trainer.py 16.1 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

99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
        # build metrics
        self.metrics = None
        validate_cfg = cfg.get('validate', None)
        if validate_cfg and 'metrics' in validate_cfg:
            self.metrics = self.model.setup_metrics(validate_cfg['metrics'])

        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

114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
        # 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
136 137 138
        if self.by_epoch:
            self.weight_interval *= self.iters_per_epoch

L
LielinJiang 已提交
139 140 141
        self.validate_interval = -1
        if cfg.get('validate', None) is not None:
            self.validate_interval = cfg.validate.get('interval', -1)
142 143

        self.time_count = {}
L
LielinJiang 已提交
144
        self.best_metric = {}
145
        self.model.set_total_iter(self.total_iters)
L
LielinJiang 已提交
146

147
    def distributed_data_parallel(self):
L
LielinJiang 已提交
148
        paddle.distributed.init_parallel_env()
149
        find_unused_parameters = self.cfg.get('find_unused_parameters', False)
150
        for net_name, net in self.model.nets.items():
151 152
            self.model.nets[net_name] = paddle.DataParallel(
                net, find_unused_parameters=find_unused_parameters)
153

L
LielinJiang 已提交
154 155 156 157 158 159 160 161 162 163 164
    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
165
    def train(self):
166 167
        reader_cost_averager = TimeAverager()
        batch_cost_averager = TimeAverager()
L
LielinJiang 已提交
168

169
        iter_loader = IterLoader(self.train_dataloader)
L
LielinJiang 已提交
170

L
LielinJiang 已提交
171 172
        # set model.is_train = True
        self.model.setup_train_mode(is_train=True)
173 174 175
        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
176

177 178 179 180 181 182 183 184 185 186 187
            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))
188 189 190

            step_start_time = time.time()

191 192 193 194 195 196 197 198 199 200 201 202
            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
203
            self.learning_rate_scheduler_step()
L
LielinJiang 已提交
204

L
LielinJiang 已提交
205
            if self.validate_interval > -1 and self.current_iter % self.validate_interval == 0:
206
                self.test()
L
fix nan  
LielinJiang 已提交
207

L
LielinJiang 已提交
208 209 210
            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
211

212
            self.current_iter += 1
L
LielinJiang 已提交
213

L
LielinJiang 已提交
214 215
    def test(self):
        if not hasattr(self, 'test_dataloader'):
216
            self.test_dataloader = build_dataloader(self.cfg.dataset.test,
217 218
                                                    is_train=False,
                                                    distributed=False)
L
lijianshe02 已提交
219 220 221
        iter_loader = IterLoader(self.test_dataloader)
        if self.max_eval_steps is None:
            self.max_eval_steps = len(self.test_dataloader)
222 223 224 225

        if self.metrics:
            for metric in self.metrics.values():
                metric.reset()
L
LielinJiang 已提交
226

L
LielinJiang 已提交
227 228 229
        # set model.is_train = False
        self.model.setup_train_mode(is_train=False)

L
lijianshe02 已提交
230 231
        for i in range(self.max_eval_steps):
            data = next(iter_loader)
232 233
            self.model.setup_input(data)
            self.model.test_iter(metrics=self.metrics)
L
LielinJiang 已提交
234 235

            visual_results = {}
L
LielinJiang 已提交
236 237 238
            current_paths = self.model.get_image_paths()
            current_visuals = self.model.get_current_visuals()

L
LielinJiang 已提交
239 240 241 242 243 244 245 246 247 248 249 250
            if len(current_visuals) > 0 and list(
                    current_visuals.values())[0].shape == 4:
                num_samples = list(current_visuals.values())[0].shape[0]
            else:
                num_samples = 1

            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)
L
LielinJiang 已提交
251 252
                for k, img_tensor in current_visuals.items():
                    name = '%s_%s' % (basename, k)
L
LielinJiang 已提交
253 254 255 256
                    if len(img_tensor.shape) == 4:
                        visual_results.update({name: img_tensor[j]})
                    else:
                        visual_results.update({name: img_tensor})
L
LielinJiang 已提交
257

郑启航 已提交
258 259 260 261
            self.visual('visual_test',
                        visual_results=visual_results,
                        step=self.batch_id,
                        is_save_image=True)
L
LielinJiang 已提交
262

L
LielinJiang 已提交
263
            if i % self.log_interval == 0:
264
                self.logger.info('Test iter: [%d/%d]' %
L
lijianshe02 已提交
265
                                 (i, self.max_eval_steps))
L
LielinJiang 已提交
266

