lightning/pytorch_lightning/callbacks/early_stopping.py

# Copyright The PyTorch Lightning team.
#
# 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.
r"""
Early Stopping
^^^^^^^^^^^^^^

Monitor a metric and stop training when it stops improving.

"""
import logging
from typing import Any, Callable, Dict, Optional, Tuple

import numpy as np
import torch

import pytorch_lightning as pl
from pytorch_lightning.callbacks.base import Callback
from pytorch_lightning.utilities import rank_zero_deprecation, rank_zero_warn
from pytorch_lightning.utilities.exceptions import MisconfigurationException

log = logging.getLogger(__name__)


class EarlyStopping(Callback):
    r"""
    Monitor a metric and stop training when it stops improving.

    Args:
        monitor: quantity to be monitored.
        min_delta: minimum change in the monitored quantity to qualify as an improvement, i.e. an absolute
            change of less than `min_delta`, will count as no improvement.
        patience: number of checks with no improvement
            after which training will be stopped. Under the default configuration, one check happens after
            every training epoch. However, the frequency of validation can be modified by setting various parameters on
            the ``Trainer``, for example ``check_val_every_n_epoch`` and ``val_check_interval``.

            .. note::

                It must be noted that the patience parameter counts the number of validation checks with
                no improvement, and not the number of training epochs. Therefore, with parameters
                ``check_val_every_n_epoch=10`` and ``patience=3``, the trainer will perform at least 40 training
                epochs before being stopped.

        verbose: verbosity mode.
        mode: one of ``'min'``, ``'max'``. In ``'min'`` mode, training will stop when the quantity
            monitored has stopped decreasing and in ``'max'`` mode it will stop when the quantity
            monitored has stopped increasing.
        strict: whether to crash the training if `monitor` is not found in the validation metrics.
        check_finite: When set ``True``, stops training when the monitor becomes NaN or infinite.
        stopping_threshold: Stop training immediately once the monitored quantity reaches this threshold.
        divergence_threshold: Stop training as soon as the monitored quantity becomes worse than this threshold.
        check_on_train_epoch_end: whether to run early stopping at the end of the training epoch.
            If this is ``False``, then the check runs at the end of the validation.

    Raises:
        MisconfigurationException:
            If ``mode`` is none of ``"min"`` or ``"max"``.
        RuntimeError:
            If the metric ``monitor`` is not available.

    Example::

        >>> from pytorch_lightning import Trainer
        >>> from pytorch_lightning.callbacks import EarlyStopping
        >>> early_stopping = EarlyStopping('val_loss')
        >>> trainer = Trainer(callbacks=[early_stopping])
    """
    mode_dict = {"min": torch.lt, "max": torch.gt}

    order_dict = {"min": "<", "max": ">"}

    def __init__(
        self,
        monitor: Optional[str] = None,
        min_delta: float = 0.0,
        patience: int = 3,
        verbose: bool = False,
        mode: str = "min",
        strict: bool = True,
        check_finite: bool = True,
        stopping_threshold: Optional[float] = None,
        divergence_threshold: Optional[float] = None,
        check_on_train_epoch_end: bool = True,
    ):
        super().__init__()
        self.min_delta = min_delta
        self.patience = patience
        self.verbose = verbose
        self.mode = mode
        self.strict = strict
        self.check_finite = check_finite
        self.stopping_threshold = stopping_threshold
        self.divergence_threshold = divergence_threshold
        self.wait_count = 0
        self.stopped_epoch = 0
        self._check_on_train_epoch_end = check_on_train_epoch_end

        if self.mode not in self.mode_dict:
            raise MisconfigurationException(f"`mode` can be {', '.join(self.mode_dict.keys())}, got {self.mode}")

        self.min_delta *= 1 if self.monitor_op == torch.gt else -1
        torch_inf = torch.tensor(np.Inf)
        self.best_score = torch_inf if self.monitor_op == torch.lt else -torch_inf

        if monitor is None:
            rank_zero_deprecation(
                "The `EarlyStopping(monitor)` argument will be required starting in v1.6."
                " For backward compatibility, setting this to `early_stop_on`."
            )
        self.monitor = monitor or "early_stop_on"

    def _validate_condition_metric(self, logs):
        monitor_val = logs.get(self.monitor)

        error_msg = (
            f"Early stopping conditioned on metric `{self.monitor}` which is not available."
            " Pass in or modify your `EarlyStopping` callback to use any of the following:"
            f' `{"`, `".join(list(logs.keys()))}`'
        )

        if monitor_val is None:
            if self.strict:
                raise RuntimeError(error_msg)
            if self.verbose > 0:
                rank_zero_warn(error_msg, RuntimeWarning)

