Initial Commit

.gitattributes

@@ -0,0 +1,41 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz 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
+*.stl filter=lfs diff=lfs merge=lfs -text
+*.glb filter=lfs diff=lfs merge=lfs -text
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.mp4 filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,4 @@
+.idea
+.DS_Store
+__pycache__
+weights

README.md

@@ -0,0 +1,16 @@
+---
+title: Stable Normal Estimation
+emoji: 🏵️
+colorFrom: blue
+colorTo: red
+sdk: gradio
+sdk_version: 4.32.2
+app_file: app.py
+pinned: true
+license: cc-by-sa-4.0
+models:
+- Stable-X/stable-normal-v0-1
+- Stable-X/yoso-normal-v0-1
+hf_oauth: true
+hf_oauth_expiration_minutes: 43200
+---

app.py

@@ -0,0 +1,506 @@
+# Copyright 2024 Anton Obukhov, ETH Zurich. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# --------------------------------------------------------------------------
+# If you find this code useful, we kindly ask you to cite our paper in your work.
+# Please find bibtex at: https://github.com/prs-eth/Marigold#-citation
+# More information about the method can be found at https://marigoldmonodepth.github.io
+# --------------------------------------------------------------------------
+from __future__ import annotations
+
+import functools
+import os
+import tempfile
+
+import diffusers
+import gradio as gr
+import imageio as imageio
+import numpy as np
+import spaces
+import torch as torch
+from PIL import Image
+from gradio_imageslider import ImageSlider
+from tqdm import tqdm
+
+from pathlib import Path
+
+import gradio
+from gradio.utils import get_cache_folder
+from stablenormal.pipeline_yoso_normal import YOSONormalsPipeline
+from stablenormal.pipeline_stablenormal import StableNormalPipeline
+from stablenormal.scheduler.heuristics_ddimsampler import HEURI_DDIMScheduler
+
+class Examples(gradio.helpers.Examples):
+    def __init__(self, *args, directory_name=None, **kwargs):
+        super().__init__(*args, **kwargs, _initiated_directly=False)
+        if directory_name is not None:
+            self.cached_folder = get_cache_folder() / directory_name
+            self.cached_file = Path(self.cached_folder) / "log.csv"
+        self.create()
+
+
+default_seed = 2024
+default_batch_size = 1
+
+default_image_processing_resolution = 768
+
+default_video_num_inference_steps = 10
+default_video_processing_resolution = 768
+default_video_out_max_frames = 450
+
+def process_image_check(path_input):
+    if path_input is None:
+        raise gr.Error(
+            "Missing image in the first pane: upload a file or use one from the gallery below."
+        )
+
+def process_image(
+    pipe,
+    path_input,
+):
+    name_base, name_ext = os.path.splitext(os.path.basename(path_input))
+    print(f"Processing image {name_base}{name_ext}")
+
+    path_output_dir = tempfile.mkdtemp()
+    # path_out_fp32 = os.path.join(path_output_dir, f"{name_base}_normal_fp32.npy")
+    path_out_png = os.path.join(path_output_dir, f"{name_base}_normal_colored.png")
+
+    input_image = Image.open(path_input)
+    input_image = center_crop(input_image)
+    
+    pipe_out = pipe(
+        input_image,
+        match_input_resolution=False,
+        return_intermediate_result=False
+    )
+
+    normal_pred = pipe_out.prediction[0, :, :]
+    normal_colored = pipe.image_processor.visualize_normals(pipe_out.prediction)
+    normal_colored[-1].save(path_out_png)
+    print(path_out_png)
+    # np.save(path_out_fp32, normal_pred)
+    # path_out_vis = os.path.join(path_output_dir, f"{name_base}_normal_refinement_process.gif")
+    # normal_colored[0].save(path_out_vis, save_all=True, 
+    #                        append_images=normal_colored[1:], 
+                        #    duration=400, loop=0)
+    return [input_image, path_out_png]
+
+def center_crop(img):
+    # Open the image file
+    img_width, img_height = img.size
+    crop_width =min(img_width, img_height)
+    # Calculate the cropping box
+    left = (img_width - crop_width) / 2
+    top = (img_height - crop_width) / 2
+    right = (img_width + crop_width) / 2
+    bottom = (img_height + crop_width) / 2
+    
+    # Crop the image
+    img_cropped = img.crop((left, top, right, bottom))
+    return img_cropped
+
+def process_video(
+    pipe,
+    path_input,
+    out_max_frames=default_video_out_max_frames,
+    target_fps=3,
+    progress=gr.Progress(),
+):
+    if path_input is None:
+        raise gr.Error(
+            "Missing video in the first pane: upload a file or use one from the gallery below."
+        )
+
+    name_base, name_ext = os.path.splitext(os.path.basename(path_input))
+    print(f"Processing video {name_base}{name_ext}")
+
+    path_output_dir = tempfile.mkdtemp()
+    path_out_vis = os.path.join(path_output_dir, f"{name_base}_normal_colored.mp4")
+
+    reader, writer = None, None
+    try:
+        reader = imageio.get_reader(path_input)
+
+        meta_data = reader.get_meta_data()
+        fps = meta_data["fps"]
+        size = meta_data["size"]
+        duration_sec = meta_data["duration"]
+
+        writer = imageio.get_writer(path_out_vis, fps=target_fps)
+
+        out_frame_id = 0
+        pbar = tqdm(desc="Processing Video", total=duration_sec)
+
+        for frame_id, frame in enumerate(reader):
+            if frame_id % (fps // target_fps) != 0:
+                continue
+            else:
+                out_frame_id += 1
+                pbar.update(1)
+            if out_frame_id > out_max_frames:
+                break
+
+            frame_pil = Image.fromarray(frame)
+            frame_pil = center_crop(frame_pil)
+            pipe_out = pipe(
+                frame_pil,
+                match_input_resolution=False,
+                return_intermediate_result=False
+            )
+
+            processed_frame = pipe.image_processor.visualize_normals(  # noqa
+                pipe_out.prediction
+            )[0]
+            processed_frame = np.array(processed_frame)
+
+            _processed_frame = imageio.core.util.Array(processed_frame)
+            writer.append_data(_processed_frame)
+            
+            yield (
+                [frame_pil, processed_frame],
+                None,
+            )
+    finally:
+
+        if writer is not None:
+            writer.close()
+
+        if reader is not None:
+            reader.close()
+
+    yield (
+        [frame_pil, processed_frame],
+        [path_out_vis,]
+    )
+
+
+def run_demo_server(pipe):
+    process_pipe_image = spaces.GPU(functools.partial(process_image, pipe))
+    process_pipe_video = spaces.GPU(
+        functools.partial(process_video, pipe), duration=120
+    )
+
+    gradio_theme = gr.themes.Default()
+
+    with gr.Blocks(
+        theme=gradio_theme,
+        title="Stable Normal Estimation",
+        css="""
+            #download {
+                height: 118px;
+            }
+            .slider .inner {
+                width: 5px;
+                background: #FFF;
+            }
+            .viewport {
+                aspect-ratio: 4/3;
+            }
+            .tabs button.selected {
+                font-size: 20px !important;
+                color: crimson !important;
+            }
+            h1 {
+                text-align: center;
+                display: block;
+            }
+            h2 {
+                text-align: center;
+                display: block;
+            }
+            h3 {
+                text-align: center;
+                display: block;
+            }
+            .md_feedback li {
+                margin-bottom: 0px !important;
+            }
+        """,
+        head="""
+            <script async src="https://www.googletagmanager.com/gtag/js?id=G-1FWSVCGZTG"></script>
+            <script>
+                window.dataLayer = window.dataLayer || [];
+                function gtag() {dataLayer.push(arguments);}
+                gtag('js', new Date());
+                gtag('config', 'G-1FWSVCGZTG');
+            </script>
+        """,
+    ) as demo:
+        gr.Markdown(
+            """
+            # StableNormal: Reducing Diffusion Variance for Stable and Sharp Normal
+            <p align="center">
+        """
+        )
+
+        with gr.Tabs(elem_classes=["tabs"]):
+            with gr.Tab("Image"):
+                with gr.Row():
+                    with gr.Column():
+                        image_input = gr.Image(
+                            label="Input Image",
+                            type="filepath",
+                        )
+                        with gr.Row():
+                            image_submit_btn = gr.Button(
+                                value="Compute Normal", variant="primary"
+                            )
+                            image_reset_btn = gr.Button(value="Reset")
+                    with gr.Column():
+                        image_output_slider = ImageSlider(
+                            label="Normal outputs",
+                            type="filepath",
+                            show_download_button=True,
+                            show_share_button=True,
+                            interactive=False,
+                            elem_classes="slider",
+                            position=0.25,
+                        )
+
+                Examples(
+                    fn=process_pipe_image,
+                    examples=sorted([
+                        os.path.join("files", "image", name)
+                        for name in os.listdir(os.path.join("files", "image"))
+                    ]),
+                    inputs=[image_input],
+                    outputs=[image_output_slider],
+                    cache_examples=True,
+                    directory_name="examples_image",
+                )
+
+            with gr.Tab("Video"):
+                with gr.Row():
+                    with gr.Column():
+                        video_input = gr.Video(
+                            label="Input Video",
+                            sources=["upload", "webcam"],
+                        )
+                        with gr.Row():
+                            video_submit_btn = gr.Button(
+                                value="Compute Normal", variant="primary"
+                            )
+                            video_reset_btn = gr.Button(value="Reset")
+                    with gr.Column():
+                        processed_frames = ImageSlider(
+                            label="Realtime Visualization",
+                            type="filepath",
+                            show_download_button=True,
+                            show_share_button=True,
+                            interactive=False,
+                            elem_classes="slider",
+                            position=0.25,
+                        )
+                        video_output_files = gr.Files(
+                            label="Normal outputs",
+                            elem_id="download",
+                            interactive=False,
+                        )
+                Examples(
+                    fn=process_pipe_video,
+                    examples=sorted([
+                        os.path.join("files", "video", name)
+                        for name in os.listdir(os.path.join("files", "video"))
+                    ]),
+                    inputs=[video_input],
+                    outputs=[processed_frames, video_output_files],
+                    directory_name="examples_video",
+                    cache_examples=True,
+                )
+                
+            with gr.Tab("Panorama"):
+                with gr.Column():
+                    gr.Markdown("Functionality coming soon on June.10th")
+
+            with gr.Tab("4K Image"):
+                with gr.Column():
+                    gr.Markdown("Functionality coming soon on June.17th")
+
+            with gr.Tab("Normal Mapping"):
+                with gr.Column():
+                    gr.Markdown("Functionality coming soon on June.24th")
+            
+            with gr.Tab("Normal SuperResolution"):
+                with gr.Column():
+                    gr.Markdown("Functionality coming soon on June.30th")
+
+        ### Image tab
+        image_submit_btn.click(
+            fn=process_image_check,
+            inputs=image_input,
+            outputs=None,
+            preprocess=False,
+            queue=False,
+        ).success(
+            fn=process_pipe_image,
+            inputs=[
+                image_input,
+            ],
+            outputs=[image_output_slider],
+            concurrency_limit=1,
+        )
+
+        image_reset_btn.click(
+            fn=lambda: (
+                None,
+                None,
+                None,
+            ),
+            inputs=[],
+            outputs=[
+                image_input,
+                image_output_slider,
+            ],
+            queue=False,
+        )
+
+        ### Video tab
+
+        video_submit_btn.click(
+            fn=process_pipe_video,
+            inputs=[video_input],
+            outputs=[processed_frames, video_output_files],
+            concurrency_limit=1,
+        )
+
+        video_reset_btn.click(
+            fn=lambda: (None, None, None),
+            inputs=[],
+            outputs=[video_input, processed_frames, video_output_files],
+            concurrency_limit=1,
+        )
+
+        ### Server launch
+
+        demo.queue(
+            api_open=False,
+        ).launch(
+            server_name="0.0.0.0",
+            server_port=7860,
+        )
+
+from einops import rearrange  
+class DINOv2_Encoder:
+    IMAGENET_DEFAULT_MEAN = [0.485, 0.456, 0.406]
+    IMAGENET_DEFAULT_STD = [0.229, 0.224, 0.225]
+
+    def __init__(
+        self,
+        model_name = 'dinov2_vitl14',
+        freeze = True,
+        antialias=True,
+        device="cuda",
+        size = 448,
+    ):
+
+        super(DINOv2_Encoder).__init__()
+
+        self.model = torch.hub.load('facebookresearch/dinov2', model_name)
+        self.model.eval()
+        self.device = device
+        self.antialias = antialias
+        self.dtype = torch.float32
+
+        self.mean = torch.Tensor(self.IMAGENET_DEFAULT_MEAN)
+        self.std = torch.Tensor(self.IMAGENET_DEFAULT_STD)
+        self.size = size
+        if freeze:
+            self.freeze()
+
+
+    def freeze(self):
+        for param in self.model.parameters():
+            param.requires_grad = False
+
+    @torch.no_grad()
+    def encoder(self, x):
+        '''
+        x: [b h w c], range from (-1, 1), rbg
+        '''
+
+        x = self.preprocess(x).to(self.device, self.dtype)
+
+        b, c, h, w = x.shape
+        patch_h, patch_w = h // 14, w // 14
+
+        embeddings = self.model.forward_features(x)['x_norm_patchtokens']
+        embeddings = rearrange(embeddings, 'b (h w) c -> b h w c', h = patch_h, w = patch_w)
+
+        return  rearrange(embeddings, 'b h w c -> b c h w')
+
+    def preprocess(self, x):
+        ''' x
+        '''
+        # normalize to [0,1],
+        x = torch.nn.functional.interpolate(
+            x,
+            size=(self.size, self.size),
+            mode='bicubic',
+            align_corners=True,
+            antialias=self.antialias,
+        )
+
+        x = (x + 1.0) / 2.0
+        # renormalize according to dino
+        mean = self.mean.view(1, 3, 1, 1).to(x.device)
+        std = self.std.view(1, 3, 1, 1).to(x.device)
+        x = (x - mean) / std
+
+        return x
+    
+    def to(self, device, dtype=None):
+        if dtype is not None:
+            self.dtype = dtype
+            self.model.to(device, dtype)
+            self.mean.to(device, dtype)
+            self.std.to(device, dtype)
+        else:
+            self.model.to(device)
+            self.mean.to(device)
+            self.std.to(device)
+        return self
+
+    def __call__(self, x, **kwargs):
+        return self.encoder(x, **kwargs)
+    
+def main():
+    os.system("pip freeze")
+
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    
+    x_start_pipeline = YOSONormalsPipeline.from_pretrained(
+        'weights/yoso-normal-v0-1', trust_remote_code=True,
+        t_start=300).to(device, torch.float16)
+    dinov2_prior = DINOv2_Encoder(size=672)
+    dinov2_prior.to(device, torch.float16)
+    
+    pipe = StableNormalPipeline.from_pretrained('weights/stable-normal-v0-1', t_start=300, trust_remote_code=True,
+                                                scheduler=HEURI_DDIMScheduler(prediction_type='sample', 
+                                                                              beta_start=0.00085, beta_end=0.0120, 
+                                                                              beta_schedule = "scaled_linear"))
+    # two stage concat
+    pipe.x_start_pipeline = x_start_pipeline
+    pipe.prior = dinov2_prior
+    pipe.to(device, torch.float16)
+    
+    try:
+        import xformers
+        pipe.enable_xformers_memory_efficient_attention()
+    except:
+        pass  # run without xformers
+
+    run_demo_server(pipe)
+
+
+if __name__ == "__main__":
+    main()

