jyoung105/sd15_perturbed_attention_guidance_i2i

Perturbed-Attention Guidance for SD 1.5 (i2i)

The original Perturbed-Attention Guidance for unconditional models and SD1.5 by Hyoungwon Cho is availiable at hyoungwoncho/sd_perturbed_attention_guidance

Project / arXiv / GitHub

This repository is just a simple implementation of the Perturbed-Attention Guidance (PAG) on Stable Diffusion 1.5 (SD 1.5) for the 🧨 diffusers library to "image-to-image".

Quickstart

Loading Custom Pipeline:

from diffusers import StableDiffusionImg2ImgPipeline
pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    custom_pipeline="jyoung105/sd15_perturbed_attention_guidance_i2i",
    torch_dtype=torch.float16
)
device="cuda"
pipe = pipe.to(device)

Unconditional sampling with PAG:

output = pipe(
        "",
        image=init_image,
        strength=0.6,
        num_inference_steps=50,
        guidance_scale=0.0,
        pag_scale=5.0,
        pag_applied_layers_index=['m0']
    ).images

Sampling with PAG and CFG:

output = pipe(
        "A man with hoodie on is looking at sky, photo",
        image=init_image,
        strength=0.6,
        num_inference_steps=50,
        guidance_scale=0.0,
        pag_scale=5.0,
        pag_applied_layers_index=['m0']
    ).images

Parameters

guidance_scale : guidance scale of CFG (ex: 7.5)

pag_scale : guidance scale of PAG (ex: 4.0)

pag_applied_layers: layer to apply perturbation (ex: ['mid'])

pag_applied_layers_index : index of the layers to apply perturbation (ex: ['m0', 'm1'])