Lightning-AI/pytorch-lightning · critical · MisconfigurationException

`configure_optimizers` must include a monitor when a `Reduce

Error message

`configure_optimizers` must include a monitor when a `ReduceLROnPlateau` scheduler is used. For example: {"optimizer": optimizer, "lr_scheduler": scheduler, "monitor": "metric_to_track"}

What it means

ReduceLROnPlateau needs a monitored metric to decide when to reduce the LR. If a ReduceLROnPlateau scheduler is configured without a 'monitor' key, Lightning raises MisconfigurationException with an example of the expected dict.

Source

Thrown at src/lightning/pytorch/core/optimizer.py:293

                "reduce_on_plateau", isinstance(scheduler["scheduler"], optim.lr_scheduler.ReduceLROnPlateau)
            )
            if scheduler["reduce_on_plateau"] and scheduler.get("monitor") is None:
                raise MisconfigurationException(
                    "The lr scheduler dict must include a monitor when a `ReduceLROnPlateau` scheduler is used."
                    ' For example: {"optimizer": optimizer, "lr_scheduler":'
                    ' {"scheduler": scheduler, "monitor": "your_loss"}}'
                )
            is_one_cycle = isinstance(scheduler["scheduler"], optim.lr_scheduler.OneCycleLR)
            if is_one_cycle and scheduler.get("interval", "epoch") == "epoch":
                rank_zero_warn(
                    "A `OneCycleLR` scheduler is using 'interval': 'epoch'."
                    " Are you sure you didn't mean 'interval': 'step'?",
                    category=RuntimeWarning,
                )
            config = LRSchedulerConfig(**scheduler)
        elif isinstance(scheduler, ReduceLROnPlateau):
            if monitor is None:
                raise MisconfigurationException(
                    "`configure_optimizers` must include a monitor when a `ReduceLROnPlateau`"
                    " scheduler is used. For example:"
                    ' {"optimizer": optimizer, "lr_scheduler": scheduler, "monitor": "metric_to_track"}'
                )
            config = LRSchedulerConfig(scheduler, reduce_on_plateau=True, monitor=monitor)
        else:
            config = LRSchedulerConfig(scheduler)
        lr_scheduler_configs.append(config)
    return lr_scheduler_configs


def _configure_schedulers_manual_opt(schedulers: list) -> list[LRSchedulerConfig]:
    """Convert each scheduler into `LRSchedulerConfig` structure with relevant information, when using manual
    optimization."""
    lr_scheduler_configs = []
    for scheduler in schedulers:
        if isinstance(scheduler, dict):
            # interval is not in this list even though the user needs to manually call the scheduler because

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Add 'monitor': '<logged_metric_name>' to the scheduler dict
2. Ensure the metric is actually logged via self.log(monitor, ...) or self.log_dict in validation/training

Example fix

# before
return {'optimizer': opt, 'lr_scheduler': torch.optim.lr_scheduler.ReduceLROnPlateau(opt)}
# after
return {
    'optimizer': opt,
    'lr_scheduler': torch.optim.lr_scheduler.ReduceLROnPlateau(opt),
    'monitor': 'val_loss',
}
# and in validation_step: self.log('val_loss', loss)

from torch.optim.lr_scheduler import ReduceLROnPlateau
if isinstance(scheduler, ReduceLROnPlateau):
    assert monitor, "ReduceLROnPlateau requires a monitor"

def needs_monitor(sched) -> bool:
    from torch.optim.lr_scheduler import ReduceLROnPlateau
    return isinstance(sched, ReduceLROnPlateau)

When it happens

Trigger: Returning {'optimizer': opt, 'lr_scheduler': ReduceLROnPlateau(opt)} with no 'monitor' key.

Common situations: Copy-pasting a StepLR config and swapping in ReduceLROnPlateau without adding a monitor; not logging the metric referenced by monitor.

Understand the failure class

Background: "missing required argument" and "the following required arguments were not provided": what required-argument errors mean and how to fix them — this error's family across 20 libraries.

Related errors

• Early stopping conditioned on metric `{self.monitor}` which
• `ModelCheckpoint(monitor={self.monitor!r})` could not find t
• The lr scheduler dict must have the key "scheduler" with its
• The "interval" key in lr scheduler dict must be "step" or "e
• The provided lr scheduler `{scheduler.__class__.__name__}` i

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/996f89cab72d45d5. Report an issue: GitHub.