LearningRateMonitor

class braintools.trainer.LearningRateMonitor(logging_interval='step', log_momentum=False)

Monitor and log learning rate during training.

Parameters:

• logging_interval (str) – When to log the learning rate. One of ‘step’ or ‘epoch’.

• log_momentum (bool) – Reserved for logging optimizer momentum values. Not yet implemented; the value is accepted but currently has no effect.

Examples

>>> lr_monitor = LearningRateMonitor(logging_interval='epoch')
>>> trainer = Trainer(callbacks=[lr_monitor])

property lr_history: List[Dict[str, float]]

History of logged learning rates.

on_train_batch_start(trainer, module, batch, batch_idx)

Log learning rate at step start.

on_train_epoch_start(trainer, module)

Log learning rate at epoch start.