Closed stweil closed 4 years ago
It allows setting the learning rate hyperparameter for the training.
The default value for finetuning is now reduced from 0.001 to 0.0001 because the old default was too large when finetuning starts with a network which is already rather good.
Signed-off-by: Stefan Weil sw@weilnetz.de
It allows setting the learning rate hyperparameter for the training.
The default value for finetuning is now reduced from 0.001 to 0.0001 because the old default was too large when finetuning starts with a network which is already rather good.
Signed-off-by: Stefan Weil sw@weilnetz.de