requirements.txt

@@ -0,0 +1,132 @@
+accelerate==0.30.1
+aiofiles==23.2.1
+aiohttp==3.9.5
+aiosignal==1.3.1
+altair==5.3.0
+annotated-types==0.7.0
+anyio==4.4.0
+async-timeout==4.0.3
+attrs==23.2.0
+Authlib==1.3.0
+certifi==2024.2.2
+cffi==1.16.0
+charset-normalizer==3.3.2
+click==8.0.4
+contourpy==1.2.1
+cryptography==42.0.7
+cycler==0.12.1
+dataclasses-json==0.6.6
+datasets==2.19.1
+Deprecated==1.2.14
+diffusers==0.28.0
+dill==0.3.8
+dnspython==2.6.1
+email_validator==2.1.1
+exceptiongroup==1.2.1
+fastapi==0.111.0
+fastapi-cli==0.0.4
+ffmpy==0.3.2
+filelock==3.14.0
+fonttools==4.53.0
+frozenlist==1.4.1
+fsspec==2024.3.1
+gradio==4.32.2
+gradio_client==0.17.0
+gradio_imageslider==0.0.20
+h11==0.14.0
+httpcore==1.0.5
+httptools==0.6.1
+httpx==0.27.0
+huggingface-hub==0.23.0
+idna==3.7
+imageio==2.34.1
+imageio-ffmpeg==0.5.0
+importlib_metadata==7.1.0
+importlib_resources==6.4.0
+itsdangerous==2.2.0
+Jinja2==3.1.4
+jsonschema==4.22.0
+jsonschema-specifications==2023.12.1
+kiwisolver==1.4.5
+markdown-it-py==3.0.0
+MarkupSafe==2.1.5
+marshmallow==3.21.2
+matplotlib==3.8.2
+mdurl==0.1.2
+mpmath==1.3.0
+multidict==6.0.5
+multiprocess==0.70.16
+mypy-extensions==1.0.0
+networkx==3.3
+numpy==1.26.4
+nvidia-cublas-cu12==12.1.3.1
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-cuda-nvrtc-cu12==12.1.105
+nvidia-cuda-runtime-cu12==12.1.105
+nvidia-cudnn-cu12==8.9.2.26
+nvidia-cufft-cu12==11.0.2.54
+nvidia-curand-cu12==10.3.2.106
+nvidia-cusolver-cu12==11.4.5.107
+nvidia-cusparse-cu12==12.1.0.106
+nvidia-nccl-cu12==2.19.3
+nvidia-nvjitlink-cu12==12.5.40
+nvidia-nvtx-cu12==12.1.105
+orjson==3.10.3
+packaging==24.0
+pandas==2.2.2
+pillow==10.3.0
+protobuf==3.20.3
+psutil==5.9.8
+pyarrow==16.0.0
+pyarrow-hotfix==0.6
+pycparser==2.22
+pydantic==2.7.2
+pydantic_core==2.18.3
+pydub==0.25.1
+pygltflib==1.16.1
+Pygments==2.18.0
+pyparsing==3.1.2
+python-dateutil==2.9.0.post0
+python-dotenv==1.0.1
+python-multipart==0.0.9
+pytz==2024.1
+PyYAML==6.0.1
+referencing==0.35.1
+regex==2024.5.15
+requests==2.31.0
+rich==13.7.1
+rpds-py==0.18.1
+ruff==0.4.7
+safetensors==0.4.3
+scipy==1.11.4
+semantic-version==2.10.0
+shellingham==1.5.4
+six==1.16.0
+sniffio==1.3.1
+spaces==0.28.3
+starlette==0.37.2
+sympy==1.12.1
+tokenizers==0.15.2
+tomlkit==0.12.0
+toolz==0.12.1
+torch==2.2.0
+tqdm==4.66.4
+transformers==4.36.1
+trimesh==4.0.5
+triton==2.2.0
+typer==0.12.3
+typing-inspect==0.9.0
+typing_extensions==4.11.0
+tzdata==2024.1
+ujson==5.10.0
+urllib3==2.2.1
+uvicorn==0.30.0
+uvloop==0.19.0
+watchfiles==0.22.0
+websockets==11.0.3
+wrapt==1.16.0
+xformers==0.0.24
+xxhash==3.4.1
+yarl==1.9.4
+zipp==3.19.1
+einops==0.7.0

requirements_min.txt

@@ -0,0 +1,17 @@
+gradio>=4.32.1
+gradio-imageslider>=0.0.20
+pygltflib==1.16.1
+trimesh==4.0.5
+imageio
+imageio-ffmpeg
+Pillow
+einops==0.7.0
+
+spaces
+accelerate
+diffusers>=0.28.0
+matplotlib==3.8.2
+scipy==1.11.4
+torch==2.0.1
+transformers==4.36.1
+xformers==0.0.21

setup.py

@@ -0,0 +1,9 @@
+from pathlib import Path
+from setuptools import setup, find_packages
+
+setup_path = Path(__file__).parent
+
+setup(
+    name = "stablenormal",
+    packages=find_packages()
+)

stablenormal/pipeline_stablenormal.py

@@ -0,0 +1,1201 @@
+# Copyright 2024 Marigold authors, PRS ETH Zurich. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# --------------------------------------------------------------------------
+# More information and citation instructions are available on the
+# --------------------------------------------------------------------------
+from dataclasses import dataclass
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import numpy as np
+import torch
+from PIL import Image
+from tqdm.auto import tqdm
+from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
+
+
+from diffusers.image_processor import PipelineImageInput
+from diffusers.models import (
+    AutoencoderKL,
+    UNet2DConditionModel,
+    ControlNetModel,
+)
+from diffusers.schedulers import (
+    DDIMScheduler
+)
+
+from diffusers.utils import (
+    BaseOutput,
+    logging,
+    replace_example_docstring,
+)
+
+from diffusers.models.unets.unet_2d_condition import UNet2DConditionOutput
+
+from diffusers.utils import USE_PEFT_BACKEND, BaseOutput, deprecate, logging, scale_lora_layers, unscale_lora_layers
+
+
+
+from diffusers.utils.torch_utils import randn_tensor
+from diffusers.pipelines.controlnet import StableDiffusionControlNetPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.marigold.marigold_image_processing import MarigoldImageProcessor
+from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
+import torch.nn.functional as F
+
+import pdb
+
+
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+EXAMPLE_DOC_STRING = """
+Examples:
+```py
+>>> import diffusers
+>>> import torch
+
+>>> pipe = diffusers.MarigoldNormalsPipeline.from_pretrained(
+...     "prs-eth/marigold-normals-lcm-v0-1", variant="fp16", torch_dtype=torch.float16
+... ).to("cuda")
+
+>>> image = diffusers.utils.load_image("https://marigoldmonodepth.github.io/images/einstein.jpg")
+>>> normals = pipe(image)
+
+>>> vis = pipe.image_processor.visualize_normals(normals.prediction)
+>>> vis[0].save("einstein_normals.png")
+```
+"""
+
+
+@dataclass
+class StableNormalOutput(BaseOutput):
+    """
+    Output class for Marigold monocular normals prediction pipeline.
+
+    Args:
+        prediction (`np.ndarray`, `torch.Tensor`):
+            Predicted normals with values in the range [-1, 1]. The shape is always $numimages \times 3 \times height
+            \times width$, regardless of whether the images were passed as a 4D array or a list.
+        uncertainty (`None`, `np.ndarray`, `torch.Tensor`):
+            Uncertainty maps computed from the ensemble, with values in the range [0, 1]. The shape is $numimages
+            \times 1 \times height \times width$.
+        latent (`None`, `torch.Tensor`):
+            Latent features corresponding to the predictions, compatible with the `latents` argument of the pipeline.
+            The shape is $numimages * numensemble \times 4 \times latentheight \times latentwidth$.
+    """
+
+    prediction: Union[np.ndarray, torch.Tensor]
+    latent: Union[None, torch.Tensor]
+
+
+class StableNormalPipeline(StableDiffusionControlNetPipeline):
+    """ Pipeline for monocular normals estimation using the Marigold method: https://marigoldmonodepth.github.io.
+    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
+    implemented for all pipelines (downloading, saving, running on a particular device, etc.).
+
+    The pipeline also inherits the following loading methods:
+        - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
+        - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
+
+    Args:
+        vae ([`AutoencoderKL`]):
+            Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
+        text_encoder ([`~transformers.CLIPTextModel`]):
+            Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)).
+        tokenizer ([`~transformers.CLIPTokenizer`]):
+            A `CLIPTokenizer` to tokenize text.
+        unet ([`UNet2DConditionModel`]):
+            A `UNet2DConditionModel` 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 more details
+            about a model's potential harms.
+        feature_extractor ([`~transformers.CLIPImageProcessor`]):
+            A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
+    """
+
+    model_cpu_offload_seq = "text_encoder->image_encoder->unet->vae"
+    _optional_components = ["safety_checker", "feature_extractor", "image_encoder"]
+    _exclude_from_cpu_offload = ["safety_checker"]
+    _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
+
+
+
+    def __init__(
+        self,
+        vae: AutoencoderKL,
+        text_encoder: CLIPTextModel,
+        tokenizer: CLIPTokenizer,
+        unet: UNet2DConditionModel,
+        controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel]],
+        dino_controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel]],
+        scheduler: Union[DDIMScheduler],
+        safety_checker: StableDiffusionSafetyChecker,
+        feature_extractor: CLIPImageProcessor,
+        image_encoder: CLIPVisionModelWithProjection = None,
+        requires_safety_checker: bool = True,
+        default_denoising_steps: Optional[int] = 10,
+        default_processing_resolution: Optional[int] = 768,
+        prompt="The normal map",
+        empty_text_embedding=None,
+        t_start: Optional[int] = 401,
+    ):
+        super().__init__(
+            vae,
+            text_encoder,
+            tokenizer,
+            unet,
+            controlnet,
+            scheduler,
+            safety_checker,
+            feature_extractor,
+            image_encoder,
+            requires_safety_checker,
+                )
+
+        self.register_modules(
+            dino_controlnet=dino_controlnet,
+        )
+
+        self.vae_scale_factor = 768
+        self.image_processor = MarigoldImageProcessor(vae_scale_factor=self.vae_scale_factor)
+        self.dino_image_processor = lambda x: x / 127.5 -1.
+
+        self.default_denoising_steps = default_denoising_steps
+        self.default_processing_resolution = default_processing_resolution
+        self.prompt = prompt
+        self.prompt_embeds = None
+        self.empty_text_embedding = empty_text_embedding
+        self.t_start= torch.tensor(t_start) # target_out latents
+
+
+    def check_inputs(
+        self,
+        image: PipelineImageInput,
+        num_inference_steps: int,
+        ensemble_size: int,
+        processing_resolution: int,
+        resample_method_input: str,
+        resample_method_output: str,
+        batch_size: int,
+        ensembling_kwargs: Optional[Dict[str, Any]],
+        latents: Optional[torch.Tensor],
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]],
+        output_type: str,
+        output_uncertainty: bool,
+    ) -> int:
+        if num_inference_steps is None:
+            raise ValueError("`num_inference_steps` is not specified and could not be resolved from the model config.")
+        if num_inference_steps < 1:
+            raise ValueError("`num_inference_steps` must be positive.")
+        if ensemble_size < 1:
+            raise ValueError("`ensemble_size` must be positive.")
+        if ensemble_size == 2:
+            logger.warning(
+                "`ensemble_size` == 2 results are similar to no ensembling (1); "
+                "consider increasing the value to at least 3."
+            )
+        if ensemble_size == 1 and output_uncertainty:
+            raise ValueError(
+                "Computing uncertainty by setting `output_uncertainty=True` also requires setting `ensemble_size` "
+                "greater than 1."
+            )
+        if processing_resolution is None:
+            raise ValueError(
+                "`processing_resolution` is not specified and could not be resolved from the model config."
+            )
+        if processing_resolution < 0:
+            raise ValueError(
+                "`processing_resolution` must be non-negative: 0 for native resolution, or any positive value for "
+                "downsampled processing."
+            )
+        if processing_resolution % self.vae_scale_factor != 0:
+            raise ValueError(f"`processing_resolution` must be a multiple of {self.vae_scale_factor}.")
+        if resample_method_input not in ("nearest", "nearest-exact", "bilinear", "bicubic", "area"):
+            raise ValueError(
+                "`resample_method_input` takes string values compatible with PIL library: "
+                "nearest, nearest-exact, bilinear, bicubic, area."
+            )
+        if resample_method_output not in ("nearest", "nearest-exact", "bilinear", "bicubic", "area"):
+            raise ValueError(
+                "`resample_method_output` takes string values compatible with PIL library: "
+                "nearest, nearest-exact, bilinear, bicubic, area."
+            )
+        if batch_size < 1:
+            raise ValueError("`batch_size` must be positive.")
+        if output_type not in ["pt", "np"]:
+            raise ValueError("`output_type` must be one of `pt` or `np`.")
+        if latents is not None and generator is not None:
+            raise ValueError("`latents` and `generator` cannot be used together.")
+        if ensembling_kwargs is not None:
+            if not isinstance(ensembling_kwargs, dict):
+                raise ValueError("`ensembling_kwargs` must be a dictionary.")
+            if "reduction" in ensembling_kwargs and ensembling_kwargs["reduction"] not in ("closest", "mean"):
+                raise ValueError("`ensembling_kwargs['reduction']` can be either `'closest'` or `'mean'`.")
+
+        # image checks
+        num_images = 0
+        W, H = None, None
+        if not isinstance(image, list):
+            image = [image]
+        for i, img in enumerate(image):
+            if isinstance(img, np.ndarray) or torch.is_tensor(img):
+                if img.ndim not in (2, 3, 4):
+                    raise ValueError(f"`image[{i}]` has unsupported dimensions or shape: {img.shape}.")
+                H_i, W_i = img.shape[-2:]
+                N_i = 1
+                if img.ndim == 4:
+                    N_i = img.shape[0]
+            elif isinstance(img, Image.Image):
+                W_i, H_i = img.size
+                N_i = 1
+            else:
+                raise ValueError(f"Unsupported `image[{i}]` type: {type(img)}.")
+            if W is None:
+                W, H = W_i, H_i
+            elif (W, H) != (W_i, H_i):
+                raise ValueError(
+                    f"Input `image[{i}]` has incompatible dimensions {(W_i, H_i)} with the previous images {(W, H)}"
+                )
+            num_images += N_i
+
+        # latents checks
+        if latents is not None:
+            if not torch.is_tensor(latents):
+                raise ValueError("`latents` must be a torch.Tensor.")
+            if latents.dim() != 4:
+                raise ValueError(f"`latents` has unsupported dimensions or shape: {latents.shape}.")
+
+            if processing_resolution > 0:
+                max_orig = max(H, W)
+                new_H = H * processing_resolution // max_orig
+                new_W = W * processing_resolution // max_orig
+                if new_H == 0 or new_W == 0:
+                    raise ValueError(f"Extreme aspect ratio of the input image: [{W} x {H}]")
+                W, H = new_W, new_H
+            w = (W + self.vae_scale_factor - 1) // self.vae_scale_factor
+            h = (H + self.vae_scale_factor - 1) // self.vae_scale_factor
+            shape_expected = (num_images * ensemble_size, self.vae.config.latent_channels, h, w)
+
+            if latents.shape != shape_expected:
+                raise ValueError(f"`latents` has unexpected shape={latents.shape} expected={shape_expected}.")
+
+        # generator checks
+        if generator is not None:
+            if isinstance(generator, list):
+                if len(generator) != num_images * ensemble_size:
+                    raise ValueError(
+                        "The number of generators must match the total number of ensemble members for all input images."
+                    )
+                if not all(g.device.type == generator[0].device.type for g in generator):
+                    raise ValueError("`generator` device placement is not consistent in the list.")
+            elif not isinstance(generator, torch.Generator):
+                raise ValueError(f"Unsupported generator type: {type(generator)}.")
+
+        return num_images
+
+    def progress_bar(self, iterable=None, total=None, desc=None, leave=True):
+        if not hasattr(self, "_progress_bar_config"):
+            self._progress_bar_config = {}
+        elif not isinstance(self._progress_bar_config, dict):
+            raise ValueError(
+                f"`self._progress_bar_config` should be of type `dict`, but is {type(self._progress_bar_config)}."
+            )
+
+        progress_bar_config = dict(**self._progress_bar_config)
+        progress_bar_config["desc"] = progress_bar_config.get("desc", desc)
+        progress_bar_config["leave"] = progress_bar_config.get("leave", leave)
+        if iterable is not None:
+            return tqdm(iterable, **progress_bar_config)
+        elif total is not None:
+            return tqdm(total=total, **progress_bar_config)
+        else:
+            raise ValueError("Either `total` or `iterable` has to be defined.")
+
+    @torch.no_grad()
+    @replace_example_docstring(EXAMPLE_DOC_STRING)
+    def __call__(
+        self,
+        image: PipelineImageInput,
+        prompt: Union[str, List[str]] = None,
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        num_inference_steps: Optional[int] = None,
+        ensemble_size: int = 1,
+        processing_resolution: Optional[int] = None,
+        return_intermediate_result: bool = False,
+        match_input_resolution: bool = True,
+        resample_method_input: str = "bilinear",
+        resample_method_output: str = "bilinear",
+        batch_size: int = 1,
+        ensembling_kwargs: Optional[Dict[str, Any]] = None,
+        latents: Optional[Union[torch.Tensor, List[torch.Tensor]]] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        num_images_per_prompt: Optional[int] = 1,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        controlnet_conditioning_scale: Union[float, List[float]] = 1.0,
+        output_type: str = "np",
+        output_uncertainty: bool = False,
+        output_latent: bool = False,
+        return_dict: bool = True,
+    ):
+        """
+        Function invoked when calling the pipeline.
+
+        Args:
+            image (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`),
+                `List[torch.Tensor]`: An input image or images used as an input for the normals estimation task. For
+                arrays and tensors, the expected value range is between `[0, 1]`. Passing a batch of images is possible
+                by providing a four-dimensional array or a tensor. Additionally, a list of images of two- or
+                three-dimensional arrays or tensors can be passed. In the latter case, all list elements must have the
+                same width and height.
+            num_inference_steps (`int`, *optional*, defaults to `None`):
+                Number of denoising diffusion steps during inference. The default value `None` results in automatic
+                selection. The number of steps should be at least 10 with the full Marigold models, and between 1 and 4
+                for Marigold-LCM models.
+            ensemble_size (`int`, defaults to `1`):
+                Number of ensemble predictions. Recommended values are 5 and higher for better precision, or 1 for
+                faster inference.
+            processing_resolution (`int`, *optional*, defaults to `None`):
+                Effective processing resolution. When set to `0`, matches the larger input image dimension. This
+                produces crisper predictions, but may also lead to the overall loss of global context. The default
+                value `None` resolves to the optimal value from the model config.
+            match_input_resolution (`bool`, *optional*, defaults to `True`):
+                When enabled, the output prediction is resized to match the input dimensions. When disabled, the longer
+                side of the output will equal to `processing_resolution`.
+            resample_method_input (`str`, *optional*, defaults to `"bilinear"`):
+                Resampling method used to resize input images to `processing_resolution`. The accepted values are:
+                `"nearest"`, `"nearest-exact"`, `"bilinear"`, `"bicubic"`, or `"area"`.
+            resample_method_output (`str`, *optional*, defaults to `"bilinear"`):
+                Resampling method used to resize output predictions to match the input resolution. The accepted values
+                are `"nearest"`, `"nearest-exact"`, `"bilinear"`, `"bicubic"`, or `"area"`.
+            batch_size (`int`, *optional*, defaults to `1`):
+                Batch size; only matters when setting `ensemble_size` or passing a tensor of images.
