Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+assets/fig.jpg filter=lfs diff=lfs merge=lfs -text

Code_of_Conduct.md

@@ -0,0 +1,52 @@
+# Yahoo Inc Open Source Code of Conduct
+
+## Summary
+This Code of Conduct is our way to encourage good behavior and discourage bad behavior in our open source projects. We invite participation from many people to bring different perspectives to our projects. We will do our part to foster a welcoming and professional environment free of harassment. We expect participants to communicate professionally and thoughtfully during their involvement with this project. 
+
+Participants may lose their good standing by engaging in misconduct. For example: insulting, threatening, or conveying unwelcome sexual content. We ask participants who observe conduct issues to report the incident directly to the project's Response Team at [email protected]. Yahoo will assign a respondent to address the issue. We may remove harassers from this project. 
+
+This code does not replace the terms of service or acceptable use policies of the websites used to support this project. We acknowledge that participants may be subject to additional conduct terms based on their employment which may govern their online expressions.
+
+## Details
+This Code of Conduct makes our expectations of participants in this community explicit.
+* We forbid harassment and abusive speech within this community.
+* We request participants to report misconduct to the project’s Response Team.
+* We urge participants to refrain from using discussion forums to play out a fight.
+
+### Expected Behaviors
+We expect participants in this community to conduct themselves professionally. Since our primary mode of communication is text on an online forum (e.g. issues, pull requests, comments, emails, or chats) devoid of vocal tone, gestures, or other context that is often vital to understanding, it is important that participants are attentive to their interaction style.
+
+* **Assume positive intent.** We ask community members to assume positive intent on the part of other people’s communications. We may disagree on details, but we expect all suggestions to be supportive of the community goals.
+* **Respect participants.** We expect occasional disagreements. Open Source projects are learning experiences. Ask, explore, challenge, and then _respectfully_ state if you agree or disagree. If your idea is rejected, be more persuasive not bitter.
+* **Welcoming to new members.** New members bring new perspectives. Some ask questions that have been addressed before. _Kindly_ point to existing discussions. Everyone is new to every project once.
+* **Be kind to beginners.** Beginners use open source projects to get experience. They might not be talented coders yet, and projects should not accept poor quality code. But we were all beginners once, and we need to engage kindly.
+* **Consider your impact on others.** Your work will be used by others, and you depend on the work of others. We expect community members to be considerate and establish a balance their self-interest with communal interest.
+* **Use words carefully.** We may not understand intent when you say something ironic. Often, people will misinterpret sarcasm in online communications. We ask community members to communicate plainly.
+* **Leave with class.** When you wish to resign from participating in this project for any reason, you are free to fork the code and create a competitive project. Open Source explicitly allows this. Your exit should not be dramatic or bitter. 
+
+### Unacceptable Behaviors
+Participants remain in good standing when they do not engage in misconduct or harassment (some examples follow). We do not list all forms of harassment, nor imply some forms of harassment are not worthy of action. Any participant who *feels* harassed or *observes* harassment, should report the incident to the Response Team. 
+* **Don't be a bigot.** Calling out project members by their identity or background in a negative or insulting manner. This includes, but is not limited to, slurs or insinuations related to protected or suspect classes e.g. race, color, citizenship, national origin, political belief, religion, sexual orientation, gender identity and expression, age, size, culture, ethnicity, genetic features, language, profession, national minority status, mental or physical ability.
+* **Don't insult.** Insulting remarks about a person’s lifestyle practices.
+* **Don't dox.** Revealing private information about other participants without explicit permission.
+* **Don't intimidate.** Threats of violence or intimidation of any project member.
+* **Don't creep.** Unwanted sexual attention or content unsuited for the subject of this project.
+* **Don't inflame.** We ask that victim of harassment not address their grievances in the public forum, as this often intensifies the problem. Report it, and let us address it off-line.
+* **Don't disrupt.** Sustained disruptions in a discussion.
+
+### Reporting Issues
+If you experience or witness misconduct, or have any other concerns about the conduct of members of this project, please report it by contacting our Response Team at [email protected] who will handle your report with discretion. Your report should include:
+* Your preferred contact information. We cannot process anonymous reports.
+* Names (real or usernames) of those involved in the incident.
+* Your account of what occurred, and if the incident is ongoing. Please provide links to or transcripts of the publicly available records (e.g. a mailing list archive or a public IRC logger), so that we can review it.
+* Any additional information that may be helpful to achieve resolution.
+
+After filing a report, a representative will contact you directly to review the incident and ask additional questions. If a member of the Yahoo Response Team is named in an incident report, that member will be recused from handling your incident. If the complaint originates from a member of the Response Team, it will be addressed by a different member of the Response Team. We will consider reports to be confidential for the purpose of protecting victims of abuse. 
+
+### Scope
+Yahoo will assign a Response Team member with admin rights on the project and legal rights on the project copyright. The Response Team is empowered to restrict some privileges to the project as needed. Since this project is governed by an open source license, any participant may fork the code under the terms of the project license. The Response Team’s goal is to preserve the project if possible, and will restrict or remove participation from those who disrupt the project. 
+
+This code does not replace the terms of service or acceptable use policies that are provided by the websites used to support this community. Nor does this code apply to communications or actions that take place outside of the context of this community. Many participants in this project are also subject to codes of conduct based on their employment. This code is a social-contract that informs participants of our social expectations. It is not a terms of service or legal contract.
+
+## License and Acknowledgment. 
+This text is shared under the [CC-BY-4.0 license](https://creativecommons.org/licenses/by/4.0/). This code is based on a study conducted by the [TODO Group](https://todogroup.org/) of many codes used in the open source community. If you have feedback about this code, contact our Response Team at the address listed above. 

LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   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.

PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,3 @@
+
+<!-- The following line must be included in your pull request -->
+I confirm that this contribution is made under the terms of the license found in the root directory of this repository's source tree and that I have the authority necessary to make this contribution on behalf of its copyright owner.

README.md

@@ -1,3 +1,65 @@
----
-license: apache-2.0
----
+# Salient Object Aware Background Generation [![Paper](assets/arxiv.svg)](https://arxiv.org/pdf/2404.10157.pdf)
+This repository accompanies our paper, [Salient Object-Aware Background Generation using Text-Guided Diffusion Models](https://arxiv.org/abs/2404.10157), which has been accepted for publication in [CVPR 2024 Generative Models for Computer Vision](https://generative-vision.github.io/workshop-CVPR-24/) workshop.
+
+The paper addresses an issue we call "object expansion" when generating backgrounds for salient objects using inpainting diffusion models.  We show that models such as [Stable Inpainting](https://huggingface.co/stabilityai/stable-diffusion-2-inpainting) can sometimes arbitrarily expand or distort the salient object, which is undesirable in applications where the object's identity should be preserved, such as e-commerce ads. We provide some examples of object expansion as follows:
+
+<div align="center">
+  <img src="assets/fig.jpg">
+</div>
+
+
+
+## Setup
+
+The dependencies are provided in `requirements.txt`, install them by:
+
+```bash
+pip install -r requirements.txt
+```
+
+## Usage
+### Training
+
+The following runs the training of text-to-image inpainting ControlNet initialized with the weights of "stable-diffusion-2-inpainting":
+```bash
+accelerate launch --multi_gpu --mixed_precision=fp16 --num_processes=8 train_controlnet_inpaint.py --pretrained_model_name_or_path "stable-diffusion-2-inpainting" --proportion_empty_prompts 0.1
+```
+
+The following runs the training of text-to-image ControlNet initialized with the weights of "stable-diffusion-2-base":
+```bash
+accelerate launch --multi_gpu --mixed_precision=fp16 --num_processes=8 train_controlnet.py --pretrained_model_name_or_path "stable-diffusion-2-base" --proportion_empty_prompts 0.1
+```
+
+### Inference
+
+Please refer to `inference.ipynb`. Tu run the code you need to download our model checkpoints. 
+
+## Models Checkpoints
+
+| Model link                                       | Datasets used                                                                                                                                                                                      |
+|--------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| [controlnet_inpainting_salient_aware.pth](https://drive.google.com/file/d/1ad4CNJqFI_HnXFFRqcS4mOD0Le2Mvd3L/view?usp=sharing)     | Salient segmentation datasets, COCO |
+
+## Citations
+
+If you found our work useful, please consider citing our paper:
+
+```bibtex
+@misc{eshratifar2024salient,
+      title={Salient Object-Aware Background Generation using Text-Guided Diffusion Models}, 
+      author={Amir Erfan Eshratifar and Joao V. B. Soares and Kapil Thadani and Shaunak Mishra and Mikhail Kuznetsov and Yueh-Ning Ku and Paloma de Juan},
+      year={2024},
+      eprint={2404.10157},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV}
+}
+```
+
+## Maintainers
+
+- Erfan Eshratifar: [email protected]
+- Joao Soares: [email protected]
+
+## License
+
+This project is licensed under the terms of the [Apache 2.0](LICENSE) open source license. Please refer to [LICENSE](LICENSE) for the full terms.

