up

* update

* update

---------

Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

.github/workflows/build_docker_images.yml

@@ -1,22 +1,58 @@
-name: Build Docker images (nightly)
+name: Test, build, and push Docker images
 
 on:
+  pull_request: # During PRs, we just check if the changes Dockerfiles can be successfully built
+    branches:
+      - main
+    paths:
+      - "docker/**"
   workflow_dispatch:
   schedule:
     - cron: "0 0 * * *" # every day at midnight
 
 concurrency:
-  group: docker-image-builds
-  cancel-in-progress: false
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
 
 env:
   REGISTRY: diffusers
   CI_SLACK_CHANNEL: ${{ secrets.CI_DOCKER_CHANNEL }}
 
 jobs:
-  build-docker-images:
+  test-build-docker-images:
     runs-on: ubuntu-latest
+    if: github.event_name == 'pull_request'
+    steps:
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v1
+      
+      - name: Check out code
+        uses: actions/checkout@v3
+      
+      - name: Find Changed Dockerfiles
+        id: file_changes
+        uses: jitterbit/get-changed-files@v1
+        with:
+          format: 'space-delimited'
+          token: ${{ secrets.GITHUB_TOKEN }}
+      
+      - name: Build Changed Docker Images
+        run: |
+          CHANGED_FILES="${{ steps.file_changes.outputs.all }}"
+          for FILE in $CHANGED_FILES; do
+            if [[ "$FILE" == docker/*Dockerfile ]]; then
+              DOCKER_PATH="${FILE%/Dockerfile}"
+              DOCKER_TAG=$(basename "$DOCKER_PATH")
+              echo "Building Docker image for $DOCKER_TAG"
+              docker build -t "$DOCKER_TAG" "$DOCKER_PATH"
+            fi
+          done
+        if: steps.file_changes.outputs.all != ''
 
+  build-and-push-docker-images:
+    runs-on: ubuntu-latest
+    if: github.event_name != 'pull_request'
+    
     permissions:
       contents: read
       packages: write

.github/workflows/nightly_tests.yml

@@ -12,6 +12,7 @@ env:
   PYTEST_TIMEOUT: 600
   RUN_SLOW: yes
   RUN_NIGHTLY: yes
+  SLACK_API_TOKEN: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
 jobs:
   run_nightly_tests:
@@ -64,6 +65,7 @@ jobs:
           python -m uv pip install -e [quality,test]
           python -m uv pip install -U transformers@git+https://github.com/huggingface/transformers
           python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate
+          python -m uv pip install pytest-reportlog
 
       - name: Environment
         run: |
@@ -78,7 +80,8 @@ jobs:
           python -m pytest -n 1 --max-worker-restart=0 --dist=loadfile \
             -s -v -k "not Flax and not Onnx" \
             --make-reports=tests_${{ matrix.config.report }} \
-            tests/
+            --report-log=${{ matrix.config.report }}.log \
+            tests/ 
 
       - name: Run nightly Flax TPU tests
         if: ${{ matrix.config.framework == 'flax' }}
@@ -89,6 +92,7 @@ jobs:
           python -m pytest -n 0 \
             -s -v -k "Flax" \
             --make-reports=tests_${{ matrix.config.report }} \
+            --report-log=${{ matrix.config.report }}.log \
             tests/
 
       - name: Run nightly ONNXRuntime CUDA tests
@@ -100,6 +104,7 @@ jobs:
           python -m pytest -n 1 --max-worker-restart=0 --dist=loadfile \
             -s -v -k "Onnx" \
             --make-reports=tests_${{ matrix.config.report }} \
+            --report-log=${{ matrix.config.report }}.log \ 
             tests/
 
       - name: Failure short reports
@@ -112,6 +117,12 @@ jobs:
         with:
           name: ${{ matrix.config.report }}_test_reports
           path: reports
+      
+      - name: Generate Report and Notify Channel
+        if: always()
+        run: |
+          pip install slack_sdk tabulate
+          python scripts/log_reports.py >> $GITHUB_STEP_SUMMARY
 
   run_nightly_tests_apple_m1:
     name: Nightly PyTorch MPS tests on MacOS
@@ -140,6 +151,7 @@ jobs:
           ${CONDA_RUN} python -m uv pip install -e [quality,test]
           ${CONDA_RUN} python -m uv pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cpu
           ${CONDA_RUN} python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate
+          ${CONDA_RUN} python -m uv pip install pytest-reportlog
 
       - name: Environment
         shell: arch -arch arm64 bash {0}
@@ -152,7 +164,9 @@ jobs:
           HF_HOME: /System/Volumes/Data/mnt/cache
           HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
         run: |
-          ${CONDA_RUN} python -m pytest -n 1 -s -v --make-reports=tests_torch_mps tests/
+          ${CONDA_RUN} python -m pytest -n 1 -s -v --make-reports=tests_torch_mps \
+            --report-log=tests_torch_mps.log \
+            tests/
 
       - name: Failure short reports
         if: ${{ failure() }}
@@ -164,3 +178,9 @@ jobs:
         with:
           name: torch_mps_test_reports
           path: reports
+
+      - name: Generate Report and Notify Channel
+        if: always()
+        run: |
+          pip install slack_sdk tabulate
+          python scripts/log_reports.py >> $GITHUB_STEP_SUMMARY

.github/workflows/notify_slack_about_release.yml

@@ -0,0 +1,23 @@
+name: Notify Slack about a release
+
+on:
+  workflow_dispatch:
+  release:
+    types: [published]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v3
+    
+    - name: Setup Python
+      uses: actions/setup-python@v4
+      with:
+        python-version: '3.8'
+    
+    - name: Notify Slack about the release
+      env:
+        SLACK_WEBHOOK_URL: ${{ secrets.SLACK_WEBHOOK_URL }}
+      run: pip install requests && python utils/notify_slack_about_release.py

.github/workflows/pypi_publish.yaml

@@ -0,0 +1,81 @@
+# Adapted from https://blog.deepjyoti30.dev/pypi-release-github-action
+
+name: PyPI release
+
+on:
+  workflow_dispatch:
+  push:
+    tags:
+      - "*"
+
+jobs:
+  find-and-checkout-latest-branch:
+    runs-on: ubuntu-latest
+    outputs:
+      latest_branch: ${{ steps.set_latest_branch.outputs.latest_branch }}
+    steps:
+      - name: Checkout Repo
+        uses: actions/checkout@v3
+
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.8'
+
+      - name: Fetch latest branch
+        id: fetch_latest_branch
+        run: |
+          pip install -U requests packaging
+          LATEST_BRANCH=$(python utils/fetch_latest_release_branch.py)
+          echo "Latest branch: $LATEST_BRANCH"
+          echo "latest_branch=$LATEST_BRANCH" >> $GITHUB_ENV
+          
+      - name: Set latest branch output
+        id: set_latest_branch
+        run: echo "::set-output name=latest_branch::${{ env.latest_branch }}"
+
+  release:
+    needs: find-and-checkout-latest-branch
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout Repo
+        uses: actions/checkout@v3
+        with:
+          ref: ${{ needs.find-and-checkout-latest-branch.outputs.latest_branch }}
+          
+      - name: Setup Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.8"
+      
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -U setuptools wheel twine
+          pip install -U torch --index-url https://download.pytorch.org/whl/cpu
+          pip install -U transformers
+      
+      - name: Build the dist files
+        run: python setup.py bdist_wheel && python setup.py sdist
+      
+      - name: Publish to the test PyPI
+        env:
+          TWINE_USERNAME: ${{ secrets.TEST_PYPI_USERNAME }}
+          TWINE_PASSWORD: ${{ secrets.TEST_PYPI_PASSWORD }}
+        run: twine upload dist/* -r pypitest --repository-url=https://test.pypi.org/legacy/    
+
+      - name: Test installing diffusers and importing
+        run: |
+          pip install diffusers && pip uninstall diffusers -y
+          pip install -i https://testpypi.python.org/pypi diffusers
+          python -c "from diffusers import __version__; print(__version__)"
+          python -c "from diffusers import DiffusionPipeline; pipe = DiffusionPipeline.from_pretrained('fusing/unet-ldm-dummy-update'); pipe()"
+          python -c "from diffusers import DiffusionPipeline; pipe = DiffusionPipeline.from_pretrained('hf-internal-testing/tiny-stable-diffusion-pipe', safety_checker=None); pipe('ah suh du')"
+          python -c "from diffusers import *"
+
+      - name: Publish to PyPI
+        env:
+          TWINE_USERNAME: ${{ secrets.PYPI_USERNAME }}
+          TWINE_PASSWORD: ${{ secrets.PYPI_PASSWORD }}
+        run: twine upload dist/* -r pypi

README.md

@@ -77,7 +77,7 @@ Please refer to the [How to use Stable Diffusion in Apple Silicon](https://huggi
 
