Lightning-AI/pytorch-lightning · error · MisconfigurationException

`train_dataloader` must be implemented to be used with the L

Error message

`train_dataloader` must be implemented to be used with the Lightning Trainer

What it means

The LightningModule/LightningDataModule base implementation of train_dataloader is a stub that raises MisconfigurationException. It exists so that calling fit() without the user overriding train_dataloader fails loudly instead of silently returning None.

Source

Thrown at src/lightning/pytorch/core/hooks.py:483

        For data processing use the following pattern:

            - download in :meth:`prepare_data`
            - process and split in :meth:`setup`

        However, the above are only necessary for distributed processing.

        .. warning:: do not assign state in prepare_data

        - :meth:`~lightning.pytorch.trainer.trainer.Trainer.fit`
        - :meth:`prepare_data`
        - :meth:`setup`

        Note:
            Lightning tries to add the correct sampler for distributed and arbitrary hardware.
            There is no need to set it yourself.

        """
        raise MisconfigurationException("`train_dataloader` must be implemented to be used with the Lightning Trainer")

    def test_dataloader(self) -> EVAL_DATALOADERS:
        r"""An iterable or collection of iterables specifying test samples.

        For more information about multiple dataloaders, see this :ref:`section <multiple-dataloaders>`.

        For data processing use the following pattern:

            - download in :meth:`prepare_data`
            - process and split in :meth:`setup`

        However, the above are only necessary for distributed processing.

        .. warning:: do not assign state in prepare_data


        - :meth:`~lightning.pytorch.trainer.trainer.Trainer.test`
        - :meth:`prepare_data`

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Implement def train_dataloader(self) in your LightningModule or DataModule returning a DataLoader/iterable
2. Or pass the loader directly: trainer.fit(model, train_dataloaders=train_dl)
3. Check spelling/signature — the override must be exactly train_dataloader

Example fix

# before
class MyModule(LightningModule):
    def training_step(self, batch, batch_idx): ...
    # no train_dataloader -> MisconfigurationException
# after
class MyModule(LightningModule):
    def training_step(self, batch, batch_idx): ...
    def train_dataloader(self):
        return DataLoader(self.dataset, batch_size=32)

def has_train_dataloader(obj) -> bool:
    from lightning.pytorch.cli import LightningModule  # or core
    m = type(obj).train_dataloader
    return getattr(m, '__owner__', None) is not type(obj) or 'train_dataloader' in type(obj).__dict__
# simplest robust check:
assert 'train_dataloader' in MyModule.__dict__ or train_dl is not None

def implements_train_dataloader(cls) -> bool:
    """True if cls itself (not the base stub) defines train_dataloader."""
    return 'train_dataloader' in cls.__dict__ or any('train_dataloader' in c.__dict__ for c in cls.__mro__[1:-1] if c.__name__ != 'Hooks')

When it happens

Trigger: trainer.fit(model) where the LightningDataModule/Module never defines train_dataloader and no train_dataloaders= argument was passed to fit.

Common situations: Forgetting to implement train_dataloader when switching from a module that only does predict/test; typos in the method name (train_dataloaders) so the override isn't picked up; passing datamodule=None accidentally.

Related errors

• `test_dataloader` must be implemented to be used with the Li
• The `CSVLogger` does not yet support logging hyperparameters
• DeepSpeed handles gradient clipping automatically within the
• The `{type(self).__name__}` does not use the `CheckpointIO`
• The `{type(self).__name__}` does not support setting a `Chec

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/0b838e7c7aca91bd. Report an issue: GitHub.