Closed TjarkMiener closed 1 year ago
When resuming the training the validation loss of the best performing checkpoint was not track. CTLearn/keras uses the default initial value which is inf. This bug fix allows to resume training properly. Related to keras issue #12803.
When resuming the training the validation loss of the best performing checkpoint was not track. CTLearn/keras uses the default initial value which is inf. This bug fix allows to resume training properly. Related to keras issue #12803.