 ## Quickstart
 
-Generating outputs is super easy with 🤗 Diffusers. To generate an image from text, use the `from_pretrained` method to load any pretrained diffusion model (browse the [Hub](https://huggingface.co/models?library=diffusers&sort=downloads) for 19000+ checkpoints):
+Generating outputs is super easy with 🤗 Diffusers. To generate an image from text, use the `from_pretrained` method to load any pretrained diffusion model (browse the [Hub](https://huggingface.co/models?library=diffusers&sort=downloads) for 22000+ checkpoints):
 
 ```python
 from diffusers import DiffusionPipeline
@@ -219,7 +219,7 @@ Also, say 👋 in our public Discord channel <a href="https://discord.gg/G7tWnz9
 - https://github.com/deep-floyd/IF
 - https://github.com/bentoml/BentoML
 - https://github.com/bmaltais/kohya_ss
-- +8000 other amazing GitHub repositories 💪
+- +9000 other amazing GitHub repositories 💪
 
 Thank you for using us ❤️.
 

docker/diffusers-pytorch-cpu/Dockerfile

@@ -40,6 +40,6 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
         numpy \
         scipy \
         tensorboard \
-        transformers
+        transformers matplotlib
 
 CMD ["/bin/bash"]

docs/source/en/_toctree.yml

@@ -18,7 +18,7 @@
   - local: tutorials/basic_training
     title: Train a diffusion model
   - local: tutorials/using_peft_for_inference
-    title: Inference with PEFT
+    title: Load LoRAs for inference
   - local: tutorials/fast_diffusion
     title: Accelerate inference of text-to-image diffusion models
   title: Tutorials
@@ -62,6 +62,8 @@
       title: Textual inversion
     - local: using-diffusers/ip_adapter
       title: IP-Adapter
+    - local: using-diffusers/merge_loras
+      title: Merge LoRAs
     - local: training/distributed_inference
       title: Distributed inference with multiple GPUs
     - local: using-diffusers/reusing_seeds
@@ -102,6 +104,8 @@
       title: Latent Consistency Model-LoRA
     - local: using-diffusers/inference_with_lcm
       title: Latent Consistency Model
+    - local: using-diffusers/inference_with_tcd_lora
+      title: Trajectory Consistency Distillation-LoRA
     - local: using-diffusers/svd
       title: Stable Video Diffusion
     title: Specific pipeline examples
@@ -302,6 +306,8 @@
       title: Latent Consistency Models
     - local: api/pipelines/latent_diffusion
       title: Latent Diffusion
+    - local: api/pipelines/ledits_pp
+      title: LEDITS++
     - local: api/pipelines/panorama
       title: MultiDiffusion
     - local: api/pipelines/musicldm
@@ -318,6 +324,8 @@
       title: Semantic Guidance
     - local: api/pipelines/shap_e
       title: Shap-E
+    - local: api/pipelines/stable_cascade
+      title: Stable Cascade
     - sections:
       - local: api/pipelines/stable_diffusion/overview
         title: Overview
@@ -392,6 +400,10 @@
       title: DPMSolverSDEScheduler
     - local: api/schedulers/singlestep_dpm_solver
       title: DPMSolverSinglestepScheduler
+    - local: api/schedulers/edm_multistep_dpm_solver
+      title: EDMDPMSolverMultistepScheduler
+    - local: api/schedulers/edm_euler
+      title: EDMEulerScheduler
     - local: api/schedulers/euler_ancestral
       title: EulerAncestralDiscreteScheduler
     - local: api/schedulers/euler
@@ -418,6 +430,8 @@
       title: ScoreSdeVeScheduler
     - local: api/schedulers/score_sde_vp
       title: ScoreSdeVpScheduler
+    - local: api/schedulers/tcd
+      title: TCDScheduler
     - local: api/schedulers/unipc
       title: UniPCMultistepScheduler
     - local: api/schedulers/vq_diffusion

docs/source/en/api/loaders/ip_adapter.md

@@ -23,3 +23,7 @@ Learn how to load an IP-Adapter checkpoint and image in the IP-Adapter [loading]
 ## IPAdapterMixin
 
 [[autodoc]] loaders.ip_adapter.IPAdapterMixin
+
+## IPAdapterMaskProcessor
+
+[[autodoc]] image_processor.IPAdapterMaskProcessor

docs/source/en/api/pipelines/ledits_pp.md

@@ -0,0 +1,54 @@
+<!--Copyright 2023 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# LEDITS++
+
+LEDITS++ was proposed in [LEDITS++: Limitless Image Editing using Text-to-Image Models](https://huggingface.co/papers/2311.16711) by Manuel Brack, Felix Friedrich, Katharina Kornmeier, Linoy Tsaban, Patrick Schramowski, Kristian Kersting, Apolinário Passos.
+
+The abstract from the paper is:
+
+*Text-to-image diffusion models have recently received increasing interest for their astonishing ability to produce high-fidelity images from solely text inputs. Subsequent research efforts aim to exploit and apply their capabilities to real image editing. However, existing image-to-image methods are often inefficient, imprecise, and of limited versatility. They either require time-consuming fine-tuning, deviate unnecessarily strongly from the input image, and/or lack support for multiple, simultaneous edits. To address these issues, we introduce LEDITS++, an efficient yet versatile and precise textual image manipulation technique. LEDITS++'s novel inversion approach requires no tuning nor optimization and produces high-fidelity results with a few diffusion steps. Second, our methodology supports multiple simultaneous edits and is architecture-agnostic. Third, we use a novel implicit masking technique that limits changes to relevant image regions. We propose the novel TEdBench++ benchmark as part of our exhaustive evaluation. Our results demonstrate the capabilities of LEDITS++ and its improvements over previous methods. The project page is available at https://leditsplusplus-project.static.hf.space .*
+
+<Tip>
+
+You can find additional information about LEDITS++ on the [project page](https://leditsplusplus-project.static.hf.space/index.html) and try it out in a [demo](https://huggingface.co/spaces/editing-images/leditsplusplus).
+
+</Tip>
+
+<Tip warning={true}>
+Due to some backward compatability issues with the current diffusers implementation of [`~schedulers.DPMSolverMultistepScheduler`] this implementation of LEdits++ can no longer guarantee perfect inversion. 
+This issue is unlikely to have any noticeable effects on applied use-cases. However, we provide an alternative implementation that guarantees perfect inversion in a dedicated [GitHub repo](https://github.com/ml-research/ledits_pp). 
+</Tip>
+
+We provide two distinct pipelines based on different pre-trained models. 
+
+## LEditsPPPipelineStableDiffusion
+[[autodoc]] pipelines.ledits_pp.LEditsPPPipelineStableDiffusion
+	- all
+	- __call__
+	- invert
+
+## LEditsPPPipelineStableDiffusionXL
+[[autodoc]] pipelines.ledits_pp.LEditsPPPipelineStableDiffusionXL
+	- all
+	- __call__
+	- invert
+
+
+
+## LEditsPPDiffusionPipelineOutput
+[[autodoc]] pipelines.ledits_pp.pipeline_output.LEditsPPDiffusionPipelineOutput
+	- all
+
+## LEditsPPInversionPipelineOutput
+[[autodoc]] pipelines.ledits_pp.pipeline_output.LEditsPPInversionPipelineOutput
+	- all

docs/source/en/api/pipelines/overview.md

@@ -57,6 +57,7 @@ The table below lists all the pipelines currently available in 🤗 Diffusers an
 | [Latent Consistency Models](latent_consistency_models) | text2image |
 | [Latent Diffusion](latent_diffusion) | text2image, super-resolution |
 | [LDM3D](stable_diffusion/ldm3d_diffusion) | text2image, text-to-3D, text-to-pano, upscaling |
+| [LEDITS++](ledits_pp) | image editing |
 | [MultiDiffusion](panorama) | text2image |
 | [MusicLDM](musicldm) | text2audio |
 | [Paint by Example](paint_by_example) | inpainting |

docs/source/en/api/pipelines/semantic_stable_diffusion.md

@@ -30,6 +30,6 @@ Make sure to check out the Schedulers [guide](../../using-diffusers/schedulers)
 	- all
 	- __call__
 
-## StableDiffusionSafePipelineOutput
+## SemanticStableDiffusionPipelineOutput
 [[autodoc]] pipelines.semantic_stable_diffusion.pipeline_output.SemanticStableDiffusionPipelineOutput
 	- all

docs/source/en/api/pipelines/stable_cascade.md

@@ -0,0 +1,229 @@
+<!--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.
+-->
+
+# Stable Cascade
+
+This model is built upon the [Würstchen](https://openreview.net/forum?id=gU58d5QeGv) architecture and its main
+difference to other models like Stable Diffusion is that it is working at a much smaller latent space. Why is this
+important? The smaller the latent space, the **faster** you can run inference and the **cheaper** the training becomes.
+How small is the latent space? Stable Diffusion uses a compression factor of 8, resulting in a 1024x1024 image being
+encoded to 128x128. Stable Cascade achieves a compression factor of 42, meaning that it is possible to encode a
+1024x1024 image to 24x24, while maintaining crisp reconstructions. The text-conditional model is then trained in the
+highly compressed latent space. Previous versions of this architecture, achieved a 16x cost reduction over Stable
+Diffusion 1.5.
+
+Therefore, this kind of model is well suited for usages where efficiency is important. Furthermore, all known extensions
+like finetuning, LoRA, ControlNet, IP-Adapter, LCM etc. are possible with this method as well.
+
+The original codebase can be found at [Stability-AI/StableCascade](https://github.com/Stability-AI/StableCascade).
+
+## Model Overview
+Stable Cascade consists of three models: Stage A, Stage B and Stage C, representing a cascade to generate images,
+hence the name "Stable Cascade".
+
+Stage A & B are used to compress images, similar to what the job of the VAE is in Stable Diffusion.
+However, with this setup, a much higher compression of images can be achieved. While the Stable Diffusion models use a
+spatial compression factor of 8, encoding an image with resolution of 1024 x 1024 to 128 x 128, Stable Cascade achieves
+a compression factor of 42. This encodes a 1024 x 1024 image to 24 x 24, while being able to accurately decode the
+image. This comes with the great benefit of cheaper training and inference. Furthermore, Stage C is responsible
+for generating the small 24 x 24 latents given a text prompt.
+
+The Stage C model operates on the small 24 x 24 latents and denoises the latents conditioned on text prompts. The model is also the largest component in the Cascade pipeline and is meant to be used with the `StableCascadePriorPipeline`
+
+The Stage B and Stage A models are used with the `StableCascadeDecoderPipeline` and are responsible for generating the final image given the small 24 x 24 latents.
+
+<Tip warning={true}>
+
+There are some restrictions on data types that can be used with the Stable Cascade models. The official checkpoints for the  `StableCascadePriorPipeline` do not support the `torch.float16` data type. Please use `torch.bfloat16` instead.
+
+In order to use the `torch.bfloat16` data type with the `StableCascadeDecoderPipeline` you need to have PyTorch 2.2.0 or higher installed. This also means that using the `StableCascadeCombinedPipeline` with `torch.bfloat16` requires PyTorch 2.2.0 or higher, since it calls the `StableCascadeDecoderPipeline` internally.
+
+If it is not possible to install PyTorch 2.2.0 or higher in your environment, the `StableCascadeDecoderPipeline` can be used on its own with the `torch.float16` data type. You can download the full precision or `bf16` variant weights for the pipeline and cast the weights to `torch.float16`.
+
+</Tip>
+
+## Usage example
+
+```python
+import torch
+from diffusers import StableCascadeDecoderPipeline, StableCascadePriorPipeline
+
+prompt = "an image of a shiba inu, donning a spacesuit and helmet"
+negative_prompt = ""
+
+prior = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", variant="bf16", torch_dtype=torch.bfloat16)
+decoder = StableCascadeDecoderPipeline.from_pretrained("stabilityai/stable-cascade", variant="bf16", torch_dtype=torch.float16)
+
+prior.enable_model_cpu_offload()
+prior_output = prior(
+    prompt=prompt,
+    height=1024,
+    width=1024,
+    negative_prompt=negative_prompt,
+    guidance_scale=4.0,
+    num_images_per_prompt=1,
+    num_inference_steps=20
+)
+
+decoder.enable_model_cpu_offload()
+decoder_output = decoder(
+    image_embeddings=prior_output.image_embeddings.to(torch.float16),
+    prompt=prompt,
+    negative_prompt=negative_prompt,
+    guidance_scale=0.0,
+    output_type="pil",
+    num_inference_steps=10
+).images[0]
+decoder_output.save("cascade.png")
+```
+
+## Using the Lite Versions of the Stage B and Stage C models
+
+```python
+import torch
+from diffusers import (
+    StableCascadeDecoderPipeline,
+    StableCascadePriorPipeline,
+    StableCascadeUNet,
+)
+
+prompt = "an image of a shiba inu, donning a spacesuit and helmet"
+negative_prompt = ""
+
+prior_unet = StableCascadeUNet.from_pretrained("stabilityai/stable-cascade-prior", subfolder="prior_lite")
+decoder_unet = StableCascadeUNet.from_pretrained("stabilityai/stable-cascade", subfolder="decoder_lite")
+
+prior = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", prior=prior_unet)
+decoder = StableCascadeDecoderPipeline.from_pretrained("stabilityai/stable-cascade", decoder=decoder_unet)
+
+prior.enable_model_cpu_offload()
+prior_output = prior(
+    prompt=prompt,
+    height=1024,
+    width=1024,
+    negative_prompt=negative_prompt,
+    guidance_scale=4.0,
+    num_images_per_prompt=1,
+    num_inference_steps=20
+)
+
+decoder.enable_model_cpu_offload()
+decoder_output = decoder(
+    image_embeddings=prior_output.image_embeddings,
+    prompt=prompt,
+    negative_prompt=negative_prompt,
+    guidance_scale=0.0,
+    output_type="pil",
+    num_inference_steps=10
+).images[0]
+decoder_output.save("cascade.png")
+```
+
+## Loading original checkpoints with `from_single_file`
+
+Loading the original format checkpoints is supported via `from_single_file` method in the StableCascadeUNet.
+
+```python
+import torch
+from diffusers import (
+    StableCascadeDecoderPipeline,
+    StableCascadePriorPipeline,
+    StableCascadeUNet,
+)
+
+prompt = "an image of a shiba inu, donning a spacesuit and helmet"
+negative_prompt = ""
+
+prior_unet = StableCascadeUNet.from_single_file(
+    "https://huggingface.co/stabilityai/stable-cascade/resolve/main/stage_c_bf16.safetensors",
+    torch_dtype=torch.bfloat16
+)
+decoder_unet = StableCascadeUNet.from_single_file(
+    "https://huggingface.co/stabilityai/stable-cascade/blob/main/stage_b_bf16.safetensors",
+    torch_dtype=torch.bfloat16
+)
+
+prior = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", prior=prior_unet, torch_dtype=torch.bfloat16)
+decoder = StableCascadeDecoderPipeline.from_pretrained("stabilityai/stable-cascade", decoder=decoder_unet, torch_dtype=torch.bfloat16)
+
+prior.enable_model_cpu_offload()
+prior_output = prior(
+    prompt=prompt,
+    height=1024,
+    width=1024,
+    negative_prompt=negative_prompt,
+    guidance_scale=4.0,
+    num_images_per_prompt=1,
+    num_inference_steps=20
+)
+
+decoder.enable_model_cpu_offload()
+decoder_output = decoder(
+    image_embeddings=prior_output.image_embeddings,
+    prompt=prompt,
+    negative_prompt=negative_prompt,
+    guidance_scale=0.0,
+    output_type="pil",
+    num_inference_steps=10
+).images[0]
+decoder_output.save("cascade-single-file.png")
+```
+
+## Uses
+
+### Direct Use
+
+The model is intended for research purposes for now. Possible research areas and tasks include
+
+- Research on generative models.
+- Safe deployment of models which have the potential to generate harmful content.
+- Probing and understanding the limitations and biases of generative models.
+- Generation of artworks and use in design and other artistic processes.
+- Applications in educational or creative tools.
+
+Excluded uses are described below.
+
+### Out-of-Scope Use
+
+The model was not trained to be factual or true representations of people or events,
+and therefore using the model to generate such content is out-of-scope for the abilities of this model.
+The model should not be used in any way that violates Stability AI's [Acceptable Use Policy](https://stability.ai/use-policy).
+
+## Limitations and Bias
+
+### Limitations
+- Faces and people in general may not be generated properly.
+- The autoencoding part of the model is lossy.
+
+
+## StableCascadeCombinedPipeline
+
+[[autodoc]] StableCascadeCombinedPipeline
+	- all
+	- __call__
+
+## StableCascadePriorPipeline
+
+[[autodoc]] StableCascadePriorPipeline
+	- all
+	- __call__
+
+## StableCascadePriorPipelineOutput
+
+[[autodoc]] pipelines.stable_cascade.pipeline_stable_cascade_prior.StableCascadePriorPipelineOutput
+
+## StableCascadeDecoderPipeline
+
+[[autodoc]] StableCascadeDecoderPipeline
+	- all
+	- __call__
+

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

@@ -172,3 +172,41 @@ inpaint = StableDiffusionInpaintPipeline(**text2img.components)
 
 # now you can use text2img(...), img2img(...), inpaint(...) just like the call methods of each respective pipeline
 ```
+
+### Create web demos using `gradio`
+
+The Stable Diffusion pipelines are automatically supported in [Gradio](https://github.com/gradio-app/gradio/), a library that makes creating beautiful and user-friendly machine learning apps on the web a breeze. First, make sure you have Gradio installed:
+
+```
+pip install -U gradio
+```
+
+Then, create a web demo around any Stable Diffusion-based pipeline. For example, you can create an image generation pipeline in a single line of code with Gradio's [`Interface.from_pipeline`](https://www.gradio.app/docs/interface#interface-from-pipeline) function:
+
+```py
+from diffusers import StableDiffusionPipeline
+import gradio as gr
+
+pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4")
+
+gr.Interface.from_pipeline(pipe).launch()
+```
+
+which opens an intuitive drag-and-drop interface in your browser:
+
+![](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/gradio-panda.png)
+
+Similarly, you could create a demo for an image-to-image pipeline with:
+
+```py
+from diffusers import StableDiffusionImg2ImgPipeline
+import gradio as gr
+
+
+pipe = StableDiffusionImg2ImgPipeline.from_pretrained("runwayml/stable-diffusion-v1-5")
+
+gr.Interface.from_pipeline(pipe).launch()
+```
+
+By default, the web demo runs on a local server. If you'd like to share it with others, you can generate a temporary public
+link by setting `share=True` in `launch()`. Or, you can host your demo on [Hugging Face Spaces](https://huggingface.co/spaces)https://huggingface.co/spaces for a permanent link. 

docs/source/en/api/schedulers/edm_euler.md

@@ -0,0 +1,22 @@
+<!--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.
+-->
+
+# EDMEulerScheduler
+
+The Karras formulation of the Euler scheduler (Algorithm 2) from the [Elucidating the Design Space of Diffusion-Based Generative Models](https://huggingface.co/papers/2206.00364) paper by Karras et al. This is a fast scheduler which can often generate good outputs in 20-30 steps. The scheduler is based on the original [k-diffusion](https://github.com/crowsonkb/k-diffusion/blob/481677d114f6ea445aa009cf5bd7a9cdee909e47/k_diffusion/sampling.py#L51) implementation by [Katherine Crowson](https://github.com/crowsonkb/).
+
+
+## EDMEulerScheduler
+[[autodoc]] EDMEulerScheduler
+
+## EDMEulerSchedulerOutput
+[[autodoc]] schedulers.scheduling_edm_euler.EDMEulerSchedulerOutput

docs/source/en/api/schedulers/edm_multistep_dpm_solver.md

@@ -0,0 +1,24 @@
+<!--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.
+-->
+
+# EDMDPMSolverMultistepScheduler
+
+`EDMDPMSolverMultistepScheduler` is a [Karras formulation](https://huggingface.co/papers/2206.00364) of `DPMSolverMultistep`, a multistep scheduler from [DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps](https://huggingface.co/papers/2206.00927) and [DPM-Solver++: Fast Solver for Guided Sampling of Diffusion Probabilistic Models](https://huggingface.co/papers/2211.01095) by Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan Li, and Jun Zhu.
+
+DPMSolver (and the improved version DPMSolver++) is a fast dedicated high-order solver for diffusion ODEs with convergence order guarantee. Empirically, DPMSolver sampling with only 20 steps can generate high-quality
+samples, and it can generate quite good samples even in 10 steps.
+
+## EDMDPMSolverMultistepScheduler
+[[autodoc]] EDMDPMSolverMultistepScheduler
+
+## SchedulerOutput
+[[autodoc]] schedulers.scheduling_utils.SchedulerOutput

docs/source/en/api/schedulers/tcd.md

@@ -0,0 +1,29 @@
+<!--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.
+-->
+
+# TCDScheduler 
+
+[Trajectory Consistency Distillation](https://huggingface.co/papers/2402.19159) by Jianbin Zheng, Minghui Hu, Zhongyi Fan, Chaoyue Wang, Changxing Ding, Dacheng Tao and Tat-Jen Cham introduced a Strategic Stochastic Sampling (Algorithm 4) that is capable of generating good samples in a small number of steps. Distinguishing it as an advanced iteration of the multistep scheduler (Algorithm 1) in the [Consistency Models](https://huggingface.co/papers/2303.01469), Strategic Stochastic Sampling specifically tailored for the trajectory consistency function.
+
+The abstract from the paper is:
+
+*Latent Consistency Model (LCM) extends the Consistency Model to the latent space and leverages the guided consistency distillation technique to achieve impressive performance in accelerating text-to-image synthesis. However, we observed that LCM struggles to generate images with both clarity and detailed intricacy. To address this limitation, we initially delve into and elucidate the underlying causes. Our investigation identifies that the primary issue stems from errors in three distinct areas. Consequently, we introduce Trajectory Consistency Distillation (TCD), which encompasses trajectory consistency function and strategic stochastic sampling. The trajectory consistency function diminishes the distillation errors by broadening the scope of the self-consistency boundary condition and endowing the TCD with the ability to accurately trace the entire trajectory of the Probability Flow ODE. Additionally, strategic stochastic sampling is specifically designed to circumvent the accumulated errors inherent in multi-step consistency sampling, which is meticulously tailored to complement the TCD model. Experiments demonstrate that TCD not only significantly enhances image quality at low NFEs but also yields more detailed results compared to the teacher model at high NFEs.*
+
+The original codebase can be found at [jabir-zheng/TCD](https://github.com/jabir-zheng/TCD).
+
+## TCDScheduler
+[[autodoc]] TCDScheduler
+
+
+## TCDSchedulerOutput
+[[autodoc]] schedulers.scheduling_tcd.TCDSchedulerOutput
+

docs/source/en/training/lora.md

@@ -77,7 +77,7 @@ accelerate config default
 
 Or if your environment doesn't support an interactive shell, like a notebook, you can use:
 
-```bash
+```py
 from accelerate.utils import write_basic_config
 
 write_basic_config()
@@ -170,7 +170,7 @@ Aside from setting up the LoRA layers, the training script is more or less the s
 
 Once you've made all your changes or you're okay with the default configuration, you're ready to launch the training script! 🚀
 
-Let's train on the [Pokémon BLIP captions](https://huggingface.co/datasets/lambdalabs/pokemon-blip-captions) dataset to generate our yown Pokémon. Set the environment variables `MODEL_NAME` and `DATASET_NAME` to the model and dataset respectively. You should also specify where to save the model in `OUTPUT_DIR`, and the name of the model to save to on the Hub with `HUB_MODEL_ID`. The script creates and saves the following files to your repository:
+Let's train on the [Pokémon BLIP captions](https://huggingface.co/datasets/lambdalabs/pokemon-blip-captions) dataset to generate our own Pokémon. Set the environment variables `MODEL_NAME` and `DATASET_NAME` to the model and dataset respectively. You should also specify where to save the model in `OUTPUT_DIR`, and the name of the model to save to on the Hub with `HUB_MODEL_ID`. The script creates and saves the following files to your repository:
 
 - saved model checkpoints
 - `pytorch_lora_weights.safetensors` (the trained LoRA weights)

docs/source/en/tutorials/using_peft_for_inference.md

@@ -14,19 +14,17 @@ specific language governing permissions and limitations under the License.
 
 # Load LoRAs for inference
 
-There are many adapters (with LoRAs being the most common type) trained in different styles to achieve different effects. You can even combine multiple adapters to create new and unique images. With the 🤗 [PEFT](https://huggingface.co/docs/peft/index) integration in 🤗 Diffusers, it is really easy to load and manage adapters for inference. In this guide, you'll learn how to use different adapters with [Stable Diffusion XL (SDXL)](../api/pipelines/stable_diffusion/stable_diffusion_xl) for inference.
+There are many adapter types (with [LoRAs](https://huggingface.co/docs/peft/conceptual_guides/adapter#low-rank-adaptation-lora) being the most popular) trained in different styles to achieve different effects. You can even combine multiple adapters to create new and unique images.
 
-Throughout this guide, you'll use LoRA as the main adapter technique, so we'll use the terms LoRA and adapter interchangeably. You should have some familiarity with LoRA, and if you don't, we welcome you to check out the [LoRA guide](https://huggingface.co/docs/peft/conceptual_guides/lora).
+In this tutorial, you'll learn how to easily load and manage adapters for inference with the 🤗 [PEFT](https://huggingface.co/docs/peft/index) integration in 🤗 Diffusers. You'll use LoRA as the main adapter technique, so you'll see the terms LoRA and adapter used interchangeably.
 
