Instructions to use nunchaku-ai/nunchaku-sdxl with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use nunchaku-ai/nunchaku-sdxl with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("nunchaku-ai/nunchaku-sdxl", 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 Settings
- Draw Things
- DiffusionBee
ComfyUI and fp4 (Blackwell) plans?
Hi there, are there any updates to compatibility with ComfyUI and/or fp4 blackwell quants? Thank you for your work.
Hi there, are there any updates to compatibility with ComfyUI and/or fp4 blackwell quants? Thank you for your work.
Hopefully they make more fp4 (Blackwell) models.
I'm seeking a Blackweel model as well.
It's easy to think SDXL is fast enough on its own to not need the benefit of fp4 inference speeds. However, SUPIR, still one of the best image upscaling tools, uses SDXL as a base and even on decent hardware takes a good deal of time to generate. If y'all have converted an SDXL model but figure it's not worth the trouble to put out please consider SUPIR and a few other tools that use SDXL.
Thx for your work!