Shortcuts

Logging

Lightning supports the most popular logging frameworks (TensorBoard, Comet, etc…). To use a logger, simply pass it into the Trainer. Lightning uses TensorBoard by default.

from pytorch_lightning import loggers as pl_loggers

tb_logger = pl_loggers.TensorBoardLogger('logs/')
trainer = Trainer(logger=tb_logger)

Choose from any of the others such as MLflow, Comet, Neptune, WandB, …

comet_logger = pl_loggers.CometLogger(save_dir='logs/')
trainer = Trainer(logger=comet_logger)

To use multiple loggers, simply pass in a list or tuple of loggers …

tb_logger = pl_loggers.TensorBoardLogger('logs/')
comet_logger = pl_loggers.CometLogger(save_dir='logs/')
trainer = Trainer(logger=[tb_logger, comet_logger])

Note

By default, lightning logs every 50 steps. Use Trainer flags to Control logging frequency.

Note

All loggers log by default to os.getcwd(). To change the path without creating a logger set Trainer(default_root_dir=’/your/path/to/save/checkpoints’)


Logging from a LightningModule

Lightning offers automatic log functionalities for logging scalars, or manual logging for anything else.

Automatic Logging

Use the log() method to log from anywhere in a lightning module and callbacks except functions with batch_start in their names.

def training_step(self, batch, batch_idx):
    self.log('my_metric', x)

Depending on where log is called from, Lightning auto-determines the correct logging mode for you. But of course you can override the default behavior by manually setting the log() parameters.

def training_step(self, batch, batch_idx):
    self.log('my_loss', loss, on_step=True, on_epoch=True, prog_bar=True, logger=True)

The log() method has a few options:

  • on_step: Logs the metric at the current step. Defaults to True in training_step(), and training_step_end().

  • on_epoch: Automatically accumulates and logs at the end of the epoch. Defaults to True anywhere in validation or test loops, and in training_epoch_end().

  • prog_bar: Logs to the progress bar.

  • logger: Logs to the logger like Tensorboard, or any other custom logger passed to the Trainer.

Note

  • Setting on_epoch=True will cache all your logged values during the full training epoch and perform a reduction in on_train_epoch_end. We recommend using the metrics API when working with custom reduction.

  • Setting both on_step=True and on_epoch=True will create two keys per metric you log with suffix _step and _epoch, respectively. You can refer to these keys e.g. in the monitor argument of ModelCheckpoint or in the graphs plotted to the logger of your choice.

If your work requires to log in an unsupported function, please open an issue with a clear description of why it is blocking you.

Manual logging

If you want to log anything that is not a scalar, like histograms, text, images, etc… you may need to use the logger object directly.

def training_step(...):
    ...
    # the logger you used (in this case tensorboard)
    tensorboard = self.logger.experiment
    tensorboard.add_image()
    tensorboard.add_histogram(...)
    tensorboard.add_figure(...)

Access your logs

Once your training starts, you can view the logs by using your favorite logger or booting up the Tensorboard logs:

tensorboard --logdir ./lightning_logs

Make a custom logger

You can implement your own logger by writing a class that inherits from LightningLoggerBase. Use the rank_zero_experiment() and rank_zero_only() decorators to make sure that only the first process in DDP training creates the experiment and logs the data respectively.

from pytorch_lightning.utilities import rank_zero_only
from pytorch_lightning.loggers import LightningLoggerBase
from pytorch_lightning.loggers.base import rank_zero_experiment

class MyLogger(LightningLoggerBase):

    @property
    def name(self):
        return 'MyLogger'

    @property
    @rank_zero_experiment
    def experiment(self):
        # Return the experiment object associated with this logger.
        pass

    @property
    def version(self):
        # Return the experiment version, int or str.
        return '0.1'

    @rank_zero_only
    def log_hyperparams(self, params):
        # params is an argparse.Namespace
        # your code to record hyperparameters goes here
        pass

    @rank_zero_only
    def log_metrics(self, metrics, step):
        # metrics is a dictionary of metric names and values
        # your code to record metrics goes here
        pass

    @rank_zero_only
    def save(self):
        # Optional. Any code necessary to save logger data goes here
        # If you implement this, remember to call `super().save()`
        # at the start of the method (important for aggregation of metrics)
        super().save()

    @rank_zero_only
    def finalize(self, status):
        # Optional. Any code that needs to be run after training
        # finishes goes here
        pass

If you write a logger that may be useful to others, please send a pull request to add it to Lightning!


