TensorBoardLogger

class pytorch_lightning.loggers.TensorBoardLogger(save_dir, name='lightning_logs', version=None, log_graph=False, default_hp_metric=True, prefix='', sub_dir=None, **kwargs)

Bases: pytorch_lightning.loggers.logger.Logger, lightning_fabric.loggers.tensorboard.TensorBoardLogger

Log to local file system in TensorBoard format.

Implemented using SummaryWriter. Logs are saved to os.path.join(save_dir, name, version). This is the default logger in Lightning, it comes preinstalled.

Example:

from pytorch_lightning import Trainer
from pytorch_lightning.loggers import TensorBoardLogger

logger = TensorBoardLogger("tb_logs", name="my_model")
trainer = Trainer(logger=logger)

Parameters:

• save_dir (Union[str, Path]) – Save directory

• name (Optional[str]) – Experiment name. Defaults to 'default'. If it is the empty string then no per-experiment subdirectory is used.

• version (Union[int, str, None]) – Experiment version. If version is not specified the logger inspects the save directory for existing versions, then automatically assigns the next available version. If it is a string then it is used as the run-specific subdirectory name, otherwise 'version_${version}' is used.

• log_graph (bool) – Adds the computational graph to tensorboard. This requires that the user has defined the self.example_input_array attribute in their model.

• default_hp_metric (bool) – Enables a placeholder metric with key hp_metric when log_hyperparams is called without a metric (otherwise calls to log_hyperparams without a metric are ignored).

• prefix (str) – A string to put at the beginning of metric keys.

• sub_dir (Union[str, Path, None]) – 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 (Any) – Additional arguments used by tensorboardX.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.

Example

>>> import shutil, tempfile
>>> tmp = tempfile.mkdtemp()
>>> tbl = TensorBoardLogger(tmp)
>>> tbl.log_hyperparams({"epochs": 5, "optimizer": "Adam"})
>>> tbl.log_metrics({"acc": 0.75})
>>> tbl.log_metrics({"acc": 0.9})
>>> tbl.finalize("success")
>>> shutil.rmtree(tmp)

after_save_checkpoint(checkpoint_callback)

Called after model checkpoint callback saves a new checkpoint.

Parameters:

checkpoint_callback (ModelCheckpoint) – the model checkpoint callback instance

Return type:

None

finalize(status)

Do any processing that is necessary to finalize an experiment.

Parameters:

status (str) – Status that the experiment finished with (e.g. success, failed, aborted)

Return type:

None

log_graph(model, input_array=None)

Record model graph.

Parameters:

• model (LightningModule) – the model with an implementation of forward.

• input_array (Optional[Tensor]) – input passes to model.forward

Return type:

None

log_hyperparams(params, metrics=None)

Record hyperparameters. TensorBoard logs with and without saved hyperparameters are incompatible, the hyperparameters are then not displayed in the TensorBoard. Please delete or move the previously saved logs to display the new ones with hyperparameters.

Parameters:

• params (Union[Dict[str, Any], Namespace]) – a dictionary-like container with the hyperparameters

• metrics (Optional[Dict[str, Any]]) – Dictionary with metric names as keys and measured quantities as values

Return type:

None

save()

Save log data.

Return type:

None

property log_dir: str

The directory for this run’s tensorboard checkpoint.

By default, it is named 'version_${self.version}' but it can be overridden by passing a string value for the constructor’s version parameter instead of None or an int.

property root_dir: str

Parent directory for all tensorboard checkpoint subdirectories.

If the experiment name parameter is an empty string, no experiment subdirectory is used and the checkpoint will be saved in “save_dir/version”

property save_dir: str

Gets the save directory where the TensorBoard experiments are saved.

Returns:

The local path to the save directory where the TensorBoard experiments are saved.