callbacks.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 time
import numbers
import warnings

import numpy as np

import paddle
from paddle.distributed import ParallelEnv
from paddle.utils import try_import

from .progressbar import ProgressBar

__all__ = []


def config_callbacks(callbacks=None,
                     model=None,
                     batch_size=None,
                     epochs=None,
                     steps=None,
                     log_freq=2,
                     verbose=2,
                     save_freq=1,
                     save_dir=None,
                     metrics=None,
                     mode='train'):
    cbks = callbacks or []
    cbks = cbks if isinstance(cbks, (list, tuple)) else [cbks]
    if not any(isinstance(k, ProgBarLogger) for k in cbks) and verbose:
        cbks = [ProgBarLogger(log_freq, verbose=verbose)] + cbks

    if not any(isinstance(k, ModelCheckpoint) for k in cbks):
        cbks = cbks + [ModelCheckpoint(save_freq, save_dir)]

    for k in cbks:
        if isinstance(k, EarlyStopping):
            k.save_dir = save_dir
    if not any(isinstance(k, LRScheduler) for k in cbks):
        cbks = cbks + [LRScheduler()]

    cbk_list = CallbackList(cbks)
    cbk_list.set_model(model)
    metrics = metrics or [] if mode != 'test' else []
    params = {
        'batch_size': batch_size,
        'epochs': epochs,
        'steps': steps,
        'verbose': verbose,
        'metrics': metrics,
    }
    cbk_list.set_params(params)
    return cbk_list


class CallbackList(object):

    def __init__(self, callbacks=None):
        # copy
        self.callbacks = [c for c in callbacks]
        self.params = {}
        self.model = None

    def append(self, callback):
        self.callbacks.append(callback)

    def __iter__(self):
        return iter(self.callbacks)

    def set_params(self, params):
        for c in self.callbacks:
            c.set_params(params)

    def set_model(self, model):
        for c in self.callbacks:
            c.set_model(model)

    def _call(self, name, *args):
        for c in self.callbacks:
            func = getattr(c, name)
            func(*args)

    def _check_mode(self, mode):
        assert mode in ['train', 'eval', 'predict'], \
            'mode should be train, eval or predict'

    def on_begin(self, mode, logs=None):
        self._check_mode(mode)
        name = 'on_{}_begin'.format(mode)
        self._call(name, logs)

    def on_end(self, mode, logs=None):
        self._check_mode(mode)
        name = 'on_{}_end'.format(mode)
        self._call(name, logs)

    def on_epoch_begin(self, epoch=None, logs=None):
        self._call('on_epoch_begin', epoch, logs)

    def on_epoch_end(self, epoch=None, logs=None):
        self._call('on_epoch_end', epoch, logs)

    def on_batch_begin(self, mode, step=None, logs=None):
        self._check_mode(mode)
        name = 'on_{}_batch_begin'.format(mode)
        self._call(name, step, logs)

    def on_batch_end(self, mode, step=None, logs=None):
        self._check_mode(mode)
        name = 'on_{}_batch_end'.format(mode)
        self._call(name, step, logs)


class Callback(object):
    """
    Base class used to build new callbacks. And new callbacks could also
    terminate training by setting `model.stop_training=True`.

    Examples:

        .. code-block:: python

            import paddle

            # build a simple model checkpoint callback
            class ModelCheckpoint(paddle.callbacks.Callback):
                def __init__(self, save_freq=1, save_dir=None):
                    self.save_freq = save_freq
                    self.save_dir = save_dir

                def on_epoch_end(self, epoch, logs=None):
                    if self.model is not None and epoch % self.save_freq == 0:
                        path = '{}/{}'.format(self.save_dir, epoch)
                        print('save checkpoint at {}'.format(path))
                        self.model.save(path)

    """

    def __init__(self):
        self.model = None
        self.params = {}

    def set_params(self, params):
        """
        Set parameters, which is dict. The keys contain:

          - 'batch_size': an integer. Number of samples per batch.
          - 'epochs': an integer. Number of epochs.
          - 'steps': an integer. Number of steps of one epoch.
          - 'verbose': an integer. Verbose mode is 0, 1 or 2. 0 = silent, 1 = progress bar, 2 = one line per epoch.
          - 'metrics': a list of str. Names of metrics, including 'loss' and the names of paddle.metric.Metric.
        """
        self.params = params

    def set_model(self, model):
        """model is instance of paddle.Model.
        """
        self.model = model

    def on_train_begin(self, logs=None):
        """Called at the start of training.

