runchat/lora-9556a6ce-b874-444b-82ea-359fe3d393eb-ds4t4d

Flux LoRA: pers style

This is a LoRA (Low-Rank Adaptation) model for Flux.1-dev fine-tuned on images with the trigger word pers style.

Files

• pytorch_lora_weights.safetensors: Diffusers format (use with diffusers library)
• pytorch_lora_weights_webui.safetensors: Kohya format (use with AUTOMATIC1111, ComfyUI, etc.)

Usage

Diffusers Library

from diffusers import FluxPipeline
import torch

# Load base model
pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-dev",
    torch_dtype=torch.bfloat16
)

# Load LoRA weights (diffusers format)
pipe.load_lora_weights("runchat/lora-9556a6ce-b874-444b-82ea-359fe3d393eb-ds4t4d", weight_name="pytorch_lora_weights.safetensors")
pipe = pipe.to("cuda")

# Generate image
prompt = "a photo of a pers style style"
image = pipe(prompt, num_inference_steps=50, guidance_scale=3.5).images[0]
image.save("output.png")

WebUI (AUTOMATIC1111, ComfyUI, etc.)

Download the pytorch_lora_weights_webui.safetensors file and place it in your WebUI's LoRA directory. Use the trigger word pers style in your prompts.

Training Details

• Base model: black-forest-labs/FLUX.1-dev
• Training steps: 500
• Learning rate: 0.001
• Batch size: 2
• LoRA rank: 16
• Trigger word: pers style

License

This model is trained on Flux.1-dev and inherits its non-commercial license. Please see the license for usage restrictions.