Lightning-AI/pytorch-lightning · error · MisconfigurationException

In tuner with method={method!r}, `dataloaders` argument shou

Error message

In tuner with method={method!r}, `dataloaders` argument should be None, please consider setting `train_dataloaders` and `val_dataloaders` instead.

What it means

When the tuner runs with method='fit', the combined `dataloaders` argument is disallowed because training needs distinct train and validation loaders. Passing `dataloaders` raises MisconfigurationException.

Source

Thrown at src/lightning/pytorch/tuner/tuning.py:217

        self._trainer.callbacks = [cb for cb in self._trainer.callbacks if cb is not lr_finder_callback]

        return lr_finder_callback.optimal_lr


def _check_tuner_configuration(
    train_dataloaders: Optional[Union[TRAIN_DATALOADERS, "pl.LightningDataModule"]] = None,
    val_dataloaders: Optional[EVAL_DATALOADERS] = None,
    dataloaders: Optional[EVAL_DATALOADERS] = None,
    method: Literal["fit", "validate", "test", "predict"] = "fit",
) -> None:
    supported_methods = ("fit", "validate", "test", "predict")
    if method not in supported_methods:
        raise ValueError(f"method {method!r} is invalid. Should be one of {supported_methods}.")

    if method == "fit":
        if dataloaders is not None:
            raise MisconfigurationException(
                f"In tuner with method={method!r}, `dataloaders` argument should be None,"
                " please consider setting `train_dataloaders` and `val_dataloaders` instead."
            )
    else:
        if train_dataloaders is not None or val_dataloaders is not None:
            raise MisconfigurationException(
                f"In tuner with `method`={method!r}, `train_dataloaders` and `val_dataloaders`"
                " arguments should be None, please consider setting `dataloaders` instead."
            )


def _check_lr_find_configuration(trainer: "pl.Trainer") -> None:
    # local import to avoid circular import
    from lightning.pytorch.callbacks.lr_finder import LearningRateFinder

    configured_callbacks = [cb for cb in trainer.callbacks if isinstance(cb, LearningRateFinder)]
    if configured_callbacks:
        raise ValueError(

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Pass train_dataloaders=... (and optionally val_dataloaders=...) instead of dataloaders
2. Or let lr_find reuse the loaders already attached to the model/trainer by omitting the argument

Example fix

# before
trainer.tuner.lr_find(model, dataloaders=train_dl)
# after
trainer.tuner.lr_find(model, train_dataloaders=train_dl, val_dataloaders=val_dl)

assert dataloaders is None, "method='fit' requires train_dataloaders/val_dataloaders, not dataloaders"

When it happens

Trigger: trainer.tuner.lr_find(model, dataloaders=dl) or tuner.scale_batch_size(..., dataloaders=dl, method='fit') (method defaults to 'fit').

Common situations: Migrating from an older API that accepted a single dataloaders argument; reusing a validate/test-style call for fitting.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• In tuner with `method`={method!r}, `train_dataloaders` and `
• method='fit' is the only valid configuration to run lr finde
• Device should be CPU, got {device} instead.
• Trying to inject custom `Sampler` into the `{dataloader_cls_
• You are trying to `self.log()` but the loop's result collect

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/bad9b3bcdc478a8d. Report an issue: GitHub.