pipeline.py

# Reference: https://github.com/cloneofsimo/sdxl_inversions/blob/master/pnp_pipeline.py

import torch
import torch.nn.functional as F
from typing import Any, Callable, Dict, List, Optional, Tuple, Union

from diffusers import (
    StableDiffusionXLImg2ImgPipeline,
    DDIMScheduler,
)

from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl import (
    rescale_noise_cfg,
    StableDiffusionXLPipelineOutput,
)

from diffusers.image_processor import PipelineImageInput

import PIL
import numpy as np
from tqdm import tqdm

def _backward_ddim(x_tm1, alpha_t, alpha_tm1, eps_xt):
    """
    let a = alpha_t, b = alpha_{t - 1}
    We have a > b,
    x_{t} - x_{t - 1} = sqrt(a) ((sqrt(1/b) - sqrt(1/a)) * x_{t-1} + (sqrt(1/a - 1) - sqrt(1/b - 1)) * eps_{t-1})
    From https://arxiv.org/pdf/2105.05233.pdf, section F.
    """

    a, b = alpha_t, alpha_tm1
    sa = a**0.5
    sb = b**0.5

    return sa * ((1 / sb) * x_tm1 + ((1 / a - 1) ** 0.5 - (1 / b - 1) ** 0.5) * eps_xt)

class SDXLDDIMPipeline(StableDiffusionXLImg2ImgPipeline):
    @torch.no_grad()
    def __call__(
        self,
        prompt: Union[str, List[str]] = None,
        prompt_2: Optional[Union[str, List[str]]] = None,
        image: Union[
            torch.FloatTensor,
            PIL.Image.Image,
            np.ndarray,
            List[torch.FloatTensor],
            List[PIL.Image.Image],
            List[np.ndarray],
        ] = None,
        strength: float = 0.3,
        num_inference_steps: int = 50,
        denoising_start: Optional[float] = None,
        denoising_end: Optional[float] = None,
        guidance_scale: float = 5.0,
        negative_prompt: Optional[Union[str, List[str]]] = None,
        negative_prompt_2: Optional[Union[str, List[str]]] = None,
        num_images_per_prompt: Optional[int] = 1,
        eta: float = 0.0,
        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
        latents: Optional[torch.FloatTensor] = None,
        prompt_embeds: Optional[torch.FloatTensor] = None,
        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
        negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
        output_type: Optional[str] = "pil",
        return_dict: bool = True,
        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
        callback_steps: int = 1,
        cross_attention_kwargs: Optional[Dict[str, Any]] = None,
        guidance_rescale: float = 0.0,
        original_size: Tuple[int, int] = None,
        crops_coords_top_left: Tuple[int, int] = (0, 0),
        target_size: Tuple[int, int] = None,
        negative_original_size: Optional[Tuple[int, int]] = None,
        negative_crops_coords_top_left: Tuple[int, int] = (0, 0),
        negative_target_size: Optional[Tuple[int, int]] = None,
        aesthetic_score: float = 6.0,
        negative_aesthetic_score: float = 2.5,
        guided_latent: Optional[torch.FloatTensor] = None,
    ):
        self.scheduler = DDIMScheduler.from_config(self.scheduler.config)

        # 1. Check inputs. Raise error if not correct
        self.check_inputs(
            prompt,
            prompt_2,
            strength,
            num_inference_steps,
            callback_steps,
            negative_prompt,
            negative_prompt_2,
            prompt_embeds,
            negative_prompt_embeds,
        )

        # 2. Define call parameters
        if prompt is not None and isinstance(prompt, str):
            batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            batch_size = prompt_embeds.shape[0]

        device = self._execution_device

        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
        # corresponds to doing no classifier free guidance.
        do_classifier_free_guidance = False
        # 3. Encode input prompt
        text_encoder_lora_scale = (
            cross_attention_kwargs.get("scale", None)
            if cross_attention_kwargs is not None
            else None
        )
        (
            prompt_embeds,
            negative_prompt_embeds,
            pooled_prompt_embeds,
            negative_pooled_prompt_embeds,
        ) = self.encode_prompt(
            prompt=prompt,
            prompt_2=prompt_2,
            device=device,
            num_images_per_prompt=num_images_per_prompt,
            do_classifier_free_guidance=do_classifier_free_guidance,
            negative_prompt=negative_prompt,
            negative_prompt_2=negative_prompt_2,
            prompt_embeds=prompt_embeds,
            negative_prompt_embeds=negative_prompt_embeds,
            pooled_prompt_embeds=pooled_prompt_embeds,
            negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
            lora_scale=text_encoder_lora_scale,
        )