+            ensembling_kwargs (`dict`, *optional*, defaults to `None`)
+                Extra dictionary with arguments for precise ensembling control. The following options are available:
+                - reduction (`str`, *optional*, defaults to `"closest"`): Defines the ensembling function applied in
+                  every pixel location, can be either `"closest"` or `"mean"`.
+            latents (`torch.Tensor`, *optional*, defaults to `None`):
+                Latent noise tensors to replace the random initialization. These can be taken from the previous
+                function call's output.
+            generator (`torch.Generator`, or `List[torch.Generator]`, *optional*, defaults to `None`):
+                Random number generator object to ensure reproducibility.
+            output_type (`str`, *optional*, defaults to `"np"`):
+                Preferred format of the output's `prediction` and the optional `uncertainty` fields. The accepted
+                values are: `"np"` (numpy array) or `"pt"` (torch tensor).
+            output_uncertainty (`bool`, *optional*, defaults to `False`):
+                When enabled, the output's `uncertainty` field contains the predictive uncertainty map, provided that
+                the `ensemble_size` argument is set to a value above 2.
+            output_latent (`bool`, *optional*, defaults to `False`):
+                When enabled, the output's `latent` field contains the latent codes corresponding to the predictions
+                within the ensemble. These codes can be saved, modified, and used for subsequent calls with the
+                `latents` argument.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~pipelines.marigold.MarigoldDepthOutput`] instead of a plain tuple.
+
+        Examples:
+
+        Returns:
+            [`~pipelines.marigold.MarigoldNormalsOutput`] or `tuple`:
+                If `return_dict` is `True`, [`~pipelines.marigold.MarigoldNormalsOutput`] is returned, otherwise a
+                `tuple` is returned where the first element is the prediction, the second element is the uncertainty
+                (or `None`), and the third is the latent (or `None`).
+        """
+
+        # 0. Resolving variables.
+        device = self._execution_device
+        dtype = self.dtype
+
+        # Model-specific optimal default values leading to fast and reasonable results.
+        if num_inference_steps is None:
+            num_inference_steps = self.default_denoising_steps
+        if processing_resolution is None:
+            processing_resolution = self.default_processing_resolution
+
+
+        image, padding, original_resolution = self.image_processor.preprocess(
+            image, processing_resolution, resample_method_input, device, dtype
+        )  # [N,3,PPH,PPW]
+
+        # 0. X_start latent obtain
+        predictor = self.x_start_pipeline(image, skip_preprocess=True)
+        x_start_latent = predictor.latent
+        gauss_latent = predictor.gauss_latent
+
+        # 1. Check inputs.
+        num_images = self.check_inputs(
+            image,
+            num_inference_steps,
+            ensemble_size,
+            processing_resolution,
+            resample_method_input,
+            resample_method_output,
+            batch_size,
+            ensembling_kwargs,
+            latents,
+            generator,
+            output_type,
+            output_uncertainty,
+        )
+
+
+        # 2. Prepare empty text conditioning.
+        # Model invocation: self.tokenizer, self.text_encoder.
+        if self.empty_text_embedding is None:
+            prompt = ""
+            text_inputs = self.tokenizer(
+                prompt,
+                padding="do_not_pad",
+                max_length=self.tokenizer.model_max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+            text_input_ids = text_inputs.input_ids.to(device)
+            self.empty_text_embedding = self.text_encoder(text_input_ids)[0]  # [1,2,1024]
+
+
+
+        # 3. prepare prompt
+        if self.prompt_embeds is None:
+            prompt_embeds, negative_prompt_embeds = self.encode_prompt(
+                self.prompt,
+                device,
+                num_images_per_prompt,
+                False,
+                negative_prompt,
+                prompt_embeds=prompt_embeds,
+                negative_prompt_embeds=None,
+                lora_scale=None,
+                clip_skip=None,
+            )
+            self.prompt_embeds = prompt_embeds
+            self.negative_prompt_embeds = negative_prompt_embeds
+
+
+
+        # 5. dino guider features obtaining
+        ## TODO different case-1
+        dino_features = self.prior(image)
+        dino_features = self.dino_controlnet.dino_controlnet_cond_embedding(dino_features)
+        dino_features = self.match_noisy(dino_features, x_start_latent)
+
+        # 6. Encode input image into latent space. At this step, each of the `N` input images is represented with `E`
+        # ensemble members. Each ensemble member is an independent diffused prediction, just initialized independently.
+        # Latents of each such predictions across all input images and all ensemble members are represented in the
+        # `pred_latent` variable. The variable `image_latent` is of the same shape: it contains each input image encoded
+        # into latent space and replicated `E` times. The latents can be either generated (see `generator` to ensure
+        # reproducibility), or passed explicitly via the `latents` argument. The latter can be set outside the pipeline
+        # code. For example, in the Marigold-LCM video processing demo, the latents initialization of a frame is taken
+        # as a convex combination of the latents output of the pipeline for the previous frame and a newly-sampled
+        # noise. This behavior can be achieved by setting the `output_latent` argument to `True`. The latent space
+        # dimensions are `(h, w)`. Encoding into latent space happens in batches of size `batch_size`.
+        # Model invocation: self.vae.encoder.
+        image_latent, pred_latent = self.prepare_latents(
+            image, latents, generator, ensemble_size, batch_size
+        )  # [N*E,4,h,w], [N*E,4,h,w]
+
+
+        del (
+                image,
+        )
+
+        # 7. denoise sampling, using heuritic sampling proposed by Ye.
+        self.scheduler.set_timesteps(num_inference_steps, device=device)
+
+        cond_scale =controlnet_conditioning_scale
+        pred_latent = x_start_latent
+
+        cur_step = 0
+
+        # dino controlnet
+        dino_down_block_res_samples, dino_mid_block_res_sample = self.dino_controlnet(
+            dino_features.detach(),
+            0, # not depend on time steps
+            encoder_hidden_states=self.prompt_embeds,
+            conditioning_scale=cond_scale,
+            guess_mode=False,
+            return_dict=False,
+        )
+        assert dino_mid_block_res_sample == None
+
+        pred_latents = []
+
+        down_block_res_samples, mid_block_res_sample = self.controlnet(
+            image_latent.detach(),
+            self.t_start,
+            encoder_hidden_states=self.prompt_embeds,
+            conditioning_scale=cond_scale,
+            guess_mode=False,
+            return_dict=False,
+        )
+        last_pred_latent = pred_latent
+        for i in range(4):
+            _dino_down_block_res_samples = [dino_down_block_res_sample for dino_down_block_res_sample in dino_down_block_res_samples]  # copy, avoid repeat quiery
+            
+            model_output = self.dino_unet_forward(
+                self.unet,
+                pred_latent,
+                self.t_start,
+                encoder_hidden_states=self.prompt_embeds,
+                down_block_additional_residuals=down_block_res_samples,
+                mid_block_additional_residual=mid_block_res_sample,
+                dino_down_block_additional_residuals= _dino_down_block_res_samples,
+                return_dict=False,
+            )[0]  # [B,4,h,w]
+            pred_latents.append(model_output)
+            pred_latent = self.scheduler.add_noise(model_output, gauss_latent, self.t_start)
+            pred_latent = 0.4 * pred_latent + 0.6 * last_pred_latent
+            last_pred_latent = pred_latent
+        pred_latents = torch.cat(pred_latents, dim=0)
+        del (
+            image_latent,
+            dino_features,
+        )
+
+
+        # decoder
+        if return_intermediate_result:
+            prediction = []
+            for _pred_latent in pred_latents:
+                _prediction = self.decode_prediction(_pred_latent.unsqueeze(dim=0))
+                prediction.append(_prediction)
+            prediction = torch.cat(prediction, dim=0)
+        else:
+            prediction = self.decode_prediction(pred_latents[-1].unsqueeze(dim=0))
+        prediction = self.image_processor.unpad_image(prediction, padding)  # [N*E,3,PH,PW]
+        
+        if match_input_resolution:
+            prediction = self.image_processor.resize_antialias(
+                prediction, original_resolution, resample_method_output, is_aa=False
+            )  # [N,3,H,W]
+        prediction = self.normalize_normals(prediction)  # [N,3,H,W]
+
+        if output_type == "np":
+            prediction = self.image_processor.pt_to_numpy(prediction)  # [N,H,W,3]
+            prediction = prediction.clip(min=-1, max=1)
+
+        # 11. Offload all models
+        self.maybe_free_model_hooks()
+
+        return StableNormalOutput(
+            prediction=prediction,
+            latent=pred_latent,
+        )
+
+    # Copied from diffusers.pipelines.marigold.pipeline_marigold_depth.MarigoldDepthPipeline.prepare_latents
+    def prepare_latents(
+        self,
+        image: torch.Tensor,
+        latents: Optional[torch.Tensor],
+        generator: Optional[torch.Generator],
+        ensemble_size: int,
+        batch_size: int,
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        def retrieve_latents(encoder_output):
+            if hasattr(encoder_output, "latent_dist"):
+                return encoder_output.latent_dist.mode()
+            elif hasattr(encoder_output, "latents"):
+                return encoder_output.latents
+            else:
+                raise AttributeError("Could not access latents of provided encoder_output")
+
+
+
+        image_latent = torch.cat(
+            [
+                retrieve_latents(self.vae.encode(image[i : i + batch_size]))
+                for i in range(0, image.shape[0], batch_size)
+            ],
+            dim=0,
+        )  # [N,4,h,w]
+        image_latent = image_latent * self.vae.config.scaling_factor
+        image_latent = image_latent.repeat_interleave(ensemble_size, dim=0)  # [N*E,4,h,w]
+
+        pred_latent = latents
+        if pred_latent is None:
+
+
+            pred_latent = randn_tensor(
+                image_latent.shape,
+                generator=generator,
+                device=image_latent.device,
+                dtype=image_latent.dtype,
+            )  # [N*E,4,h,w]
+
+        return image_latent, pred_latent
+
+    def decode_prediction(self, pred_latent: torch.Tensor) -> torch.Tensor:
+        if pred_latent.dim() != 4 or pred_latent.shape[1] != self.vae.config.latent_channels:
+            raise ValueError(
+                f"Expecting 4D tensor of shape [B,{self.vae.config.latent_channels},H,W]; got {pred_latent.shape}."
+            )
+
+        prediction = self.vae.decode(pred_latent / self.vae.config.scaling_factor, return_dict=False)[0]  # [B,3,H,W]
+
+        return prediction  # [B,3,H,W]
+
+    @staticmethod
+    def normalize_normals(normals: torch.Tensor, eps: float = 1e-6) -> torch.Tensor:
+        if normals.dim() != 4 or normals.shape[1] != 3:
+            raise ValueError(f"Expecting 4D tensor of shape [B,3,H,W]; got {normals.shape}.")
+
+        norm = torch.norm(normals, dim=1, keepdim=True)
+        normals /= norm.clamp(min=eps)
+
+        return normals
+
+    @staticmethod
+    def match_noisy(dino, noisy):
+        _, __, dino_h, dino_w =  dino.shape
+        _, __, h, w =  noisy.shape
+
+        if h == dino_h and w == dino_w:
+            return dino
+        else:
+            return F.interpolate(dino, (h, w), mode='bilinear')
+
+
+
+
+
+
+
+
+
+
+    @staticmethod
+    def dino_unet_forward(
+        self,  # NOTE that repurpose to UNet
+        sample: torch.Tensor,
+        timestep: Union[torch.Tensor, float, int],
+        encoder_hidden_states: torch.Tensor,
+        class_labels: Optional[torch.Tensor] = None,
+        timestep_cond: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        cross_attention_kwargs: Optional[Dict[str, Any]] = None,
+        added_cond_kwargs: Optional[Dict[str, torch.Tensor]] = None,
+        down_block_additional_residuals: Optional[Tuple[torch.Tensor]] = None,
+        mid_block_additional_residual: Optional[torch.Tensor] = None,
+        dino_down_block_additional_residuals: Optional[torch.Tensor] = None,
+        down_intrablock_additional_residuals: Optional[Tuple[torch.Tensor]] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        return_dict: bool = True,
+    ) -> Union[UNet2DConditionOutput, Tuple]:
+        r"""
+        The [`UNet2DConditionModel`] forward method.
+
+        Args:
+            sample (`torch.Tensor`):
+                The noisy input tensor with the following shape `(batch, channel, height, width)`.
+            timestep (`torch.Tensor` or `float` or `int`): The number of timesteps to denoise an input.
+            encoder_hidden_states (`torch.Tensor`):
+                The encoder hidden states with shape `(batch, sequence_length, feature_dim)`.
+            class_labels (`torch.Tensor`, *optional*, defaults to `None`):
+                Optional class labels for conditioning. Their embeddings will be summed with the timestep embeddings.
+            timestep_cond: (`torch.Tensor`, *optional*, defaults to `None`):
+                Conditional embeddings for timestep. If provided, the embeddings will be summed with the samples passed
+                through the `self.time_embedding` layer to obtain the timestep embeddings.
+            attention_mask (`torch.Tensor`, *optional*, defaults to `None`):
+                An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. If `1` the mask
+                is kept, otherwise if `0` it is discarded. Mask will be converted into a bias, which adds large
+                negative values to the attention scores corresponding to "discard" tokens.
+            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).
+            added_cond_kwargs: (`dict`, *optional*):
+                A kwargs dictionary containing additional embeddings that if specified are added to the embeddings that
+                are passed along to the UNet blocks.
+            down_block_additional_residuals: (`tuple` of `torch.Tensor`, *optional*):
+                A tuple of tensors that if specified are added to the residuals of down unet blocks.
+            mid_block_additional_residual: (`torch.Tensor`, *optional*):
+                A tensor that if specified is added to the residual of the middle unet block.
+            down_intrablock_additional_residuals (`tuple` of `torch.Tensor`, *optional*):
+                additional residuals to be added within UNet down blocks, for example from T2I-Adapter side model(s)
+            encoder_attention_mask (`torch.Tensor`):
+                A cross-attention mask of shape `(batch, sequence_length)` is applied to `encoder_hidden_states`. If
+                `True` the mask is kept, otherwise if `False` it is discarded. Mask will be converted into a bias,
+                which adds large negative values to the attention scores corresponding to "discard" tokens.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~models.unets.unet_2d_condition.UNet2DConditionOutput`] instead of a plain
+                tuple.
+
+        Returns:
+            [`~models.unets.unet_2d_condition.UNet2DConditionOutput`] or `tuple`:
+                If `return_dict` is True, an [`~models.unets.unet_2d_condition.UNet2DConditionOutput`] is returned,
+                otherwise a `tuple` is returned where the first element is the sample tensor.
+        """
+        # By default samples have to be AT least a multiple of the overall upsampling factor.
+        # The overall upsampling factor is equal to 2 ** (# num of upsampling layers).
+        # However, the upsampling interpolation output size can be forced to fit any upsampling size
+        # on the fly if necessary.
+
+
+        default_overall_up_factor = 2**self.num_upsamplers
+
+        # upsample size should be forwarded when sample is not a multiple of `default_overall_up_factor`
+        forward_upsample_size = False
+        upsample_size = None
+
+        for dim in sample.shape[-2:]:
+            if dim % default_overall_up_factor != 0:
+                # Forward upsample size to force interpolation output size.
+                forward_upsample_size = True
+                break
+