 Let's first install all the required libraries.
 
 ```bash
-!pip install -q transformers accelerate
-!pip install peft
-!pip install diffusers
+!pip install -q transformers accelerate peft diffusers
 ```
 
-Now, let's load a pipeline with a SDXL checkpoint:
+Now, load a pipeline with a [Stable Diffusion XL (SDXL)](../api/pipelines/stable_diffusion/stable_diffusion_xl) checkpoint:
 
 ```python
 from diffusers import DiffusionPipeline
@@ -36,16 +34,13 @@ pipe_id = "stabilityai/stable-diffusion-xl-base-1.0"
 pipe = DiffusionPipeline.from_pretrained(pipe_id, torch_dtype=torch.float16).to("cuda")
 ```
 
-
-Next, load a LoRA checkpoint with the [`~diffusers.loaders.StableDiffusionXLLoraLoaderMixin.load_lora_weights`] method.
-
-With the 🤗 PEFT integration, you can assign a specific `adapter_name` to the checkpoint, which let's you easily switch between different LoRA checkpoints. Let's call this adapter `"toy"`.
+Next, load a [CiroN2022/toy-face](https://huggingface.co/CiroN2022/toy-face) adapter with the [`~diffusers.loaders.StableDiffusionXLLoraLoaderMixin.load_lora_weights`] method. With the 🤗 PEFT integration, you can assign a specific `adapter_name` to the checkpoint, which let's you easily switch between different LoRA checkpoints. Let's call this adapter `"toy"`.
 
 ```python
 pipe.load_lora_weights("CiroN2022/toy-face", weight_name="toy_face_sdxl.safetensors", adapter_name="toy")
 ```
 
-And then perform inference:
+Make sure to include the token `toy_face` in the prompt and then you can perform inference:
 
 ```python
 prompt = "toy_face of a hacker with a hoodie"
@@ -59,17 +54,16 @@ image
 
 ![toy-face](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/peft_integration/diffusers_peft_lora_inference_8_1.png)
 
+With the `adapter_name` parameter, it is really easy to use another adapter for inference! Load the [nerijs/pixel-art-xl](https://huggingface.co/nerijs/pixel-art-xl) adapter that has been fine-tuned to generate pixel art images and call it `"pixel"`.
 
-With the `adapter_name` parameter, it is really easy to use another adapter for inference! Load the [nerijs/pixel-art-xl](https://huggingface.co/nerijs/pixel-art-xl) adapter that has been fine-tuned to generate pixel art images, and let's call it `"pixel"`.
-
-The pipeline automatically sets the first loaded adapter (`"toy"`) as the active adapter. But you can activate the `"pixel"` adapter with the [`~diffusers.loaders.UNet2DConditionLoadersMixin.set_adapters`] method as shown below:
+The pipeline automatically sets the first loaded adapter (`"toy"`) as the active adapter, but you can activate the `"pixel"` adapter with the [`~diffusers.loaders.UNet2DConditionLoadersMixin.set_adapters`] method:
 
 ```python
 pipe.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
 pipe.set_adapters("pixel")
 ```
 
-Let's now generate an image with the second adapter and check the result:
+Make sure you include the token `pixel art` in your prompt to generate a pixel art image:
 
 ```python
 prompt = "a hacker with a hoodie, pixel art"
@@ -81,29 +75,25 @@ image
 
 ![pixel-art](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/peft_integration/diffusers_peft_lora_inference_12_1.png)
 
-## Combine multiple adapters
+## Merge adapters
 
-You can also perform multi-adapter inference where you combine different adapter checkpoints for inference.
+You can also merge different adapter checkpoints for inference to blend their styles together.
 
-Once again, use the [`~diffusers.loaders.UNet2DConditionLoadersMixin.set_adapters`] method to activate two LoRA checkpoints and specify the weight for how the checkpoints should be combined.
+Once again, use the [`~diffusers.loaders.UNet2DConditionLoadersMixin.set_adapters`] method to activate the `pixel` and `toy` adapters and specify the weights for how they should be merged.
 
 ```python
 pipe.set_adapters(["pixel", "toy"], adapter_weights=[0.5, 1.0])
 ```
 
-Now that we have set these two adapters, let's generate an image from the combined adapters!
-
 <Tip>
 
 LoRA checkpoints in the diffusion community are almost always obtained with [DreamBooth](https://huggingface.co/docs/diffusers/main/en/training/dreambooth). DreamBooth training often relies on "trigger" words in the input text prompts in order for the generation results to look as expected. When you combine multiple LoRA checkpoints, it's important to ensure the trigger words for the corresponding LoRA checkpoints are present in the input text prompts.
 
 </Tip>
 
-The trigger words for [CiroN2022/toy-face](https://hf.co/CiroN2022/toy-face) and [nerijs/pixel-art-xl](https://hf.co/nerijs/pixel-art-xl) are found in their repositories.
-
+Remember to use the trigger words for [CiroN2022/toy-face](https://hf.co/CiroN2022/toy-face) and [nerijs/pixel-art-xl](https://hf.co/nerijs/pixel-art-xl) (these are found in their repositories) in the prompt to generate an image.
 
 ```python
-# Notice how the prompt is constructed.
 prompt = "toy_face of a hacker with a hoodie, pixel art"
 image = pipe(
     prompt, num_inference_steps=30, cross_attention_kwargs={"scale": 1.0}, generator=torch.manual_seed(0)
@@ -113,15 +103,16 @@ image
 
 ![toy-face-pixel-art](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/peft_integration/diffusers_peft_lora_inference_16_1.png)
 
-Impressive! As you can see, the model was able to generate an image that mixes the characteristics of both adapters.
+Impressive! As you can see, the model generated an image that mixed the characteristics of both adapters.
 
-If you want to go back to using only one adapter, use the [`~diffusers.loaders.UNet2DConditionLoadersMixin.set_adapters`] method to activate the `"toy"` adapter:
+> [!TIP]
+> Through its PEFT integration, Diffusers also offers more efficient merging methods which you can learn about in the [Merge LoRAs](../using-diffusers/merge_loras) guide!
+
+To return to only using one adapter, use the [`~diffusers.loaders.UNet2DConditionLoadersMixin.set_adapters`] method to activate the `"toy"` adapter:
 
 ```python
-# First, set the adapter.
 pipe.set_adapters("toy")
 
-# Then, run inference.
 prompt = "toy_face of a hacker with a hoodie"
 lora_scale= 0.9
 image = pipe(
@@ -130,11 +121,7 @@ image = pipe(
 image
 ```
 
-![toy-face-again](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/peft_integration/diffusers_peft_lora_inference_18_1.png)
-
-
-If you want to switch to only the base model, disable all LoRAs with the [`~diffusers.loaders.UNet2DConditionLoadersMixin.disable_lora`] method.
-
+Or to disable all adapters entirely, use the [`~diffusers.loaders.UNet2DConditionLoadersMixin.disable_lora`] method to return the base model.
 
 ```python
 pipe.disable_lora()
@@ -145,11 +132,9 @@ image = pipe(prompt, num_inference_steps=30, generator=torch.manual_seed(0)).ima
 image
 ```
 
-![no-lora](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/peft_integration/diffusers_peft_lora_inference_20_1.png)
+## Manage active adapters
 
-## Monitoring active adapters
-
-You have attached multiple adapters in this tutorial, and if you're feeling a bit lost on what adapters have been attached to the pipeline's components, you can easily check the list of active adapters using the [`~diffusers.loaders.LoraLoaderMixin.get_active_adapters`] method:
+You have attached multiple adapters in this tutorial, and if you're feeling a bit lost on what adapters have been attached to the pipeline's components, use the [`~diffusers.loaders.LoraLoaderMixin.get_active_adapters`] method to check the list of active adapters:
 
 ```py
 active_adapters = pipe.get_active_adapters()
@@ -164,78 +149,3 @@ list_adapters_component_wise = pipe.get_list_adapters()
 list_adapters_component_wise
 {"text_encoder": ["toy", "pixel"], "unet": ["toy", "pixel"], "text_encoder_2": ["toy", "pixel"]}
 ```
-
-## Compatibility with `torch.compile`
-
-If you want to compile your model with `torch.compile` make sure to first fuse the LoRA weights into the base model and unload them.
-
-```diff
-pipe.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
-pipe.load_lora_weights("CiroN2022/toy-face", weight_name="toy_face_sdxl.safetensors", adapter_name="toy")
-
-pipe.set_adapters(["pixel", "toy"], adapter_weights=[0.5, 1.0])
-# Fuses the LoRAs into the Unet
-pipe.fuse_lora()
-pipe.unload_lora_weights()
-
-+ pipe.unet.to(memory_format=torch.channels_last)
-+ pipe.unet = torch.compile(pipe.unet, mode="reduce-overhead", fullgraph=True)
-
-prompt = "toy_face of a hacker with a hoodie, pixel art"
-image = pipe(prompt, num_inference_steps=30, generator=torch.manual_seed(0)).images[0]
-```
-
-> [!TIP]
-> You can refer to the `torch.compile()` section [here](https://huggingface.co/docs/diffusers/main/en/optimization/torch2.0#torchcompile) and [here](https://huggingface.co/docs/diffusers/main/en/tutorials/fast_diffusion#torchcompile) for more elaborate examples.
-
-## Fusing adapters into the model
-
-You can use PEFT to easily fuse/unfuse multiple adapters directly into the model weights (both UNet and text encoder) using the [`~diffusers.loaders.LoraLoaderMixin.fuse_lora`] method, which can lead to a speed-up in inference and lower VRAM usage.
-
-```py
-pipe.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
-pipe.load_lora_weights("CiroN2022/toy-face", weight_name="toy_face_sdxl.safetensors", adapter_name="toy")
-
-pipe.set_adapters(["pixel", "toy"], adapter_weights=[0.5, 1.0])
-# Fuses the LoRAs into the Unet
-pipe.fuse_lora()
-
-prompt = "toy_face of a hacker with a hoodie, pixel art"
-image = pipe(prompt, num_inference_steps=30, generator=torch.manual_seed(0)).images[0]
-
-# Gets the Unet back to the original state
-pipe.unfuse_lora()
-```
-
-You can also fuse some adapters using `adapter_names` for faster generation:
-
-```py
-pipe.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
-pipe.load_lora_weights("CiroN2022/toy-face", weight_name="toy_face_sdxl.safetensors", adapter_name="toy")
-
-pipe.set_adapters(["pixel"], adapter_weights=[0.5, 1.0])
-# Fuses the LoRAs into the Unet
-pipe.fuse_lora(adapter_names=["pixel"])
-
-prompt = "a hacker with a hoodie, pixel art"
-image = pipe(prompt, num_inference_steps=30, generator=torch.manual_seed(0)).images[0]
-
-# Gets the Unet back to the original state
-pipe.unfuse_lora()
-
-# Fuse all adapters
-pipe.fuse_lora(adapter_names=["pixel", "toy"])
-
-prompt = "toy_face of a hacker with a hoodie, pixel art"
-image = pipe(prompt, num_inference_steps=30, generator=torch.manual_seed(0)).images[0]
-```
-
-## Saving a pipeline after fusing the adapters
-
-To properly save a pipeline after it's been loaded with the adapters, it should be serialized like so:
-
-```python
-pipe.fuse_lora(lora_scale=1.0)
-pipe.unload_lora_weights()
-pipe.save_pretrained("path-to-pipeline")
-```

docs/source/en/using-diffusers/callback.md

@@ -12,13 +12,18 @@ specific language governing permissions and limitations under the License.
 
 # Pipeline callbacks
 
-The denoising loop of a pipeline can be modified with custom defined functions using the `callback_on_step_end` parameter. This can be really useful for *dynamically* adjusting certain pipeline attributes, or modifying tensor variables. The flexibility of callbacks opens up some interesting use-cases such as changing the prompt embeddings at each timestep, assigning different weights to the prompt embeddings, and editing the guidance scale.
+The denoising loop of a pipeline can be modified with custom defined functions using the `callback_on_step_end` parameter. The callback function is executed at the end of each step, and modifies the pipeline attributes and variables for the next step. This is really useful for *dynamically* adjusting certain pipeline attributes or modifying tensor variables. This versatility allows for interesting use-cases such as changing the prompt embeddings at each timestep, assigning different weights to the prompt embeddings, and editing the guidance scale. With callbacks, you can implement new features without modifying the underlying code!
 
-This guide will show you how to use the `callback_on_step_end` parameter to disable classifier-free guidance (CFG) after 40% of the inference steps to save compute with minimal cost to performance.
+> [!TIP]
+> 🤗 Diffusers currently only supports `callback_on_step_end`, but feel free to open a [feature request](https://github.com/huggingface/diffusers/issues/new/choose) if you have a cool use-case and require a callback function with a different execution point!
 
-The callback function should have the following arguments:
+This guide will demonstrate how callbacks work by a few features you can implement with them.
 
-* `pipe` (or the pipeline instance) provides access to useful properties such as `num_timesteps` and `guidance_scale`. You can modify these properties by updating the underlying attributes. For this example, you'll disable CFG by setting `pipe._guidance_scale=0.0`.
+## Dynamic classifier-free guidance
+
+Dynamic classifier-free guidance (CFG) is a feature that allows you to disable CFG after a certain number of inference steps which can help you save compute with minimal cost to performance. The callback function for this should have the following arguments:
+
+* `pipeline` (or the pipeline instance) provides access to important properties such as `num_timesteps` and `guidance_scale`. You can modify these properties by updating the underlying attributes. For this example, you'll disable CFG by setting `pipeline._guidance_scale=0.0`.
 * `step_index` and `timestep` tell you where you are in the denoising loop. Use `step_index` to turn off CFG after reaching 40% of `num_timesteps`.
 * `callback_kwargs` is a dict that contains tensor variables you can modify during the denoising loop. It only includes variables specified in the `callback_on_step_end_tensor_inputs` argument, which is passed to the pipeline's `__call__` method. Different pipelines may use different sets of variables, so please check a pipeline's `_callback_tensor_inputs` attribute for the list of variables you can modify. Some common variables include `latents` and `prompt_embeds`. For this function, change the batch size of `prompt_embeds` after setting `guidance_scale=0.0` in order for it to work properly.
 
@@ -27,13 +32,13 @@ Your callback function should look something like this:
 ```python
 def callback_dynamic_cfg(pipe, step_index, timestep, callback_kwargs):
         # adjust the batch_size of prompt_embeds according to guidance_scale
-        if step_index == int(pipe.num_timesteps * 0.4):
+        if step_index == int(pipeline.num_timesteps * 0.4):
                 prompt_embeds = callback_kwargs["prompt_embeds"]
                 prompt_embeds = prompt_embeds.chunk(2)[-1]
 
-        # update guidance_scale and prompt_embeds
-        pipe._guidance_scale = 0.0
-        callback_kwargs["prompt_embeds"] = prompt_embeds
+                # update guidance_scale and prompt_embeds
+                pipeline._guidance_scale = 0.0
+                callback_kwargs["prompt_embeds"] = prompt_embeds
         return callback_kwargs
 ```
 
@@ -43,58 +48,134 @@ Now, you can pass the callback function to the `callback_on_step_end` parameter
 import torch
 from diffusers import StableDiffusionPipeline
 
-pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16)
-pipe = pipe.to("cuda")
+pipeline = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16)
+pipeline = pipeline.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
 
 generator = torch.Generator(device="cuda").manual_seed(1)
-out = pipe(prompt, generator=generator, callback_on_step_end=callback_dynamic_cfg, callback_on_step_end_tensor_inputs=['prompt_embeds'])
+out = pipeline(
+    prompt,
+    generator=generator,
+    callback_on_step_end=callback_dynamic_cfg,
+    callback_on_step_end_tensor_inputs=['prompt_embeds']
+)
 
 out.images[0].save("out_custom_cfg.png")
 ```
 
-The callback function is executed at the end of each denoising step, and modifies the pipeline attributes and tensor variables for the next denoising step.
-
-With callbacks, you can implement features such as dynamic CFG without having to modify the underlying code at all!
-
-<Tip>
-
-🤗 Diffusers currently only supports `callback_on_step_end`, but feel free to open a [feature request](https://github.com/huggingface/diffusers/issues/new/choose) if you have a cool use-case and require a callback function with a different execution point!
-
-</Tip>
-
 ## Interrupt the diffusion process
 
-Interrupting the diffusion process is particularly useful when building UIs that work with Diffusers because it allows users to stop the generation process if they're unhappy with the intermediate results. You can incorporate this into your pipeline with a callback.
+> [!TIP]
+> The interruption callback is supported for text-to-image, image-to-image, and inpainting for the [StableDiffusionPipeline](../api/pipelines/stable_diffusion/overview) and [StableDiffusionXLPipeline](../api/pipelines/stable_diffusion/stable_diffusion_xl).
 
-<Tip>
+Stopping the diffusion process early is useful when building UIs that work with Diffusers because it allows users to stop the generation process if they're unhappy with the intermediate results. You can incorporate this into your pipeline with a callback.
 
-The interruption callback is supported for text-to-image, image-to-image, and inpainting for the [StableDiffusionPipeline](../api/pipelines/stable_diffusion/overview) and [StableDiffusionXLPipeline](../api/pipelines/stable_diffusion/stable_diffusion_xl).
-
-</Tip>
-
-This callback function should take the following arguments: `pipe`, `i`, `t`, and `callback_kwargs` (this must be returned). Set the pipeline's `_interrupt` attribute to `True` to stop the diffusion process after a certain number of steps. You are also free to implement your own custom stopping logic inside the callback.
+This callback function should take the following arguments: `pipeline`, `i`, `t`, and `callback_kwargs` (this must be returned). Set the pipeline's `_interrupt` attribute to `True` to stop the diffusion process after a certain number of steps. You are also free to implement your own custom stopping logic inside the callback.
 