        Args:
            logs (dict): The logs is a dict or None.
        """

    def on_train_end(self, logs=None):
        """Called at the end of training.

        Args:
            logs (dict): The logs is a dict or None. The keys of logs
                passed by paddle.Model contains 'loss', metric names and
                `batch_size`.
        """

    def on_eval_begin(self, logs=None):
        """Called at the start of evaluation.

        Args:
            logs (dict): The logs is a dict or None. The keys of logs
                passed by paddle.Model contains 'steps' and 'metrics',
                The `steps` is number of total steps of validation dataset.
                The `metrics` is a list of str including 'loss' and the names
                of paddle.metric.Metric.
        """

    def on_eval_end(self, logs=None):
        """Called at the end of evaluation.

        Args:
            logs (dict): The logs is a dict or None. The `logs` passed by
                paddle.Model is a dict contains 'loss', metrics and 'batch_size'
                of last batch of validation dataset.
        """

    def on_predict_begin(self, logs=None):
        """Called at the beginning of predict.

        Args:
            logs (dict): The logs is a dict or None.
        """

    def on_predict_end(self, logs=None):
        """Called at the end of predict.

        Args:
            logs (dict): The logs is a dict or None.
        """

    def on_epoch_begin(self, epoch, logs=None):
        """Called at the beginning of each epoch.

        Args:
            epoch (int): The index of epoch.
            logs (dict): The logs is a dict or None. The `logs` passed by
                paddle.Model is None.
        """

    def on_epoch_end(self, epoch, logs=None):
        """Called at the end of each epoch.

        Args:
            epoch (int): The index of epoch.
            logs (dict): The logs is a dict or None. The `logs` passed by
                paddle.Model is a dict, contains 'loss', metrics and 'batch_size'
                of last batch.
        """

    def on_train_batch_begin(self, step, logs=None):
        """Called at the beginning of each batch in training.

        Args:
            step (int): The index of step (or iteration).
            logs (dict): The logs is a dict or None. The `logs` passed by
                paddle.Model is empty.
        """

    def on_train_batch_end(self, step, logs=None):
        """Called at the end of each batch in training.

        Args:
            step (int): The index of step (or iteration).
            logs (dict): The logs is a dict or None. The `logs` passed by
                paddle.Model is a dict, contains 'loss', metrics and 'batch_size'
                of current batch.
        """

    def on_eval_batch_begin(self, step, logs=None):
        """Called at the beginning of each batch in evaluation.

        Args:
            step (int): The index of step (or iteration).
            logs (dict): The logs is a dict or None. The `logs` passed by
                paddle.Model is empty.
        """

    def on_eval_batch_end(self, step, logs=None):
        """Called at the end of each batch in evaluation.

        Args:
            step (int): The index of step (or iteration).
            logs (dict): The logs is a dict or None. The `logs` passed by
                paddle.Model is a dict, contains 'loss', metrics and 'batch_size'
                of current batch.
        """

    def on_predict_batch_begin(self, step, logs=None):
        """Called at the beginning of each batch in predict.

        Args:
            step (int): The index of step (or iteration).
            logs (dict): The logs is a dict or None.
        """

    def on_predict_batch_end(self, step, logs=None):
        """Called at the end of each batch in predict.

        Args:
            step (int): The index of step (or iteration).
            logs (dict): The logs is a dict or None.
        """


class ProgBarLogger(Callback):
    """
    Logger callback function to print loss and metrics to stdout. It supports
    silent mode (not print), progress bar or one line per each printing,
    see arguments for more detailed.

    Args:
        log_freq (int): The frequency, in number of steps,
            the logs such as loss, metrics are printed. Default: 1.
        verbose (int): The verbosity mode, should be 0, 1, or 2.
            0 = silent, 1 = progress bar, 2 = one line each printing, 3 = 2 +
            time counter, such as average reader cost, samples per second.
            Default: 2.

