Instructions to use nitrosocke/Arcane-Diffusion with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use nitrosocke/Arcane-Diffusion with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("nitrosocke/Arcane-Diffusion", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps
- Draw Things
- DiffusionBee
What kind of regularization images do you use?
#8
by tapeters - opened
I am trying to understand the theory behind the use of the use of regularization images. What images are used here? Generated images with the original model with the prompt 'Arcane characters' or something totally different?
The reg images where generated with the model I trained on (SD 1.4) with the DDIM sampler and the prompt "illustration style". You can check them out here: https://drive.google.com/folderview?id=1gfp7MRnF2ZjSgsXk12q920ZK5bosBuEw
The theory behind them is to prevent the training from "bleeding" into other classes. Like training a style I don't want other styles except the "arcane style" to be altered.