+        # ensure attention_mask is a bias, and give it a singleton query_tokens dimension
+        # expects mask of shape:
+        #   [batch, key_tokens]
+        # adds singleton query_tokens dimension:
+        #   [batch,                    1, key_tokens]
+        # this helps to broadcast it as a bias over attention scores, which will be in one of the following shapes:
+        #   [batch,  heads, query_tokens, key_tokens] (e.g. torch sdp attn)
+        #   [batch * heads, query_tokens, key_tokens] (e.g. xformers or classic attn)
+        if attention_mask is not None:
+            # assume that mask is expressed as:
+            #   (1 = keep,      0 = discard)
+            # convert mask into a bias that can be added to attention scores:
+            #       (keep = +0,     discard = -10000.0)
+            attention_mask = (1 - attention_mask.to(sample.dtype)) * -10000.0
+            attention_mask = attention_mask.unsqueeze(1)
+
+        # convert encoder_attention_mask to a bias the same way we do for attention_mask
+        if encoder_attention_mask is not None:
+            encoder_attention_mask = (1 - encoder_attention_mask.to(sample.dtype)) * -10000.0
+            encoder_attention_mask = encoder_attention_mask.unsqueeze(1)
+
+        # 0. center input if necessary
+        if self.config.center_input_sample:
+            sample = 2 * sample - 1.0
+
+        # 1. time
+        t_emb = self.get_time_embed(sample=sample, timestep=timestep)
+        emb = self.time_embedding(t_emb, timestep_cond)
+        aug_emb = None
+
+        class_emb = self.get_class_embed(sample=sample, class_labels=class_labels)
+        if class_emb is not None:
+            if self.config.class_embeddings_concat:
+                emb = torch.cat([emb, class_emb], dim=-1)
+            else:
+                emb = emb + class_emb
+
+        aug_emb = self.get_aug_embed(
+            emb=emb, encoder_hidden_states=encoder_hidden_states, added_cond_kwargs=added_cond_kwargs
+        )
+        if self.config.addition_embed_type == "image_hint":
+            aug_emb, hint = aug_emb
+            sample = torch.cat([sample, hint], dim=1)
+
+        emb = emb + aug_emb if aug_emb is not None else emb
+
+        if self.time_embed_act is not None:
+            emb = self.time_embed_act(emb)
+
+        encoder_hidden_states = self.process_encoder_hidden_states(
+            encoder_hidden_states=encoder_hidden_states, added_cond_kwargs=added_cond_kwargs
+        )
+
+        # 2. pre-process
+        sample = self.conv_in(sample)
+
+        # 2.5 GLIGEN position net
+        if cross_attention_kwargs is not None and cross_attention_kwargs.get("gligen", None) is not None:
+            cross_attention_kwargs = cross_attention_kwargs.copy()
+            gligen_args = cross_attention_kwargs.pop("gligen")
+            cross_attention_kwargs["gligen"] = {"objs": self.position_net(**gligen_args)}
+
+        # 3. down
+        # we're popping the `scale` instead of getting it because otherwise `scale` will be propagated
+        # to the internal blocks and will raise deprecation warnings. this will be confusing for our users.
+        if cross_attention_kwargs is not None:
+            cross_attention_kwargs = cross_attention_kwargs.copy()
+            lora_scale = cross_attention_kwargs.pop("scale", 1.0)
+        else:
+            lora_scale = 1.0
+
+        if USE_PEFT_BACKEND:
+            # weight the lora layers by setting `lora_scale` for each PEFT layer
+            scale_lora_layers(self, lora_scale)
+
+        is_controlnet = mid_block_additional_residual is not None and down_block_additional_residuals is not None
+        # using new arg down_intrablock_additional_residuals for T2I-Adapters, to distinguish from controlnets
+        is_adapter = down_intrablock_additional_residuals is not None
+        # maintain backward compatibility for legacy usage, where
+        #       T2I-Adapter and ControlNet both use down_block_additional_residuals arg
+        #       but can only use one or the other
+        if not is_adapter and mid_block_additional_residual is None and down_block_additional_residuals is not None:
+            deprecate(
+                "T2I should not use down_block_additional_residuals",
+                "1.3.0",
+                "Passing intrablock residual connections with `down_block_additional_residuals` is deprecated \
+                       and will be removed in diffusers 1.3.0.  `down_block_additional_residuals` should only be used \
+                       for ControlNet. Please make sure use `down_intrablock_additional_residuals` instead. ",
+                standard_warn=False,
+            )
+            down_intrablock_additional_residuals = down_block_additional_residuals
+            is_adapter = True
+
+
+
+        def residual_downforward(
+            self, hidden_states: torch.Tensor, temb: Optional[torch.Tensor] = None,
+            additional_residuals: Optional[torch.Tensor] = None,
+            *args, **kwargs,
+        ) -> Tuple[torch.Tensor, Tuple[torch.Tensor, ...]]:
+            if len(args) > 0 or kwargs.get("scale", None) is not None:
+                deprecation_message = "The `scale` argument is deprecated and will be ignored. Please remove it, as passing it will raise an error in the future. `scale` should directly be passed while calling the underlying pipeline component i.e., via `cross_attention_kwargs`."
+                deprecate("scale", "1.0.0", deprecation_message)
+
+            output_states = ()
+
+            for resnet in self.resnets:
+                if self.training and self.gradient_checkpointing:
+
+                    def create_custom_forward(module):
+                        def custom_forward(*inputs):
+                            return module(*inputs)
+
+                        return custom_forward
+
+                    if is_torch_version(">=", "1.11.0"):
+                        hidden_states = torch.utils.checkpoint.checkpoint(
+                            create_custom_forward(resnet), hidden_states, temb, use_reentrant=False
+                        )
+                    else:
+                        hidden_states = torch.utils.checkpoint.checkpoint(
+                            create_custom_forward(resnet), hidden_states, temb
+                        )
+                else:
+                    hidden_states = resnet(hidden_states, temb)
+                    hidden_states += additional_residuals.pop(0)
+
+
+                output_states = output_states + (hidden_states,)
+
+            if self.downsamplers is not None:
+                for downsampler in self.downsamplers:
+                    hidden_states = downsampler(hidden_states)
+                    hidden_states += additional_residuals.pop(0)
+
+                output_states = output_states + (hidden_states,)
+
+            return hidden_states, output_states
+
+
+        def residual_blockforward(
+            self,  ## NOTE that repurpose to unet_blocks
+            hidden_states: torch.Tensor,
+            temb: Optional[torch.Tensor] = None,
+            encoder_hidden_states: Optional[torch.Tensor] = None,
+            attention_mask: Optional[torch.Tensor] = None,
+            cross_attention_kwargs: Optional[Dict[str, Any]] = None,
+            encoder_attention_mask: Optional[torch.Tensor] = None,
+            additional_residuals: Optional[torch.Tensor] = None,
+        ) -> Tuple[torch.Tensor, Tuple[torch.Tensor, ...]]:
+            if cross_attention_kwargs is not None:
+                if cross_attention_kwargs.get("scale", None) is not None:
+                    logger.warning("Passing `scale` to `cross_attention_kwargs` is deprecated. `scale` will be ignored.")
+
+
+
+            output_states = ()
+
+            blocks = list(zip(self.resnets, self.attentions))
+
+            for i, (resnet, attn) in enumerate(blocks):
+                if self.training and self.gradient_checkpointing:
+
+                    def create_custom_forward(module, return_dict=None):
+                        def custom_forward(*inputs):
+                            if return_dict is not None:
+                                return module(*inputs, return_dict=return_dict)
+                            else:
+                                return module(*inputs)
+
+                        return custom_forward
+
+                    ckpt_kwargs: Dict[str, Any] = {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
+                    hidden_states = torch.utils.checkpoint.checkpoint(
+                        create_custom_forward(resnet),
+                        hidden_states,
+                        temb,
+                        **ckpt_kwargs,
+                    )
+                    hidden_states = attn(
+                        hidden_states,
+                        encoder_hidden_states=encoder_hidden_states,
+                        cross_attention_kwargs=cross_attention_kwargs,
+                        attention_mask=attention_mask,
+                        encoder_attention_mask=encoder_attention_mask,
+                        return_dict=False,
+                    )[0]
+                else:
+                    hidden_states = resnet(hidden_states, temb)
+                    hidden_states = attn(
+                        hidden_states,
+                        encoder_hidden_states=encoder_hidden_states,
+                        cross_attention_kwargs=cross_attention_kwargs,
+                        attention_mask=attention_mask,
+                        encoder_attention_mask=encoder_attention_mask,
+                        return_dict=False,
+                    )[0]
+
+                hidden_states += additional_residuals.pop(0)
+
+                output_states = output_states + (hidden_states,)
+
+            if self.downsamplers is not None:
+                for downsampler in self.downsamplers:
+                    hidden_states = downsampler(hidden_states)
+                    hidden_states += additional_residuals.pop(0)
+
+                output_states = output_states + (hidden_states,)
+
+            return hidden_states, output_states
+
+
+        down_intrablock_additional_residuals = dino_down_block_additional_residuals
+
+        sample += down_intrablock_additional_residuals.pop(0)
+        down_block_res_samples = (sample,)
+
+        for downsample_block in self.down_blocks:
+            if hasattr(downsample_block, "has_cross_attention") and downsample_block.has_cross_attention:
+
+                sample, res_samples = residual_blockforward(
+                    downsample_block,
+                    hidden_states=sample,
+                    temb=emb,
+                    encoder_hidden_states=encoder_hidden_states,
+                    attention_mask=attention_mask,
+                    cross_attention_kwargs=cross_attention_kwargs,
+                    encoder_attention_mask=encoder_attention_mask,
+                    additional_residuals = down_intrablock_additional_residuals,
+                )
+
+            else:
+                sample, res_samples = residual_downforward(
+                    downsample_block,
+                    hidden_states=sample,
+                    temb=emb,
+                    additional_residuals = down_intrablock_additional_residuals,
+                        )
+
+
+            down_block_res_samples += res_samples
+
+
+        if is_controlnet:
+            new_down_block_res_samples = ()
+
+            for down_block_res_sample, down_block_additional_residual in zip(
+                down_block_res_samples, down_block_additional_residuals
+            ):
+                down_block_res_sample = down_block_res_sample + down_block_additional_residual
+                new_down_block_res_samples = new_down_block_res_samples + (down_block_res_sample,)
+
+            down_block_res_samples = new_down_block_res_samples
+
+        # 4. mid
+        if self.mid_block is not None:
+            if hasattr(self.mid_block, "has_cross_attention") and self.mid_block.has_cross_attention:
+                sample = self.mid_block(
+                    sample,
+                    emb,
+                    encoder_hidden_states=encoder_hidden_states,
+                    attention_mask=attention_mask,
+                    cross_attention_kwargs=cross_attention_kwargs,
+                    encoder_attention_mask=encoder_attention_mask,
+                )
+            else:
+                sample = self.mid_block(sample, emb)
+
+            # To support T2I-Adapter-XL
+            if (
+                is_adapter
+                and len(down_intrablock_additional_residuals) > 0
+                and sample.shape == down_intrablock_additional_residuals[0].shape
+            ):
+                sample += down_intrablock_additional_residuals.pop(0)
+
+        if is_controlnet:
+            sample = sample + mid_block_additional_residual
+
+        # 5. up
+        for i, upsample_block in enumerate(self.up_blocks):
+            is_final_block = i == len(self.up_blocks) - 1
+
+            res_samples = down_block_res_samples[-len(upsample_block.resnets) :]
+            down_block_res_samples = down_block_res_samples[: -len(upsample_block.resnets)]
+
+            # if we have not reached the final block and need to forward the
+            # upsample size, we do it here
+            if not is_final_block and forward_upsample_size:
+                upsample_size = down_block_res_samples[-1].shape[2:]
+
+            if hasattr(upsample_block, "has_cross_attention") and upsample_block.has_cross_attention:
+                sample = upsample_block(
+                    hidden_states=sample,
+                    temb=emb,
+                    res_hidden_states_tuple=res_samples,
+                    encoder_hidden_states=encoder_hidden_states,
+                    cross_attention_kwargs=cross_attention_kwargs,
+                    upsample_size=upsample_size,
+                    attention_mask=attention_mask,
+                    encoder_attention_mask=encoder_attention_mask,
+                )
+            else:
+                sample = upsample_block(
+                    hidden_states=sample,
+                    temb=emb,
+                    res_hidden_states_tuple=res_samples,
+                    upsample_size=upsample_size,
+                )
+
+        # 6. post-process
+        if self.conv_norm_out:
+            sample = self.conv_norm_out(sample)
+            sample = self.conv_act(sample)
+        sample = self.conv_out(sample)
+
+        if USE_PEFT_BACKEND:
+            # remove `lora_scale` from each PEFT layer
+            unscale_lora_layers(self, lora_scale)
+
+        if not return_dict:
+            return (sample,)
+
+        return UNet2DConditionOutput(sample=sample)
+
+
+
+    @staticmethod
+    def ensemble_normals(
+        normals: torch.Tensor, output_uncertainty: bool, reduction: str = "closest"
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
+        """
+        Ensembles the normals maps represented by the `normals` tensor with expected shape `(B, 3, H, W)`, where B is
+        the number of ensemble members for a given prediction of size `(H x W)`.
+
+        Args:
+            normals (`torch.Tensor`):
+                Input ensemble normals maps.
+            output_uncertainty (`bool`, *optional*, defaults to `False`):
+                Whether to output uncertainty map.
+            reduction (`str`, *optional*, defaults to `"closest"`):
+                Reduction method used to ensemble aligned predictions. The accepted values are: `"closest"` and
+                `"mean"`.
+
+        Returns:
+            A tensor of aligned and ensembled normals maps with shape `(1, 3, H, W)` and optionally a tensor of
+            uncertainties of shape `(1, 1, H, W)`.
+        """
+        if normals.dim() != 4 or normals.shape[1] != 3:
+            raise ValueError(f"Expecting 4D tensor of shape [B,3,H,W]; got {normals.shape}.")
+        if reduction not in ("closest", "mean"):
+            raise ValueError(f"Unrecognized reduction method: {reduction}.")
+
+        mean_normals = normals.mean(dim=0, keepdim=True)  # [1,3,H,W]
+        mean_normals = MarigoldNormalsPipeline.normalize_normals(mean_normals)  # [1,3,H,W]
+
+        sim_cos = (mean_normals * normals).sum(dim=1, keepdim=True)  # [E,1,H,W]
+        sim_cos = sim_cos.clamp(-1, 1)  # required to avoid NaN in uncertainty with fp16
+
+        uncertainty = None
+        if output_uncertainty:
+            uncertainty = sim_cos.arccos()  # [E,1,H,W]
+            uncertainty = uncertainty.mean(dim=0, keepdim=True) / np.pi  # [1,1,H,W]
+
+        if reduction == "mean":
+            return mean_normals, uncertainty  # [1,3,H,W], [1,1,H,W]
+
+        closest_indices = sim_cos.argmax(dim=0, keepdim=True)  # [1,1,H,W]
+        closest_indices = closest_indices.repeat(1, 3, 1, 1)  # [1,3,H,W]
+        closest_normals = torch.gather(normals, 0, closest_indices)  # [1,3,H,W]
+
+        return closest_normals, uncertainty  # [1,3,H,W], [1,1,H,W]
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    """
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps

