ReduceLROnPlateau.Config¶

Component: ReduceLROnPlateau

class ReduceLROnPlateau.Config[source]: Bases: Scheduler.Config

All Attributes (including base classes)

lower_is_better: bool = True

This indicates the desirable direction in which we would like the training to proceed. If set to true, learning rate will be reduce when quantity being monitored stops going down

factor: float = 0.1

Factor by which the learning rate will be reduced. new_lr = lr * factor

patience: int = 5

Number of epochs with no improvement after which learning rate will be reduced

min_lr: float = 0

Lower bound on the learning rate of all param groups

threshold: float = 0.0001

Threshold for measuring the new optimum, to only focus on significant changes.

threshold_is_absolute: bool = True

One of rel, abs. In rel mode, dynamic_threshold = best * ( 1 + threshold ) in ‘max’ mode or best * ( 1 - threshold ) in min mode. In abs mode, dynamic_threshold = best + threshold in max mode or best - threshold in min mode.

cooldown: int = 0

Number of epochs to wait before resuming normal operation after lr has been reduced.

Default JSON

{
    "lower_is_better": true,
    "factor": 0.1,
    "patience": 5,
    "min_lr": 0,
    "threshold": 0.0001,
    "threshold_is_absolute": true,
    "cooldown": 0
}