trainer.py 17.4 KB
Newer Older
Q
qingqing01 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
#   Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# 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.

L
LielinJiang 已提交
15 16
import os
import time
L
LielinJiang 已提交
17
import copy
L
LielinJiang 已提交
18

L
LielinJiang 已提交
19
import logging
L
LielinJiang 已提交
20
import datetime
L
LielinJiang 已提交
21

L
LielinJiang 已提交
22
import paddle
L
LielinJiang 已提交
23
from paddle.distributed import ParallelEnv
L
LielinJiang 已提交
24 25 26 27

from ..datasets.builder import build_dataloader
from ..models.builder import build_model
from ..utils.visual import tensor2img, save_image
L
LielinJiang 已提交
28
from ..utils.filesystem import makedirs, save, load
29
from ..utils.timer import TimeAverager
L
lzzyzlbb 已提交
30
from ..utils.profiler import add_profiler_step
L
fix nan  
LielinJiang 已提交
31

32

33 34 35 36 37
class IterLoader:
    def __init__(self, dataloader):
        self._dataloader = dataloader
        self.iter_loader = iter(self._dataloader)
        self._epoch = 1
L
LielinJiang 已提交
38

39 40 41 42 43 44 45 46 47 48 49
    @property
    def epoch(self):
        return self._epoch

    def __next__(self):
        try:
            data = next(self.iter_loader)
        except StopIteration:
            self._epoch += 1
            self.iter_loader = iter(self._dataloader)
            data = next(self.iter_loader)
L
LielinJiang 已提交
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
        return data

    def __len__(self):
        return len(self._dataloader)


class Trainer:
    """
    # trainer calling logic:
    #
    #                build_model                               ||    model(BaseModel)
    #                     |                                    ||
    #               build_dataloader                           ||    dataloader
    #                     |                                    ||
    #               model.setup_lr_schedulers                  ||    lr_scheduler
    #                     |                                    ||
    #               model.setup_optimizers                     ||    optimizers
    #                     |                                    ||
    #     train loop (model.setup_input + model.train_iter)    ||    train loop
    #                     |                                    ||
    #         print log (model.get_current_losses)             ||
    #                     |                                    ||
    #         save checkpoint (model.nets)                     \/
    """
    def __init__(self, cfg):
76 77 78 79 80 81 82
        # base config
        self.logger = logging.getLogger(__name__)
        self.cfg = cfg
        self.output_dir = cfg.output_dir
        self.max_eval_steps = cfg.model.get('max_eval_steps', None)

        self.local_rank = ParallelEnv().local_rank
83
        self.world_size = ParallelEnv().nranks
84 85 86 87 88 89 90 91 92 93
        self.log_interval = cfg.log_config.interval
        self.visual_interval = cfg.log_config.visiual_interval
        self.weight_interval = cfg.snapshot_config.interval

        self.start_epoch = 1
        self.current_epoch = 1
        self.current_iter = 1
        self.inner_iter = 1
        self.batch_id = 0
        self.global_steps = 0
L
LielinJiang 已提交
94

L
LielinJiang 已提交
95
        # build model
96
        self.model = build_model(cfg.model)
97 98 99
        # multiple gpus prepare
        if ParallelEnv().nranks > 1:
            self.distributed_data_parallel()
L
LielinJiang 已提交
100

101 102
        # build metrics
        self.metrics = None
L
LielinJiang 已提交
103
        self.is_save_img = True
104 105 106
        validate_cfg = cfg.get('validate', None)
        if validate_cfg and 'metrics' in validate_cfg:
            self.metrics = self.model.setup_metrics(validate_cfg['metrics'])
107 108
        if validate_cfg and 'save_img' in validate_cfg:
            self.is_save_img = validate_cfg['save_img']
109 110 111 112 113 114 115 116 117 118

        self.enable_visualdl = cfg.get('enable_visualdl', False)
        if self.enable_visualdl:
            import visualdl
            self.vdl_logger = visualdl.LogWriter(logdir=cfg.output_dir)

