program.py 19.5 KB
Newer Older
M
refine  
MissPenguin 已提交
1
# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
L
LDOUBLEV 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

W
WenmuZhou 已提交
19
import os
L
LDOUBLEV 已提交
20
import sys
21
import platform
L
LDOUBLEV 已提交
22 23
import yaml
import time
24
import datetime
W
WenmuZhou 已提交
25 26 27 28 29
import paddle
import paddle.distributed as dist
from tqdm import tqdm
from argparse import ArgumentParser, RawDescriptionHelpFormatter

L
LDOUBLEV 已提交
30 31
from ppocr.utils.stats import TrainingStats
from ppocr.utils.save_load import save_model
32
from ppocr.utils.utility import print_dict, AverageMeter
D
dyning 已提交
33
from ppocr.utils.logging import get_logger
L
LDOUBLEV 已提交
34
from ppocr.utils import profiler
D
dyning 已提交
35
from ppocr.data import build_dataloader
L
LDOUBLEV 已提交
36

D
dyning 已提交
37

L
LDOUBLEV 已提交
38 39 40 41 42 43 44
class ArgsParser(ArgumentParser):
    def __init__(self):
        super(ArgsParser, self).__init__(
            formatter_class=RawDescriptionHelpFormatter)
        self.add_argument("-c", "--config", help="configuration file to use")
        self.add_argument(
            "-o", "--opt", nargs='+', help="set configuration options")
L
LDOUBLEV 已提交
45 46 47 48 49
        self.add_argument(
            '-p',
            '--profiler_options',
            type=str,
            default=None,
50 51
            help='The option of profiler, which should be in format ' \
                 '\"key1=value1;key2=value2;key3=value3\".'
L
LDOUBLEV 已提交
52
        )
L
LDOUBLEV 已提交
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

    def parse_args(self, argv=None):
        args = super(ArgsParser, self).parse_args(argv)
        assert args.config is not None, \
            "Please specify --config=configure_file_path."
        args.opt = self._parse_opt(args.opt)
        return args

    def _parse_opt(self, opts):
        config = {}
        if not opts:
            return config
        for s in opts:
            s = s.strip()
            k, v = s.split('=')
            config[k] = yaml.load(v, Loader=yaml.Loader)
        return config


def load_config(file_path):
    """
    Load config from yml/yaml file.
    Args:
        file_path (str): Path of the config file to be loaded.
    Returns: global config
    """
    _, ext = os.path.splitext(file_path)
    assert ext in ['.yml', '.yaml'], "only support yaml files for now"
81 82
    config = yaml.load(open(file_path, 'rb'), Loader=yaml.Loader)
    return config
L
LDOUBLEV 已提交
83 84


85
def merge_config(config, opts):
L
LDOUBLEV 已提交
86 87 88 89 90 91
    """
    Merge config into global config.
    Args:
        config (dict): Config to be merged.
    Returns: global config
    """
92
    for key, value in opts.items():
L
LDOUBLEV 已提交
93
        if "." not in key:
94 95
            if isinstance(value, dict) and key in config:
                config[key].update(value)
L
LDOUBLEV 已提交
96
            else:
97
                config[key] = value
L
LDOUBLEV 已提交
98 99
        else:
            sub_keys = key.split('.')
T
tink2123 已提交
100
            assert (
101
                sub_keys[0] in config
102 103
            ), "the sub_keys can only be one of global_config: {}, but get: " \
               "{}, please check your running command".format(
104 105
                config.keys(), sub_keys[0])
            cur = config[sub_keys[0]]
L
LDOUBLEV 已提交
106 107 108 109 110
            for idx, sub_key in enumerate(sub_keys[1:]):
                if idx == len(sub_keys) - 2:
                    cur[sub_key] = value
                else:
                    cur = cur[sub_key]
111
    return config
L
LDOUBLEV 已提交
112 113 114 115 116 117 118 119 120 121 122 123 124 125


def check_gpu(use_gpu):
    """
    Log error and exit when set use_gpu=true in paddlepaddle
    cpu version.
    """
    err = "Config use_gpu cannot be set as true while you are " \
          "using paddlepaddle cpu version ! \nPlease try: \n" \
          "\t1. Install paddlepaddle-gpu to run model on GPU \n" \
          "\t2. Set use_gpu as false in config file to run " \
          "model on CPU"

    try:
W
WenmuZhou 已提交
126
        if use_gpu and not paddle.is_compiled_with_cuda():
W
WenmuZhou 已提交
127
            print(err)
L
LDOUBLEV 已提交
128 129 130 131 132
            sys.exit(1)
    except Exception as e:
        pass


W
WenmuZhou 已提交
133
def train(config,
D
dyning 已提交
134 135 136
          train_dataloader,
          valid_dataloader,
          device,
W
WenmuZhou 已提交
137 138 139 140 141 142 143 144
          model,
          loss_class,
          optimizer,
          lr_scheduler,
          post_process_class,
          eval_class,
          pre_best_model_dict,
          logger,
S
stephon 已提交
145 146
          vdl_writer=None,
          scaler=None):
W
WenmuZhou 已提交
147 148
    cal_metric_during_train = config['Global'].get('cal_metric_during_train',
                                                   False)
149
    calc_epoch_interval = config['Global'].get('calc_epoch_interval', 1)
L
LDOUBLEV 已提交
150 151 152 153
    log_smooth_window = config['Global']['log_smooth_window']
    epoch_num = config['Global']['epoch_num']
    print_batch_step = config['Global']['print_batch_step']
    eval_batch_step = config['Global']['eval_batch_step']
L
LDOUBLEV 已提交
154
    profiler_options = config['profiler_options']
W
WenmuZhou 已提交
155

D
dyning 已提交
156
    global_step = 0
157 158
    if 'global_step' in pre_best_model_dict:
        global_step = pre_best_model_dict['global_step']
L
LDOUBLEV 已提交
159 160 161 162
    start_eval_step = 0
    if type(eval_batch_step) == list and len(eval_batch_step) >= 2:
        start_eval_step = eval_batch_step[0]
        eval_batch_step = eval_batch_step[1]
W
WenmuZhou 已提交
163 164
        if len(valid_dataloader) == 0:
            logger.info(
165 166
                'No Images in eval dataset, evaluation during training ' \
                'will be disabled'
W
WenmuZhou 已提交
167 168
            )
            start_eval_step = 1e111
L
LDOUBLEV 已提交
169
        logger.info(
170 171
            "During the training process, after the {}th iteration, " \
            "an evaluation is run every {} iterations".
L
LDOUBLEV 已提交
172
            format(start_eval_step, eval_batch_step))
L
LDOUBLEV 已提交
173 174
    save_epoch_step = config['Global']['save_epoch_step']
    save_model_dir = config['Global']['save_model_dir']
175 176
    if not os.path.exists(save_model_dir):
        os.makedirs(save_model_dir)
W
WenmuZhou 已提交
177 178 179 180
    main_indicator = eval_class.main_indicator
    best_model_dict = {main_indicator: 0}
    best_model_dict.update(pre_best_model_dict)
    train_stats = TrainingStats(log_smooth_window, ['lr'])
T
tink2123 已提交
181
    model_average = False
W
WenmuZhou 已提交
182 183
    model.train()

T
tink2123 已提交
184
    use_srn = config['Architecture']['algorithm'] == "SRN"
T
tink2123 已提交
185
    extra_input = config['Architecture'][
L
LDOUBLEV 已提交
186
        'algorithm'] in ["SRN", "NRTR", "SAR", "SEED"]
187
    try:
L
fix bug  
LDOUBLEV 已提交
188
        model_type = config['Architecture']['model_type']
189
    except:
L
fix bug  
LDOUBLEV 已提交
190
        model_type = None
T
tink2123 已提交
191
    algorithm = config['Architecture']['algorithm']
T
tink2123 已提交
192

193 194 195 196
    start_epoch = best_model_dict[
        'start_epoch'] if 'start_epoch' in best_model_dict else 1

    total_samples = 0
197 198
    train_reader_cost = 0.0
    train_batch_cost = 0.0
199
    reader_start = time.time()
200
    eta_meter = AverageMeter()
201 202 203

    max_iter = len(train_dataloader) - 1 if platform.system(
    ) == "Windows" else len(train_dataloader)
W
WenmuZhou 已提交
204

