JamesTissot-Flux-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: FlowMatchEulerDiscreteScheduler
Seed: 42
Resolution: 968x1280
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
In the style of a James Tissot painting, a woman in a black dress with white ruffled underlayers sits in a red chair, her posture relaxed. A black cat rests beside her, and a vase of white flowers sits on a nearby table. The room features a mirror and framed artwork.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of a James Tissot painting, two women in light blue ruffled dresses stand in a luxurious room with large windows overlooking tropical plants. One pours tea at a small table while another sits nearby. The room contains ornate furniture, an intricate carpet, and a samovar.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of a James Tissot painting, a woman wearing a checkered dress sits at a breakfast table with a carafe and fruit, reading a letter. A man holds up a newspaper while ships are visible through large windows behind them.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of a James Tissot painting, a young woman practices piano in a conservatory, sunlight streaming through art nouveau windows onto her emerald green dress. Potted orchids line the walls, and sheet music scattered across the floor catches the late afternoon light.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of a James Tissot painting, two sisters prepare for a masquerade ball, one adjusting the other's venetian mask while standing before a gilt mirror. Their elaborate dresses in complementary shades of burgundy and navy reflect in the candlelit room.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of a James Tissot painting, a lady artist works at her easel in a sunny studio, her paint-stained apron contrasting with her formal Victorian dress. Through the window, hot air balloons float above a cityscape of chimneys and spires.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of a James Tissot painting, a woman astronomer in a midnight blue Victorian dress with silver buttons studies the night sky through a brass telescope on an observatory balcony. Her detailed skirt catches moonlight as she leans forward, while star charts and astronomical instruments rest on a marble-topped table nearby. Through the domed ceiling's opening, the Pleiades cluster shimmers above.

Negative Prompt
blurry, cropped, ugly

Prompt
In the style of a James Tissot painting, an elegant Japanese geisha in a coral and gold kimono serves tea to a Victorian lady wearing a lavender bustle dress in a fusion parlor. Wisteria cascades through the open shoji screens, while European oil paintings hang above Japanese tatami mats. A peacock fan rests on a lacquered table beside an English silver tea service.

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: 9
Training steps: 5000
Learning rate: 0.0004
- Learning rate schedule: polynomial
- Warmup steps: 200
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=['flux_schedule_auto_shift', 'shift=0.0', 'flux_guidance_mode=constant', 'flux_guidance_value=1.0', 'flux_beta_schedule_alpha=10.0', 'flux_beta_schedule_beta=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

ab-512

Repeats: 11
Total number of images: 29
Total number of aspect buckets: 7
Resolution: 0.262144 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

ab-768

Repeats: 11
Total number of images: 29
Total number of aspect buckets: 9
Resolution: 0.589824 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

ab-1024

Repeats: 5
Total number of images: 29
Total number of aspect buckets: 11
Resolution: 1.048576 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

ab-crops-512

Repeats: 7
Total number of images: 29
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: True
Crop style: random
Crop aspect: square
Used for regularisation data: No

ab-1024-crop

Repeats: 7
Total number of images: 29
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 = 'davidrd123/JamesTissot-Flux-LoKr'
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=968,
    height=1280,
    guidance_scale=3.0,
).images[0]
image.save("output.png", format="PNG")

davidrd123
/

JamesTissot-Flux-LoKr