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
hello. i saw in a paper that they combined two tokenizer, one of them was trained by authors and the second was a pretrained one.
is it possible to do this without having their training data?
if yes, how should we put merges in a correct way?
The text was updated successfully, but these errors were encountered:
hello. i saw in a paper that they combined two tokenizer, one of them was trained by authors and the second was a pretrained one.
is it possible to do this without having their training data?
if yes, how should we put merges in a correct way?
The text was updated successfully, but these errors were encountered: