diffusers/pipelines/versatile_diffusion/pipeline_versatile_diffusion_image_variation.py

# Copyright 2023 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import inspect
from typing import Callable, List, Optional, Union

import numpy as np
import PIL.Image
import torch
import torch.utils.checkpoint
from transformers import CLIPImageProcessor, CLIPVisionModelWithProjection

from ...image_processor import VaeImageProcessor
from ...models import AutoencoderKL, UNet2DConditionModel
from ...schedulers import KarrasDiffusionSchedulers
from ...utils import deprecate, logging
from ...utils.torch_utils import randn_tensor
from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput


logger = logging.get_logger(__name__)  # pylint: disable=invalid-name


class VersatileDiffusionImageVariationPipeline(DiffusionPipeline):
    r"""
    Pipeline for image variation using Versatile Diffusion.

    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
    implemented for all pipelines (downloading, saving, running on a particular device, etc.).

    Parameters:
        vqvae ([`VQModel`]):
            Vector-quantized (VQ) model to encode and decode images to and from latent representations.
        bert ([`LDMBertModel`]):
            Text-encoder model based on [`~transformers.BERT`].
        tokenizer ([`~transformers.BertTokenizer`]):
            A `BertTokenizer` to tokenize text.
        unet ([`UNet2DConditionModel`]):
            A `UNet2DConditionModel` 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`].
    """
    model_cpu_offload_seq = "bert->unet->vqvae"

    image_feature_extractor: CLIPImageProcessor
    image_encoder: CLIPVisionModelWithProjection
    image_unet: UNet2DConditionModel
    vae: AutoencoderKL
    scheduler: KarrasDiffusionSchedulers

    def __init__(
        self,
        image_feature_extractor: CLIPImageProcessor,
        image_encoder: CLIPVisionModelWithProjection,
        image_unet: UNet2DConditionModel,
        vae: AutoencoderKL,
        scheduler: KarrasDiffusionSchedulers,
    ):
        super().__init__()
        self.register_modules(
            image_feature_extractor=image_feature_extractor,
            image_encoder=image_encoder,
            image_unet=image_unet,
            vae=vae,
            scheduler=scheduler,
        )
        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)

    def _encode_prompt(self, prompt, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt):
        r"""
        Encodes the prompt into text encoder hidden states.

        Args:
            prompt (`str` or `List[str]`):
                prompt to be encoded
            device: (`torch.device`):
                torch device
            num_images_per_prompt (`int`):
                number of images that should be generated per prompt
            do_classifier_free_guidance (`bool`):
                whether to use classifier free guidance or not
            negative_prompt (`str` or `List[str]`):
                The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
                if `guidance_scale` is less than `1`).
        """

        def normalize_embeddings(encoder_output):
            embeds = self.image_encoder.vision_model.post_layernorm(encoder_output.last_hidden_state)
            embeds = self.image_encoder.visual_projection(embeds)
            embeds_pooled = embeds[:, 0:1]
            embeds = embeds / torch.norm(embeds_pooled, dim=-1, keepdim=True)
            return embeds

        if isinstance(prompt, torch.Tensor) and len(prompt.shape) == 4:
            prompt = list(prompt)

        batch_size = len(prompt) if isinstance(prompt, list) else 1

        # get prompt text embeddings
        image_input = self.image_feature_extractor(images=prompt, return_tensors="pt")
        pixel_values = image_input.pixel_values.to(device).to(self.image_encoder.dtype)
        image_embeddings = self.image_encoder(pixel_values)
        image_embeddings = normalize_embeddings(image_embeddings)

        # duplicate image embeddings for each generation per prompt, using mps friendly method
        bs_embed, seq_len, _ = image_embeddings.shape
        image_embeddings = image_embeddings.repeat(1, num_images_per_prompt, 1)
        image_embeddings = image_embeddings.view(bs_embed * num_images_per_prompt, seq_len, -1)

        # get unconditional embeddings for classifier free guidance
        if do_classifier_free_guidance:
            uncond_images: List[str]
            if negative_prompt is None:
                uncond_images = [np.zeros((512, 512, 3)) + 0.5] * batch_size
            elif type(prompt) is not type(negative_prompt):
                raise TypeError(
                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
                    f" {type(prompt)}."
                )
            elif isinstance(negative_prompt, PIL.Image.Image):
                uncond_images = [negative_prompt]
            elif batch_size != len(negative_prompt):
                raise ValueError(
                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
                    " the batch size of `prompt`."
                )
            else:
                uncond_images = negative_prompt

            uncond_images = self.image_feature_extractor(images=uncond_images, return_tensors="pt")
            pixel_values = uncond_images.pixel_values.to(device).to(self.image_encoder.dtype)
            negative_prompt_embeds = self.image_encoder(pixel_values)
            negative_prompt_embeds = normalize_embeddings(negative_prompt_embeds)

            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
            seq_len = negative_prompt_embeds.shape[1]
            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)

            # For classifier free guidance, we need to do two forward passes.
            # Here we concatenate the unconditional and conditional embeddings into a single batch
            # to avoid doing two forward passes
            image_embeddings = torch.cat([negative_prompt_embeds, image_embeddings])

        return image_embeddings

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
    def decode_latents(self, latents):
        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)

        latents = 1 / self.vae.config.scaling_factor * latents
        image = self.vae.decode(latents, return_dict=False)[0]
        image = (image / 2 + 0.5).clamp(0, 1)
        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
        return image

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
    def prepare_extra_step_kwargs(self, generator, eta):
        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
        # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
        # and should be between [0, 1]

        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
        extra_step_kwargs = {}
        if accepts_eta:
            extra_step_kwargs["eta"] = eta

