- Jan 18, 2019
- 183
- 363
well 30 was just a lowball number, I had someone who couldn't save anymore after page 90. It depends on how much text you have per page
It depends, some pages have very short lines, other much bigger and intricated ones. But i will try.
Last edited:
Game Version 1.0.1
Also noticed a fun bug where the window will just blank and your only option is to close the game. Could be related to running my model locally though. (RTX 3090 with Ollama using mistral:7b-instruct with 512k memory and 1024 token) sometimes it will work flawless, then randomly the screen just blanks sometimes at 1 page, other times well after 90+.
just saw it a small update will try it out see if the same issue persists on 1.0.3
Edited for Game Version number
Also noticed a fun bug where the window will just blank and your only option is to close the game. Could be related to running my model locally though. (RTX 3090 with Ollama using mistral:7b-instruct with 512k memory and 1024 token) sometimes it will work flawless, then randomly the screen just blanks sometimes at 1 page, other times well after 90+.
just saw it a small update will try it out see if the same issue persists on 1.0.3
Edited for Game Version number
Last edited:
hello, someone to explain to me with some pictures what i have to do in the guide ?
(i don't understand and i think i can understand better with some pictures of how to do)
[sorry if my english is bad]
edit : i'm talking about the guide to set up another AI in the game
(i don't understand and i think i can understand better with some pictures of how to do)
[sorry if my english is bad]
edit : i'm talking about the guide to set up another AI in the game
I'm certainly intrigued. I ran the latest version (online) to around page 300 last night. There were a few json errors, but resubmitting the prompt pushed it through.
I'm curious on running it locally, but that's beyond my knowledge base.
Btw, does anyone know if there is a way to modify the character model beyond the limits set in the world? For example, I created a clone of the slime world and tried to set breastsize to 200 maximum. It didn't alter anything in the loaded clone world, so I have no idea if it'll render gigantic tits or not.
I'm curious on running it locally, but that's beyond my knowledge base.
Btw, does anyone know if there is a way to modify the character model beyond the limits set in the world? For example, I created a clone of the slime world and tried to set breastsize to 200 maximum. It didn't alter anything in the loaded clone world, so I have no idea if it'll render gigantic tits or not.
Last edited:
- Jan 21, 2023
- 20
- 54
The slime pauses, its tendrils still wriggling, but no longer moving towards your foot.'You will kill yourself,' you say, your voice filled with command. 'Now.'
I like this game and idea as a whole and when I type in something and the end result if not the outcome I want I can just rollback and fix my writing to be better its like a grammer cop but with porn its weird but funny
Honestly it feels like a more interactive version of the old text based games from the DOS days and i'm here for it. Love being able to just type random thoughts in to change the dynamic, or situation, ect. Just hope it gets more stable also might be a big ask from the dev but it would be awesome if the LLM could be integrated into the game so you don't have to install an external one like Ollama.
Yeah it's not that simple but the developer of 'AI Dungeon' did something like this. The game is made in Godot, but has a server component written in Python that runs a local LLM. It's a bit janky, supports a limited number of models and frequently craps itself - but Python does have a very well developed suite of tools to let you wrap a large number of different models and APIs without too much work.
How does one figure out a good model to use for the game? The llama 3.3 70b is very good, but way too large for local use. The mistral nemo instruct is good, but it forgets whats going on all the time. Is downloading larger 'Q' versions of models worth it? I don't know much about the technicalities of AI models.
- Jan 18, 2019
- 183
- 363
nemo instruct has a massive memory limit of 128k so it shouldn't forget things tho?
Maybe it 'forgetting' happens because I don't enter the offered prompts so it starts a new scenario
- Jan 18, 2019
- 183
- 363
if you host nemo locally then you need to also make sure to set the memory limit in the game settings higher, by default the game assumes only 2k limit, so past 2k characters and it will start forgetting
- Jan 18, 2019
- 183
- 363
Perfect, thank you very much! The new version works fine, the save function also works fine. How hig is the memory limit that can be set for this version?
- Jan 18, 2019
- 183
- 363
its still 2k by default if you use the default AI (this is because it is hosted on my potato PC). If you follow my guide and use the free metallama model from OpenRouter, the memory limit can be set to 120k
Can you make a visual tutorial for that, some photos of the steps shoul be enough i think.