 In this example, the diffusion process is stopped after 10 steps even though `num_inference_steps` is set to 50.
 
 ```python
 from diffusers import StableDiffusionPipeline
 
-pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5")
-pipe.enable_model_cpu_offload()
+pipeline = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5")
+pipeline.enable_model_cpu_offload()
 num_inference_steps = 50
 
-def interrupt_callback(pipe, i, t, callback_kwargs):
+def interrupt_callback(pipeline, i, t, callback_kwargs):
     stop_idx = 10
     if i == stop_idx:
-        pipe._interrupt = True
+        pipeline._interrupt = True
 
     return callback_kwargs
 
-pipe(
+pipeline(
     "A photo of a cat",
     num_inference_steps=num_inference_steps,
     callback_on_step_end=interrupt_callback,
 )
 ```
+
+## Display image after each generation step
+
+> [!TIP]
+> This tip was contributed by [asomoza](https://github.com/asomoza).
+
+Display an image after each generation step by accessing and converting the latents after each step into an image. The latent space is compressed to 128x128, so the images are also 128x128 which is useful for a quick preview.
+
+1. Use the function below to convert the SDXL latents (4 channels) to RGB tensors (3 channels) as explained in the [Explaining the SDXL latent space](https://huggingface.co/blog/TimothyAlexisVass/explaining-the-sdxl-latent-space) blog post.
+
+```py
+def latents_to_rgb(latents):
+    weights = (
+        (60, -60, 25, -70),
+        (60,  -5, 15, -50),
+        (60,  10, -5, -35)
+    )
+
+    weights_tensor = torch.t(torch.tensor(weights, dtype=latents.dtype).to(latents.device))
+    biases_tensor = torch.tensor((150, 140, 130), dtype=latents.dtype).to(latents.device)
+    rgb_tensor = torch.einsum("...lxy,lr -> ...rxy", latents, weights_tensor) + biases_tensor.unsqueeze(-1).unsqueeze(-1)
+    image_array = rgb_tensor.clamp(0, 255)[0].byte().cpu().numpy()
+    image_array = image_array.transpose(1, 2, 0)
+
+    return Image.fromarray(image_array)
+```
+
+2. Create a function to decode and save the latents into an image.
+
+```py
+def decode_tensors(pipe, step, timestep, callback_kwargs):
+    latents = callback_kwargs["latents"]
+    
+    image = latents_to_rgb(latents)
+    image.save(f"{step}.png")
+
+    return callback_kwargs
+```
+
+3. Pass the `decode_tensors` function to the `callback_on_step_end` parameter to decode the tensors after each step. You also need to specify what you want to modify in the `callback_on_step_end_tensor_inputs` parameter, which in this case are the latents.
+
+```py
+from diffusers import AutoPipelineForText2Image
+import torch
+from PIL import Image
+
+pipeline = AutoPipelineForText2Image.from_pretrained(
+    "stabilityai/stable-diffusion-xl-base-1.0",
+    torch_dtype=torch.float16,
+    variant="fp16",
+    use_safetensors=True
+).to("cuda")
+
+image = pipe(
+    prompt = "A croissant shaped like a cute bear."
+    negative_prompt = "Deformed, ugly, bad anatomy"
+    callback_on_step_end=decode_tensors,
+    callback_on_step_end_tensor_inputs=["latents"],
+).images[0]
+```
+
+<div class="flex gap-4 justify-center">
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/tips_step_0.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">step 0</figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/tips_step_19.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">step 19
+    </figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/tips_step_29.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">step 29</figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/tips_step_39.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">step 39</figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/tips_step_49.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">step 49</figcaption>
+  </div>
+</div>

docs/source/en/using-diffusers/controlnet.md

@@ -429,6 +429,27 @@ image = pipe(
 make_image_grid([original_image, canny_image, image], rows=1, cols=3)
 ```
 
+<Tip>
+
+You can use a refiner model with `StableDiffusionXLControlNetPipeline` to improve image quality, just like you can with a regular `StableDiffusionXLPipeline`.
+See the [Refine image quality](./sdxl#refine-image-quality) section to learn how to use the refiner model.
+Make sure to use `StableDiffusionXLControlNetPipeline` and pass `image` and `controlnet_conditioning_scale`.
+
+```py
+base = StableDiffusionXLControlNetPipeline(...)
+image = base(
+    prompt=prompt,
+    controlnet_conditioning_scale=0.5,
+    image=canny_image,
+    num_inference_steps=40,
+    denoising_end=0.8,
+    output_type="latent",
+).images
+# rest exactly as with StableDiffusionXLPipeline
+```
+
+</Tip>
+
 ## MultiControlNet
 
 <Tip>

docs/source/en/using-diffusers/freeu.md

@@ -128,7 +128,7 @@ seed = 2023
 # The values come from
 # https://github.com/lyn-rgb/FreeU_Diffusers#video-pipelines
 pipe.enable_freeu(b1=1.2, b2=1.4, s1=0.9, s2=0.2)
-video_frames = pipe(prompt, height=320, width=576, num_frames=30, generator=torch.manual_seed(seed)).frames
+video_frames = pipe(prompt, height=320, width=576, num_frames=30, generator=torch.manual_seed(seed)).frames[0]
 export_to_video(video_frames, "astronaut_rides_horse.mp4")
 ```
 

docs/source/en/using-diffusers/inference_with_tcd_lora.md

@@ -0,0 +1,438 @@
+<!--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.
+-->
+
+[[open-in-colab]]
+
+# Trajectory Consistency Distillation-LoRA
+
+Trajectory Consistency Distillation (TCD) enables a model to generate higher quality and more detailed images with fewer steps. Moreover, owing to the effective error mitigation during the distillation process, TCD demonstrates superior performance even under conditions of large inference steps.
+
+The major advantages of TCD are:
+
+- Better than Teacher: TCD demonstrates superior generative quality at both small and large inference steps and exceeds the performance of [DPM-Solver++(2S)](../../api/schedulers/multistep_dpm_solver) with Stable Diffusion XL (SDXL). There is no additional discriminator or LPIPS supervision included during TCD training.
+
+- Flexible Inference Steps: The inference steps for TCD sampling can be freely adjusted without adversely affecting the image quality.
+
+- Freely change detail level: During inference, the level of detail in the image can be adjusted with a single hyperparameter, *gamma*.
+
+> [!TIP]
+> For more technical details of TCD, please refer to the [paper](https://arxiv.org/abs/2402.19159) or official [project page](https://mhh0318.github.io/tcd/)).
+
+For large models like SDXL, TCD is trained with [LoRA](https://huggingface.co/docs/peft/conceptual_guides/adapter#low-rank-adaptation-lora) to reduce memory usage. This is also useful because you can reuse LoRAs between different finetuned models, as long as they share the same base model, without further training.
+
+
+
+This guide will show you how to perform inference with TCD-LoRAs for a variety of tasks like text-to-image and inpainting, as well as how you can easily combine TCD-LoRAs with other adapters. Choose one of the supported base model and it's corresponding TCD-LoRA checkpoint from the table below to get started.
+
+| Base model                                                                                      | TCD-LoRA checkpoint                                            |
+|-------------------------------------------------------------------------------------------------|----------------------------------------------------------------|
+| [stable-diffusion-v1-5](https://huggingface.co/runwayml/stable-diffusion-v1-5)                  | [TCD-SD15](https://huggingface.co/h1t/TCD-SD15-LoRA)           |
+| [stable-diffusion-2-1-base](https://huggingface.co/stabilityai/stable-diffusion-2-1-base)       | [TCD-SD21-base](https://huggingface.co/h1t/TCD-SD21-base-LoRA) |
+| [stable-diffusion-xl-base-1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0) | [TCD-SDXL](https://huggingface.co/h1t/TCD-SDXL-LoRA)           |
+
+
+Make sure you have [PEFT](https://github.com/huggingface/peft) installed for better LoRA support.
+
+```bash
+pip install -U peft
+```
+
+## General tasks
+
+In this guide, let's use the [`StableDiffusionXLPipeline`] and the [`TCDScheduler`]. Use the [`~StableDiffusionPipeline.load_lora_weights`] method to load the SDXL-compatible TCD-LoRA weights.
+
+A few tips to keep in mind for TCD-LoRA inference are to:
+
+- Keep the `num_inference_steps` between 4 and 50
+- Set `eta` (used to control stochasticity at each step) between 0 and 1. You should use a higher `eta` when increasing the number of inference steps, but the downside is that a larger `eta` in [`TCDScheduler`] leads to blurrier images. A value of 0.3 is recommended to produce good results.
+
+<hfoptions id="tasks">
+<hfoption id="text-to-image">
+
+```python
+import torch
+from diffusers import StableDiffusionXLPipeline, TCDScheduler
+
+device = "cuda"
+base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
+tcd_lora_id = "h1t/TCD-SDXL-LoRA"
+
+pipe = StableDiffusionXLPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to(device)
+pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+
+pipe.load_lora_weights(tcd_lora_id)
+pipe.fuse_lora()
+
+prompt = "Painting of the orange cat Otto von Garfield, Count of Bismarck-Schönhausen, Duke of Lauenburg, Minister-President of Prussia. Depicted wearing a Prussian Pickelhaube and eating his favorite meal - lasagna."
+
+image = pipe(
+    prompt=prompt,
+    num_inference_steps=4,
+    guidance_scale=0,
+    eta=0.3, 
+    generator=torch.Generator(device=device).manual_seed(0),
+).images[0]
+```
+
+![](https://github.com/jabir-zheng/TCD/raw/main/assets/demo_image.png)
+
+</hfoption>
+
+<hfoption id="inpainting">
+
+```python
+import torch
+from diffusers import AutoPipelineForInpainting, TCDScheduler
+from diffusers.utils import load_image, make_image_grid
+
+device = "cuda"
+base_model_id = "diffusers/stable-diffusion-xl-1.0-inpainting-0.1"
+tcd_lora_id = "h1t/TCD-SDXL-LoRA"
+
+pipe = AutoPipelineForInpainting.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to(device)
+pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+
+pipe.load_lora_weights(tcd_lora_id)
+pipe.fuse_lora()
+
+img_url = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png"
+mask_url = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo_mask.png"
+
+init_image = load_image(img_url).resize((1024, 1024))
+mask_image = load_image(mask_url).resize((1024, 1024))
+
+prompt = "a tiger sitting on a park bench"
+
+image = pipe(
+  prompt=prompt,
+  image=init_image,
+  mask_image=mask_image,
+  num_inference_steps=8,
+  guidance_scale=0,
+  eta=0.3,
+  strength=0.99,  # make sure to use `strength` below 1.0
+  generator=torch.Generator(device=device).manual_seed(0),
+).images[0]
+
+grid_image = make_image_grid([init_image, mask_image, image], rows=1, cols=3)
+```
+
+![](https://github.com/jabir-zheng/TCD/raw/main/assets/inpainting_tcd.png)
+
+
+</hfoption>
+</hfoptions>
+
+## Community models
+
+TCD-LoRA also works with many community finetuned models and plugins. For example, load the [animagine-xl-3.0](https://huggingface.co/cagliostrolab/animagine-xl-3.0) checkpoint which is a community finetuned version of SDXL for generating anime images.
+
+```python
+import torch
+from diffusers import StableDiffusionXLPipeline, TCDScheduler
+
+device = "cuda"
+base_model_id = "cagliostrolab/animagine-xl-3.0"
+tcd_lora_id = "h1t/TCD-SDXL-LoRA"
+
+pipe = StableDiffusionXLPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to(device)
+pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+
+pipe.load_lora_weights(tcd_lora_id)
+pipe.fuse_lora()
+
+prompt = "A man, clad in a meticulously tailored military uniform, stands with unwavering resolve. The uniform boasts intricate details, and his eyes gleam with determination. Strands of vibrant, windswept hair peek out from beneath the brim of his cap."
+
+image = pipe(
+    prompt=prompt,
+    num_inference_steps=8,
+    guidance_scale=0,
+    eta=0.3, 
+    generator=torch.Generator(device=device).manual_seed(0),
+).images[0]
+```
+
+![](https://github.com/jabir-zheng/TCD/raw/main/assets/animagine_xl.png)
+
+TCD-LoRA also supports other LoRAs trained on different styles. For example, let's load the [TheLastBen/Papercut_SDXL](https://huggingface.co/TheLastBen/Papercut_SDXL) LoRA and fuse it with the TCD-LoRA with the [`~loaders.UNet2DConditionLoadersMixin.set_adapters`] method. 
+
+> [!TIP]
+> Check out the [Merge LoRAs](merge_loras) guide to learn more about efficient merging methods.
+
+```python
+import torch
+from diffusers import StableDiffusionXLPipeline
+from scheduling_tcd import TCDScheduler 
+
+device = "cuda"
+base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
+tcd_lora_id = "h1t/TCD-SDXL-LoRA"
+styled_lora_id = "TheLastBen/Papercut_SDXL"
+
+pipe = StableDiffusionXLPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to(device)
+pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+
+pipe.load_lora_weights(tcd_lora_id, adapter_name="tcd")
+pipe.load_lora_weights(styled_lora_id, adapter_name="style")
+pipe.set_adapters(["tcd", "style"], adapter_weights=[1.0, 1.0])
+
+prompt = "papercut of a winter mountain, snow"
+
+image = pipe(
+    prompt=prompt,
+    num_inference_steps=4,
+    guidance_scale=0,
+    eta=0.3, 
+    generator=torch.Generator(device=device).manual_seed(0),
+).images[0]
+```
+
+![](https://github.com/jabir-zheng/TCD/raw/main/assets/styled_lora.png)
+
+
+## Adapters
+
+TCD-LoRA is very versatile, and it can be combined with other adapter types like ControlNets, IP-Adapter, and AnimateDiff.
+
+<hfoptions id="adapters">
+<hfoption id="ControlNet">
+
+### Depth ControlNet
+
+```python
+import torch
+import numpy as np
+from PIL import Image
+from transformers import DPTFeatureExtractor, DPTForDepthEstimation
+from diffusers import ControlNetModel, StableDiffusionXLControlNetPipeline
+from diffusers.utils import load_image, make_image_grid
+from scheduling_tcd import TCDScheduler 
+
+device = "cuda"
+depth_estimator = DPTForDepthEstimation.from_pretrained("Intel/dpt-hybrid-midas").to(device)
+feature_extractor = DPTFeatureExtractor.from_pretrained("Intel/dpt-hybrid-midas")
+
+def get_depth_map(image):
+    image = feature_extractor(images=image, return_tensors="pt").pixel_values.to(device)
+    with torch.no_grad(), torch.autocast(device):
+        depth_map = depth_estimator(image).predicted_depth
+
+    depth_map = torch.nn.functional.interpolate(
+        depth_map.unsqueeze(1),
+        size=(1024, 1024),
+        mode="bicubic",
+        align_corners=False,
+    )
+    depth_min = torch.amin(depth_map, dim=[1, 2, 3], keepdim=True)
+    depth_max = torch.amax(depth_map, dim=[1, 2, 3], keepdim=True)
+    depth_map = (depth_map - depth_min) / (depth_max - depth_min)
+    image = torch.cat([depth_map] * 3, dim=1)
+
+    image = image.permute(0, 2, 3, 1).cpu().numpy()[0]
+    image = Image.fromarray((image * 255.0).clip(0, 255).astype(np.uint8))
+    return image
+
+base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
+controlnet_id = "diffusers/controlnet-depth-sdxl-1.0"
+tcd_lora_id = "h1t/TCD-SDXL-LoRA"
+
+controlnet = ControlNetModel.from_pretrained(
+    controlnet_id,
+    torch_dtype=torch.float16,
+    variant="fp16",
+).to(device)
+pipe = StableDiffusionXLControlNetPipeline.from_pretrained(
+    base_model_id,
+    controlnet=controlnet,
+    torch_dtype=torch.float16,
+    variant="fp16",
+).to(device)
+pipe.enable_model_cpu_offload()
+
+pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+
+pipe.load_lora_weights(tcd_lora_id)
+pipe.fuse_lora()
+
+prompt = "stormtrooper lecture, photorealistic"
+
+image = load_image("https://huggingface.co/lllyasviel/sd-controlnet-depth/resolve/main/images/stormtrooper.png")
+depth_image = get_depth_map(image)
+
+controlnet_conditioning_scale = 0.5  # recommended for good generalization
+
+image = pipe(
+    prompt, 
+    image=depth_image, 
+    num_inference_steps=4, 
+    guidance_scale=0,
+    eta=0.3,
+    controlnet_conditioning_scale=controlnet_conditioning_scale,
+    generator=torch.Generator(device=device).manual_seed(0),
+).images[0]
+
+grid_image = make_image_grid([depth_image, image], rows=1, cols=2)
+```
+
+![](https://github.com/jabir-zheng/TCD/raw/main/assets/controlnet_depth_tcd.png)
+
+### Canny ControlNet
+```python
+import torch
+from diffusers import ControlNetModel, StableDiffusionXLControlNetPipeline
+from diffusers.utils import load_image, make_image_grid
+from scheduling_tcd import TCDScheduler 
+
+device = "cuda"
+base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
+controlnet_id = "diffusers/controlnet-canny-sdxl-1.0"
+tcd_lora_id = "h1t/TCD-SDXL-LoRA"
+
+controlnet = ControlNetModel.from_pretrained(
+    controlnet_id,
+    torch_dtype=torch.float16,
+    variant="fp16",
+).to(device)
+pipe = StableDiffusionXLControlNetPipeline.from_pretrained(
+    base_model_id,
+    controlnet=controlnet,
+    torch_dtype=torch.float16,
+    variant="fp16",
+).to(device)
+pipe.enable_model_cpu_offload()
+
+pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+
+pipe.load_lora_weights(tcd_lora_id)
+pipe.fuse_lora()
+
+prompt = "ultrarealistic shot of a furry blue bird"
+
+canny_image = load_image("https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd_controlnet/bird_canny.png")
+
+controlnet_conditioning_scale = 0.5  # recommended for good generalization
+
+image = pipe(
+    prompt, 
+    image=canny_image, 
+    num_inference_steps=4, 
+    guidance_scale=0,
+    eta=0.3,
+    controlnet_conditioning_scale=controlnet_conditioning_scale,
+    generator=torch.Generator(device=device).manual_seed(0),
+).images[0]
+
+grid_image = make_image_grid([canny_image, image], rows=1, cols=2)
+```
+![](https://github.com/jabir-zheng/TCD/raw/main/assets/controlnet_canny_tcd.png)
+
+<Tip>
+The inference parameters in this example might not work for all examples, so we recommend you to try different values for `num_inference_steps`, `guidance_scale`, `controlnet_conditioning_scale` and `cross_attention_kwargs` parameters and choose the best one. 
+</Tip>
+
+</hfoption>
+<hfoption id="IP-Adapter">
+
+This example shows how to use the TCD-LoRA with the [IP-Adapter](https://github.com/tencent-ailab/IP-Adapter/tree/main) and SDXL.
+
+```python
+import torch
+from diffusers import StableDiffusionXLPipeline
+from diffusers.utils import load_image, make_image_grid
+
+from ip_adapter import IPAdapterXL
+from scheduling_tcd import TCDScheduler 
+
+device = "cuda"
+base_model_path = "stabilityai/stable-diffusion-xl-base-1.0"
+image_encoder_path = "sdxl_models/image_encoder"
+ip_ckpt = "sdxl_models/ip-adapter_sdxl.bin"
+tcd_lora_id = "h1t/TCD-SDXL-LoRA"
+
+pipe = StableDiffusionXLPipeline.from_pretrained(
+    base_model_path, 
+    torch_dtype=torch.float16, 
+    variant="fp16"
+)
+pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+
+pipe.load_lora_weights(tcd_lora_id)
+pipe.fuse_lora()
+
+ip_model = IPAdapterXL(pipe, image_encoder_path, ip_ckpt, device)
+
+ref_image = load_image("https://raw.githubusercontent.com/tencent-ailab/IP-Adapter/main/assets/images/woman.png").resize((512, 512))
+
+prompt = "best quality, high quality, wearing sunglasses"
+
+image = ip_model.generate(
+    pil_image=ref_image, 
+    prompt=prompt,
+    scale=0.5,
+    num_samples=1, 
+    num_inference_steps=4, 
+    guidance_scale=0,
+    eta=0.3, 
+    seed=0,
+)[0]
+
+grid_image = make_image_grid([ref_image, image], rows=1, cols=2)
+```
+
+![](https://github.com/jabir-zheng/TCD/raw/main/assets/ip_adapter.png)
+
+
+
+</hfoption>
+<hfoption id="AnimateDiff">
+
+[`AnimateDiff`] allows animating images using Stable Diffusion models. TCD-LoRA can substantially accelerate the process without degrading image quality. The quality of animation with TCD-LoRA and AnimateDiff has a more lucid outcome.
+
+```python
+import torch
+from diffusers import MotionAdapter, AnimateDiffPipeline, DDIMScheduler
+from scheduling_tcd import TCDScheduler
+from diffusers.utils import export_to_gif
+
+adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5")
+pipe = AnimateDiffPipeline.from_pretrained(
+    "frankjoshua/toonyou_beta6",
+    motion_adapter=adapter,
+).to("cuda")
+
+# set TCDScheduler
+pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
+
+# load TCD LoRA
+pipe.load_lora_weights("h1t/TCD-SD15-LoRA", adapter_name="tcd")
+pipe.load_lora_weights("guoyww/animatediff-motion-lora-zoom-in", weight_name="diffusion_pytorch_model.safetensors", adapter_name="motion-lora")
+
+pipe.set_adapters(["tcd", "motion-lora"], adapter_weights=[1.0, 1.2])
+
+prompt = "best quality, masterpiece, 1girl, looking at viewer, blurry background, upper body, contemporary, dress"
+generator = torch.manual_seed(0)
+frames = pipe(
+    prompt=prompt,
+    num_inference_steps=5,
+    guidance_scale=0,
+    cross_attention_kwargs={"scale": 1},
+    num_frames=24,
+    eta=0.3,
+    generator=generator
+).frames[0]
+export_to_gif(frames, "animation.gif")
+```
+
+![](https://github.com/jabir-zheng/TCD/raw/main/assets/animation_example.gif)
+
+</hfoption>
+</hfoptions>

docs/source/en/using-diffusers/ip_adapter.md

@@ -25,6 +25,9 @@ Let's take a look at how to use IP-Adapter's image prompting capabilities with t
 