        # evaluate only
        if not cfg.is_train:
            return

119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
        # build train dataloader
        self.train_dataloader = build_dataloader(cfg.dataset.train)
        self.iters_per_epoch = len(self.train_dataloader)

        # build lr scheduler
        # TODO: has a better way?
        if 'lr_scheduler' in cfg and 'iters_per_epoch' in cfg.lr_scheduler:
            cfg.lr_scheduler.iters_per_epoch = self.iters_per_epoch
        self.lr_schedulers = self.model.setup_lr_schedulers(cfg.lr_scheduler)

        # build optimizers
        self.optimizers = self.model.setup_optimizers(self.lr_schedulers,
                                                      cfg.optimizer)

        self.epochs = cfg.get('epochs', None)
        if self.epochs:
            self.total_iters = self.epochs * self.iters_per_epoch
            self.by_epoch = True
        else:
            self.by_epoch = False
            self.total_iters = cfg.total_iters

L
LielinJiang 已提交
141 142 143
        if self.by_epoch:
            self.weight_interval *= self.iters_per_epoch

L
LielinJiang 已提交
144 145 146
        self.validate_interval = -1
        if cfg.get('validate', None) is not None:
            self.validate_interval = cfg.validate.get('interval', -1)
147 148

        self.time_count = {}
L
LielinJiang 已提交
149
        self.best_metric = {}
150
        self.model.set_total_iter(self.total_iters)
L
lzzyzlbb 已提交
151
        self.profiler_options = cfg.profiler_options
L
LielinJiang 已提交
152

153
    def distributed_data_parallel(self):
L
LielinJiang 已提交
154
        paddle.distributed.init_parallel_env()
155
        find_unused_parameters = self.cfg.get('find_unused_parameters', False)
156
        for net_name, net in self.model.nets.items():
157 158
            self.model.nets[net_name] = paddle.DataParallel(
                net, find_unused_parameters=find_unused_parameters)
159

L
LielinJiang 已提交
160 161 162 163 164 165 166 167 168 169 170
    def learning_rate_scheduler_step(self):
        if isinstance(self.model.lr_scheduler, dict):
            for lr_scheduler in self.model.lr_scheduler.values():
                lr_scheduler.step()
        elif isinstance(self.model.lr_scheduler,
                        paddle.optimizer.lr.LRScheduler):
            self.model.lr_scheduler.step()
        else:
            raise ValueError(
                'lr schedulter must be a dict or an instance of LRScheduler')

L
LielinJiang 已提交
171
    def train(self):
172 173
        reader_cost_averager = TimeAverager()
        batch_cost_averager = TimeAverager()
L
LielinJiang 已提交
174

175
        iter_loader = IterLoader(self.train_dataloader)
L
LielinJiang 已提交
176

L
LielinJiang 已提交
177 178
        # set model.is_train = True
        self.model.setup_train_mode(is_train=True)
179 180 181
        while self.current_iter < (self.total_iters + 1):
            self.current_epoch = iter_loader.epoch
            self.inner_iter = self.current_iter % self.iters_per_epoch
L
LielinJiang 已提交
182

L
lzzyzlbb 已提交
183 184
            add_profiler_step(self.profiler_options)

185 186 187 188 189 190 191 192
            start_time = step_start_time = time.time()
            data = next(iter_loader)
            reader_cost_averager.record(time.time() - step_start_time)
            # unpack data from dataset and apply preprocessing
            # data input should be dict
            self.model.setup_input(data)
            self.model.train_iter(self.optimizers)

W
wangna11BD 已提交
193 194 195
            batch_cost_averager.record(
                time.time() - step_start_time,
                num_samples=self.cfg['dataset']['train'].get('batch_size', 1))
196 197 198

            step_start_time = time.time()

199 200 201 202 203 204 205 206 207
            if self.current_iter % self.log_interval == 0:
                self.data_time = reader_cost_averager.get_average()
                self.step_time = batch_cost_averager.get_average()
                self.ips = batch_cost_averager.get_ips_average()
                self.print_log()

                reader_cost_averager.reset()
                batch_cost_averager.reset()

