Add InstructPix2Pix pipeline (#2040)

* being pix2pix

* ifx

* cfg image_latents

* fix some docstr

* fix

* fix

* hack

* fix

* Apply suggestions from code review

Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

* add comments to explain the hack

* move __call__ to the top

* doc

* remove height and width

* remove depreications

* fix doc str

* quality

* fast tests

* chnage model id

* fast tests

* fix test

* address Pedro's comments

* copyright

* Simple doc page.

* Apply suggestions from code review

* style

* Remove import

* address some review comments

* Apply suggestions from code review

Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* style

Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

docs/source/en/_toctree.yml

@@ -137,6 +137,8 @@
         title: Image-Variation
       - local: api/pipelines/stable_diffusion/upscale
         title: Super-Resolution
+      - local: api/pipelines/stable_diffusion/pix2pix
+        title: InstructPix2Pix
       title: Stable Diffusion
     - local: api/pipelines/stable_diffusion_2
       title: Stable Diffusion 2

docs/source/en/api/pipelines/stable_diffusion/overview.mdx

@@ -31,6 +31,7 @@ For more details about how Stable Diffusion works and how it differs from the ba
 | [StableDiffusionDepth2ImgPipeline](./depth2img) | **Experimental** – *Depth-to-Image Text-Guided Generation * | | Coming soon
 | [StableDiffusionImageVariationPipeline](./image_variation) | **Experimental** – *Image Variation Generation * | | [🤗 Stable Diffusion Image Variations](https://huggingface.co/spaces/lambdalabs/stable-diffusion-image-variations)
 | [StableDiffusionUpscalePipeline](./upscale) | **Experimental** – *Text-Guided Image Super-Resolution * | | Coming soon
+| [StableDiffusionInstructPix2PixPipeline](./pix2pix) | **Experimental** – *Text-Based Image Editing * | | [InstructPix2Pix: Learning to Follow Image Editing Instructions](https://huggingface.co/spaces/timbrooks/instruct-pix2pix)
 
 
 

docs/source/en/api/pipelines/stable_diffusion/pix2pix.mdx

@@ -0,0 +1,72 @@
+<!--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.
+-->
+
+# InstructPix2Pix: Learning to Follow Image Editing Instructions
+
+## Overview
+
+[InstructPix2Pix: Learning to Follow Image Editing Instructions](https://arxiv.org/abs/2211.09800) by Tim Brooks, Aleksander Holynski and Alexei A. Efros.
+
+The abstract of the paper is the following:
+
+*We propose a method for editing images from human instructions: given an input image and a written instruction that tells the model what to do, our model follows these instructions to edit the image. To obtain training data for this problem, we combine the knowledge of two large pretrained models -- a language model (GPT-3) and a text-to-image model (Stable Diffusion) -- to generate a large dataset of image editing examples. Our conditional diffusion model, InstructPix2Pix, is trained on our generated data, and generalizes to real images and user-written instructions at inference time. Since it performs edits in the forward pass and does not require per example fine-tuning or inversion, our model edits images quickly, in a matter of seconds. We show compelling editing results for a diverse collection of input images and written instructions.*
+
+Resources:
+
+* [Project Page](https://www.timothybrooks.com/instruct-pix2pix).
+* [Paper](https://arxiv.org/abs/2211.09800).
+* [Original Code](https://github.com/timothybrooks/instruct-pix2pix).
+* [Demo](https://huggingface.co/spaces/timbrooks/instruct-pix2pix).
+
+
+## Available Pipelines:
+
+| Pipeline | Tasks | Demo
+|---|---|:---:|
+| [StableDiffusionInstructPix2PixPipeline](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py) | *Text-Based Image Editing* | [🤗 Space](https://huggingface.co/spaces/timbrooks/instruct-pix2pix) |
+
+<!-- TODO: add Colab -->
+
+## Usage example
+
+```python
+import PIL
+import requests
+import torch
+from diffusers import StableDiffusionInstructPix2PixPipeline
+
+model_id = "timbrooks/instruct-pix2pix"
+pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+    model_id, torch_dtype=torch.float16, revision="fp16", safety_checker=None
+).to("cuda")
+
+url = "https://raw.githubusercontent.com/timothybrooks/instruct-pix2pix/main/imgs/example.jpg"
+
+
+def download_image(url):
+    image = PIL.Image.open(requests.get(url, stream=True).raw)
+    image = PIL.ImageOps.exif_transpose(image)
+    image = image.convert("RGB")
+    return image
+
+
+image = download_image(url)
+
+prompt = "turn him into a cyborg"
+images = pipe(prompt, image=image, num_inference_steps=10, guidance_scale=1.1, image_guidance_scale=1).images
+images[0].save("david_cyborg.png")
+```
+
+## StableDiffusionInstructPix2PixPipeline
+[[autodoc]] StableDiffusionInstructPix2PixPipeline
+	- __call__
+	- all