pipeline_controlnet_inpaint.py

@@ -0,0 +1,1352 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2024, Yahoo Research
+#
+# 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.
+
+# This model implementation is heavily inspired by https://github.com/haofanwang/ControlNet-for-Diffusers/
+
+import inspect
+import warnings
+from typing import Any, Callable, Dict, List, Optional, Tuple, Union
+
+import numpy as np
+import PIL.Image
+import torch
+import torch.nn.functional as F
+from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
+
+from diffusers.image_processor import VaeImageProcessor
+from diffusers.loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
+from diffusers.models import AutoencoderKL, ControlNetModel, UNet2DConditionModel
+from diffusers.schedulers import KarrasDiffusionSchedulers
+from diffusers.utils import (
+    is_accelerate_available,
+    is_accelerate_version,
+    logging,
+    replace_example_docstring,
+)
+from diffusers.utils.torch_utils import is_compiled_module, randn_tensor
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
+from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
+from diffusers.pipelines.controlnet.multicontrolnet import MultiControlNetModel
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+EXAMPLE_DOC_STRING = """
+    Examples:
+        ```py
+        >>> # !pip install transformers accelerate
+        >>> from diffusers import StableDiffusionControlNetInpaintPipeline, ControlNetModel, DDIMScheduler
+        >>> from diffusers.utils import load_image
+        >>> import numpy as np
+        >>> import torch
+
+        >>> init_image = load_image(
+        ...     "https://huggingface.co/datasets/diffusers/test-arrays/resolve/main/stable_diffusion_inpaint/boy.png"
+        ... )
+        >>> init_image = init_image.resize((512, 512))
+
+        >>> generator = torch.Generator(device="cpu").manual_seed(1)
+
+        >>> mask_image = load_image(
+        ...     "https://huggingface.co/datasets/diffusers/test-arrays/resolve/main/stable_diffusion_inpaint/boy_mask.png"
+        ... )
+        >>> mask_image = mask_image.resize((512, 512))
+
+
+        >>> def make_inpaint_condition(image, image_mask):
+        ...     image = np.array(image.convert("RGB")).astype(np.float32) / 255.0
+        ...     image_mask = np.array(image_mask.convert("L")).astype(np.float32) / 255.0
+
+        ...     assert image.shape[0:1] == image_mask.shape[0:1], "image and image_mask must have the same image size"
+        ...     image[image_mask > 0.5] = -1.0  # set as masked pixel
+        ...     image = np.expand_dims(image, 0).transpose(0, 3, 1, 2)
+        ...     image = torch.from_numpy(image)
+        ...     return image
+
+
+        >>> control_image = make_inpaint_condition(init_image, mask_image)
+
+        >>> controlnet = ControlNetModel.from_pretrained(
+        ...     "lllyasviel/control_v11p_sd15_inpaint", torch_dtype=torch.float16
+        ... )
+        >>> pipe = StableDiffusionControlNetInpaintPipeline.from_pretrained(
+        ...     "runwayml/stable-diffusion-v1-5", controlnet=controlnet, torch_dtype=torch.float16
+        ... )
+
+        >>> pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config)
+        >>> pipe.enable_model_cpu_offload()
+
+        >>> # generate image
+        >>> image = pipe(
+        ...     "a handsome man with ray-ban sunglasses",
+        ...     num_inference_steps=20,
+        ...     generator=generator,
+        ...     eta=1.0,
+        ...     image=init_image,
+        ...     mask_image=mask_image,
+        ...     control_image=control_image,
+        ... ).images[0]
+        ```
+"""
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_inpaint.prepare_mask_and_masked_image
+def prepare_mask_and_masked_image(image, mask, height, width, return_image=False):
+    """
+    Prepares a pair (image, mask) to be consumed by the Stable Diffusion pipeline. This means that those inputs will be
+    converted to ``torch.Tensor`` with shapes ``batch x channels x height x width`` where ``channels`` is ``3`` for the
+    ``image`` and ``1`` for the ``mask``.
+
+    The ``image`` will be converted to ``torch.float32`` and normalized to be in ``[-1, 1]``. The ``mask`` will be
+    binarized (``mask > 0.5``) and cast to ``torch.float32`` too.
+
+    Args:
+        image (Union[np.array, PIL.Image, torch.Tensor]): The image to inpaint.
+            It can be a ``PIL.Image``, or a ``height x width x 3`` ``np.array`` or a ``channels x height x width``
+            ``torch.Tensor`` or a ``batch x channels x height x width`` ``torch.Tensor``.
+        mask (_type_): The mask to apply to the image, i.e. regions to inpaint.
+            It can be a ``PIL.Image``, or a ``height x width`` ``np.array`` or a ``1 x height x width``
+            ``torch.Tensor`` or a ``batch x 1 x height x width`` ``torch.Tensor``.
+
+
+    Raises:
+        ValueError: ``torch.Tensor`` images should be in the ``[-1, 1]`` range. ValueError: ``torch.Tensor`` mask
+        should be in the ``[0, 1]`` range. ValueError: ``mask`` and ``image`` should have the same spatial dimensions.
+        TypeError: ``mask`` is a ``torch.Tensor`` but ``image`` is not
+            (ot the other way around).
+
+    Returns:
+        tuple[torch.Tensor]: The pair (mask, masked_image) as ``torch.Tensor`` with 4
+            dimensions: ``batch x channels x height x width``.
+    """
+
+    if image is None:
+        raise ValueError("`image` input cannot be undefined.")
+
+    if mask is None:
+        raise ValueError("`mask_image` input cannot be undefined.")
+
+    if isinstance(image, torch.Tensor):
+        if not isinstance(mask, torch.Tensor):
+            raise TypeError(f"`image` is a torch.Tensor but `mask` (type: {type(mask)} is not")
+
+        # Batch single image
+        if image.ndim == 3:
+            assert image.shape[0] == 3, "Image outside a batch should be of shape (3, H, W)"
+            image = image.unsqueeze(0)
+
+        # Batch and add channel dim for single mask
+        if mask.ndim == 2:
+            mask = mask.unsqueeze(0).unsqueeze(0)
+
+        # Batch single mask or add channel dim
+        if mask.ndim == 3:
+            # Single batched mask, no channel dim or single mask not batched but channel dim
+            if mask.shape[0] == 1:
+                mask = mask.unsqueeze(0)
+
+            # Batched masks no channel dim
+            else:
+                mask = mask.unsqueeze(1)
+
+        assert image.ndim == 4 and mask.ndim == 4, "Image and Mask must have 4 dimensions"
+        assert image.shape[-2:] == mask.shape[-2:], "Image and Mask must have the same spatial dimensions"
+        assert image.shape[0] == mask.shape[0], "Image and Mask must have the same batch size"
+
+        # Check image is in [-1, 1]
+        if image.min() < -1 or image.max() > 1:
+            raise ValueError("Image should be in [-1, 1] range")
+
+        # Check mask is in [0, 1]
+        if mask.min() < 0 or mask.max() > 1:
+            raise ValueError("Mask should be in [0, 1] range")
+
+        # Binarize mask
+        mask[mask < 0.5] = 0
+        mask[mask >= 0.5] = 1
+
+        # Image as float32
+        image = image.to(dtype=torch.float32)
+    elif isinstance(mask, torch.Tensor):
+        raise TypeError(f"`mask` is a torch.Tensor but `image` (type: {type(image)} is not")
+    else:
+        # preprocess image
+        if isinstance(image, (PIL.Image.Image, np.ndarray)):
+            image = [image]
+        if isinstance(image, list) and isinstance(image[0], PIL.Image.Image):
+            # resize all images w.r.t passed height an width
+            image = [i.resize((width, height), resample=PIL.Image.LANCZOS) for i in image]
+            image = [np.array(i.convert("RGB"))[None, :] for i in image]
+            image = np.concatenate(image, axis=0)
+        elif isinstance(image, list) and isinstance(image[0], np.ndarray):
+            image = np.concatenate([i[None, :] for i in image], axis=0)
+
+        image = image.transpose(0, 3, 1, 2)
+        image = torch.from_numpy(image).to(dtype=torch.float32) / 127.5 - 1.0
+
+        # preprocess mask
+        if isinstance(mask, (PIL.Image.Image, np.ndarray)):
+            mask = [mask]
+
+        if isinstance(mask, list) and isinstance(mask[0], PIL.Image.Image):
+            mask = [i.resize((width, height), resample=PIL.Image.LANCZOS) for i in mask]
+            mask = np.concatenate([np.array(m.convert("L"))[None, None, :] for m in mask], axis=0)
+            mask = mask.astype(np.float32) / 255.0
+        elif isinstance(mask, list) and isinstance(mask[0], np.ndarray):
+            mask = np.concatenate([m[None, None, :] for m in mask], axis=0)
+
+        mask[mask < 0.5] = 0
+        mask[mask >= 0.5] = 1
+        mask = torch.from_numpy(mask)
+
+    masked_image = image * (mask < 0.5)
+
+    # n.b. ensure backwards compatibility as old function does not return image
+    if return_image:
+        return mask, masked_image, image
+
+    return mask, masked_image
+
+
+class StableDiffusionControlNetInpaintPipeline(
+    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
+):
+    r"""
+    Pipeline for text-to-image generation using Stable Diffusion with ControlNet guidance.
+
+    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.)
+
+    In addition the pipeline inherits the following loading methods:
+        - *Textual-Inversion*: [`loaders.TextualInversionLoaderMixin.load_textual_inversion`]
+
+    <Tip>
+
+    This pipeline can be used both with checkpoints that have been specifically fine-tuned for inpainting, such as
+    [runwayml/stable-diffusion-inpainting](https://huggingface.co/runwayml/stable-diffusion-inpainting)
+     as well as default text-to-image stable diffusion checkpoints, such as
+     [runwayml/stable-diffusion-v1-5](https://huggingface.co/runwayml/stable-diffusion-v1-5).
+    Default text-to-image stable diffusion checkpoints might be preferable for controlnets that have been fine-tuned on
+    those, such as [lllyasviel/control_v11p_sd15_inpaint](https://huggingface.co/lllyasviel/control_v11p_sd15_inpaint).
+
+    </Tip>
+
+    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.
+        controlnet ([`ControlNetModel`] or `List[ControlNetModel]`):
+            Provides additional conditioning to the unet during the denoising process. If you set multiple ControlNets
+            as a list, the outputs from each ControlNet are added together to create one combined additional
+            conditioning.
+        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 ([`CLIPImageProcessor`]):
+            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,
+        controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel], MultiControlNetModel],
+        scheduler: KarrasDiffusionSchedulers,
+        safety_checker: StableDiffusionSafetyChecker,
+        feature_extractor: CLIPImageProcessor,
+        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."
+            )
+
+        if isinstance(controlnet, (list, tuple)):
+            controlnet = MultiControlNetModel(controlnet)
+
+        self.register_modules(
+            vae=vae,
+            text_encoder=text_encoder,
+            tokenizer=tokenizer,
+            unet=unet,
+            controlnet=controlnet,
+            scheduler=scheduler,
+            safety_checker=safety_checker,
+            feature_extractor=feature_extractor,
+        )
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
+        self.control_image_processor = VaeImageProcessor(
+            vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
+        )
+        self.register_to_config(requires_safety_checker=requires_safety_checker)
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
+    def enable_vae_slicing(self):
+        r"""
+        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
+        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
+        """
+        self.vae.enable_slicing()
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
+    def disable_vae_slicing(self):
+        r"""
+        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
+        computing decoding in one step.
+        """
+        self.vae.disable_slicing()
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
+    def enable_vae_tiling(self):
+        r"""
+        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
+        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
+        processing larger images.
+        """
+        self.vae.enable_tiling()
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
+    def disable_vae_tiling(self):
+        r"""
+        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
+        computing decoding in one step.
+        """
+        self.vae.disable_tiling()
+
+    def enable_model_cpu_offload(self, gpu_id=0):
+        r"""
+        Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
+        to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
+        method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
+        `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
+        """
+        if is_accelerate_available() and is_accelerate_version(">=", "0.17.0.dev0"):
+            from accelerate import cpu_offload_with_hook
+        else:
+            raise ImportError("`enable_model_cpu_offload` requires `accelerate v0.17.0` or higher.")
+
+        device = torch.device(f"cuda:{gpu_id}")
+
+        hook = None
+        for cpu_offloaded_model in [self.text_encoder, self.unet, self.vae]:
+            _, hook = cpu_offload_with_hook(cpu_offloaded_model, device, prev_module_hook=hook)
+
+        if self.safety_checker is not None:
+            # the safety checker can offload the vae again
+            _, hook = cpu_offload_with_hook(self.safety_checker, device, prev_module_hook=hook)
+
+        # control net hook has be manually offloaded as it alternates with unet
+        cpu_offload_with_hook(self.controlnet, device)
+
+        # We'll offload the last model manually.
+        self.final_offload_hook = hook
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    def _encode_prompt(
+        self,
+        prompt,
+        device,
+        num_images_per_prompt,
+        do_classifier_free_guidance,
+        negative_prompt=None,
+        prompt_embeds: Optional[torch.FloatTensor] = None,
+        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        lora_scale: Optional[float] = None,
+    ):
+        r"""
+        Encodes the prompt into text encoder hidden states.
+
+        Args:
+             prompt (`str` or `List[str]`, *optional*):
+                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]`, *optional*):
+                The prompt or prompts not to guide the image generation. If not defined, one has to pass
+                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
+                less than `1`).
+            prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+            lora_scale (`float`, *optional*):
+                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
+        """
+        # set lora scale so that monkey patched LoRA
+        # function of text encoder can correctly access it
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
+            self._lora_scale = lora_scale
+
+        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]
+
+        if prompt_embeds is None:
+            # textual inversion: procecss multi-vector tokens if necessary
+            if isinstance(self, TextualInversionLoaderMixin):
+                prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
+
+            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
+
+            prompt_embeds = self.text_encoder(
+                text_input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            prompt_embeds = prompt_embeds[0]
+
+        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+
+        bs_embed, seq_len, _ = prompt_embeds.shape
+        # duplicate text embeddings for each generation per prompt, using mps friendly method
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
+
+        # get unconditional embeddings for classifier free guidance
+        if do_classifier_free_guidance and negative_prompt_embeds is None:
+            uncond_tokens: List[str]
+            if negative_prompt is None:
+                uncond_tokens = [""] * batch_size
+            elif prompt is not None and 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
+
+            # textual inversion: procecss multi-vector tokens if necessary
+            if isinstance(self, TextualInversionLoaderMixin):
+                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
+
+            max_length = prompt_embeds.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
+
