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
|
#!/usr/bin/python3
import os
import os.path
import sys
from subprocess import Popen, PIPE
from argparse import ArgumentParser
import utils
from myconfig import MyConfig
config = MyConfig()
#change cwd to the libpinyin utils/training directory
libpinyin_dir = config.getToolsDir()
libpinyin_sub_dir = os.path.join(libpinyin_dir, 'data')
os.chdir(libpinyin_sub_dir)
#chdir done
def handleError(error):
sys.exit(error)
#Note: all file passed here should be trained.
def generateOneText(infile, modelfile, reportfile):
infilestatuspath = infile + config.getStatusPostfix()
infilestatus = utils.load_status(infilestatuspath)
if not utils.check_epoch(infilestatus, 'Segment'):
raise utils.EpochError('Please segment first.\n')
if utils.check_epoch(infilestatus, 'Generate'):
return False
#begin processing
cmdline = ['../utils/training/gen_k_mixture_model', \
'--maximum-occurs-allowed', \
str(config.getMaximumOccursAllowed()), \
'--maximum-increase-rates-allowed', \
str(config.getMaximumIncreaseRatesAllowed()), \
'--k-mixture-model-file', \
modelfile, infile + \
config.getSegmentPostfix()]
subprocess = Popen(cmdline, shell=False, stderr=PIPE, \
close_fds=True)
lines = subprocess.stderr.readlines()
if lines:
print('found error report')
with open(reportfile, 'ab') as f:
f.writelines(lines)
f.close()
(pid, status) = os.waitpid(subprocess.pid, 0)
if status != 0:
sys.exit('gen_k_mixture_model encounters error.')
#end processing
utils.sign_epoch(infilestatus, 'Generate')
utils.store_status(infilestatuspath, infilestatus)
return True
#Note: should check the corpus file size, and skip the too small text file.
def handleOneIndex(indexpath, subdir, indexname):
def cleanupFiles(modelnum):
modeldir = os.path.join(config.getModelDir(), subdir, indexname)
modelfile = os.path.join( \
modeldir, config.getCandidateModelName(modelnum))
reportfile = modelfile + config.getReportPostfix()
if os.access(modelfile, os.F_OK):
os.unlink(modelfile)
if os.access(reportfile, os.F_OK):
os.unlink(reportfile)
def storeModelStatus(modelfile, textnum, nexttextnum):
#store model info in status file
modelstatuspath = modelfile + config.getStatusPostfix()
#create None status
modelstatus = {}
modelstatus['GenerateStart'] = textnum
modelstatus['GenerateEnd'] = nexttextnum
utils.sign_epoch(modelstatus, 'Generate')
utils.store_status(modelstatuspath, modelstatus)
print(indexpath, subdir, indexname)
indexstatuspath = indexpath + config.getStatusPostfix()
indexstatus = utils.load_status(indexstatuspath)
if not utils.check_epoch(indexstatus, 'Segment'):
raise utils.EpochError('Please segment first.\n')
if utils.check_epoch(indexstatus, 'Generate'):
return
#continue generating
textnum, modelnum, aggmodelsize = 0, 0, 0
if 'GenerateTextEnd' in indexstatus:
textnum = indexstatus['GenerateTextEnd']
if 'GenerateModelEnd' in indexstatus:
modelnum = indexstatus['GenerateModelEnd']
#clean up previous file
cleanupFiles(modelnum)
#begin processing
indexfile = open(indexpath, 'r')
for i, oneline in enumerate(indexfile.readlines()):
#continue last generating
if i < textnum:
continue
#remove trailing '\n'
oneline = oneline.rstrip(os.linesep)
(title, textpath) = oneline.split('#')
infile = config.getTextDir() + textpath
infilesize = utils.get_file_length(infile + config.getSegmentPostfix())
if infilesize < config.getMinimumFileSize():
print("Skipping " + title + '#' + textpath)
continue
modeldir = os.path.join(config.getModelDir(), subdir, indexname)
os.makedirs(modeldir, exist_ok=True)
modelfile = os.path.join(modeldir, \
config.getCandidateModelName(modelnum))
reportfile = modelfile + config.getReportPostfix()
print("Proccessing " + title + '#' + textpath)
if generateOneText(infile, modelfile, reportfile):
aggmodelsize += infilesize
print("Processed " + title + '#' + textpath)
if aggmodelsize > config.getCandidateModelSize():
nexttextnum = i + 1
storeModelStatus(modelfile, textnum, nexttextnum)
#new model candidate
aggmodelsize = 0
textnum = nexttextnum
modelnum += 1
#clean up next file
cleanupFiles(modelnum)
#save current progress in status file
indexstatus['GenerateTextEnd'] = nexttextnum
indexstatus['GenerateModelEnd'] = modelnum
utils.store_status(indexstatuspath, indexstatus)
nexttextnum = i + 1
storeModelStatus(modelfile, textnum, nexttextnum)
indexfile.close()
#end processing
#save current progress in status file
modelnum += 1
indexstatus['GenerateTextEnd'] = nexttextnum
indexstatus['GenerateModelEnd'] = modelnum
utils.sign_epoch(indexstatus, 'Generate')
utils.store_status(indexstatuspath, indexstatus)
def walkThroughIndex(path):
for root, dirs, files in os.walk(path, topdown=True, onerror=handleError):
for onefile in files:
filepath = os.path.join(root, onefile)
indexpostfix = config.getIndexPostfix()
if onefile.endswith(indexpostfix):
subdir = os.path.relpath(root, path)
indexname = onefile[:-len(indexpostfix)]
handleOneIndex(filepath, subdir, indexname)
elif onefile.endswith(config.getStatusPostfix()):
pass
else:
print('Unexpected file:' + filepath)
if __name__ == '__main__':
parser = ArgumentParser(description='Generate model candidates.')
parser.add_argument('--indexdir', action='store', \
help='index directory', \
default=os.path.join(config.getTextDir(), 'index'))
args = parser.parse_args()
print(args)
walkThroughIndex(args.indexdir)
print('done')
|