src/diffusers/__init__.py

@@ -114,6 +114,7 @@ else:
         StableDiffusionImg2ImgPipeline,
         StableDiffusionInpaintPipeline,
         StableDiffusionInpaintPipelineLegacy,
+        StableDiffusionInstructPix2PixPipeline,
         StableDiffusionPipeline,
         StableDiffusionPipelineSafe,
         StableDiffusionUpscalePipeline,

src/diffusers/pipelines/__init__.py

@@ -51,6 +51,7 @@ else:
         StableDiffusionImg2ImgPipeline,
         StableDiffusionInpaintPipeline,
         StableDiffusionInpaintPipelineLegacy,
+        StableDiffusionInstructPix2PixPipeline,
         StableDiffusionPipeline,
         StableDiffusionUpscalePipeline,
     )

src/diffusers/pipelines/stable_diffusion/__init__.py

@@ -43,6 +43,7 @@ if is_transformers_available() and is_torch_available():
     from .pipeline_stable_diffusion_img2img import StableDiffusionImg2ImgPipeline
     from .pipeline_stable_diffusion_inpaint import StableDiffusionInpaintPipeline
     from .pipeline_stable_diffusion_inpaint_legacy import StableDiffusionInpaintPipelineLegacy
+    from .pipeline_stable_diffusion_instruct_pix2pix import StableDiffusionInstructPix2PixPipeline
     from .pipeline_stable_diffusion_upscale import StableDiffusionUpscalePipeline
     from .safety_checker import StableDiffusionSafetyChecker
 

