Deprecate logger=bool

docs/source-pytorch/extensions/logging.rst

@@ -378,7 +378,7 @@ When Lightning creates a checkpoint, it stores a key ``"hyper_parameters"`` with
 
 Some loggers also allow logging the hyperparams used in the experiment. For instance,
 when using the ``TensorBoardLogger``, all hyperparams will show
-in the `hparams tab <https://pytorch.org/docs/stable/tensorboard.html#torch.utils.tensorboard.writer.SummaryWriter.add_hparams>`_.
+in the hparams tab at :meth:`torch.utils.tensorboard.writer.SummaryWriter.add_hparams`.
 
 .. note::
     If you want to track a metric in the tensorboard hparams tab, log scalars to the key ``hp_metric``. If tracking multiple metrics, initialize ``TensorBoardLogger`` with ``default_hp_metric=False`` and call ``log_hyperparams`` only once with your metric keys and initial values. Subsequent updates can simply be logged to the metric keys. Refer to the examples below for setting up proper hyperparams metrics tracking within the :doc:`LightningModule <../common/lightning_module>`.

requirements/pytorch/base.txt

@@ -6,7 +6,6 @@ torch>=1.10.*, <1.13.0
 tqdm>=4.57.0, <4.65.0
 PyYAML>=5.4, <=6.0
 fsspec[http]>2021.06.0, <2022.8.0
-tensorboard>=2.9.1, <2.11.0
 torchmetrics>=0.7.0, <0.10.1  # needed for using fixed compare_version
 packaging>=17.0, <=21.3
 typing-extensions>=4.0.0, <=4.4.0

requirements/pytorch/test.txt

@@ -12,5 +12,6 @@ scikit-learn>0.22.1, <1.1.3
 onnxruntime<1.13.0
 psutil<5.9.4 # for `DeviceStatsMonitor`
 pandas>1.0, <1.5.2  # needed in benchmarks
-fastapi<0.87.0
-uvicorn<0.19.1
+tensorboard>=2.9.1, <2.11.0  # for `TensorBoardLogger`
+fastapi<0.87.0  # for `ServableModuleValidator`
+uvicorn<0.19.1  # for `ServableModuleValidator`

src/pytorch_lightning/callbacks/device_stats_monitor.py

@@ -64,7 +64,7 @@ def setup(
             return
 
         if not trainer.loggers:
-            raise MisconfigurationException("Cannot use `DeviceStatsMonitor` callback with `Trainer(logger=False)`.")
+            raise MisconfigurationException("Cannot use `DeviceStatsMonitor` callback with `Trainer(logger=None)`.")
 
         # warn in setup to warn once
         device = trainer.strategy.root_device

src/pytorch_lightning/cli.py

@@ -25,6 +25,7 @@
 import pytorch_lightning as pl
 from lightning_lite.utilities.cloud_io import get_filesystem
 from pytorch_lightning import Callback, LightningDataModule, LightningModule, seed_everything, Trainer
+from pytorch_lightning.trainer.connectors.logger_connector.logger_connector import _NoneSentinel
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.model_helpers import is_overridden
 from pytorch_lightning.utilities.rank_zero import rank_zero_deprecation, rank_zero_warn
@@ -550,9 +551,17 @@ def _instantiate_trainer(self, config: Dict[str, Any], callbacks: List[Callback]
                 value = self.trainer_defaults[key]
                 config[key] += value if isinstance(value, list) else [value]
             if self.save_config_callback and not config.get("fast_dev_run", False):
+                to_save = self.config.get(str(self.subcommand), self.config)
+                # TODO: remove this override in v2.0.0
+                if (
+                    "trainer" in to_save
+                    and "logger" in to_save.trainer
+                    and isinstance(to_save.trainer.logger, _NoneSentinel)
+                ):
+                    to_save.trainer.logger = None
                 config_callback = self.save_config_callback(
                     self._parser(self.subcommand),
-                    self.config.get(str(self.subcommand), self.config),
+                    to_save,
                     **self.save_config_kwargs,
                 )
                 config[key].append(config_callback)

