Lightning-AI/pytorch-lightning · error · ValueError

A dataloader should be passed only once to the `setup_datalo

Error message

A dataloader should be passed only once to the `setup_dataloaders` method.

What it means

Raised when a dataloader passed to `fabric.setup_dataloaders()` is already a `_FabricDataLoader`, i.e. it was already wrapped by a previous setup call. Double-wrapping would break iteration and distributed sharding, so Fabric rejects it.

Source

Thrown at src/lightning/fabric/fabric.py:1250

            raise ValueError("`setup_optimizers` requires at least one optimizer as input.")

        if any(isinstance(opt, _FabricOptimizer) for opt in optimizers):
            raise ValueError("An optimizer should be passed only once to the `setup_optimizers` method.")

        if any(_has_meta_device_parameters_or_buffers(optimizer) for optimizer in optimizers):
            raise RuntimeError(
                "The optimizer has references to the model's meta-device parameters. Materializing them is"
                " is currently not supported. Create the optimizer after setting up the model, then call"
                " `fabric.setup_optimizers(optimizer)`."
            )

    def _validate_setup_dataloaders(self, dataloaders: Sequence[DataLoader]) -> None:
        self._validate_launched()
        if not dataloaders:
            raise ValueError("`setup_dataloaders` requires at least one dataloader as input.")

        if any(isinstance(dl, _FabricDataLoader) for dl in dataloaders):
            raise ValueError("A dataloader should be passed only once to the `setup_dataloaders` method.")

        if any(not isinstance(dl, DataLoader) for dl in dataloaders):
            raise TypeError("Only PyTorch DataLoader are currently supported in `setup_dataloaders`.")

    @staticmethod
    def _configure_callbacks(callbacks: Optional[Union[list[Any], Any]]) -> list[Any]:
        callbacks = callbacks if callbacks is not None else []
        callbacks = callbacks if isinstance(callbacks, list) else [callbacks]
        callbacks.extend(_load_external_callbacks("lightning.fabric.callbacks_factory"))
        return callbacks

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Keep the raw `DataLoader` objects and pass each only once to setup_dataloaders; use the wrapped return value for training/evaluation only
2. Consolidate into a single call: `train_dl, val_dl = fabric.setup_dataloaders(train_dl, val_dl)`
3. Check helper utilities for hidden setup calls

Example fix

# before
train_dl = fabric.setup_dataloaders(train_dl)
train_dl, val_dl = fabric.setup_dataloaders(train_dl, val_dl)  # ValueError

# after
train_dl, val_dl = fabric.setup_dataloaders(train_dl, val_dl)

from lightning.fabric.utilities.data import _FabricDataLoader
assert all(not isinstance(dl, _FabricDataLoader) for dl in dls), 'already wrapped'

from torch.utils.data import DataLoader
from lightning.fabric.utilities.data import _FabricDataLoader
def is_raw_dataloader(dl) -> bool:
    return isinstance(dl, DataLoader) and not isinstance(dl, _FabricDataLoader)

When it happens

Trigger: Calling `fabric.setup_dataloaders(dl)` twice with the same dataloader; keeping a reference to the wrapped return value and passing it to a second setup call (e.g. setting up train and val loaders that share one already-wrapped loader).

Common situations: Setting up multiple phases (train/val/test) in separate calls and accidentally reusing an already-wrapped loader; helper functions that call setup_dataloaders internally being called on already-prepared inputs.

Related errors

• An optimizer should be passed only once to the `setup_optimi
• `setup_dataloaders` requires at least one dataloader as inpu
• Only PyTorch DataLoader are currently supported in `setup_da
• A model should be passed only once to the `setup` method.
• An optimizer should be passed only once to the `setup` metho

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/f660bdab0014b6d9. Report an issue: GitHub.