The problem with that one is that it only runs on powerful online AI's, so while it reads better, you are gatekept by having to acquire credits.
- Jan 18, 2019
- 269
- 471
how do i do that exactly cuz i can handle alot but if i need to trim some tokens to get a bit faster respones time then i will cuz i relly have no clue what im doing in lms other then more tokens= more story. so like what exactly is gpu offload out of 30 and cpu thread pool size mean
Last edited:
- Jan 18, 2019
- 269
- 471
I'm not experienced with LM Studio, only used it a few times. Hopefully you can get help with the LM Studio parameters on their discord or forum
A "continue" button would go a long way, sometimes it just stops where it'd be better to keep going. Even with multiple paragraphs enabled.
- Jan 18, 2019
- 269
- 471
you can type 'continue'
It's not really possible to allow the AI to continue writing from the previous game text, most AI endpoints don't support this unfortunately.
One thing fierylion should consider is to add a few small validations on game load, to make sure the save being loaded is from the active world.
Adding versioning to the world definition and validating the save's 'world version' should also be considered
Just a warning will suffice when detecting a mismatched world or version (at least for now)
Adding versioning to the world definition and validating the save's 'world version' should also be considered
Just a warning will suffice when detecting a mismatched world or version (at least for now)
- Jan 18, 2019
- 269
- 471
thats output tokens correct cuz ya that other story takes a good chuck to come up with it respones and i have no idea if i mess with the world setting and remove how many promts it tires to give cuz it gives like more then 5 auto respones
- Jan 18, 2019
- 269
- 471
What determines how long of a response the AI will write? Is that memory/token configs alone or is there something in the world/game settings too?
- Jan 18, 2019
- 269
- 471
max output tokens in the settings
However, by default AI will only produce one paragraph of gametext. You can disable this and allow long responses by unchecking One Paragraph Responses in the setting
I'm a little confused by that setting, what's the difference between "max memory" and "max output tokens"? It seems that max output tokens depends on being smaller than max memory, why have two separate settings?