src/pytorch_lightning/demos/boring_classes.py

@@ -156,11 +156,8 @@ def predict_dataloader(self) -> DataLoader:
 
 
 class BoringDataModule(LightningDataModule):
-    def __init__(self, data_dir: str = "./"):
+    def __init__(self) -> None:
         super().__init__()
-        self.data_dir = data_dir
-        self.non_picklable = None
-        self.checkpoint_state: Optional[str] = None
         self.random_full = RandomDataset(32, 64 * 4)
 
     def setup(self, stage: str) -> None:

src/pytorch_lightning/loggers/tensorboard.py

@@ -19,12 +19,11 @@
 import logging
 import os
 from argparse import Namespace
-from typing import Any, Dict, Mapping, Optional, Union
+from typing import Any, Dict, Mapping, Optional, TYPE_CHECKING, Union
 
 import numpy as np
+from lightning_utilities.core.imports import RequirementCache
 from torch import Tensor
-from torch.utils.tensorboard import SummaryWriter
-from torch.utils.tensorboard.summary import hparams
 
 import pytorch_lightning as pl
 from lightning_lite.utilities.cloud_io import get_filesystem
@@ -34,19 +33,23 @@
 from pytorch_lightning.utilities.imports import _OMEGACONF_AVAILABLE
 from pytorch_lightning.utilities.logger import _add_prefix, _convert_params, _flatten_dict
 from pytorch_lightning.utilities.logger import _sanitize_params as _utils_sanitize_params
-from pytorch_lightning.utilities.rank_zero import rank_zero_only, rank_zero_warn
+from pytorch_lightning.utilities.rank_zero import rank_zero_deprecation, rank_zero_only, rank_zero_warn
 
 log = logging.getLogger(__name__)
 
 if _OMEGACONF_AVAILABLE:
     from omegaconf import Container, OmegaConf
 
+_TENSORBOARD_AVAILABLE = RequirementCache("tensorboard>=2.9.1")
+if TYPE_CHECKING and _TENSORBOARD_AVAILABLE:
+    from torch.utils.tensorboard import SummaryWriter
+
 
 class TensorBoardLogger(Logger):
     r"""
     Log to local file system in `TensorBoard <https://www.tensorflow.org/tensorboard>`_ format.
 
-    Implemented using :class:`~torch.utils.tensorboard.SummaryWriter`. Logs are saved to
+    Implemented using :class:`torch.utils.tensorboard.SummaryWriter`. Logs are saved to
     ``os.path.join(save_dir, name, version)``. This is the default logger in Lightning, it comes
     preinstalled.
 
@@ -77,7 +80,7 @@ class TensorBoardLogger(Logger):
         sub_dir: Sub-directory to group TensorBoard logs. If a sub_dir argument is passed
             then logs are saved in ``/save_dir/name/version/sub_dir/``. Defaults to ``None`` in which
             logs are saved in ``/save_dir/name/version/``.
-        \**kwargs: Additional arguments used by :class:`SummaryWriter` can be passed as keyword
+        \**kwargs: Additional arguments used by :class:`torch.utils.tensorboard.SummaryWriter` can be passed as keyword
             arguments in this logger. To automatically flush to disk, `max_queue` sets the size
             of the queue for pending logs before flushing. `flush_secs` determines how many seconds
             elapses before flushing.
@@ -97,6 +100,17 @@ def __init__(
         sub_dir: Optional[_PATH] = None,
         **kwargs: Any,
     ):
+        if not _TENSORBOARD_AVAILABLE:
+            # TODO: replace this with `raise ModuleNotFoundError` in v2.0.0
+            rank_zero_deprecation(
+                "`tensorboard` has been removed as a dependency of `pytorch_lightning` in v1.9.0. Until v2.0.0 we will"
+                f" try to install it for you automatically. This deprecation is caused by {_TENSORBOARD_AVAILABLE!s}"
+            )
+            import pip
+
+            retcode = pip.main(["install", _TENSORBOARD_AVAILABLE.requirement])
+            assert retcode
+
         super().__init__()
         save_dir = os.fspath(save_dir)
         self._save_dir = save_dir
@@ -157,7 +171,7 @@ def sub_dir(self) -> Optional[str]:
 