 In all the following examples, you'll see the [`~loaders.IPAdapterMixin.set_ip_adapter_scale`] method. This method controls the amount of text or image conditioning to apply to the model. A value of `1.0` means the model is only conditioned on the image prompt. Lowering this value encourages the model to produce more diverse images, but they may not be as aligned with the image prompt. Typically, a value of `0.5` achieves a good balance between the two prompt types and produces good results.
 
+> [!TIP]
+> In the examples below, try adding `low_cpu_mem_usage=True` to the [`~loaders.IPAdapterMixin.load_ip_adapter`] method to speed up the loading time.
+
 <hfoptions id="tasks">
 <hfoption id="Text-to-image">
 
@@ -231,10 +234,21 @@ export_to_gif(frames, "gummy_bear.gif")
 </hfoption>
 </hfoptions>
 
-> [!TIP]
-> While calling `load_ip_adapter()`, pass `low_cpu_mem_usage=True` to speed up the loading time.
+## Configure parameters
 
-All the pipelines supporting IP-Adapter accept a `ip_adapter_image_embeds` argument. If you need to run the IP-Adapter multiple times with the same image, you can encode the image once and save the embedding to the disk.
+There are a couple of IP-Adapter parameters that are useful to know about and can help you with your image generation tasks. These parameters can make your workflow more efficient or give you more control over image generation.
+
+### Image embeddings
+
+IP-Adapter enabled pipelines provide the `ip_adapter_image_embeds` parameter to accept precomputed image embeddings. This is particularly useful in scenarios where you need to run the IP-Adapter pipeline multiple times because you have more than one image. For example, [multi IP-Adapter](#multi-ip-adapter) is a specific use case where you provide multiple styling images to generate a specific image in a specific style. Loading and encoding multiple images each time you use the pipeline would be inefficient. Instead, you can precompute and save the image embeddings to disk (which can save a lot of space if you're using high-quality images) and load them when you need them.
+
+> [!TIP]
+> This parameter also gives you the flexibility to load embeddings from other sources. For example, ComfyUI image embeddings for IP-Adapters are compatible with Diffusers and should work ouf-of-the-box!
+
+Call the [`~StableDiffusionPipeline.prepare_ip_adapter_image_embeds`] method to encode and generate the image embeddings. Then you can save them to disk with `torch.save`.
+
+> [!TIP]
+> If you're using IP-Adapter with `ip_adapter_image_embedding` instead of `ip_adapter_image`', you can set `load_ip_adapter(image_encoder_folder=None,...)` because you don't need to load an encoder to generate the image embeddings.
 
 ```py
 image_embeds = pipeline.prepare_ip_adapter_image_embeds(
@@ -248,10 +262,7 @@ image_embeds = pipeline.prepare_ip_adapter_image_embeds(
 torch.save(image_embeds, "image_embeds.ipadpt")
 ```
 
-Load the image embedding and pass it to the pipeline as `ip_adapter_image_embeds`
-
-> [!TIP]
-> ComfyUI image embeddings for IP-Adapters are fully compatible in Diffusers and should work out-of-box.
+Now load the image embeddings by passing them to the `ip_adapter_image_embeds` parameter.
 
 ```py
 image_embeds = torch.load("image_embeds.ipadpt")
@@ -264,8 +275,86 @@ images = pipeline(
 ).images
 ```
 
-> [!TIP]
-> If you use IP-Adapter with `ip_adapter_image_embedding` instead of `ip_adapter_image`, you can choose not to load an image encoder by passing `image_encoder_folder=None` to `load_ip_adapter()`. 
+### IP-Adapter masking
+
+Binary masks specify which portion of the output image should be assigned to an IP-Adapter. This is useful for composing more than one IP-Adapter image. For each input IP-Adapter image, you must provide a binary mask an an IP-Adapter.
+
+To start, preprocess the input IP-Adapter images with the [`~image_processor.IPAdapterMaskProcessor.preprocess()`] to generate their masks. For optimal results, provide the output height and width to [`~image_processor.IPAdapterMaskProcessor.preprocess()`]. This ensures masks with different aspect ratios are appropriately stretched. If the input masks already match the aspect ratio of the generated image, you don't have to set the `height` and `width`.
+
+```py
+from diffusers.image_processor import IPAdapterMaskProcessor
+
+mask1 = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ip_mask_mask1.png")
+mask2 = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ip_mask_mask2.png")
+
+output_height = 1024
+output_width = 1024
+
+processor = IPAdapterMaskProcessor()
+masks = processor.preprocess([mask1, mask2], height=output_height, width=output_width)
+```
+
+<div class="flex flex-row gap-4">
+  <div class="flex-1">
+    <img class="rounded-xl" src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_mask1.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">mask one</figcaption>
+  </div>
+  <div class="flex-1">
+    <img class="rounded-xl" src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_mask2.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">mask two</figcaption>
+  </div>
+</div>
+
+When there is more than one input IP-Adapter image, load them as a list to ensure each image is assigned to a different IP-Adapter. Each of the input IP-Adapter images here correspond to the masks generated above.
+
+```py
+face_image1 = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ip_mask_girl1.png")
+face_image2 = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ip_mask_girl2.png")
+
+ip_images = [[face_image1], [face_image2]]
+```
+
+<div class="flex flex-row gap-4">
+  <div class="flex-1">
+    <img class="rounded-xl" src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_girl1.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">IP-Adapter image one</figcaption>
+  </div>
+  <div class="flex-1">
+    <img class="rounded-xl" src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_girl2.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">IP-Adapter image two</figcaption>
+  </div>
+</div>
+
+Now pass the preprocessed masks to `cross_attention_kwargs` in the pipeline call.
+
+```py
+pipeline.load_ip_adapter("h94/IP-Adapter", subfolder="sdxl_models", weight_name=["ip-adapter-plus-face_sdxl_vit-h.safetensors"] * 2)
+pipeline.set_ip_adapter_scale([0.7] * 2)
+generator = torch.Generator(device="cpu").manual_seed(0)
+num_images = 1
+
+image = pipeline(
+    prompt="2 girls",
+    ip_adapter_image=ip_images,
+    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality",
+    num_inference_steps=20,
+    num_images_per_prompt=num_images,
+    generator=generator,
+    cross_attention_kwargs={"ip_adapter_masks": masks}
+).images[0]
+image
+```
+
+<div class="flex flex-row gap-4">
+  <div class="flex-1">
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ip_adapter_attention_mask_result_seed_0.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">IP-Adapter masking applied</figcaption>
+  </div>
+  <div class="flex-1">
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ip_adapter_no_attention_mask_result_seed_0.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">no IP-Adapter masking applied</figcaption>
+  </div>
+</div>
 
 ## Specific use cases
 
@@ -279,6 +368,7 @@ Generating accurate faces is challenging because they are complex and nuanced. D
 * [ip-adapter-plus-face_sd15.safetensors](https://huggingface.co/h94/IP-Adapter/blob/main/models/ip-adapter-plus-face_sd15.safetensors) uses patch embeddings and is conditioned with images of cropped faces
 
 > [!TIP]
+>
 > [IP-Adapter-FaceID](https://huggingface.co/h94/IP-Adapter-FaceID) is a face-specific IP-Adapter trained with face ID embeddings instead of CLIP image embeddings, allowing you to generate more consistent faces in different contexts and styles. Try out this popular [community pipeline](https://github.com/huggingface/diffusers/tree/main/examples/community#ip-adapter-face-id) and see how it compares to the other face IP-Adapters.
 
 For face models, use the [h94/IP-Adapter](https://huggingface.co/h94/IP-Adapter) checkpoint. It is also recommended to use [`DDIMScheduler`] or [`EulerDiscreteScheduler`] for face models.
@@ -502,82 +592,3 @@ image
 <div class="flex justify-center">
     <img src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ipa-controlnet-out.png" />
 </div>
-
-### IP-Adapter masking
-
-Binary masks can be used to specify which portion of the output image should be assigned to an IP-Adapter.
-For each input IP-Adapter image, a binary mask and an IP-Adapter must be provided.
-
-Before passing the masks to the pipeline, it's essential to preprocess them using [`IPAdapterMaskProcessor.preprocess()`].
-
-> [!TIP]
-> For optimal results, provide the output height and width to [`IPAdapterMaskProcessor.preprocess()`]. This ensures that masks with differing aspect ratios are appropriately stretched. If the input masks already match the aspect ratio of the generated image, specifying height and width can be omitted.
-
-Here an example with two masks:
-
-```py
-from diffusers.image_processor import IPAdapterMaskProcessor
-
-mask1 = load_image("https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_mask1.png")
-mask2 = load_image("https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_mask2.png")
-
-output_height = 1024
-output_width = 1024
-
-processor = IPAdapterMaskProcessor()
-masks = processor.preprocess([mask1, mask2], height=output_height, width=output_width)
-```
-
-<div class="flex flex-row gap-4">
-  <div class="flex-1">
-    <img class="rounded-xl" src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_mask1.png"/>
-    <figcaption class="mt-2 text-center text-sm text-gray-500">mask one</figcaption>
-  </div>
-  <div class="flex-1">
-    <img class="rounded-xl" src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_mask2.png"/>
-    <figcaption class="mt-2 text-center text-sm text-gray-500">mask two</figcaption>
-  </div>
-</div>
-
-If you have more than one IP-Adapter image, load them into a list, ensuring each image is assigned to a different IP-Adapter.
-
-```py
-face_image1 = load_image("https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_girl1.png")
-face_image2 = load_image("https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_girl2.png")
-
-ip_images = [[face_image1], [face_image2]]
-```
-
-<div class="flex flex-row gap-4">
-  <div class="flex-1">
-    <img class="rounded-xl" src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_girl1.png"/>
-    <figcaption class="mt-2 text-center text-sm text-gray-500">ip adapter image one</figcaption>
-  </div>
-  <div class="flex-1">
-    <img class="rounded-xl" src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip_mask_girl2.png"/>
-    <figcaption class="mt-2 text-center text-sm text-gray-500">ip adapter image two</figcaption>
-  </div>
-</div>
-
-Pass preprocessed masks to the pipeline using `cross_attention_kwargs` as shown below:
-
-```py
-pipeline.load_ip_adapter("h94/IP-Adapter", subfolder="sdxl_models", weight_name=["ip-adapter-plus-face_sdxl_vit-h.safetensors"] * 2)
-pipeline.set_ip_adapter_scale([0.7] * 2)
-generator = torch.Generator(device="cpu").manual_seed(0)
-num_images = 1
-
-image = pipeline(
-    prompt="2 girls",
-    ip_adapter_image=ip_images,
-    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality",
-    num_inference_steps=20, num_images_per_prompt=num_images,
-    generator=generator, cross_attention_kwargs={"ip_adapter_masks": masks}
-).images[0]
-image
-```
-
-<div class="flex justify-center">
-    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ip_adapter_attention_mask_result_seed_0.png" />
-   <figcaption class="mt-2 text-center text-sm text-gray-500">output image</figcaption>
-</div>

docs/source/en/using-diffusers/loading_adapters.md

@@ -103,7 +103,7 @@ image
 
 <Tip>
 
-LoRA is a very general training technique that can be used with other training methods. For example, it is common to train a model with DreamBooth and LoRA.
+LoRA is a very general training technique that can be used with other training methods. For example, it is common to train a model with DreamBooth and LoRA. It is also increasingly common to load and merge multiple LoRAs to create new and unique images. You can learn more about it in the in-depth [Merge LoRAs](merge_loras) guide since merging is outside the scope of this loading guide.
 
 </Tip>
 
@@ -165,101 +165,14 @@ To unload the LoRA weights, use the [`~loaders.LoraLoaderMixin.unload_lora_weigh
 pipeline.unload_lora_weights()
 ```
 
-### Load multiple LoRAs
-
-It can be fun to use multiple LoRAs together to create something entirely new and unique. The [`~loaders.LoraLoaderMixin.fuse_lora`] method allows you to fuse the LoRA weights with the original weights of the underlying model.
-
-<Tip>
-
-Fusing the weights can lead to a speedup in inference latency because you don't need to separately load the base model and LoRA! You can save your fused pipeline with [`~DiffusionPipeline.save_pretrained`] to avoid loading and fusing the weights every time you want to use the model.
-
-</Tip>
-
-Load an initial model:
-
-```py
-from diffusers import StableDiffusionXLPipeline, AutoencoderKL
-import torch
-
-vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
-pipeline = StableDiffusionXLPipeline.from_pretrained(
-    "stabilityai/stable-diffusion-xl-base-1.0",
-    vae=vae,
-    torch_dtype=torch.float16,
-).to("cuda")
-```
-
-Next, load the LoRA checkpoint and fuse it with the original weights. The `lora_scale` parameter controls how much to scale the output by with the LoRA weights. It is important to make the `lora_scale` adjustments in the [`~loaders.LoraLoaderMixin.fuse_lora`] method because it won't work if you try to pass `scale` to the `cross_attention_kwargs` in the pipeline.
-
-If you need to reset the original model weights for any reason (use a different `lora_scale`), you should use the [`~loaders.LoraLoaderMixin.unfuse_lora`] method.
-
-```py
-pipeline.load_lora_weights("ostris/ikea-instructions-lora-sdxl")
-pipeline.fuse_lora(lora_scale=0.7)
-
-# to unfuse the LoRA weights
-pipeline.unfuse_lora()
-```
-
-Then fuse this pipeline with the next set of LoRA weights:
-
-```py
-pipeline.load_lora_weights("ostris/super-cereal-sdxl-lora")
-pipeline.fuse_lora(lora_scale=0.7)
-```
-
-<Tip warning={true}>
-
-You can't unfuse multiple LoRA checkpoints, so if you need to reset the model to its original weights, you'll need to reload it.
-
-</Tip>
-
-Now you can generate an image that uses the weights from both LoRAs:
-
-```py
-prompt = "A cute brown bear eating a slice of pizza, stunning color scheme, masterpiece, illustration"
-image = pipeline(prompt).images[0]
-image
-```
-
-### 🤗 PEFT
-
-<Tip>
-
-Read the [Inference with 🤗 PEFT](../tutorials/using_peft_for_inference) tutorial to learn more about its integration with 🤗 Diffusers and how you can easily work with and juggle multiple adapters. You'll need to install 🤗 Diffusers and PEFT from source to run the example in this section.
-
-</Tip>
-
-Another way you can load and use multiple LoRAs is to specify the `adapter_name` parameter in [`~loaders.LoraLoaderMixin.load_lora_weights`]. This method takes advantage of the 🤗 PEFT integration. For example, load and name both LoRA weights:
-
-```py
-from diffusers import DiffusionPipeline
-import torch
-
-pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16).to("cuda")
-pipeline.load_lora_weights("ostris/ikea-instructions-lora-sdxl", weight_name="ikea_instructions_xl_v1_5.safetensors", adapter_name="ikea")
-pipeline.load_lora_weights("ostris/super-cereal-sdxl-lora", weight_name="cereal_box_sdxl_v1.safetensors", adapter_name="cereal")
-```
-
-Now use the [`~loaders.UNet2DConditionLoadersMixin.set_adapters`] to activate both LoRAs, and you can configure how much weight each LoRA should have on the output:
-
-```py
-pipeline.set_adapters(["ikea", "cereal"], adapter_weights=[0.7, 0.5])
-```
-
-Then, generate an image:
-
-```py
-prompt = "A cute brown bear eating a slice of pizza, stunning color scheme, masterpiece, illustration"
-image = pipeline(prompt, num_inference_steps=30, cross_attention_kwargs={"scale": 1.0}).images[0]
-image
-```
-
 ### Kohya and TheLastBen
 
 Other popular LoRA trainers from the community include those by [Kohya](https://github.com/kohya-ss/sd-scripts/) and [TheLastBen](https://github.com/TheLastBen/fast-stable-diffusion). These trainers create different LoRA checkpoints than those trained by 🤗 Diffusers, but they can still be loaded in the same way.
 
-Let's download the [Blueprintify SD XL 1.0](https://civitai.com/models/150986/blueprintify-sd-xl-10) checkpoint from [Civitai](https://civitai.com/):
+<hfoptions id="other-trainers">
+<hfoption id="Kohya">
+
+To load a Kohya LoRA, let's download the [Blueprintify SD XL 1.0](https://civitai.com/models/150986/blueprintify-sd-xl-10) checkpoint from [Civitai](https://civitai.com/) as an example:
 