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py

@@ -0,0 +1,619 @@
+# Copyright 2023 The InstructPix2Pix Authors and 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 torch
+
+import PIL
+from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer
+
+from ...models import AutoencoderKL, UNet2DConditionModel
+from ...schedulers import KarrasDiffusionSchedulers
+from ...utils import PIL_INTERPOLATION, deprecate, is_accelerate_available, logging, randn_tensor
+from ..pipeline_utils import DiffusionPipeline
+from . import StableDiffusionPipelineOutput
+from .safety_checker import StableDiffusionSafetyChecker
+
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.preprocess
+def preprocess(image):
+    if isinstance(image, torch.Tensor):
+        return image
+    elif isinstance(image, PIL.Image.Image):
+        image = [image]
+
+    if isinstance(image[0], PIL.Image.Image):
+        w, h = image[0].size
+        w, h = map(lambda x: x - x % 8, (w, h))  # resize to integer multiple of 8
+
+        image = [np.array(i.resize((w, h), resample=PIL_INTERPOLATION["lanczos"]))[None, :] for i in image]
+        image = np.concatenate(image, axis=0)
+        image = np.array(image).astype(np.float32) / 255.0
+        image = image.transpose(0, 3, 1, 2)
+        image = 2.0 * image - 1.0
+        image = torch.from_numpy(image)
+    elif isinstance(image[0], torch.Tensor):
+        image = torch.cat(image, dim=0)
+    return image
+
+
+class StableDiffusionInstructPix2PixPipeline(DiffusionPipeline):
+    r"""
+    Pipeline for pixel-level image editing by following text instructions. Based on 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.)
+
+    Args:
+        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](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically
+            the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.
+        tokenizer (`CLIPTokenizer`):
+            Tokenizer of class
+            [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.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](https://huggingface.co/runwayml/stable-diffusion-v1-5) for details.
+        feature_extractor ([`CLIPFeatureExtractor`]):
+            Model that extracts features from generated images to be used as inputs for the `safety_checker`.
+    """
+    _optional_components = ["safety_checker", "feature_extractor"]
+
+    def __init__(
+        self,
+        vae: AutoencoderKL,
+        text_encoder: CLIPTextModel,
+        tokenizer: CLIPTokenizer,
+        unet: UNet2DConditionModel,
+        scheduler: KarrasDiffusionSchedulers,
+        safety_checker: StableDiffusionSafetyChecker,
+        feature_extractor: CLIPFeatureExtractor,
+        requires_safety_checker: bool = True,
+    ):
+        super().__init__()
+
+        if safety_checker is None and requires_safety_checker:
+            logger.warning(
+                f"You have disabled the safety checker for {self.__class__} by passing `safety_checker=None`. Ensure"
+                " that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered"
+                " results in services or applications open to the public. Both the diffusers team and Hugging Face"
+                " strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling"
+                " it only for use-cases that involve analyzing network behavior or auditing its results. For more"
+                " information, please have a look at https://github.com/huggingface/diffusers/pull/254 ."
+            )
+
+        if safety_checker is not None and feature_extractor is None:
+            raise ValueError(
+                "Make sure to define a feature extractor when loading {self.__class__} if you want to use the safety"
+                " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
+            )
+
+        self.register_modules(
+            vae=vae,
+            text_encoder=text_encoder,
+            tokenizer=tokenizer,
+            unet=unet,
+            scheduler=scheduler,
+            safety_checker=safety_checker,
+            feature_extractor=feature_extractor,
+        )
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.register_to_config(requires_safety_checker=requires_safety_checker)
+
+    @torch.no_grad()
+    def __call__(
+        self,
+        prompt: Union[str, List[str]],
+        image: Union[torch.FloatTensor, PIL.Image.Image],
+        num_inference_steps: int = 100,
+        guidance_scale: float = 7.5,
+        image_guidance_scale: float = 1.5,
+        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: Optional[int] = 1,
+    ):
+        r"""
+        Function invoked when calling the pipeline for generation.
+
+        Args:
+            prompt (`str` or `List[str]`):
+                The prompt or prompts to guide the image generation.
+            image (`PIL.Image.Image`):
+                `Image`, or tensor representing an image batch which will be repainted according to `prompt`.
+            num_inference_steps (`int`, *optional*, defaults to 100):
+                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](https://arxiv.org/abs/2207.12598).
+                `guidance_scale` is defined as `w` of equation 2. of [Imagen
+                Paper](https://arxiv.org/pdf/2205.11487.pdf). 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. This pipeline requires a value of at least `1`.
+            image_guidance_scale (`float`, *optional*, defaults to 1.5):
+                Image guidance scale is to push the generated image towards the inital image `image`. Image guidance
+                scale is enabled by setting `image_guidance_scale > 1`. Higher image guidance scale encourages to
+                generate images that are closely linked to the source image `image`, usually at the expense of lower
+                image quality. This pipeline requires a value of at least `1`.
+            negative_prompt (`str` or `List[str]`, *optional*):
+                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`).
+            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`, *optional*):
+                One or a list of [torch generator(s)](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 will ge generated by sampling using the supplied random `generator`.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generate image. Choose between
+                [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.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 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.
+
+        Examples:
+
+        ```py
+        >>> import PIL
+        >>> import requests
+        >>> import torch
+        >>> from io import BytesIO
+
+        >>> from diffusers import StableDiffusionInstructPix2PixPipeline
+
+
+        >>> def download_image(url):
+        ...     response = requests.get(url)
+        ...     return PIL.Image.open(BytesIO(response.content)).convert("RGB")
+
+
+        >>> img_url = "https://raw.githubusercontent.com/timothybrooks/instruct-pix2pix/main/imgs/example.jpg"