     @property
     @rank_zero_experiment
-    def experiment(self) -> SummaryWriter:
+    def experiment(self) -> "SummaryWriter":
         r"""
         Actual tensorboard object. To use TensorBoard features in your
         :class:`~pytorch_lightning.core.module.LightningModule` do the following.
@@ -173,6 +187,8 @@ def experiment(self) -> SummaryWriter:
         assert rank_zero_only.rank == 0, "tried to init log dirs in non global_rank=0"
         if self.root_dir:
             self._fs.makedirs(self.root_dir, exist_ok=True)
+        from torch.utils.tensorboard import SummaryWriter
+
         self._experiment = SummaryWriter(log_dir=self.log_dir, **self._kwargs)
         return self._experiment
 
@@ -209,6 +225,8 @@ def log_hyperparams(
 
         if metrics:
             self.log_metrics(metrics, 0)
+            from torch.utils.tensorboard.summary import hparams
+
             exp, ssi, sei = hparams(params, metrics)
             writer = self.experiment._get_file_writer()
             writer.add_summary(exp)

src/pytorch_lightning/trainer/connectors/logger_connector/logger_connector.py

@@ -22,6 +22,15 @@
 from pytorch_lightning.loggers import Logger, TensorBoardLogger
 from pytorch_lightning.trainer.connectors.logger_connector.result import _METRICS, _OUT_DICT, _PBAR_DICT
 from pytorch_lightning.utilities.metrics import metrics_to_scalars
+from pytorch_lightning.utilities.rank_zero import rank_zero_deprecation
+
+
+# TODO: remove in v2.0.0
+class _NoneSentinel:
+    """Used as a sentinel value for ``None`` in the depreaction for ``Trainer(logger=bool)``.
+
+    Remove this class with the deprecation.
+    """
 
 
 class LoggerConnector:
@@ -37,7 +46,7 @@ def __init__(self, trainer: "pl.Trainer") -> None:
 
     def on_trainer_init(
         self,
-        logger: Union[bool, Logger, Iterable[Logger]],
+        logger: Optional[Union[Logger, Iterable[Logger], _NoneSentinel]],
         log_every_n_steps: int,
         move_metrics_to_cpu: bool,
     ) -> None:
@@ -51,12 +60,22 @@ def should_update_logs(self) -> bool:
         should_log = (self.trainer.fit_loop.epoch_loop._batches_that_stepped + 1) % self.trainer.log_every_n_steps == 0
         return should_log or self.trainer.should_stop
 
-    def configure_logger(self, logger: Union[bool, Logger, Iterable[Logger]]) -> None:
-        if not logger:
-            # logger is None or logger is False
+    def configure_logger(self, logger: Optional[Union[Logger, Iterable[Logger], _NoneSentinel]]) -> None:
+        if isinstance(logger, _NoneSentinel) or not logger:
+            if logger is False:
+                # TODO: remove in v2.0.0
+                rank_zero_deprecation(
+                    "`Trainer(logger=False)` has been deprecated in favor of `Trainer(logger=None)` in v1.9.0 and will"
+                    " be removed in v2.0.0."
+                )
             self.trainer.loggers = []
         elif logger is True:
-            # default logger
+            # TODO: remove in v2.0.0
+            rank_zero_deprecation(
+                "`Trainer(logger=True)` has been deprecated in favor of `Trainer(logger=TensorBoardLogger())` in v1.9.0"
+                " and will be removed in v2.0.0. Additionally, the `tensorboard` dependency will not be installed"
+                " when installing `pytorch_lightning`."
+            )
             self.trainer.loggers = [
                 TensorBoardLogger(save_dir=self.trainer.default_root_dir, version=SLURMEnvironment.job_id())
             ]