 ```sh
 !wget https://civitai.com/api/download/models/168776 -O blueprintify-sd-xl-10.safetensors
@@ -293,6 +206,9 @@ Some limitations of using Kohya LoRAs with 🤗 Diffusers include:
 
 </Tip>
 
+</hfoption>
+<hfoption id="TheLastBen">
+
 Loading a checkpoint from TheLastBen is very similar. For example, to load the [TheLastBen/William_Eggleston_Style_SDXL](https://huggingface.co/TheLastBen/William_Eggleston_Style_SDXL) checkpoint:
 
 ```py
@@ -308,6 +224,9 @@ image = pipeline(prompt=prompt).images[0]
 image
 ```
 
+</hfoption>
+</hfoptions>
+
 ## IP-Adapter
 
 [IP-Adapter](https://ip-adapter.github.io/) is a lightweight adapter that enables image prompting for any diffusion model. This adapter works by decoupling the cross-attention layers of the image and text features. All the other model components are frozen and only the embedded image features in the UNet are trained. As a result, IP-Adapter files are typically only ~100MBs.

docs/source/en/using-diffusers/merge_loras.md

@@ -0,0 +1,266 @@
+<!--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.
+-->
+
+# Merge LoRAs
+
+It can be fun and creative to use multiple [LoRAs]((https://huggingface.co/docs/peft/conceptual_guides/adapter#low-rank-adaptation-lora)) together to generate something entirely new and unique. This works by merging multiple LoRA weights together to produce images that are a blend of different styles. Diffusers provides a few methods to merge LoRAs depending on *how* you want to merge their weights, which can affect image quality.
+
+This guide will show you how to merge LoRAs using the [`~loaders.UNet2DConditionLoadersMixin.set_adapters`] and [`~peft.LoraModel.add_weighted_adapter`] methods. To improve inference speed and reduce memory-usage of merged LoRAs, you'll also see how to use the [`~loaders.LoraLoaderMixin.fuse_lora`] method to fuse the LoRA weights with the original weights of the underlying model.
+
+For this guide, load a Stable Diffusion XL (SDXL) checkpoint and the [KappaNeuro/studio-ghibli-style]() and [Norod78/sdxl-chalkboarddrawing-lora]() LoRAs with the [`~loaders.LoraLoaderMixin.load_lora_weights`] method. You'll need to assign each LoRA an `adapter_name` to combine them later.
+
+```py
+from diffusers import DiffusionPipeline
+import torch
+
+pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16).to("cuda")
+pipeline.load_lora_weights("ostris/ikea-instructions-lora-sdxl", weight_name="ikea_instructions_xl_v1_5.safetensors", adapter_name="ikea")
+pipeline.load_lora_weights("lordjia/by-feng-zikai", weight_name="fengzikai_v1.0_XL.safetensors", adapter_name="feng")
+```
+
+## set_adapters
+
+The [`~loaders.UNet2DConditionLoadersMixin.set_adapters`] method merges LoRA adapters by concatenating their weighted matrices. Use the adapter name to specify which LoRAs to merge, and the `adapter_weights` parameter to control the scaling for each LoRA. For example, if `adapter_weights=[0.5, 0.5]`, then the merged LoRA output is an average of both LoRAs. Try adjusting the adapter weights to see how it affects the generated image!
+
+```py
+pipeline.set_adapters(["ikea", "feng"], adapter_weights=[0.7, 0.8])
+
+generator = torch.manual_seed(0)
+prompt = "A bowl of ramen shaped like a cute kawaii bear, by Feng Zikai"
+image = pipeline(prompt, generator=generator, cross_attention_kwargs={"scale": 1.0}).images[0]
+image
+```
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/lora_merge_set_adapters.png"/>
+</div>
+
+## add_weighted_adapter
+
+> [!WARNING]
+> This is an experimental method that adds PEFTs [`~peft.LoraModel.add_weighted_adapter`] method to Diffusers to enable more efficient merging methods. Check out this [issue](https://github.com/huggingface/diffusers/issues/6892) if you're interested in learning more about the motivation and design behind this integration.
+
+The [`~peft.LoraModel.add_weighted_adapter`] method provides access to more efficient merging method such as [TIES and DARE](https://huggingface.co/docs/peft/developer_guides/model_merging). To use these merging methods, make sure you have the latest stable version of Diffusers and PEFT installed.
+
+```bash
+pip install -U diffusers peft
+```
+
+There are three steps to merge LoRAs with the [`~peft.LoraModel.add_weighted_adapter`] method:
+
+1. Create a [`~peft.PeftModel`] from the underlying model and LoRA checkpoint.
+2. Load a base UNet model and the LoRA adapters.
+3. Merge the adapters using the [`~peft.LoraModel.add_weighted_adapter`] method and the merging method of your choice.
+
+Let's dive deeper into what these steps entail.
+
+1. Load a UNet that corresponds to the UNet in the LoRA checkpoint. In this case, both LoRAs use the SDXL UNet as their base model.
+
+```python
+from diffusers import UNet2DConditionModel
+import torch
+
+unet = UNet2DConditionModel.from_pretrained(
+    "stabilityai/stable-diffusion-xl-base-1.0",
+    torch_dtype=torch.float16, 
+    use_safetensors=True,
+    variant="fp16",
+    subfolder="unet",
+).to("cuda")
+```
+
+Load the SDXL pipeline and the LoRA checkpoints, starting with the [ostris/ikea-instructions-lora-sdxl](https://huggingface.co/ostris/ikea-instructions-lora-sdxl) LoRA.
+
+```python
+from diffusers import DiffusionPipeline
+
+pipeline = DiffusionPipeline.from_pretrained(
+    "stabilityai/stable-diffusion-xl-base-1.0",
+    variant="fp16",
+    torch_dtype=torch.float16,
+    unet=unet
+).to("cuda")
+pipeline.load_lora_weights("ostris/ikea-instructions-lora-sdxl", weight_name="ikea_instructions_xl_v1_5.safetensors", adapter_name="ikea")
+```
+
+Now you'll create a [`~peft.PeftModel`] from the loaded LoRA checkpoint by combining the SDXL UNet and the LoRA UNet from the pipeline.
+
+```python
+from peft import get_peft_model, LoraConfig
+import copy
+
+sdxl_unet = copy.deepcopy(unet)
+ikea_peft_model = get_peft_model(
+    sdxl_unet,
+    pipeline.unet.peft_config["ikea"],
+    adapter_name="ikea"
+)
+
+original_state_dict = {f"base_model.model.{k}": v for k, v in pipeline.unet.state_dict().items()}
+ikea_peft_model.load_state_dict(original_state_dict, strict=True)
+```
+
+> [!TIP]
+> You can optionally push the ikea_peft_model to the Hub by calling `ikea_peft_model.push_to_hub("ikea_peft_model", token=TOKEN)`.
+
+Repeat this process to create a [`~peft.PeftModel`] from the [lordjia/by-feng-zikai](https://huggingface.co/lordjia/by-feng-zikai) LoRA.
+
+```python
+pipeline.delete_adapters("ikea")
+sdxl_unet.delete_adapters("ikea")
+
+pipeline.load_lora_weights("lordjia/by-feng-zikai", weight_name="fengzikai_v1.0_XL.safetensors", adapter_name="feng")
+pipeline.set_adapters(adapter_names="feng")
+
+feng_peft_model = get_peft_model(
+    sdxl_unet,
+    pipeline.unet.peft_config["feng"],
+    adapter_name="feng"
+)
+
+original_state_dict = {f"base_model.model.{k}": v for k, v in pipe.unet.state_dict().items()}
+feng_peft_model.load_state_dict(original_state_dict, strict=True)
+```
+
+2. Load a base UNet model and then load the adapters onto it.
+
+```python
+from peft import PeftModel
+
+base_unet = UNet2DConditionModel.from_pretrained(
+    "stabilityai/stable-diffusion-xl-base-1.0",
+    torch_dtype=torch.float16, 
+    use_safetensors=True,
+    variant="fp16",
+    subfolder="unet",
+).to("cuda")
+
+model = PeftModel.from_pretrained(base_unet, "stevhliu/ikea_peft_model", use_safetensors=True, subfolder="ikea", adapter_name="ikea")
+model.load_adapter("stevhliu/feng_peft_model", use_safetensors=True, subfolder="feng", adapter_name="feng")
+```
+
+3. Merge the adapters using the [`~peft.LoraModel.add_weighted_adapter`] method and the merging method of your choice (learn more about other merging methods in this [blog post](https://huggingface.co/blog/peft_merging)). For this example, let's use the `"dare_linear"` method to merge the LoRAs.
+
+> [!WARNING]
+> Keep in mind the LoRAs need to have the same rank to be merged!
+
+```python
+model.add_weighted_adapter(
+    adapters=["ikea", "feng"],
+    weights=[1.0, 1.0],
+    combination_type="dare_linear",
+    adapter_name="ikea-feng"
+)
+model.set_adapters("ikea-feng")
+```
+
+Now you can generate an image with the merged LoRA.
+
+```python
+model = model.to(dtype=torch.float16, device="cuda")
+
+pipeline = DiffusionPipeline.from_pretrained(
+    "stabilityai/stable-diffusion-xl-base-1.0", unet=model, variant="fp16", torch_dtype=torch.float16,
+).to("cuda")
+
+image = pipeline("A bowl of ramen shaped like a cute kawaii bear, by Feng Zikai", generator=torch.manual_seed(0)).images[0]
+image
+```
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ikea-feng-dare-linear.png"/>
+</div>
+
+## fuse_lora
+
+Both the [`~loaders.UNet2DConditionLoadersMixin.set_adapters`] and [`~peft.LoraModel.add_weighted_adapter`] methods require loading the base model and the LoRA adapters separately which incurs some overhead. The [`~loaders.LoraLoaderMixin.fuse_lora`] method allows you to fuse the LoRA weights directly with the original weights of the underlying model. This way, you're only loading the model once which can increase inference and lower memory-usage.
+
+You can use PEFT to easily fuse/unfuse multiple adapters directly into the model weights (both UNet and text encoder) using the [`~loaders.LoraLoaderMixin.fuse_lora`] method, which can lead to a speed-up in inference and lower VRAM usage.
+
+For example, if you have a base model and adapters loaded and set as active with the following adapter weights:
+
+```py
+from diffusers import DiffusionPipeline
+import torch
+
+pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16).to("cuda")
+pipeline.load_lora_weights("ostris/ikea-instructions-lora-sdxl", weight_name="ikea_instructions_xl_v1_5.safetensors", adapter_name="ikea")
+pipeline.load_lora_weights("lordjia/by-feng-zikai", weight_name="fengzikai_v1.0_XL.safetensors", adapter_name="feng")
+
+pipeline.set_adapters(["ikea", "feng"], adapter_weights=[0.7, 0.8])
+```
+
+Fuse these LoRAs into the UNet with the [`~loaders.LoraLoaderMixin.fuse_lora`] method. The `lora_scale` parameter controls how much to scale the output by with the LoRA weights. It is important to make the `lora_scale` adjustments in the [`~loaders.LoraLoaderMixin.fuse_lora`] method because it won’t work if you try to pass `scale` to the `cross_attention_kwargs` in the pipeline.
+
+```py
+pipeline.fuse_lora(adapter_names=["ikea", "feng"], lora_scale=1.0)
+```
+
+Then you should use [`~loaders.LoraLoaderMixin.unload_lora_weights`] to unload the LoRA weights since they've already been fused with the underlying base model. Finally, call [`~DiffusionPipeline.save_pretrained`] to save the fused pipeline locally or you could call [`~DiffusionPipeline.push_to_hub`] to push the fused pipeline to the Hub.
+
+```py
+pipeline.unload_lora_weights()
+# save locally
+pipeline.save_pretrained("path/to/fused-pipeline")
+# save to the Hub
+pipeline.push_to_hub("fused-ikea-feng")
+```
+
+Now you can quickly load the fused pipeline and use it for inference without needing to separately load the LoRA adapters.
+
+```py
+pipeline = DiffusionPipeline.from_pretrained(
+    "username/fused-ikea-feng", torch_dtype=torch.float16,
+).to("cuda")
+
+image = pipeline("A bowl of ramen shaped like a cute kawaii bear, by Feng Zikai", generator=torch.manual_seed(0)).images[0]
+image
+```
+
+You can call [`~loaders.LoraLoaderMixin.unfuse_lora`] to restore the original model's weights (for example, if you want to use a different `lora_scale` value). However, this only works if you've only fused one LoRA adapter to the original model. If you've fused multiple LoRAs, you'll need to reload the model.
+
+```py
+pipeline.unfuse_lora()
+```
+
+### torch.compile
+
+[torch.compile](../optimization/torch2.0#torchcompile) can speed up your pipeline even more, but the LoRA weights must be fused first and then unloaded. Typically, the UNet is compiled because it is such a computationally intensive component of the pipeline.
+
+```py
+from diffusers import DiffusionPipeline
+import torch
+
+# load base model and LoRAs
+pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16).to("cuda")
+pipeline.load_lora_weights("ostris/ikea-instructions-lora-sdxl", weight_name="ikea_instructions_xl_v1_5.safetensors", adapter_name="ikea")
+pipeline.load_lora_weights("lordjia/by-feng-zikai", weight_name="fengzikai_v1.0_XL.safetensors", adapter_name="feng")
+
+# activate both LoRAs and set adapter weights
+pipeline.set_adapters(["ikea", "feng"], adapter_weights=[0.7, 0.8])
+
+# fuse LoRAs and unload weights
+pipeline.fuse_lora(adapter_names=["ikea", "feng"], lora_scale=1.0)
+pipeline.unload_lora_weights()
+
+# torch.compile
+pipeline.unet.to(memory_format=torch.channels_last)
+pipeline.unet = torch.compile(pipeline.unet, mode="reduce-overhead", fullgraph=True)
+
+image = pipeline("A bowl of ramen shaped like a cute kawaii bear, by Feng Zikai", generator=torch.manual_seed(0)).images[0]
+```
+
+Learn more about torch.compile in the [Accelerate inference of text-to-image diffusion models](../tutorials/fast_diffusion#torchcompile) guide.
+
+## Next steps
+
+For more conceptual details about how each merging method works, take a look at the [🤗 PEFT welcomes new merging methods](https://huggingface.co/blog/peft_merging#concatenation-cat) blog post!

docs/source/en/using-diffusers/write_own_pipeline.md

@@ -273,7 +273,6 @@ Lastly, convert the image to a `PIL.Image` to see your generated image!
 ```py
 >>> image = (image / 2 + 0.5).clamp(0, 1).squeeze()
 >>> image = (image.permute(1, 2, 0) * 255).to(torch.uint8).cpu().numpy()
->>> image = (image * 255).round().astype("uint8")
 >>> image = Image.fromarray(image)
 >>> image
 ```

examples/advanced_diffusion_training/README.md

@@ -259,6 +259,50 @@ pip install git+https://github.com/huggingface/peft.git
 **Inference** 
 The inference is the same as if you train a regular LoRA 🤗
 
+## Conducting EDM-style training
+
+It's now possible to perform EDM-style training as proposed in [Elucidating the Design Space of Diffusion-Based Generative Models](https://arxiv.org/abs/2206.00364). 
+
+simply set:
+
+```diff
++  --do_edm_style_training \
+```
+
+Other SDXL-like models that use the EDM formulation, such as [playgroundai/playground-v2.5-1024px-aesthetic](https://huggingface.co/playgroundai/playground-v2.5-1024px-aesthetic), can also be DreamBooth'd with the script. Below is an example command:
+
+```bash
+accelerate launch train_dreambooth_lora_sdxl_advanced.py \
+  --pretrained_model_name_or_path="playgroundai/playground-v2.5-1024px-aesthetic"  \
+  --dataset_name="linoyts/3d_icon" \
+  --instance_prompt="3d icon in the style of TOK" \
+  --validation_prompt="a TOK icon of an astronaut riding a horse, in the style of TOK" \
+  --output_dir="3d-icon-SDXL-LoRA" \
+  --do_edm_style_training \
+  --caption_column="prompt" \
+  --mixed_precision="bf16" \
+  --resolution=1024 \
+  --train_batch_size=3 \
+  --repeats=1 \
+  --report_to="wandb"\
+  --gradient_accumulation_steps=1 \
+  --gradient_checkpointing \
+  --learning_rate=1.0 \
+  --text_encoder_lr=1.0 \
+  --optimizer="prodigy"\
+  --train_text_encoder_ti\
+  --train_text_encoder_ti_frac=0.5\
+  --lr_scheduler="constant" \
+  --lr_warmup_steps=0 \
+  --rank=8 \
+  --max_train_steps=1000 \
+  --checkpointing_steps=2000 \
+  --seed="0" \
+  --push_to_hub
+```
+
+> [!CAUTION]
+> Min-SNR gamma is not supported with the EDM-style training yet. When training with the PlaygroundAI model, it's recommended to not pass any "variant".
 