+            negative_prompt_embeds = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            negative_prompt_embeds = negative_prompt_embeds[0]
+
+        if do_classifier_free_guidance:
+            # 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.to(dtype=self.text_encoder.dtype, device=device)
+
+            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 text embeddings into a single batch
+            # to avoid doing two forward passes
+            prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])
+
+        return prompt_embeds
+
+    # 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 None:
+            has_nsfw_concept = None
+        else:
+            if torch.is_tensor(image):
+                feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
+            else:
+                feature_extractor_input = self.image_processor.numpy_to_pil(image)
+            safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
+            image, has_nsfw_concept = self.safety_checker(
+                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
+            )
+        return image, has_nsfw_concept
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    def decode_latents(self, latents):
+        warnings.warn(
+            "The decode_latents method is deprecated and will be removed in a future version. Please"
+            " use VaeImageProcessor instead",
+            FutureWarning,
+        )
+        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_img2img.StableDiffusionImg2ImgPipeline.get_timesteps
+    def get_timesteps(self, num_inference_steps, strength, device):
+        # get the original timestep using init_timestep
+        init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
+
+        t_start = max(num_inference_steps - init_timestep, 0)
+        timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
+
+        return timesteps, num_inference_steps - t_start
+
+    def check_inputs(
+        self,
+        prompt,
+        image,
+        height,
+        width,
+        callback_steps,
+        negative_prompt=None,
+        prompt_embeds=None,
+        negative_prompt_embeds=None,
+        controlnet_conditioning_scale=1.0,
+        control_guidance_start=0.0,
+        control_guidance_end=1.0,
+    ):
+        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)}."
+            )
+
+        if prompt is not None and prompt_embeds is not None:
+            raise ValueError(
+                f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
+                " only forward one of the two."
+            )
+        elif prompt is None and prompt_embeds is None:
+            raise ValueError(
+                "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
+            )
+        elif prompt is not None and (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 negative_prompt is not None and negative_prompt_embeds is not None:
+            raise ValueError(
+                f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
+                f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
+            )
+
+        if prompt_embeds is not None and negative_prompt_embeds is not None:
+            if prompt_embeds.shape != negative_prompt_embeds.shape:
+                raise ValueError(
+                    "`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
+                    f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
+                    f" {negative_prompt_embeds.shape}."
+                )
+
+        # `prompt` needs more sophisticated handling when there are multiple
+        # conditionings.
+        if isinstance(self.controlnet, MultiControlNetModel):
+            if isinstance(prompt, list):
+                logger.warning(
+                    f"You have {len(self.controlnet.nets)} ControlNets and you have passed {len(prompt)}"
+                    " prompts. The conditionings will be fixed across the prompts."
+                )
+
+        # Check `image`
+        is_compiled = hasattr(F, "scaled_dot_product_attention") and isinstance(
+            self.controlnet, torch._dynamo.eval_frame.OptimizedModule
+        )
+        if (
+            isinstance(self.controlnet, ControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, ControlNetModel)
+        ):
+            self.check_image(image, prompt, prompt_embeds)
+        elif (
+            isinstance(self.controlnet, MultiControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, MultiControlNetModel)
+        ):
+            if not isinstance(image, list):
+                raise TypeError("For multiple controlnets: `image` must be type `list`")
+
+            # When `image` is a nested list:
+            # (e.g. [[canny_image_1, pose_image_1], [canny_image_2, pose_image_2]])
+            elif any(isinstance(i, list) for i in image):
+                raise ValueError("A single batch of multiple conditionings are supported at the moment.")
+            elif len(image) != len(self.controlnet.nets):
+                raise ValueError(
+                    f"For multiple controlnets: `image` must have the same length as the number of controlnets, but got {len(image)} images and {len(self.controlnet.nets)} ControlNets."
+                )
+
+            for image_ in image:
+                self.check_image(image_, prompt, prompt_embeds)
+        else:
+            assert False
+
+        # Check `controlnet_conditioning_scale`
+        if (
+            isinstance(self.controlnet, ControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, ControlNetModel)
+        ):
+            if not isinstance(controlnet_conditioning_scale, float):
+                raise TypeError("For single controlnet: `controlnet_conditioning_scale` must be type `float`.")
+        elif (
+            isinstance(self.controlnet, MultiControlNetModel)
+            or is_compiled
+            and isinstance(self.controlnet._orig_mod, MultiControlNetModel)
+        ):
+            if isinstance(controlnet_conditioning_scale, list):
+                if any(isinstance(i, list) for i in controlnet_conditioning_scale):
+                    raise ValueError("A single batch of multiple conditionings are supported at the moment.")
+            elif isinstance(controlnet_conditioning_scale, list) and len(controlnet_conditioning_scale) != len(
+                self.controlnet.nets
+            ):
+                raise ValueError(
+                    "For multiple controlnets: When `controlnet_conditioning_scale` is specified as `list`, it must have"
+                    " the same length as the number of controlnets"
+                )
+        else:
+            assert False
+
+        if len(control_guidance_start) != len(control_guidance_end):
+            raise ValueError(
+                f"`control_guidance_start` has {len(control_guidance_start)} elements, but `control_guidance_end` has {len(control_guidance_end)} elements. Make sure to provide the same number of elements to each list."
+            )
+
+        if isinstance(self.controlnet, MultiControlNetModel):
+            if len(control_guidance_start) != len(self.controlnet.nets):
+                raise ValueError(
+                    f"`control_guidance_start`: {control_guidance_start} has {len(control_guidance_start)} elements but there are {len(self.controlnet.nets)} controlnets available. Make sure to provide {len(self.controlnet.nets)}."
+                )
+
+        for start, end in zip(control_guidance_start, control_guidance_end):
+            if start >= end:
+                raise ValueError(
+                    f"control guidance start: {start} cannot be larger or equal to control guidance end: {end}."
+                )
+            if start < 0.0:
+                raise ValueError(f"control guidance start: {start} can't be smaller than 0.")
+            if end > 1.0:
+                raise ValueError(f"control guidance end: {end} can't be larger than 1.0.")
+
+    # Copied from diffusers.pipelines.controlnet.pipeline_controlnet.StableDiffusionControlNetPipeline.check_image
+    def check_image(self, image, prompt, prompt_embeds):
+        image_is_pil = isinstance(image, PIL.Image.Image)
+        image_is_tensor = isinstance(image, torch.Tensor)
+        image_is_np = isinstance(image, np.ndarray)
+        image_is_pil_list = isinstance(image, list) and isinstance(image[0], PIL.Image.Image)
+        image_is_tensor_list = isinstance(image, list) and isinstance(image[0], torch.Tensor)
+        image_is_np_list = isinstance(image, list) and isinstance(image[0], np.ndarray)
+
+        if (
+            not image_is_pil
+            and not image_is_tensor
+            and not image_is_np
+            and not image_is_pil_list
+            and not image_is_tensor_list
+            and not image_is_np_list
+        ):
+            raise TypeError(
+                f"image must be passed and be one of PIL image, numpy array, torch tensor, list of PIL images, list of numpy arrays or list of torch tensors, but is {type(image)}"
+            )
+
+        if image_is_pil:
+            image_batch_size = 1
+        else:
+            image_batch_size = len(image)
+
+        if prompt is not None and isinstance(prompt, str):
+            prompt_batch_size = 1
+        elif prompt is not None and isinstance(prompt, list):
+            prompt_batch_size = len(prompt)
+        elif prompt_embeds is not None:
+            prompt_batch_size = prompt_embeds.shape[0]
+
+        if image_batch_size != 1 and image_batch_size != prompt_batch_size:
+            raise ValueError(
+                f"If image batch size is not 1, image batch size must be same as prompt batch size. image batch size: {image_batch_size}, prompt batch size: {prompt_batch_size}"
+            )
+
+    # Copied from diffusers.pipelines.controlnet.pipeline_controlnet.StableDiffusionControlNetPipeline.prepare_image
+    def prepare_control_image(
+        self,
+        image,
+        width,
+        height,
+        batch_size,
+        num_images_per_prompt,
+        device,
+        dtype,
+        do_classifier_free_guidance=False,
+        guess_mode=False,
+    ):
+        image = self.control_image_processor.preprocess(image, height=height, width=width).to(dtype=torch.float32)
+        image_batch_size = image.shape[0]
+
+        if image_batch_size == 1:
+            repeat_by = batch_size
+        else:
+            # image batch size is the same as prompt batch size
+            repeat_by = num_images_per_prompt
+
+        image = image.repeat_interleave(repeat_by, dim=0)
+
+        image = image.to(device=device, dtype=dtype)
+
+        if do_classifier_free_guidance and not guess_mode:
+            image = torch.cat([image] * 2)
+
+        return image
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_inpaint.StableDiffusionInpaintPipeline.prepare_latents
+    def prepare_latents(
+        self,
+        batch_size,
+        num_channels_latents,
+        height,
+        width,
+        dtype,
+        device,
+        generator,
+        latents=None,
+        image=None,
+        timestep=None,
+        is_strength_max=True,
+        return_noise=False,
+        return_image_latents=False,
+    ):
+        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 (image is None or timestep is None) and not is_strength_max:
+            raise ValueError(
+                "Since strength < 1. initial latents are to be initialised as a combination of Image + Noise."
+                "However, either the image or the noise timestep has not been provided."
+            )
+
+        if return_image_latents or (latents is None and not is_strength_max):
+            image = image.to(device=device, dtype=dtype)
+            image_latents = self._encode_vae_image(image=image, generator=generator)
+
+        if latents is None:
+            noise = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
+            # if strength is 1. then initialise the latents to noise, else initial to image + noise
+            latents = noise if is_strength_max else self.scheduler.add_noise(image_latents, noise, timestep)
+            # if pure noise then scale the initial latents by the  Scheduler's init sigma
+            latents = latents * self.scheduler.init_noise_sigma if is_strength_max else latents
+        else:
+            noise = latents.to(device)
+            latents = noise * self.scheduler.init_noise_sigma
+
+        outputs = (latents,)
+
+        if return_noise:
+            outputs += (noise,)
+
+        if return_image_latents:
+            outputs += (image_latents,)
+
+        return outputs
+
+    def _default_height_width(self, height, width, image):
+        # NOTE: It is possible that a list of images have different
+        # dimensions for each image, so just checking the first image
+        # is not _exactly_ correct, but it is simple.
+        while isinstance(image, list):
+            image = image[0]
+
+        if height is None:
+            if isinstance(image, PIL.Image.Image):
+                height = image.height
+            elif isinstance(image, torch.Tensor):
+                height = image.shape[2]
+
+            height = (height // 8) * 8  # round down to nearest multiple of 8
+
+        if width is None:
+            if isinstance(image, PIL.Image.Image):
+                width = image.width
+            elif isinstance(image, torch.Tensor):
+                width = image.shape[3]
+
+            width = (width // 8) * 8  # round down to nearest multiple of 8
+
+        return height, width
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_inpaint.StableDiffusionInpaintPipeline.prepare_mask_latents
+    def prepare_mask_latents(
+        self, mask, masked_image, batch_size, height, width, dtype, device, generator, do_classifier_free_guidance
+    ):
+        # resize the mask to latents shape as we concatenate the mask to the latents
+        # we do that before converting to dtype to avoid breaking in case we're using cpu_offload
+        # and half precision
+        mask = torch.nn.functional.interpolate(
+            mask, size=(height // self.vae_scale_factor, width // self.vae_scale_factor)
+        )
+        mask = mask.to(device=device, dtype=dtype)
+
+        masked_image = masked_image.to(device=device, dtype=dtype)
+        masked_image_latents = self._encode_vae_image(masked_image, generator=generator)
+
+        # duplicate mask and masked_image_latents for each generation per prompt, using mps friendly method
+        if mask.shape[0] < batch_size:
+            if not batch_size % mask.shape[0] == 0:
+                raise ValueError(
+                    "The passed mask and the required batch size don't match. Masks are supposed to be duplicated to"
+                    f" a total batch size of {batch_size}, but {mask.shape[0]} masks were passed. Make sure the number"
+                    " of masks that you pass is divisible by the total requested batch size."
+                )
+            mask = mask.repeat(batch_size // mask.shape[0], 1, 1, 1)
+        if masked_image_latents.shape[0] < batch_size:
+            if not batch_size % masked_image_latents.shape[0] == 0:
+                raise ValueError(
+                    "The passed images and the required batch size don't match. Images are supposed to be duplicated"
+                    f" to a total batch size of {batch_size}, but {masked_image_latents.shape[0]} images were passed."
+                    " Make sure the number of images that you pass is divisible by the total requested batch size."
+                )
+            masked_image_latents = masked_image_latents.repeat(batch_size // masked_image_latents.shape[0], 1, 1, 1)
+
+        mask = torch.cat([mask] * 2) if do_classifier_free_guidance else mask
+        masked_image_latents = (
+            torch.cat([masked_image_latents] * 2) if do_classifier_free_guidance else masked_image_latents
+        )
+
+        # aligning device to prevent device errors when concating it with the latent model input
+        masked_image_latents = masked_image_latents.to(device=device, dtype=dtype)
+        return mask, masked_image_latents
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_inpaint.StableDiffusionInpaintPipeline._encode_vae_image
+    def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
+        if isinstance(generator, list):
+            image_latents = [
+                self.vae.encode(image[i : i + 1]).latent_dist.sample(generator=generator[i])
+                for i in range(image.shape[0])
+            ]
+            image_latents = torch.cat(image_latents, dim=0)
+        else:
+            image_latents = self.vae.encode(image).latent_dist.sample(generator=generator)
+
+        image_latents = self.vae.config.scaling_factor * image_latents
+
