Instructions to use cxxai/scfm with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use cxxai/scfm with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("cxxai/scfm") prompt = "-" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Update README.md
Browse files
README.md
CHANGED
|
@@ -20,9 +20,12 @@ license: mit
|
|
| 20 |
|
| 21 |
This lora checkpoint allows you to generate images within 3-8 steps by applying the weight as a LoRA on the FLUX series checkpoints.
|
| 22 |
|
| 23 |
-
Recommended settings: lora strength 1.0-1.75, cfg >=
|
| 24 |
|
| 25 |
For more details, please do not hesitate to check our paper: [Shortcutting Pre-trained Flow Matching Diffusion Models is Almost Free Lunch](https://arxiv.org/pdf/2510.17858)
|
|
|
|
|
|
|
|
|
|
| 26 |
|
| 27 |
## Trigger words
|
| 28 |
|
|
|
|
| 20 |
|
| 21 |
This lora checkpoint allows you to generate images within 3-8 steps by applying the weight as a LoRA on the FLUX series checkpoints.
|
| 22 |
|
| 23 |
+
Recommended settings: lora strength 1.0-1.75, cfg >=3.5. Lower steps generally require higher strength.
|
| 24 |
|
| 25 |
For more details, please do not hesitate to check our paper: [Shortcutting Pre-trained Flow Matching Diffusion Models is Almost Free Lunch](https://arxiv.org/pdf/2510.17858)
|
| 26 |
+
and the [GitHub Repo](https://github.com/caitree/scfm).
|
| 27 |
+
|
| 28 |
+
For SD3.5, the only recommended negative prompt is null (i.e., an empty string ""), since the checkpoint was trained exclusively with this setting.
|
| 29 |
|
| 30 |
## Trigger words
|
| 31 |
|