If I just have only 2 sets: train and val then should I use beam search to get the best model or use greedy search.
If choose greedy search then is it correct that we won't ensure that beam search always give better performance than greedy search right?