I have a bunch of stories which make use of checkpoints to essentially give deterministic loop.
However, i find that i need to increase the memoization policy size quite a bit in order to memorise all the stories. My understanding that the tracker, when in a given state, would survey the given stories to see if one matches. I understand that this may be complicated if the memory history was large, but should it not reduce to a state machine in the case of the history length being basically zero? If not, how would i achieve this?
EDIT: to clarify, i have had cases where i start with ‘hi’ and it goes to fallback unless i train it a bunch more, despite having a bunch of stories that begin with greet, all with same next step, so it shouldn’t fall over there