It's good that the Dev is constantly working on the game and that the game is getting constant updates, but it would be nice if some of those updates moved the main story forward and were not only improvements to the sandbox.
Others Multiic [v0.3.67] [Stvoler]
- Thread starter aleksandr unconditional
- Start date
A few local models I've tried:
I suggest first trying to make koboldcpp itself work with a very small model;
As a reference, with a similar 11B model I have here on a smaller quant (Fimbulvetr-11B-v2.1-16K.i1-Q3_K_S.gguf), I can get the first Laura reply after ~75s, and ~1 word per second during generation. And I have a five year old AMD iGPU (Vulkan backend).
-
You must be registered to see the links(Q3_K_S) worked very well in story mode, but tended to be a bit liberal with the emojis on sandbox mode;
-
You must be registered to see the links(Q4_K_S) and variants seemed to have better overall consistency;
-
You must be registered to see the links(Q4_K_M) had some of the best writing on a few of scenes, but it pretty much avoided using most emojis.
You must be registered to see the links
). For instance, I have 8G VRAM, so I only try GGUF files under 6GB, and most under 5GB.I suggest first trying to make koboldcpp itself work with a very small model;
You must be registered to see the links
is one of the smallest I know that I can still have fun roleplaying with. Load that one in Koboldcpp, open the web interface, click Scenarios, and chat a little bit with Tiff or Nail. CPU usage should be fairly low.That particular quant is very big; it does need either a powerful GPU or CPU, and plenty of RAM. I'd guess it's falling back to CPU.
As a reference, with a similar 11B model I have here on a smaller quant (Fimbulvetr-11B-v2.1-16K.i1-Q3_K_S.gguf), I can get the first Laura reply after ~75s, and ~1 word per second during generation. And I have a five year old AMD iGPU (Vulkan backend).
gemini-1.5-pro works best, but for me it stops working after a few minutes. In that case I switch either to gemini-1.5-flash or to a local model.
Have you got Koboldcpp itself working on
You must be registered to see the links
as described in
You must be registered to see the links
?
Multiic-0.3.67
rpdl torrents are unaffiliated with F95Zone and the game developer.
Please note that we do not provide support for games.
For torrent-related issues use here, or join us on
You don't have permission to view the spoiler content.
Log in or register now.
Please note that we do not provide support for games.
For torrent-related issues use here, or join us on
You must be registered to see the links
!
You must be registered to see the links
,
You must be registered to see the links
. Downloading issues? Look here.
I have used the koboldcpp and it still never worked, Like it would install and i could turn on kobold but the localhost would not even when i used a ggup that was around 1 gigabyte
got a nice run with laura in bath .)
but at the end it triggers the end of the prologe
so josie, dina and fiona events are cut out
intendet or bug ?
2nd
i got around 10(not sure, but more then 5) of this mutiple choice thingies to choose from
these choices need around 2-3 minutes to pop up, each
but at the end it triggers the end of the prologe
so josie, dina and fiona events are cut out
intendet or bug ?
2nd
i got around 10(not sure, but more then 5) of this mutiple choice thingies to choose from
these choices need around 2-3 minutes to pop up, each
- Apr 17, 2020
- 444
- 549
What an innovative game, but the experience is somehow frustrating
it is super lagging; I need to wait minutes for the AI's reply, the conversation is confusing, the pronouns are messy(due to AI again, I guess), and it is challenging to tell who says what when it comes to a multi-character scenario. Despite those, the story is intriguing even though it is more like a premise than a story.
it is super lagging; I need to wait minutes for the AI's reply, the conversation is confusing, the pronouns are messy(due to AI again, I guess), and it is challenging to tell who says what when it comes to a multi-character scenario. Despite those, the story is intriguing even though it is more like a premise than a story.
- Apr 17, 2020
- 444
- 549
almost certainly, it is based on AI; your input will be sent to a super-computer somewhere.
- Jan 18, 2018
- 2
- 0
I guess when you have enough experience running LLMs on your own GPU and chat with different bots, learning how Kobold and SillyTavern works, you begin to understand a lot of things about this game. About how it works. And a few words for the developer: Слава Україні, шановний!
What exactly triggers the girls to undress (change image) from a technical standpoint? This seems to work well with the Gemini models, but not at all with other models.
Dalsia hra ktora neakceptuje hracov co nevedia anglicky
This text is real reaktion and real information for dev!!!!!
This text is real reaktion and real information for dev!!!!!
Im using local model, working good with text etc ...
But can we with local model have new ai generated images ? like infinity possibility ? Because i can trigger only the same images of the character every time
But can we with local model have new ai generated images ? like infinity possibility ? Because i can trigger only the same images of the character every time
- Oct 19, 2019
- 19
- 9
У меня в начале 1 дня случается ядерный взрыв (После того как отказываюсь от тренировки или сразу после тренировки с Лорой) и игра начинается сначало.
At the beginning of day 1, I have a nuclear explosion (After I refuse training or immediately after training with Laura) and the game starts all over again.
At the beginning of day 1, I have a nuclear explosion (After I refuse training or immediately after training with Laura) and the game starts all over again.
- May 27, 2023
- 9
- 36
play a bit longer after the boom, somewhere before you get to Fiona bath scene you can choose to play day 0 again or go for next chapter
The regenerate button doesnt exist or is supporters only? Sometimes the model goes crazy =)
Anyway, I think it would be good to have an icon showing "waiting for the model" or something. When I am using my own I can see it is generating, but for people using the interwebz they have no clue. I will for sure keep an eye on this game for the future, this looks great.
Anyway, I think it would be good to have an icon showing "waiting for the model" or something. When I am using my own I can see it is generating, but for people using the interwebz they have no clue. I will for sure keep an eye on this game for the future, this looks great.
Even better, a generic OpenAI-compatible API. Many local and remote providers could be supported that way.
Last edited: