goonsai-com

/

civitaiprompts

like 31

Follow

Goonsai.com 13

I am not familiar with LMStudio. does it support gguf? if that is possible then i can make a GGUF for you

I am not familiar with LMStudio. does it support gguf? if that is possible then i can make a GGUF for you

Yes it does support GGUF. Would be fantastic if you could make these models GGUF!

ok i uploaded for the 10k model. use only the qwen2.5 version. Do not use the qwen3 at all.
uploading the 100k version now but together with gguf.

hey folks, sorry this current gguf is wrong. I will upload the correct one. I used the wrong system and it made the gguf fp16 which is bad for quality. I will reupload after I convert it on a the right machine.

doen and uploaded. please lmk if it works with your regular tools

Tested in LMStudio with qwen2.5-3B-goonsai-nsfw-100k-BF16.gguf. Looks good!

I was able to adapt the Modelfile settings into the LMStudio UI with relative ease.

Example:

prompt: a woman smoking a cigarette

answer: quality 80s anime, long black hair flowing from forehead to shoulders, white blouse with floral pattern over dark blue sweater, light pink lipstick and eye shadow, crouching pose while holding lit cigarrette in right hand, sunny afternoon setting under bright yellow sunlights, warm golden lighting highlights her face

Thanks for your work!

Will try out any other versions you upload too (within reason as I have a 6GB VRAM GPU :))

fantastic. thank you for sharing the feedback. what gpu do you have so i can better test the next udpate. Can I ask what prompt template you've used?

fantastic. thank you for sharing the feedback. what gpu do you have so i can better test the next udpate. Can I ask what prompt template you've used?

I have a RTX 2060 mobile and here's the template:

You are goonsai. Generate CivitAI-style prompts using comma-separated tags with spaces (not underscores).

Format: quality tags, subject description(s), clothing, pose/action, setting, lighting,

Keep tags concise. No repetition. Use spaces between words in tags. Use danbooru tags only

As you've mentioned elsewhere the model doesn't really know about Danbooru tags but it was worth a try

how to load this in HF space , GGUF not supported , I am trying to load via pipeline which version is best to try

how to load this in HF space , GGUF not supported , I am trying to load via pipeline which version is best to try

you should use the safentensors. it is no different than loading any other model in space.

https://huggingface.co/spaces/rahul7star/QwenChat see this any special textencoder i need ? Take a look model should respond hmm

I got this response for "a woman dancing in swimsuit "

Happy girl

• Summer vibes
• Beach girl
• Fun and carefree
• Beautiful
• Cool girl
• Just having fun!
• Good vibes only