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
|
#!/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
#Please `make -f Makefile.data prepare` first
config = MyConfig()
#change cwd to the libpinyin data directory
libpinyin_dir = config.getEvalsDir()
libpinyin_sub_dir = os.path.join(libpinyin_dir, 'data')
os.chdir(libpinyin_sub_dir)
#chdir done
datafiles = [config.getFinalModelFileName(), \
config.getEvalsTextFileName(), \
'deleted_bigram.db']
def checkData():
#check data files
for onefile in datafiles:
if not os.access(onefile, os.F_OK):
sys.exit('missing one data file:' + onefile)
def cleanUpData():
#begin processing
cmdline = ['/usr/bin/make', 'clean']
subprocess = Popen(cmdline, shell=False, close_fds=True)
(pid, status) = os.waitpid(subprocess.pid, 0)
if status != 0:
sys.exit('make clean for data files failed.')
#end processing
def buildData():
#begin processing
cmdline = ['/usr/bin/make']
subprocess = Popen(cmdline, shell=False, close_fds=True)
(pid, status) = os.waitpid(subprocess.pid, 0)
if status != 0:
sys.exit('make build for data files failed.')
#end processing
def estimateModel(reportfile):
result_line_prefix = "average lambda:"
avg_lambda = 0.
#begin processing
cmdline = ['../utils/training/estimate_interpolation']
subprocess = Popen(cmdline, shell=False, stdout=PIPE, \
close_fds=True)
reporthandle = open(reportfile, 'wb')
for line in subprocess.stdout.readlines():
reporthandle.writelines([line])
#remove trailing '\n'
line = line.decode('utf-8')
line = line.rstrip(os.linesep)
if line.startswith(result_line_prefix):
avg_lambda = float(line[len(result_line_prefix):])
reporthandle.close()
(pid, status) = os.waitpid(subprocess.pid, 0)
if status != 0:
sys.exit('estimate interpolation model returns error.')
#end processing
return avg_lambda
def modifyLambda(lambdaparam):
#begin processing
cmdline = ['/usr/bin/make', 'modify', \
'LAMBDA_PARAMETER=' + str(lambdaparam)]
subprocess = Popen(cmdline, shell=False, close_fds=True)
(pid, status) = os.waitpid(subprocess.pid, 0)
if status != 0:
sys.exit('make rebuild for data files failed.')
#end processing
def evaluateModel(reportfile):
result_line_prefix = "correction rate:"
rate = 0.
#begin processing
cmdline = '../utils/training/eval_correction_rate 2>"' + reportfile + '"'
subprocess = Popen(cmdline, shell=True, stdout=PIPE, \
close_fds=True)
for line in subprocess.stdout.readlines():
#remove training '\n'
line = line.decode('utf-8')
line = line.rstrip(os.linesep)
if line.startswith(result_line_prefix):
rate = float(line[len(result_line_prefix):])
os.waitpid(subprocess.pid, 0)
#end processing
return rate
if __name__ == '__main__':
parser = ArgumentParser(description='Evaluate correction rate.')
parser.add_argument('--finaldir', action='store', \
help='final directory', \
default=config.getFinalModelDir())
parser.add_argument('tryname', action='store', \
help='the storage directory')
args = parser.parse_args()
print(args)
tryname = 'try' + args.tryname
trydir = os.path.join(args.finaldir, tryname)
if not os.access(trydir, os.F_OK):
sys.exit(tryname + "doesn't exist.")
cwdstatuspath = os.path.join(trydir, config.getFinalStatusFileName())
cwdstatus = utils.load_status(cwdstatuspath)
if not utils.check_epoch(cwdstatus, 'Prune'):
raise utils.EpochError('Please tryprune first.')
if utils.check_epoch(cwdstatus, 'Evaluate'):
sys.exit('already evaluated.')
print('checking')
checkData()
modelfile = os.path.join(trydir, config.getFinalModelFileName())
destfile = os.path.join(libpinyin_dir, 'data', \
config.getFinalModelFileName())
utils.copyfile(modelfile, destfile)
print('cleaning')
cleanUpData()
print('building')
buildData()
print('estimating')
reportfile = os.path.join(trydir, 'estimate' + config.getReportPostfix())
avg_lambda = estimateModel(reportfile)
print('average lambda:', avg_lambda)
cwdstatus['EvaluateAverageLambda'] = avg_lambda
utils.store_status(cwdstatuspath, cwdstatus)
print('modifying lambda')
modifyLambda(avg_lambda)
print('evaluating')
reportfile = os.path.join(trydir, 'evaluate' + config.getReportPostfix())
rate = evaluateModel(reportfile)
print(tryname + "'s correction rate:", rate)
cwdstatus['EvaluateCorrectionRate'] = rate
utils.store_status(cwdstatuspath, cwdstatus)
utils.sign_epoch(cwdstatus, 'Evaluate')
utils.store_status(cwdstatuspath, cwdstatus)
print('done')
|
