File size: 11,608 Bytes
52da96f |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 |
import abc
import os
import time
import sys
from tqdm import tqdm
from math import ceil
class MultipleProcessRunner:
"""
Abstarct class for running tasks with multiple process
There are three abstract methods that should be implemented:
1. __len__() : return the length of data
2. _target() : target function for each process
3. _aggregate() : aggregate results from each process
"""
def __init__(self,
data,
save_path=None,
n_process=1,
verbose=True,
total_only=True,
log_step=1,
start_method='fork'):
"""
Args:
data : data to be processed that can be sliced
path : final output path
n_process: number of process
verbose : if True, display progress bar
total_only: If True, only total progress bar is displayed
log_step : For total progress bar, Next log will be printed when
``current iteration`` - ``last log iteration`` >= log_step
start_method: start method for multiprocessing
"""
self.data = data
self.save_path = save_path
self.n_process = n_process
self.verbose = verbose
self.total_only = total_only
self.log_step = log_step
self.start_method = start_method
# get terminal width to format output
try:
self.terminal_y = os.get_terminal_size()[0]
except Exception as e:
print(e)
print("Can't get terminal size, set terminal_y = None")
self.terminal_y = None
def _s2hms(self, seconds: float):
"""
convert second format of time into hour:minute:second format
"""
m, s = divmod(seconds, 60)
h, m = divmod(m, 60)
return "%02d:%02d:%02d" % (h, m, s)
def _display_time(self, st_time, now, total):
ed_time = time.time()
running_time = ed_time - st_time
rest_time = running_time * (total - now) / now
iter_sec = f"{now / running_time:.2f}it/s" if now > running_time else f"{running_time / now:.2f}s/it"
return f' [{self._s2hms(running_time)} < {self._s2hms(rest_time)}, {iter_sec}]'
def _display_bar(self, now, total, length):
now = now if now <= total else total
num = now * length // total
progress_bar = '[' + '#' * num + '_' * (length - num) + ']'
return progress_bar
def _display_all(self, now, total, desc, st_time):
# make a progress bar
length = 50
progress_bar = self._display_bar(now, total, length)
time_display = self._display_time(st_time, now, total)
display = f'{desc}{progress_bar} {int(now / total * 100):02d}% {now}/{total}{time_display}'
# Clean a line
width = self.terminal_y if self.terminal_y is not None else 100
num_space = width - len(display)
if num_space > 0:
display += ' ' * num_space
else:
length += num_space
progress_bar = self._display_bar(now, total, length)
display = f'{desc}{progress_bar} {int(now / total * 100):02d}% {now}/{total}{time_display}'
# Set color
display = f"\033[31m{display}\033[0m"
return display
# Print progress bar at specific position in terminal
def terminal_progress_bar(self,
process_id: int,
now: int,
total: int,
desc: str = ''):
"""
Args:
process_id: process id
now: now iteration number
total: total iteration number
desc: description
"""
st_time = self.process_st_time[process_id]
# Aggregate total information
self.counts[process_id] = now
self._total_display(self.process_st_time["total"])
if not self.total_only:
process_display = self._display_all(now, total, desc, st_time)
if self.terminal_y is not None:
sys.stdout.write(f"\x1b7\x1b[{process_id + 1};{0}f{process_display}\x1b8")
sys.stdout.flush()
else:
print(f"\x1b7\x1b[{process_id + 1};{0}f{process_display}\x1b8", flush=True)
# Print global information
def _total_display(self, st_time):
if self.total_display_callable.value == 1:
self.total_display_callable.value = 0
cnt = sum([self.counts[i] for i in range(self.n_process)])
if cnt - self.last_cnt.value >= self.log_step:
total_display = self._display_all(cnt, self.__len__(), f"Total: ", st_time)
self.last_cnt.value = cnt
x = self.n_process + 1 if not self.total_only else 0
# if self.terminal_y is not None:
# sys.stdout.write(f"\x1b7\x1b[{x};{0}f{total_display}\x1b8")
# sys.stdout.flush()
# else:
# print(f"\x1b7\x1b[{x};{0}f{total_display}\x1b8", flush=True)
print(f"\r\x1b7\x1b[{x};{0}f{total_display}\x1b8", flush=True, end="")
self.total_display_callable.value = 1
def run(self):
"""
The function is used to run a multi-process task
Returns: return the result of function '_aggregate()'
"""
import multiprocess as mp
mp.set_start_method(self.start_method, force=True)
# total number of data that is already processed
self.counts = mp.Manager().dict({i: 0 for i in range(self.n_process)})
# record start time for each process
self.process_st_time = {"total": time.time()}
# set a lock to call total number display
self.total_display_callable = mp.Value('d', 1)
# Save last log iteration number
self.last_cnt = mp.Value('d', 0)
num_per_process = ceil(self.__len__() / self.n_process)
if self.save_path is not None:
file_name, suffix = os.path.splitext(self.save_path)
process_list = []
sub_paths = []
for i in range(self.n_process):
st = i * num_per_process
ed = st + num_per_process