src/pytorch_lightning/trainer/trainer.py

@@ -72,7 +72,7 @@
 from pytorch_lightning.trainer.connectors.callback_connector import CallbackConnector
 from pytorch_lightning.trainer.connectors.checkpoint_connector import CheckpointConnector
 from pytorch_lightning.trainer.connectors.data_connector import DataConnector
-from pytorch_lightning.trainer.connectors.logger_connector import LoggerConnector
+from pytorch_lightning.trainer.connectors.logger_connector.logger_connector import _NoneSentinel, LoggerConnector
 from pytorch_lightning.trainer.connectors.logger_connector.result import _OUT_DICT, _PBAR_DICT, _ResultCollection
 from pytorch_lightning.trainer.connectors.signal_connector import SignalConnector
 from pytorch_lightning.trainer.states import RunningStage, TrainerFn, TrainerState, TrainerStatus
@@ -112,7 +112,8 @@ class Trainer:
     @_defaults_from_env_vars
     def __init__(
         self,
-        logger: Union[Logger, Iterable[Logger], bool] = True,
+        # TODO: Replace default for `None` in 2.0
+        logger: Optional[Union[Logger, Iterable[Logger], _NoneSentinel]] = _NoneSentinel(),
         enable_checkpointing: bool = True,
         callbacks: Optional[Union[List[Callback], Callback]] = None,
         default_root_dir: Optional[_PATH] = None,
@@ -270,11 +271,10 @@ def __init__(
             limit_predict_batches: How much of prediction dataset to check (float = fraction, int = num_batches).
                 Default: ``1.0``.
 
-            logger: Logger (or iterable collection of loggers) for experiment tracking. A ``True`` value uses
-                the default ``TensorBoardLogger``. ``False`` will disable logging. If multiple loggers are
-                provided, local files (checkpoints, profiler traces, etc.) are saved in the ``log_dir`` of
-                the first logger.
-                Default: ``True``.
+            logger: Logger (or iterable collection of loggers) for experiment tracking. ``None`` will disable logging.
+                If multiple loggers are provided, local files (checkpoints, profiler traces, etc.) are saved in the
+                ``log_dir`` of the first logger.
+                Default: ``None``.
 
             log_every_n_steps: How often to log within steps.
                 Default: ``50``.

src/pytorch_lightning/utilities/argparse.py

@@ -51,7 +51,7 @@ def from_argparse_args(
         >>> parser = Trainer.add_argparse_args(parser)
         >>> parser.add_argument('--my_custom_arg', default='something')  # doctest: +SKIP
         >>> args = Trainer.parse_argparser(parser.parse_args(""))
-        >>> trainer = Trainer.from_argparse_args(args, logger=False)
+        >>> trainer = Trainer.from_argparse_args(args)
     """
     if isinstance(args, ArgumentParser):
         args = cls.parse_argparser(args)

tests/tests_pytorch/benchmarks/test_basic_parity.py

@@ -160,7 +160,6 @@ def lightning_loop(cls_model, idx, device_type: str = "cuda", num_epochs=10):
         enable_checkpointing=False,
         accelerator="gpu" if device_type == "cuda" else "cpu",
         devices=1,
-        logger=False,
         replace_sampler_ddp=False,
         benchmark=False,
     )

tests/tests_pytorch/callbacks/progress/test_rich_progress_bar.py

@@ -189,7 +189,6 @@ def test_rich_progress_bar_leave(tmpdir, leave, reset_call_count):
             limit_val_batches=0,
             max_epochs=4,
             callbacks=progress_bar,
-            logger=False,
             enable_checkpointing=False,
             enable_model_summary=False,
         )
@@ -429,7 +428,6 @@ def test_rich_progress_bar_can_be_pickled():
         limit_val_batches=1,
         limit_test_batches=1,
         limit_predict_batches=1,
-        logger=False,
         enable_model_summary=False,
     )
     model = BoringModel()