Control logging frequency

Logging frequency

It may slow training down to log every single batch. By default, Lightning logs every 50 rows, or 50 training steps. To change this behaviour, set the log_every_n_steps Trainer flag.

k = 10
trainer = Trainer(log_every_n_steps=k)

Log writing frequency

Writing to a logger can be expensive, so by default Lightning write logs to disc or to the given logger every 100 training steps. To change this behaviour, set the interval at which you wish to flush logs to the filesystem using log_every_n_steps Trainer flag.

k = 100
trainer = Trainer(flush_logs_every_n_steps=k)

Unlike the log_every_n_steps, this argument does not apply to all loggers. The example shown here works with TensorBoardLogger, which is the default logger in Lightning.


Progress Bar

You can add any metric to the progress bar using log() method, setting prog_bar=True.

def training_step(self, batch, batch_idx):
    self.log('my_loss', loss, prog_bar=True)

Modifying the progress bar

The progress bar by default already includes the training loss and version number of the experiment if you are using a logger. These defaults can be customized by overriding the get_progress_bar_dict() hook in your module.

def get_progress_bar_dict(self):
    # don't show the version number
    items = super().get_progress_bar_dict()
    items.pop("v_num", None)
    return items

Configure console logging

Lightning logs useful information about the training process and user warnings to the console. You can retrieve the Lightning logger and change it to your liking. For example, adjust the logging level or redirect output for certain modules to log files:

import logging

# configure logging at the root level of lightning
logging.getLogger("pytorch_lightning").setLevel(logging.ERROR)

# configure logging on module level, redirect to file
logger = logging.getLogger("pytorch_lightning.core")
logger.addHandler(logging.FileHandler("core.log"))

Read more about custom Python logging here.


Logging hyperparameters

When training a model, it’s useful to know what hyperparams went into that model. When Lightning creates a checkpoint, it stores a key “hyper_parameters” with the hyperparams.

lightning_checkpoint = torch.load(filepath, map_location=lambda storage, loc: storage)
hyperparams = lightning_checkpoint['hyper_parameters']

Some loggers also allow logging the hyperparams used in the experiment. For instance, when using the TestTubeLogger or the TensorBoardLogger, all hyperparams will show in the hparams tab.

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 following for examples on how to setup proper hyperparams metrics tracking within LightningModule.

# Using default_hp_metric
def validation_step(self, batch, batch_idx):
    self.log("hp_metric", some_scalar)

# Using custom or multiple metrics (default_hp_metric=False)
def on_train_start(self):
    self.logger.log_hyperparams(self.hparams, {"hp/metric_1": 0, "hp/metric_2": 0})

def validation_step(self, batch, batch_idx):
    self.log("hp/metric_1", some_scalar_1)
    self.log("hp/metric_2", some_scalar_2)

In the example, using hp/ as a prefix allows for the metrics to be grouped under “hp” in the tensorboard scalar tab where you can collapse them.


Snapshot code

Loggers also allow you to snapshot a copy of the code used in this experiment. For example, TestTubeLogger does this with a flag:

from pytorch_lightning.loggers import TestTubeLogger
logger = TestTubeLogger('.', create_git_tag=True)

Supported Loggers

The following are loggers we support

Note

The following loggers will normally plot an additional chart (global_step VS epoch).

Note

postfix _step and _epoch will be appended to the name you logged if on_step and on_epoch are set to True in self.log().

Note

Depending on the loggers you use, there might be some additional charts.

CometLogger

Log using Comet.ml.

CSVLogger

Log to local file system in yaml and CSV format.

MLFlowLogger

Log using MLflow.

NeptuneLogger

Log using Neptune.

TensorBoardLogger

Log to local file system in TensorBoard format.

TestTubeLogger

Log to local file system in TensorBoard format but using a nicer folder structure (see full docs).

WandbLogger

Log using Weights and Biases.