davinci-sdxl-lora-01

This is a LyCORIS adapter derived from stabilityai/stable-diffusion-xl-base-1.0.

The main validation prompt used during training was:

DaVinciXL, One mechanical device with gears and levers, no human subjects, one item in the image.

Validation settings

CFG: 4.2
CFG Rescale: 0.0
Steps: 30
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
ggn_style painting of a hipster making a chair

Negative Prompt
blurry, cropped, ugly

Prompt
ggn_style painting of a hamster

Negative Prompt
blurry, cropped, ugly

Prompt
in the style of ggn_style, A painting of a woman stands near the water holding an object. Another woman swims in the water. A tree with twisted branches is at the foreground left. Flowers and vegetation are near the lower center. Hills with vegetation are in the background. Text 'Parau na te Varua ino' at the bottom left and artist's signature at the lower right.

Negative Prompt
blurry, cropped, ugly

Prompt
ggn_style, A seated woman with long dark hair is depicted in a front-facing view. She is wearing a dress with a white collar and appears to be in her thirties. Her hands are on her lap. Green leaves and flowers surround her.

Negative Prompt
blurry, cropped, ugly

Prompt
ggm_style, tropical fruits and flowers, bold outlines, non-naturalistic colors, decorative composition

Negative Prompt
blurry, cropped, ugly

Prompt
DaVinciXL, One mechanical device with gears and levers, no human subjects, one item in the image.

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: 3000
Learning rate: 7e-05
Effective batch size: 8
- Micro-batch size: 8
- Gradient accumulation steps: 1
- Number of GPUs: 1
Prediction type: epsilon
Rescaled betas zero SNR: False
Optimizer: optimi-lionweight_decay=1e-3
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

davinci-sdxl-512

Repeats: 10
Total number of images: 50
Total number of aspect buckets: 8
Resolution: 0.262144 megapixels
Cropped: False
Crop style: None
Crop aspect: None

davinci-sdxl-1024

Repeats: 10
Total number of images: 50
Total number of aspect buckets: 16
Resolution: 1.048576 megapixels
Cropped: False
Crop style: None
Crop aspect: None

davinci-sdxl-512-crop

Repeats: 10
Total number of images: 50
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: True
Crop style: random
Crop aspect: square

davinci-sdxl-1024-crop

Repeats: 10
Total number of images: 50
Total number of aspect buckets: 1
Resolution: 1.048576 megapixels
Cropped: True
Crop style: random
Crop aspect: square

Inference

import torch
from diffusers import DiffusionPipeline
from lycoris import create_lycoris_from_weights

model_id = 'stabilityai/stable-diffusion-xl-base-1.0'
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 = "DaVinciXL, One mechanical device with gears and levers, no human subjects, one item in the image."
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=30,
    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=4.2,
    guidance_rescale=0.0,
).images[0]
image.save("output.png", format="PNG")

davidrd123
/

davinci-sdxl-lora-01