stablenormal/pipeline_yoso_normal.py

@@ -0,0 +1,723 @@
+# Copyright 2024 Marigold authors, PRS ETH Zurich. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# --------------------------------------------------------------------------
+# More information and citation instructions are available on the
+# --------------------------------------------------------------------------
+from dataclasses import dataclass
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import numpy as np
+import torch
+from PIL import Image
+from tqdm.auto import tqdm
+from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
+
+
+from diffusers.image_processor import PipelineImageInput
+from diffusers.models import (
+    AutoencoderKL,
+    UNet2DConditionModel,
+	ControlNetModel,
+)
+from diffusers.schedulers import (
+	DDIMScheduler
+)
+
+from diffusers.utils import (
+    BaseOutput,
+    logging,
+    replace_example_docstring,
+)
+
+
+from diffusers.utils.torch_utils import randn_tensor
+from diffusers.pipelines.controlnet import StableDiffusionControlNetPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.marigold.marigold_image_processing import MarigoldImageProcessor
+from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
+
+import pdb
+
+
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+EXAMPLE_DOC_STRING = """
+Examples:
+```py
+>>> import diffusers
+>>> import torch
+
+>>> pipe = diffusers.MarigoldNormalsPipeline.from_pretrained(
+...     "prs-eth/marigold-normals-lcm-v0-1", variant="fp16", torch_dtype=torch.float16
+... ).to("cuda")
+
+>>> image = diffusers.utils.load_image("https://marigoldmonodepth.github.io/images/einstein.jpg")
+>>> normals = pipe(image)
+
+>>> vis = pipe.image_processor.visualize_normals(normals.prediction)
+>>> vis[0].save("einstein_normals.png")
+```
+"""
+
+
+@dataclass
+class YosoNormalsOutput(BaseOutput):
+    """
+    Output class for Marigold monocular normals prediction pipeline.
+
+    Args:
+        prediction (`np.ndarray`, `torch.Tensor`):
+            Predicted normals with values in the range [-1, 1]. The shape is always $numimages \times 3 \times height
+            \times width$, regardless of whether the images were passed as a 4D array or a list.
+        uncertainty (`None`, `np.ndarray`, `torch.Tensor`):
+            Uncertainty maps computed from the ensemble, with values in the range [0, 1]. The shape is $numimages
+            \times 1 \times height \times width$.
+        latent (`None`, `torch.Tensor`):
+            Latent features corresponding to the predictions, compatible with the `latents` argument of the pipeline.
+            The shape is $numimages * numensemble \times 4 \times latentheight \times latentwidth$.
+    """
+
+    prediction: Union[np.ndarray, torch.Tensor]
+    latent: Union[None, torch.Tensor]
+    gauss_latent: Union[None, torch.Tensor]
+
+
+class YOSONormalsPipeline(StableDiffusionControlNetPipeline):
+    """ Pipeline for monocular normals estimation using the Marigold method: https://marigoldmonodepth.github.io.
+    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
+    implemented for all pipelines (downloading, saving, running on a particular device, etc.).
+
+    The pipeline also inherits the following loading methods:
+        - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
+        - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
+
+    Args:
+        vae ([`AutoencoderKL`]):
+            Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
+        text_encoder ([`~transformers.CLIPTextModel`]):
+            Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)).
+        tokenizer ([`~transformers.CLIPTokenizer`]):
+            A `CLIPTokenizer` to tokenize text.
+        unet ([`UNet2DConditionModel`]):
+            A `UNet2DConditionModel` 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 more details
+            about a model's potential harms.
+        feature_extractor ([`~transformers.CLIPImageProcessor`]):
+            A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
+    """
+
+    model_cpu_offload_seq = "text_encoder->image_encoder->unet->vae"
+    _optional_components = ["safety_checker", "feature_extractor", "image_encoder"]
+    _exclude_from_cpu_offload = ["safety_checker"]
+    _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
+
+
+
+    def __init__(
+        self,
+        vae: AutoencoderKL,
+        text_encoder: CLIPTextModel,
+        tokenizer: CLIPTokenizer,
+        unet: UNet2DConditionModel,
+        controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel]],
+        scheduler: Union[DDIMScheduler],
+        safety_checker: StableDiffusionSafetyChecker,
+        feature_extractor: CLIPImageProcessor,
+        image_encoder: CLIPVisionModelWithProjection = None,
+        requires_safety_checker: bool = True,
+        default_denoising_steps: Optional[int] = 1,
+		default_processing_resolution: Optional[int] = 768,
+        prompt="",
+        empty_text_embedding=None,
+        t_start: Optional[int] = 401,
+    ):
+        super().__init__(
+            vae,
+            text_encoder,
+            tokenizer,
+            unet,
+            controlnet,
+            scheduler,
+            safety_checker,
+            feature_extractor,
+            image_encoder,
+            requires_safety_checker,
+                )
+
+        # TODO yoso ImageProcessor
+        self.image_processor = MarigoldImageProcessor(vae_scale_factor=self.vae_scale_factor)
+        self.control_image_processor = MarigoldImageProcessor(vae_scale_factor=self.vae_scale_factor)
+        self.default_denoising_steps = default_denoising_steps
+        self.default_processing_resolution = default_processing_resolution
+        self.prompt = prompt
+        self.prompt_embeds = None
+        self.empty_text_embedding = empty_text_embedding
+        self.t_start= t_start # target_out latents
+        self.gauss_latent = None
+
+    def check_inputs(
+        self,
+        image: PipelineImageInput,
+        num_inference_steps: int,
+        ensemble_size: int,
+        processing_resolution: int,
+        resample_method_input: str,
+        resample_method_output: str,
+        batch_size: int,
+        ensembling_kwargs: Optional[Dict[str, Any]],
+        latents: Optional[torch.Tensor],
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]],
+        output_type: str,
+        output_uncertainty: bool,
+    ) -> int:
+        if num_inference_steps is None:
+            raise ValueError("`num_inference_steps` is not specified and could not be resolved from the model config.")
+        if num_inference_steps < 1:
+            raise ValueError("`num_inference_steps` must be positive.")
+        if ensemble_size < 1:
+            raise ValueError("`ensemble_size` must be positive.")
+        if ensemble_size == 2:
+            logger.warning(
+                "`ensemble_size` == 2 results are similar to no ensembling (1); "
+                "consider increasing the value to at least 3."
+            )
+        if ensemble_size == 1 and output_uncertainty:
+            raise ValueError(
+                "Computing uncertainty by setting `output_uncertainty=True` also requires setting `ensemble_size` "
+                "greater than 1."
+            )
+        if processing_resolution is None:
+            raise ValueError(
+                "`processing_resolution` is not specified and could not be resolved from the model config."
+            )
+        if processing_resolution < 0:
+            raise ValueError(
+                "`processing_resolution` must be non-negative: 0 for native resolution, or any positive value for "
+                "downsampled processing."
+            )
+        if processing_resolution % self.vae_scale_factor != 0:
+            raise ValueError(f"`processing_resolution` must be a multiple of {self.vae_scale_factor}.")
+        if resample_method_input not in ("nearest", "nearest-exact", "bilinear", "bicubic", "area"):
+            raise ValueError(
+                "`resample_method_input` takes string values compatible with PIL library: "
+                "nearest, nearest-exact, bilinear, bicubic, area."
+            )
+        if resample_method_output not in ("nearest", "nearest-exact", "bilinear", "bicubic", "area"):
+            raise ValueError(
+                "`resample_method_output` takes string values compatible with PIL library: "
+                "nearest, nearest-exact, bilinear, bicubic, area."
+            )
+        if batch_size < 1:
+            raise ValueError("`batch_size` must be positive.")
+        if output_type not in ["pt", "np"]:
+            raise ValueError("`output_type` must be one of `pt` or `np`.")
+        if latents is not None and generator is not None:
+            raise ValueError("`latents` and `generator` cannot be used together.")
+        if ensembling_kwargs is not None:
+            if not isinstance(ensembling_kwargs, dict):
+                raise ValueError("`ensembling_kwargs` must be a dictionary.")
+            if "reduction" in ensembling_kwargs and ensembling_kwargs["reduction"] not in ("closest", "mean"):
+                raise ValueError("`ensembling_kwargs['reduction']` can be either `'closest'` or `'mean'`.")
+
+        # image checks
+        num_images = 0
+        W, H = None, None
+        if not isinstance(image, list):
+            image = [image]
+        for i, img in enumerate(image):
+            if isinstance(img, np.ndarray) or torch.is_tensor(img):
+                if img.ndim not in (2, 3, 4):
+                    raise ValueError(f"`image[{i}]` has unsupported dimensions or shape: {img.shape}.")
+                H_i, W_i = img.shape[-2:]
+                N_i = 1
+                if img.ndim == 4:
+                    N_i = img.shape[0]
+            elif isinstance(img, Image.Image):
+                W_i, H_i = img.size
+                N_i = 1
+            else:
+                raise ValueError(f"Unsupported `image[{i}]` type: {type(img)}.")
+            if W is None:
+                W, H = W_i, H_i
+            elif (W, H) != (W_i, H_i):
+                raise ValueError(
+                    f"Input `image[{i}]` has incompatible dimensions {(W_i, H_i)} with the previous images {(W, H)}"
+                )
+            num_images += N_i
+
+        # latents checks
+        if latents is not None:
+            if not torch.is_tensor(latents):
+                raise ValueError("`latents` must be a torch.Tensor.")
+            if latents.dim() != 4:
+                raise ValueError(f"`latents` has unsupported dimensions or shape: {latents.shape}.")
+
+            if processing_resolution > 0:
+                max_orig = max(H, W)
+                new_H = H * processing_resolution // max_orig
+                new_W = W * processing_resolution // max_orig
+                if new_H == 0 or new_W == 0:
+                    raise ValueError(f"Extreme aspect ratio of the input image: [{W} x {H}]")
+                W, H = new_W, new_H
+            w = (W + self.vae_scale_factor - 1) // self.vae_scale_factor
+            h = (H + self.vae_scale_factor - 1) // self.vae_scale_factor
+            shape_expected = (num_images * ensemble_size, self.vae.config.latent_channels, h, w)
+
+            if latents.shape != shape_expected:
+                raise ValueError(f"`latents` has unexpected shape={latents.shape} expected={shape_expected}.")
+
+        # generator checks
+        if generator is not None:
+            if isinstance(generator, list):
+                if len(generator) != num_images * ensemble_size:
+                    raise ValueError(
+                        "The number of generators must match the total number of ensemble members for all input images."
+                    )
+                if not all(g.device.type == generator[0].device.type for g in generator):
+                    raise ValueError("`generator` device placement is not consistent in the list.")
+            elif not isinstance(generator, torch.Generator):
+                raise ValueError(f"Unsupported generator type: {type(generator)}.")
+
+        return num_images
+
+    def progress_bar(self, iterable=None, total=None, desc=None, leave=True):
+        if not hasattr(self, "_progress_bar_config"):
+            self._progress_bar_config = {}
+        elif not isinstance(self._progress_bar_config, dict):
+            raise ValueError(
+                f"`self._progress_bar_config` should be of type `dict`, but is {type(self._progress_bar_config)}."
+            )
+
+        progress_bar_config = dict(**self._progress_bar_config)
+        progress_bar_config["desc"] = progress_bar_config.get("desc", desc)
+        progress_bar_config["leave"] = progress_bar_config.get("leave", leave)
+        if iterable is not None:
+            return tqdm(iterable, **progress_bar_config)
+        elif total is not None:
+            return tqdm(total=total, **progress_bar_config)
+        else:
+            raise ValueError("Either `total` or `iterable` has to be defined.")
+
+    @torch.no_grad()
+    @replace_example_docstring(EXAMPLE_DOC_STRING)
+    def __call__(
+        self,
+        image: PipelineImageInput,
+        prompt: Union[str, List[str]] = None,
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        num_inference_steps: Optional[int] = None,
+        ensemble_size: int = 1,
+        processing_resolution: Optional[int] = None,
+        match_input_resolution: bool = True,
+        resample_method_input: str = "bilinear",
+        resample_method_output: str = "bilinear",
+        batch_size: int = 1,
+        ensembling_kwargs: Optional[Dict[str, Any]] = None,
+        latents: Optional[Union[torch.Tensor, List[torch.Tensor]]] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        num_images_per_prompt: Optional[int] = 1,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        controlnet_conditioning_scale: Union[float, List[float]] = 1.0,
+        output_type: str = "np",
+        output_uncertainty: bool = False,
+        output_latent: bool = False,
+        skip_preprocess: bool = False,
+        return_dict: bool = True,
+        **kwargs,
+    ):
+        """
+        Function invoked when calling the pipeline.
+
+        Args:
+            image (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`),
+                `List[torch.Tensor]`: An input image or images used as an input for the normals estimation task. For
+                arrays and tensors, the expected value range is between `[0, 1]`. Passing a batch of images is possible
+                by providing a four-dimensional array or a tensor. Additionally, a list of images of two- or
+                three-dimensional arrays or tensors can be passed. In the latter case, all list elements must have the
+                same width and height.
+            num_inference_steps (`int`, *optional*, defaults to `None`):
