inner wigeon
Im trying to use gpt-3.5-turbo in order to generate json for a minecraft mod datapack. In order to do this i need the model to know the format the the mod uses, so what i have been doing is feeding it pages from the mods wiki as system prompts, which works fine for small tasks, but i can only give it a few pages from the wiki before the tokens max out, which is a problem, because i need it to be aware of the whole wiki. ive also attempted using fine tuning on the gpt-3 model curie, but that model is only able to do completions, which is not what i am looking for.
