parameters for learning rate reduction upon loss plateau mutable by CLI

fizyr / keras-retinanet

Keras implementation of RetinaNet object detection.

Apache License 2.0

4.38k stars 1.96k forks source link

parameters for learning rate reduction upon loss plateau mutable by CLI #1370

Closed mooratov closed 4 years ago

mooratov commented 4 years ago

the default parameters for ReduceLROnPlateau are fairly aggressive. Probably fine for most cases, but can be a problem if number of steps per epoch is low or validation set is limited.

hgaiser commented 4 years ago

LGTM, thanks :)