+                Number of denoising diffusion steps during inference. The default value `None` results in automatic
+                selection. The number of steps should be at least 10 with the full Marigold models, and between 1 and 4
+                for Marigold-LCM models.
+            ensemble_size (`int`, defaults to `1`):
+                Number of ensemble predictions. Recommended values are 5 and higher for better precision, or 1 for
+                faster inference.
+            processing_resolution (`int`, *optional*, defaults to `None`):
+                Effective processing resolution. When set to `0`, matches the larger input image dimension. This
+                produces crisper predictions, but may also lead to the overall loss of global context. The default
+                value `None` resolves to the optimal value from the model config.
+            match_input_resolution (`bool`, *optional*, defaults to `True`):
+                When enabled, the output prediction is resized to match the input dimensions. When disabled, the longer
+                side of the output will equal to `processing_resolution`.
+            resample_method_input (`str`, *optional*, defaults to `"bilinear"`):
+                Resampling method used to resize input images to `processing_resolution`. The accepted values are:
+                `"nearest"`, `"nearest-exact"`, `"bilinear"`, `"bicubic"`, or `"area"`.
+            resample_method_output (`str`, *optional*, defaults to `"bilinear"`):
+                Resampling method used to resize output predictions to match the input resolution. The accepted values
+                are `"nearest"`, `"nearest-exact"`, `"bilinear"`, `"bicubic"`, or `"area"`.
+            batch_size (`int`, *optional*, defaults to `1`):
+                Batch size; only matters when setting `ensemble_size` or passing a tensor of images.
+            ensembling_kwargs (`dict`, *optional*, defaults to `None`)
+                Extra dictionary with arguments for precise ensembling control. The following options are available:
+                - reduction (`str`, *optional*, defaults to `"closest"`): Defines the ensembling function applied in
+                  every pixel location, can be either `"closest"` or `"mean"`.
+            latents (`torch.Tensor`, *optional*, defaults to `None`):
+                Latent noise tensors to replace the random initialization. These can be taken from the previous
+                function call's output.
+            generator (`torch.Generator`, or `List[torch.Generator]`, *optional*, defaults to `None`):
+                Random number generator object to ensure reproducibility.
+            output_type (`str`, *optional*, defaults to `"np"`):
+                Preferred format of the output's `prediction` and the optional `uncertainty` fields. The accepted
+                values are: `"np"` (numpy array) or `"pt"` (torch tensor).
+            output_uncertainty (`bool`, *optional*, defaults to `False`):
+                When enabled, the output's `uncertainty` field contains the predictive uncertainty map, provided that
+                the `ensemble_size` argument is set to a value above 2.
+            output_latent (`bool`, *optional*, defaults to `False`):
+                When enabled, the output's `latent` field contains the latent codes corresponding to the predictions
+                within the ensemble. These codes can be saved, modified, and used for subsequent calls with the
+                `latents` argument.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~pipelines.marigold.MarigoldDepthOutput`] instead of a plain tuple.
+
+        Examples:
+
+        Returns:
+            [`~pipelines.marigold.MarigoldNormalsOutput`] or `tuple`:
+                If `return_dict` is `True`, [`~pipelines.marigold.MarigoldNormalsOutput`] is returned, otherwise a
+                `tuple` is returned where the first element is the prediction, the second element is the uncertainty
+                (or `None`), and the third is the latent (or `None`).
+        """
+
+        # 0. Resolving variables.
+        device = self._execution_device
+        dtype = self.dtype
+
+        # Model-specific optimal default values leading to fast and reasonable results.
+        if num_inference_steps is None:
+            num_inference_steps = self.default_denoising_steps
+        if processing_resolution is None:
+            processing_resolution = self.default_processing_resolution
+
+        # 1. Check inputs.
+        num_images = self.check_inputs(
+            image,
+            num_inference_steps,
+            ensemble_size,
+            processing_resolution,
+            resample_method_input,
+            resample_method_output,
+            batch_size,
+            ensembling_kwargs,
+            latents,
+            generator,
+            output_type,
+            output_uncertainty,
+        )
+
+
+        # 2. Prepare empty text conditioning.
+        # Model invocation: self.tokenizer, self.text_encoder.
+        if self.empty_text_embedding is None:
+            prompt = ""
+            text_inputs = self.tokenizer(
+                prompt,
+                padding="do_not_pad",
+                max_length=self.tokenizer.model_max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+            text_input_ids = text_inputs.input_ids.to(device)
+            self.empty_text_embedding = self.text_encoder(text_input_ids)[0]  # [1,2,1024]
+
+
+
+        # 3. prepare prompt
+        if self.prompt_embeds is None:
+            prompt_embeds, negative_prompt_embeds = self.encode_prompt(
+                self.prompt,
+                device,
+                num_images_per_prompt,
+                False,
+                negative_prompt,
+                prompt_embeds=prompt_embeds,
+                negative_prompt_embeds=None,
+                lora_scale=None,
+                clip_skip=None,
+            )
+            self.prompt_embeds = prompt_embeds
+            self.negative_prompt_embeds = negative_prompt_embeds
+
+
+
+        # 4. Preprocess input images. This function loads input image or images of compatible dimensions `(H, W)`,
+        # optionally downsamples them to the `processing_resolution` `(PH, PW)`, where
+        # `max(PH, PW) == processing_resolution`, and pads the dimensions to `(PPH, PPW)` such that these values are
+        # divisible by the latent space downscaling factor (typically 8 in Stable Diffusion). The default value `None`
+        # of `processing_resolution` resolves to the optimal value from the model config. It is a recommended mode of
+        # operation and leads to the most reasonable results. Using the native image resolution or any other processing
+        # resolution can lead to loss of either fine details or global context in the output predictions.
+        if not skip_preprocess:
+            image, padding, original_resolution = self.image_processor.preprocess(
+                image, processing_resolution, resample_method_input, device, dtype
+            )  # [N,3,PPH,PPW]
+        else:
+            padding = (0, 0)
+            original_resolution = image.shape[2:]
+        # 5. Encode input image into latent space. At this step, each of the `N` input images is represented with `E`
+        # ensemble members. Each ensemble member is an independent diffused prediction, just initialized independently.
+        # Latents of each such predictions across all input images and all ensemble members are represented in the
+        # `pred_latent` variable. The variable `image_latent` is of the same shape: it contains each input image encoded
+        # into latent space and replicated `E` times. The latents can be either generated (see `generator` to ensure
+        # reproducibility), or passed explicitly via the `latents` argument. The latter can be set outside the pipeline
+        # code. For example, in the Marigold-LCM video processing demo, the latents initialization of a frame is taken
+        # as a convex combination of the latents output of the pipeline for the previous frame and a newly-sampled
+        # noise. This behavior can be achieved by setting the `output_latent` argument to `True`. The latent space
+        # dimensions are `(h, w)`. Encoding into latent space happens in batches of size `batch_size`.
+        # Model invocation: self.vae.encoder.
+        image_latent, gauss_latent = self.prepare_latents(
+            image, latents, generator, ensemble_size, batch_size
+        )  # [N*E,4,h,w], [N*E,4,h,w]
+
+        del image
+
+
+        # 6. obtain control_output
+
+        cond_scale =controlnet_conditioning_scale
+        down_block_res_samples, mid_block_res_sample = self.controlnet(
+            image_latent.detach(),
+            self.t_start,
+            encoder_hidden_states=self.prompt_embeds,
+            conditioning_scale=cond_scale,
+            guess_mode=False,
+            return_dict=False,
+        )
+
+
+        # 7. YOSO sampling
+        latent_x_t = self.unet(
+            gauss_latent,
+            self.t_start,
+            encoder_hidden_states=self.prompt_embeds,
+            down_block_additional_residuals=down_block_res_samples,
+            mid_block_additional_residual=mid_block_res_sample,
+            return_dict=False,
+        )[0]
+
+
+        del (
+            image_latent,
+        )
+
+        # decoder
+        prediction = self.decode_prediction(latent_x_t)
+        prediction = self.image_processor.unpad_image(prediction, padding)  # [N*E,3,PH,PW]
+
+        prediction = self.image_processor.resize_antialias(
+            prediction, original_resolution, resample_method_output, is_aa=False
+        )  # [N,3,H,W]
+        prediction = self.normalize_normals(prediction)  # [N,3,H,W]
+
+        if output_type == "np":
+            prediction = self.image_processor.pt_to_numpy(prediction)  # [N,H,W,3]
+
+        # 11. Offload all models
+        self.maybe_free_model_hooks()
+
+        return YosoNormalsOutput(
+            prediction=prediction,
+            latent=latent_x_t,
+            gauss_latent=gauss_latent,
+        )
+
+    # Copied from diffusers.pipelines.marigold.pipeline_marigold_depth.MarigoldDepthPipeline.prepare_latents
+    def prepare_latents(
+        self,
+        image: torch.Tensor,
+        latents: Optional[torch.Tensor],
+        generator: Optional[torch.Generator],
+        ensemble_size: int,
+        batch_size: int,
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        def retrieve_latents(encoder_output):
+            if hasattr(encoder_output, "latent_dist"):
+                return encoder_output.latent_dist.mode()
+            elif hasattr(encoder_output, "latents"):
+                return encoder_output.latents
+            else:
+                raise AttributeError("Could not access latents of provided encoder_output")
+
+
+
+        image_latent = torch.cat(
+            [
+                retrieve_latents(self.vae.encode(image[i : i + batch_size]))
+                for i in range(0, image.shape[0], batch_size)
+            ],
+            dim=0,
+        )  # [N,4,h,w]
+        image_latent = image_latent * self.vae.config.scaling_factor
+        image_latent = image_latent.repeat_interleave(ensemble_size, dim=0)  # [N*E,4,h,w]
+
+        pred_latent = self.gauss_latent
+        if pred_latent is None:
+            self.gauss_latent = torch.randn_like(image_latent)
+            pred_latent = self.gauss_latent
+
+        return image_latent, pred_latent
+
+    def decode_prediction(self, pred_latent: torch.Tensor) -> torch.Tensor:
+        if pred_latent.dim() != 4 or pred_latent.shape[1] != self.vae.config.latent_channels:
+            raise ValueError(
+                f"Expecting 4D tensor of shape [B,{self.vae.config.latent_channels},H,W]; got {pred_latent.shape}."
+            )
+
+        prediction = self.vae.decode(pred_latent / self.vae.config.scaling_factor, return_dict=False)[0]  # [B,3,H,W]
+
+        prediction = self.normalize_normals(prediction)  # [B,3,H,W]
+
+        return prediction  # [B,3,H,W]
+
+    @staticmethod
+    def normalize_normals(normals: torch.Tensor, eps: float = 1e-6) -> torch.Tensor:
+        if normals.dim() != 4 or normals.shape[1] != 3:
+            raise ValueError(f"Expecting 4D tensor of shape [B,3,H,W]; got {normals.shape}.")
+
+        norm = torch.norm(normals, dim=1, keepdim=True)
+        normals /= norm.clamp(min=eps)
+
+        return normals
+
+    @staticmethod
+    def ensemble_normals(
+        normals: torch.Tensor, output_uncertainty: bool, reduction: str = "closest"
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
+        """
+        Ensembles the normals maps represented by the `normals` tensor with expected shape `(B, 3, H, W)`, where B is
+        the number of ensemble members for a given prediction of size `(H x W)`.
+
+        Args:
+            normals (`torch.Tensor`):
+                Input ensemble normals maps.
+            output_uncertainty (`bool`, *optional*, defaults to `False`):
+                Whether to output uncertainty map.
+            reduction (`str`, *optional*, defaults to `"closest"`):
+                Reduction method used to ensemble aligned predictions. The accepted values are: `"closest"` and
+                `"mean"`.
+
+        Returns:
+            A tensor of aligned and ensembled normals maps with shape `(1, 3, H, W)` and optionally a tensor of
+            uncertainties of shape `(1, 1, H, W)`.
+        """
+        if normals.dim() != 4 or normals.shape[1] != 3:
+            raise ValueError(f"Expecting 4D tensor of shape [B,3,H,W]; got {normals.shape}.")
+        if reduction not in ("closest", "mean"):
+            raise ValueError(f"Unrecognized reduction method: {reduction}.")
+
+        mean_normals = normals.mean(dim=0, keepdim=True)  # [1,3,H,W]
+        mean_normals = MarigoldNormalsPipeline.normalize_normals(mean_normals)  # [1,3,H,W]
+
+        sim_cos = (mean_normals * normals).sum(dim=1, keepdim=True)  # [E,1,H,W]
+        sim_cos = sim_cos.clamp(-1, 1)  # required to avoid NaN in uncertainty with fp16
+
+        uncertainty = None
+        if output_uncertainty:
+            uncertainty = sim_cos.arccos()  # [E,1,H,W]
+            uncertainty = uncertainty.mean(dim=0, keepdim=True) / np.pi  # [1,1,H,W]
+
+        if reduction == "mean":
+            return mean_normals, uncertainty  # [1,3,H,W], [1,1,H,W]
+
+        closest_indices = sim_cos.argmax(dim=0, keepdim=True)  # [1,1,H,W]
+        closest_indices = closest_indices.repeat(1, 3, 1, 1)  # [1,3,H,W]
+        closest_normals = torch.gather(normals, 0, closest_indices)  # [1,3,H,W]
+
+        return closest_normals, uncertainty  # [1,3,H,W], [1,1,H,W]
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    """
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps

