/
encode.py
175 lines (145 loc) · 6.05 KB
/
encode.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
import os
import time
import sys
import glob
import argparse
import logging as log
import multiprocessing
from ergo.project import Project
from ergo.core.queue import TaskQueue
def parse_args(argv):
parser = argparse.ArgumentParser(prog="ergo encode", description="Encode one or more files to vectors and create or update a csv dataset for training.",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument("project", help="The path containing the model definition.")
parser.add_argument("path", help="Path of a single file or of a folder of files.")
parser.add_argument( "-l", "--label", dest="label", default='auto',
help="The class of the file(s) or 'auto' to use the name of the containing folder.")
parser.add_argument( "-o", "--output", dest="output", default='dataset.csv',
help="Output CSV file names.")
parser.add_argument( "-f", "--filter", dest="filter", default='*.*',
help="If PATH is a folder, this flag determines which files are going to be selected.")
parser.add_argument( "-m", "--multi", dest="multi", default=False, action="store_true",
help="If PATH is a single file, this flag will enable line by line reading of multiple inputs from it.")
parser.add_argument( "-w", "--workers", dest="workers", default=0, type=int,
help="Number of concurrent workers to use for encoding, or zero to run two workers per available CPU core.")
parser.add_argument( "-d", "--delete", dest="delete", default=False, action="store_true",
help="Delete each file after encoding it.")
args = parser.parse_args(argv)
return args
# given the arguments setup and a path, return the label
def label_of(args, path):
if args.label == 'auto':
return os.path.basename(os.path.dirname(path))
else:
return args.label
progress_at = None
prev_done = 0
prev_speed = 0
def get_speed(done):
global progress_at, prev_done, prev_speed
speed = 0
now = time.time()
if progress_at is not None:
delta_v = int(done - prev_done)
if delta_v >= 10:
delta_t = now - progress_at
speed = delta_v / delta_t
else:
speed = prev_speed
now = progress_at
done = prev_done
progress_at = now
prev_done = done
prev_speed = speed
return speed
# simple progress bar
def on_progress(done, total):
maxsz = 80
perc = done / total
sp = " " if perc > 0 else ""
bar = ("█" * int(maxsz * perc)) + sp
sys.stdout.write("\r%d/%d (%d/s) %s%.1f%%" % (done, total, get_speed(done), bar, perc * 100.0))
if perc == 1.0:
sys.stdout.write("\n")
# wait for lines on the results queue and append them to filepath
def appender(filepath, total, results):
log.debug("file appender for %s started ...", filepath)
print()
with open(filepath, 'a+t') as fp:
done = 0
while True:
on_progress(done, total)
line = results.get()
if line is None:
break
fp.write(line.strip() + "\n")
done += 1
# use the project prepare_input function to encode a single
# input into a CSV line
def parse_input(prj, inp, label, results, do_delete = False):
log.debug("encoding '%s' as '%s' ...", inp, label)
x = prj.logic.prepare_input(inp, is_encoding = True)
results.put("%s,%s" % (str(label), ','.join(map(str,x))))
if do_delete and os.path.isfile(inp):
os.remove(inp)
def action_encode(argc, argv):
args = parse_args(argv)
if not os.path.exists(args.path):
log.error("%s does not exist.", args.path)
quit()
prj = Project(args.project)
err = prj.load()
if err is not None:
log.error("error while loading project: %s", err)
quit()
args.label = args.label.strip().lower()
log.info("using %s labeling", 'auto' if args.label == 'auto' else 'hardcoded')
inputs = []
if os.path.isdir(args.path):
in_files = []
if args.label == 'auto':
# the label is inferred from the dirname, so we expect
# args.path to contain multiple subfolders
for subfolder in glob.glob(os.path.join(args.path, "*")):
log.info("enumerating %s ...", subfolder)
in_filter = os.path.join(subfolder, args.filter)
in_sub = glob.glob(in_filter)
n_sub = len(in_sub)
if n_sub > 0:
log.info("collected %d inputs from %s", n_sub, subfolder)
in_files.extend(in_sub)
else:
# grab files directly from args.path
in_filter = os.path.join(args.path, args.filter)
in_files.extend(glob.glob(in_filter))
log.info("collected %d inputs from %s", len(in_files), args.path)
log.info("labeling %d files ...", len(in_files))
for filepath in in_files:
if os.path.isfile(filepath):
inputs.append((label_of(args, filepath), filepath))
elif args.multi:
log.info("parsing multiple inputs from %s ...", args.path)
label = label_of(args, args.path)
with open(args.path, 'rt') as fp:
for line in fp:
inputs.append((label, line))
else:
label = label_of(args, args.path)
inputs.append((label, args.path))
# one encoding queue that pushes to another queue that centralizes
# append operations to a single writer process
num_in = len(inputs)
enc_q = TaskQueue('encoding', args.workers)
res_q = multiprocessing.Queue()
app_p = multiprocessing.Process(target=appender, args=(args.output, num_in, res_q))
# open the output file and start waiting for lines to append
app_p.start()
log.info("encoding %d inputs to %s ...", num_in, args.output)
for (y, x) in inputs:
enc_q.add_task(parse_input, prj, x, y, res_q, args.delete)
# wait for all inputs to be encoded
enc_q.join()
# let the writer know there are no more inputs to read
res_q.put(None)
# wait for the writer to finish
app_p.join()