Flux-Georges-Barbier-LoKr

This is a LyCORIS adapter derived from black-forest-labs/FLUX.1-dev.

No validation prompt was used during training.

None

Validation settings

CFG: 3.0
CFG Rescale: 0.0
Steps: 20
Sampler: None
Seed: 42
Resolution: 1024x1024

Note: The validation settings are not necessarily the same as the training settings.

You can find some example images in the following gallery:

Prompt
unconditional (blank prompt)

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of b4rb13r Art Deco illustration, Three women in a luxuriant garden setting during sunset. The central figure wears a peacock-inspired gown with trailing feathers and gold embroidery, while her companions wear complementary dresses with geometric patterns and oriental motifs. Japanese-inspired cherry blossoms, ornamental fountains, and stylized clouds frame the elegant scene.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of b4rb13r Art Deco illustration, A sophisticated woman poses beneath an Art Deco theater marquee at night. Her black velvet dress features bold geometric beading and a dramatic cape. She holds a fan of ostrich feathers while golden light spills from curved architectural elements, creating dramatic shadows across the geometric patterns of the entrance.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of b4rb13r Art Deco illustration, Two fashionable women on a coastal promenade at dusk. One stands gracefully in a striped sailing costume with a silk scarf flowing in the breeze, while her companion reclines on a chaise in a flowing paisley dress. Stylized waves, seabirds, and Art Deco lampposts frame the Mediterranean scene.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of b4rb13r Art Deco illustration, A winter evening gathering showing three elegant women in fur-trimmed coats and evening gowns against a snowy palace backdrop. Their dresses feature intricate beadwork and geometric patterns in silver and blue. Stylized snowflakes and angular ice sculptures decorate the terrace, while northern lights dance in the geometric sky.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of b4rb13r Art Deco illustration, A futuristic beauty salon where a woman in a neo-Victorian dress with fiber optic embellishments receives a beauty treatment from elegant androids. Her dress combines traditional Art Deco patterns with glowing circuits. Holographic mirrors and neon-lit architectural elements maintain the geometric Art Deco aesthetic.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of b4rb13r Art Deco illustration, An underwater palace ballroom where mermaids in Art Deco-patterned scales dance through geometric currents. They wear pearl-studded tops with flowing silk elements that ripple in the water. Crystal chandeliers cast prismatic light through stylized schools of fish and coral formations.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of b4rb13r Art Deco illustration, Three women in flowing desert robes gather around a geometric oasis pool. Their garments blend traditional Middle Eastern patterns with bold Art Deco designs in gold and turquoise. Stylized palm trees, peacocks, and Islamic architectural elements frame the scene, while the setting sun creates dramatic shadows.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of b4rb13r Art Deco illustration, A retrofuturistic space lounge where elegant women in zero-gravity gowns float gracefully. Their dresses combine 1920s silhouettes with transparent bubble helmets and metallic Art Deco patterns. Through curved viewport windows, Earth and geometric constellations provide a cosmic backdrop to the sophisticated gathering.

Negative Prompt
blurry, cropped, ugly

The text encoder was not trained. You may reuse the base model text encoder for inference.

Training settings

Training epochs: 17
Training steps: 8000
Learning rate: 0.0008
Max grad norm: 2.0
Effective batch size: 4
- Micro-batch size: 4
- Gradient accumulation steps: 1
- Number of GPUs: 1
Prediction type: flow-matching (flux parameters=['flux_guidance_value=1.0'])
Rescaled betas zero SNR: False
Optimizer: adamw_bf16
Precision: Pure BF16
Quantised: Yes: int8-quanto
Xformers: Not used
LyCORIS Config:

{
    "algo": "lokr",
    "multiplier": 1.0,
    "linear_dim": 10000,
    "linear_alpha": 1,
    "factor": 16,
    "apply_preset": {
        "target_module": [
            "Attention",
            "FeedForward"
        ],
        "module_algo_map": {
            "Attention": {
                "factor": 16
            },
            "FeedForward": {
                "factor": 8
            }
        }
    }
}

Datasets

georges-barbier-512

Repeats: 12
Total number of images: 43
Total number of aspect buckets: 2
Resolution: 0.262144 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

georges-barbier-1024

Repeats: 4
Total number of images: 43
Total number of aspect buckets: 11
Resolution: 1.048576 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

georges-barbier-512-crop

Repeats: 12
Total number of images: 43
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: True
Crop style: random
Crop aspect: square
Used for regularisation data: No

georges-barbier-1024-crop

Repeats: 4
Total number of images: 43
Total number of aspect buckets: 1
Resolution: 1.048576 megapixels
Cropped: True
Crop style: random
Crop aspect: square
Used for regularisation data: No

Inference

import torch
from diffusers import DiffusionPipeline
from lycoris import create_lycoris_from_weights

model_id = 'black-forest-labs/FLUX.1-dev'
adapter_id = 'pytorch_lora_weights.safetensors' # you will have to download this manually
lora_scale = 1.0
wrapper, _ = create_lycoris_from_weights(lora_scale, adapter_id, pipeline.transformer)
wrapper.merge_to()

prompt = "An astronaut is riding a horse through the jungles of Thailand."

pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
image = pipeline(
    prompt=prompt,
    num_inference_steps=20,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(1641421826),
    width=1024,
    height=1024,
    guidance_scale=3.0,
).images[0]
image.save("output.png", format="PNG")

davidrd123
/

Flux-Georges-Barbier-LoKr