stablenormal/stablecontrolnet.py

@@ -0,0 +1,1354 @@
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import inspect
+from typing import 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, CLIPVisionModelWithProjection
+
+from ...callbacks import MultiPipelineCallbacks, PipelineCallback
+from ...image_processor import PipelineImageInput, VaeImageProcessor
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
+from ...models import AutoencoderKL, ControlNetModel, ImageProjection, UNet2DConditionModel
+from ...models.lora import adjust_lora_scale_text_encoder
+from ...schedulers import KarrasDiffusionSchedulers
+from ...utils import (
+    USE_PEFT_BACKEND,
+    deprecate,
+    logging,
+    replace_example_docstring,
+    scale_lora_layers,
+    unscale_lora_layers,
+)
+from ...utils.torch_utils import is_compiled_module, is_torch_version, randn_tensor
+from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
+from ..stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
+from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
+from .multicontrolnet import MultiControlNetModel
+
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+EXAMPLE_DOC_STRING = """
+    Examples:
+        ```py
+        >>> # !pip install opencv-python transformers accelerate
+        >>> from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultistepScheduler
+        >>> from diffusers.utils import load_image
+        >>> import numpy as np
+        >>> import torch
+
+        >>> import cv2
+        >>> from PIL import Image
+
+        >>> # download an image
+        >>> image = load_image(
+        ...     "https://hf.co/datasets/huggingface/documentation-images/resolve/main/diffusers/input_image_vermeer.png"
+        ... )
+        >>> image = np.array(image)
+
+        >>> # get canny image
+        >>> image = cv2.Canny(image, 100, 200)
+        >>> image = image[:, :, None]
+        >>> image = np.concatenate([image, image, image], axis=2)
+        >>> canny_image = Image.fromarray(image)
+
+        >>> # load control net and stable diffusion v1-5
+        >>> controlnet = ControlNetModel.from_pretrained("lllyasviel/sd-controlnet-canny", torch_dtype=torch.float16)
+        >>> pipe = StableDiffusionControlNetPipeline.from_pretrained(
+        ...     "runwayml/stable-diffusion-v1-5", controlnet=controlnet, torch_dtype=torch.float16
+        ... )
+
+        >>> # speed up diffusion process with faster scheduler and memory optimization
+        >>> pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)
+        >>> # remove following line if xformers is not installed
+        >>> pipe.enable_xformers_memory_efficient_attention()
+
+        >>> pipe.enable_model_cpu_offload()
+
+        >>> # generate image
+        >>> generator = torch.manual_seed(0)
+        >>> image = pipe(
+        ...     "futuristic-looking woman", num_inference_steps=20, generator=generator, image=canny_image
+        ... ).images[0]
+        ```
+"""
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    """
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+
+class StableDiffusionControlNetPipeline(
+    DiffusionPipeline,
+    StableDiffusionMixin,
+    TextualInversionLoaderMixin,
+    LoraLoaderMixin,
+    IPAdapterMixin,
+    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
+    implemented for all pipelines (downloading, saving, running on a particular device, etc.).
+
+    The pipeline also inherits the following loading methods:
+        - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
+        - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
+
+    Args:
+        vae ([`AutoencoderKL`]):
+            Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
+        text_encoder ([`~transformers.CLIPTextModel`]):
+            Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)).
+        tokenizer ([`~transformers.CLIPTokenizer`]):
+            A `CLIPTokenizer` to tokenize text.
+        unet ([`UNet2DConditionModel`]):
+            A `UNet2DConditionModel` 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 more details
+            about a model's potential harms.
+        feature_extractor ([`~transformers.CLIPImageProcessor`]):
+            A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
+    """
+
+    model_cpu_offload_seq = "text_encoder->image_encoder->unet->vae"
+    _optional_components = ["safety_checker", "feature_extractor", "image_encoder"]
+    _exclude_from_cpu_offload = ["safety_checker"]
+    _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
+
+    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,
+        image_encoder: CLIPVisionModelWithProjection = None,
+        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,
+            image_encoder=image_encoder,
+        )
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True)
+        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._encode_prompt
+    def _encode_prompt(
+        self,
+        prompt,
+        device,
+        num_images_per_prompt,
+        do_classifier_free_guidance,
+        negative_prompt=None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        lora_scale: Optional[float] = None,
+        **kwargs,
+    ):
+        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
+        deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
+
+        prompt_embeds_tuple = self.encode_prompt(
+            prompt=prompt,
+            device=device,
+            num_images_per_prompt=num_images_per_prompt,
+            do_classifier_free_guidance=do_classifier_free_guidance,
+            negative_prompt=negative_prompt,
+            prompt_embeds=prompt_embeds,
+            negative_prompt_embeds=negative_prompt_embeds,
+            lora_scale=lora_scale,
+            **kwargs,
+        )
+
+        # concatenate for backwards comp
+        prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
+
+        return prompt_embeds
+
+    # 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.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        lora_scale: Optional[float] = None,
+        clip_skip: Optional[int] = 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.Tensor`, *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.Tensor`, *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.
+            clip_skip (`int`, *optional*):
+                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
+                the output of the pre-final layer will be used for computing the prompt embeddings.
+        """
+        # 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
+
+            # dynamically adjust the LoRA scale
+            if not USE_PEFT_BACKEND:
+                adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
+            else:
+                scale_lora_layers(self.text_encoder, 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: process 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
+
+            if clip_skip is None:
+                prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
+                prompt_embeds = prompt_embeds[0]
+            else:
+                prompt_embeds = self.text_encoder(
+                    text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
+                )
+                # Access the `hidden_states` first, that contains a tuple of
+                # all the hidden states from the encoder layers. Then index into
+                # the tuple to access the hidden states from the desired layer.
+                prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
+                # We also need to apply the final LayerNorm here to not mess with the
+                # representations. The `last_hidden_states` that we typically use for
+                # obtaining the final prompt representations passes through the LayerNorm
+                # layer.
+                prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
+
+        if self.text_encoder is not None:
+            prompt_embeds_dtype = self.text_encoder.dtype
+        elif self.unet is not None:
+            prompt_embeds_dtype = self.unet.dtype
+        else:
+            prompt_embeds_dtype = prompt_embeds.dtype
+
+        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_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: process 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=prompt_embeds_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)
+
+        if self.text_encoder is not None:
+            if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder, lora_scale)
+
+        return prompt_embeds, negative_prompt_embeds
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
+    def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
+        dtype = next(self.image_encoder.parameters()).dtype
+
+        if not isinstance(image, torch.Tensor):
+            image = self.feature_extractor(image, return_tensors="pt").pixel_values
+
+        image = image.to(device=device, dtype=dtype)
+        if output_hidden_states:
+            image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
+            image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
+            uncond_image_enc_hidden_states = self.image_encoder(
+                torch.zeros_like(image), output_hidden_states=True
+            ).hidden_states[-2]
+            uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
+                num_images_per_prompt, dim=0
+            )
+            return image_enc_hidden_states, uncond_image_enc_hidden_states
+        else:
+            image_embeds = self.image_encoder(image).image_embeds
+            image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
+            uncond_image_embeds = torch.zeros_like(image_embeds)
+
+            return image_embeds, uncond_image_embeds
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
+
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
+
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
+
+                if do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
+
+                image_embeds.append(single_image_embeds)
+        else:
+            repeat_dims = [1]
+            image_embeds = []
+            for single_image_embeds in ip_adapter_image_embeds:
+                if do_classifier_free_guidance:
+                    single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
+                    single_image_embeds = single_image_embeds.repeat(
+                        num_images_per_prompt, *(repeat_dims * len(single_image_embeds.shape[1:]))
+                    )
+                    single_negative_image_embeds = single_negative_image_embeds.repeat(
+                        num_images_per_prompt, *(repeat_dims * len(single_negative_image_embeds.shape[1:]))
+                    )
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                else:
+                    single_image_embeds = single_image_embeds.repeat(
+                        num_images_per_prompt, *(repeat_dims * len(single_image_embeds.shape[1:]))
+                    )
+                image_embeds.append(single_image_embeds)
+
+        return image_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):
+        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
+        deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
+
+        latents = 1 / self.vae.config.scaling_factor * latents
+        image = self.vae.decode(latents, return_dict=False)[0]
+        image = (image / 2 + 0.5).clamp(0, 1)
+        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
+        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
+        return image
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    def prepare_extra_step_kwargs(self, generator, eta):
+        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
+        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
+        # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
+        # and should be between [0, 1]
+
+        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        extra_step_kwargs = {}
+        if accepts_eta:
+            extra_step_kwargs["eta"] = eta
+
+        # check if the scheduler accepts generator
+        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        if accepts_generator:
+            extra_step_kwargs["generator"] = generator
+        return extra_step_kwargs
+
+    def check_inputs(
+        self,
+        prompt,
+        image,
+        callback_steps,
+        negative_prompt=None,
+        prompt_embeds=None,
+        negative_prompt_embeds=None,
+        ip_adapter_image=None,
+        ip_adapter_image_embeds=None,
+        controlnet_conditioning_scale=1.0,
+        control_guidance_start=0.0,
+        control_guidance_end=1.0,
+        callback_on_step_end_tensor_inputs=None,
+    ):
+        if 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 callback_on_step_end_tensor_inputs is not None and not all(
+            k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
+        ):
+            raise ValueError(
+                f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
+            )
+
+        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}."
+                )
+
+        # 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):
+                transposed_image = [list(t) for t in zip(*image)]
+                if len(transposed_image) != len(self.controlnet.nets):
+                    raise ValueError(
+                        f"For multiple controlnets: if you pass`image` as a list of list, each sublist must have the same length as the number of controlnets, but the sublists in `image` got {len(transposed_image)} images and {len(self.controlnet.nets)} ControlNets."
+                    )
+                for image_ in transposed_image:
+                    self.check_image(image_, prompt, prompt_embeds)
+            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."
+                )
+            else:
+                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 varying conditioning scale settings (e.g. [[1.0, 0.5], [0.2, 0.8]]) is not supported at the moment. "
+                        "The conditioning scale must be fixed across the batch."
+                    )
+            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 not isinstance(control_guidance_start, (tuple, list)):
+            control_guidance_start = [control_guidance_start]
+
+        if not isinstance(control_guidance_end, (tuple, list)):
+            control_guidance_end = [control_guidance_end]
+
+        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.")
+
+        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
+            raise ValueError(
+                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
+            )
+
+        if ip_adapter_image_embeds is not None:
+            if not isinstance(ip_adapter_image_embeds, list):
+                raise ValueError(
+                    f"`ip_adapter_image_embeds` has to be of type `list` but is {type(ip_adapter_image_embeds)}"
+                )
+            elif ip_adapter_image_embeds[0].ndim not in [3, 4]:
+                raise ValueError(
+                    f"`ip_adapter_image_embeds` has to be a list of 3D or 4D tensors but is {ip_adapter_image_embeds[0].ndim}D"
+                )
+
+    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}"
+            )
+
+    def prepare_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.StableDiffusionPipeline.prepare_latents
+    def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
+        shape = (
+            batch_size,
+            num_channels_latents,
+            int(height) // self.vae_scale_factor,
+            int(width) // self.vae_scale_factor,
+        )
+        if isinstance(generator, list) and len(generator) != batch_size:
+            raise ValueError(
+                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
+                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
+            )
+
+        if latents is None:
+            latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
+        else:
+            latents = latents.to(device)
+
+        # scale the initial noise by the standard deviation required by the scheduler
+        latents = latents * self.scheduler.init_noise_sigma
+        return latents
+
+    # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
+    def get_guidance_scale_embedding(
+        self, w: torch.Tensor, embedding_dim: int = 512, dtype: torch.dtype = torch.float32
+    ) -> torch.Tensor:
+        """
+        See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
+
+        Args:
+            w (`torch.Tensor`):
+                Generate embedding vectors with a specified guidance scale to subsequently enrich timestep embeddings.
+            embedding_dim (`int`, *optional*, defaults to 512):
+                Dimension of the embeddings to generate.
+            dtype (`torch.dtype`, *optional*, defaults to `torch.float32`):
+                Data type of the generated embeddings.
+
+        Returns:
+            `torch.Tensor`: Embedding vectors with shape `(len(w), embedding_dim)`.
+        """
+        assert len(w.shape) == 1
+        w = w * 1000.0
+
+        half_dim = embedding_dim // 2
+        emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)
+        emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)
+        emb = w.to(dtype)[:, None] * emb[None, :]
+        emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
+        if embedding_dim % 2 == 1:  # zero pad
+            emb = torch.nn.functional.pad(emb, (0, 1))
+        assert emb.shape == (w.shape[0], embedding_dim)
+        return emb
+
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def clip_skip(self):
+        return self._clip_skip
+
+    # 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.
+    @property
+    def do_classifier_free_guidance(self):
+        return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
+
+    @property
+    def cross_attention_kwargs(self):
+        return self._cross_attention_kwargs
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
+    @torch.no_grad()
+    @replace_example_docstring(EXAMPLE_DOC_STRING)
+    def __call__(
+        self,
+        prompt: Union[str, List[str]] = None,
+        image: PipelineImageInput = None,
+        height: Optional[int] = None,
+        width: Optional[int] = None,
+        num_inference_steps: int = 50,
+        timesteps: List[int] = None,
+        sigmas: List[float] = None,
+        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.Tensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
+        output_type: Optional[str] = "pil",
+        return_dict: bool = True,
+        cross_attention_kwargs: Optional[Dict[str, Any]] = None,
+        controlnet_conditioning_scale: Union[float, List[float]] = 1.0,
+        guess_mode: bool = False,
+        control_guidance_start: Union[float, List[float]] = 0.0,
+        control_guidance_end: Union[float, List[float]] = 1.0,
+        clip_skip: Optional[int] = None,
+        callback_on_step_end: Optional[
+            Union[Callable[[int, int, Dict], None], PipelineCallback, MultiPipelineCallbacks]
+        ] = None,
+        callback_on_step_end_tensor_inputs: List[str] = ["latents"],
+        **kwargs,
+    ):
+        r"""
+        The call function to the pipeline for generation.
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts to guide image generation. If not defined, you need to pass `prompt_embeds`.
+            image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, `List[np.ndarray]`,:
+                    `List[List[torch.Tensor]]`, `List[List[np.ndarray]]` or `List[List[PIL.Image.Image]]`):
+                The ControlNet input condition to provide guidance to the `unet` for generation. If the type is
+                specified as `torch.Tensor`, 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 accordingly. 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. When `prompt` is a list, and if a list of images is passed for a single
+                ControlNet, each will be paired with each prompt in the `prompt` list. This also applies to multiple
+                ControlNets, where a list of image lists can be passed to batch for each prompt and each 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.
+            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.
+            timesteps (`List[int]`, *optional*):
+                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
+                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
+                passed will be used. Must be in descending order.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
+            guidance_scale (`float`, *optional*, defaults to 7.5):
+                A higher guidance scale value encourages the model to generate images closely linked to the text
+                `prompt` at the expense of lower image quality. Guidance scale is enabled when `guidance_scale > 1`.
+            negative_prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts to guide what to not include in image generation. If not defined, you need to
+                pass `negative_prompt_embeds` instead. Ignored when not using guidance (`guidance_scale < 1`).
+            num_images_per_prompt (`int`, *optional*, defaults to 1):
+                The number of images to generate per prompt.
+            eta (`float`, *optional*, defaults to 0.0):
+                Corresponds to parameter eta (η) from the [DDIM](https://arxiv.org/abs/2010.02502) paper. Only applies
+                to the [`~schedulers.DDIMScheduler`], and is ignored in other schedulers.
+            generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
+                A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make
+                generation deterministic.
+            latents (`torch.Tensor`, *optional*):
+                Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
+                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
+                tensor is generated by sampling using the supplied random `generator`.
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not
+                provided, text embeddings are generated from the `prompt` input argument.
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs (prompt weighting). If
+                not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
+            ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
+                IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. It should
+                contain the negative image embedding if `do_classifier_free_guidance` is set to `True`. If not
+                provided, embeddings are computed from the `ip_adapter_image` input argument.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generated image. Choose between `PIL.Image` or `np.array`.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
+                plain tuple.
+            callback (`Callable`, *optional*):
+                A function that calls every `callback_steps` steps during inference. The function is called with the
+                following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
+            callback_steps (`int`, *optional*, defaults to 1):
+                The frequency at which the `callback` function is called. If not specified, the callback is called at
+                every step.
+            cross_attention_kwargs (`dict`, *optional*):
+                A kwargs dictionary that if specified is passed along to the [`AttentionProcessor`] as defined in
+                [`self.processor`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
+            controlnet_conditioning_scale (`float` or `List[float]`, *optional*, defaults to 1.0):
+                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.
+            guess_mode (`bool`, *optional*, defaults to `False`):
+                The ControlNet encoder tries to recognize the content of the input image even if you remove all
+                prompts. A `guidance_scale` value 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.
+            clip_skip (`int`, *optional*):
+                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
+                the output of the pre-final layer will be used for computing the prompt embeddings.
+            callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
+                A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
+                each denoising step during the inference. with the following arguments: `callback_on_step_end(self:
+                DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)`. `callback_kwargs` will include a
+                list of all tensors as specified by `callback_on_step_end_tensor_inputs`.
+            callback_on_step_end_tensor_inputs (`List`, *optional*):
+                The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
+                will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
+                `._callback_tensor_inputs` attribute of your pipeline class.
+
+        Examples:
+
+        Returns:
+            [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
+                If `return_dict` is `True`, [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] is returned,
+                otherwise a `tuple` is returned where the first element is a list with the generated images and the
+                second element is a list of `bool`s indicating whether the corresponding generated image contains
+                "not-safe-for-work" (nsfw) content.
+        """
+
+        callback = kwargs.pop("callback", None)
+        callback_steps = kwargs.pop("callback_steps", None)
+
+        if callback is not None:
+            deprecate(
+                "callback",
+                "1.0.0",
+                "Passing `callback` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
+            )
+        if callback_steps is not None:
+            deprecate(
+                "callback_steps",
+                "1.0.0",
+                "Passing `callback_steps` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
+            )
+
+        if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
+            callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs
+
+        controlnet = self.controlnet._orig_mod if is_compiled_module(self.controlnet) else self.controlnet
+
+        # 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,
+            image,
+            callback_steps,
+            negative_prompt,
+            prompt_embeds,
+            negative_prompt_embeds,
+            ip_adapter_image,
+            ip_adapter_image_embeds,
+            controlnet_conditioning_scale,
+            control_guidance_start,
+            control_guidance_end,
+            callback_on_step_end_tensor_inputs,
+        )
+
+        self._guidance_scale = guidance_scale
+        self._clip_skip = clip_skip
+        self._cross_attention_kwargs = cross_attention_kwargs
+
+        # 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
+
+        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 = (
+            self.cross_attention_kwargs.get("scale", None) if self.cross_attention_kwargs is not None else None
+        )
+        prompt_embeds, negative_prompt_embeds = self.encode_prompt(
+            prompt,
+            device,
+            num_images_per_prompt,
+            self.do_classifier_free_guidance,
+            negative_prompt,
+            prompt_embeds=prompt_embeds,
+            negative_prompt_embeds=negative_prompt_embeds,
+            lora_scale=text_encoder_lora_scale,
+            clip_skip=self.clip_skip,
+        )
+        # 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
+        if self.do_classifier_free_guidance:
+            prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])
+
+        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
+            image_embeds = self.prepare_ip_adapter_image_embeds(
+                ip_adapter_image,
+                ip_adapter_image_embeds,
+                device,
+                batch_size * num_images_per_prompt,
+                self.do_classifier_free_guidance,
+            )
+
+        # 4. Prepare image
+        if isinstance(controlnet, ControlNetModel):
+            image = self.prepare_image(
+                image=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=self.do_classifier_free_guidance,
+                guess_mode=guess_mode,
+            )
+            height, width = image.shape[-2:]
+        elif isinstance(controlnet, MultiControlNetModel):
+            images = []
+
+            # Nested lists as ControlNet condition
+            if isinstance(image[0], list):
+                # Transpose the nested image list
+                image = [list(t) for t in zip(*image)]
+
+            for image_ in image:
+                image_ = self.prepare_image(
+                    image=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=self.do_classifier_free_guidance,
+                    guess_mode=guess_mode,
+                )
+
+                images.append(image_)
+
+            image = images
+            height, width = image[0].shape[-2:]
+        else:
+            assert False
+
+        # 5. Prepare timesteps
+        timesteps, num_inference_steps = retrieve_timesteps(
+            self.scheduler, num_inference_steps, device, timesteps, sigmas
+        )
+        self._num_timesteps = len(timesteps)
+
+        # 6. Prepare latent variables
+        num_channels_latents = self.unet.config.in_channels
+        latents = self.prepare_latents(
+            batch_size * num_images_per_prompt,
+            num_channels_latents,
+            height,
+            width,
+            prompt_embeds.dtype,
+            device,
+            generator,
+            latents,
+        )
+
+        # 6.5 Optionally get Guidance Scale Embedding
+        timestep_cond = None
+        if self.unet.config.time_cond_proj_dim is not None:
+            guidance_scale_tensor = torch.tensor(self.guidance_scale - 1).repeat(batch_size * num_images_per_prompt)
+            timestep_cond = self.get_guidance_scale_embedding(
+                guidance_scale_tensor, embedding_dim=self.unet.config.time_cond_proj_dim
+            ).to(device=device, dtype=latents.dtype)
+
+        # 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 Add image embeds for IP-Adapter
+        added_cond_kwargs = (
+            {"image_embeds": image_embeds}
+            if ip_adapter_image is not None or ip_adapter_image_embeds is not None
+            else None
+        )
+
+        # 7.2 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
+        is_unet_compiled = is_compiled_module(self.unet)
+        is_controlnet_compiled = is_compiled_module(self.controlnet)
+        is_torch_higher_equal_2_1 = is_torch_version(">=", "2.1")
+        with self.progress_bar(total=num_inference_steps) as progress_bar:
+            for i, t in enumerate(timesteps):
+                # Relevant thread:
+                # https://dev-discuss.pytorch.org/t/cudagraphs-in-pytorch-2-0/1428
+                if (is_unet_compiled and is_controlnet_compiled) and is_torch_higher_equal_2_1:
+                    torch._inductor.cudagraph_mark_step_begin()
+                # expand the latents if we are doing classifier free guidance
+                latent_model_input = torch.cat([latents] * 2) if self.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 self.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:
+                    controlnet_cond_scale = controlnet_conditioning_scale
+                    if isinstance(controlnet_cond_scale, list):
+                        controlnet_cond_scale = controlnet_cond_scale[0]
+                    cond_scale = controlnet_cond_scale * controlnet_keep[i]
+
+                down_block_res_samples, mid_block_res_sample = self.controlnet(
+                    control_model_input,
+                    t,
+                    encoder_hidden_states=controlnet_prompt_embeds,
+                    controlnet_cond=image,
+                    conditioning_scale=cond_scale,
+                    guess_mode=guess_mode,
+                    return_dict=False,
+                )
+
+                if guess_mode and self.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])
+
+                # predict the noise residual
+                noise_pred = self.unet(
+                    latent_model_input,
+                    t,
+                    encoder_hidden_states=prompt_embeds,
+                    timestep_cond=timestep_cond,
+                    cross_attention_kwargs=self.cross_attention_kwargs,
+                    down_block_additional_residuals=down_block_res_samples,
+                    mid_block_additional_residual=mid_block_res_sample,
+                    added_cond_kwargs=added_cond_kwargs,
+                    return_dict=False,
+                )[0]
+
+                # perform guidance
+                if self.do_classifier_free_guidance:
+                    noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
+                    noise_pred = noise_pred_uncond + self.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 callback_on_step_end is not None:
+                    callback_kwargs = {}
+                    for k in callback_on_step_end_tensor_inputs:
+                        callback_kwargs[k] = locals()[k]
+                    callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)
+
+                    latents = callback_outputs.pop("latents", latents)
+                    prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
+                    negative_prompt_embeds = callback_outputs.pop("negative_prompt_embeds", negative_prompt_embeds)
+
+                # call the callback, if provided
+                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
+                    progress_bar.update()
+                    if callback is not None and i % callback_steps == 0:
+                        step_idx = i // getattr(self.scheduler, "order", 1)
+                        callback(step_idx, t, latents)
+
+        # If 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, generator=generator)[
+                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 all models
+        self.maybe_free_model_hooks()
+
+        if not return_dict:
+            return (image, has_nsfw_concept)
+
+        return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)