• curbstickle_lw@lemmy.worldEnglish
    14 days

    Definitely a new one for here…

    That said, your post is missing the [AIT] tag - AI Topic. Please update asap.

    Edit: Locking until updated.

  • iceberg314@slrpnk.netEnglish
    15 days

    That’s too bad you can’t discuss the tianamen square massacre with your sex/roleplay chat bot.

    Lol what’s your tokens per second not the server with no GPU running those sizes of models?

    • matti@seksi.jatai.deEnglish
      14 days

      I’m getting like 10-20tps with some other similar-sized models, but didn’t benchmark qwen yet. Note that they are mixture-of-experts, so they are evaluating only a few billion parameters each time.

      Edit:

      I benchmarked the model now:

      prompt eval count:    11 token(s)  
prompt eval duration: 485.605ms  
prompt eval rate:     22.65 tokens/s  
eval count:           207 token(s)  
eval duration:        21.392268s  
eval rate:            9.68 tokens/s

      Not bad for a good CPU model.

  • ZeldaFreak@lemmy.worldEnglish
    15 days

    There is an ERP (erotic role play) Benchmark: https://huggingface.co/spaces/overhead520/Unhinged-ERP-Benchmark?not-for-all-audiences=true You may also interested in software that is more ERP capable. 2 options would be Marinara Engine and SillyTavern. There you pair a model, a character card, your persona card, lorebooks and presets. What is that? A character card are the character, the AI plays. Being it a specific character, a world or a scenario. A persona is structured like a character card, but the purpose is that you are this character. Best practice is only to fill this card with stuff the character knows or sees about you.

    A lorebook is additional information about the world or items. For example Marinara Engine can be used for RPG roleplay and in a lorebook you can explain weapons and items in details. This saves tokens, as these information only get pulled from the lorebook and send to the AI, when it get’s triggered by keywords.

    Presets are the structure. It tells the AI how to output stuff and so on. These can contain soft jailbreaks. https://rentry.org/freaky-frankenstein-presets Freaky Frankenstein is one I use (FF5 Micro) and there you can set up a lot of things. But of course you can download other presets, that do fit you more or you write your own.

    You must do research which presets work with which model and how far you can go and what you want and what you like. I use GLM 5.1 with FF5 Micro. I tried Deepseek but it was slower and I didn’t liked the output as much.

    And now a comment to people who are disliking this: As long as you do treat it more as porn replacement and not as replacement for real human interactions, I find this good. Porn videos did killed the imagination and goes more back to erotic literature, which does require you to do more mentally. With AI you can choose your setting and boundaries and aren’t on other humans to do ERP. I never did ERP with other humans, but I’m pretty sure you can’t just close the chat window, when you are satisfied and let the other person hanging.

    Also with both Marinara Engine and Silly Tavern, you don’t need to go sexual. Marinara does support games. I didn’t used it much yet. I tried the RPG function once. I didn’t had a char card, that was optimized to it. And the thing is, I have zero RPG experience. I never had friends who played DnD or pen and paper. But apparently Marinara does support this and they added a Multiplayer function. This way the AI can do the GM or other players. I don’t have much information for this, as I didn’t found a good resource jet, which explains me stuff and does recommend me fitting char cards to use.

  • [object Object]@lemmy.caEnglish
    15 days

    I don’t have experience nor do I want to get into it, but openclaw is a very token heavy application and you can probably do a lot better performance wise.

    The reason it takes so long for first message is the amount of prompt openclaw requires. It’s almost certainly overkill, and with low prompt processing it will spend most of its time reading memory and soul files and config info for all of its tools.

    I think you could just install earandil pi, set a prompt, and add a memory plugin and this will be faster. This is what openclaw uses if you strip back all the extras.

    Also look for mixture of experts, they’re far less GPU intensive, so you’ll get better mileage. No idea what models are out there, but like even an a3b or a4b will be slow but maybe usable.

    Again, I do not want to get into what this use case lacks or doesn’t work for, that’s your business and I don’t want to get into it. I’m just saying openclaw is probably like driving a 12 ton Humvee to the grocery store.

  • trojanbark@lemmy.worldEnglish
    15 days

    as someone who wants the global population to be reduced drastically, i support this

    but also this is sad

    cognitive dissonance sucks