Diffusers documentation

Self-Attention Guidance (SAG)

You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v0.14.0).
Join the Hugging Face community

to get started

# Self-Attention Guidance (SAG)

## Overview

Self-Attention Guidance by Susung Hong et al.

The abstract of the paper is the following:

Denoising diffusion models (DDMs) have been drawing much attention for their appreciable sample quality and diversity. Despite their remarkable performance, DDMs remain black boxes on which further study is necessary to take a profound step. Motivated by this, we delve into the design of conventional U-shaped diffusion models. More specifically, we investigate the self-attention modules within these models through carefully designed experiments and explore their characteristics. In addition, inspired by the studies that substantiate the effectiveness of the guidance schemes, we present plug-and-play diffusion guidance, namely Self-Attention Guidance (SAG), that can drastically boost the performance of existing diffusion models. Our method, SAG, extracts the intermediate attention map from a diffusion model at every iteration and selects tokens above a certain attention score for masking and blurring to obtain a partially blurred input. Subsequently, we measure the dissimilarity between the predicted noises obtained from feeding the blurred and original input to the diffusion model and leverage it as guidance. With this guidance, we observe apparent improvements in a wide range of diffusion models, e.g., ADM, IDDPM, and Stable Diffusion, and show that the results further improve by combining our method with the conventional guidance scheme. We provide extensive ablation studies to verify our choices.

Resources:

## Available Pipelines:

StableDiffusionSAGPipeline Text-to-Image Generation Colab

## Usage example

import torch
from diffusers import StableDiffusionSAGPipeline
from accelerate.utils import set_seed

pipe = StableDiffusionSAGPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", torch_dtype=torch.float16)
pipe = pipe.to("cuda")

seed = 8978
prompt = "."
guidance_scale = 7.5
num_images_per_prompt = 1

sag_scale = 1.0

set_seed(seed)
images = pipe(
prompt, num_images_per_prompt=num_images_per_prompt, guidance_scale=guidance_scale, sag_scale=sag_scale
).images
images[0].save("example.png")

## StableDiffusionSAGPipeline

### class diffusers.StableDiffusionSAGPipeline

< >

( vae: AutoencoderKL text_encoder: CLIPTextModel tokenizer: CLIPTokenizer unet: UNet2DConditionModel scheduler: KarrasDiffusionSchedulers safety_checker: StableDiffusionSafetyChecker feature_extractor: CLIPImageProcessor requires_safety_checker: bool = True )

Parameters

• vae (AutoencoderKL) — Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
• text_encoder (CLIPTextModel) — Frozen text-encoder. Stable Diffusion uses the text portion of CLIP, specifically the clip-vit-large-patch14 variant.
• tokenizer (CLIPTokenizer) — Tokenizer of class CLIPTokenizer.
• unet (UNet2DConditionModel) — Conditional U-Net architecture to denoise the encoded image latents.
• scheduler (SchedulerMixin) — A scheduler to be used in combination with unet to denoise the encoded image latents. Can be one of DDIMScheduler, LMSDiscreteScheduler, or PNDMScheduler.
• safety_checker (StableDiffusionSafetyChecker) — Classification module that estimates whether generated images could be considered offensive or harmful. Please, refer to the model card for details.
• feature_extractor (CLIPImageProcessor) — Model that extracts features from generated images to be used as inputs for the safety_checker.

Pipeline for text-to-image generation using Stable Diffusion.

This model inherits from DiffusionPipeline. Check the superclass documentation for the generic methods the library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)

#### __call__

< >

( prompt: typing.Union[str, typing.List[str]] = None height: typing.Optional[int] = None width: typing.Optional[int] = None num_inference_steps: int = 50 guidance_scale: float = 7.5 sag_scale: float = 0.75 negative_prompt: typing.Union[str, typing.List[str], NoneType] = None num_images_per_prompt: typing.Optional[int] = 1 eta: float = 0.0 generator: typing.Union[torch._C.Generator, typing.List[torch._C.Generator], NoneType] = None latents: typing.Optional[torch.FloatTensor] = None prompt_embeds: typing.Optional[torch.FloatTensor] = None negative_prompt_embeds: typing.Optional[torch.FloatTensor] = None output_type: typing.Optional[str] = 'pil' return_dict: bool = True callback: typing.Union[typing.Callable[[int, int, torch.FloatTensor], NoneType], NoneType] = None callback_steps: typing.Optional[int] = 1 cross_attention_kwargs: typing.Union[typing.Dict[str, typing.Any], NoneType] = None ) StableDiffusionPipelineOutput or tuple