        # check if the scheduler accepts generator
        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
        if accepts_generator:
            extra_step_kwargs["generator"] = generator
        return extra_step_kwargs

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_image_variation.StableDiffusionImageVariationPipeline.check_inputs
    def check_inputs(self, image, height, width, callback_steps):
        if (
            not isinstance(image, torch.Tensor)
            and not isinstance(image, PIL.Image.Image)
            and not isinstance(image, list)
        ):
            raise ValueError(
                "`image` has to be of type `torch.FloatTensor` or `PIL.Image.Image` or `List[PIL.Image.Image]` but is"
                f" {type(image)}"
            )

        if height % 8 != 0 or width % 8 != 0:
            raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")

        if (callback_steps is None) or (
            callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0)
        ):
            raise ValueError(
                f"`callback_steps` has to be a positive integer but is {callback_steps} of type"
                f" {type(callback_steps)}."
            )

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
    def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
        shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
        if isinstance(generator, list) and len(generator) != batch_size:
            raise ValueError(
                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
            )

        if latents is None:
            latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
        else:
            latents = latents.to(device)

        # scale the initial noise by the standard deviation required by the scheduler
        latents = latents * self.scheduler.init_noise_sigma
        return latents

    @torch.no_grad()
    def __call__(
        self,
        image: Union[PIL.Image.Image, List[PIL.Image.Image], torch.Tensor],
        height: Optional[int] = None,
        width: Optional[int] = None,
        num_inference_steps: int = 50,
        guidance_scale: float = 7.5,
        negative_prompt: 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,
        output_type: Optional[str] = "pil",
        return_dict: bool = True,
        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
        callback_steps: int = 1,
        **kwargs,
    ):
        r"""
        The call function to the pipeline for generation.

        Args:
            image (`PIL.Image.Image`, `List[PIL.Image.Image]` or `torch.Tensor`):
                The image prompt or prompts to guide the image generation.
            height (`int`, *optional*, defaults to `self.image_unet.config.sample_size * self.vae_scale_factor`):
                The height in pixels of the generated image.
            width (`int`, *optional*, defaults to `self.image_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):
                A higher guidance scale value encourages the model to generate images closely linked to the text
                `prompt` at the expense of lower image quality. Guidance scale is enabled when `guidance_scale > 1`.
            negative_prompt (`str` or `List[str]`, *optional*):
                The prompt or prompts to guide what to not include in image generation. If not defined, you need to
                pass `negative_prompt_embeds` instead. Ignored when not using guidance (`guidance_scale < 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) from the [DDIM](https://arxiv.org/abs/2010.02502) paper. Only applies
                to the [`~schedulers.DDIMScheduler`], and is ignored in other schedulers.
            generator (`torch.Generator`, *optional*):
                A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) 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 is generated by sampling using the supplied random `generator`.
            output_type (`str`, *optional*, defaults to `"pil"`):
                The output format of the generated image. Choose between `PIL.Image` or `np.array`.
            return_dict (`bool`, *optional*, defaults to `True`):
                Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
                plain tuple.
            callback (`Callable`, *optional*):
                A function that calls every `callback_steps` steps during inference. The function is 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 is called. If not specified, the callback is called at
                every step.

        Examples:

        ```py
        >>> from diffusers import VersatileDiffusionImageVariationPipeline
        >>> import torch
        >>> import requests
        >>> from io import BytesIO
        >>> from PIL import Image

        >>> # let's download an initial image
        >>> url = "https://huggingface.co/datasets/diffusers/images/resolve/main/benz.jpg"

        >>> response = requests.get(url)
        >>> image = Image.open(BytesIO(response.content)).convert("RGB")

        >>> pipe = VersatileDiffusionImageVariationPipeline.from_pretrained(
        ...     "shi-labs/versatile-diffusion", torch_dtype=torch.float16
        ... )
        >>> pipe = pipe.to("cuda")

        >>> generator = torch.Generator(device="cuda").manual_seed(0)
        >>> image = pipe(image, generator=generator).images[0]
        >>> image.save("./car_variation.png")
        ```

        Returns:
            [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
                If `return_dict` is `True`, [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] is returned,
                otherwise a `tuple` is returned where the first element is a list with the generated images.
        """
        # 0. Default height and width to unet
        height = height or self.image_unet.config.sample_size * self.vae_scale_factor
        width = width or self.image_unet.config.sample_size * self.vae_scale_factor

        # 1. Check inputs. Raise error if not correct
        self.check_inputs(image, height, width, callback_steps)

        # 2. Define call parameters
        batch_size = 1 if isinstance(image, PIL.Image.Image) else len(image)
        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
        image_embeddings = self._encode_prompt(
            image, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt
        )

        # 4. Prepare timesteps
        self.scheduler.set_timesteps(num_inference_steps, device=device)
        timesteps = self.scheduler.timesteps

        # 5. Prepare latent variables
        num_channels_latents = self.image_unet.config.in_channels
        latents = self.prepare_latents(
            batch_size * num_images_per_prompt,
            num_channels_latents,
            height,
            width,
            image_embeddings.dtype,
            device,
            generator,
            latents,
        )

        # 6. Prepare extra step kwargs.
        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)

        # 7. Denoising loop
        for i, t in enumerate(self.progress_bar(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
            noise_pred = self.image_unet(latent_model_input, t, encoder_hidden_states=image_embeddings).sample

            # 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)

            # compute the previous noisy sample x_t -> x_t-1
            latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample

            # call the callback, if provided
            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":
            image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False)[0]
        else:
            image = latents

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

        if not return_dict:
            return (image,)

        return ImagePipelineOutput(images=image)