-
Notifications
You must be signed in to change notification settings - Fork 0
/
gen_labelled_data.py
213 lines (196 loc) · 8.3 KB
/
gen_labelled_data.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
import tempfile
import datetime
import time
import os
import sys
import tensorflow as tf
import numpy as np
import ray
from cnf_util import *
from tftd import TFDC, tfdc_to_example
import sr
from util import *
from gen_fmlas import *
@ray.remote
def worker(writer, get_fmla, tmpdir):
while True:
try:
writer.write_example.remote(tfdc_to_example(gen_tfdc(get_fmla(), tmpdir)))
return 0
except FileNotFoundError:
continue
@ray.remote
class Worker:
def __init__(self, writer, get_fmla, tmpdir):
self.datapoint = None
self.get_fmla = get_fmla
self.tmpdir = tmpdir
self.writer = writer
def _gen_datapoint(self):
while True:
try:
self.datapoint = gen_tfdc(self.get_fmla(), self.tmpdir)
break
except FileNotFoundError: # TODO(jesse): fix this upstream
continue
def serialize_datapoint(self):
return tfdc_to_example(self.datapoint)
def main(self):
self._gen_datapoint()
self.writer.write_example.remote(self.serialize_datapoint())
return 0
@ray.remote
class Writer:
def __init__(self, data_dir, n_tfrs_per_file):
self.start_time = time.time()
self.opts = mk_data_writer_opts(data_dir, n_tfrs_per_file)
self.wtr = data_writer(self.opts)
self.write_count = 0
datestring = datetime.date.strftime(datetime.datetime.now(),"%Y-%m-%d-%H-%M")
self.log_path = os.path.join(data_dir,"logs/", "datagen_" + datestring + ".log")
if not os.path.exists(os.path.dirname(self.log_path)):
os.makedirs(os.path.dirname(self.log_path))
with open(self.log_path, "a") as f:
f.write(f"{datetime.datetime.now()}: writer initialized\n")
def write_log(self, arg=None):
with open(self.log_path, "a") as f:
if arg is None:
f.write(f"{datetime.datetime.now()}: current count {self.write_count}" + "\n")
f.write(f"elapsed time: {datetime.timedelta(seconds=(time.time() - self.start_time))}" + "\n")
else:
f.write(arg + "\n")
def write_example(self, e):
print("got something, writing example")
self.wtr.write_example(e)
self.write_count += 1
return 0
def finalize(self):
self.write_log("finalizing")
self.write_log()
self.wtr.finalize()
def count(self):
return self.write_count
def test_ray():
data_dir = os.path.join(PROJECT_DIR, "train_data", "scratch")
log_dir = os.path.join(data_dir, "log")
n_tfrs_per_file = 150
writer = Writer.remote(data_dir, n_tfrs_per_file)
count = 0
while count < 300:
with tempfile.TemporaryDirectory() as tmpdir:
tmpdir = tmpdir + "/"
jobs = []
for _ in range(4):
worker = Worker.remote(writer, lambda : get_unsat_randkcnf(3,40), tmpdir)
jobs += [worker.main.remote() for _ in range(50)]
ray.get(jobs)
ray.get(writer.write_log.remote())
count += ray.get(writer.count.remote())
print("count: ", count)
ray.get(writer.finalize.remote())
def gen_ramsey(s=4, k=4, N=18, c=30, data_dir = os.path.join(PROJECT_DIR, "train_data", "ramsey2", "test"), n_tfrs_per_file=100, n_datapoints=1000, batch_size=100, num_threads=6):
check_make_path(data_dir)
get_fmla = lambda: gen_ramsey_fragment(s,k,N,c)
writer = Writer.remote(data_dir, n_tfrs_per_file)
count = 0
while count < n_datapoints:
with tempfile.TemporaryDirectory() as tmpdir:
tmpdir = tmpdir + "/"
jobs = []
jobs_per_worker = int(np.ceil((np.minimum(abs(n_datapoints - count), batch_size))/num_threads))
for _ in range(num_threads):
worker = Worker.remote(writer, get_fmla, tmpdir)
# jobs = jobs + [worker.remote(writer, get_fmla, tmpdir) for _ in range(jobs_per_worker)]
jobs = jobs + [worker.main.remote() for _ in range(jobs_per_worker)]
ray.get(jobs)
ray.get(writer.write_log.remote())
count = ray.get(writer.count.remote())
print("count: ", count)
ray.get(writer.finalize.remote())
print("done")
def gen_randkcnf(k=3, n=40, data_dir = os.path.join(PROJECT_DIR, "train_data", "randkcnf", "train"), n_tfrs_per_file=5000, n_datapoints = 65000, batch_size = 500, num_threads = 4):
check_make_path(data_dir)
get_fmla = lambda: get_unsat_randkcnf(k,n)
