Lightning-AI/pytorch-lightning · warning

Starting from v1.9.0, `tensorboardX` has been removed as a d

Error message

Starting from v1.9.0, `tensorboardX` has been removed as a dependency of the `lightning.pytorch` package, due to potential conflicts with other packages in the ML ecosystem. For this reason, `logger=True` will use `CSVLogger` as the default logger, unless the `tensorboard` or `tensorboardX` packages are found. Please `pip install lightning[extra]` or one of them to enable TensorBoard support by default

What it means

Since v1.9.0 tensorboardX is not a dependency of lightning.pytorch. If Trainer(logger=True) (the default) finds neither tensorboard nor tensorboardX installed, it warns and falls back to CSVLogger.

Source

Thrown at src/lightning/pytorch/trainer/connectors/logger_connector/logger_connector.py:76

            # `+ 1` because it can be checked before a step is executed, for example, in `on_train_batch_start`
            step = loop.epoch_loop._batches_that_stepped + 1
        elif isinstance(loop, (pl.loops._EvaluationLoop, pl.loops._PredictionLoop)):
            step = loop.batch_progress.current.ready
        else:
            raise NotImplementedError(loop)
        should_log = step % trainer.log_every_n_steps == 0
        return should_log or trainer.should_stop

    def configure_logger(self, logger: Union[bool, Logger, Iterable[Logger]]) -> None:
        if not logger:
            # logger is None or logger is False
            self.trainer.loggers = []
        elif logger is True:
            # default logger
            if _TENSORBOARD_AVAILABLE or _TENSORBOARDX_AVAILABLE:
                logger_ = TensorBoardLogger(save_dir=self.trainer.default_root_dir, version=SLURMEnvironment.job_id())
            else:
                warning_cache.warn(
                    "Starting from v1.9.0, `tensorboardX` has been removed as a dependency of the `lightning.pytorch`"
                    " package, due to potential conflicts with other packages in the ML ecosystem. For this reason,"
                    " `logger=True` will use `CSVLogger` as the default logger, unless the `tensorboard`"
                    " or `tensorboardX` packages are found."
                    " Please `pip install lightning[extra]` or one of them to enable TensorBoard support by default"
                )
                logger_ = CSVLogger(save_dir=self.trainer.default_root_dir)  # type: ignore[assignment]
            self.trainer.loggers = [logger_]
        elif isinstance(logger, Iterable):
            self.trainer.loggers = list(logger)
        else:
            self.trainer.loggers = [logger]

        if (
            not any(isinstance(logger, LitLogger) for logger in self.trainer.loggers)
            and self.trainer.suggest_integrations
        ):
            rank_zero_info(

View on GitHub (pinned to 9fed5c27d2)

Solutions

pip install lightning[extra] or pip install tensorboard tensorboardX to get TensorBoardLogger by default
Or explicitly pass logger=TensorBoardLogger(...) / CSVLogger(...) to remove ambiguity
Point tensorboard at the CSV fallback or use the CSV logs directly

Example fix

# before
trainer = Trainer()  # warns, uses CSVLogger
# after
# pip install lightning[extra]
trainer = Trainer(logger=TensorBoardLogger('logs'))

Defensive patterns

Strategy: validation

Validate before calling

try:
    import tensorboard  # noqa
    tb = True
except ImportError:
    tb = False
logger = TensorBoardLogger('logs') if tb else CSVLogger('logs')
trainer = Trainer(logger=logger)

Prevention

Always pass an explicit logger instead of relying on logger=True
Install lightning[extra] in images that expect TensorBoard

When it happens

Trigger: Fresh install of lightning without extras; Trainer() with defaults; then checking logs expecting TensorBoard event files.

Common situations: CI environments or slim Docker images without tensorboard; upgrading Lightning and finding logs as CSV instead of tfevents.

Related errors

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/150d76109d9f495a. Report an issue: GitHub.