L
LielinJiang 已提交
208
            if self.current_iter % self.visual_interval == 0 and self.local_rank == 0:
209 210
                self.visual('visual_train')

L
LielinJiang 已提交
211
            self.learning_rate_scheduler_step()
L
LielinJiang 已提交
212

L
LielinJiang 已提交
213
            if self.validate_interval > -1 and self.current_iter % self.validate_interval == 0:
214
                self.test()
L
fix nan  
LielinJiang 已提交
215

L
LielinJiang 已提交
216 217 218
            if self.current_iter % self.weight_interval == 0:
                self.save(self.current_iter, 'weight', keep=-1)
                self.save(self.current_iter)
L
LielinJiang 已提交
219

220
            self.current_iter += 1
L
LielinJiang 已提交
221

L
LielinJiang 已提交
222 223
    def test(self):
        if not hasattr(self, 'test_dataloader'):
224
            self.test_dataloader = build_dataloader(self.cfg.dataset.test,
225
                                                    is_train=False)
L
lijianshe02 已提交
226 227 228
        iter_loader = IterLoader(self.test_dataloader)
        if self.max_eval_steps is None:
            self.max_eval_steps = len(self.test_dataloader)
229 230 231 232

        if self.metrics:
            for metric in self.metrics.values():
                metric.reset()
L
LielinJiang 已提交
233

L
LielinJiang 已提交
234 235 236
        # set model.is_train = False
        self.model.setup_train_mode(is_train=False)

L
lijianshe02 已提交
237
        for i in range(self.max_eval_steps):
238 239
            if self.max_eval_steps < self.log_interval or i % self.log_interval == 0:
                self.logger.info('Test iter: [%d/%d]' %
W
wangna11BD 已提交
240 241
                                 (i * self.world_size,
                                  self.max_eval_steps * self.world_size))
242

L
lijianshe02 已提交
243
            data = next(iter_loader)
244 245
            self.model.setup_input(data)
            self.model.test_iter(metrics=self.metrics)
L
LielinJiang 已提交
246

247 248 249 250
            if self.is_save_img:
                visual_results = {}
                current_paths = self.model.get_image_paths()
                current_visuals = self.model.get_current_visuals()
L
LielinJiang 已提交
251

252 253 254
                if len(current_visuals) > 0 and list(
                        current_visuals.values())[0].shape == 4:
                    num_samples = list(current_visuals.values())[0].shape[0]
L
LielinJiang 已提交
255
                else:
256
                    num_samples = 1
L
LielinJiang 已提交
257

258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274
                for j in range(num_samples):
                    if j < len(current_paths):
                        short_path = os.path.basename(current_paths[j])
                        basename = os.path.splitext(short_path)[0]
                    else:
                        basename = '{:04d}_{:04d}'.format(i, j)
                    for k, img_tensor in current_visuals.items():
                        name = '%s_%s' % (basename, k)
                        if len(img_tensor.shape) == 4:
                            visual_results.update({name: img_tensor[j]})
                        else:
                            visual_results.update({name: img_tensor})

                self.visual('visual_test',
                            visual_results=visual_results,
                            step=self.batch_id,
                            is_save_image=True)
L
LielinJiang 已提交
275

276 277 278 279 280
        if self.metrics:
            for metric_name, metric in self.metrics.items():
                self.logger.info("Metric {}: {:.4f}".format(
                    metric_name, metric.accumulate()))

L
LielinJiang 已提交
281 282
    def print_log(self):
        losses = self.model.get_current_losses()
L
LielinJiang 已提交
283