    Examples:
        .. code-block:: python

            import paddle
            import paddle.vision.transforms as T
            from paddle.vision.datasets import MNIST
            from paddle.static import InputSpec

            inputs = [InputSpec([-1, 1, 28, 28], 'float32', 'image')]
            labels = [InputSpec([None, 1], 'int64', 'label')]

            transform = T.Compose([
                T.Transpose(),
                T.Normalize([127.5], [127.5])
            ])
            train_dataset = MNIST(mode='train', transform=transform)

            lenet = paddle.vision.models.LeNet()
            model = paddle.Model(lenet,
                inputs, labels)

            optim = paddle.optimizer.Adam(0.001, parameters=lenet.parameters())
            model.prepare(optimizer=optim,
                        loss=paddle.nn.CrossEntropyLoss(),
                        metrics=paddle.metric.Accuracy())

            callback = paddle.callbacks.ProgBarLogger(log_freq=10)
            model.fit(train_dataset, batch_size=64, callbacks=callback)
    """

    def __init__(self, log_freq=1, verbose=2):
        self.epochs = None
        self.steps = None
        self.progbar = None
        self.verbose = verbose
        self.log_freq = log_freq

    def _is_print(self):
        return self.verbose and ParallelEnv().local_rank == 0

    def on_train_begin(self, logs=None):
        self.epochs = self.params['epochs']
        assert self.epochs
        self.train_metrics = self.params['metrics']
        assert self.train_metrics

        self._train_timer = {
            'data_time': 0,
            'batch_time': 0,
            'count': 0,
            'samples': 0,
        }
        if self._is_print():
            print(
                "The loss value printed in the log is the current step, and the metric is the average value of previous steps."
            )

    def on_epoch_begin(self, epoch=None, logs=None):
        self.steps = self.params['steps']
        self.epoch = epoch
        self.train_step = 0
        if self.epochs and self._is_print():
            print('Epoch %d/%d' % (epoch + 1, self.epochs))
        self.train_progbar = ProgressBar(num=self.steps, verbose=self.verbose)

        self._train_timer['batch_start_time'] = time.time()

    def _updates(self, logs, mode):
        values = []
        metrics = getattr(self, '%s_metrics' % (mode))
        progbar = getattr(self, '%s_progbar' % (mode))
        steps = getattr(self, '%s_step' % (mode))

        for k in metrics:
            if k in logs:
                values.append((k, logs[k]))

        if self.verbose == 3 and hasattr(self, '_%s_timer' % (mode)):
            timer = getattr(self, '_%s_timer' % (mode))
            cnt = timer['count'] if timer['count'] > 0 else 1.0
            samples = timer['samples'] if timer['samples'] > 0 else 1.0
            values.append(
                ('avg_reader_cost', "%.5f sec" % (timer['data_time'] / cnt)))
            values.append(
                ('avg_batch_cost', "%.5f sec" % (timer['batch_time'] / cnt)))
            values.append(
                ('ips', "%.5f samples/sec" %
                 (samples / (timer['data_time'] + timer['batch_time']))))
            timer['count'] = 0
            timer['samples'] = 0
            timer['data_time'] = 0.
            timer['batch_time'] = 0.

        progbar.update(steps, values)

    def on_train_batch_begin(self, step, logs=None):
        self._train_timer['batch_data_end_time'] = time.time()
        self._train_timer['data_time'] += (
            self._train_timer['batch_data_end_time'] -
            self._train_timer['batch_start_time'])

    def on_train_batch_end(self, step, logs=None):
        logs = logs or {}
        self.train_step += 1

        self._train_timer['batch_time'] += (
            time.time() - self._train_timer['batch_data_end_time'])
        self._train_timer['count'] += 1
        samples = logs.get('batch_size', 1)
        self._train_timer['samples'] += samples
        if self._is_print() and self.train_step % self.log_freq == 0:
            if self.steps is None or self.train_step < self.steps:
                self._updates(logs, 'train')
        self._train_timer['batch_start_time'] = time.time()

    def on_epoch_end(self, epoch, logs=None):
        logs = logs or {}
        if self._is_print() and (self.steps is not None):
            self._updates(logs, 'train')

    def on_eval_begin(self, logs=None):
        self.eval_steps = logs.get('steps', None)
        self.eval_metrics = logs.get('metrics', [])
        self.eval_step = 0
        self.evaled_samples = 0

        self._eval_timer = {
            'data_time': 0,
            'batch_time': 0,
            'count': 0,
            'samples': 0,
        }

        self.eval_progbar = ProgressBar(num=self.eval_steps,
                                        verbose=self.verbose)
        if self._is_print():
            print('Eval begin...')

