DeepSpell
DeepSpell copied to clipboard
Trailing Periods
I've noticed that the model's predictions all have trailing periods. For example:
Q Possibly even for good reasons. A Possibly even for good reasons. ☒ Possibly even for good rensuss..............................
Is this possibly a bug?
Figured it out. I ended up adding an end of sequence token to the training data so that the model ended up learning when to terminate outputs.
@xueharry Can you share the new working code please? or maybe send a pull request.
@xueharry Can you please explain more about it? What did you use as an end of sequence and how did the model know where to terminate outputs?
@MajorTal any word of advice?