-
Notifications
You must be signed in to change notification settings - Fork 0
/
benchmark_all.py
197 lines (160 loc) · 7.54 KB
/
benchmark_all.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
import argparse
import datetime
import json
import memory_profiler
import os
import shutil
import subprocess
import threading
import psutil
import time as t
import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
models = ['tpot']
time = 30 # in minutes
n_runs = 3
split_seed = 1
python_bin = '/home/Eldeeb/anaconda3/bin/python3.7'
java_bin = 'java'
resource_log_dir = '/home/Eldeeb/logs'
java_benchmark_jar = 'java/automl-benchmarking-0.0.1-SNAPSHOT-jar-with-dependencies.jar'
autoweka_jar = '/home/Eldeeb/autoweka-2.6/autoweka.jar'
ml_plan_dir = '/home/Eldeeb/ailibs-0.9.0'
mlplan_jar = f'{ml_plan_dir}/AILibs-0.0.1-SNAPSHOT-all.jar'
jars = '"' + ':'.join([os.path.abspath(java_benchmark_jar), autoweka_jar, mlplan_jar]) + '"'
python2_bin = '/home/Eldeeb/anaconda3/envs/py27/bin/python'
recipe_dir = '/home/Eldeeb/Recipe'
def split(dataset_file: str, output_dir: str, p: float = 0.75):
name = os.path.splitext(os.path.basename(dataset_file))[0]
df = pd.read_csv(dataset_file)
train, test = train_test_split(df, train_size=p, random_state=split_seed)
train.to_csv(os.path.join(output_dir, f'{name}_train.csv'), index=False)
test.to_csv(os.path.join(output_dir, f'{name}_test.csv'), index=False)
def benchmark(model: str, train_file: str, test_file: str, output_file: str):
cmd = None
if model in ['autosklearn', 'autosklearn-v', 'autosklearn-m', 'autosklearn-e', 'tpot']:
cmd = ' '.join([python_bin, '-u', 'python/main.py', train_file, output_file,
'-t', str(time), '-m', model, '-te', test_file])
elif model == 'recipe':
cmd = ' '.join([python_bin, '-u', 'python/main.py', train_file, output_file,
'-t', str(time), '-m', model, '-te', test_file,
'-c', python2_bin, recipe_dir])
elif model == 'autoweka':
cmd = ' '.join(['java', '-Xmx6g', '-cp', jars, 'ee.ut.bigdata.Main',
model, train_file, test_file, output_file, str(time)])
elif model == 'mlplan':
cmd = ' '.join(['cd', ml_plan_dir, '&&', 'java', '-Xmx6g', '-cp', jars, 'ee.ut.bigdata.Main',
model, os.path.abspath(train_file), os.path.abspath(test_file),
os.path.abspath(output_file), str(time)])
if cmd:
try:
proc = subprocess.Popen("exec " + cmd, shell=True, stderr=subprocess.STDOUT)
logger(proc.pid)
proc.wait(timeout=time*60*1.5)
except TimeoutExpired:
print("Timeout Expired")
#exit()
finally:
proc.terminate()
def collect(output_dir: str, collect_dir: str):
data = {}
for benchmark in os.listdir(output_dir):
benchmark_dir = os.path.join(output_dir, benchmark)
if os.path.isdir(benchmark_dir):
for dataset in os.listdir(benchmark_dir):
dataset_dir = os.path.join(benchmark_dir, dataset)
if os.path.isdir(dataset_dir):
dataset_name = os.path.splitext(dataset)[0]
for run in os.listdir(dataset_dir):
run_file = os.path.join(dataset_dir, run)
if os.path.getsize(run_file) > 0:
with open(run_file) as f:
run_result = json.load(f)
data.setdefault(benchmark, {}).setdefault(dataset_name, []).append(run_result)
if not os.path.isdir(collect_dir):
os.makedirs(collect_dir)
for benchmark, datasets in data.items():
benchmark_data = {}
for dataset, runs in datasets.items():
dataset_data = {}
run_values = {}
for i, run in enumerate(runs):
for k, v in run.items():
dataset_data['{}_{}'.format(k, i + 1)] = v
if k not in ['time', 'error', 'model']:
run_values.setdefault(k, []).append(v)
for k, vs in run_values.items():
dataset_data[k + '_mean'] = np.mean(vs)
dataset_data[k + '_std'] = np.std(vs)
benchmark_data[dataset] = dataset_data
df = pd.DataFrame.from_dict(benchmark_data, orient='index')
df.to_csv(os.path.join(collect_dir, benchmark + '.csv'))
def makedirs(path):
if not os.path.exists(os.path.dirname(path)):
os.makedirs(os.path.dirname(path))
def cpu_logger(pid, filename):
print("CPU logger start\n")
makedirs(filename)
process = psutil.Process(pid)
with open(filename, "a+") as f:
f.write(str(datetime.datetime.now()) + '\n')
for index in range(1, 360000):
message = f'{index}, memory_percent: {process.memory_percent()}, ' \
f'cpu_percent: {psutil.cpu_percent(percpu=False)}'
f.write(message + "\n")
t.sleep(1)
def mem_logger(pid, filename):
print("Memory logger start\n")
makedirs(filename)
with open(filename, "a") as f:
f.write(str(datetime.datetime.now()) + '\n')
memory_profiler.memory_usage(pid, interval=1, timeout=360000, retval=False, stream=f)
def logger(pid):
try:
threading.Thread(target=cpu_logger, args=(pid, os.path.join(resource_log_dir, 'cpu.log')), daemon=True).start()
threading.Thread(target=mem_logger, args=(pid, os.path.join(resource_log_dir, 'mem.log')), daemon=True).start()
except:
print("Error while running the logger service")
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('input_dir', help='Datasets directory')
parser.add_argument('output_dir', help='Benchmark results directory')
args = parser.parse_args()
input_dir, output_dir = args.input_dir, args.output_dir
tmp_splits = '/home/Eldeeb/AutoMLBenchmarking/splits'
tmp_output = '/home/Eldeeb/AutoMLBenchmarking/output'
datasets = [file for file in os.listdir(input_dir) if file.endswith('.csv')]
runs = list(range(1, n_runs + 1))
if not datasets:
print('No datasets supplied')
else:
if not os.path.isdir(tmp_splits):
os.makedirs(tmp_splits)
for dataset in datasets:
split(os.path.join(input_dir, dataset), tmp_splits, 0.75)
params = [(model, dataset, run) for model in models for dataset in datasets for run in runs]
completed = True
for model, dataset, run in params:
dataset_name = os.path.splitext(dataset)[0]
full_output_dir = os.path.join(tmp_output, model, dataset_name)
output_file = os.path.join(full_output_dir, '{}.json'.format(run))
if not os.path.isfile(output_file):
train_file = os.path.join(tmp_splits, f'{dataset_name}_train.csv')
test_file = os.path.join(tmp_splits, f'{dataset_name}_test.csv')
if not os.path.isdir(full_output_dir):
os.makedirs(full_output_dir)
open(output_file, 'w').close()
print('{} <{}> <{}> run {} start'.format(datetime.datetime.now(), model, dataset, run))
benchmark(model, train_file, test_file, output_file)
print('{} <{}> <{}> run {} end'.format(datetime.datetime.now(), model, dataset, run))
completed = False
break
if completed:
collect(tmp_output, output_dir)
print('{} completed!'.format(datetime.datetime.now()))
shutil.rmtree(tmp_splits)
shutil.rmtree(tmp_output)
else:
print('{} rebooting...'.format(datetime.datetime.now()))
os.system('reboot')