T
tink2123 已提交
205
    for epoch in range(start_epoch, epoch_num + 1):
206 207 208 209 210
        if train_dataloader.dataset.need_reset:
            train_dataloader = build_dataloader(
                config, 'Train', device, logger, seed=epoch)
            max_iter = len(train_dataloader) - 1 if platform.system(
            ) == "Windows" else len(train_dataloader)
W
WenmuZhou 已提交
211
        for idx, batch in enumerate(train_dataloader):
L
LDOUBLEV 已提交
212
            profiler.add_profiler_step(profiler_options)
文幕地方's avatar
文幕地方 已提交
213
            train_reader_cost += time.time() - reader_start
J
Jane-Ding 已提交
214
            if idx >= max_iter:
W
WenmuZhou 已提交
215 216 217
                break
            lr = optimizer.get_lr()
            images = batch[0]
T
tink2123 已提交
218
            if use_srn:
T
tink2123 已提交
219
                model_average = True
S
stephon 已提交
220 221 222 223 224 225 226 227

            # use amp
            if scaler:
                with paddle.amp.auto_cast():
                    if model_type == 'table' or extra_input:
                        preds = model(images, data=batch[1:])
                    else:
                        preds = model(images)
T
tink2123 已提交
228
            else:
S
stephon 已提交
229 230
                if model_type == 'table' or extra_input:
                    preds = model(images, data=batch[1:])
231
                elif model_type in ["kie", 'vqa']:
L
LDOUBLEV 已提交
232
                    preds = model(batch)
S
stephon 已提交
233 234
                else:
                    preds = model(images)
235

W
WenmuZhou 已提交
236 237
            loss = loss_class(preds, batch)
            avg_loss = loss['loss']
S
stephon 已提交
238 239 240 241 242 243 244 245

            if scaler:
                scaled_avg_loss = scaler.scale(avg_loss)
                scaled_avg_loss.backward()
                scaler.minimize(optimizer, scaled_avg_loss)
            else:
                avg_loss.backward()
                optimizer.step()
W
WenmuZhou 已提交
246
            optimizer.clear_grad()
W
WenmuZhou 已提交
247

248 249 250 251 252 253 254 255 256 257
            if cal_metric_during_train and epoch % calc_epoch_interval == 0:  # only rec and cls need
                batch = [item.numpy() for item in batch]
                if model_type in ['table', 'kie']:
                    eval_class(preds, batch)
                else:
                    post_result = post_process_class(preds, batch[1])
                    eval_class(post_result, batch)
                metric = eval_class.get_metric()
                train_stats.update(metric)

258 259 260
            train_batch_time = time.time() - reader_start
            train_batch_cost += train_batch_time
            eta_meter.update(train_batch_time)
261
            global_step += 1
文幕地方's avatar
文幕地方 已提交
262
            total_samples += len(images)
W
WenmuZhou 已提交
263

D
dyning 已提交
264 265
            if not isinstance(lr_scheduler, float):
                lr_scheduler.step()
W
WenmuZhou 已提交
266 267 268 269 270 271 272 273 274 275 276

            # logger and visualdl
            stats = {k: v.numpy().mean() for k, v in loss.items()}
            stats['lr'] = lr
            train_stats.update(stats)

            if vdl_writer is not None and dist.get_rank() == 0:
                for k, v in train_stats.get().items():
                    vdl_writer.add_scalar('TRAIN/{}'.format(k), v, global_step)
                vdl_writer.add_scalar('TRAIN/lr', lr, global_step)

277 278 279
            if dist.get_rank() == 0 and (
                (global_step > 0 and global_step % print_batch_step == 0) or
                (idx >= len(train_dataloader) - 1)):
W
WenmuZhou 已提交
280
                logs = train_stats.log()
L
LDOUBLEV 已提交
281

