napoleon-lokr-multi

This is a LyCORIS adapter derived from stabilityai/stable-diffusion-3.5-large.

The main validation prompt used during training was:

a photo-realistic image of a Napoleon Dynamite man wearing a bedazzled gymnast's leotard, standing triumphantly on the winner's podium with a large gold medal hanging from it's blue ribbon displayed proudly on his chest. power pose, smile, high quality

Validation settings

CFG: 4.0
CFG Rescale: 0.0
Steps: 20
Sampler: None
Seed: 1404
Resolutions: 1024x1024, 896x1152, 1216x832

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
a photo-realistic image of a Napoleon Dynamite man wearing a bedazzled gymnast's leotard, standing triumphantly on the winner's podium with a large gold medal hanging from it's blue ribbon displayed proudly on his chest. power pose, smile, high quality

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: 5000
Learning rate: 0.00016
Max grad norm: 0.01
Effective batch size: 4
- Micro-batch size: 1
- Gradient accumulation steps: 1
- Number of GPUs: 4
Prediction type: flow-matching
Rescaled betas zero SNR: False
Optimizer: adamw_bf16
Precision: Pure BF16
Quantised: No
Xformers: Not used
LyCORIS Config:

{
    "bypass_mode": true,
    "algo": "lokr",
    "multiplier": 1.0,
    "full_matrix": true,
    "linear_dim": 10000,
    "linear_alpha": 1,
    "factor": 12,
    "apply_preset": {
        "target_module": [
            "Attention"
        ],
        "module_algo_map": {
            "Attention": {
                "factor": 6
            }
        }
    }
}

Datasets

napoleon-512

Repeats: 10
Total number of images: ~84
Total number of aspect buckets: 6
Resolution: 0.262144 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

napoleon-1024

Repeats: 10
Total number of images: ~60
Total number of aspect buckets: 3
Resolution: 1.048576 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

napoleon-512-crop

Repeats: 10
Total number of images: ~72
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: True
Crop style: random
Crop aspect: square
Used for regularisation data: No

napoleon-1024-crop

Repeats: 10
Total number of images: ~44
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 = 'stabilityai/stable-diffusion-3.5-large'
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 = "a photo-realistic image of a Napoleon Dynamite man wearing a bedazzled gymnast's leotard, standing triumphantly on the winner's podium with a large gold medal hanging from it's blue ribbon displayed proudly on his chest. power pose, smile, high quality"
negative_prompt = 'blurry, cropped, ugly'
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
image = pipeline(
    prompt=prompt,
    negative_prompt=negative_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=4.0,
).images[0]
image.save("output.png", format="PNG")

zaksynack
/

napoleon-lokr-multi