Closest thing you can do is set
checkpoint_model: True under DIET, TED, or ResponseSelector, which saves the best performing model during training.
evaluate_on_number_of_examples > 0 and
evaluate_every_number_of_epochs > 0 .
@chensaics For more details do visit these Github links:
and the one which Chris suggested is this one:
Your recommendation are already considered
@chensaics No worries close this thread with the solution for others.