        self._eval_timer['batch_start_time'] = time.time()

    def on_eval_batch_begin(self, step, logs=None):
        self._eval_timer['batch_data_end_time'] = time.time()
        self._eval_timer['data_time'] += (
            self._eval_timer['batch_data_end_time'] -
            self._eval_timer['batch_start_time'])

    def on_eval_batch_end(self, step, logs=None):
        logs = logs or {}
        self.eval_step += 1
        samples = logs.get('batch_size', 1)
        self.evaled_samples += samples

        self._eval_timer['batch_time'] += (
            time.time() - self._eval_timer['batch_data_end_time'])
        self._eval_timer['count'] += 1
        samples = logs.get('batch_size', 1)
        self._eval_timer['samples'] += samples

        if self._is_print() and self.eval_step % self.log_freq == 0:
            if self.eval_steps is None or self.eval_step < self.eval_steps:
                self._updates(logs, 'eval')

        self._eval_timer['batch_start_time'] = time.time()

    def on_predict_begin(self, logs=None):
        self.test_steps = logs.get('steps', None)
        self.test_metrics = logs.get('metrics', [])
        self.test_step = 0
        self.tested_samples = 0

        self._test_timer = {
            'data_time': 0,
            'batch_time': 0,
            'count': 0,
            'samples': 0,
        }

        self.test_progbar = ProgressBar(num=self.test_steps,
                                        verbose=self.verbose)
        if self._is_print():
            print('Predict begin...')

        self._test_timer['batch_start_time'] = time.time()

    def on_predict_batch_begin(self, step, logs=None):
        self._test_timer['batch_data_end_time'] = time.time()
        self._test_timer['data_time'] += (
            self._test_timer['batch_data_end_time'] -
            self._test_timer['batch_start_time'])

    def on_predict_batch_end(self, step, logs=None):
        logs = logs or {}
        self.test_step += 1
        samples = logs.get('batch_size', 1)
        self.tested_samples += samples

        self._test_timer['batch_time'] += (
            time.time() - self._test_timer['batch_data_end_time'])
        self._test_timer['count'] += 1
        samples = logs.get('batch_size', 1)
        self._test_timer['samples'] += samples

        if self.test_step % self.log_freq == 0 and self._is_print():
            if self.test_steps is None or self.test_step < self.test_steps:
                self._updates(logs, 'test')

        self._test_timer['batch_start_time'] = time.time()

    def on_eval_end(self, logs=None):
        logs = logs or {}
        if self._is_print() and (self.eval_steps is not None):
            self._updates(logs, 'eval')
            print('Eval samples: %d' % (self.evaled_samples))

    def on_predict_end(self, logs=None):
        logs = logs or {}
        if self._is_print():
            if self.test_step % self.log_freq != 0 or self.verbose == 1:
                self._updates(logs, 'test')
            print('Predict samples: %d' % (self.tested_samples))


class ModelCheckpoint(Callback):
    """
    Model checkpoint callback function to save model weights and optimizer
    state during training in conjunction with model.fit(). Currently,
    ModelCheckpoint only supports saving after a fixed number of epochs.

    Args:
        save_freq(int): The frequency, in number of epochs, the model checkpoint
            are saved. Default: 1.
        save_dir(str|None): The directory to save checkpoint during training.
            If None, will not save checkpoint. Default: None.

    Examples:
        .. code-block:: python

            import paddle
            import paddle.vision.transforms as T
            from paddle.vision.datasets import MNIST
            from paddle.static import InputSpec

            inputs = [InputSpec([-1, 1, 28, 28], 'float32', 'image')]
            labels = [InputSpec([None, 1], 'int64', 'label')]

            transform = T.Compose([
                T.Transpose(),
                T.Normalize([127.5], [127.5])
            ])
            train_dataset = MNIST(mode='train', transform=transform)

            lenet = paddle.vision.models.LeNet()
            model = paddle.Model(lenet,
                inputs, labels)

            optim = paddle.optimizer.Adam(0.001, parameters=lenet.parameters())
            model.prepare(optimizer=optim,
                        loss=paddle.nn.CrossEntropyLoss(),
                        metrics=paddle.metric.Accuracy())

            callback = paddle.callbacks.ModelCheckpoint(save_dir='./temp')
            model.fit(train_dataset, batch_size=64, callbacks=callback)
    """

    def __init__(self, save_freq=1, save_dir=None):
        self.save_freq = save_freq
        self.save_dir = save_dir

    def on_epoch_begin(self, epoch=None, logs=None):
        self.epoch = epoch

    def _is_save(self):
        return self.model and self.save_dir and ParallelEnv().local_rank == 0

    def on_epoch_end(self, epoch, logs=None):
        if self._is_save() and self.epoch % self.save_freq == 0:
            path = '{}/{}'.format(self.save_dir, epoch)
            print('save checkpoint at {}'.format(os.path.abspath(path)))
            self.model.save(path)

    def on_train_end(self, logs=None):
        if self._is_save():
            path = '{}/final'.format(self.save_dir)
            print('save checkpoint at {}'.format(os.path.abspath(path)))
            self.model.save(path)


class LRScheduler(Callback):
    """Lr scheduler callback function

    Args:
        by_step(bool, optional): whether to update learning rate scheduler
            by step. Default: True.
        by_epoch(bool, optional): whether to update learning rate scheduler
            by epoch. Default: False.

