valueblocked-complexmarco-6e-4

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: FlowMatchEulerDiscreteScheduler
Seed: 42
Resolution: 1408x768
Skip-layer guidance:

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
Value-blocked conceptual background study in the style of Marco Bucci, A series of industrial machines are arranged in rows inside a large, spacious warehouse. Bright natural light streams in from expansive windows, casting shadows across the wooden floor. The interior is filled with structural elements like beams and supports, suggesting a manufacturing environment.

Negative Prompt
blurry, cropped, ugly

Prompt
Value-blocked conceptual background study in the style of Marco Bucci, A moonlit alley with clothes hanging on a line and dimly lit buildings. The sky is overcast with clouds partially covering the moon. Balconies and beams create shadows across the scene.

Negative Prompt
blurry, cropped, ugly

Prompt
Value-blocked conceptual background study in the style of Marco Bucci, A dark, starry night sky with swirling clouds over a mountainous landscape. A small, illuminated caravan sits in an open field dotted with white flowers.

Negative Prompt
blurry, cropped, ugly

Prompt
Value-blocked conceptual background study in the style of Marco Bucci, A green chalkboard with handwritten text partially covered by shadows cast from a window. The window frame and sunlight create distinct lines and patterns on the board. Artwork pages are pinned at the top.

Negative Prompt
blurry, cropped, ugly

Prompt
Value-blocked conceptual background study in the style of Marco Bucci, A misty morning harbor with fishing boats gently bobbing in the water. The rising sun casts long shadows across weathered wooden docks, while seabirds circle overhead. Stacked crates and coiled ropes line the pier.

Negative Prompt
blurry, cropped, ugly

Prompt
Value-blocked conceptual background study in the style of Marco Bucci, A two-story library interior with spiral staircases and towering wooden bookshelves. Autumn sunlight filters through stained glass windows, creating colorful patterns on leather armchairs and scattered open books.

Negative Prompt
blurry, cropped, ugly

Prompt
Value-blocked conceptual background study in the style of Marco Bucci, An abandoned Victorian greenhouse with broken glass panels and overgrown vines. Shafts of afternoon light pierce through the dusty air, illuminating scattered terra cotta pots and rusted gardening tools.

Negative Prompt
blurry, cropped, ugly

Prompt
Value-blocked conceptual background study in the style of Marco Bucci, A rural train platform at dusk with a wooden waiting shelter. Paper lanterns cast a warm glow on the wooden planks, while steam from a distant locomotive drifts across the purple-orange sky.

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: 6
Training steps: 8000
Learning rate: 0.0006
- Learning rate schedule: polynomial
- Warmup steps: 100
Max grad norm: 0.1
Effective batch size: 3
- Micro-batch size: 3
- Gradient accumulation steps: 1
- Number of GPUs: 1
Gradient checkpointing: True
Prediction type: flow-matching (extra parameters=['shift=3', 'flux_guidance_mode=constant', 'flux_guidance_value=1.0', 'flow_matching_loss=compatible'])
Optimizer: adamw_bf16
Trainable parameter precision: Pure BF16
Caption dropout probability: 10.0%

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

marco-background-512

Repeats: 22
Total number of images: 34
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

marco-background-768

Repeats: 22
Total number of images: 34
Total number of aspect buckets: 1
Resolution: 0.589824 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

marco-background-1024

Repeats: 11
Total number of images: 34
Total number of aspect buckets: 2
Resolution: 1.048576 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

marco-background-1536

Repeats: 5
Total number of images: 34
Total number of aspect buckets: 1
Resolution: 2.359296 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

marco-background-512-crop

Repeats: 11
Total number of images: 34
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: True
Crop style: random
Crop aspect: square
Used for regularisation data: No

marco-background-768-crop

Repeats: 11
Total number of images: 34
Total number of aspect buckets: 1
Resolution: 0.589824 megapixels
Cropped: True
Crop style: random
Crop aspect: square
Used for regularisation data: No

marco-background-1024-crop

Repeats: 5
Total number of images: 34
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


def download_adapter(repo_id: str):
    import os
    from huggingface_hub import hf_hub_download
    adapter_filename = "pytorch_lora_weights.safetensors"
    cache_dir = os.environ.get('HF_PATH', os.path.expanduser('~/.cache/huggingface/hub/models'))
    cleaned_adapter_path = repo_id.replace("/", "_").replace("\\", "_").replace(":", "_")
    path_to_adapter = os.path.join(cache_dir, cleaned_adapter_path)
    path_to_adapter_file = os.path.join(path_to_adapter, adapter_filename)
    os.makedirs(path_to_adapter, exist_ok=True)
    hf_hub_download(
        repo_id=repo_id, filename=adapter_filename, local_dir=path_to_adapter
    )

    return path_to_adapter_file
    
model_id = 'black-forest-labs/FLUX.1-dev'
adapter_repo_id = 'mipat12/valueblocked-complexmarco-6e-4'
adapter_filename = 'pytorch_lora_weights.safetensors'
adapter_file_path = download_adapter(repo_id=adapter_repo_id)
pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.bfloat16) # loading directly in bf16
lora_scale = 1.0
wrapper, _ = create_lycoris_from_weights(lora_scale, adapter_file_path, pipeline.transformer)
wrapper.merge_to()

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


## Optional: quantise the model to save on vram.
## Note: The model was quantised during training, and so it is recommended to do the same during inference time.
from optimum.quanto import quantize, freeze, qint8
quantize(pipeline.transformer, weights=qint8)
freeze(pipeline.transformer)
    
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu') # the pipeline is already in its target precision level
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(42),
    width=1408,
    height=768,
    guidance_scale=3.0,
).images[0]
image.save("output.png", format="PNG")

mipat12
/

valueblocked-complexmarco-6e-4