284 285 286 287 288 289 290 291 292
        message = ''
        if self.by_epoch:
            message += 'Epoch: %d/%d, iter: %d/%d ' % (
                self.current_epoch, self.epochs, self.inner_iter,
                self.iters_per_epoch)
        else:
            message += 'Iter: %d/%d ' % (self.current_iter, self.total_iters)

        message += f'lr: {self.current_learning_rate:.3e} '
L
LielinJiang 已提交
293 294 295

        for k, v in losses.items():
            message += '%s: %.3f ' % (k, v)
郑启航 已提交
296 297
            if self.enable_visualdl:
                self.vdl_logger.add_scalar(k, v, step=self.global_steps)
L
LielinJiang 已提交
298

299 300 301
        if hasattr(self, 'step_time'):
            message += 'batch_cost: %.5f sec ' % self.step_time

302
        if hasattr(self, 'data_time'):
303
            message += 'reader_cost: %.5f sec ' % self.data_time
304

305
        if hasattr(self, 'ips'):
L
LielinJiang 已提交
306 307 308
            message += 'ips: %.5f images/s ' % self.ips

        if hasattr(self, 'step_time'):
L
LielinJiang 已提交
309 310 311
            eta = self.step_time * (self.total_iters - self.current_iter)
            eta = eta if eta > 0 else 0

L
LielinJiang 已提交
312 313
            eta_str = str(datetime.timedelta(seconds=int(eta)))
            message += f'eta: {eta_str}'
314

L
LielinJiang 已提交
315 316 317 318 319
        # print the message
        self.logger.info(message)

    @property
    def current_learning_rate(self):
L
LielinJiang 已提交
320 321
        for optimizer in self.model.optimizers.values():
            return optimizer.get_lr()
L
LielinJiang 已提交
322

郑启航 已提交
323 324 325 326 327 328 329 330 331 332 333 334 335
    def visual(self,
               results_dir,
               visual_results=None,
               step=None,
               is_save_image=False):
        """
        visual the images, use visualdl or directly write to the directory
        Parameters:
            results_dir (str)     --  directory name which contains saved images
            visual_results (dict) --  the results images dict
            step (int)            --  global steps, used in visualdl
            is_save_image (bool)  --  weather write to the directory or visualdl
        """
L
LielinJiang 已提交
336 337 338 339 340
        self.model.compute_visuals()

        if visual_results is None:
            visual_results = self.model.get_current_visuals()

L
LielinJiang 已提交
341 342 343
        min_max = self.cfg.get('min_max', None)
        if min_max is None:
            min_max = (-1., 1.)
344

郑启航 已提交
345 346 347
        image_num = self.cfg.get('image_num', None)
        if (image_num is None) or (not self.enable_visualdl):
            image_num = 1
L
LielinJiang 已提交
348
        for label, image in visual_results.items():
郑启航 已提交
349 350 351 352 353 354 355 356 357
            image_numpy = tensor2img(image, min_max, image_num)
            if (not is_save_image) and self.enable_visualdl:
                self.vdl_logger.add_image(
                    results_dir + '/' + label,
                    image_numpy,
                    step=step if step else self.global_steps,
                    dataformats="HWC" if image_num == 1 else "NCHW")
            else:
                if self.cfg.is_train:
W
wangna11BD 已提交
358 359 360 361
                    if self.by_epoch:
                        msg = 'epoch%.3d_' % self.current_epoch
                    else:
                        msg = 'iter%.3d_' % self.current_iter
郑启航 已提交
362 363 364 365 366 367
                else:
                    msg = ''
                makedirs(os.path.join(self.output_dir, results_dir))
                img_path = os.path.join(self.output_dir, results_dir,
                                        msg + '%s.png' % (label))
                save_image(image_numpy, img_path)
L
LielinJiang 已提交
368 369 370 371

    def save(self, epoch, name='checkpoint', keep=1):
        if self.local_rank != 0:
            return
L
LielinJiang 已提交
372

L
LielinJiang 已提交
373 374 375
        assert name in ['checkpoint', 'weight']

        state_dicts = {}
L
LielinJiang 已提交
376 377 378 379 380 381
        if self.by_epoch:
            save_filename = 'epoch_%s_%s.pdparams' % (
                epoch // self.iters_per_epoch, name)
        else:
            save_filename = 'iter_%s_%s.pdparams' % (epoch, name)