+        return image_latents
+
+    @torch.no_grad()
+    @replace_example_docstring(EXAMPLE_DOC_STRING)
+    def __call__(
+        self,
+        prompt: Union[str, List[str]] = None,
+        image: Union[torch.Tensor, PIL.Image.Image] = None,
+        mask_image: Union[torch.Tensor, PIL.Image.Image] = None,
+        control_image: Union[
+            torch.FloatTensor,
+            PIL.Image.Image,
+            np.ndarray,
+            List[torch.FloatTensor],
+            List[PIL.Image.Image],
+            List[np.ndarray],
+        ] = None,
+        height: Optional[int] = None,
+        width: Optional[int] = None,
+        strength: float = 1.0,
+        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,
+        prompt_embeds: Optional[torch.FloatTensor] = None,
+        negative_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,
+        controlnet_conditioning_scale: Union[float, List[float]] = 0.5,
+        guess_mode: bool = False,
+        control_guidance_start: Union[float, List[float]] = 0.0,
+        control_guidance_end: Union[float, List[float]] = 1.0,
+    ):
+        r"""
+        Function invoked when calling the pipeline for generation.
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
+                instead.
+            image (`torch.FloatTensor`, `PIL.Image.Image`, `List[torch.FloatTensor]`, `List[PIL.Image.Image]`,
+                    `List[List[torch.FloatTensor]]`, or `List[List[PIL.Image.Image]]`):
+                The ControlNet input condition. ControlNet uses this input condition to generate guidance to Unet. If
+                the type is specified as `Torch.FloatTensor`, it is passed to ControlNet as is. `PIL.Image.Image` can
+                also be accepted as an image. The dimensions of the output image defaults to `image`'s dimensions. If
+                height and/or width are passed, `image` is resized according to them. If multiple ControlNets are
+                specified in init, images must be passed as a list such that each element of the list can be correctly
+                batched for input to a single controlnet.
+            height (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
+                The height in pixels of the generated image.
+            width (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
+                The width in pixels of the generated image.
+            strength (`float`, *optional*, defaults to 1.):
+                Conceptually, indicates how much to transform the masked portion of the reference `image`. Must be
+                between 0 and 1. `image` will be used as a starting point, adding more noise to it the larger the
+                `strength`. The number of denoising steps depends on the amount of noise initially added. When
+                `strength` is 1, added noise will be maximum and the denoising process will run for the full number of
+                iterations specified in `num_inference_steps`. A value of 1, therefore, essentially ignores the masked
+                portion of the reference `image`.
+            num_inference_steps (`int`, *optional*, defaults to 50):
+                The number of denoising steps. More denoising steps usually lead to a higher quality image at the
+                expense of slower inference.
+            guidance_scale (`float`, *optional*, defaults to 7.5):
+                Guidance scale as defined in [Classifier-Free Diffusion Guidance](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.
+            negative_prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts not to guide the image generation. If not defined, one has to pass
+                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
+                less than `1`).
+            num_images_per_prompt (`int`, *optional*, defaults to 1):
+                The number of images to generate per prompt.
+            eta (`float`, *optional*, defaults to 0.0):
+                Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
+                [`schedulers.DDIMScheduler`], will be ignored for others.
+            generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
+                One or a list of [torch generator(s)](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`.
+            prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generate image. Choose between
+                [PIL](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.
+            cross_attention_kwargs (`dict`, *optional*):
+                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
+                `self.processor` in
+                [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
+            controlnet_conditioning_scale (`float` or `List[float]`, *optional*, defaults to 0.5):
+                The outputs of the controlnet are multiplied by `controlnet_conditioning_scale` before they are added
+                to the residual in the original unet. If multiple ControlNets are specified in init, you can set the
+                corresponding scale as a list. Note that by default, we use a smaller conditioning scale for inpainting
+                than for [`~StableDiffusionControlNetPipeline.__call__`].
+            guess_mode (`bool`, *optional*, defaults to `False`):
+                In this mode, the ControlNet encoder will try best to recognize the content of the input image even if
+                you remove all prompts. The `guidance_scale` between 3.0 and 5.0 is recommended.
+            control_guidance_start (`float` or `List[float]`, *optional*, defaults to 0.0):
+                The percentage of total steps at which the controlnet starts applying.
+            control_guidance_end (`float` or `List[float]`, *optional*, defaults to 1.0):
+                The percentage of total steps at which the controlnet stops applying.
+
+        Examples:
+
+        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`.
+        """
+        controlnet = self.controlnet._orig_mod if is_compiled_module(self.controlnet) else self.controlnet
+
+        # 0. Default height and width to unet
+        height, width = self._default_height_width(height, width, image)
+
+        # align format for control guidance
+        if not isinstance(control_guidance_start, list) and isinstance(control_guidance_end, list):
+            control_guidance_start = len(control_guidance_end) * [control_guidance_start]
+        elif not isinstance(control_guidance_end, list) and isinstance(control_guidance_start, list):
+            control_guidance_end = len(control_guidance_start) * [control_guidance_end]
+        elif not isinstance(control_guidance_start, list) and not isinstance(control_guidance_end, list):
+            mult = len(controlnet.nets) if isinstance(controlnet, MultiControlNetModel) else 1
+            control_guidance_start, control_guidance_end = mult * [control_guidance_start], mult * [
+                control_guidance_end
+            ]
+
+        # 1. Check inputs. Raise error if not correct
+        self.check_inputs(
+            prompt,
+            control_image,
+            height,
+            width,
+            callback_steps,
+            negative_prompt,
+            prompt_embeds,
+            negative_prompt_embeds,
+            controlnet_conditioning_scale,
+            control_guidance_start,
+            control_guidance_end,
+        )
+
+        # 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
+
+        if isinstance(controlnet, MultiControlNetModel) and isinstance(controlnet_conditioning_scale, float):
+            controlnet_conditioning_scale = [controlnet_conditioning_scale] * len(controlnet.nets)
+
+        global_pool_conditions = (
+            controlnet.config.global_pool_conditions
+            if isinstance(controlnet, ControlNetModel)
+            else controlnet.nets[0].config.global_pool_conditions
+        )
+        guess_mode = guess_mode or global_pool_conditions
+
+        # 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 = self._encode_prompt(
+            prompt,
+            device,
+            num_images_per_prompt,
+            do_classifier_free_guidance,
+            negative_prompt,
+            prompt_embeds=prompt_embeds,
+            negative_prompt_embeds=negative_prompt_embeds,
+            lora_scale=text_encoder_lora_scale,
+        )
+
+        # 4. Prepare image
+        if isinstance(controlnet, ControlNetModel):
+            control_image = self.prepare_control_image(
+                image=control_image,
+                width=width,
+                height=height,
+                batch_size=batch_size * num_images_per_prompt,
+                num_images_per_prompt=num_images_per_prompt,
+                device=device,
+                dtype=controlnet.dtype,
+                do_classifier_free_guidance=do_classifier_free_guidance,
+                guess_mode=guess_mode,
+            )
+        elif isinstance(controlnet, MultiControlNetModel):
+            control_images = []
+
+            for control_image_ in control_image:
+                control_image_ = self.prepare_control_image(
+                    image=control_image_,
+                    width=width,
+                    height=height,
+                    batch_size=batch_size * num_images_per_prompt,
+                    num_images_per_prompt=num_images_per_prompt,
+                    device=device,
+                    dtype=controlnet.dtype,
+                    do_classifier_free_guidance=do_classifier_free_guidance,
+                    guess_mode=guess_mode,
+                )
+
+                control_images.append(control_image_)
+
+            control_image = control_images
+        else:
+            assert False
+
+        # 4. Preprocess mask and image - resizes image and mask w.r.t height and width
+        mask, masked_image, init_image = prepare_mask_and_masked_image(
+            image, mask_image, height, width, return_image=True
+        )
+
+        # 5. Prepare timesteps
+        self.scheduler.set_timesteps(num_inference_steps, device=device)
+        timesteps, num_inference_steps = self.get_timesteps(
+            num_inference_steps=num_inference_steps, strength=strength, device=device
+        )
+        # at which timestep to set the initial noise (n.b. 50% if strength is 0.5)
+        latent_timestep = timesteps[:1].repeat(batch_size * num_images_per_prompt)
+        # create a boolean to check if the strength is set to 1. if so then initialise the latents with pure noise
+        is_strength_max = strength == 1.0
+
+        # 6. Prepare latent variables
+        num_channels_latents = self.vae.config.latent_channels
+        num_channels_unet = self.unet.config.in_channels
+        return_image_latents = num_channels_unet == 4
+        latents_outputs = self.prepare_latents(
+            batch_size * num_images_per_prompt,
+            num_channels_latents,
+            height,
+            width,
+            prompt_embeds.dtype,
+            device,
+            generator,
+            latents,
+            image=init_image,
+            timestep=latent_timestep,
+            is_strength_max=is_strength_max,
+            return_noise=True,
+            return_image_latents=return_image_latents,
+        )
+
+        if return_image_latents:
+            latents, noise, image_latents = latents_outputs
+        else:
+            latents, noise = latents_outputs
+
+        # 7. Prepare mask latent variables
+        mask, masked_image_latents = self.prepare_mask_latents(
+            mask,
+            masked_image,
+            batch_size * num_images_per_prompt,
+            height,
+            width,
+            prompt_embeds.dtype,
+            device,
+            generator,
+            do_classifier_free_guidance,
+        )
+
+        # 7. 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)
+
+        # 7.1 Create tensor stating which controlnets to keep
+        controlnet_keep = []
+        for i in range(len(timesteps)):
+            keeps = [
+                1.0 - float(i / len(timesteps) < s or (i + 1) / len(timesteps) > e)
+                for s, e in zip(control_guidance_start, control_guidance_end)
+            ]
+            controlnet_keep.append(keeps[0] if isinstance(controlnet, ControlNetModel) else keeps)
+
+        # 8. 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
+                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)
+
+                # controlnet(s) inference
+                if guess_mode and do_classifier_free_guidance:
+                    # Infer ControlNet only for the conditional batch.
+                    control_model_input = latents
+                    control_model_input = self.scheduler.scale_model_input(control_model_input, t)
+                    controlnet_prompt_embeds = prompt_embeds.chunk(2)[1]
+                else:
+                    control_model_input = latent_model_input
+                    controlnet_prompt_embeds = prompt_embeds
+
+                if isinstance(controlnet_keep[i], list):
+                    cond_scale = [c * s for c, s in zip(controlnet_conditioning_scale, controlnet_keep[i])]
+                else:
+                    cond_scale = controlnet_conditioning_scale * controlnet_keep[i]
+
+                # predict the noise residual
+                if num_channels_unet == 9:
+                    latent_model_input = torch.cat([latent_model_input, mask, masked_image_latents], dim=1)
+
+                down_block_res_samples, mid_block_res_sample = self.controlnet(
+                    latent_model_input, #control_model_input,
+                    t,
+                    encoder_hidden_states=controlnet_prompt_embeds,
+                    controlnet_cond=control_image,
+                    conditioning_scale=cond_scale,
+                    guess_mode=guess_mode,
+                    return_dict=False,
+                )
+
+                if guess_mode and do_classifier_free_guidance:
+                    # Infered ControlNet only for the conditional batch.
+                    # To apply the output of ControlNet to both the unconditional and conditional batches,
+                    # add 0 to the unconditional batch to keep it unchanged.
+                    down_block_res_samples = [torch.cat([torch.zeros_like(d), d]) for d in down_block_res_samples]
+                    mid_block_res_sample = torch.cat([torch.zeros_like(mid_block_res_sample), mid_block_res_sample])
+
+                noise_pred = self.unet(
+                    latent_model_input,
+                    t,
+                    encoder_hidden_states=prompt_embeds,
+                    cross_attention_kwargs=cross_attention_kwargs,
+                    down_block_additional_residuals=down_block_res_samples,
+                    mid_block_additional_residual=mid_block_res_sample,
+                    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)
+
+                # 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]
+
+                if num_channels_unet == 4:
+                    init_latents_proper = image_latents[:1]
+                    init_mask = mask[:1]
+
+                    if i < len(timesteps) - 1:
+                        noise_timestep = timesteps[i + 1]
+                        init_latents_proper = self.scheduler.add_noise(
+                            init_latents_proper, noise, torch.tensor([noise_timestep])
+                        )
+
+                    latents = (1 - init_mask) * init_latents_proper + init_mask * latents
+
+                # 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)
+
+        # If we do sequential model offloading, let's offload unet and controlnet
+        # manually for max memory savings
+        if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None:
+            self.unet.to("cpu")
+            self.controlnet.to("cpu")
+            torch.cuda.empty_cache()
+
+        if not output_type == "latent":
+            image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False)[0]
+            image, has_nsfw_concept = self.run_safety_checker(image, device, prompt_embeds.dtype)
+        else:
+            image = latents
+            has_nsfw_concept = None
+
+        if has_nsfw_concept is None:
+            do_denormalize = [True] * image.shape[0]
+        else:
+            do_denormalize = [not has_nsfw for has_nsfw in has_nsfw_concept]
+
+        image = self.image_processor.postprocess(image, output_type=output_type, do_denormalize=do_denormalize)
+
+        # Offload last model to CPU
+        if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None:
+            self.final_offload_hook.offload()
+
+        if not return_dict:
+            return (image, has_nsfw_concept)
+
+        return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)