282 283 284 285 286
                eta_sec = ((epoch_num + 1 - epoch) * \
                    len(train_dataloader) - idx - 1) * eta_meter.avg
                eta_sec_format = str(datetime.timedelta(seconds=int(eta_sec)))
                strs = 'epoch: [{}/{}], global_step: {}, {}, avg_reader_cost: ' \
                       '{:.5f} s, avg_batch_cost: {:.5f} s, avg_samples: {}, ' \
L
LDOUBLEV 已提交
287
                       'ips: {:.5f} samples/s, eta: {}'.format(
288 289 290 291 292
                    epoch, epoch_num, global_step, logs,
                    train_reader_cost / print_batch_step,
                    train_batch_cost / print_batch_step,
                    total_samples / print_batch_step,
                    total_samples / train_batch_cost, eta_sec_format)
W
WenmuZhou 已提交
293
                logger.info(strs)
294

文幕地方's avatar
文幕地方 已提交
295
                total_samples = 0
296 297
                train_reader_cost = 0.0
                train_batch_cost = 0.0
W
WenmuZhou 已提交
298 299
            # eval
            if global_step > start_eval_step and \
300 301
                    (global_step - start_eval_step) % eval_batch_step == 0 \
                    and dist.get_rank() == 0:
T
tink2123 已提交
302 303 304 305 306 307 308
                if model_average:
                    Model_Average = paddle.incubate.optimizer.ModelAverage(
                        0.15,
                        parameters=model.parameters(),
                        min_average_window=10000,
                        max_average_window=15625)
                    Model_Average.apply()
T
tink2123 已提交
309 310 311 312 313
                cur_metric = eval(
                    model,
                    valid_dataloader,
                    post_process_class,
                    eval_class,
M
refine  
MissPenguin 已提交
314
                    model_type,
T
tink2123 已提交
315
                    extra_input=extra_input)
L
LDOUBLEV 已提交
316 317 318
                cur_metric_str = 'cur metric, {}'.format(', '.join(
                    ['{}: {}'.format(k, v) for k, v in cur_metric.items()]))
                logger.info(cur_metric_str)
W
WenmuZhou 已提交
319 320 321

                # logger metric
                if vdl_writer is not None:
L
LDOUBLEV 已提交
322
                    for k, v in cur_metric.items():
W
WenmuZhou 已提交
323 324
                        if isinstance(v, (float, int)):
                            vdl_writer.add_scalar('EVAL/{}'.format(k),
L
LDOUBLEV 已提交
325 326
                                                  cur_metric[k], global_step)
                if cur_metric[main_indicator] >= best_model_dict[
W
WenmuZhou 已提交
327
                        main_indicator]:
L
LDOUBLEV 已提交
328
                    best_model_dict.update(cur_metric)
W
WenmuZhou 已提交
329 330 331 332 333 334
                    best_model_dict['best_epoch'] = epoch
                    save_model(
                        model,
                        optimizer,
                        save_model_dir,
                        logger,
335
                        config,
W
WenmuZhou 已提交
336 337 338
                        is_best=True,
                        prefix='best_accuracy',
                        best_model_dict=best_model_dict,
339 340
                        epoch=epoch,
                        global_step=global_step)
L
LDOUBLEV 已提交
341
                best_str = 'best metric, {}'.format(', '.join([
W
WenmuZhou 已提交
342 343 344 345 346 347 348 349
                    '{}: {}'.format(k, v) for k, v in best_model_dict.items()
                ]))
                logger.info(best_str)
                # logger best metric
                if vdl_writer is not None:
                    vdl_writer.add_scalar('EVAL/best_{}'.format(main_indicator),
                                          best_model_dict[main_indicator],
                                          global_step)
350