 ### Tips and Tricks
 Check out [these recommended practices](https://huggingface.co/blog/sdxl_lora_advanced_script#additional-good-practices)

examples/advanced_diffusion_training/train_dreambooth_lora_sd15_advanced.py

@@ -70,13 +70,14 @@ from diffusers.utils.import_utils import is_xformers_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
 
 def save_model_card(
     repo_id: str,
+    use_dora: bool,
     images=None,
     base_model=str,
     train_text_encoder=False,
@@ -88,6 +89,7 @@ def save_model_card(
     vae_path=None,
 ):
     img_str = "widget:\n"
+    lora = "lora" if not use_dora else "dora"
     for i, image in enumerate(images):
         image.save(os.path.join(repo_folder, f"image_{i}.png"))
         img_str += f"""
@@ -139,9 +141,10 @@ to trigger concept `{key}` → use `{tokens}` in your prompt \n
 tags:
 - stable-diffusion
 - stable-diffusion-diffusers
+- diffusers-training
 - text-to-image
 - diffusers
-- lora
+- {lora}
 - template:sd-lora
 {img_str}
 base_model: {base_model}
@@ -1212,7 +1215,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, "
                     "please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
@@ -1363,14 +1366,14 @@ def main(args):
 
     # Optimizer creation
     if not (args.optimizer.lower() == "prodigy" or args.optimizer.lower() == "adamw"):
-        logger.warn(
+        logger.warning(
             f"Unsupported choice of optimizer: {args.optimizer}.Supported optimizers include [adamW, prodigy]."
             "Defaulting to adamW"
         )
         args.optimizer = "adamw"
 
     if args.use_8bit_adam and not args.optimizer.lower() == "adamw":
-        logger.warn(
+        logger.warning(
             f"use_8bit_adam is ignored when optimizer is not set to 'AdamW'. Optimizer was "
             f"set to {args.optimizer.lower()}"
         )
@@ -1404,11 +1407,11 @@ def main(args):
         optimizer_class = prodigyopt.Prodigy
 
         if args.learning_rate <= 0.1:
-            logger.warn(
+            logger.warning(
                 "Learning rate is too low. When using prodigy, it's generally better to set learning rate around 1.0"
             )
         if args.train_text_encoder and args.text_encoder_lr:
-            logger.warn(
+            logger.warning(
                 f"Learning rates were provided both for the unet and the text encoder- e.g. text_encoder_lr:"
                 f" {args.text_encoder_lr} and learning_rate: {args.learning_rate}. "
                 f"When using prodigy only learning_rate is used as the initial learning rate."
@@ -1967,6 +1970,7 @@ def main(args):
 
         save_model_card(
             model_id if not args.push_to_hub else repo_id,
+            use_dora=args.use_dora,
             images=images,
             base_model=args.pretrained_model_name_or_path,
             train_text_encoder=args.train_text_encoder,

examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py

@@ -14,9 +14,11 @@
 # See the License for the specific language governing permissions and
 
 import argparse
+import contextlib
 import gc
 import hashlib
 import itertools
+import json
 import logging
 import math
 import os
@@ -37,7 +39,7 @@ import transformers
 from accelerate import Accelerator
 from accelerate.logging import get_logger
 from accelerate.utils import DistributedDataParallelKwargs, ProjectConfiguration, set_seed
-from huggingface_hub import create_repo, upload_folder
+from huggingface_hub import create_repo, hf_hub_download, upload_folder
 from packaging import version
 from peft import LoraConfig, set_peft_model_state_dict
 from peft.utils import get_peft_model_state_dict
@@ -55,6 +57,8 @@ from diffusers import (
     AutoencoderKL,
     DDPMScheduler,
     DPMSolverMultistepScheduler,
+    EDMEulerScheduler,
+    EulerDiscreteScheduler,
     StableDiffusionXLPipeline,
     UNet2DConditionModel,
 )
@@ -74,13 +78,28 @@ from diffusers.utils.torch_utils import is_compiled_module
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
 
+def determine_scheduler_type(pretrained_model_name_or_path, revision):
+    model_index_filename = "model_index.json"
+    if os.path.isdir(pretrained_model_name_or_path):
+        model_index = os.path.join(pretrained_model_name_or_path, model_index_filename)
+    else:
+        model_index = hf_hub_download(
+            repo_id=pretrained_model_name_or_path, filename=model_index_filename, revision=revision
+        )
+
+    with open(model_index, "r") as f:
+        scheduler_type = json.load(f)["scheduler"][1]
+    return scheduler_type
+
+
 def save_model_card(
     repo_id: str,
+    use_dora: bool,
     images=None,
     base_model=str,
     train_text_encoder=False,
@@ -92,6 +111,7 @@ def save_model_card(
     vae_path=None,
 ):
     img_str = "widget:\n"
+    lora = "lora" if not use_dora else "dora"
     for i, image in enumerate(images):
         image.save(os.path.join(repo_folder, f"image_{i}.png"))
         img_str += f"""
@@ -144,9 +164,10 @@ to trigger concept `{key}` → use `{tokens}` in your prompt \n
 tags:
 - stable-diffusion-xl
 - stable-diffusion-xl-diffusers
+- diffusers-training
 - text-to-image
 - diffusers
-- lora
+- {lora}
 - template:sd-lora
 {img_str}
 base_model: {base_model}
@@ -367,6 +388,11 @@ def parse_args(input_args=None):
             " `args.validation_prompt` multiple times: `args.num_validation_images`."
         ),
     )
+    parser.add_argument(
+        "--do_edm_style_training",
+        action="store_true",
+        help="Flag to conduct training using the EDM formulation as introduced in https://arxiv.org/abs/2206.00364.",
+    )
     parser.add_argument(
         "--with_prior_preservation",
         default=False,
@@ -663,7 +689,6 @@ def parse_args(input_args=None):
     )
     parser.add_argument(
         "--use_dora",
-        type=bool,
         action="store_true",
         default=False,
         help=(
@@ -1115,6 +1140,8 @@ def main(args):
             "You cannot use both --report_to=wandb and --hub_token due to a security risk of exposing your token."
             " Please use `huggingface-cli login` to authenticate with the Hub."
         )
+    if args.do_edm_style_training and args.snr_gamma is not None:
+        raise ValueError("Min-SNR formulation is not supported when conducting EDM-style training.")
 
     logging_dir = Path(args.output_dir, args.logging_dir)
 
@@ -1232,7 +1259,19 @@ def main(args):
     )
 
     # Load scheduler and models
-    noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+    scheduler_type = determine_scheduler_type(args.pretrained_model_name_or_path, args.revision)
+    if "EDM" in scheduler_type:
+        args.do_edm_style_training = True
+        noise_scheduler = EDMEulerScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+        logger.info("Performing EDM-style training!")
+    elif args.do_edm_style_training:
+        noise_scheduler = EulerDiscreteScheduler.from_pretrained(
+            args.pretrained_model_name_or_path, subfolder="scheduler"
+        )
+        logger.info("Performing EDM-style training!")
+    else:
+        noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+
     text_encoder_one = text_encoder_cls_one.from_pretrained(
         args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant
     )
@@ -1250,7 +1289,12 @@ def main(args):
         revision=args.revision,
         variant=args.variant,
     )
-    vae_scaling_factor = vae.config.scaling_factor
+    latents_mean = latents_std = None
+    if hasattr(vae.config, "latents_mean") and vae.config.latents_mean is not None:
+        latents_mean = torch.tensor(vae.config.latents_mean).view(1, 4, 1, 1)
+    if hasattr(vae.config, "latents_std") and vae.config.latents_std is not None:
+        latents_std = torch.tensor(vae.config.latents_std).view(1, 4, 1, 1)
+
     unet = UNet2DConditionModel.from_pretrained(
         args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision, variant=args.variant
     )
@@ -1315,7 +1359,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, "
                     "please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
@@ -1520,14 +1564,14 @@ def main(args):
 
     # Optimizer creation
     if not (args.optimizer.lower() == "prodigy" or args.optimizer.lower() == "adamw"):
-        logger.warn(
+        logger.warning(
             f"Unsupported choice of optimizer: {args.optimizer}.Supported optimizers include [adamW, prodigy]."
             "Defaulting to adamW"
         )
         args.optimizer = "adamw"
 
     if args.use_8bit_adam and not args.optimizer.lower() == "adamw":
-        logger.warn(
+        logger.warning(
             f"use_8bit_adam is ignored when optimizer is not set to 'AdamW'. Optimizer was "
             f"set to {args.optimizer.lower()}"
         )
@@ -1561,11 +1605,11 @@ def main(args):
         optimizer_class = prodigyopt.Prodigy
 
         if args.learning_rate <= 0.1:
-            logger.warn(
+            logger.warning(
                 "Learning rate is too low. When using prodigy, it's generally better to set learning rate around 1.0"
             )
         if args.train_text_encoder and args.text_encoder_lr:
-            logger.warn(
+            logger.warning(
                 f"Learning rates were provided both for the unet and the text encoder- e.g. text_encoder_lr:"
                 f" {args.text_encoder_lr} and learning_rate: {args.learning_rate}. "
                 f"When using prodigy only learning_rate is used as the initial learning rate."
@@ -1788,6 +1832,19 @@ def main(args):
         disable=not accelerator.is_local_main_process,
     )
 
+    def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
+        # TODO: revisit other sampling algorithms
+        sigmas = noise_scheduler.sigmas.to(device=accelerator.device, dtype=dtype)
+        schedule_timesteps = noise_scheduler.timesteps.to(accelerator.device)
+        timesteps = timesteps.to(accelerator.device)
+
+        step_indices = [(schedule_timesteps == t).nonzero().item() for t in timesteps]
+
+        sigma = sigmas[step_indices].flatten()
+        while len(sigma.shape) < n_dim:
+            sigma = sigma.unsqueeze(-1)
+        return sigma
+
     if args.train_text_encoder:
         num_train_epochs_text_encoder = int(args.train_text_encoder_frac * args.num_train_epochs)
     elif args.train_text_encoder_ti:  # args.train_text_encoder_ti
@@ -1839,9 +1896,15 @@ def main(args):
                     pixel_values = batch["pixel_values"].to(dtype=vae.dtype)
                     model_input = vae.encode(pixel_values).latent_dist.sample()
 
-                model_input = model_input * vae_scaling_factor
-                if args.pretrained_vae_model_name_or_path is None:
-                    model_input = model_input.to(weight_dtype)
+                if latents_mean is None and latents_std is None:
+                    model_input = model_input * vae.config.scaling_factor
+                    if args.pretrained_vae_model_name_or_path is None:
+                        model_input = model_input.to(weight_dtype)
+                else:
+                    latents_mean = latents_mean.to(device=model_input.device, dtype=model_input.dtype)
+                    latents_std = latents_std.to(device=model_input.device, dtype=model_input.dtype)
+                    model_input = (model_input - latents_mean) * vae.config.scaling_factor / latents_std
+                    model_input = model_input.to(dtype=weight_dtype)
 
                 # Sample noise that we'll add to the latents
                 noise = torch.randn_like(model_input)
@@ -1852,15 +1915,32 @@ def main(args):
                     )
 
                 bsz = model_input.shape[0]
+
                 # Sample a random timestep for each image
-                timesteps = torch.randint(
-                    0, noise_scheduler.config.num_train_timesteps, (bsz,), device=model_input.device
-                )
-                timesteps = timesteps.long()
+                if not args.do_edm_style_training:
+                    timesteps = torch.randint(
+                        0, noise_scheduler.config.num_train_timesteps, (bsz,), device=model_input.device
+                    )
+                    timesteps = timesteps.long()
+                else:
+                    # in EDM formulation, the model is conditioned on the pre-conditioned noise levels
+                    # instead of discrete timesteps, so here we sample indices to get the noise levels
+                    # from `scheduler.timesteps`
+                    indices = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,))
+                    timesteps = noise_scheduler.timesteps[indices].to(device=model_input.device)
 
                 # Add noise to the model input according to the noise magnitude at each timestep
                 # (this is the forward diffusion process)
                 noisy_model_input = noise_scheduler.add_noise(model_input, noise, timesteps)
+                # For EDM-style training, we first obtain the sigmas based on the continuous timesteps.
+                # We then precondition the final model inputs based on these sigmas instead of the timesteps.
+                # Follow: Section 5 of https://arxiv.org/abs/2206.00364.
+                if args.do_edm_style_training:
+                    sigmas = get_sigmas(timesteps, len(noisy_model_input.shape), noisy_model_input.dtype)
+                    if "EDM" in scheduler_type:
+                        inp_noisy_latents = noise_scheduler.precondition_inputs(noisy_model_input, sigmas)
+                    else:
+                        inp_noisy_latents = noisy_model_input / ((sigmas**2 + 1) ** 0.5)
 
                 # time ids
                 add_time_ids = torch.cat(
@@ -1886,7 +1966,7 @@ def main(args):
                     }
                     prompt_embeds_input = prompt_embeds.repeat(elems_to_repeat_text_embeds, 1, 1)
                     model_pred = unet(
-                        noisy_model_input,
+                        inp_noisy_latents if args.do_edm_style_training else noisy_model_input,
                         timesteps,
                         prompt_embeds_input,
                         added_cond_kwargs=unet_added_conditions,
@@ -1904,14 +1984,42 @@ def main(args):
                     )
                     prompt_embeds_input = prompt_embeds.repeat(elems_to_repeat_text_embeds, 1, 1)
                     model_pred = unet(
-                        noisy_model_input, timesteps, prompt_embeds_input, added_cond_kwargs=unet_added_conditions
+                        inp_noisy_latents if args.do_edm_style_training else noisy_model_input,
+                        timesteps,
+                        prompt_embeds_input,
+                        added_cond_kwargs=unet_added_conditions,
                     ).sample
 
+                weighting = None
+                if args.do_edm_style_training:
+                    # Similar to the input preconditioning, the model predictions are also preconditioned
+                    # on noised model inputs (before preconditioning) and the sigmas.
+                    # Follow: Section 5 of https://arxiv.org/abs/2206.00364.
+                    if "EDM" in scheduler_type:
+                        model_pred = noise_scheduler.precondition_outputs(noisy_model_input, model_pred, sigmas)
+                    else:
+                        if noise_scheduler.config.prediction_type == "epsilon":
+                            model_pred = model_pred * (-sigmas) + noisy_model_input
+                        elif noise_scheduler.config.prediction_type == "v_prediction":
+                            model_pred = model_pred * (-sigmas / (sigmas**2 + 1) ** 0.5) + (
+                                noisy_model_input / (sigmas**2 + 1)
+                            )
+                    # We are not doing weighting here because it tends result in numerical problems.
+                    # See: https://github.com/huggingface/diffusers/pull/7126#issuecomment-1968523051
+                    # There might be other alternatives for weighting as well:
+                    # https://github.com/huggingface/diffusers/pull/7126#discussion_r1505404686
+                    if "EDM" not in scheduler_type:
+                        weighting = (sigmas**-2.0).float()
+
                 # Get the target for loss depending on the prediction type
                 if noise_scheduler.config.prediction_type == "epsilon":
-                    target = noise
+                    target = model_input if args.do_edm_style_training else noise
                 elif noise_scheduler.config.prediction_type == "v_prediction":
-                    target = noise_scheduler.get_velocity(model_input, noise, timesteps)
+                    target = (
+                        model_input
+                        if args.do_edm_style_training
+                        else noise_scheduler.get_velocity(model_input, noise, timesteps)
+                    )
                 else:
                     raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
 
@@ -1921,10 +2029,28 @@ def main(args):
                     target, target_prior = torch.chunk(target, 2, dim=0)
 
                     # Compute prior loss
-                    prior_loss = F.mse_loss(model_pred_prior.float(), target_prior.float(), reduction="mean")
+                    if weighting is not None:
+                        prior_loss = torch.mean(
+                            (weighting.float() * (model_pred_prior.float() - target_prior.float()) ** 2).reshape(
+                                target_prior.shape[0], -1
+                            ),
+                            1,
+                        )
+                        prior_loss = prior_loss.mean()
+                    else:
+                        prior_loss = F.mse_loss(model_pred_prior.float(), target_prior.float(), reduction="mean")
 
                 if args.snr_gamma is None:
-                    loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
+                    if weighting is not None:
+                        loss = torch.mean(
+                            (weighting.float() * (model_pred.float() - target.float()) ** 2).reshape(
+                                target.shape[0], -1
+                            ),
+                            1,
+                        )
+                        loss = loss.mean()
+                    else:
+                        loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
                 else:
                     # Compute loss-weights as per Section 3.4 of https://arxiv.org/abs/2303.09556.
                     # Since we predict the noise instead of x_0, the original formulation is slightly changed.
@@ -2047,17 +2173,18 @@ def main(args):
                 # We train on the simplified learning objective. If we were previously predicting a variance, we need the scheduler to ignore it
                 scheduler_args = {}
 
-                if "variance_type" in pipeline.scheduler.config:
-                    variance_type = pipeline.scheduler.config.variance_type
+                if not args.do_edm_style_training:
+                    if "variance_type" in pipeline.scheduler.config:
+                        variance_type = pipeline.scheduler.config.variance_type
 
-                    if variance_type in ["learned", "learned_range"]:
-                        variance_type = "fixed_small"
+                        if variance_type in ["learned", "learned_range"]:
+                            variance_type = "fixed_small"
 
-                    scheduler_args["variance_type"] = variance_type
+                        scheduler_args["variance_type"] = variance_type
 
-                pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
-                    pipeline.scheduler.config, **scheduler_args
-                )
+                    pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
+                        pipeline.scheduler.config, **scheduler_args
+                    )
 
                 pipeline = pipeline.to(accelerator.device)
                 pipeline.set_progress_bar_config(disable=True)
@@ -2065,8 +2192,13 @@ def main(args):
                 # run inference
                 generator = torch.Generator(device=accelerator.device).manual_seed(args.seed) if args.seed else None
                 pipeline_args = {"prompt": args.validation_prompt}
+                inference_ctx = (
+                    contextlib.nullcontext()
+                    if "playground" in args.pretrained_model_name_or_path
+                    else torch.cuda.amp.autocast()
+                )
 
