train.py

# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# 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.

import os

import paddle
from paddle.distributed import ParallelEnv
from paddle.distributed import init_parallel_env
from paddle.io import DistributedBatchSampler
from paddle.io import DataLoader
import paddle.nn.functional as F

import paddleseg.utils.logger as logger
from paddleseg.utils import load_pretrained_model
from paddleseg.utils import resume
from paddleseg.utils import Timer, calculate_eta
from .val import evaluate


def check_logits_losses(logits, losses):
    len_logits = len(logits)
    len_losses = len(losses['types'])
    if len_logits != len_losses:
        raise RuntimeError(
            'The length of logits should equal to the types of loss config: {} != {}.'
            .format(len_logits, len_losses))


def loss_computation(logits, label, losses):
    check_logits_losses(logits, losses)
    loss = 0
    for i in range(len(logits)):
        logit = logits[i]
        if logit.shape[-2:] != label.shape[-2:]:
            logit = F.resize_bilinear(logit, label.shape[-2:])
        loss_i = losses['types'][i](logit, label)
        loss += losses['coef'][i] * loss_i
    return loss


def train(model,
          train_dataset,
          places=None,
          eval_dataset=None,
          optimizer=None,
          save_dir='output',
          iters=10000,
          batch_size=2,
          resume_model=None,
          save_interval_iters=1000,
          log_iters=10,
          num_classes=None,
          num_workers=8,
          use_vdl=False,
          losses=None,
          ignore_index=255):

    nranks = ParallelEnv().nranks

    start_iter = 0
    if resume_model is not None:
        start_iter = resume(model, optimizer, resume_model)

    if not os.path.isdir(save_dir):
        if os.path.exists(save_dir):
            os.remove(save_dir)
        os.makedirs(save_dir)

    if nranks > 1:
        # Initialize parallel training environment.
        init_parallel_env()
        strategy = paddle.distributed.prepare_context()
        ddp_model = paddle.DataParallel(model, strategy)

    batch_sampler = DistributedBatchSampler(
        train_dataset, batch_size=batch_size, shuffle=True, drop_last=True)

    loader = DataLoader(
        train_dataset,
        batch_sampler=batch_sampler,
        places=places,
        num_workers=num_workers,
        return_list=True,
    )

    if use_vdl:
        from visualdl import LogWriter
        log_writer = LogWriter(save_dir)

    timer = Timer()
    avg_loss = 0.0
    iters_per_epoch = len(batch_sampler)
    best_mean_iou = -1.0
    best_model_iter = -1
    train_reader_cost = 0.0
    train_batch_cost = 0.0
    timer.start()

    iter = start_iter
    while iter < iters:
        for data in loader:
            iter += 1
            if iter > iters:
                break
            train_reader_cost += timer.elapsed_time()
            images = data[0]
            labels = data[1].astype('int64')
            if nranks > 1:
                logits = ddp_model(images)
                loss = loss_computation(logits, labels, losses)
                # apply_collective_grads sum grads over multiple gpus.
                loss = ddp_model.scale_loss(loss)
                loss.backward()
                ddp_model.apply_collective_grads()
            else:
                logits = model(images)
                loss = loss_computation(logits, labels, losses)
                # loss = model(images, labels)
                loss.backward()
            # optimizer.minimize(loss)
            optimizer.step()
            if isinstance(optimizer._learning_rate,
                          paddle.optimizer._LRScheduler):
                optimizer._learning_rate.step()
            model.clear_gradients()
            # Sum loss over all ranks
            if nranks > 1:
                paddle.distributed.all_reduce(loss)
            avg_loss += loss.numpy()[0]
            lr = optimizer.get_lr()
            train_batch_cost += timer.elapsed_time()
            if (iter) % log_iters == 0 and ParallelEnv().local_rank == 0:
                avg_loss /= log_iters
                avg_train_reader_cost = train_reader_cost / log_iters
                avg_train_batch_cost = train_batch_cost / log_iters
                train_reader_cost = 0.0
                train_batch_cost = 0.0
                remain_iters = iters - iter
                eta = calculate_eta(remain_iters, avg_train_batch_cost)
                logger.info(
                    "[TRAIN] epoch={}, iter={}/{}, loss={:.4f}, lr={:.6f}, batch_cost={:.4f}, reader_cost={:.4f} | ETA {}"
                    .format((iter - 1) // iters_per_epoch + 1, iter, iters,
                            avg_loss, lr, avg_train_batch_cost,
                            avg_train_reader_cost, eta))
                if use_vdl:
                    log_writer.add_scalar('Train/loss', avg_loss, iter)
                    log_writer.add_scalar('Train/lr', lr, iter)
                    log_writer.add_scalar('Train/batch_cost',
                                          avg_train_batch_cost, iter)
                    log_writer.add_scalar('Train/reader_cost',
                                          avg_train_reader_cost, iter)
                avg_loss = 0.0

            if (iter % save_interval_iters == 0
                    or iter == iters) and ParallelEnv().local_rank == 0:
                current_save_dir = os.path.join(save_dir,
                                                "iter_{}".format(iter))
                if not os.path.isdir(current_save_dir):
                    os.makedirs(current_save_dir)
                paddle.save(model.state_dict(),
                            os.path.join(current_save_dir, 'model'))
                paddle.save(optimizer.state_dict(),
                            os.path.join(current_save_dir, 'model'))

                if eval_dataset is not None:
                    mean_iou, avg_acc = evaluate(
                        model,
                        eval_dataset,
                        model_dir=current_save_dir,
                        num_classes=num_classes,
                        ignore_index=ignore_index,
                        iter_id=iter)
                    if mean_iou > best_mean_iou:
                        best_mean_iou = mean_iou
                        best_model_iter = iter
                        best_model_dir = os.path.join(save_dir, "best_model")
                        paddle.save(model.state_dict(),
                                    os.path.join(best_model_dir, 'model'))
                    logger.info(
                        'Current evaluated best model in eval_dataset is iter_{}, miou={:4f}'
                        .format(best_model_iter, best_mean_iou))

                    if use_vdl:
                        log_writer.add_scalar('Evaluate/mIoU', mean_iou, iter)
                        log_writer.add_scalar('Evaluate/aAcc', avg_acc, iter)
                    model.train()
            timer.restart()
    if use_vdl:
        log_writer.close()