Source code for optuna.integration.pytorch_lightning

import warnings

import optuna

with optuna._imports.try_import() as _imports:
    from pytorch_lightning import LightningModule
    from pytorch_lightning import Trainer
    from pytorch_lightning.callbacks import Callback

if not _imports.is_successful():
    Callback = object  # type: ignore # NOQA
    LightningModule = object  # type: ignore # NOQA
    Trainer = object  # type: ignore # NOQA

[docs]class PyTorchLightningPruningCallback(Callback): """PyTorch Lightning callback to prune unpromising trials. See `the example < main/pytorch/>`__ if you want to add a pruning callback which observes accuracy. Args: trial: A :class:`~optuna.trial.Trial` corresponding to the current evaluation of the objective function. monitor: An evaluation metric for pruning, e.g., ``val_loss`` or ``val_acc``. The metrics are obtained from the returned dictionaries from e.g. ``pytorch_lightning.LightningModule.training_step`` or ``pytorch_lightning.LightningModule.validation_epoch_end`` and the names thus depend on how this dictionary is formatted. """ def __init__(self, trial: optuna.trial.Trial, monitor: str) -> None: _imports.check() super().__init__() self._trial = trial self.monitor = monitor def on_validation_end(self, trainer: Trainer, pl_module: LightningModule) -> None: epoch = pl_module.current_epoch current_score = trainer.callback_metrics.get(self.monitor) if current_score is None: message = ( "The metric '{}' is not in the evaluation logs for pruning. " "Please make sure you set the correct metric name.".format(self.monitor) ) warnings.warn(message) return, step=epoch) if self._trial.should_prune(): message = "Trial was pruned at epoch {}.".format(epoch) raise optuna.TrialPruned(message)