Parameters

• prompt (str or List[str], optional) — The prompt or prompts to guide the image generation. If not defined, one has to pass prompt_embeds. instead.
• height (int, optional, defaults to self.unet.config.sample_size * self.vae_scale_factor) — The height in pixels of the generated image.
• width (int, optional, defaults to self.unet.config.sample_size * self.vae_scale_factor) — The width in pixels of the generated image.
• num_inference_steps (int, optional, defaults to 50) — The number of denoising steps. More denoising steps usually lead to a higher quality image at the expense of slower inference.
• guidance_scale (float, optional, defaults to 7.5) — Guidance scale as defined in Classifier-Free Diffusion Guidance. guidance_scale is defined as w of equation 2. of Imagen Paper. Guidance scale is enabled by setting guidance_scale > 1. Higher guidance scale encourages to generate images that are closely linked to the text prompt, usually at the expense of lower image quality.
• sag_scale (float, optional, defaults to 0.75) — SAG scale as defined in [Improving Sample Quality of Diffusion Models Using Self-Attention Guidance] (https://arxiv.org/abs/2210.00939). sag_scale is defined as s_s of equation (24) of SAG paper: https://arxiv.org/pdf/2210.00939.pdf. Typically chosen between [0, 1.0] for better quality.
• negative_prompt (str or List[str], optional) — The prompt or prompts not to guide the image generation. If not defined, one has to pass negative_prompt_embeds instead. Ignored when not using guidance (i.e., ignored if guidance_scale is less than 1).
• num_images_per_prompt (int, optional, defaults to 1) — The number of images to generate per prompt.
• eta (float, optional, defaults to 0.0) — Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to schedulers.DDIMScheduler, will be ignored for others.
• generator (torch.Generator or List[torch.Generator], optional) — One or a list of torch generator(s) to make generation deterministic.
• latents (torch.FloatTensor, optional) — Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image generation. Can be used to tweak the same generation with different prompts. If not provided, a latents tensor will ge generated by sampling using the supplied random generator.
• prompt_embeds (torch.FloatTensor, optional) — Pre-generated text embeddings. Can be used to easily tweak text inputs, e.g. prompt weighting. If not provided, text embeddings will be generated from prompt input argument.
• negative_prompt_embeds (torch.FloatTensor, optional) — Pre-generated negative text embeddings. Can be used to easily tweak text inputs, e.g. prompt weighting. If not provided, negative_prompt_embeds will be generated from negative_prompt input argument.
• output_type (str, optional, defaults to "pil") — The output format of the generate image. Choose between PIL: PIL.Image.Image or np.array.
• return_dict (bool, optional, defaults to True) — Whether or not to return a StableDiffusionPipelineOutput instead of a plain tuple.
• callback (Callable, optional) — A function that will be called every callback_steps steps during inference. The function will be called with the following arguments: callback(step: int, timestep: int, latents: torch.FloatTensor).
• callback_steps (int, optional, defaults to 1) — The frequency at which the callback function will be called. If not specified, the callback will be called at every step.
• cross_attention_kwargs (dict, optional) — A kwargs dictionary that if specified is passed along to the AttentionProcessor as defined under self.processor in diffusers.cross_attention.

Returns

StableDiffusionPipelineOutput or tuple

StableDiffusionPipelineOutput if return_dict is True, otherwise a tuple. When returning a tuple, the first element is a list with the generated images, and the second element is a list of bools denoting whether the corresponding generated image likely represents "not-safe-for-work" (nsfw) content, according to the safety_checker.

Function invoked when calling the pipeline for generation.

Examples:

>>> import torch
>>> from diffusers import StableDiffusionSAGPipeline

>>> pipe = StableDiffusionSAGPipeline.from_pretrained(
...     "runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16
... )
>>> pipe = pipe.to("cuda")

>>> prompt = "a photo of an astronaut riding a horse on mars"
>>> image = pipe(prompt, sag_scale=0.75).images[0]

#### disable_vae_slicing

< >

( )

Disable sliced VAE decoding. If enable_vae_slicing was previously invoked, this method will go back to computing decoding in one step.

< >

( gpu_id = 0 )

Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet, text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a torch.device('meta') and loaded to GPU only when their specific submodule has its forwardmethod called. Note that offloading happens on a submodule basis. Memory savings are higher than withenable_model_cpu_offload, but performance is lower.

#### enable_vae_slicing

< >

( )

Enable sliced VAE decoding.

When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.