forked from gray-stanton/spamGAN
-
Notifications
You must be signed in to change notification settings - Fork 0
/
final_runner.py
167 lines (135 loc) · 6.32 KB
/
final_runner.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
import os
import sys
import csv
import importlib
import spamGAN_train
import texar
import random
BASEDIR = '/home/yankun/spamGAN_output/'
def get_config_file(trp, usp):
if usp == -1:
return 'stepGAN_base_config_nogan'
if usp == 0.0:
return 'stepGAN_base_config_nounsup'
if usp == 0.5 or usp == 0.6:
return 'stepGAN_base_config_smallunsup'
if usp == 0.7 or usp == 0.8:
return 'stepGAN_base_config_smallunsup'
if usp == 0.9 or usp == 1.0:
return 'stepGAN_base_config_smallunsup'
unsup_revs_path = '/home/yankun/spamGAN_output/unlabeled_review.txt'
train_revs = '/home/yankun/spamGAN_output/train_review.txt'
train_labs = '/home/yankun/spamGAN_output/train_label.txt'
test_revs = '/home/yankun/spamGAN_output/test_review.txt'
test_labs = '/home/yankun/spamGAN_output/test_label.txt'
def make_data(trp, usp, run):
nogan = False
if usp == -1:
usp = 0.0
nogan = True
with open(train_revs, 'r') as f:
revs = f.readlines()
with open(train_labs, 'r') as f:
labs = f.readlines()
shfl_idx = random.sample(list(range(len(revs))), len(revs))
revs = [str(revs[i]) for i in shfl_idx]
labs = [str(labs[i]) for i in shfl_idx]
tr = revs[:round(trp * len(revs) * 0.9)]
vr = revs[round(0.9 * trp * len(revs)): round(trp * len(revs))]
tl = labs[:round(trp * len(revs) * 0.9)]
vl = labs[round(0.9 * trp * len(revs)): round(trp * len(revs))]
if len(vr) == 0 :
# just add a fake as a workaround
vr = revs[0:100]
vl = labs[0:100]
with open(unsup_revs_path, 'r') as f:
unsup_revs_full = f.readlines()
random.shuffle(unsup_revs_full)
unsup_revs = unsup_revs_full[:round(usp * len(unsup_revs_full))]
unsup_labs = ['-1\n'] * len(unsup_revs)
dir_name = 'tr{}_usp{}_{}'.format(int(trp*100), int(usp * 100), run)
result_name = 'tr{}_usp{}'.format(int(trp*100), int(usp * 100))
if nogan:
dir_name = dir_name + '_nogan/'
result_name = result_name + '_nogan'
os.mkdir(os.path.join(BASEDIR, dir_name))
curdir = os.path.join(BASEDIR, dir_name)
resultdir = os.path.join(BASEDIR, "result")
result_file = os.path.join(resultdir, result_name)
data_paths = {
'train_data_reviews' : os.path.join(curdir, 'trevs.txt'),
'train_data_labels' : os.path.join(curdir, 'tlabs.txt'),
'val_data_reviews' : os.path.join(curdir, 'vrevs.txt'),
'val_data_labels' : os.path.join(curdir, 'vlabs.txt'),
'unsup_train_data_reviews' : os.path.join(curdir, 'unsup_trevs.txt'),
'unsup_train_data_labels' : os.path.join(curdir, 'unsup_tlabs.txt'),
'vocab' : os.path.join(curdir, 'vocab.txt'),
'clas_test_ckpt' : os.path.join(curdir, 'ckpt-bestclas'),
'clas_pred_output' : os.path.join(curdir, 'testpreds.txt'),
'dir' : curdir,
'result_file' : result_file,
'clas_pretrain_save' : nogan
}
with open(data_paths['train_data_reviews'], 'w') as f:
for x in tr:
f.write(x)
with open(data_paths['train_data_labels'], 'w') as f:
for x in tl:
f.write(str(x))
with open(data_paths['unsup_train_data_reviews'], 'w') as f:
for x in unsup_revs:
f.write(x)
with open(data_paths['unsup_train_data_labels'], 'w') as f:
for x in unsup_labs:
f.write(str(x))
with open(data_paths['val_data_reviews'], 'w') as f:
for x in vr:
f.write(x)
with open(data_paths['val_data_labels'], 'w') as f:
for x in vl:
f.write(str(x))
vocab = texar.data.make_vocab([train_revs, test_revs, data_paths['unsup_train_data_reviews']], 10000)
with open(data_paths['vocab'], 'w') as f:
for v in vocab:
f.write(v + '\n')
return data_paths
# 0.5, 0.8 x 0.5, 0.8
for train_pcent in [0.9]:
for unsup_pcent in [-1]:
for run in range(0, 5):
base_config_file = 'spamGAN_config_smallunsup'
data_paths = make_data(train_pcent, unsup_pcent, run)
importlib.invalidate_caches()
base_config = importlib.import_module(base_config_file)
base_config = importlib.reload(base_config)
# inject file paths
base_config.train_data['datasets'][0]['files'] = [data_paths['train_data_reviews'],
data_paths['unsup_train_data_reviews']]
base_config.train_data['datasets'][1]['files' ] = [data_paths['train_data_labels'],
data_paths['unsup_train_data_labels']]
base_config.clas_train_data['datasets'][0]['files'] = data_paths['train_data_reviews']
base_config.clas_train_data['datasets'][1]['files'] = data_paths['train_data_labels']
base_config.val_data['datasets'][0]['files'] = data_paths['val_data_reviews']
base_config.val_data['datasets'][1]['files'] = data_paths['val_data_labels']
base_config.test_data['datasets'][0]['files'] = test_revs
base_config.test_data['datasets'][1]['files'] = test_labs
base_config.train_data['datasets'][0]['vocab_file'] = data_paths['vocab']
base_config.clas_train_data['datasets'][0]['vocab_file'] = data_paths['vocab']
base_config.val_data['datasets'][0]['vocab_file'] = data_paths['vocab']
base_config.test_data['datasets'][0]['vocab_file'] = data_paths['vocab']
base_config.clas_test_ckpt = data_paths['clas_test_ckpt']
base_config.clas_pred_output = data_paths['clas_pred_output']
base_config.log_dir = data_paths['dir']
base_config.checkpoint_dir = data_paths['dir']
print(base_config.train_data['datasets'][0]['files'])
print('Train Pcent {} Unsup Pcent {} Run {}'.format(train_pcent, unsup_pcent, run))
# Run
dict_res = spamGAN_train.main(base_config)
file_exists = os.path.isfile(data_paths["result_file"])
f = open(data_paths["result_file"],'a')
w = csv.DictWriter(f, dict_res.keys())
if not file_exists:
print("writing header")
w.writeheader()
w.writerow(dict_res)
f.close()