        # 4. Preprocess image
        image = self.image_processor.preprocess(image)

        self.scheduler.set_timesteps(num_inference_steps, device=device)
       
        # 6. Prepare latent variables
        latents = self.prepare_latents(
            image,
            None,
            batch_size,
            num_images_per_prompt,
            prompt_embeds.dtype,
            device,
            generator,
            False,
        )
       
        height, width = latents.shape[-2:]
        height = height * self.vae_scale_factor
        width = width * self.vae_scale_factor

        original_size = original_size or (height, width)
        target_size = target_size or (height, width)

        # 8. Prepare added time ids & embeddings
        if negative_original_size is None:
            negative_original_size = original_size
        if negative_target_size is None:
            negative_target_size = target_size

        add_text_embeds = pooled_prompt_embeds

        if self.text_encoder_2 is None:
            text_encoder_projection_dim = int(pooled_prompt_embeds.shape[-1])
        else:
            text_encoder_projection_dim = self.text_encoder_2.config.projection_dim

        add_time_ids, add_neg_time_ids = self._get_add_time_ids(
            original_size,
            crops_coords_top_left,
            target_size,
            aesthetic_score,
            negative_aesthetic_score,
            negative_original_size,
            negative_crops_coords_top_left,
            negative_target_size,
            dtype=prompt_embeds.dtype,
            text_encoder_projection_dim=text_encoder_projection_dim,
        )
        add_time_ids = add_time_ids.repeat(batch_size * num_images_per_prompt, 1)

        prompt_embeds = prompt_embeds.to(device)
        add_text_embeds = add_text_embeds.to(device)
        add_time_ids = add_time_ids.to(device)

        added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids, "latent_embeds": guided_latent}
        prev_timestep = None

        timesteps = self.scheduler.timesteps
        # 8.1 Apply denoising_end
        if denoising_end is not None and isinstance(denoising_end, float) and denoising_end > 0 and denoising_end < 1:
            discrete_timestep_cutoff = int(
                round(
                    self.scheduler.config.num_train_timesteps
                    - (denoising_end * self.scheduler.config.num_train_timesteps)
                )
            )
            num_inference_steps = len(list(filter(lambda ts: ts >= discrete_timestep_cutoff, timesteps)))
            timesteps = reversed(timesteps)[:num_inference_steps]
        else:
            timesteps = reversed(timesteps)
        ###

        for t in tqdm(timesteps):
            latent_model_input = latents
            noise_pred = self.unet(
                latent_model_input,
                t,
                encoder_hidden_states=prompt_embeds,
                cross_attention_kwargs=cross_attention_kwargs,
                added_cond_kwargs=added_cond_kwargs,
                return_dict=False,
            )[0]

            alpha_prod_t = self.scheduler.alphas_cumprod[t]
            alpha_prod_t_prev = (
                self.scheduler.alphas_cumprod[prev_timestep]
                if prev_timestep is not None
                else self.scheduler.final_alpha_cumprod
            )
            prev_timestep = t

            latents = _backward_ddim(
                x_tm1=latents,
                alpha_t=alpha_prod_t,
                alpha_tm1=alpha_prod_t_prev,
                eps_xt=noise_pred,
            )

        image = latents
        return StableDiffusionXLPipelineOutput(images=image)

class SDXLImg2ImgPipeline(StableDiffusionXLImg2ImgPipeline):
    def minimize_changes_l0(self, noise_pred_text, noise_pred_uncond, quantile=0.7):
        score_delta = noise_pred_text - noise_pred_uncond
        score_delta = score_delta.float()
        if quantile > 0:
            threshold = score_delta.abs().quantile(quantile)
        else:
            threshold = -quantile  # if quantile is negative, use it as a fixed threshold
        score_delta -= score_delta.clamp(-threshold, threshold)
        return score_delta.half()
    
    def minimize_changes_l1(self, noise_pred_text, noise_pred_uncond, quantile=0.7):
        score_delta = noise_pred_text - noise_pred_uncond
        score_delta = score_delta.float()
        if quantile > 0:
            threshold = score_delta.abs().quantile(quantile)
        else:
            threshold = -quantile  # if quantile is negative, use it as a fixed threshold
        score_delta -= score_delta.clamp(-threshold, threshold)
        score_delta = torch.where(score_delta > 0, score_delta-threshold, score_delta)
        score_delta = torch.where(score_delta < 0, score_delta+threshold, score_delta)
        return score_delta.half()

