early_stopper.py

import torch


class EarlyStopping:
    """Early stops the training if validation loss doesn't improve after a given patience."""

    def __init__(
        self,
        path: str,
        patience=7,
        verbose=False,
        delta=0,
        trace_func=print,
        load_best: bool = True,
    ):
        """
        Args:
            patience (int): How long to wait after last time validation loss improved.
                            Default: 7
            verbose (bool): If True, prints a message for each validation loss improvement.
                            Default: False
            delta (float): Minimum change in the monitored quantity to qualify as an improvement.
                            Default: 0
            path (str): Path for the checkpoint to be saved to.
                            Default: 'checkpoint.pt'
            trace_func (function): trace print function.
                            Default: print
        """
        self.patience = patience
        self.verbose = verbose
        self.counter = 0
        self.best_score = None
        self.early_stop = False
        self.val_loss_min = -torch.inf
        self.delta = delta
        self.path = path
        self.trace_func = trace_func
        self.load_best = load_best

    def __call__(self, val_loss, model):
        score = val_loss

        if self.best_score is None:
            self.best_score = score
            self.save_checkpoint(val_loss, model)
        elif score <= self.best_score + self.delta:
            self.counter += 1
            self.trace_func(
                f"EarlyStopping counter: {self.counter} out of {self.patience}"
            )
            if self.counter >= self.patience:
                self.early_stop = True
        else:
            self.best_score = score
            self.save_checkpoint(val_loss, model)
            self.counter = 0

    def save_checkpoint(self, val_loss, model):
        """Saves model when validation loss decrease."""
        if self.verbose:
            self.trace_func(
                f"Validation loss increased ({self.val_loss_min:.6f} --> {val_loss:.6f}).  Saving model ..."
            )
        torch.save(model.state_dict(), self.path)
        self.val_loss_min = val_loss

    def load_checkpoint(self, model):
        """Reloads best model when called"""
        if self.verbose:
            self.trace_func(f"Early stopping occurred. Will reload the best model...")
        model.load_state_dict(torch.load(self.path))
        return model