So, a couple years ago, i found something called AI dungeon, it was my first contact with an AI model. It was very good until it got censored af. The version i used was probably using GPT 2 or 1 (idk)
Is there any model that can compare or is better for text based RPGS that isn’t very censored and preferably 7b (im running on oobagooba with my CPU)
Note: Sorry if i used the wrong terminology, I’m still a total noob lol. I discovered you can run your own language model just 2 days ago, and don’t worry, I’m not trying to fulfill some weird fetish 💀.
I have a dungeon crawl character card that always just keeps my “inventory” in a list after the narration on every prompt, but the model I use might be too heavy since I use a dual 3090 setup, Dolphin 2.2 70B.
Damnnn that’s a sick setup, btw how can i get these characters?
I’m working on something similar, but different: https://github.com/neph1/LlamaTale
It will never allow the kind of freedom that a pure storytelling model will give you. But the goal is to provide a more stable experience. I’ve made a number of posts about it here (and the next one is due, soon), but here’s the first:
https://www.reddit.com/r/LocalLLaMA/comments/152w71n/mud_llm_for_a_stronger_roleplaying_experience/
It runs fairly well with a 7b, https://huggingface.co/TokenBender/llama2-7b-chat-hf-codeCherryPop-qLoRA-merged being the one I play with, mostly.