A LoRA fine-tune for FLUX.1 [dev] that specializes in generating fractals. Trained using Fluxgym by cocktail peanut, utilizing my own Fractal collection that I have made over the years with UltraFractal plotting software, deploying novel complex series for an experimental fractal dataset. (borch's ultra fractal scripts repo coming...)
Overview
This LoRA enhances FLUX.1 [dev]'s ability to generate detailed fractals patterns with vibrant colors, and symmetrical details. Use the trigger word `fractalLand` to activate the fractal pattern capabilities.
Installation
First, ensure you have FLUX.1 [dev] installed:
# Install
pip install -U diffusers
# Loading with Diffusersimport torch
from diffusers import FluxPipeline
# Load base model
pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16)
pipe.enable_model_cpu_offload()
# Load and apply LoRA weights
pipe.load_lora_weights("Borcherding/FLUX.1-dev-LoRA-FractalLand-v0.1")
# Generate image
prompt = "fractalLand A detailed fractal with vibrant red, black and green spirals."
image = pipe(
prompt,
height=1024,
width=1024,
guidance_scale=3.5,
num_inference_steps=50,
max_sequence_length=512,
generator=torch.Generator("cpu").manual_seed(0)
).images[0]
image.save("autumn-scene.png")
# Unload LoRA weights if needed
pipe.unload_lora_weights()
# Optional: Merge LoRA weights permanently
pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16)
pipe.load_lora_weights("Borcherding/FLUX.1-dev-LoRA-AutumnSpringTrees")
pipe.merge_lora_weights()
pipe.save_pretrained("FLUX.1-dev-LoRA-AutumnSpringTrees-merged")
Trigger Word
Use fractalLand in your prompts to activate the fractal elements.
Best results achieved when placing the trigger word at the start of the prompt
Prompt Examples
"fractalLand A detailed fractal with vibrant red, black and green spirals."
"fractalLand A symetrical round fractal with vibrant yellow, black and pink waves."
"fractalLand A detailed fractal with vibrant blue, black and orange blocks."
Limitations
This is a LoRA for FLUX.1 [dev] and requires the base model to function
Limited training data, and poorly annotated data, leads this model to be confused when prompted, next generations will deploy more specific references to the names of specific fractal formulas, shapes, colors, and coordinates in the complex plane.
Inherits base limitations from FLUX.1 [dev]
License
This LoRA falls under the same licensing terms as FLUX.1 [dev]. Please refer to the base model's license for usage terms.
Credits
Base model: FLUX.1 [dev] by Black Forest Labs
LoRA training and development: [Borcherding at BorchInk]
Trigger words
You should use fractalLand to trigger the image generation.
Download model
Weights for this model are available in Safetensors format.