+
+        >>> image = download_image(img_url).resize((512, 512))
+
+        >>> pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+        ...     "timbrooks/instruct-pix2pix", torch_dtype=torch.float16
+        ... )
+        >>> pipe = pipe.to("cuda")
+
+        >>> prompt = "turn him into cyborg"
+        >>> image = pipe(prompt=prompt, image=image).images[0]
+        ```
+
+        Returns:
+            [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
+            [`~pipelines.stable_diffusion.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 `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
+            (nsfw) content, according to the `safety_checker`.
+        """
+        # 0. Check inputs
+        self.check_inputs(prompt, callback_steps)
+
+        # 1. Define call parameters
+        batch_size = 1 if isinstance(prompt, str) else len(prompt)
+        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 and image_guidance_scale >= 1.0
+        # check if scheduler is in sigmas space
+        scheduler_is_in_sigma_space = hasattr(self.scheduler, "sigmas")
+
+        # 2. Encode input prompt
+        text_embeddings = self._encode_prompt(
+            prompt, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt
+        )
+
+        # 3. Preprocess image
+        image = preprocess(image)
+        height, width = image.shape[-2:]
+
+        # 4. set timesteps
+        self.scheduler.set_timesteps(num_inference_steps, device=device)
+        timesteps = self.scheduler.timesteps
+
+        # 5. Prepare Image latents
+        image_latents = self.prepare_image_latents(
+            image,
+            batch_size,
+            num_images_per_prompt,
+            text_embeddings.dtype,
+            device,
+            do_classifier_free_guidance,
+            generator,
+        )
+
+        # 6. Prepare latent variables
+        num_channels_latents = self.vae.config.latent_channels
+        latents = self.prepare_latents(
+            batch_size * num_images_per_prompt,
+            num_channels_latents,
+            height,
+            width,
+            text_embeddings.dtype,
+            device,
+            generator,
+            latents,
+        )
+
+        # 7. Check that shapes of latents and image match the UNet channels
+        num_channels_image = image_latents.shape[1]
+        if num_channels_latents + num_channels_image != self.unet.config.in_channels:
+            raise ValueError(
+                f"Incorrect configuration settings! The config of `pipeline.unet`: {self.unet.config} expects"
+                f" {self.unet.config.in_channels} but received `num_channels_latents`: {num_channels_latents} +"
+                f" `num_channels_image`: {num_channels_image} "
+                f" = {num_channels_latents+num_channels_image}. Please verify the config of"
+                " `pipeline.unet` or your `image` input."
+            )
+
+        # 8. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
+        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
+
+        # 9. Denoising loop
+        num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
+        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.
+                # The latents are expanded 3 times because for pix2pix the guidance\
+                # is applied for both the text and the input image.
+                latent_model_input = torch.cat([latents] * 3) if do_classifier_free_guidance else latents
+
+                # concat latents, image_latents in the channel dimension
+                scaled_latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
+                scaled_latent_model_input = torch.cat([scaled_latent_model_input, image_latents], dim=1)
+
+                # predict the noise residual
+                noise_pred = self.unet(scaled_latent_model_input, t, encoder_hidden_states=text_embeddings).sample
+
+                # Hack:
+                # For karras style schedulers the model does classifer free guidance using the
+                # predicted_original_sample instead of the noise_pred. So we need to compute the
+                # predicted_original_sample here if we are using a karras style scheduler.
+                if scheduler_is_in_sigma_space:
+                    step_index = (self.scheduler.timesteps == t).nonzero().item()
+                    sigma = self.scheduler.sigmas[step_index]
+                    noise_pred = latent_model_input - sigma * noise_pred
+
+                # perform guidance
+                if do_classifier_free_guidance:
+                    noise_pred_text, noise_pred_image, noise_pred_uncond = noise_pred.chunk(3)
+                    noise_pred = (
+                        noise_pred_uncond
+                        + guidance_scale * (noise_pred_text - noise_pred_image)
+                        + image_guidance_scale * (noise_pred_image - noise_pred_uncond)
+                    )
+
+                # Hack:
+                # For karras style schedulers the model does classifer free guidance using the
+                # predicted_original_sample instead of the noise_pred. But the scheduler.step function
+                # expects the noise_pred and computes the predicted_original_sample internally. So we
+                # need to overwrite the noise_pred here such that the value of the computed
+                # predicted_original_sample is correct.
+                if scheduler_is_in_sigma_space:
+                    noise_pred = (noise_pred - latents) / (-sigma)
+
+                # 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 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:
+                        callback(i, t, latents)
+
+        # 10. Post-processing
+        image = self.decode_latents(latents)
+
+        # 11. Run safety checker
+        image, has_nsfw_concept = self.run_safety_checker(image, device, text_embeddings.dtype)
+
+        # 12. Convert to PIL
+        if output_type == "pil":
+            image = self.numpy_to_pil(image)
+
+        if not return_dict:
+            return (image, has_nsfw_concept)
+
+        return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_sequential_cpu_offload
+    def enable_sequential_cpu_offload(self, gpu_id=0):
+        r"""
+        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 `forward` method called.
+        """
+        if is_accelerate_available():
+            from accelerate import cpu_offload
+        else:
+            raise ImportError("Please install accelerate via `pip install accelerate`")
+
+        device = torch.device(f"cuda:{gpu_id}")
+
+        for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
+            cpu_offload(cpu_offloaded_model, device)
+
+        if self.safety_checker is not None:
+            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
+
+    @property
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device
+    def _execution_device(self):
+        r"""
+        Returns the device on which the pipeline's models will be executed. After calling
