Instructions to use ostris/OpenFLUX.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use ostris/OpenFLUX.1 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("ostris/OpenFLUX.1", 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
instructions for lora training?
I tried kohya_ss using the same settings that work for flux dev, but it has errors with the vae.
Have you tried to route /pipeline in the regular FLUX vae? It's not like you're training the vae anyways. Depending on what you're using to launch/config the Kohya training and how, you might be able to either add a path to the original vae as a training argument (if, say, you're training on the Diffusers format weights from here). Alternately you could try that with the transformer only OpenFlux Safetensors of OpenFlux. Here's a link to it: https://huggingface.co/Kijai/OpenFLUX-comfy/blob/main/OpenFlux-fp8_e4m3fn.safetensors (For extra clarity: the larger Safetensors in Ostris' repo is all-in-one, I believe. Which makes it harder to replace the vae on the fly. Try it with the Unet though. Perhaps try to launch the trainer through ComfyUi.)