Instructions to use DeverStyle/Krea2-Loras with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use DeverStyle/Krea2-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-Turbo", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("DeverStyle/Krea2-Loras") prompt = "-" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
metadata
tags:
- text-to-image
- lora
- diffusers
- template:diffusion-lora
widget:
- output:
url: images/test_00425_.png
text: '-'
base_model: krea/Krea-2-Turbo
instance_prompt: null
license: apache-2.0
NotFal

- Prompt
- -
NOT TO BE USED - Just an example
A repository to show how bad the recently uploaded Fal Krea2 loras really can be and why you should not copy their training parameters or ever call a lora trained on 8 images a style lora.
Two models are available just for your comparison pleasure, use the trigger word n0t_f4l for both.
- a
n0t_f4l_000001000(1000 steps / LR 0.0001) and a fal versionn0t_f4l_bad_100(100 steps / LR 0.00035).
Download model
Download them in the Files & versions tab.