+        `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module
+        hooks.
+        """
+        if self.device != torch.device("meta") or not hasattr(self.unet, "_hf_hook"):
+            return self.device
+        for module in self.unet.modules():
+            if (
+                hasattr(module, "_hf_hook")
+                and hasattr(module._hf_hook, "execution_device")
+                and module._hf_hook.execution_device is not None
+            ):
+                return torch.device(module._hf_hook.execution_device)
+        return self.device
+
+    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(int)`):
+                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`).
+        """
+        batch_size = len(prompt) if isinstance(prompt, list) else 1
+
+        text_inputs = self.tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=self.tokenizer.model_max_length,
+            truncation=True,
+            return_tensors="pt",
+        )
+        text_input_ids = text_inputs.input_ids
+        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+
+        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
+            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1])
+            logger.warning(
+                "The following part of your input was truncated because CLIP can only handle sequences up to"
+                f" {self.tokenizer.model_max_length} tokens: {removed_text}"
+            )
+
+        if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+            attention_mask = text_inputs.attention_mask.to(device)
+        else:
+            attention_mask = None
+
+        text_embeddings = self.text_encoder(
+            text_input_ids.to(device),
+            attention_mask=attention_mask,
+        )
+        text_embeddings = text_embeddings[0]
+
+        # duplicate text embeddings for each generation per prompt, using mps friendly method
+        bs_embed, seq_len, _ = text_embeddings.shape
+        text_embeddings = text_embeddings.repeat(1, num_images_per_prompt, 1)
+        text_embeddings = text_embeddings.view(bs_embed * num_images_per_prompt, seq_len, -1)
+
+        # get unconditional embeddings for classifier free guidance
+        if do_classifier_free_guidance:
+            uncond_tokens: List[str]
+            if negative_prompt is None:
+                uncond_tokens = [""] * 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, str):
+                uncond_tokens = [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_tokens = negative_prompt
+
+            max_length = text_input_ids.shape[-1]
+            uncond_input = self.tokenizer(
+                uncond_tokens,
+                padding="max_length",
+                max_length=max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = uncond_input.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            uncond_embeddings = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            uncond_embeddings = uncond_embeddings[0]
+
+            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
+            seq_len = uncond_embeddings.shape[1]
+            uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt, 1)
+            uncond_embeddings = uncond_embeddings.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 text embeddings into a single batch
+            # to avoid doing two forward passes
+            text_embeddings = torch.cat([text_embeddings, uncond_embeddings, uncond_embeddings])
+
+        return text_embeddings
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    def run_safety_checker(self, image, device, dtype):
+        if self.safety_checker is not None:
+            safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(device)
+            image, has_nsfw_concept = self.safety_checker(
+                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
+            )
+        else:
+            has_nsfw_concept = None
+        return image, has_nsfw_concept
+
+    # 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.StableDiffusionPipeline.decode_latents
+    def decode_latents(self, latents):
+        latents = 1 / 0.18215 * latents
+        image = self.vae.decode(latents).sample
+        image = (image / 2 + 0.5).clamp(0, 1)
+        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloa16
+        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
+        return image
+
+    def check_inputs(self, prompt, callback_steps):
+        if not isinstance(prompt, str) and not isinstance(prompt, list):
+            raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")
+
+        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
+
+    def prepare_image_latents(
+        self, image, batch_size, num_images_per_prompt, dtype, device, do_classifier_free_guidance, generator=None
+    ):
+        if not isinstance(image, (torch.Tensor, PIL.Image.Image, list)):
+            raise ValueError(
+                f"`image` has to be of type `torch.Tensor`, `PIL.Image.Image` or list but is {type(image)}"
+            )
+
+        image = image.to(device=device, dtype=dtype)
+
+        batch_size = batch_size * num_images_per_prompt
+        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 isinstance(generator, list):
+            image_latents = [self.vae.encode(image[i : i + 1]).latent_dist.mode() for i in range(batch_size)]
+            image_latents = torch.cat(image_latents, dim=0)
+        else:
+            image_latents = self.vae.encode(image).latent_dist.mode()
+
+        if batch_size > image_latents.shape[0] and batch_size % image_latents.shape[0] == 0:
+            # expand image_latents for batch_size
+            deprecation_message = (
+                f"You have passed {batch_size} text prompts (`prompt`), but only {image_latents.shape[0]} initial"
+                " images (`image`). Initial images are now duplicating to match the number of text prompts. Note"
+                " that this behavior is deprecated and will be removed in a version 1.0.0. Please make sure to update"
+                " your script to pass as many initial images as text prompts to suppress this warning."
+            )
+            deprecate("len(prompt) != len(image)", "1.0.0", deprecation_message, standard_warn=False)
+            additional_image_per_prompt = batch_size // image_latents.shape[0]
+            image_latents = torch.cat([image_latents] * additional_image_per_prompt, dim=0)
+        elif batch_size > image_latents.shape[0] and batch_size % image_latents.shape[0] != 0:
+            raise ValueError(
+                f"Cannot duplicate `image` of batch size {image_latents.shape[0]} to {batch_size} text prompts."
+            )
+        else:
+            image_latents = torch.cat([image_latents], dim=0)
+
+        if do_classifier_free_guidance:
+            uncond_image_latents = torch.zeros_like(image_latents)
+            image_latents = torch.cat([image_latents, image_latents, uncond_image_latents], dim=0)
+
+        return image_latents