    @torch.no_grad()
    def __call__(
        self,
        prompt: Union[str, List[str]] = None,
        prompt_2: Optional[Union[str, List[str]]] = None,
        image: PipelineImageInput = None,
        strength: float = 0.3,
        num_inference_steps: int = 50,
        denoising_start: Optional[float] = None,
        denoising_end: Optional[float] = None,
        guidance_scale: float = 5.0,
        negative_prompt: Optional[Union[str, List[str]]] = None,
        negative_prompt_2: Optional[Union[str, List[str]]] = None,
        num_images_per_prompt: Optional[int] = 1,
        eta: float = 0.0,
        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
        latents: Optional[torch.FloatTensor] = None,
        prompt_embeds: Optional[torch.FloatTensor] = None,
        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
        negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
        output_type: Optional[str] = "pil",
        return_dict: bool = True,
        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
        callback_steps: int = 1,
        cross_attention_kwargs: Optional[Dict[str, Any]] = None,
        guidance_rescale: float = 0.0,
        original_size: Tuple[int, int] = None,
        crops_coords_top_left: Tuple[int, int] = (0, 0),
        target_size: Tuple[int, int] = None,
        negative_original_size: Optional[Tuple[int, int]] = None,
        negative_crops_coords_top_left: Tuple[int, int] = (0, 0),
        negative_target_size: Optional[Tuple[int, int]] = None,
        aesthetic_score: float = 6.0,
        negative_aesthetic_score: float = 2.5,
        clip_skip: Optional[int] = None,
        guided_latent: Optional[torch.FloatTensor] = None,
        regularization: Optional[str] = None,
    ):
        # 1. Check inputs. Raise error if not correct
        self.check_inputs(
            prompt,
            prompt_2,
            strength,
            num_inference_steps,
            callback_steps,
            negative_prompt,
            negative_prompt_2,
            prompt_embeds,
            negative_prompt_embeds,
        )

        # 2. Define call parameters
        if prompt is not None and isinstance(prompt, str):
            batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            batch_size = prompt_embeds.shape[0]

        device = self._execution_device

        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
        # corresponds to doing no classifier free guidance.
        do_classifier_free_guidance = guidance_scale > 1.0

        # 3. Encode input prompt
        text_encoder_lora_scale = (
            cross_attention_kwargs.get("scale", None) if cross_attention_kwargs is not None else None
        )
        (
            prompt_embeds,
            negative_prompt_embeds,
            pooled_prompt_embeds,
            negative_pooled_prompt_embeds,
        ) = self.encode_prompt(
            prompt=prompt,
            prompt_2=prompt_2,
            device=device,
            num_images_per_prompt=num_images_per_prompt,
            do_classifier_free_guidance=do_classifier_free_guidance,
            negative_prompt=negative_prompt,
            negative_prompt_2=negative_prompt_2,
            prompt_embeds=prompt_embeds,
            negative_prompt_embeds=negative_prompt_embeds,
            pooled_prompt_embeds=pooled_prompt_embeds,
            negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
            lora_scale=text_encoder_lora_scale,
            clip_skip=clip_skip,
        )

        # 4. Preprocess image
        image = self.image_processor.preprocess(image)

        # 5. Prepare timesteps
        def denoising_value_valid(dnv):
            return isinstance(denoising_end, float) and 0 < dnv < 1

        self.scheduler.set_timesteps(num_inference_steps, device=device)
        timesteps, num_inference_steps = self.get_timesteps(
            num_inference_steps, strength, device, denoising_start=denoising_start if denoising_value_valid else None
        )
        latent_timestep = timesteps[:1].repeat(batch_size * num_images_per_prompt)

        add_noise = True if denoising_start is None else False
        # 6. Prepare latent variables
        latents = self.prepare_latents(
            image,
            latent_timestep,
            batch_size,
            num_images_per_prompt,
            prompt_embeds.dtype,
            device,
            generator,
            add_noise,
        )
        # 7. Prepare extra step kwargs.
        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)

        height, width = latents.shape[-2:]
        height = height * self.vae_scale_factor
        width = width * self.vae_scale_factor

        original_size = original_size or (height, width)
        target_size = target_size or (height, width)