文幕地方's avatar
文幕地方 已提交
351
            reader_start = time.time()
W
WenmuZhou 已提交
352 353 354 355 356 357
        if dist.get_rank() == 0:
            save_model(
                model,
                optimizer,
                save_model_dir,
                logger,
358
                config,
W
WenmuZhou 已提交
359 360 361
                is_best=False,
                prefix='latest',
                best_model_dict=best_model_dict,
362 363
                epoch=epoch,
                global_step=global_step)
W
WenmuZhou 已提交
364 365 366 367 368 369
        if dist.get_rank() == 0 and epoch > 0 and epoch % save_epoch_step == 0:
            save_model(
                model,
                optimizer,
                save_model_dir,
                logger,
370
                config,
W
WenmuZhou 已提交
371 372 373
                is_best=False,
                prefix='iter_epoch_{}'.format(epoch),
                best_model_dict=best_model_dict,
374 375
                epoch=epoch,
                global_step=global_step)
L
LDOUBLEV 已提交
376
    best_str = 'best metric, {}'.format(', '.join(
W
WenmuZhou 已提交
377 378 379 380
        ['{}: {}'.format(k, v) for k, v in best_model_dict.items()]))
    logger.info(best_str)
    if dist.get_rank() == 0 and vdl_writer is not None:
        vdl_writer.close()
L
LDOUBLEV 已提交
381 382 383
    return


M
refine  
MissPenguin 已提交
384 385 386 387
def eval(model,
         valid_dataloader,
         post_process_class,
         eval_class,
L
LDOUBLEV 已提交
388
         model_type=None,
T
tink2123 已提交
389
         extra_input=False):
W
WenmuZhou 已提交
390 391 392 393
    model.eval()
    with paddle.no_grad():
        total_frame = 0.0
        total_time = 0.0
文幕地方's avatar
文幕地方 已提交
394 395 396 397 398
        pbar = tqdm(
            total=len(valid_dataloader),
            desc='eval model:',
            position=0,
            leave=True)
399 400
        max_iter = len(valid_dataloader) - 1 if platform.system(
        ) == "Windows" else len(valid_dataloader)
W
WenmuZhou 已提交
401
        for idx, batch in enumerate(valid_dataloader):
402
            if idx >= max_iter:
W
WenmuZhou 已提交
403
                break
W
fix bug  
WenmuZhou 已提交
404
            images = batch[0]
W
WenmuZhou 已提交
405
            start = time.time()
T
tink2123 已提交
406
            if model_type == 'table' or extra_input:
M
refine  
MissPenguin 已提交
407
                preds = model(images, data=batch[1:])
408
            elif model_type in ["kie", 'vqa']:
L
LDOUBLEV 已提交
409
                preds = model(batch)
X
xiaoting 已提交
410
            else:
L
LDOUBLEV 已提交
411
                preds = model(images)
412 413 414 415 416 417 418

            batch_numpy = []
            for item in batch:
                if isinstance(item, paddle.Tensor):
                    batch_numpy.append(item.numpy())
                else:
                    batch_numpy.append(item)
W
WenmuZhou 已提交
419 420 421
            # Obtain usable results from post-processing methods
            total_time += time.time() - start
            # Evaluate the results of the current batch
L
LDOUBLEV 已提交
422
            if model_type in ['table', 'kie']:
423 424 425 426
                eval_class(preds, batch_numpy)
            elif model_type in ['vqa']:
                post_result = post_process_class(preds, batch_numpy)
                eval_class(post_result, batch_numpy)
M
MissPenguin 已提交
427
            else:
428 429
                post_result = post_process_class(preds, batch_numpy[1])
                eval_class(post_result, batch_numpy)
L
LDOUBLEV 已提交
430

W
fix bug  
WenmuZhou 已提交
431
            pbar.update(1)
W
WenmuZhou 已提交
432
            total_frame += len(images)
L
LDOUBLEV 已提交
433 434
        # Get final metric,eg. acc or hmean
        metric = eval_class.get_metric()
D
dyning 已提交
435

W
fix bug  
WenmuZhou 已提交
436
    pbar.close()
W
WenmuZhou 已提交
437
    model.train()
L
LDOUBLEV 已提交
438 439
    metric['fps'] = total_frame / total_time
    return metric
L
licx 已提交
440

