trainer.py 16.3 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
LielinJiang 已提交
30

L
fix nan  
LielinJiang 已提交
31

32 33 34 35 36
class IterLoader:
    def __init__(self, dataloader):
        self._dataloader = dataloader
        self.iter_loader = iter(self._dataloader)
        self._epoch = 1
L
LielinJiang 已提交
37

38 39 40 41 42 43 44 45 46 47 48
    @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
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
        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):
75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
        # 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
        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
92

L
LielinJiang 已提交
93
        # build model
94
        self.model = build_model(cfg.model)
95 96 97
        # multiple gpus prepare
        if ParallelEnv().nranks > 1:
            self.distributed_data_parallel()
L
LielinJiang 已提交
98

99 100
        # build metrics
        self.metrics = None
L
LielinJiang 已提交
101
        self.is_save_img = True
102 103 104
        validate_cfg = cfg.get('validate', None)
        if validate_cfg and 'metrics' in validate_cfg:
            self.metrics = self.model.setup_metrics(validate_cfg['metrics'])
105 106
        if validate_cfg and 'save_img' in validate_cfg:
            self.is_save_img = validate_cfg['save_img']
107 108 109 110 111 112 113 114 115 116

        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

117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
        # 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
139 140 141
        if self.by_epoch:
            self.weight_interval *= self.iters_per_epoch

L
LielinJiang 已提交
142 143 144
        self.validate_interval = -1
        if cfg.get('validate', None) is not None:
            self.validate_interval = cfg.validate.get('interval', -1)
145 146

        self.time_count = {}
L
LielinJiang 已提交
147
        self.best_metric = {}
148
        self.model.set_total_iter(self.total_iters)
L
LielinJiang 已提交
149

150
    def distributed_data_parallel(self):
L
LielinJiang 已提交
151
        paddle.distributed.init_parallel_env()
152
        find_unused_parameters = self.cfg.get('find_unused_parameters', False)
153
        for net_name, net in self.model.nets.items():
154 155
            self.model.nets[net_name] = paddle.DataParallel(
                net, find_unused_parameters=find_unused_parameters)
156

L
LielinJiang 已提交
157 158 159 160 161 162 163 164 165 166 167
    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
168
    def train(self):
169 170
        reader_cost_averager = TimeAverager()
        batch_cost_averager = TimeAverager()
L
LielinJiang 已提交
171

172
        iter_loader = IterLoader(self.train_dataloader)
L
LielinJiang 已提交
173

L
LielinJiang 已提交
174 175
        # set model.is_train = True
        self.model.setup_train_mode(is_train=True)
176 177 178
        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
179

180 181 182 183 184 185 186 187 188 189 190
            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)

            batch_cost_averager.record(time.time() - step_start_time,
                                       num_samples=self.cfg.get(
                                           'batch_size', 1))
191 192 193

            step_start_time = time.time()

194 195 196 197 198 199 200 201 202 203 204 205
            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()

            if self.current_iter % self.visual_interval == 0:
                self.visual('visual_train')

L
LielinJiang 已提交
206
            self.learning_rate_scheduler_step()
L
LielinJiang 已提交
207

L
LielinJiang 已提交
208
            if self.validate_interval > -1 and self.current_iter % self.validate_interval == 0:
209
                self.test()
L
fix nan  
LielinJiang 已提交
210

L
LielinJiang 已提交
211 212 213
            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
214

215
            self.current_iter += 1
L
LielinJiang 已提交
216

L
LielinJiang 已提交
217 218
    def test(self):
        if not hasattr(self, 'test_dataloader'):
219
            self.test_dataloader = build_dataloader(self.cfg.dataset.test,
220 221
                                                    is_train=False,
                                                    distributed=False)
L
lijianshe02 已提交
222 223 224
        iter_loader = IterLoader(self.test_dataloader)
        if self.max_eval_steps is None:
            self.max_eval_steps = len(self.test_dataloader)
225 226 227 228

        if self.metrics:
            for metric in self.metrics.values():
                metric.reset()
L
LielinJiang 已提交
229

L
LielinJiang 已提交
230 231 232
        # set model.is_train = False
        self.model.setup_train_mode(is_train=False)

L
lijianshe02 已提交
233 234
        for i in range(self.max_eval_steps):
            data = next(iter_loader)
235 236
            self.model.setup_input(data)
            self.model.test_iter(metrics=self.metrics)
L
LielinJiang 已提交
237

238 239 240 241
            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
242

243 244 245
                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
246
                else:
247
                    num_samples = 1
L
LielinJiang 已提交
248

249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265
                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
266