src/diffusers/utils/dummy_torch_and_transformers_objects.py

@@ -154,6 +154,21 @@ class StableDiffusionInpaintPipelineLegacy(metaclass=DummyObject):
         requires_backends(cls, ["torch", "transformers"])
 
 
+class StableDiffusionInstructPix2PixPipeline(metaclass=DummyObject):
+    _backends = ["torch", "transformers"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch", "transformers"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+
 class StableDiffusionPipeline(metaclass=DummyObject):
     _backends = ["torch", "transformers"]
 

tests/pipelines/stable_diffusion/test_stable_diffusion_instruction_pix2pix.py

@@ -0,0 +1,373 @@
+# coding=utf-8
+# Copyright 2023 HuggingFace Inc.
+#
+# 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 gc
+import random
+import unittest
+
+import numpy as np
+import torch
+
+from diffusers import (
+    AutoencoderKL,
+    DDIMScheduler,
+    EulerAncestralDiscreteScheduler,
+    LMSDiscreteScheduler,
+    PNDMScheduler,
+    StableDiffusionInstructPix2PixPipeline,
+    UNet2DConditionModel,
+)
+from diffusers.utils import floats_tensor, load_image, slow, torch_device
+from diffusers.utils.testing_utils import require_torch_gpu
+from PIL import Image
+from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer
+
+from ...test_pipelines_common import PipelineTesterMixin
+
+
+torch.backends.cuda.matmul.allow_tf32 = False
+
+
+class StableDiffusionInstructPix2PixPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+    pipeline_class = StableDiffusionInstructPix2PixPipeline
+
+    def get_dummy_components(self):
+        torch.manual_seed(0)
+        unet = UNet2DConditionModel(
+            block_out_channels=(32, 64),
+            layers_per_block=2,
+            sample_size=32,
+            in_channels=8,
+            out_channels=4,
+            down_block_types=("DownBlock2D", "CrossAttnDownBlock2D"),
+            up_block_types=("CrossAttnUpBlock2D", "UpBlock2D"),
+            cross_attention_dim=32,
+        )
+        scheduler = PNDMScheduler(skip_prk_steps=True)
+        torch.manual_seed(0)
+        vae = AutoencoderKL(
+            block_out_channels=[32, 64],
+            in_channels=3,
+            out_channels=3,
+            down_block_types=["DownEncoderBlock2D", "DownEncoderBlock2D"],
+            up_block_types=["UpDecoderBlock2D", "UpDecoderBlock2D"],
+            latent_channels=4,
+        )
+        torch.manual_seed(0)
+        text_encoder_config = CLIPTextConfig(
+            bos_token_id=0,
+            eos_token_id=2,
+            hidden_size=32,
+            intermediate_size=37,
+            layer_norm_eps=1e-05,
+            num_attention_heads=4,
+            num_hidden_layers=5,
+            pad_token_id=1,
+            vocab_size=1000,
+        )
+        text_encoder = CLIPTextModel(text_encoder_config)
+        tokenizer = CLIPTokenizer.from_pretrained("hf-internal-testing/tiny-random-clip")
+
+        components = {
+            "unet": unet,
+            "scheduler": scheduler,
+            "vae": vae,
+            "text_encoder": text_encoder,
+            "tokenizer": tokenizer,
+            "safety_checker": None,
+            "feature_extractor": None,
+        }
+        return components
+
+    def get_dummy_inputs(self, device, seed=0):
+        image = floats_tensor((1, 3, 32, 32), rng=random.Random(seed)).to(device)
+        image = image.cpu().permute(0, 2, 3, 1)[0]
+        image = Image.fromarray(np.uint8(image)).convert("RGB")
+        if str(device).startswith("mps"):
+            generator = torch.manual_seed(seed)
+        else:
+            generator = torch.Generator(device=device).manual_seed(seed)
+        inputs = {
+            "prompt": "A painting of a squirrel eating a burger",
+            "image": image,
+            "generator": generator,
+            "num_inference_steps": 2,
+            "guidance_scale": 6.0,
+            "image_guidance_scale": 1,
+            "output_type": "numpy",
+        }
+        return inputs
+
+    def test_stable_diffusion_pix2pix_default_case(self):
+        device = "cpu"  # ensure determinism for the device-dependent torch.Generator
+        components = self.get_dummy_components()
+        sd_pipe = StableDiffusionInstructPix2PixPipeline(**components)
+        sd_pipe = sd_pipe.to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs(device)
+        image = sd_pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1]
+        assert image.shape == (1, 32, 32, 3)
+        expected_slice = np.array([0.7318, 0.3723, 0.4662, 0.623, 0.5770, 0.5014, 0.4281, 0.5550, 0.4813])
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-3
+
+    def test_stable_diffusion_pix2pix_negative_prompt(self):
+        device = "cpu"  # ensure determinism for the device-dependent torch.Generator
+        components = self.get_dummy_components()
+        sd_pipe = StableDiffusionInstructPix2PixPipeline(**components)
+        sd_pipe = sd_pipe.to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs(device)