            return False

        return True

    @property
    def monitor_op(self) -> Callable:
        return self.mode_dict[self.mode]

    def on_save_checkpoint(
        self, trainer: "pl.Trainer", pl_module: "pl.LightningModule", checkpoint: Dict[str, Any]
    ) -> Dict[str, Any]:
        return {
            "wait_count": self.wait_count,
            "stopped_epoch": self.stopped_epoch,
            "best_score": self.best_score,
            "patience": self.patience,
        }

    def on_load_checkpoint(self, callback_state: Dict[str, Any]) -> None:
        self.wait_count = callback_state["wait_count"]
        self.stopped_epoch = callback_state["stopped_epoch"]
        self.best_score = callback_state["best_score"]
        self.patience = callback_state["patience"]

    def _should_skip_check(self, trainer) -> bool:
        from pytorch_lightning.trainer.states import TrainerFn

        return trainer.state.fn != TrainerFn.FITTING or trainer.sanity_checking

    def on_train_epoch_end(self, trainer: "pl.Trainer", pl_module: "pl.LightningModule") -> None:
        if not self._check_on_train_epoch_end or self._should_skip_check(trainer):
            return
        self._run_early_stopping_check(trainer)

    def on_validation_end(self, trainer, pl_module) -> None:
        if self._check_on_train_epoch_end or self._should_skip_check(trainer):
            return
        self._run_early_stopping_check(trainer)

    def _run_early_stopping_check(self, trainer: "pl.Trainer") -> None:
        """
        Checks whether the early stopping condition is met
        and if so tells the trainer to stop the training.
        """
        logs = trainer.callback_metrics

        if trainer.fast_dev_run or not self._validate_condition_metric(  # disable early_stopping with fast_dev_run
            logs
        ):  # short circuit if metric not present
            return

        current = logs.get(self.monitor)

        # when in dev debugging
        trainer.dev_debugger.track_early_stopping_history(self, current)

        should_stop, reason = self._evalute_stopping_criteria(current)

        # stop every ddp process if any world process decides to stop
        should_stop = trainer.training_type_plugin.reduce_boolean_decision(should_stop)
        trainer.should_stop = trainer.should_stop or should_stop
        if should_stop:
            self.stopped_epoch = trainer.current_epoch
        if reason and self.verbose:
            self._log_info(trainer, reason)

    def _evalute_stopping_criteria(self, current: torch.Tensor) -> Tuple[bool, str]:
        should_stop = False
        reason = None
        if self.check_finite and not torch.isfinite(current):
            should_stop = True
            reason = (
                f"Monitored metric {self.monitor} = {current} is not finite."
                f" Previous best value was {self.best_score:.3f}. Signaling Trainer to stop."
            )
        elif self.stopping_threshold is not None and self.monitor_op(current, self.stopping_threshold):
            should_stop = True
            reason = (
                "Stopping threshold reached:"
                f" {self.monitor} = {current} {self.order_dict[self.mode]} {self.stopping_threshold}."
                " Signaling Trainer to stop."
            )
        elif self.divergence_threshold is not None and self.monitor_op(-current, -self.divergence_threshold):
            should_stop = True
            reason = (
                "Divergence threshold reached:"
                f" {self.monitor} = {current} {self.order_dict[self.mode]} {self.divergence_threshold}."
                " Signaling Trainer to stop."
            )
        elif self.monitor_op(current - self.min_delta, self.best_score.to(current.device)):
            should_stop = False
            reason = self._improvement_message(current)
            self.best_score = current
            self.wait_count = 0
        else:
            self.wait_count += 1
            if self.wait_count >= self.patience:
                should_stop = True
                reason = (
                    f"Monitored metric {self.monitor} did not improve in the last {self.wait_count} records."
                    f" Best score: {self.best_score:.3f}. Signaling Trainer to stop."
                )

        return should_stop, reason

    def _improvement_message(self, current: torch.Tensor) -> str:
        """Formats a log message that informs the user about an improvement in the monitored score."""
        if torch.isfinite(self.best_score):
            msg = (
                f"Metric {self.monitor} improved by {abs(self.best_score - current):.3f} >="
                f" min_delta = {abs(self.min_delta)}. New best score: {current:.3f}"
            )
        else:
            msg = f"Metric {self.monitor} improved. New best score: {current:.3f}"
        return msg

    @staticmethod
    def _log_info(trainer: Optional["pl.Trainer"], message: str) -> None:
        if trainer is not None and trainer.world_size > 1:
            log.info(f"[rank: {trainer.global_rank}] {message}")
        else:
            log.info(message)