DiffSynth-Studio/LoRAFusion-preview-FLUX.1-dev

LoRA Fuser (FLUX.1-dev)

This model is designed to harmonize conflicting relationships when multiple LoRA models are loaded simultaneously, preventing image degradation.

For details, please refer to our paper: AutoLoRA: Automatic LoRA Retrieval and Fine-Grained Gated Fusion for Text-to-Image Generation

We are still optimizing this model. The current version may produce unstable results with certain LoRAs. Please stay tuned for future updates.

Result Demonstration

Prompt: a cat

	LoRA 1	LoRA 2	LoRA 3	LoRA 4
LoRA 1
LoRA 2
LoRA 3
LoRA 4

Prompt: a car

	LoRA 1	LoRA 2	LoRA 3	LoRA 4
LoRA 1
LoRA 2
LoRA 3
LoRA 4

Inference Code

git clone https://github.com/modelscope/DiffSynth-Studio.git  
cd DiffSynth-Studio
pip install -e .

import torch
from diffsynth.pipelines.flux_image_new import FluxImagePipeline, ModelConfig

        
pipe = FluxImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="black-forest-labs/FLUX.1-dev", origin_file_pattern="flux1-dev.safetensors"),
        ModelConfig(model_id="black-forest-labs/FLUX.1-dev", origin_file_pattern="text_encoder/model.safetensors"),
        ModelConfig(model_id="black-forest-labs/FLUX.1-dev", origin_file_pattern="text_encoder_2/"),
        ModelConfig(model_id="black-forest-labs/FLUX.1-dev", origin_file_pattern="ae.safetensors"),
        ModelConfig(model_id="DiffSynth-Studio/LoRAFusion-preview-FLUX.1-dev", origin_file_pattern="model.safetensors"),
    ],
)
pipe.enable_lora_magic()

pipe.load_lora(
    pipe.dit,
    ModelConfig(model_id="cancel13/cxsk", origin_file_pattern="30.safetensors"),
    hotload=True,
)
pipe.load_lora(
    pipe.dit,
    ModelConfig(model_id="DiffSynth-Studio/ArtAug-lora-FLUX.1dev-v1", origin_file_pattern="merged_lora.safetensors"),
    hotload=True,
)
image = pipe(prompt="a cat", seed=0)
image.save("image_fused.jpg")