tests/tests_pytorch/callbacks/progress/test_tqdm_progress_bar.py

@@ -315,7 +315,6 @@ def on_validation_epoch_end(self, *args):
         limit_train_batches=1,
         limit_val_batches=limit_val_batches,
         callbacks=[pbar],
-        logger=False,
         enable_checkpointing=False,
     )
     trainer.fit(model)
@@ -371,7 +370,6 @@ def test_main_progress_bar_update_amount(
         limit_train_batches=train_batches,
         limit_val_batches=val_batches,
         callbacks=[progress_bar],
-        logger=False,
         enable_checkpointing=False,
     )
     with mock.patch("pytorch_lightning.callbacks.progress.tqdm_progress.Tqdm", MockTqdm):
@@ -392,7 +390,6 @@ def test_test_progress_bar_update_amount(tmpdir, test_batches: int, refresh_rate
         max_epochs=1,
         limit_test_batches=test_batches,
         callbacks=[progress_bar],
-        logger=False,
         enable_checkpointing=False,
     )
     with mock.patch("pytorch_lightning.callbacks.progress.tqdm_progress.Tqdm", MockTqdm):
@@ -410,9 +407,7 @@ def training_step(self, batch, batch_idx):
             self.log("c", {"c1": 2}, prog_bar=True, on_epoch=False)
             return super().training_step(batch, batch_idx)
 
-    trainer = Trainer(
-        default_root_dir=tmpdir, max_epochs=1, limit_train_batches=2, logger=False, enable_checkpointing=False
-    )
+    trainer = Trainer(default_root_dir=tmpdir, max_epochs=1, limit_train_batches=2, enable_checkpointing=False)
     trainer.fit(TestModel())
 
     torch.testing.assert_close(trainer.progress_bar_metrics["a"], 0.123)
@@ -546,7 +541,6 @@ def test_tqdm_progress_bar_can_be_pickled():
         limit_val_batches=1,
         limit_test_batches=1,
         limit_predict_batches=1,
-        logger=False,
         enable_model_summary=False,
     )
     model = BoringModel()

tests/tests_pytorch/callbacks/test_device_stats_monitor.py

@@ -137,7 +137,6 @@ def test_device_stats_monitor_no_logger(tmpdir):
         default_root_dir=tmpdir,
         callbacks=[device_stats],
         max_epochs=1,
-        logger=False,
         enable_checkpointing=False,
         enable_progress_bar=False,
     )

tests/tests_pytorch/callbacks/test_early_stopping.py

@@ -308,7 +308,6 @@ def training_step(self, batch, batch_idx):
         limit_train_batches=limit_train_batches,
         min_epochs=min_epochs,
         min_steps=min_steps,
-        logger=False,
         enable_checkpointing=False,
         enable_progress_bar=False,
         enable_model_summary=False,

tests/tests_pytorch/callbacks/test_lr_monitor.py

@@ -192,7 +192,7 @@ def test_lr_monitor_no_logger(tmpdir):
     model = BoringModel()
 
     lr_monitor = LearningRateMonitor()
-    trainer = Trainer(default_root_dir=tmpdir, max_epochs=1, callbacks=[lr_monitor], logger=False)
+    trainer = Trainer(default_root_dir=tmpdir, max_epochs=1, callbacks=[lr_monitor])
 
     with pytest.raises(MisconfigurationException, match="`Trainer` that has no logger"):
         trainer.fit(model)

tests/tests_pytorch/callbacks/test_pruning.py

@@ -108,7 +108,6 @@ def train_with_pruning_callback(
         enable_progress_bar=False,
         enable_model_summary=False,
         enable_checkpointing=False,
-        logger=False,
         limit_train_batches=10,
         limit_val_batches=2,
         max_epochs=10,
@@ -214,7 +213,6 @@ def apply_lottery_ticket_hypothesis(self):
         enable_progress_bar=False,
         enable_model_summary=False,
         enable_checkpointing=False,
-        logger=False,
         limit_train_batches=10,
         limit_val_batches=2,
         max_epochs=5,
@@ -241,7 +239,6 @@ def test_multiple_pruning_callbacks(tmpdir, caplog, make_pruning_permanent: bool
         enable_progress_bar=False,
         enable_model_summary=False,
         enable_checkpointing=False,
-        logger=False,
         limit_train_batches=10,
         limit_val_batches=2,
         max_epochs=3,