writer = Writer.remote(data_dir, n_tfrs_per_file)
count = 0
while count < n_datapoints:
with tempfile.TemporaryDirectory() as tmpdir:
tmpdir = tmpdir + "/"
jobs = []
jobs_per_worker = int(np.ceil((np.minimum(abs(n_datapoints - count), batch_size))/num_threads))
for _ in range(num_threads):
worker = Worker.remote(writer, get_fmla, tmpdir)
# jobs = jobs + [worker.remote(writer, get_fmla, tmpdir) for _ in range(jobs_per_worker)]
jobs = jobs + [worker.main.remote() for _ in range(jobs_per_worker)]
ray.get(jobs)
ray.get(writer.write_log.remote())
count = ray.get(writer.count.remote())
print("count: ", count)
ray.get(writer.finalize.remote())
print("done")
def gen_sr(n1=10, n2=40, min_cls_len=2, data_dir = os.path.join(PROJECT_DIR, "train_data", "sr", "train"), n_tfrs_per_file=5000, n_datapoints = 185000, batch_size = 500, num_threads = 4):
check_make_path(data_dir)
get_fmla = lambda: get_unsat_sr(n1,n2,min_cls_len)
writer = Writer.remote(data_dir, n_tfrs_per_file)
count = 0
while count < n_datapoints:
with tempfile.TemporaryDirectory() as tmpdir:
tmpdir = tmpdir + "/"
jobs = []
jobs_per_worker = int(np.ceil((np.minimum(abs(n_datapoints - count), batch_size))/num_threads))
for _ in range(num_threads):
worker = Worker.remote(writer, get_fmla, tmpdir)
# jobs = jobs + [worker.remote(writer, get_fmla, tmpdir) for _ in range(jobs_per_worker)]
jobs = jobs + [worker.main.remote() for _ in range(jobs_per_worker)]
ray.get(jobs)
ray.get(writer.write_log.remote())
count = ray.get(writer.count.remote())
print("count: ", count)
ray.get(writer.finalize.remote())
print("done")
def gen_src(n1=20, n2=100, min_cls_len=2, data_dir = os.path.join(PROJECT_DIR, "train_data", "src", "train"), n_tfrs_per_file=5000, n_datapoints = 195000, batch_size = 500, num_threads = 4):
check_make_path(data_dir)
get_fmla = lambda: get_unsat_src(n1,n2,min_cls_len)
writer = Writer.remote(data_dir, n_tfrs_per_file)
count = 0
while count < n_datapoints:
with tempfile.TemporaryDirectory() as tmpdir:
tmpdir = tmpdir + "/"
jobs = []
jobs_per_worker = int(np.ceil((np.minimum(abs(n_datapoints - count), batch_size))/num_threads))
for _ in range(num_threads):
worker = Worker.remote(writer, get_fmla, tmpdir)
# jobs = jobs + [worker.remote(writer, get_fmla, tmpdir) for _ in range(jobs_per_worker)]
jobs = jobs + [worker.main.remote() for _ in range(jobs_per_worker)]
ray.get(jobs)
ray.get(writer.write_log.remote())
count = ray.get(writer.count.remote())
print("count: ", count)
ray.get(writer.finalize.remote())
print("done")
def ray_gen_argparse():
import argparse
parser = argparse.ArgumentParser()
parser.add_argument("--sr", action="store_true", dest="sr")
parser.add_argument("--src", action="store_true", dest="src")
parser.add_argument("--randkcnf", action="store_true", dest="randkcnf")
parser.add_argument("--ramsey", action="store_true", dest="ramsey")
parser.add_argument("--datapoints", action="store", dest="n_datapoints", type=int)
parser.add_argument("--test", action="store_true", dest="test")
parser.add_argument("--threads", action="store", dest="num_threads", type=int)
parser.add_argument("--batch-size", action="store", dest="batch_size", type=int, default=80)
return parser.parse_args()
if __name__ == "__main__":
opts = ray_gen_argparse()
assert exactly_one([opts.sr, opts.src, opts.randkcnf, opts.ramsey, opts.test])
ray.init()
if opts.test:
gen_sr(n_datapoints=300, num_threads=3, data_dir = os.path.join(PROJECT_DIR, "train_data_test/sr/train/"), batch_size=opts.batch_size)
elif opts.sr:
gen_sr(n_datapoints=opts.n_datapoints, num_threads=opts.num_threads, batch_size=opts.batch_size)
elif opts.src:
gen_src(n_datapoints=opts.n_datapoints, num_threads=opts.num_threads, batch_size=opts.batch_size)
elif opts.ramsey:
gen_ramsey(n_datapoints=opts.n_datapoints, num_threads=opts.num_threads, batch_size=opts.batch_size)
elif opts.randkcnf:
gen_randkcnf(n_datapoints=opts.n_datapoints, num_threads=opts.num_threads, batch_size=opts.batch_size)