requirements.txt

@@ -0,0 +1,10 @@
+accelerate==0.28.0
+transformers==4.39.3
+pyarrow==15.0.2
+ftfy==6.2.0
+tensorboard==2.14.0
+datasets==2.18.0
+torchvision==0.17.2
+jupyterlab==4.1.6
+diffusers==0.27.2
+transparent-background==1.2.12

screwdriver.yaml

@@ -0,0 +1,16 @@
+jobs:
+  validate-semgrep-sast:
+    template: ProdSec/validate_semgrep@stable
+    image: alma8
+    environment:
+      YAHOO_SEMGREP_ENFORCING: False #(If you choose to fail builds for validation failures in Semgrep, then you should set this value to True)
+      YAHOO_SEMGREP_ONLINE: True
+
+  checkov:
+    requires: [~pr, ~commit]
+    image: docker.ouroath.com:4443/containers/python3:latest
+    steps:
+      - run: |
+          sd-cmd exec ProdSec/checkov@stable -d $SD_SOURCE_DIR
+    environment:
+      CHECKOV_HARD_FAIL_ON_FINDINGS: false

train_controlnet.py

@@ -0,0 +1,1255 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2024, Yahoo Research
+#
+# 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 argparse
+import logging
+import math
+import os
+import random
+import shutil
+from pathlib import Path
+
+import cv2
+import accelerate
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+import transformers
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import ProjectConfiguration, set_seed
+from datasets import load_dataset
+from huggingface_hub import create_repo, upload_folder
+from packaging import version
+from PIL import Image, ImageOps
+from torchvision import transforms
+from tqdm.auto import tqdm
+from transformers import AutoTokenizer, PretrainedConfig
+import diffusers
+from diffusers import (
+    AutoencoderKL,
+    ControlNetModel,
+    DDPMScheduler,
+    StableDiffusionControlNetInpaintPipeline,
+    UNet2DConditionModel,
+    UniPCMultistepScheduler,
+)
+from diffusers.optimization import get_scheduler
+from diffusers.utils import check_min_version, is_wandb_available
+from diffusers.utils.import_utils import is_xformers_available
+
+
+if is_wandb_available():
+    import wandb
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.20.0.dev0")
+
+logger = get_logger(__name__)
+
+
+def image_grid(imgs, rows, cols):
+    assert len(imgs) == rows * cols
+
+    w, h = imgs[0].size
+    grid = Image.new("RGB", size=(cols * w, rows * h))
+
+    for i, img in enumerate(imgs):
+        grid.paste(img, box=(i % cols * w, i // cols * h))
+    return grid
+
+
+def resize_with_padding(img, expected_size):
+    img.thumbnail((expected_size[0], expected_size[1]))
+    # print(img.size)
+    delta_width = expected_size[0] - img.size[0]
+    delta_height = expected_size[1] - img.size[1]
+    pad_width = delta_width // 2
+    pad_height = delta_height // 2
+    padding = (pad_width, pad_height, delta_width - pad_width, delta_height - pad_height)
+    return ImageOps.expand(img, padding)
+
+def log_validation(vae, text_encoder, tokenizer, unet, controlnet, args, accelerator, weight_dtype, step):
+    logger.info("Running validation... ")
+
+    controlnet = accelerator.unwrap_model(controlnet)
+
+    pipeline = StableDiffusionControlNetInpaintPipeline.from_pretrained(
+        args.pretrained_model_name_or_path,
+        vae=vae,
+        text_encoder=text_encoder,
+        tokenizer=tokenizer,
+        unet=unet,
+        controlnet=controlnet,
+        safety_checker=None,
+        revision=args.revision,
+        torch_dtype=weight_dtype,
+    )
+    pipeline.scheduler = UniPCMultistepScheduler.from_config(pipeline.scheduler.config)
+    pipeline = pipeline.to(accelerator.device)
+    pipeline.set_progress_bar_config(disable=True)
+
+    if args.enable_xformers_memory_efficient_attention:
+        pipeline.enable_xformers_memory_efficient_attention()
+
+    if args.seed is None:
+        generator = None
+    else:
+        generator = torch.Generator(device=accelerator.device).manual_seed(args.seed)
+
+    if len(args.validation_image) == len(args.validation_prompt):
+        validation_images = args.validation_image
+        validation_inpainting_images = args.validation_inpainting_image
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_image) == 1:
+        validation_images = args.validation_image * len(args.validation_prompt)
+        validation_inpainting_images = args.validation_inpainting_image * len(args.validation_prompt)
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_prompt) == 1:
+        validation_images = args.validation_image
+        validation_inpainting_images = args.validation_inpainting_image
+        validation_prompts = args.validation_prompt * len(args.validation_image)
+    else:
+        raise ValueError(
+            "number of `args.validation_image` and `args.validation_prompt` should be checked in `parse_args`"
+        )
+
+    image_logs = []
+
+    for validation_prompt, validation_image, validation_inpainting_image in zip(validation_prompts, validation_images, validation_inpainting_images):
+        validation_image = Image.open(validation_image).convert("RGB")
+        validation_image = resize_with_padding(validation_image, (512,512))
+        validation_inpainting_image = Image.open(validation_inpainting_image).convert("RGB")
+        validation_inpainting_image = resize_with_padding(validation_inpainting_image, (512,512))
+        images = []
+
+        for _ in range(args.num_validation_images):
+            with torch.autocast("cuda"):
+                mask = ImageOps.invert(validation_image)
+                control_image = ImageOps.invert(validation_image)
+                #control_image.paste(validation_inpainting_image, box=(0,0), mask=ImageOps.invert(control_image).convert('L'))
+#                control_image.save('cont_img_val.jpeg')
+                image = pipeline(
+                    prompt=validation_prompt, image=validation_inpainting_image, mask_image=mask, control_image=control_image, num_inference_steps=20, guess_mode=False, controlnet_conditioning_scale=1.0, generator=generator
+                ).images[0]
+
+            images.append(image)
+
+        image_logs.append(
+            {"validation_image": validation_image, "images": images, "validation_prompt": validation_prompt}
+        )
+
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images = []
+
+                formatted_images.append(np.asarray(validation_image))
+
+                for image in images:
+                    formatted_images.append(np.asarray(image))
+
+                formatted_images = np.stack(formatted_images)
+
+                tracker.writer.add_images(validation_prompt, formatted_images, step, dataformats="NHWC")
+        elif tracker.name == "wandb":
+            formatted_images = []
+
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images.append(wandb.Image(validation_image, caption="Controlnet conditioning"))
+
+                for image in images:
+                    image = wandb.Image(image, caption=validation_prompt)
+                    formatted_images.append(image)
+
+            tracker.log({"validation": formatted_images})
+        else:
+            logger.warn(f"image logging not implemented for {tracker.name}")
+
+        return image_logs
+
+
+def import_model_class_from_model_name_or_path(pretrained_model_name_or_path: str, revision: str):
+    text_encoder_config = PretrainedConfig.from_pretrained(
+        pretrained_model_name_or_path,
+        subfolder="text_encoder",
+        revision=revision,
+    )
+    model_class = text_encoder_config.architectures[0]
+
+    if model_class == "CLIPTextModel":
+        from transformers import CLIPTextModel
+
+        return CLIPTextModel
+    elif model_class == "RobertaSeriesModelWithTransformation":
+        from diffusers.pipelines.alt_diffusion.modeling_roberta_series import RobertaSeriesModelWithTransformation
+
+        return RobertaSeriesModelWithTransformation
+    else:
+        raise ValueError(f"{model_class} is not supported.")
+
+
+def save_model_card(repo_id: str, image_logs=None, base_model=str, repo_folder=None):
+    img_str = ""
+    if image_logs is not None:
+        img_str = "You can find some example images below.\n"
+        for i, log in enumerate(image_logs):
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+            validation_image.save(os.path.join(repo_folder, "image_control.png"))
+            img_str += f"prompt: {validation_prompt}\n"
+            images = [validation_image] + images
+            image_grid(images, 1, len(images)).save(os.path.join(repo_folder, f"images_{i}.png"))
+            img_str += f"![images_{i})](./images_{i}.png)\n"
+
+    yaml = f"""
+---
+license: creativeml-openrail-m
+base_model: {base_model}
+tags:
+- stable-diffusion
+- stable-diffusion-diffusers
+- text-to-image
+- diffusers
+- controlnet
+inference: true
+---
+    """
+    model_card = f"""
+# controlnet-{repo_id}
+
+These are controlnet weights trained on {base_model} with new type of conditioning.
+{img_str}
+"""
+    with open(os.path.join(repo_folder, "README.md"), "w") as f:
+        f.write(yaml + model_card)
+
+
+def parse_args(input_args=None):
+    parser = argparse.ArgumentParser(description="Simple example of a ControlNet training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        default=None,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--controlnet_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to pretrained controlnet model or model identifier from huggingface.co/models."
+        " If not specified controlnet weights are initialized from unet.",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        required=False,
+        help=(
+            "Revision of pretrained model identifier from huggingface.co/models. Trainable model components should be"
+            " float32 precision."
+        ),
+    )
+    parser.add_argument(
+        "--tokenizer_name",
+        type=str,
+        default=None,
+        help="Pretrained tokenizer name or path if not the same as model_name",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="controlnet-model",
+        help="The output directory where the model predictions and checkpoints will be written.",
+    )
+    parser.add_argument(
+        "--cache_dir",
+        type=str,
+        default=None,
+        help="The directory where the downloaded models and datasets will be stored.",
+    )
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=1)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=500,
+        help=(
+            "Save a checkpoint of the training state every X updates. Checkpoints can be used for resuming training via `--resume_from_checkpoint`. "
+            "In the case that the checkpoint is better than the final trained model, the checkpoint can also be used for inference."
+            "Using a checkpoint for inference requires separate loading of the original pipeline and the individual checkpointed model components."
+            "See https://huggingface.co/docs/diffusers/main/en/training/dreambooth#performing-inference-using-a-saved-checkpoint for step by step"
+            "instructions."
+        ),
+    )
+    parser.add_argument(
+        "--checkpoints_total_limit",
+        type=int,
+        default=None,
+        help=("Max number of checkpoints to store."),
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help=(
+            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
+            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
+        ),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--gradient_checkpointing",
+        action="store_true",
+        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=5e-6,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        default=False,
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument(
+        "--lr_num_cycles",
+        type=int,
+        default=1,
+        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
+    )
+    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
+    parser.add_argument(
+        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
+    )
+    parser.add_argument(
+        "--dataloader_num_workers",
+        type=int,
+        default=0,
+        help=(
+            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
+        ),
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_dir",
+        type=str,
+        default="logs",
+        help=(
+            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
+            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
+        ),
+    )
+    parser.add_argument(
+        "--allow_tf32",
+        action="store_true",
+        help=(
+            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
+            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
+        ),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="tensorboard",
+        help=(
+            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
+            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
+        ),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
+            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
+        ),
+    )
+    parser.add_argument(
+        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
+    )
+    parser.add_argument(
+        "--set_grads_to_none",
+        action="store_true",
+        help=(
+            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
+            " behaviors, so disable this argument if it causes any problems. More info:"
+            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
+        ),
+    )
+    parser.add_argument(
+        "--dataset_name",
+        type=str,
+        default=None,
+        help=(
+            "The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
+            " dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
+            " or to a folder containing files that 🤗 Datasets can understand."
+        ),
+    )
+    parser.add_argument(
+        "--dataset_config_name",
+        type=str,
+        default=None,
+        help="The config of the Dataset, leave as None if there's only one config.",
+    )
+    parser.add_argument(
+        "--train_data_dir",
+        type=str,
+        default=None,
+        help=(
+            "A folder containing the training data. Folder contents must follow the structure described in"
+            " https://huggingface.co/docs/datasets/image_dataset#imagefolder. In particular, a `metadata.jsonl` file"
+            " must exist to provide the captions for the images. Ignored if `dataset_name` is specified."
+        ),
+    )
+    parser.add_argument(
+        "--image_column", type=str, default="image", help="The column of the dataset containing the target image."
+    )
+    parser.add_argument(
+        "--conditioning_image_column",
+        type=str,
+        default="conditioning_image",
+        help="The column of the dataset containing the controlnet conditioning image.",
+    )
+    parser.add_argument(
+        "--caption_column",
+        type=str,
+        default="text",
+        help="The column of the dataset containing a caption or a list of captions.",
+    )
+    parser.add_argument(
+        "--max_train_samples",
+        type=int,
+        default=None,
+        help=(
+            "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set."
+        ),
+    )
+    parser.add_argument(
+        "--proportion_empty_prompts",
+        type=float,
+        default=0,
+        help="Proportion of image prompts to be replaced with empty strings. Defaults to 0 (no prompt replacement).",
+    )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of prompts evaluated every `--validation_steps` and logged to `--report_to`."
+            " Provide either a matching number of `--validation_image`s, a single `--validation_image`"
+            " to be used with all prompts, or a single prompt that will be used with all `--validation_image`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_inpainting_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+
+    parser.add_argument(
+        "--validation_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=4,
+        help="Number of images to be generated for each `--validation_image`, `--validation_prompt` pair",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` multiple times: `args.num_validation_images`"
+            " and logging the images."
+        ),
+    )
+    parser.add_argument(
+        "--tracker_project_name",
+        type=str,
+        default="train_controlnet",
+        help=(
+            "The `project_name` argument passed to Accelerator.init_trackers for"
+            " more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
+        ),
+    )
+
+    if input_args is not None:
+        args = parser.parse_args(input_args)
+    else:
+        args = parser.parse_args()
+
+    if args.dataset_name is None and args.train_data_dir is None:
+        raise ValueError("Specify either `--dataset_name` or `--train_data_dir`")
+
+    if args.dataset_name is not None and args.train_data_dir is not None:
+        raise ValueError("Specify only one of `--dataset_name` or `--train_data_dir`")
+
+    if args.proportion_empty_prompts < 0 or args.proportion_empty_prompts > 1:
+        raise ValueError("`--proportion_empty_prompts` must be in the range [0, 1].")
+
+    if args.validation_prompt is not None and args.validation_image is None:
+        raise ValueError("`--validation_image` must be set if `--validation_prompt` is set")
+
+    if args.validation_prompt is None and args.validation_image is not None:
+        raise ValueError("`--validation_prompt` must be set if `--validation_image` is set")
+
+    if (
+        args.validation_image is not None
+        and args.validation_prompt is not None
+        and len(args.validation_image) != 1
+        and len(args.validation_prompt) != 1
+        and len(args.validation_image) != len(args.validation_prompt)
+    ):
+        raise ValueError(
+            "Must provide either 1 `--validation_image`, 1 `--validation_prompt`,"
+            " or the same number of `--validation_prompt`s and `--validation_image`s"
+        )
+
+    if args.resolution % 8 != 0:
+        raise ValueError(
+            "`--resolution` must be divisible by 8 for consistently sized encoded images between the VAE and the controlnet encoder."
+        )
+
+    return args
+
+
+def make_train_dataset(args, tokenizer, accelerator):
+    # Get the datasets: you can either provide your own training and evaluation files (see below)
+    # or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
+
+    # In distributed training, the load_dataset function guarantees that only one local process can concurrently
+    # download the dataset.
+    if args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        dataset = load_dataset(
+            args.dataset_name,
+            args.dataset_config_name,
+            cache_dir=args.cache_dir,
+        )
+    else:
+        if args.train_data_dir is not None:
+            dataset = load_dataset(
+                args.train_data_dir,
+                cache_dir=args.cache_dir,
+            )
+        # See more about loading custom images at
+        # https://huggingface.co/docs/datasets/v2.0.0/en/dataset_script
+
+    # Preprocessing the datasets.
+    # We need to tokenize inputs and targets.
+    column_names = dataset["train"].column_names
+
+    # 6. Get the column names for input/target.
+    if args.image_column is None:
+        image_column = column_names[0]
+        logger.info(f"image column defaulting to {image_column}")
+    else:
+        image_column = args.image_column
+        if image_column not in column_names:
+            raise ValueError(
+                f"`--image_column` value '{args.image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.caption_column is None:
+        caption_column = column_names[1]
+        logger.info(f"caption column defaulting to {caption_column}")
+    else:
+        caption_column = args.caption_column
+        if caption_column not in column_names:
+            raise ValueError(
+                f"`--caption_column` value '{args.caption_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.conditioning_image_column is None:
+        conditioning_image_column = column_names[2]
+        logger.info(f"conditioning image column defaulting to {conditioning_image_column}")
+    else:
+        conditioning_image_column = args.conditioning_image_column
+        if conditioning_image_column not in column_names:
+            raise ValueError(
+                f"`--conditioning_image_column` value '{args.conditioning_image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    def tokenize_captions(examples, is_train=True):
