Lightning-AI/pytorch-lightning · warning · UserWarning

LitLogger does not support `log_graph`

Error message

LitLogger does not support `log_graph`

What it means

LitLogger (local file logger) does not implement graph logging; calling log_graph(model) or Trainer(log_graph=True) with this logger emits a UserWarning and does nothing.

Source

Thrown at src/lightning/pytorch/loggers/litlogger.py:274

    @override
    @rank_zero_only
    def log_hyperparams(
        self,
        params: Union[dict[str, Any], Namespace],
        metrics: Optional[dict[str, Any]] = None,
    ) -> None:
        """Log hyperparams."""
        if isinstance(params, Namespace):
            params = params.__dict__
        experiment = self._require_experiment()
        for key, value in params.items():
            experiment[key] = str(value)

    @override
    @rank_zero_only
    def log_graph(self, model: Module, input_array: Optional[Tensor] = None) -> None:
        warnings.warn("LitLogger does not support `log_graph`", UserWarning, stacklevel=2)

    @override
    @rank_zero_only
    def save(self) -> None:
        pass

    @override
    @rank_zero_only
    def finalize(self, status: Optional[str] = None) -> None:
        if self._experiment is not None:
            # log checkpoints as artifacts before finalizing
            if self._checkpoint_callback:
                self._scan_and_log_checkpoints(self._checkpoint_callback)
            self._experiment.finalize(status)

    # ──────────────────────────────────────────────────────────────────────────────
    # Public methods
    # ──────────────────────────────────────────────────────────────────────────────

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Use TensorBoardLogger (pip install lightning[extra] or tensorboard) for log_graph support
2. Remove log_graph=True from the Logger init when using LitLogger/CSVLogger
3. Ignore the warning — everything else works

Example fix

# before
logger = LitLogger('logs')
# log_graph=True somewhere -> warning
# after
from lightning.pytorch.loggers import TensorBoardLogger
logger = TensorBoardLogger('logs', log_graph=True)

from lightning.pytorch.loggers.tensorboard import _TENSORBOARD_AVAILABLE
if want_graph and not _TENSORBOARD_AVAILABLE:
    raise SystemExit('install tensorboard for log_graph')

When it happens

Trigger: CSV/LitLogger-based Trainer(log_graph=True) (e.g. when tensorboard is not installed, since logger=True falls back to CSV), or explicit logger.log_graph(model).

Common situations: Copying a TensorBoard-oriented config while tensorboard isn't installed; wanting model graph visualization with a file-based logger.

Related errors

• Experiment is not initialized
• `synchronous` requires mlflow>=2.8.0
• NeptuneLogger is no longer supported. Neptune has been sunse
• Providing log_model={log_model} and offline={offline} is an
• Starting from v1.9.0, `tensorboardX` has been removed as a d

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/0ebdfb6c781c650b. Report an issue: GitHub.