SurenaSa/IOAI2024_CV_Problem_V6_LoRA

Giraffe-Zebra Concept Swap LoRA

This LoRA adapter swaps the concepts of giraffes and zebras in Stable Diffusion.

Training Details

• Base Model: lambdalabs/miniSD-diffusers
• LoRA Rank: 64
• Training Epochs: 30
• Learning Rate: 0.0002

Usage

from diffusers import StableDiffusionPipeline
from peft import PeftModel
import torch

# Load base model
pipe = StableDiffusionPipeline.from_pretrained(
    "lambdalabs/miniSD-diffusers",
    torch_dtype=torch.float16
)

# Load LoRA weights
pipe.unet = PeftModel.from_pretrained(
    pipe.unet,
    "SurenaSa/IOAI2024_CV_Problem_V6_LoRA",
    is_trainable=False
)

pipe.to("cuda")

# Generate
prompt = "a photo of a giraffe"  # Will generate a zebra!
image = pipe(prompt, num_inference_steps=30, guidance_scale=7.5).images[0]
image.save("output.png")

Examples

• Prompt: "a photo of a giraffe" → Generates zebra
• Prompt: "a photo of a zebra" → Generates giraffe