    Examples:
        .. code-block:: python

            import paddle
            import paddle.vision.transforms as T
            from paddle.static import InputSpec

            inputs = [InputSpec([-1, 1, 28, 28], 'float32', 'image')]
            labels = [InputSpec([None, 1], 'int64', 'label')]

            transform = T.Compose([
                T.Transpose(),
                T.Normalize([127.5], [127.5])
            ])
            train_dataset = paddle.vision.datasets.MNIST(mode='train', transform=transform)

            lenet = paddle.vision.models.LeNet()
            model = paddle.Model(lenet,
                inputs, labels)

            base_lr = 1e-3
            boundaries = [5, 8]
            wamup_steps = 4

            def make_optimizer(parameters=None):
                momentum = 0.9
                weight_decay = 5e-4
                values = [base_lr * (0.1**i) for i in range(len(boundaries) + 1)]
                learning_rate = paddle.optimizer.lr.PiecewiseDecay(
                    boundaries=boundaries, values=values)
                learning_rate = paddle.optimizer.lr.LinearWarmup(
                    learning_rate=learning_rate,
                    warmup_steps=wamup_steps,
                    start_lr=base_lr / 5.,
                    end_lr=base_lr,
                    verbose=True)
                optimizer = paddle.optimizer.Momentum(
                    learning_rate=learning_rate,
                    weight_decay=weight_decay,
                    momentum=momentum,
                    parameters=parameters)
                return optimizer

            optim = make_optimizer(parameters=lenet.parameters())
            model.prepare(optimizer=optim,
                        loss=paddle.nn.CrossEntropyLoss(),
                        metrics=paddle.metric.Accuracy())

            # if LRScheduler callback not set, an instance LRScheduler update by step
            # will be created auto.
            model.fit(train_dataset, batch_size=64)

            # create a learning rate scheduler update by epoch
            callback = paddle.callbacks.LRScheduler(by_step=False, by_epoch=True)
            model.fit(train_dataset, batch_size=64, callbacks=callback)
    """

    def __init__(self, by_step=True, by_epoch=False):
        if by_step and by_epoch:
            raise ValueError(
                "by_step option is mutually exclusive with by_epoch")

        self.by_step = by_step
        self.by_epoch = by_epoch

    def on_epoch_end(self, epoch, logs=None):
        if self.by_epoch:
            if self.model._optimizer and \
                hasattr(self.model._optimizer, '_learning_rate') and \
                isinstance(self.model._optimizer._learning_rate,
                           paddle.optimizer.lr.LRScheduler):
                self.model._optimizer._learning_rate.step()

    def on_train_batch_end(self, step, logs=None):
        if self.by_step:
            if self.model._optimizer and \
                hasattr(self.model._optimizer, '_learning_rate') and \
                isinstance(self.model._optimizer._learning_rate,
                           paddle.optimizer.lr.LRScheduler):
                self.model._optimizer._learning_rate.step()


class EarlyStopping(Callback):
    """Stop training when the given monitor stopped improving during evaluation
    by setting `model.stop_training=True`.

    Args:
        monitor(str): Quantity to be monitored. Default: 'loss'.
        mode(str|None): Mode should be one of 'auto', 'min' or 'max'. In 'min'
            mode, training will stop until monitored quantity stops decreasing.
            In 'max' mode, training will stop until monitored quantity stops
            increasing. In 'auto' mode, exact mode can be inferred by the name
            of monitor. If 'acc' in monitor, the mode will be considered as
            'max', otherwise the mode will be set to 'min'. Default: 'auto'.
        patience(int): Number of epochs with no improvement after which
            training will be stopped. Default: 0.
        verbose(int): The verbosity mode, should be 0 or 1. When verbose=0,
            logs will not be printed. When verbose=1, logs will be printed.
            Default: 1.
        min_delta(int|float): The minimum change of monitored quantity. If
            the change is less than min_delta, model could be considered as no
            improvement. Default: 0.
        baseline(int|float|None): Baseline value for the monitored quantity.
            Training will stop if the model doesn't show improvement over the
            baseline. Default: None.
        save_best_model(bool): Whether to save best model. Default: True.

