()
if as_text:
from_label_to_text(pred_filename, output_dir + "/pred.txt", "temp/predict.txt")
def get_loss():
with open("log", 'r') as file:
lines = file.read().split('\n')
for idx in range(len(lines) - 1, -1, -1):
if lines[idx].find("Loss for final step") != -1:
return float(lines[idx].split(' ')[-1][:-1])
def init_BE():
with open(Config.data_fn, 'r', encoding="utf-8") as file:
lines = [line for line in file.read().split("\n") if len(line) > 0]
with open("data/data.txt", 'w', encoding="utf-8") as file:
for i, line in enumerate(lines):
file.write(line)
if i != len(lines) - 1:
file.write("\n")
BE.init(corpus="", fn="data/data.txt", redo=False, use_jieba=False)