+        captions = []
+        for caption in examples[caption_column]:
+            if random.random() < args.proportion_empty_prompts:
+                captions.append("")
+            elif isinstance(caption, str):
+                captions.append(caption)
+            elif isinstance(caption, (list, np.ndarray)):
+                # take a random caption if there are multiple
+                captions.append(random.choice(caption) if is_train else caption[0])
+            else:
+                raise ValueError(
+                    f"Caption column `{caption_column}` should contain either strings or lists of strings."
+                )
+        inputs = tokenizer(
+            captions, max_length=tokenizer.model_max_length, padding="max_length", truncation=True, return_tensors="pt"
+        )
+        return inputs.input_ids
+
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+        ]
+    )
+
+    def preprocess_train(examples):
+        examples["pixel_values"] = examples[image_column] #images
+        examples["conditioning_pixel_values"] =  examples[conditioning_image_column] #conditioning_images
+        examples["input_ids"] = tokenize_captions(examples)
+
+        return examples
+
+    with accelerator.main_process_first():
+        if args.max_train_samples is not None:
+            dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
+        # Set the training transforms
+        train_dataset = dataset["train"].with_transform(preprocess_train)
+
+    return train_dataset
+
+
+def prepare_mask_and_masked_image(image, mask):
+    image = np.array(image.convert("RGB"))
+    image = image[None].transpose(0, 3, 1, 2)
+    image = torch.from_numpy(image).to(dtype=torch.float32) / 127.5 - 1.0
+
+    mask = np.array(mask.convert("L"))
+    mask = mask.astype(np.float32) / 255.0
+    mask = mask[None, None]
+    mask[mask < 0.5] = 0
+    mask[mask >= 0.5] = 1
+    mask = torch.from_numpy(mask)
+
+    masked_image = image * (mask < 0.5)
+
+    return mask, masked_image
+
+
+def collate_fn(examples):
+
+    pixel_values = [example["pixel_values"].convert("RGB") for example in examples]
+    conditioning_images = [ImageOps.invert(example["conditioning_pixel_values"].convert("RGB")) for example in examples]
+    masks = []
+    masked_images = []
+
+    # Resize and random crop images
+    for i in range(len(pixel_values)):
+        image = np.array(pixel_values[i])
+        mask = np.array(conditioning_images[i])
+        dim_min_ind = np.argmin(image.shape[0:2])
+        dim = [0, 0]
+
+        resize_len = 768.0
+        ratio = resize_len / image.shape[0:2][dim_min_ind]
+        dim[1-dim_min_ind] = int(resize_len)
+        dim[dim_min_ind] = int(ratio * image.shape[0:2][1-dim_min_ind])
+        dim = tuple(dim)
+
+        # resize image
+        image = cv2.resize(image, dim, interpolation = cv2.INTER_AREA)
+        mask = cv2.resize(mask, dim, interpolation = cv2.INTER_AREA)
+        max_x = image.shape[1] - 512
+        max_y = image.shape[0] - 512
+        x = np.random.randint(0, max_x)
+        y = np.random.randint(0, max_y)
+        image = image[y: y + 512, x: x + 512]
+        mask = mask[y: y + 512, x: x + 512]
+
+        # fix for bluish outputs
+        r= np.copy(image[:,:,0])
+        image[:,:,0] = image[:,:,2]
+        image[:,:,2] = r
+        image = Image.fromarray(image)
+        b, g, r = image.split()
+        image = Image.merge("RGB", (r, g, b))
+        pixel_values[i] = image
+
+        conditioning_images[i] = Image.fromarray(mask)
+        mask, masked_image = prepare_mask_and_masked_image(pixel_values[i], conditioning_images[i])
+        masks.append(mask)
+        masked_images.append(masked_image)
+
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+        ]
+    )
+
+    pixel_values = [image_transforms(image) for image in pixel_values]
+    pixel_values = torch.stack(pixel_values)
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_images = [conditioning_image_transforms(image) for image in conditioning_images]
+    conditioning_pixel_values = torch.stack(conditioning_images)
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    input_ids = torch.stack([example["input_ids"] for example in examples])
+
+    masks = torch.stack(masks)
+    masked_images = torch.stack(masked_images)
+
+    return {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "input_ids": input_ids,
+        "masks": masks, "masked_images": masked_images
+    }
+
+
+def main(args):
+    logging_dir = Path(args.output_dir, args.logging_dir)
+
+    accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=logging_dir)
+
+    accelerator = Accelerator(
+        gradient_accumulation_steps=args.gradient_accumulation_steps,
+        mixed_precision=args.mixed_precision,
+        log_with=args.report_to,
+        project_config=accelerator_project_config,
+    )
+
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+    if accelerator.is_local_main_process:
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+        diffusers.utils.logging.set_verbosity_error()
+
+    # If passed along, set the training seed now.
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizer
+    if args.tokenizer_name:
+        tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_name, revision=args.revision, use_fast=False)
+    elif args.pretrained_model_name_or_path:
+        tokenizer = AutoTokenizer.from_pretrained(
+            args.pretrained_model_name_or_path,
+            subfolder="tokenizer",
+            revision=args.revision,
+            use_fast=False,
+        )
+
+    # import correct text encoder class
+    text_encoder_cls = import_model_class_from_model_name_or_path(args.pretrained_model_name_or_path, args.revision)
+
+    # Load scheduler and models
+    noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+    text_encoder = text_encoder_cls.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision
+    )
+    vae = AutoencoderKL.from_pretrained(args.pretrained_model_name_or_path, subfolder="vae", revision=args.revision)
+    unet = UNet2DConditionModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision
+    )
+    if args.controlnet_model_name_or_path:
+        logger.info("Loading existing controlnet weights")
+        controlnet = ControlNetModel.from_pretrained(args.controlnet_model_name_or_path)
+    else:
+        logger.info("Initializing controlnet weights from unet")
+        controlnet = ControlNetModel.from_unet(UNet2DConditionModel.from_pretrained(args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision))
+
+    # `accelerate` 0.16.0 will have better support for customized saving
+    if version.parse(accelerate.__version__) >= version.parse("0.16.0"):
+        # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
+        def save_model_hook(models, weights, output_dir):
+            i = len(weights) - 1
+
+            while len(weights) > 0:
+                weights.pop()
+                model = models[i]
+
+                sub_dir = "controlnet"
+                model.save_pretrained(os.path.join(output_dir, sub_dir))
+
+                i -= 1
+
+        def load_model_hook(models, input_dir):
+            while len(models) > 0:
+                # pop models so that they are not loaded again
+                model = models.pop()
+
+                # load diffusers style into model
+                load_model = ControlNetModel.from_pretrained(input_dir, subfolder="controlnet")
+                model.register_to_config(**load_model.config)
+
+                model.load_state_dict(load_model.state_dict())
+                del load_model
+
+        accelerator.register_save_state_pre_hook(save_model_hook)
+        accelerator.register_load_state_pre_hook(load_model_hook)
+
+    vae.requires_grad_(False)
+    unet.requires_grad_(False)
+    text_encoder.requires_grad_(False)
+    controlnet.train()
+
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            import xformers
+
+            xformers_version = version.parse(xformers.__version__)
+            if xformers_version == version.parse("0.0.16"):
+                logger.warn(
+                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
+                )
+            unet.enable_xformers_memory_efficient_attention()
+            controlnet.enable_xformers_memory_efficient_attention()
+        else:
+            raise ValueError("xformers is not available. Make sure it is installed correctly")
+
+    if args.gradient_checkpointing:
+        controlnet.enable_gradient_checkpointing()
+
+    # Check that all trainable models are in full precision
+    low_precision_error_string = (
+        " Please make sure to always have all model weights in full float32 precision when starting training - even if"
+        " doing mixed precision training, copy of the weights should still be float32."
+    )
+
+    if accelerator.unwrap_model(controlnet).dtype != torch.float32:
+        raise ValueError(
+            f"Controlnet loaded as datatype {accelerator.unwrap_model(controlnet).dtype}. {low_precision_error_string}"
+        )
+
+    # Enable TF32 for faster training on Ampere GPUs,
+    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
+    if args.allow_tf32:
+        torch.backends.cuda.matmul.allow_tf32 = True
+
+    if args.scale_lr:
+        args.learning_rate = (
+            args.learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
+    if args.use_8bit_adam:
+        try:
+            import bitsandbytes as bnb
+        except ImportError:
+            raise ImportError(
+                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
+            )
+
+        optimizer_class = bnb.optim.AdamW8bit
+    else:
+        optimizer_class = torch.optim.AdamW
+
+    # Optimizer creation
+    params_to_optimize = controlnet.parameters()
+    optimizer = optimizer_class(
+        params_to_optimize,
+        lr=args.learning_rate,
+        betas=(args.adam_beta1, args.adam_beta2),
+        weight_decay=args.adam_weight_decay,
+        eps=args.adam_epsilon,
+    )
+
+    train_dataset = make_train_dataset(args, tokenizer, accelerator)
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        shuffle=True,
+        collate_fn=collate_fn,
+        batch_size=args.train_batch_size,
+        num_workers=args.dataloader_num_workers,
+    )
+
+    # Scheduler and math around the number of training steps.
+    overrode_max_train_steps = False
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        overrode_max_train_steps = True
+
+    lr_scheduler = get_scheduler(
+        args.lr_scheduler,
+        optimizer=optimizer,
+        num_warmup_steps=args.lr_warmup_steps * accelerator.num_processes,
+        num_training_steps=args.max_train_steps * accelerator.num_processes,
+        num_cycles=args.lr_num_cycles,
+        power=args.lr_power,
+    )
+
+    # Prepare everything with our `accelerator`.
+    controlnet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+        controlnet, optimizer, train_dataloader, lr_scheduler
+    )
+
+    # For mixed precision training we cast the text_encoder and vae weights to half-precision
+    # as these models are only used for inference, keeping weights in full precision is not required.
+    weight_dtype = torch.float32
+    if accelerator.mixed_precision == "fp16":
+        weight_dtype = torch.float16
+    elif accelerator.mixed_precision == "bf16":
+        weight_dtype = torch.bfloat16
+
+    # Move vae, unet and text_encoder to device and cast to weight_dtype
+    vae.to(accelerator.device, dtype=weight_dtype)
+    unet.to(accelerator.device, dtype=weight_dtype)
+    text_encoder.to(accelerator.device, dtype=weight_dtype)
+
+    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if overrode_max_train_steps:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+    # Afterwards we recalculate our number of training epochs
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    # We need to initialize the trackers we use, and also store our configuration.
+    # The trackers initializes automatically on the main process.
+    if accelerator.is_main_process:
+        tracker_config = dict(vars(args))
+
+        # tensorboard cannot handle list types for config
+        tracker_config.pop("validation_prompt")
+        tracker_config.pop("validation_image")
+        tracker_config.pop("validation_inpainting_image")
+
+        accelerator.init_trackers(args.tracker_project_name, config=tracker_config)
+
+    # Train!
+    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    global_step = 0
+    first_epoch = 0
+
+    # Potentially load in the weights and states from a previous save
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint != "latest":
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            # Get the most recent checkpoint
+            dirs = os.listdir(args.output_dir)
+            dirs = [d for d in dirs if d.startswith("checkpoint")]
+            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
+            path = dirs[-1] if len(dirs) > 0 else None
+
+        if path is None:
+            accelerator.print(
+                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
+            )
+            args.resume_from_checkpoint = None
+            initial_global_step = 0
+        else:
+            accelerator.print(f"Resuming from checkpoint {path}")
+            accelerator.load_state(os.path.join(args.output_dir, path))
+            global_step = int(path.split("-")[1])
+
+            initial_global_step = global_step
+            first_epoch = global_step // num_update_steps_per_epoch
+    else:
+        initial_global_step = 0
+
+    progress_bar = tqdm(
+        range(0, args.max_train_steps),
+        initial=initial_global_step,
+        desc="Steps",
+        # Only show the progress bar once on each machine.
+        disable=not accelerator.is_local_main_process,
+    )
+
+    image_logs = None
+    for epoch in range(first_epoch, args.num_train_epochs):
+        for param_group in optimizer.param_groups:
+            param_group['lr'] = 0.00001
+        for step, batch in enumerate(train_dataloader):
+            with accelerator.accumulate(controlnet):
+                # Convert images to latent space
+                latents = vae.encode(batch["pixel_values"].to(dtype=weight_dtype)).latent_dist.sample()
+                latents = latents * vae.config.scaling_factor
+                # Convert masked images to latent space
+                masked_latents = vae.encode(
+                    batch["masked_images"].reshape(batch["pixel_values"].shape).to(dtype=weight_dtype)
+                ).latent_dist.sample()
+                masked_latents = masked_latents * vae.config.scaling_factor
+                masks = batch["masks"]
+                # resize the mask to latents shape as we concatenate the mask to the latents
+                mask = torch.stack(
+                    [
+                        torch.nn.functional.interpolate(mask, size=(args.resolution // 8, args.resolution // 8))
+                        for mask in masks
+                    ]
+                )
+                mask = mask.reshape(-1, 1, args.resolution // 8, args.resolution // 8)
+                # Sample noise that we'll add to the latents
+                noise = torch.randn_like(latents)
+                bsz = latents.shape[0]
+                # Sample a random timestep for each image
+                timesteps = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)
+                timesteps = timesteps.long()
+
+                # Add noise to the latents according to the noise magnitude at each timestep
+                # (this is the forward diffusion process)
+                noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)
+
+                # concatenate the noised latents with the mask and the masked latents
+                latent_model_input = torch.cat([noisy_latents, mask, masked_latents], dim=1)
+                # Get the text embedding for conditioning
+                encoder_hidden_states = text_encoder(batch["input_ids"])[0]
+
+                controlnet_image = batch["conditioning_pixel_values"].to(dtype=weight_dtype)
+
+                down_block_res_samples, mid_block_res_sample = controlnet(
+                    latent_model_input,
+                    timesteps,
+                    encoder_hidden_states=encoder_hidden_states,
+                    controlnet_cond=controlnet_image,
+                    return_dict=False,
+                )
+
+                # Predict the noise residual
+                model_pred = unet(
+                    latent_model_input.to(dtype=weight_dtype),
+                    timesteps.to(dtype=weight_dtype),
+                    encoder_hidden_states=encoder_hidden_states.to(dtype=weight_dtype),
+                    down_block_additional_residuals=[
+                        sample.to(dtype=weight_dtype) for sample in down_block_res_samples
+                    ],
+                    mid_block_additional_residual=mid_block_res_sample.to(dtype=weight_dtype),
+                ).sample
+
+                # Get the target for loss depending on the prediction type
+                if noise_scheduler.config.prediction_type == "epsilon":
+                    target = noise
+                elif noise_scheduler.config.prediction_type == "v_prediction":
+                    target = noise_scheduler.get_velocity(latents, noise, timesteps)
+                else:
+                    raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
+                loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
+
+                accelerator.backward(loss)
+                if accelerator.sync_gradients:
+                    params_to_clip = controlnet.parameters()
+                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad(set_to_none=args.set_grads_to_none)
+
+            # Checks if the accelerator has performed an optimization step behind the scenes
+            if accelerator.sync_gradients:
+                progress_bar.update(1)
+                global_step += 1
+
+                if accelerator.is_main_process:
+                    if global_step % args.checkpointing_steps == 0:
+                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
+                        if args.checkpoints_total_limit is not None:
+                            checkpoints = os.listdir(args.output_dir)
+                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
+                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))
+
+                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
+                            if len(checkpoints) >= args.checkpoints_total_limit:
+                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
+                                removing_checkpoints = checkpoints[0:num_to_remove]
+
+                                logger.info(
+                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
+                                )
+                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")
+
+                                for removing_checkpoint in removing_checkpoints:
+                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
+                                    shutil.rmtree(removing_checkpoint)
+
+                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
+                        accelerator.save_state(save_path)
+                        logger.info(f"Saved state to {save_path}")
+
+                    if args.validation_prompt is not None and global_step % args.validation_steps == 0:
+                        image_logs = log_validation(
+                            vae,
+                            text_encoder,
+                            tokenizer,
+                            unet,
+                            controlnet,
+                            args,
+                            accelerator,
+                            weight_dtype,
+                            global_step,
+                        )
+
+            logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
+            progress_bar.set_postfix(**logs)
+            accelerator.log(logs, step=global_step)
+
+            if global_step >= args.max_train_steps:
+                break
+
+    # Create the pipeline using using the trained modules and save it.
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        controlnet = accelerator.unwrap_model(controlnet)
+        controlnet.save_pretrained(args.output_dir)
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                image_logs=image_logs,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)