    Examples:
        .. code-block:: python

            import paddle
            from paddle import Model
            from paddle.static import InputSpec
            from paddle.vision.models import LeNet
            from paddle.vision.datasets import MNIST
            from paddle.metric import Accuracy
            from paddle.nn import CrossEntropyLoss
            import paddle.vision.transforms as T

            device = paddle.set_device('cpu')
            sample_num = 200
            save_dir = './best_model_checkpoint'
            transform = T.Compose(
                [T.Transpose(), T.Normalize([127.5], [127.5])])
            train_dataset = MNIST(mode='train', transform=transform)
            val_dataset = MNIST(mode='test', transform=transform)
            net = LeNet()
            optim = paddle.optimizer.Adam(
                learning_rate=0.001, parameters=net.parameters())

            inputs = [InputSpec([None, 1, 28, 28], 'float32', 'x')]
            labels = [InputSpec([None, 1], 'int64', 'label')]

            model = Model(net, inputs=inputs, labels=labels)
            model.prepare(
                optim,
                loss=CrossEntropyLoss(reduction="sum"),
                metrics=[Accuracy()])
            callbacks = paddle.callbacks.EarlyStopping(
                'loss',
                mode='min',
                patience=1,
                verbose=1,
                min_delta=0,
                baseline=None,
                save_best_model=True)
            model.fit(train_dataset,
                      val_dataset,
                      batch_size=64,
                      log_freq=200,
                      save_freq=10,
                      save_dir=save_dir,
                      epochs=20,
                      callbacks=[callbacks])
    """

    def __init__(self,
                 monitor='loss',
                 mode='auto',
                 patience=0,
                 verbose=1,
                 min_delta=0,
                 baseline=None,
                 save_best_model=True):
        super(EarlyStopping, self).__init__()
        self.monitor = monitor
        self.patience = patience
        self.verbose = verbose
        self.baseline = baseline
        self.min_delta = abs(min_delta)
        self.wait_epoch = 0
        self.best_weights = None
        self.stopped_epoch = 0
        self.save_best_model = save_best_model
        # The value of `save_dir` is set in function `config_callbacks`
        self.save_dir = None
        if mode not in ['auto', 'min', 'max']:
            warnings.warn('EarlyStopping mode %s is unknown, '
                          'fallback to auto mode.' % mode)
            mode = 'auto'
        if mode == 'min':
            self.monitor_op = np.less
        elif mode == 'max':
            self.monitor_op = np.greater
        # When mode == 'auto', the mode should be inferred by `self.monitor`
        else:
            if 'acc' in self.monitor:
                self.monitor_op = np.greater
            else:
                self.monitor_op = np.less

        if self.monitor_op == np.greater:
            self.min_delta *= 1
        else:
            self.min_delta *= -1

    def on_train_begin(self, logs=None):
        self.wait_epoch = 0
        if self.baseline is not None:
            self.best_value = self.baseline
        else:
            self.best_value = np.inf if self.monitor_op == np.less else -np.inf
            self.best_weights = None

    def on_eval_end(self, logs=None):
        if logs is None or self.monitor not in logs:
            warnings.warn(
                'Monitor of EarlyStopping should be loss or metric name.')
            return
        current = logs[self.monitor]
        if isinstance(current, (list, tuple)):
            current = current[0]
        elif isinstance(current, numbers.Number):
            current = current
        else:
            return

        if self.monitor_op(current - self.min_delta, self.best_value):
            self.best_value = current
            self.wait_epoch = 0
            if self.save_best_model and self.save_dir is not None:
                path = os.path.join(self.save_dir, 'best_model')
                self.model.save(path)
        else:
            self.wait_epoch += 1
        if self.wait_epoch >= self.patience:
            self.model.stop_training = True
            if self.verbose > 0:
                print('Epoch %d: Early stopping.' % (self.stopped_epoch + 1))
                if self.save_best_model and self.save_dir is not None:
                    print('Best checkpoint has been saved at %s' %
                          (os.path.abspath(
                              os.path.join(self.save_dir, 'best_model'))))
        self.stopped_epoch += 1


class VisualDL(Callback):
    """
    VisualDL callback function.

    Args:
        log_dir (str): The directory to save visualdl log file.