L
LielinJiang 已提交
267
            if i % self.log_interval == 0:
268
                self.logger.info('Test iter: [%d/%d]' %
L
lijianshe02 已提交
269
                                 (i, self.max_eval_steps))
L
LielinJiang 已提交
270

271 272 273 274 275
        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
276 277
    def print_log(self):
        losses = self.model.get_current_losses()
L
LielinJiang 已提交
278

279 280 281 282 283 284 285 286 287
        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
288 289 290

        for k, v in losses.items():
            message += '%s: %.3f ' % (k, v)
郑启航 已提交
291 292
            if self.enable_visualdl:
                self.vdl_logger.add_scalar(k, v, step=self.global_steps)
L
LielinJiang 已提交
293

294 295 296
        if hasattr(self, 'step_time'):
            message += 'batch_cost: %.5f sec ' % self.step_time

297
        if hasattr(self, 'data_time'):
298
            message += 'reader_cost: %.5f sec ' % self.data_time
299

300
        if hasattr(self, 'ips'):
L
LielinJiang 已提交
301 302 303
            message += 'ips: %.5f images/s ' % self.ips

        if hasattr(self, 'step_time'):
L
LielinJiang 已提交
304 305 306
            eta = self.step_time * (self.total_iters - self.current_iter)
            eta = eta if eta > 0 else 0

L
LielinJiang 已提交
307 308
            eta_str = str(datetime.timedelta(seconds=int(eta)))
            message += f'eta: {eta_str}'
309

L
LielinJiang 已提交
310 311 312 313 314
        # print the message
        self.logger.info(message)

    @property
    def current_learning_rate(self):
L
LielinJiang 已提交
315 316
        for optimizer in self.model.optimizers.values():
            return optimizer.get_lr()
L
LielinJiang 已提交
317

郑启航 已提交
318 319 320 321 322 323 324 325 326 327 328 329 330 331
    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
332 333 334 335 336
        self.model.compute_visuals()

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

L
LielinJiang 已提交
337 338 339
        min_max = self.cfg.get('min_max', None)
        if min_max is None:
            min_max = (-1., 1.)
340

郑启航 已提交
341 342 343
        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
344
        for label, image in visual_results.items():
郑启航 已提交
345 346 347 348 349 350 351 352 353
            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
354 355 356 357
                    if self.by_epoch:
                        msg = 'epoch%.3d_' % self.current_epoch
                    else:
                        msg = 'iter%.3d_' % self.current_iter
郑启航 已提交
358 359 360 361 362 363
                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
364 365 366 367

    def save(self, epoch, name='checkpoint', keep=1):
        if self.local_rank != 0:
            return
L
LielinJiang 已提交
368

L
LielinJiang 已提交
369 370 371
        assert name in ['checkpoint', 'weight']

        state_dicts = {}
L
LielinJiang 已提交
372 373 374 375 376 377
        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
378
        os.makedirs(self.output_dir, exist_ok=True)
L
LielinJiang 已提交
379
        save_path = os.path.join(self.output_dir, save_filename)
L
LielinJiang 已提交
380 381
        for net_name, net in self.model.nets.items():
            state_dicts[net_name] = net.state_dict()
L
LielinJiang 已提交
382 383 384 385 386 387 388

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

        state_dicts['epoch'] = epoch

L
LielinJiang 已提交
389 390
        for opt_name, opt in self.model.optimizers.items():
            state_dicts[opt_name] = opt.state_dict()
L
LielinJiang 已提交
391 392 393 394 395

        save(state_dicts, save_path)

        if keep > 0:
            try:
L
LielinJiang 已提交
396 397 398 399 400 401 402 403 404 405
                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
406 407 408 409 410 411 412 413 414 415
                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
416
            self.global_steps = self.iters_per_epoch * state_dicts['epoch']
L
LielinJiang 已提交
417

L
lijianshe02 已提交
418 419
            self.current_iter = state_dicts['epoch'] + 1

L
LielinJiang 已提交
420
        for net_name, net in self.model.nets.items():
421
            net.set_state_dict(state_dicts[net_name])
L
LielinJiang 已提交
422

L
LielinJiang 已提交
423
        for opt_name, opt in self.model.optimizers.items():
424
            opt.set_state_dict(state_dicts[opt_name])
L
LielinJiang 已提交
425 426 427

    def load(self, weight_path):
        state_dicts = load(weight_path)
L
LielinJiang 已提交
428

L
LielinJiang 已提交
429
        for net_name, net in self.model.nets.items():
430 431 432 433 434 435 436 437
            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))
郑启航 已提交
438 439 440 441 442 443 444 445

    def close(self):
        """
        when finish the training need close file handler or other.

        """
        if self.enable_visualdl:
            self.vdl_logger.close()