L
lijianshe02 已提交
382
        os.makedirs(self.output_dir, exist_ok=True)
L
LielinJiang 已提交
383
        save_path = os.path.join(self.output_dir, save_filename)
L
LielinJiang 已提交
384 385
        for net_name, net in self.model.nets.items():
            state_dicts[net_name] = net.state_dict()
L
LielinJiang 已提交
386 387 388 389 390 391 392

        if name == 'weight':
            save(state_dicts, save_path)
            return

        state_dicts['epoch'] = epoch

L
LielinJiang 已提交
393 394
        for opt_name, opt in self.model.optimizers.items():
            state_dicts[opt_name] = opt.state_dict()
L
LielinJiang 已提交
395 396 397 398 399

        save(state_dicts, save_path)

        if keep > 0:
            try:
L
LielinJiang 已提交
400 401 402 403 404 405 406 407 408 409
                if self.by_epoch:
                    checkpoint_name_to_be_removed = os.path.join(
                        self.output_dir, 'epoch_%s_%s.pdparams' %
                        ((epoch - keep * self.weight_interval) //
                         self.iters_per_epoch, name))
                else:
                    checkpoint_name_to_be_removed = os.path.join(
                        self.output_dir, 'iter_%s_%s.pdparams' %
                        (epoch - keep * self.weight_interval, name))

L
LielinJiang 已提交
410 411 412 413 414 415 416 417 418 419
                if os.path.exists(checkpoint_name_to_be_removed):
                    os.remove(checkpoint_name_to_be_removed)

            except Exception as e:
                self.logger.info('remove old checkpoints error: {}'.format(e))

    def resume(self, checkpoint_path):
        state_dicts = load(checkpoint_path)
        if state_dicts.get('epoch', None) is not None:
            self.start_epoch = state_dicts['epoch'] + 1
L
LielinJiang 已提交
420
            self.global_steps = self.iters_per_epoch * state_dicts['epoch']
L
LielinJiang 已提交
421

L
lijianshe02 已提交
422 423
            self.current_iter = state_dicts['epoch'] + 1

L
LielinJiang 已提交
424
        for net_name, net in self.model.nets.items():
425
            net.set_state_dict(state_dicts[net_name])
L
LielinJiang 已提交
426

L
LielinJiang 已提交
427
        for opt_name, opt in self.model.optimizers.items():
428
            opt.set_state_dict(state_dicts[opt_name])
L
LielinJiang 已提交
429 430 431

    def load(self, weight_path):
        state_dicts = load(weight_path)
L
LielinJiang 已提交
432

433 434 435 436 437 438 439
        def is_dict_in_dict_weight(state_dict):
            if isinstance(state_dict, dict) and len(state_dict) > 0:
                val = list(state_dict.values())[0]
                if isinstance(val, dict):
                    return True
                else:
                    return False
440
            else:
441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
                return False

        if is_dict_in_dict_weight(state_dicts):
            for net_name, net in self.model.nets.items():
                if net_name in state_dicts:
                    net.set_state_dict(state_dicts[net_name])
                    self.logger.info(
                        'Loaded pretrained weight for net {}'.format(net_name))
                else:
                    self.logger.warning(
                        'Can not find state dict of net {}. Skip load pretrained weight for net {}'
                        .format(net_name, net_name))
        else:
            assert len(self.model.nets
                       ) == 1, 'checkpoint only contain weight of one net, \
                                                but model contains more than one net!'

            net_name, net = list(self.model.nets.items())[0]
            net.set_state_dict(state_dicts)
            self.logger.info(
                'Loaded pretrained weight for net {}'.format(net_name))
郑启航 已提交
462 463 464 465 466 467

    def close(self):
        """
        when finish the training need close file handler or other.
        """
        if self.enable_visualdl:
W
wangna11BD 已提交
468
            self.vdl_logger.close()