    Examples:
        .. code-block:: python

            import paddle
            import paddle.vision.transforms as T
            from paddle.static import InputSpec

            inputs = [InputSpec([-1, 1, 28, 28], 'float32', 'image')]
            labels = [InputSpec([None, 1], 'int64', 'label')]

            transform = T.Compose([
                T.Transpose(),
                T.Normalize([127.5], [127.5])
            ])
            train_dataset = paddle.vision.datasets.MNIST(mode='train', transform=transform)
            eval_dataset = paddle.vision.datasets.MNIST(mode='test', transform=transform)

            net = paddle.vision.models.LeNet()
            model = paddle.Model(net, inputs, labels)

            optim = paddle.optimizer.Adam(0.001, parameters=net.parameters())
            model.prepare(optimizer=optim,
                        loss=paddle.nn.CrossEntropyLoss(),
                        metrics=paddle.metric.Accuracy())

            ## uncomment following lines to fit model with visualdl callback function
            # callback = paddle.callbacks.VisualDL(log_dir='visualdl_log_dir')
            # model.fit(train_dataset, eval_dataset, batch_size=64, callbacks=callback)

    """

    def __init__(self, log_dir):
        self.log_dir = log_dir
        self.epochs = None
        self.steps = None
        self.epoch = 0

    def _is_write(self):
        return ParallelEnv().local_rank == 0

    def on_train_begin(self, logs=None):
        self.epochs = self.params['epochs']
        assert self.epochs
        self.train_metrics = self.params['metrics']
        assert self.train_metrics
        self._is_fit = True
        self.train_step = 0

    def on_epoch_begin(self, epoch=None, logs=None):
        self.steps = self.params['steps']
        self.epoch = epoch

    def _updates(self, logs, mode):
        if not self._is_write():
            return
        if not hasattr(self, 'writer'):
            visualdl = try_import('visualdl')
            self.writer = visualdl.LogWriter(self.log_dir)

        metrics = getattr(self, '%s_metrics' % (mode))
        current_step = getattr(self, '%s_step' % (mode))

        if mode == 'train':
            total_step = current_step
        else:
            total_step = self.epoch

        for k in metrics:
            if k in logs:
                temp_tag = mode + '/' + k

                if isinstance(logs[k], (list, tuple)):
                    temp_value = logs[k][0]
                elif isinstance(logs[k], numbers.Number):
                    temp_value = logs[k]
                else:
                    continue

                self.writer.add_scalar(tag=temp_tag,
                                       step=total_step,
                                       value=temp_value)

    def on_train_batch_end(self, step, logs=None):
        logs = logs or {}
        self.train_step += 1

        if self._is_write():
            self._updates(logs, 'train')

    def on_eval_begin(self, logs=None):
        self.eval_steps = logs.get('steps', None)
        self.eval_metrics = logs.get('metrics', [])
        self.eval_step = 0
        self.evaled_samples = 0

    def on_train_end(self, logs=None):
        if hasattr(self, 'writer'):
            self.writer.close()
            delattr(self, 'writer')

    def on_eval_end(self, logs=None):
        if self._is_write():
            self._updates(logs, 'eval')

            if (not hasattr(self, '_is_fit')) and hasattr(self, 'writer'):
                self.writer.close()
                delattr(self, 'writer')


class ReduceLROnPlateau(Callback):
    """Reduce learning rate when a metric of evaluation has stopped improving.
    Models often benefit from reducing the learning rate by a factor
    of 2-10 once learning stagnates. This callback monitors a
    quantity and if no improvement is seen for a 'patience' number
    of epochs, the learning rate is reduced.

    Args:
        monitor(str, optional): Quantity to be monitored. Default: 'loss'.
        factor(float, optional): factor by which the learning rate will be reduced.
            `new_lr = lr * factor`. Default: 0.1.
        patience(int, optional): Number of epochs with no improvement after which
            learning rate will be reduced. Default: 10.
        verbose(int, optional): The verbosity mode. 0: quiet, 1: update messages.
            Default: 1.
        mode(str, optional): one of `{'auto', 'min', 'max'}`. In `'min'` mode,
            the learning rate will be reduced when the quantity monitored has
            stopped decreasing. In 'max' mode, learning rate will reduce until
            monitored quantity stops increasing. In 'auto' mode, exact mode
            can be inferred by the name of monitor. If 'acc' in monitor, the
            mode will be considered as 'max', otherwise the mode will be set
            to 'min'. Default: 'auto'.
        min_delta(int|float, optional): threshold for measuring the new optimum,
            to only focus on significant changes. Default: 0.
        cooldown(int, optional): number of epochs to wait before resuming normal operation after
            lr has been reduced. Default: 0.
        min_lr(float, optional): lower bound on the learning rate. Default: 0.