train_controlnet_inpaint.py

@@ -0,0 +1,1244 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2024, Yahoo Research
+#
+# 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 argparse
+import logging
+import math
+import os
+import random
+import shutil
+from pathlib import Path
+import cv2
+from PIL import Image, ImageOps
+import accelerate
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+import transformers
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import ProjectConfiguration, set_seed
+from datasets import load_dataset
+from huggingface_hub import create_repo, upload_folder
+from packaging import version
+from PIL import Image
+from torchvision import transforms
+from tqdm.auto import tqdm
+from transformers import AutoTokenizer, PretrainedConfig
+
+import diffusers
+from diffusers import (
+    AutoencoderKL,
+    ControlNetModel,
+    DDPMScheduler,
+    StableDiffusionControlNetPipeline,
+    UNet2DConditionModel,
+    UniPCMultistepScheduler,
+)
+from diffusers.optimization import get_scheduler
+from diffusers.utils import check_min_version, is_wandb_available
+from diffusers.utils.import_utils import is_xformers_available
+
+
+if is_wandb_available():
+    import wandb
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.20.0.dev0")
+
+logger = get_logger(__name__)
+
+
+def image_grid(imgs, rows, cols):
+    assert len(imgs) == rows * cols
+
+    w, h = imgs[0].size
+    grid = Image.new("RGB", size=(cols * w, rows * h))
+
+    for i, img in enumerate(imgs):
+        grid.paste(img, box=(i % cols * w, i // cols * h))
+    return grid
+
+
+def log_validation(vae, text_encoder, tokenizer, unet, controlnet, args, accelerator, weight_dtype, step):
+    logger.info("Running validation... ")
+
+    controlnet = accelerator.unwrap_model(controlnet)
+
+    pipeline = StableDiffusionControlNetPipeline.from_pretrained(
+        args.pretrained_model_name_or_path,
+        vae=vae,
+        text_encoder=text_encoder,
+        tokenizer=tokenizer,
+        unet=unet,
+        controlnet=controlnet,
+        safety_checker=None,
+        revision=args.revision,
+        torch_dtype=weight_dtype,
+    )
+    pipeline.scheduler = UniPCMultistepScheduler.from_config(pipeline.scheduler.config)
+    pipeline = pipeline.to(accelerator.device)
+    pipeline.set_progress_bar_config(disable=True)
+
+    if args.enable_xformers_memory_efficient_attention:
+        pipeline.enable_xformers_memory_efficient_attention()
+
+    if args.seed is None:
+        generator = None
+    else:
+        generator = torch.Generator(device=accelerator.device).manual_seed(args.seed)
+
+    if len(args.validation_image) == len(args.validation_prompt):
+        validation_images = args.validation_image
+        validation_inpainting_images = args.validation_inpainting_image
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_image) == 1:
+        validation_images = args.validation_image * len(args.validation_prompt)
+        validation_inpainting_images = args.validation_inpainting_image * len(args.validation_prompt)
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_prompt) == 1:
+        validation_images = args.validation_image
+        validation_inpainting_images = args.validation_inpainting_image
+        validation_prompts = args.validation_prompt * len(args.validation_image)
+    else:
+        raise ValueError(
+            "number of `args.validation_image` and `args.validation_prompt` should be checked in `parse_args`"
+        )
+
+    image_logs = []
+
+    for validation_prompt, validation_image, validation_inpainting_image in zip(validation_prompts, validation_images, validation_inpainting_images):
+        mask = Image.open(validation_image)
+        mask = resize_with_padding(mask, (512,512))
+
+        inpainting_image  = Image.open(validation_inpainting_image).convert("RGB")
+        inpainting_image = resize_with_padding(inpainting_image, (512,512))
+
+        validation_image = Image.composite(inpainting_image, mask, mask.convert('L')).convert('RGB')
+        images = []
+        for _ in range(args.num_validation_images):
+            with torch.autocast("cuda"):
+                image = pipeline(
+                    validation_prompt, validation_image, num_inference_steps=20, generator=generator
+                ).images[0]
+            images.append(image)
+
+        image_logs.append(
+            {"validation_image": validation_image, "images": images, "validation_prompt": validation_prompt}
+        )
+
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images = []
+
+                formatted_images.append(np.asarray(validation_image))
+
+                for image in images:
+                    formatted_images.append(np.asarray(image))
+                formatted_images = np.stack(formatted_images)
+
+                tracker.writer.add_images(validation_prompt, formatted_images, step, dataformats="NHWC")
+        elif tracker.name == "wandb":
+            formatted_images = []
+
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images.append(wandb.Image(validation_image, caption="Controlnet conditioning"))
+
+                for image in images:
+                    image = wandb.Image(image, caption=validation_prompt)
+                    formatted_images.append(image)
+
+            tracker.log({"validation": formatted_images})
+        else:
+            logger.warn(f"image logging not implemented for {tracker.name}")
+
+        return image_logs
+
+
+def import_model_class_from_model_name_or_path(pretrained_model_name_or_path: str, revision: str):
+    text_encoder_config = PretrainedConfig.from_pretrained(
+        pretrained_model_name_or_path,
+        subfolder="text_encoder",
+        revision=revision,
+    )
+    model_class = text_encoder_config.architectures[0]
+
+    if model_class == "CLIPTextModel":
+        from transformers import CLIPTextModel
+
+        return CLIPTextModel
+    elif model_class == "RobertaSeriesModelWithTransformation":
+        from diffusers.pipelines.alt_diffusion.modeling_roberta_series import RobertaSeriesModelWithTransformation
+
+        return RobertaSeriesModelWithTransformation
+    else:
+        raise ValueError(f"{model_class} is not supported.")
+
+
+def save_model_card(repo_id: str, image_logs=None, base_model=str, repo_folder=None):
+    img_str = ""
+    if image_logs is not None:
+        img_str = "You can find some example images below.\n"
+        for i, log in enumerate(image_logs):
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+            validation_image.save(os.path.join(repo_folder, "image_control.png"))
+            img_str += f"prompt: {validation_prompt}\n"
+            images = [validation_image] + images
+            image_grid(images, 1, len(images)).save(os.path.join(repo_folder, f"images_{i}.png"))
+            img_str += f"![images_{i})](./images_{i}.png)\n"
+
+    yaml = f"""
+---
+license: creativeml-openrail-m
+base_model: {base_model}
+tags:
+- stable-diffusion
+- stable-diffusion-diffusers
+- text-to-image
+- diffusers
+- controlnet
+inference: true
+---
+    """
+    model_card = f"""
+# controlnet-{repo_id}
+
+These are controlnet weights trained on {base_model} with new type of conditioning.
+{img_str}
+"""
+    with open(os.path.join(repo_folder, "README.md"), "w") as f:
+        f.write(yaml + model_card)
+
+
+def parse_args(input_args=None):
+    parser = argparse.ArgumentParser(description="Simple example of a ControlNet training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        default=None,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--controlnet_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to pretrained controlnet model or model identifier from huggingface.co/models."
+        " If not specified controlnet weights are initialized from unet.",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        required=False,
+        help=(
+            "Revision of pretrained model identifier from huggingface.co/models. Trainable model components should be"
+            " float32 precision."
+        ),
+    )
+    parser.add_argument(
+        "--tokenizer_name",
+        type=str,
+        default=None,
+        help="Pretrained tokenizer name or path if not the same as model_name",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="controlnet-model",
+        help="The output directory where the model predictions and checkpoints will be written.",
+    )
+    parser.add_argument(
+        "--cache_dir",
+        type=str,
+        default=None,
+        help="The directory where the downloaded models and datasets will be stored.",
+    )
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=1)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=500,
+        help=(
+            "Save a checkpoint of the training state every X updates. Checkpoints can be used for resuming training via `--resume_from_checkpoint`. "
+            "In the case that the checkpoint is better than the final trained model, the checkpoint can also be used for inference."
+            "Using a checkpoint for inference requires separate loading of the original pipeline and the individual checkpointed model components."
+            "See https://huggingface.co/docs/diffusers/main/en/training/dreambooth#performing-inference-using-a-saved-checkpoint for step by step"
+            "instructions."
+        ),
+    )
+    parser.add_argument(
+        "--checkpoints_total_limit",
+        type=int,
+        default=None,
+        help=("Max number of checkpoints to store."),
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help=(
+            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
+            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
+        ),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--gradient_checkpointing",
+        action="store_true",
+        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=5e-6,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        default=False,
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument(
+        "--lr_num_cycles",
+        type=int,
+        default=1,
+        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
+    )
+    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
+    parser.add_argument(
+        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
+    )
+    parser.add_argument(
+        "--dataloader_num_workers",
+        type=int,
+        default=0,
+        help=(
+            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
+        ),
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_dir",
+        type=str,
+        default="logs",
+        help=(
+            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
+            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
+        ),
+    )
+    parser.add_argument(
+        "--allow_tf32",
+        action="store_true",
+        help=(
+            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
+            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
+        ),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="tensorboard",
+        help=(
+            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
+            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
+        ),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
+            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
+        ),
+    )
+    parser.add_argument(
+        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
+    )
+    parser.add_argument(
+        "--set_grads_to_none",
+        action="store_true",
+        help=(
+            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
+            " behaviors, so disable this argument if it causes any problems. More info:"
+            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
+        ),
+    )
+    parser.add_argument(
+        "--dataset_name",
+        type=str,
+        default=None,
+        help=(
+            "The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
+            " dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
+            " or to a folder containing files that 🤗 Datasets can understand."
+        ),
+    )
+    parser.add_argument(
+        "--dataset_config_name",
+        type=str,
+        default=None,
+        help="The config of the Dataset, leave as None if there's only one config.",
+    )
+    parser.add_argument(
+        "--train_data_dir",
+        type=str,
+        default=None,
+        help=(
+            "A folder containing the training data. Folder contents must follow the structure described in"
+            " https://huggingface.co/docs/datasets/image_dataset#imagefolder. In particular, a `metadata.jsonl` file"
+            " must exist to provide the captions for the images. Ignored if `dataset_name` is specified."
+        ),
+    )
+    parser.add_argument(
+        "--image_column", type=str, default="image", help="The column of the dataset containing the target image."
+    )
+    parser.add_argument(
+        "--conditioning_image_column",
+        type=str,
+        default="conditioning_image",
+        help="The column of the dataset containing the controlnet conditioning image.",
+    )
+    parser.add_argument(
+        "--caption_column",
+        type=str,
+        default="text",
+        help="The column of the dataset containing a caption or a list of captions.",
+    )
+    parser.add_argument(
+        "--max_train_samples",
+        type=int,
+        default=None,
+        help=(
+            "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set."
+        ),
+    )
+    parser.add_argument(
+        "--proportion_empty_prompts",
+        type=float,
+        default=0,
+        help="Proportion of image prompts to be replaced with empty strings. Defaults to 0 (no prompt replacement).",
+    )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of prompts evaluated every `--validation_steps` and logged to `--report_to`."
+            " Provide either a matching number of `--validation_image`s, a single `--validation_image`"
+            " to be used with all prompts, or a single prompt that will be used with all `--validation_image`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_inpainting_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=4,
+        help="Number of images to be generated for each `--validation_image`, `--validation_prompt` pair",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` multiple times: `args.num_validation_images`"
+            " and logging the images."
+        ),
+    )
+    parser.add_argument(
+        "--tracker_project_name",
+        type=str,
+        default="train_controlnet",
+        help=(
+            "The `project_name` argument passed to Accelerator.init_trackers for"
+            " more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
+        ),
+    )
+
+    if input_args is not None:
+        args = parser.parse_args(input_args)
+    else:
+        args = parser.parse_args()
+
+    if args.dataset_name is None and args.train_data_dir is None:
+        raise ValueError("Specify either `--dataset_name` or `--train_data_dir`")
+
+    if args.dataset_name is not None and args.train_data_dir is not None:
+        raise ValueError("Specify only one of `--dataset_name` or `--train_data_dir`")
+
+    if args.proportion_empty_prompts < 0 or args.proportion_empty_prompts > 1:
+        raise ValueError("`--proportion_empty_prompts` must be in the range [0, 1].")
+
+    if args.validation_prompt is not None and args.validation_image is None:
+        raise ValueError("`--validation_image` must be set if `--validation_prompt` is set")
+
+    if args.validation_prompt is None and args.validation_image is not None:
+        raise ValueError("`--validation_prompt` must be set if `--validation_image` is set")
+
+    if (
+        args.validation_image is not None
+        and args.validation_prompt is not None
+        and len(args.validation_image) != 1
+        and len(args.validation_prompt) != 1
+        and len(args.validation_image) != len(args.validation_prompt)
+    ):
+        raise ValueError(
+            "Must provide either 1 `--validation_image`, 1 `--validation_prompt`,"
+            " or the same number of `--validation_prompt`s and `--validation_image`s"
+        )
+
+    if args.resolution % 8 != 0:
+        raise ValueError(
+            "`--resolution` must be divisible by 8 for consistently sized encoded images between the VAE and the controlnet encoder."
+        )
+
+    return args
+
+
+def make_train_dataset(args, tokenizer, accelerator):
+    # Get the datasets: you can either provide your own training and evaluation files (see below)
+    # or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
+
+    # In distributed training, the load_dataset function guarantees that only one local process can concurrently
+    # download the dataset.
+    if args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        dataset = load_dataset(
+            args.dataset_name,
+            args.dataset_config_name,
+            cache_dir=args.cache_dir,
+        )
+    else:
+        if args.train_data_dir is not None:
+            dataset = load_dataset(
+                args.train_data_dir,
+                cache_dir=args.cache_dir,
+            )
+        # See more about loading custom images at
+        # https://huggingface.co/docs/datasets/v2.0.0/en/dataset_script
+
+    # Preprocessing the datasets.
+    # We need to tokenize inputs and targets.
+    column_names = dataset["train"].column_names
+
+    # 6. Get the column names for input/target.
+    if args.image_column is None:
+        image_column = column_names[0]
+        logger.info(f"image column defaulting to {image_column}")
+    else:
+        image_column = args.image_column
+        if image_column not in column_names:
+            raise ValueError(
+                f"`--image_column` value '{args.image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.caption_column is None:
+        caption_column = column_names[1]
+        logger.info(f"caption column defaulting to {caption_column}")
+    else:
+        caption_column = args.caption_column
+        if caption_column not in column_names:
+            raise ValueError(
+                f"`--caption_column` value '{args.caption_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.conditioning_image_column is None:
+        conditioning_image_column = column_names[2]
+        logger.info(f"conditioning image column defaulting to {conditioning_image_column}")
+    else:
+        conditioning_image_column = args.conditioning_image_column
+        if conditioning_image_column not in column_names:
+            raise ValueError(
+                f"`--conditioning_image_column` value '{args.conditioning_image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    def tokenize_captions(examples, is_train=True):
+        captions = []
+        for caption in examples[caption_column]:
+            if random.random() < args.proportion_empty_prompts:
+                captions.append("")
+            elif isinstance(caption, str):
+                captions.append(caption)
+            elif isinstance(caption, (list, np.ndarray)):
+                # take a random caption if there are multiple
+                captions.append(random.choice(caption) if is_train else caption[0])
+            else:
+                raise ValueError(
+                    f"Caption column `{caption_column}` should contain either strings or lists of strings."
+                )
+        inputs = tokenizer(
