Mani Flux Dress

🚀 mani - flux - dress

這是一個基於[black - forest - labs/FLUX.1 - dev](https://huggingface.co/black - forest - labs/FLUX.1 - dev)的LyCORIS適配器。它可用於特定風格的圖像生成，在時尚攝影場景圖像生成方面具有一定優勢。

🚀 快速開始

推理示例

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 = 'PixelWormhole/mani-flux-dress'
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 = "A film still of a high fashion photoshoot featuring the _8FSL39XP1_ Dress, an elegant off-the-shoulder design in delicate white lace with floral motifs. The dress is form-fitting, mini-length, with long sleeves that flare slightly at the cuffs. The scene is set in a luxurious, modern Mediterranean architectural space with a sea view. A beautiful, young, natural-looking European woman with long hair is captured in a dynamic mid-motion pose, either mid-step or gently turning. She stands against a backdrop of clean white stone and curated designer furniture in neutral tones, with the calm, open sea visible in the distance. Her expression is serene and self-assured, her posture poised and refined. Natural daylight enhances the soft textures and casts gentle shadows across the scene. The color palette blends creamy whites, stone beiges, and hints of ocean blue."


## Optional: quantise the model to save on vram.
## Note: The model was not quantised during training, so it is not necessary to quantise it 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
model_output = pipeline(
    prompt=prompt,
    num_inference_steps=28,
    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=3.0,
).images[0]

model_output.save("output.png", format="PNG")

✨ 主要特性

• 基於LyCORIS技術，可有效適配基礎模型，提升特定場景的圖像生成效果。
• 支持多種推理設置，可靈活調整以滿足不同需求。

📚 詳細文檔

驗證設置

• CFG：3.0
• CFG Rescale：0.0
• 步數：28
• 採樣器：FlowMatchEulerDiscreteScheduler
• 隨機種子：42
• 分辨率：1024x1024
• 跳過層引導：無

注意：驗證設置不一定與訓練設置相同。

你可以在以下圖庫中找到一些示例圖像：

文本編碼器未進行訓練。你可以在推理時重用基礎模型的文本編碼器。

訓練設置

LyCORIS配置

{
    "algo": "lora",
    "multiplier": 1.0,
    "linear_dim": 64,
    "linear_alpha": 32,
    "apply_preset": {
        "target_module": [
            "Attention",
            "FeedForward"
        ],
        "module_algo_map": {
            "Attention": {
                "factor": 16
            },
            "FeedForward": {
                "factor": 8
            }
        }
    }
}

數據集

📄 許可證

許可證類型：other