+        negative_prompt = "french fries"
+        output = sd_pipe(**inputs, negative_prompt=negative_prompt)
+        image = output.images
+        image_slice = image[0, -3:, -3:, -1]
+
+        assert image.shape == (1, 32, 32, 3)
+        expected_slice = np.array([0.7323, 0.3688, 0.4611, 0.6255, 0.5746, 0.5017, 0.433, 0.5553, 0.4827])
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-3
+
+    def test_stable_diffusion_pix2pix_multiple_init_images(self):
+        device = "cpu"  # ensure determinism for the device-dependent torch.Generator
+        components = self.get_dummy_components()
+        sd_pipe = StableDiffusionInstructPix2PixPipeline(**components)
+        sd_pipe = sd_pipe.to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs(device)
+        inputs["prompt"] = [inputs["prompt"]] * 2
+
+        image = np.array(inputs["image"]).astype(np.float32) / 255.0
+        image = torch.from_numpy(image).unsqueeze(0).to(device)
+        image = image.permute(0, 3, 1, 2)
+        inputs["image"] = image.repeat(2, 1, 1, 1)
+
+        image = sd_pipe(**inputs).images
+        image_slice = image[-1, -3:, -3:, -1]
+
+        assert image.shape == (2, 32, 32, 3)
+        expected_slice = np.array([0.606, 0.5712, 0.5099, 0.598, 0.5805, 0.7205, 0.6793, 0.554, 0.5607])
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-3
+
+    def test_stable_diffusion_pix2pix_euler(self):
+        device = "cpu"  # ensure determinism for the device-dependent torch.Generator
+        components = self.get_dummy_components()
+        components["scheduler"] = EulerAncestralDiscreteScheduler(
+            beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear"
+        )
+        sd_pipe = StableDiffusionInstructPix2PixPipeline(**components)
+        sd_pipe = sd_pipe.to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs(device)
+        image = sd_pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1]
+
+        slice = [round(x, 4) for x in image_slice.flatten().tolist()]
+        print(",".join([str(x) for x in slice]))
+
+        assert image.shape == (1, 32, 32, 3)
+        expected_slice = np.array([0.726, 0.3902, 0.4868, 0.585, 0.5672, 0.511, 0.3906, 0.551, 0.4846])
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-3
+
+    def test_stable_diffusion_pix2pix_num_images_per_prompt(self):
+        device = "cpu"  # ensure determinism for the device-dependent torch.Generator
+        components = self.get_dummy_components()
+        sd_pipe = StableDiffusionInstructPix2PixPipeline(**components)
+        sd_pipe = sd_pipe.to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        # test num_images_per_prompt=1 (default)
+        inputs = self.get_dummy_inputs(device)
+        images = sd_pipe(**inputs).images
+
+        assert images.shape == (1, 32, 32, 3)
+
+        # test num_images_per_prompt=1 (default) for batch of prompts
+        batch_size = 2
+        inputs = self.get_dummy_inputs(device)
+        inputs["prompt"] = [inputs["prompt"]] * batch_size
+        images = sd_pipe(**inputs).images
+
+        assert images.shape == (batch_size, 32, 32, 3)
+
+        # test num_images_per_prompt for single prompt
+        num_images_per_prompt = 2
+        inputs = self.get_dummy_inputs(device)
+        images = sd_pipe(**inputs, num_images_per_prompt=num_images_per_prompt).images
+
+        assert images.shape == (num_images_per_prompt, 32, 32, 3)
+
+        # test num_images_per_prompt for batch of prompts
+        batch_size = 2
+        inputs = self.get_dummy_inputs(device)
+        inputs["prompt"] = [inputs["prompt"]] * batch_size
+        images = sd_pipe(**inputs, num_images_per_prompt=num_images_per_prompt).images
+
+        assert images.shape == (batch_size * num_images_per_prompt, 32, 32, 3)
+
+
+@slow
+@require_torch_gpu
+class StableDiffusionInstructPix2PixPipelineSlowTests(unittest.TestCase):
+    def tearDown(self):
+        super().tearDown()
+        gc.collect()
+        torch.cuda.empty_cache()
+
+    def get_inputs(self, device, dtype=torch.float32, seed=0):
+        generator = torch.Generator(device=device).manual_seed(seed)
+        image = load_image(
+            "https://huggingface.co/datasets/diffusers/test-arrays/resolve/main/stable_diffusion_pix2pix/example.jpg"
+        )
+        inputs = {
+            "prompt": "turn him into a cyborg",
+            "image": image,
+            "generator": generator,
+            "num_inference_steps": 3,
+            "guidance_scale": 7.5,
+            "image_guidance_scale": 1.0,
+            "output_type": "numpy",
+        }
+        return inputs
+
+    def test_stable_diffusion_pix2pix_default(self):
+        pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+            "timbrooks/instruct-pix2pix", safety_checker=None
+        )
+        pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+        pipe.enable_attention_slicing()
+
+        inputs = self.get_inputs(torch_device)
+        image = pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1].flatten()
+
+        assert image.shape == (1, 512, 512, 3)
+        expected_slice = np.array([0.3214, 0.3252, 0.3313, 0.3261, 0.3332, 0.3351, 0.324, 0.3296, 0.3206])
+        assert np.abs(expected_slice - image_slice).max() < 1e-3
+
+    def test_stable_diffusion_pix2pix_k_lms(self):
+        pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+            "timbrooks/instruct-pix2pix", safety_checker=None
+        )
+        pipe.scheduler = LMSDiscreteScheduler.from_config(pipe.scheduler.config)
+        pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+        pipe.enable_attention_slicing()
+
+        inputs = self.get_inputs(torch_device)
+        image = pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1].flatten()
+
+        assert image.shape == (1, 512, 512, 3)
+        expected_slice = np.array([0.3893, 0.393, 0.3997, 0.4196, 0.4239, 0.4307, 0.4268, 0.4317, 0.419])
+        assert np.abs(expected_slice - image_slice).max() < 1e-3
+
+    def test_stable_diffusion_pix2pix_ddim(self):
+        pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+            "timbrooks/instruct-pix2pix", safety_checker=None
+        )
+        pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config)
+        pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+        pipe.enable_attention_slicing()
+
+        inputs = self.get_inputs(torch_device)
+        image = pipe(**inputs).images
+        image_slice = image[0, -3:, -3:, -1].flatten()
+
+        assert image.shape == (1, 512, 512, 3)
+        expected_slice = np.array([0.5151, 0.5186, 0.5133, 0.5176, 0.5147, 0.5198, 0.522, 0.5122, 0.5244])
+        assert np.abs(expected_slice - image_slice).max() < 1e-3
+
+    def test_stable_diffusion_pix2pix_intermediate_state(self):
+        number_of_steps = 0
+
+        def callback_fn(step: int, timestep: int, latents: torch.FloatTensor) -> None:
+            callback_fn.has_been_called = True
+            nonlocal number_of_steps
+            number_of_steps += 1
+            if step == 1:
+                latents = latents.detach().cpu().numpy()
+                assert latents.shape == (1, 4, 64, 64)
+                latents_slice = latents[0, -3:, -3:, -1]
+                expected_slice = np.array([-0.7178, -0.9165, -1.3906, 1.8174, 1.9482, 1.3652, 1.1533, 1.542, 1.2461])
+                assert np.abs(latents_slice.flatten() - expected_slice).max() < 1e-3
+            elif step == 2:
+                latents = latents.detach().cpu().numpy()
+                assert latents.shape == (1, 4, 64, 64)
+                latents_slice = latents[0, -3:, -3:, -1]
+                expected_slice = np.array([-0.7183, -0.9253, -1.3857, 1.8174, 1.9766, 1.3574, 1.1533, 1.5244, 1.2539])
+                assert np.abs(latents_slice.flatten() - expected_slice).max() < 1e-3
+
+        callback_fn.has_been_called = False
+
+        pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+            "timbrooks/instruct-pix2pix", safety_checker=None, torch_dtype=torch.float16
+        )
+        pipe = pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+        pipe.enable_attention_slicing()
+
+        inputs = self.get_inputs(torch_device, dtype=torch.float16)
+        pipe(**inputs, callback=callback_fn, callback_steps=1)
+        assert callback_fn.has_been_called
+        assert number_of_steps == 3
+
+    def test_stable_diffusion_pipeline_with_sequential_cpu_offloading(self):
+        torch.cuda.empty_cache()
+        torch.cuda.reset_max_memory_allocated()
+        torch.cuda.reset_peak_memory_stats()
+
+        pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+            "timbrooks/instruct-pix2pix", safety_checker=None, torch_dtype=torch.float16
+        )
+        pipe = pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+        pipe.enable_attention_slicing(1)
+        pipe.enable_sequential_cpu_offload()
+
+        inputs = self.get_inputs(torch_device, dtype=torch.float16)
+        _ = pipe(**inputs)
+
+        mem_bytes = torch.cuda.max_memory_allocated()
+        # make sure that less than 2.2 GB is allocated
+        assert mem_bytes < 2.2 * 10**9
+
+    def test_stable_diffusion_pix2pix_pipeline_multiple_of_8(self):
+        inputs = self.get_inputs(torch_device)
+        # resize to resolution that is divisible by 8 but not 16 or 32
+        inputs["image"] = inputs["image"].resize((504, 504))
+
+        model_id = "timbrooks/instruct-pix2pix"
+        pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+            model_id,
+            safety_checker=None,
+        )
+        pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+        pipe.enable_attention_slicing()
+
+        output = pipe(**inputs)
+        image = output.images[0]
+
+        image_slice = image[255:258, 383:386, -1]
+
+        assert image.shape == (504, 504, 3)
+        expected_slice = np.array([0.1834, 0.2046, 0.2429, 0.1825, 0.2201, 0.2576, 0.1968, 0.2185, 0.2487])
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-3