        # 8. Prepare added time ids & embeddings
        if negative_original_size is None:
            negative_original_size = original_size
        if negative_target_size is None:
            negative_target_size = target_size

        add_text_embeds = pooled_prompt_embeds
        if self.text_encoder_2 is None:
            text_encoder_projection_dim = int(pooled_prompt_embeds.shape[-1])
        else:
            text_encoder_projection_dim = self.text_encoder_2.config.projection_dim

        add_time_ids, add_neg_time_ids = self._get_add_time_ids(
            original_size,
            crops_coords_top_left,
            target_size,
            aesthetic_score,
            negative_aesthetic_score,
            negative_original_size,
            negative_crops_coords_top_left,
            negative_target_size,
            dtype=prompt_embeds.dtype,
            text_encoder_projection_dim=text_encoder_projection_dim,
        )
        add_time_ids = add_time_ids.repeat(batch_size * num_images_per_prompt, 1)

        if do_classifier_free_guidance:
            prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds], dim=0)
            add_text_embeds = torch.cat([negative_pooled_prompt_embeds, add_text_embeds], dim=0)
            add_neg_time_ids = add_neg_time_ids.repeat(batch_size * num_images_per_prompt, 1)
            add_time_ids = torch.cat([add_neg_time_ids, add_time_ids], dim=0)
            guided_latent = torch.stack([guided_latent, guided_latent], dim=0)

        prompt_embeds = prompt_embeds.to(device)
        add_text_embeds = add_text_embeds.to(device)
        add_time_ids = add_time_ids.to(device)

        # 9. Denoising loop
        num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)

        # 9.1 Apply denoising_end
        if (
            denoising_end is not None
            and denoising_start is not None
            and denoising_value_valid(denoising_end)
            and denoising_value_valid(denoising_start)
            and denoising_start >= denoising_end
        ):
            raise ValueError(
                f"`denoising_start`: {denoising_start} cannot be larger than or equal to `denoising_end`: "
                + f" {denoising_end} when using type float."
            )
        elif denoising_end is not None and denoising_value_valid(denoising_end):
            discrete_timestep_cutoff = int(
                round(
                    self.scheduler.config.num_train_timesteps
                    - (denoising_end * self.scheduler.config.num_train_timesteps)
                )
            )
            num_inference_steps = len(list(filter(lambda ts: ts >= discrete_timestep_cutoff, timesteps)))
            timesteps = timesteps[:num_inference_steps]

        with self.progress_bar(total=num_inference_steps) as progress_bar:
            for i, t in enumerate(timesteps):
                # expand the latents if we are doing classifier free guidance
                latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents

                latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)

                # predict the noise residual
                added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids, "latent_embeds": guided_latent}
                noise_pred = self.unet(
                    latent_model_input,
                    t,
                    encoder_hidden_states=prompt_embeds,
                    cross_attention_kwargs=cross_attention_kwargs,
                    added_cond_kwargs=added_cond_kwargs,
                    return_dict=False,
                )[0]

                # perform guidance
                if do_classifier_free_guidance:
                    noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                    # noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
                    if regularization == "l0":
                        sparse_diff = self.minimize_changes_l0(noise_pred_text, noise_pred_uncond)
                    elif regularization == "l1":
                        sparse_diff = self.minimize_changes_l1(noise_pred_text, noise_pred_uncond)
                    else:
                        sparse_diff = noise_pred_text - noise_pred_uncond
                    noise_pred = noise_pred_uncond + guidance_scale * sparse_diff

                if do_classifier_free_guidance and guidance_rescale > 0.0:
                    # Based on 3.4. in https://arxiv.org/pdf/2305.08891.pdf
                    noise_pred = rescale_noise_cfg(noise_pred, noise_pred_text, guidance_rescale=guidance_rescale)

                # compute the previous noisy sample x_t -> x_t-1
                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]

                # call the callback, if provided
                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
                    progress_bar.update()
                    if callback is not None and i % callback_steps == 0:
                        step_idx = i // getattr(self.scheduler, "order", 1)
                        callback(step_idx, t, latents)

        if not output_type == "latent":
            # make sure the VAE is in float32 mode, as it overflows in float16
            needs_upcasting = self.vae.dtype == torch.float16 and self.vae.config.force_upcast

            if needs_upcasting:
                self.upcast_vae()
                latents = latents.to(next(iter(self.vae.post_quant_conv.parameters())).dtype)

            image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False)[0]

            # cast back to fp16 if needed
            if needs_upcasting:
                self.vae.to(dtype=torch.float16)
        else:
            image = latents
            return StableDiffusionXLPipelineOutput(images=image)

        # apply watermark if available
        if self.watermark is not None:
            image = self.watermark.apply_watermark(image)

        image = self.image_processor.postprocess(image, output_type=output_type)

        # Offload all models
        self.maybe_free_model_hooks()

        if not return_dict:
            return (image,)

        return StableDiffusionXLPipelineOutput(images=image)