simpletuner-sana-lora-mikrei-3e-4

This is a LyCORIS adapter derived from terminusresearch/sana-1.6b-1024px.

No validation prompt was used during training.

None

Validation settings

CFG: 4.0
CFG Rescale: 0.0
Steps: 30
Sampler: sana
Seed: 42
Resolutions: 1024x1024,1280x768

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 of mikrei eating a delicious slice of pizza at a restaurant

Negative Prompt
blurry, cropped, ugly

Prompt
A photo of mikrei feeding a dinousaur in a jungle with lush vegetation and cinematic lighting

Negative Prompt
blurry, cropped, ugly

Prompt
professional portrait of serious mikrei in a cockpit with instruments piloting a luxurious private jet in dramatic weather

Negative Prompt
blurry, cropped, ugly

Prompt
close-up portrait of mikrei on a throne in a magnificent palace wearing a crown and royal attire

Negative Prompt
blurry, cropped, ugly

Prompt
a high-quality, detailed photograph of mikrei as a sous-chef, immersed in the art of culinary creation

Negative Prompt
blurry, cropped, ugly

Prompt
a lifelike and intimate portrait of mikrei, showcasing his unique personality and charm

Negative Prompt
blurry, cropped, ugly

Prompt
a cinematic, visually stunning photo of mikrei, emphasizing his dramatic and captivating presence

Negative Prompt
blurry, cropped, ugly

Prompt
an elegant and timeless portrait of mikrei, exuding grace and sophistication

Negative Prompt
blurry, cropped, ugly

Prompt
a dynamic and adventurous photo of mikrei, captured in an exciting, action-filled moment

Negative Prompt
blurry, cropped, ugly

Prompt
a mysterious and enigmatic portrait of mikrei, shrouded in shadows and intrigue

Negative Prompt
blurry, cropped, ugly

Prompt
a vintage-style portrait of mikrei, evoking the charm and nostalgia of a bygone era

Negative Prompt
blurry, cropped, ugly

Prompt
an artistic and abstract representation of mikrei, blending creativity with visual storytelling

Negative Prompt
blurry, cropped, ugly

Prompt
a futuristic and cutting-edge portrayal of mikrei, set against a backdrop of advanced technology

Negative Prompt
blurry, cropped, ugly

Prompt
a beautifully crafted portrait of a woman, highlighting her natural beauty and unique features

Negative Prompt
blurry, cropped, ugly

Prompt
a powerful and striking portrait of a man, capturing his strength and character

Negative Prompt
blurry, cropped, ugly

Prompt
a playful and spirited portrait of a boy, capturing youthful energy and innocence

Negative Prompt
blurry, cropped, ugly

Prompt
a charming and vibrant portrait of a girl, emphasizing her bright personality and joy

Negative Prompt
blurry, cropped, ugly

Prompt
a heartwarming and cohesive family portrait, showcasing the bonds and connections between loved ones

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: 5
Training steps: 10000
Learning rate: 0.0003
- Learning rate schedule: polynomial
- Warmup steps: 100
Max grad norm: 2.0
Effective batch size: 1
- Micro-batch size: 1
- Gradient accumulation steps: 1
- Number of GPUs: 1
Gradient checkpointing: False
Prediction type: epsilon (extra parameters=['training_scheduler_timestep_spacing=trailing', 'inference_scheduler_timestep_spacing=trailing'])
Optimizer: optimi-stableadamw
Trainable parameter precision: Pure BF16
Caption dropout probability: 10.0%

LyCORIS Config:

{
    "algo": "lokr",
    "multiplier": 1.0,
    "linear_dim": 10000,
    "linear_alpha": 1,
    "factor": 8,
    "apply_preset": {
        "target_module": [
            "Attention",
            "FeedForward"
        ],
        "module_algo_map": {
            "Attention": {
                "factor": 8
            },
            "FeedForward": {
                "factor": 4
            }
        }
    }
}

Datasets

mikrei-data-512px

Repeats: 100
Total number of images: 17
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: False
Crop style: None
Crop aspect: None
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 = 'terminusresearch/sana-1.6b-1024px'
adapter_repo_id = 'mtreinik/simpletuner-sana-lora-mikrei-3e-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."
negative_prompt = 'blurry, cropped, ugly'

## 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,
    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(42),
    width=1024,
    height=1024,
    guidance_scale=4.0,
    guidance_rescale=0.0,
).images[0]
image.save("output.png", format="PNG")

Exponential Moving Average (EMA)

SimpleTuner generates a safetensors variant of the EMA weights and a pt file.

The safetensors file is intended to be used for inference, and the pt file is for continuing finetuning.

The EMA model may provide a more well-rounded result, but typically will feel undertrained compared to the full model as it is a running decayed average of the model weights.

mtreinik
/

simpletuner-sana-lora-mikrei-3e-4