You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In translation task, I have to choice "best_checkpoint_metric".
But, my goal is to get the best BLEU score model in validation dataset.
So, I want to choice BLEU (not "loss") as "best _checkpoint_metric", because the lowest "loss" score model in validation dataset does not necessarily perform the best from the viewpoint of BLEU.
But I cannot choose BLEU as best_checkpoint_metric.
Or, is it natural to use the lowest loss model's BLEU score as metric in NLP conference's paper??
The text was updated successfully, but these errors were encountered:
In translation task, I have to choice "best_checkpoint_metric".
But, my goal is to get the best BLEU score model in validation dataset.
So, I want to choice BLEU (not "loss") as "best _checkpoint_metric", because the lowest "loss" score model in validation dataset does not necessarily perform the best from the viewpoint of BLEU.
But I cannot choose BLEU as best_checkpoint_metric.
Or, is it natural to use the lowest loss model's BLEU score as metric in NLP conference's paper??
The text was updated successfully, but these errors were encountered: