-
Yih-Dar authored
* create branch for issue #6968 * First attempt to fix incorrect tf trainer loss calculation * Fix training loss in metric * fix tf trainer evaluation loss * apply count_instances_in_batch() for eval and test datasets * prototype of using a new argument in trainer_tf.py to fix loss issue * some renaming and fix, in particular for evaluation methods * fix bugs to have a running version * change to @staticmethod * apply style
cb061e78