+            captions, max_length=tokenizer.model_max_length, padding="max_length", truncation=True, return_tensors="pt"
+        )
+        return inputs.input_ids
+
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+        ]
+    )
+
+    def preprocess_train(examples):
+        examples["pixel_values"] = examples[image_column] #images
+        examples["conditioning_pixel_values"] =  examples[conditioning_image_column] #conditioning_images
+        examples["input_ids"] = tokenize_captions(examples)
+
+        return examples
+
+    with accelerator.main_process_first():
+        if args.max_train_samples is not None:
+            dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
+        # Set the training transforms
+        train_dataset = dataset["train"].with_transform(preprocess_train)
+
+    return train_dataset
+
+
+def resize_with_padding(img, expected_size):
+    img.thumbnail((expected_size[0], expected_size[1]))
+    # print(img.size)
+    delta_width = expected_size[0] - img.size[0]
+    delta_height = expected_size[1] - img.size[1]
+    pad_width = delta_width // 2
+    pad_height = delta_height // 2
+    padding = (pad_width, pad_height, delta_width - pad_width, delta_height - pad_height)
+    return ImageOps.expand(img, padding)
+
+def prepare_mask_and_masked_image(image, mask):
+    image = np.array(image.convert("RGB"))
+    image = image[None].transpose(0, 3, 1, 2)
+    image = torch.from_numpy(image).to(dtype=torch.float32) / 127.5 - 1.0
+
+    mask = np.array(mask.convert("L"))
+    mask = mask.astype(np.float32) / 255.0
+    mask = mask[None, None]
+    mask[mask < 0.5] = 0
+    mask[mask >= 0.5] = 1
+    #mask = torch.from_numpy(mask)
+
+    masked_image = image * (mask < 0.5)
+
+    return mask, masked_image
+
+def collate_fn(examples):
+    pixel_values = [example["pixel_values"].convert("RGB") for example in examples]
+    conditioning_images = [example["conditioning_pixel_values"].convert("RGB") for example in examples]
+    masks = []
+    masked_images = []
+
+    # Resize and random crop images
+    for i in range(len(pixel_values)):
+        image = np.array(pixel_values[i])
+        mask = np.array(conditioning_images[i])
+        dim_min_ind = np.argmin(image.shape[0:2])
+        dim = [0, 0]
+
+        resize_len = 768.0
+        ratio = resize_len / image.shape[0:2][dim_min_ind]
+        dim[1-dim_min_ind] = int(resize_len)
+        dim[dim_min_ind] = int(ratio * image.shape[0:2][1-dim_min_ind])
+        dim = tuple(dim)
+
+        # resize image
+        image = cv2.resize(image, dim, interpolation = cv2.INTER_AREA)
+        mask = cv2.resize(mask, dim, interpolation = cv2.INTER_AREA)
+        max_x = image.shape[1] - 512
+        max_y = image.shape[0] - 512
+        x = np.random.randint(0, max_x)
+        y = np.random.randint(0, max_y)
+        image = image[y: y + 512, x: x + 512]
+        mask = mask[y: y + 512, x: x + 512]
+
+        # fix for bluish outputs
+        r = np.copy(image[:,:,0])
+        image[:,:,0] = image[:,:,2]
+        image[:,:,2] = r
+        image = Image.fromarray(image)
+        b, g, r = image.split()
+        image = Image.merge("RGB", (r, g, b))
+        pixel_values[i] = image
+        conditioning_images[i] = Image.composite(image, Image.fromarray(mask), Image.fromarray(mask).convert('L')).convert('RGB')
+
+
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5])
+        ]
+    )
+
+    pixel_values = [image_transforms(image) for image in pixel_values]
+    pixel_values = torch.stack(pixel_values)
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_images = [conditioning_image_transforms(image) for image in conditioning_images]
+    conditioning_pixel_values = torch.stack(conditioning_images)
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    input_ids = torch.stack([example["input_ids"] for example in examples])
+
+ #   masks = torch.stack(masks)
+ #   masked_images = torch.stack(masked_images)
+
+    return {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "input_ids": input_ids,
+  #      "masks": masks, "masked_images": masked_images
+    }
+
+#    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+#    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+#    conditioning_pixel_values = torch.stack([example["conditioning_pixel_values"] for example in examples])
+#    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+ #   input_ids = torch.stack([example["input_ids"] for example in examples])
+
+  #  return {
+  #      "pixel_values": pixel_values,
+  #      "conditioning_pixel_values": conditioning_pixel_values,
+  #      "input_ids": input_ids,
+  #  }
+
+
+def main(args):
+    logging_dir = Path(args.output_dir, args.logging_dir)
+
+    accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=logging_dir)
+
+    accelerator = Accelerator(
+        gradient_accumulation_steps=args.gradient_accumulation_steps,
+        mixed_precision=args.mixed_precision,
+        log_with=args.report_to,
+        project_config=accelerator_project_config,
+    )
+
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+    if accelerator.is_local_main_process:
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+        diffusers.utils.logging.set_verbosity_error()
+
+    # If passed along, set the training seed now.
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizer
+    if args.tokenizer_name:
+        tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_name, revision=args.revision, use_fast=False)
+    elif args.pretrained_model_name_or_path:
+        tokenizer = AutoTokenizer.from_pretrained(
+            args.pretrained_model_name_or_path,
+            subfolder="tokenizer",
+            revision=args.revision,
+            use_fast=False,
+        )
+
+    # import correct text encoder class
+    text_encoder_cls = import_model_class_from_model_name_or_path(args.pretrained_model_name_or_path, args.revision)
+
+    # Load scheduler and models
+    noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+    text_encoder = text_encoder_cls.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision
+    )
+    vae = AutoencoderKL.from_pretrained(args.pretrained_model_name_or_path, subfolder="vae", revision=args.revision)
+    unet = UNet2DConditionModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision
+    )
+
+    if args.controlnet_model_name_or_path:
+        logger.info("Loading existing controlnet weights")
+        controlnet = ControlNetModel.from_pretrained(args.controlnet_model_name_or_path)
+    else:
+        logger.info("Initializing controlnet weights from unet")
+        controlnet = ControlNetModel.from_unet(unet)
+
+    # `accelerate` 0.16.0 will have better support for customized saving
+    if version.parse(accelerate.__version__) >= version.parse("0.16.0"):
+        # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
+        def save_model_hook(models, weights, output_dir):
+            i = len(weights) - 1
+
+            while len(weights) > 0:
+                weights.pop()
+                model = models[i]
+
+                sub_dir = "controlnet"
+                model.save_pretrained(os.path.join(output_dir, sub_dir))
+
+                i -= 1
+
+        def load_model_hook(models, input_dir):
+            while len(models) > 0:
+                # pop models so that they are not loaded again
+                model = models.pop()
+
+                # load diffusers style into model
+                load_model = ControlNetModel.from_pretrained(input_dir, subfolder="controlnet")
+                model.register_to_config(**load_model.config)
+
+                model.load_state_dict(load_model.state_dict())
+                del load_model
+
+        accelerator.register_save_state_pre_hook(save_model_hook)
+        accelerator.register_load_state_pre_hook(load_model_hook)
+
+    vae.requires_grad_(False)
+    unet.requires_grad_(False)
+    text_encoder.requires_grad_(False)
+    controlnet.train()
+
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            import xformers
+
+            xformers_version = version.parse(xformers.__version__)
+            if xformers_version == version.parse("0.0.16"):
+                logger.warn(
+                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
+                )
+            unet.enable_xformers_memory_efficient_attention()
+            controlnet.enable_xformers_memory_efficient_attention()
+        else:
+            raise ValueError("xformers is not available. Make sure it is installed correctly")
+
+    if args.gradient_checkpointing:
+        controlnet.enable_gradient_checkpointing()
+
+    # Check that all trainable models are in full precision
+    low_precision_error_string = (
+        " Please make sure to always have all model weights in full float32 precision when starting training - even if"
+        " doing mixed precision training, copy of the weights should still be float32."
+    )
+
+    if accelerator.unwrap_model(controlnet).dtype != torch.float32:
+        raise ValueError(
+            f"Controlnet loaded as datatype {accelerator.unwrap_model(controlnet).dtype}. {low_precision_error_string}"
+        )
+
+    # Enable TF32 for faster training on Ampere GPUs,
+    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
+    if args.allow_tf32:
+        torch.backends.cuda.matmul.allow_tf32 = True
+
+    if args.scale_lr:
+        args.learning_rate = (
+            args.learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
+    if args.use_8bit_adam:
+        try:
+            import bitsandbytes as bnb
+        except ImportError:
+            raise ImportError(
+                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
+            )
+
+        optimizer_class = bnb.optim.AdamW8bit
+    else:
+        optimizer_class = torch.optim.AdamW
+
+    # Optimizer creation
+    params_to_optimize = controlnet.parameters()
+    optimizer = optimizer_class(
+        params_to_optimize,
+        lr=args.learning_rate,
+        betas=(args.adam_beta1, args.adam_beta2),
+        weight_decay=args.adam_weight_decay,
+        eps=args.adam_epsilon,
+    )
+
+    train_dataset = make_train_dataset(args, tokenizer, accelerator)
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        shuffle=True,
+        collate_fn=collate_fn,
+        batch_size=args.train_batch_size,
+        num_workers=args.dataloader_num_workers,
+    )
+
+    # Scheduler and math around the number of training steps.
+    overrode_max_train_steps = False
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        overrode_max_train_steps = True
+
+    lr_scheduler = get_scheduler(
+        args.lr_scheduler,
+        optimizer=optimizer,
+        num_warmup_steps=args.lr_warmup_steps * accelerator.num_processes,
+        num_training_steps=args.max_train_steps * accelerator.num_processes,
+        num_cycles=args.lr_num_cycles,
+        power=args.lr_power,
+    )
+
+    # Prepare everything with our `accelerator`.
+    controlnet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+        controlnet, optimizer, train_dataloader, lr_scheduler
+    )
+
+    # For mixed precision training we cast the text_encoder and vae weights to half-precision
+    # as these models are only used for inference, keeping weights in full precision is not required.
+    weight_dtype = torch.float32
+    if accelerator.mixed_precision == "fp16":
+        weight_dtype = torch.float16
+    elif accelerator.mixed_precision == "bf16":
+        weight_dtype = torch.bfloat16
+
+    # Move vae, unet and text_encoder to device and cast to weight_dtype
+    vae.to(accelerator.device, dtype=weight_dtype)
+    unet.to(accelerator.device, dtype=weight_dtype)
+    text_encoder.to(accelerator.device, dtype=weight_dtype)
+
+    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if overrode_max_train_steps:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+    # Afterwards we recalculate our number of training epochs
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    # We need to initialize the trackers we use, and also store our configuration.
+    # The trackers initializes automatically on the main process.
+    if accelerator.is_main_process:
+        tracker_config = dict(vars(args))
+
+        # tensorboard cannot handle list types for config
+        tracker_config.pop("validation_prompt")
+        tracker_config.pop("validation_image")
+        tracker_config.pop("validation_inpainting_image")
+
+        accelerator.init_trackers(args.tracker_project_name, config=tracker_config)
+
+    # Train!
+    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    global_step = 0
+    first_epoch = 0
+
+    # Potentially load in the weights and states from a previous save
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint != "latest":
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            # Get the most recent checkpoint
+            dirs = os.listdir(args.output_dir)
+            dirs = [d for d in dirs if d.startswith("checkpoint")]
+            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
+            path = dirs[-1] if len(dirs) > 0 else None
+
+        if path is None:
+            accelerator.print(
+                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
+            )
+            args.resume_from_checkpoint = None
+            initial_global_step = 0
+        else:
+            accelerator.print(f"Resuming from checkpoint {path}")
+            accelerator.load_state(os.path.join(args.output_dir, path))
+            global_step = int(path.split("-")[1])
+
+            initial_global_step = global_step
+            first_epoch = global_step // num_update_steps_per_epoch
+    else:
+        initial_global_step = 0
+
+    progress_bar = tqdm(
+        range(0, args.max_train_steps),
+        initial=initial_global_step,
+        desc="Steps",
+        # Only show the progress bar once on each machine.
+        disable=not accelerator.is_local_main_process,
+    )
+
+    image_logs = None
+    for epoch in range(first_epoch, args.num_train_epochs):
+        for step, batch in enumerate(train_dataloader):
+            with accelerator.accumulate(controlnet):
+                # Convert images to latent space
+                latents = vae.encode(batch["pixel_values"].to(dtype=weight_dtype)).latent_dist.sample()
+                latents = latents * vae.config.scaling_factor
+
+                # Sample noise that we'll add to the latents
+                noise = torch.randn_like(latents)
+                bsz = latents.shape[0]
+                # Sample a random timestep for each image
+                timesteps = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)
+                timesteps = timesteps.long()
+
+                # Add noise to the latents according to the noise magnitude at each timestep
+                # (this is the forward diffusion process)
+                noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)
+
+                # Get the text embedding for conditioning
+                encoder_hidden_states = text_encoder(batch["input_ids"])[0]
+
+                controlnet_image = batch["conditioning_pixel_values"].to(dtype=weight_dtype)
+
+                down_block_res_samples, mid_block_res_sample = controlnet(
+                    noisy_latents,
+                    timesteps,
+                    encoder_hidden_states=encoder_hidden_states,
+                    controlnet_cond=controlnet_image,
+                    return_dict=False,
+                )
+
+                # Predict the noise residual
+                model_pred = unet(
+                    noisy_latents,
+                    timesteps,
+                    encoder_hidden_states=encoder_hidden_states,
+                    down_block_additional_residuals=[
+                        sample.to(dtype=weight_dtype) for sample in down_block_res_samples
+                    ],
+                    mid_block_additional_residual=mid_block_res_sample.to(dtype=weight_dtype),
+                ).sample
+
+                # Get the target for loss depending on the prediction type
+                if noise_scheduler.config.prediction_type == "epsilon":
+                    target = noise
+                elif noise_scheduler.config.prediction_type == "v_prediction":
+                    target = noise_scheduler.get_velocity(latents, noise, timesteps)
+                else:
+                    raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
+                loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
+
+                accelerator.backward(loss)
+                if accelerator.sync_gradients:
+                    params_to_clip = controlnet.parameters()
+                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad(set_to_none=args.set_grads_to_none)
+
+            # Checks if the accelerator has performed an optimization step behind the scenes
+            if accelerator.sync_gradients:
+                progress_bar.update(1)
+                global_step += 1
+
+                if accelerator.is_main_process:
+                    if global_step % args.checkpointing_steps == 0:
+                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
+                        if args.checkpoints_total_limit is not None:
+                            checkpoints = os.listdir(args.output_dir)
+                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
+                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))
+
+                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
+                            if len(checkpoints) >= args.checkpoints_total_limit:
+                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
+                                removing_checkpoints = checkpoints[0:num_to_remove]
+
+                                logger.info(
+                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
+                                )
+                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")
+
+                                for removing_checkpoint in removing_checkpoints:
+                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
+                                    shutil.rmtree(removing_checkpoint)
+
+                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
+                        accelerator.save_state(save_path)
+                        logger.info(f"Saved state to {save_path}")
+
+                    if args.validation_prompt is not None and global_step % args.validation_steps == 0:
+                        image_logs = log_validation(
+                            vae,
+                            text_encoder,
+                            tokenizer,
+                            unet,
+                            controlnet,
+                            args,
+                            accelerator,
+                            weight_dtype,
+                            global_step,
+                        )
+
+            logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
+            progress_bar.set_postfix(**logs)
+            accelerator.log(logs, step=global_step)
+
+            if global_step >= args.max_train_steps:
+                break
+
+    # Create the pipeline using using the trained modules and save it.
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        controlnet = accelerator.unwrap_model(controlnet)
+        controlnet.save_pretrained(args.output_dir)
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                image_logs=image_logs,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)