    Examples:
          .. code-block:: python

              import paddle
              from paddle import Model
              from paddle.static import InputSpec
              from paddle.vision.models import LeNet
              from paddle.vision.datasets import MNIST
              from paddle.metric import Accuracy
              from paddle.nn.layer.loss import CrossEntropyLoss
              import paddle.vision.transforms as T
              sample_num = 200
              transform = T.Compose(
                  [T.Transpose(), T.Normalize([127.5], [127.5])])
              train_dataset = MNIST(mode='train', transform=transform)
              val_dataset = MNIST(mode='test', transform=transform)
              net = LeNet()
              optim = paddle.optimizer.Adam(
                  learning_rate=0.001, parameters=net.parameters())
              inputs = [InputSpec([None, 1, 28, 28], 'float32', 'x')]
              labels = [InputSpec([None, 1], 'int64', 'label')]
              model = Model(net, inputs=inputs, labels=labels)
              model.prepare(
                  optim,
                  loss=CrossEntropyLoss(),
                  metrics=[Accuracy()])
              callbacks = paddle.callbacks.ReduceLROnPlateau(patience=3, verbose=1)
              model.fit(train_dataset,
                          val_dataset,
                          batch_size=64,
                          log_freq=200,
                          save_freq=10,
                          epochs=20,
                          callbacks=[callbacks])

    """

    def __init__(self,
                 monitor='loss',
                 factor=0.1,
                 patience=10,
                 verbose=1,
                 mode='auto',
                 min_delta=1e-4,
                 cooldown=0,
                 min_lr=0):
        super(ReduceLROnPlateau, self).__init__()

        self.monitor = monitor
        if factor >= 1.0:
            raise ValueError('ReduceLROnPlateau '
                             'does not support a factor >= 1.0.')

        self.factor = factor
        self.min_lr = min_lr
        self.min_delta = min_delta
        self.patience = patience
        self.verbose = verbose
        self.cooldown = cooldown
        self.cooldown_counter = 0  # Cooldown counter.
        self.wait = 0
        self.best = 0
        self.mode = mode
        self.monitor_op = None
        self.epoch = 0
        self._reset()

    def _reset(self):
        """Resets wait counter and cooldown counter.
        """
        if self.mode not in ['auto', 'min', 'max']:
            warnings.warn('Learning rate reduction mode %s is unknown, '
                          'fallback to auto mode.' % self.mode)
            self.mode = 'auto'
        if (self.mode == 'min'
                or (self.mode == 'auto' and 'acc' not in self.monitor)):
            self.monitor_op = lambda a, b: np.less(a, b - self.min_delta)
            self.best = np.Inf
        else:
            self.monitor_op = lambda a, b: np.greater(a, b + self.min_delta)
            self.best = -np.Inf
        self.cooldown_counter = 0
        self.wait = 0

    def on_train_begin(self, logs=None):
        self._reset()

    def on_eval_end(self, logs=None):
        if logs is None or self.monitor not in logs:
            warnings.warn(
                'Monitor of ReduceLROnPlateau should be loss or metric name.')
            return
        else:
            try:
                lr = self.model._optimizer._learning_rate
                if not isinstance(lr, float):
                    warnings.warn(
                        'Expected learning_rate be float, bug got {}.'.format(
                            type(lr)))
                    return
            except Exception as e:
                warnings.warn(
                    'There are something wrong when get learning_rate from optimizer: {}.'
                    .format(e))
                return

        current = logs[self.monitor]
        if isinstance(current, (list, tuple)):
            current = current[0]
        elif isinstance(current, numbers.Number):
            current = current
        else:
            return

        if self.in_cooldown():
            self.cooldown_counter -= 1
            self.wait = 0

        if self.monitor_op(current, self.best):
            self.best = current
            self.wait = 0
        elif not self.in_cooldown():
            self.wait += 1
            if self.wait >= self.patience:
                old_lr = self.model._optimizer.get_lr()
                if old_lr > np.float32(self.min_lr):
                    new_lr = old_lr * self.factor
                    new_lr = max(new_lr, self.min_lr)
                    self.model._optimizer._learning_rate = new_lr
                    if self.verbose > 0 and ParallelEnv().local_rank == 0:
                        print('\nEpoch %d: ReduceLROnPlateau reducing learning '
                              'rate to %s.' % (self.epoch + 1, new_lr))
                    self.cooldown_counter = self.cooldown
                    self.wait = 0
        self.epoch += 1

    def in_cooldown(self):
        return self.cooldown_counter > 0