Instructions to use pmczip/Krea2Raw_Loras with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use pmczip/Krea2Raw_Loras with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("krea/Krea-2-Raw", torch_dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("pmczip/Krea2Raw_Loras") prompt = "a high-resolution side-view photograph featuring 4j133 with long, wavy dark brown hair, styled in loose waves, she has a pleasant, inviting smile and striking brown eyes, she is wearing a black strapless bandeau top, and a pair of light blue jeans that are pulled down to her waist, revealing her midriff, her outfit is accessorized with large gold hoop earrings and a black belt with a silver buckle, adding a touch of sophistication, the background is a solid, bright pink, which contrasts nicely with her dark hair and outfit, making her the focal point of the image, the lighting is soft and even, highlighting her smooth skin, the overall mood of casual and relaxed, with a focus on the subject's confident demeanor" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
any change to include the trigger in the filename?
these are great, but I'm having to grab the triggers from your samples as they aren't in the filename
I might be able to. It might not be until the next update though. But woman or female work just as well as the trigger, if that works.
You're welcome BrknSoul. ucren, I added the trigger word to the Krea2 loras. I get the need for trigger words, especially for putting 2 people in a scene. I will not be adding the trigger words to the other lora models but will continue to do so for Krea2 loras.
By any chance what training model do you use and settings of the model?
I just use AI-Toolkit and the config file from here: https://www.reddit.com/r/StableDiffusion/comments/1ueacq2/krea_2_character_lora_training_for_16_gb_vram/
I just change the filename and trigger word.