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
I used the antislop sampler a lot, but sometimes the banning of phrases gets frustrating in how models keep getting around it. See the following:
shivers down his spine
shivers down her spine
shivers down their spines
I don't know how this sampler works or what backtracking really means, but is it possible to accept phrases with placeholders where we can substitute values? Just like how string interpolation in programming languages work (I.E. f"shivers down {word} spine" in python).
I imagine that the big challenge here is we're no longer identifying raw strings, instead we're identifying string templates, but then again I've no idea how this sampler works.
What do you think?
The text was updated successfully, but these errors were encountered:
Hi,
I used the antislop sampler a lot, but sometimes the banning of phrases gets frustrating in how models keep getting around it. See the following:
I don't know how this sampler works or what backtracking really means, but is it possible to accept phrases with placeholders where we can substitute values? Just like how string interpolation in programming languages work (I.E. f"shivers down {word} spine" in python).
I imagine that the big challenge here is we're no longer identifying raw strings, instead we're identifying string templates, but then again I've no idea how this sampler works.
What do you think?
The text was updated successfully, but these errors were encountered: