LRMonitor#

class composer.callbacks.LRMonitor[source]#

Logs the learning rate.

This callback iterates over all optimizers and their parameter groups to log learning rate under the lr-{OPTIMIZER_NAME}/group{GROUP_NUMBER} key.

Example

>>> from composer import Trainer
>>> from composer.callbacks import LRMonitor
>>> # constructing trainer object with this callback
>>> trainer = Trainer(
...     model=model,
...     train_dataloader=train_dataloader,
...     eval_dataloader=eval_dataloader,
...     optimizers=optimizer,
...     max_duration="1ep",
...     callbacks=[LRMonitor()],
... )

The learning rate is logged by the Logger to the following key as described below.

Key

Logged data

lr-{OPTIMIZER_NAME}/group{GROUP_NUMBER}

Learning rate for each optimizer and parameter group for that optimizer is logged to a separate key.