# construct slice and sub path for sub process
data_slice = self.data[st: ed]
sub_path = None
# Create a directory to save sub-results
if self.save_path is not None:
save_dir = f"{file_name}{suffix}_temp"
os.makedirs(save_dir, exist_ok=True)
sub_path = f"{save_dir}/temp_{i}{suffix}"
# construct sub process
input_args = (i, data_slice, sub_path)
self.process_st_time[i] = time.time()
p = mp.Process(target=self._target, args=input_args)
p.start()
process_list.append(p)
sub_paths.append(sub_path)
for p in process_list:
p.join()
# aggregate results and remove temporary directory
results = self._aggregate(self.save_path, sub_paths)
if self.save_path is not None:
save_dir = f"{file_name}{suffix}_temp"
os.rmdir(save_dir)
return results
def parallel_run(self):
import multiprocess as mp
from joblib import Parallel, delayed
# total number of data that is already processed
self.counts = mp.Manager().dict({i: 0 for i in range(self.n_process)})
# record start time for each process
self.process_st_time = {"total": time.time()}
# set a lock to call total number display
self.total_display_callable = mp.Value('d', 1)
# Save last log iteration number
self.last_cnt = mp.Value('d', 0)
num_per_process = ceil(self.__len__() / self.n_process)
if self.save_path is not None:
file_name, suffix = os.path.splitext(self.save_path)
sub_paths = []
input_arg_list = []
for i in range(self.n_process):
st = i * num_per_process
ed = st + num_per_process
# construct slice and sub path for sub process
data_slice = self.data[st: ed]
sub_path = None
# Create a directory to save sub-results
if self.save_path is not None:
save_dir = f"{file_name}{suffix}_temp"
os.makedirs(save_dir, exist_ok=True)
sub_path = f"{save_dir}/temp_{i}{suffix}"
# construct sub process
input_args = (i, data_slice, sub_path)
self.process_st_time[i] = time.time()
sub_paths.append(sub_path)
input_arg_list.append(input_args)
# Start parallel processing
Parallel(n_jobs=self.n_process)(delayed(self._target)(input_args) for input_args in input_arg_list)
# aggregate results and remove temporary directory
results = self._aggregate(self.save_path, sub_paths)
if self.save_path is not None:
save_dir = f"{file_name}{suffix}_temp"
os.rmdir(save_dir)
return results
@abc.abstractmethod
def _aggregate(self, final_path: str, sub_paths):
"""
This function is used to aggregate results from sub processes into a file
Args:
final_path: path to save final results
sub_paths : list of sub paths
Returns: None or desirable results specified by user
"""
raise NotImplementedError
@abc.abstractmethod
def _target(self, process_id, data, sub_path):
"""
The main body to operate data in one process
Args:
i : process id
data : data slice
sub_path: sub path to save results
"""
raise NotImplementedError
@abc.abstractmethod
def __len__(self):
raise NotImplementedError
class MultipleProcessRunnerSimplifier(MultipleProcessRunner):
"""
A simplified version of MultipleProcessRunner.
User only need to implement the function 'do', then it will be automatically executed
in every iteration after call the function 'run'.
If 'save_path' is specified, it will open a file in the 'sub_path' into which
user can write results, and results will be aggregated into 'save_path'.
The procedure would be like:
...
with open(sub_path, 'w') as w:
for i, d in enumerate(data):
self.do(process_id, i, d, w) # You can write results into the file.
...
The 'do' function should be like:
def do(process_id, idx, data, writer):
...
If 'save_path' is None, the argument 'writer' will be set to None.
"""
def __init__(self,
data,
do,
save_path=None,
n_process=1,
verbose=True,
total_only=True,
log_step=1,
return_results=False,
start_method='fork'):
super().__init__(data=data,
save_path=save_path,
n_process=n_process,
verbose=verbose,
total_only=total_only,
log_step=log_step,
start_method=start_method)
self.do = do
self.return_results = return_results
def run(self):
self.start_time = time.time()
return super().run()
def _aggregate(self, final_path: str, sub_paths):
results = []
w = open(final_path, 'w') if final_path is not None else None
if self.verbose:
iterator = tqdm(enumerate(sub_paths), "Aggregating results...")
else:
iterator = enumerate(sub_paths)
for i, sub_path in iterator:
if sub_path is None and self.return_results:
sub_path = f"MultipleProcessRunnerSimplifier_{self.start_time}_{i}.tmp"
if sub_path is not None:
with open(sub_path, 'r') as r:
for line in r:
if w is not None:
w.write(line)
if self.return_results:
results.append(line[:-1])
os.remove(sub_path)
return results
def _target(self, process_id, data, sub_path):
if sub_path is None and self.return_results:
sub_path = f"MultipleProcessRunnerSimplifier_{self.start_time}_{process_id}.tmp"
w = open(sub_path, 'w') if sub_path is not None else None
for i, d in enumerate(data):
self.do(process_id, i, d, w)
if self.verbose:
self.terminal_progress_bar(process_id, i + 1, len(data), f"Process{process_id} running...")
if w is not None:
w.close()
def __len__(self):
return len(self.data)
|