267 268 269 270 271
        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
272 273
    def print_log(self):
        losses = self.model.get_current_losses()
L
LielinJiang 已提交
274

275 276 277 278 279 280 281 282 283
        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
284 285 286

        for k, v in losses.items():
            message += '%s: %.3f ' % (k, v)
郑启航 已提交
287 288
            if self.enable_visualdl:
                self.vdl_logger.add_scalar(k, v, step=self.global_steps)
L
LielinJiang 已提交
289

290 291 292
        if hasattr(self, 'step_time'):
            message += 'batch_cost: %.5f sec ' % self.step_time

293
        if hasattr(self, 'data_time'):
294
            message += 'reader_cost: %.5f sec ' % self.data_time
295

296
        if hasattr(self, 'ips'):
L
LielinJiang 已提交
297 298 299
            message += 'ips: %.5f images/s ' % self.ips

        if hasattr(self, 'step_time'):
L
LielinJiang 已提交
300 301 302
            eta = self.step_time * (self.total_iters - self.current_iter)
            eta = eta if eta > 0 else 0

L
LielinJiang 已提交
303 304
            eta_str = str(datetime.timedelta(seconds=int(eta)))
            message += f'eta: {eta_str}'
305

L
LielinJiang 已提交
306 307 308 309 310
        # print the message
        self.logger.info(message)

    @property
    def current_learning_rate(self):
L
LielinJiang 已提交
311 312
        for optimizer in self.model.optimizers.values():
            return optimizer.get_lr()
L
LielinJiang 已提交
313

郑启航 已提交
314 315 316 317 318 319 320 321 322 323 324 325 326 327
    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
328 329 330 331 332
        self.model.compute_visuals()

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

L
LielinJiang 已提交
333 334 335
        min_max = self.cfg.get('min_max', None)
        if min_max is None:
            min_max = (-1., 1.)
336

郑启航 已提交
337 338 339
        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
340
        for label, image in visual_results.items():
郑启航 已提交
341 342 343 344 345 346 347 348 349
            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
350 351 352 353
                    if self.by_epoch:
                        msg = 'epoch%.3d_' % self.current_epoch
                    else:
                        msg = 'iter%.3d_' % self.current_iter
郑启航 已提交
354 355 356 357 358 359
                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
360 361 362 363

    def save(self, epoch, name='checkpoint', keep=1):
        if self.local_rank != 0:
            return
L
LielinJiang 已提交
364

L
LielinJiang 已提交
365 366 367
        assert name in ['checkpoint', 'weight']

        state_dicts = {}
L
LielinJiang 已提交
368 369 370 371 372 373
        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
374
        os.makedirs(self.output_dir, exist_ok=True)
L
LielinJiang 已提交
375
        save_path = os.path.join(self.output_dir, save_filename)
L
LielinJiang 已提交
376 377
        for net_name, net in self.model.nets.items():
            state_dicts[net_name] = net.state_dict()
L
LielinJiang 已提交
378 379 380 381 382 383 384

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

        state_dicts['epoch'] = epoch

L
LielinJiang 已提交
385 386
        for opt_name, opt in self.model.optimizers.items():
            state_dicts[opt_name] = opt.state_dict()
L
LielinJiang 已提交
387 388 389 390 391

        save(state_dicts, save_path)

        if keep > 0:
            try:
L
LielinJiang 已提交
392 393 394 395 396 397 398 399 400 401
                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
402 403 404 405 406 407 408 409 410 411
                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
412
            self.global_steps = self.iters_per_epoch * state_dicts['epoch']
L
LielinJiang 已提交
413

L
lijianshe02 已提交
414 415
            self.current_iter = state_dicts['epoch'] + 1

L
LielinJiang 已提交
416
        for net_name, net in self.model.nets.items():
417
            net.set_state_dict(state_dicts[net_name])
L
LielinJiang 已提交
418

L
LielinJiang 已提交
419
        for opt_name, opt in self.model.optimizers.items():
420
            opt.set_state_dict(state_dicts[opt_name])
L
LielinJiang 已提交
421 422 423

    def load(self, weight_path):
        state_dicts = load(weight_path)
L
LielinJiang 已提交
424

L
LielinJiang 已提交
425
        for net_name, net in self.model.nets.items():
426 427 428 429 430 431 432 433
            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
434 435 436 437 438 439 440 441

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

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