T
tink2123 已提交
441

B
Bin Lu 已提交
442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490
def update_center(char_center, post_result, preds):
    result, label = post_result
    feats, logits = preds
    logits = paddle.argmax(logits, axis=-1)
    feats = feats.numpy()
    logits = logits.numpy()

    for idx_sample in range(len(label)):
        if result[idx_sample][0] == label[idx_sample][0]:
            feat = feats[idx_sample]
            logit = logits[idx_sample]
            for idx_time in range(len(logit)):
                index = logit[idx_time]
                if index in char_center.keys():
                    char_center[index][0] = (
                        char_center[index][0] * char_center[index][1] +
                        feat[idx_time]) / (char_center[index][1] + 1)
                    char_center[index][1] += 1
                else:
                    char_center[index] = [feat[idx_time], 1]
    return char_center


def get_center(model, eval_dataloader, post_process_class):
    pbar = tqdm(total=len(eval_dataloader), desc='get center:')
    max_iter = len(eval_dataloader) - 1 if platform.system(
    ) == "Windows" else len(eval_dataloader)
    char_center = dict()
    for idx, batch in enumerate(eval_dataloader):
        if idx >= max_iter:
            break
        images = batch[0]
        start = time.time()
        preds = model(images)

        batch = [item.numpy() for item in batch]
        # Obtain usable results from post-processing methods
        post_result = post_process_class(preds, batch[1])

        #update char_center
        char_center = update_center(char_center, post_result, preds)
        pbar.update(1)

    pbar.close()
    for key in char_center.keys():
        char_center[key] = char_center[key][0]
    return char_center


491
def preprocess(is_train=False):
L
licx 已提交
492
    FLAGS = ArgsParser().parse_args()
L
LDOUBLEV 已提交
493
    profiler_options = FLAGS.profiler_options
L
licx 已提交
494
    config = load_config(FLAGS.config)
495
    config = merge_config(config, FLAGS.opt)
L
LDOUBLEV 已提交
496
    profile_dic = {"profiler_options": FLAGS.profiler_options}
497
    config = merge_config(config, profile_dic)
L
licx 已提交
498

W
WenmuZhou 已提交
499 500 501 502 503 504 505 506 507 508 509
    if is_train:
        # save_config
        save_model_dir = config['Global']['save_model_dir']
        os.makedirs(save_model_dir, exist_ok=True)
        with open(os.path.join(save_model_dir, 'config.yml'), 'w') as f:
            yaml.dump(
                dict(config), f, default_flow_style=False, sort_keys=False)
        log_file = '{}/train.log'.format(save_model_dir)
    else:
        log_file = None
    logger = get_logger(name='root', log_file=log_file)
L
licx 已提交
510 511 512 513 514

    # check if set use_gpu=True in paddlepaddle cpu version
    use_gpu = config['Global']['use_gpu']
    check_gpu(use_gpu)

W
WenmuZhou 已提交
515 516
    alg = config['Architecture']['algorithm']
    assert alg in [
J
Jethong 已提交
517
        'EAST', 'DB', 'SAST', 'Rosetta', 'CRNN', 'STARNet', 'RARE', 'SRN',
T
tink2123 已提交
518
        'CLS', 'PGNet', 'Distillation', 'NRTR', 'TableAttn', 'SAR', 'PSE',
519
        'SEED', 'SDMGR', 'LayoutXLM', 'LayoutLM'
W
WenmuZhou 已提交
520
    ]
L
licx 已提交
521

W
WenmuZhou 已提交
522 523
    device = 'gpu:{}'.format(dist.ParallelEnv().dev_id) if use_gpu else 'cpu'
    device = paddle.set_device(device)
D
dyning 已提交
524

D
dyning 已提交
525
    config['Global']['distributed'] = dist.get_world_size() != 1
W
WenmuZhou 已提交
526

littletomatodonkey's avatar
littletomatodonkey 已提交
527
    if config['Global']['use_visualdl'] and dist.get_rank() == 0:
D
dyning 已提交
528
        from visualdl import LogWriter
L
fix bug  
LDOUBLEV 已提交
529
        save_model_dir = config['Global']['save_model_dir']
D
dyning 已提交
530 531 532 533 534 535 536 537 538
        vdl_writer_path = '{}/vdl/'.format(save_model_dir)
        os.makedirs(vdl_writer_path, exist_ok=True)
        vdl_writer = LogWriter(logdir=vdl_writer_path)
    else:
        vdl_writer = None
    print_dict(config, logger)
    logger.info('train with paddle {} and device {}'.format(paddle.__version__,
                                                            device))
    return config, device, logger, vdl_writer