-                with torch.cuda.amp.autocast():
+                with inference_ctx:
                     images = [
                         pipeline(**pipeline_args, generator=generator).images[0]
                         for _ in range(args.num_validation_images)
@@ -2142,15 +2274,18 @@ def main(args):
             # We train on the simplified learning objective. If we were previously predicting a variance, we need the scheduler to ignore it
             scheduler_args = {}
 
-            if "variance_type" in pipeline.scheduler.config:
-                variance_type = pipeline.scheduler.config.variance_type
+            if not args.do_edm_style_training:
+                if "variance_type" in pipeline.scheduler.config:
+                    variance_type = pipeline.scheduler.config.variance_type
 
-                if variance_type in ["learned", "learned_range"]:
-                    variance_type = "fixed_small"
+                    if variance_type in ["learned", "learned_range"]:
+                        variance_type = "fixed_small"
 
-                scheduler_args["variance_type"] = variance_type
+                    scheduler_args["variance_type"] = variance_type
 
-            pipeline.scheduler = DPMSolverMultistepScheduler.from_config(pipeline.scheduler.config, **scheduler_args)
+                pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
+                    pipeline.scheduler.config, **scheduler_args
+                )
 
             # load attention processors
             pipeline.load_lora_weights(args.output_dir)
@@ -2204,6 +2339,7 @@ def main(args):
 
         save_model_card(
             model_id if not args.push_to_hub else repo_id,
+            use_dora=args.use_dora,
             images=images,
             base_model=args.pretrained_model_name_or_path,
             train_text_encoder=args.train_text_encoder,

examples/community/README.md

@@ -105,7 +105,7 @@ pipeline_output = pipe(
     # processing_res=768,     # (optional) Maximum resolution of processing. If set to 0: will not resize at all. Defaults to 768.
     # match_input_res=True,   # (optional) Resize depth prediction to match input resolution.
     # batch_size=0,           # (optional) Inference batch size, no bigger than `num_ensemble`. If set to 0, the script will automatically decide the proper batch size. Defaults to 0.
-    # color_map="Spectral",   # (optional) Colormap used to colorize the depth map. Defaults to "Spectral".
+    # color_map="Spectral",   # (optional) Colormap used to colorize the depth map. Defaults to "Spectral". Set to `None` to skip colormap generation.
     # show_progress_bar=True, # (optional) If true, will show progress bars of the inference progress.
 )
 
@@ -750,7 +750,7 @@ This example produces the following images:
 ![image](https://user-images.githubusercontent.com/4313860/198328706-295824a4-9856-4ce5-8e66-278ceb42fd29.png)
 
 ### GlueGen Stable Diffusion Pipeline
-GlueGen is a minimal adapter that allow alignment between any encoder (Text Encoder of different language, Multilingual Roberta, AudioClip) and CLIP text encoder used in standard Stable Diffusion model. This method allows easy language adaptation to available english Stable Diffusion checkpoints without the need of an image captioning dataset as well as long training hours. 
+GlueGen is a minimal adapter that allow alignment between any encoder (Text Encoder of different language, Multilingual Roberta, AudioClip) and CLIP text encoder used in standard Stable Diffusion model. This method allows easy language adaptation to available english Stable Diffusion checkpoints without the need of an image captioning dataset as well as long training hours.
 
 Make sure you downloaded `gluenet_French_clip_overnorm_over3_noln.ckpt` for French (there are also pre-trained weights for Chinese, Italian, Japanese, Spanish or train your own) at [GlueGen's official repo](https://github.com/salesforce/GlueGen/tree/main)
 
@@ -782,9 +782,9 @@ if __name__ == "__main__":
     ).to(device)
     pipeline.load_language_adapter("gluenet_French_clip_overnorm_over3_noln.ckpt", num_token=token_max_length, dim=1024, dim_out=768, tensor_norm=tensor_norm)
 
-    prompt = "une voiture sur la plage" 
+    prompt = "une voiture sur la plage"
 
-    generator = torch.Generator(device=device).manual_seed(42) 
+    generator = torch.Generator(device=device).manual_seed(42)
     image = pipeline(prompt, generator=generator).images[0]
     image.save("gluegen_output_fr.png")
 ```
@@ -1755,7 +1755,7 @@ with torch.cpu.amp.autocast(enabled=True, dtype=torch.bfloat16):
 ```
 
 The following code compares the performance of the original stable diffusion xl pipeline with the ipex-optimized pipeline.
-By using this optimized pipeline, we can get about 1.4-2 times performance boost with BFloat16 on fourth generation of Intel Xeon CPUs, 
+By using this optimized pipeline, we can get about 1.4-2 times performance boost with BFloat16 on fourth generation of Intel Xeon CPUs,
 code-named Sapphire Rapids.
 
 ```python
@@ -1826,7 +1826,7 @@ This approach is using (optional) CoCa model to avoid writing image description.
 
 This SDXL pipeline support unlimited length prompt and negative prompt, compatible with A1111 prompt weighted style.
 
-You can provide both `prompt` and `prompt_2`. If only one prompt is provided, `prompt_2` will be a copy of the provided `prompt`. Here is a sample code to use this pipeline. 
+You can provide both `prompt` and `prompt_2`. If only one prompt is provided, `prompt_2` will be a copy of the provided `prompt`. Here is a sample code to use this pipeline.
 
 ```python
 from diffusers import DiffusionPipeline
@@ -3397,7 +3397,7 @@ invert_prompt = "A lying cat"
 input_image = "siamese.jpg"
 steps = 50
 
-# Provide prompt used for generation. Same if reconstruction 
+# Provide prompt used for generation. Same if reconstruction
 prompt = "A lying cat"
 # or different if editing.
 prompt = "A lying dog"
@@ -3414,15 +3414,13 @@ pipeline(prompt, uncond, inverted_latent, guidance_scale=7.5, num_inference_step
 
 ### Rerender A Video
 
-This is the Diffusers implementation of zero-shot video-to-video translation pipeline [Rerender A Video](https://github.com/williamyang1991/Rerender_A_Video) (without Ebsynth postprocessing). To run the code, please install gmflow. Then modify the path in `examples/community/rerender_a_video.py`:
+This is the Diffusers implementation of zero-shot video-to-video translation pipeline [Rerender A Video](https://github.com/williamyang1991/Rerender_A_Video) (without Ebsynth postprocessing). To run the code, please install gmflow. Then modify the path in `gmflow_dir`. After that, you can run the pipeline with:
 
 ```py
+import sys
 gmflow_dir = "/path/to/gmflow"
-```
+sys.path.insert(0, gmflow_dir)
 
-After that, you can run the pipeline with:
-
-```py
 from diffusers import ControlNetModel, AutoencoderKL, DDIMScheduler
 from diffusers.utils import export_to_video
 import numpy as np
@@ -3493,7 +3491,7 @@ output_frames = pipe(
     mask_end=0.8,
     mask_strength=0.5,
     negative_prompt='longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality'
-).frames
+).frames[0]
 
 export_to_video(
     output_frames, "/path/to/video.mp4", 5)
@@ -3636,8 +3634,8 @@ image = torch.from_numpy(faces[0].normed_embedding).unsqueeze(0)
 images = pipeline(
     prompt="A photo of a girl wearing a black dress, holding red roses in hand, upper body, behind is the Eiffel Tower",
     image_embeds=image,
-    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality", 
-    num_inference_steps=20, num_images_per_prompt=num_images, width=512, height=704, 
+    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality",
+    num_inference_steps=20, num_images_per_prompt=num_images, width=512, height=704,
     generator=generator
 ).images
 

examples/community/latent_consistency_img2img.py

@@ -513,9 +513,7 @@ class LCMSchedulerWithTimestamp(SchedulerMixin, ConfigMixin):
             there is no previous alpha. When this option is `True` the previous alpha product is fixed to `1`,
             otherwise it uses the alpha value at step 0.
         steps_offset (`int`, defaults to 0):
-            An offset added to the inference steps. You can use a combination of `offset=1` and
-            `set_alpha_to_one=False` to make the last step use step 0 for the previous alpha product like in Stable
-            Diffusion.
+            An offset added to the inference steps, as required by some model families.
         prediction_type (`str`, defaults to `epsilon`, *optional*):
             Prediction type of the scheduler function; can be `epsilon` (predicts the noise of the diffusion process),
             `sample` (directly predicts the noisy sample`) or `v_prediction` (see section 2.4 of [Imagen

examples/community/latent_consistency_txt2img.py

@@ -418,9 +418,7 @@ class LCMScheduler(SchedulerMixin, ConfigMixin):
             there is no previous alpha. When this option is `True` the previous alpha product is fixed to `1`,
             otherwise it uses the alpha value at step 0.
         steps_offset (`int`, defaults to 0):
-            An offset added to the inference steps. You can use a combination of `offset=1` and
-            `set_alpha_to_one=False` to make the last step use step 0 for the previous alpha product like in Stable
-            Diffusion.
+            An offset added to the inference steps, as required by some model families.
         prediction_type (`str`, defaults to `epsilon`, *optional*):
             Prediction type of the scheduler function; can be `epsilon` (predicts the noise of the diffusion process),
             `sample` (directly predicts the noisy sample`) or `v_prediction` (see section 2.4 of [Imagen

examples/community/marigold_depth_estimation.py

@@ -40,7 +40,7 @@ from diffusers.utils import BaseOutput, check_min_version
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 
 class MarigoldDepthOutput(BaseOutput):
@@ -50,14 +50,14 @@ class MarigoldDepthOutput(BaseOutput):
     Args:
         depth_np (`np.ndarray`):
             Predicted depth map, with depth values in the range of [0, 1].
-        depth_colored (`PIL.Image.Image`):
+        depth_colored (`None` or `PIL.Image.Image`):
             Colorized depth map, with the shape of [3, H, W] and values in [0, 1].
         uncertainty (`None` or `np.ndarray`):
             Uncalibrated uncertainty(MAD, median absolute deviation) coming from ensembling.
     """
 
     depth_np: np.ndarray
-    depth_colored: Image.Image
+    depth_colored: Union[None, Image.Image]
     uncertainty: Union[None, np.ndarray]
 
 
@@ -139,14 +139,15 @@ class MarigoldPipeline(DiffusionPipeline):
                 If set to 0, the script will automatically decide the proper batch size.
             show_progress_bar (`bool`, *optional*, defaults to `True`):
                 Display a progress bar of diffusion denoising.
-            color_map (`str`, *optional*, defaults to `"Spectral"`):
+            color_map (`str`, *optional*, defaults to `"Spectral"`, pass `None` to skip colorized depth map generation):
                 Colormap used to colorize the depth map.
             ensemble_kwargs (`dict`, *optional*, defaults to `None`):
                 Arguments for detailed ensembling settings.
         Returns:
             `MarigoldDepthOutput`: Output class for Marigold monocular depth prediction pipeline, including:
             - **depth_np** (`np.ndarray`) Predicted depth map, with depth values in the range of [0, 1]
-            - **depth_colored** (`PIL.Image.Image`) Colorized depth map, with the shape of [3, H, W] and values in [0, 1]
+            - **depth_colored** (`None` or `PIL.Image.Image`) Colorized depth map, with the shape of [3, H, W] and
+                    values in [0, 1]. None if `color_map` is `None`
             - **uncertainty** (`None` or `np.ndarray`) Uncalibrated uncertainty(MAD, median absolute deviation)
                     coming from ensembling. None if `ensemble_size = 1`
         """
@@ -233,12 +234,15 @@ class MarigoldPipeline(DiffusionPipeline):
         depth_pred = depth_pred.clip(0, 1)
 
         # Colorize
-        depth_colored = self.colorize_depth_maps(
-            depth_pred, 0, 1, cmap=color_map
-        ).squeeze()  # [3, H, W], value in (0, 1)
-        depth_colored = (depth_colored * 255).astype(np.uint8)
-        depth_colored_hwc = self.chw2hwc(depth_colored)
-        depth_colored_img = Image.fromarray(depth_colored_hwc)
+        if color_map is not None:
+            depth_colored = self.colorize_depth_maps(
+                depth_pred, 0, 1, cmap=color_map
+            ).squeeze()  # [3, H, W], value in (0, 1)
+            depth_colored = (depth_colored * 255).astype(np.uint8)
+            depth_colored_hwc = self.chw2hwc(depth_colored)
+            depth_colored_img = Image.fromarray(depth_colored_hwc)
+        else:
+            depth_colored_img = None
         return MarigoldDepthOutput(
             depth_np=depth_pred,
             depth_colored=depth_colored_img,

examples/community/pipeline_animatediff_controlnet.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import inspect
-from dataclasses import dataclass
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import numpy as np
@@ -27,6 +26,7 @@ from diffusers.loaders import IPAdapterMixin, LoraLoaderMixin, TextualInversionL
 from diffusers.models import AutoencoderKL, ControlNetModel, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.models.unets.unet_motion_model import MotionAdapter
+from diffusers.pipelines.animatediff.pipeline_output import AnimateDiffPipelineOutput
 from diffusers.pipelines.controlnet.multicontrolnet import MultiControlNetModel
 from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.schedulers import (
@@ -37,7 +37,7 @@ from diffusers.schedulers import (
     LMSDiscreteScheduler,
     PNDMScheduler,
 )
-from diffusers.utils import USE_PEFT_BACKEND, BaseOutput, deprecate, logging, scale_lora_layers, unscale_lora_layers
+from diffusers.utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
 from diffusers.utils.torch_utils import is_compiled_module, randn_tensor
 
 
@@ -91,10 +91,8 @@ EXAMPLE_DOC_STRING = """
 """
 
 
+# Copied from diffusers.pipelines.animatediff.pipeline_animatediff.tensor2vid
 def tensor2vid(video: torch.Tensor, processor, output_type="np"):
-    # Based on:
-    # https://github.com/modelscope/modelscope/blob/1509fdb973e5871f37148a4b5e5964cafd43e64d/modelscope/pipelines/multi_modal/text_to_video_synthesis_pipeline.py#L78
-
     batch_size, channels, num_frames, height, width = video.shape
     outputs = []
     for batch_idx in range(batch_size):
@@ -103,14 +101,18 @@ def tensor2vid(video: torch.Tensor, processor, output_type="np"):
 
         outputs.append(batch_output)
 
+    if output_type == "np":
+        outputs = np.stack(outputs)
+
+    elif output_type == "pt":
+        outputs = torch.stack(outputs)
+
+    elif not output_type == "pil":
+        raise ValueError(f"{output_type} does not exist. Please choose one of ['np', 'pt', 'pil']")
+
     return outputs
 
 
-@dataclass
-class AnimateDiffControlNetPipelineOutput(BaseOutput):
-    frames: Union[torch.Tensor, np.ndarray]
-
-
 class AnimateDiffControlNetPipeline(
     DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, IPAdapterMixin, LoraLoaderMixin
 ):
@@ -843,8 +845,8 @@ class AnimateDiffControlNetPipeline(
         Examples:
 
         Returns:
-            [`~pipelines.text_to_video_synthesis.TextToVideoSDPipelineOutput`] or `tuple`:
-                If `return_dict` is `True`, [`~pipelines.text_to_video_synthesis.TextToVideoSDPipelineOutput`] is
+            [`~pipelines.animatediff.pipeline_output.AnimateDiffPipelineOutput`] or `tuple`:
+                If `return_dict` is `True`, [`~pipelines.animatediff.pipeline_output.AnimateDiffPipelineOutput`] is
                 returned, otherwise a `tuple` is returned where the first element is a list with the generated frames.
         """
 
@@ -1020,7 +1022,7 @@ class AnimateDiffControlNetPipeline(
             ]
             controlnet_keep.append(keeps[0] if isinstance(controlnet, ControlNetModel) else keeps)
 
-        # Denoising loop
+        # 8. Denoising loop
         num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
         with self.progress_bar(total=num_inference_steps) as progress_bar:
             for i, t in enumerate(timesteps):
@@ -1096,21 +1098,17 @@ class AnimateDiffControlNetPipeline(
                     if callback is not None and i % callback_steps == 0:
                         callback(i, t, latents)
 
+        # 9. Post processing
         if output_type == "latent":
-            return AnimateDiffControlNetPipelineOutput(frames=latents)
-
-        # Post-processing
-        video_tensor = self.decode_latents(latents)
-
-        if output_type == "pt":
-            video = video_tensor
+            video = latents
         else:
+            video_tensor = self.decode_latents(latents)
             video = tensor2vid(video_tensor, self.image_processor, output_type=output_type)
 
-        # Offload all models
+        # 10. Offload all models
         self.maybe_free_model_hooks()
 
         if not return_dict:
             return (video,)
 
-        return AnimateDiffControlNetPipelineOutput(frames=video)
+        return AnimateDiffPipelineOutput(frames=video)

examples/community/pipeline_animatediff_img2video.py

@@ -158,10 +158,8 @@ def slerp(
     return v2
 
 
+# Copied from diffusers.pipelines.animatediff.pipeline_animatediff.tensor2vid
 def tensor2vid(video: torch.Tensor, processor, output_type="np"):
-    # Based on:
-    # https://github.com/modelscope/modelscope/blob/1509fdb973e5871f37148a4b5e5964cafd43e64d/modelscope/pipelines/multi_modal/text_to_video_synthesis_pipeline.py#L78
-
     batch_size, channels, num_frames, height, width = video.shape
     outputs = []
     for batch_idx in range(batch_size):
@@ -170,6 +168,15 @@ def tensor2vid(video: torch.Tensor, processor, output_type="np"):
 
         outputs.append(batch_output)
 
+    if output_type == "np":
+        outputs = np.stack(outputs)
+
+    elif output_type == "pt":
+        outputs = torch.stack(outputs)
+
+    elif not output_type == "pil":
+        raise ValueError(f"{output_type} does not exist. Please choose one of ['np', 'pt', 'pil']")
+
     return outputs
 
 
@@ -826,8 +833,8 @@ class AnimateDiffImgToVideoPipeline(
         Examples:
 
         Returns:
-            [`AnimateDiffPipelineOutput`] or `tuple`:
-                If `return_dict` is `True`, [`AnimateDiffPipelineOutput`] is
+            [`~pipelines.animatediff.pipeline_output.AnimateDiffPipelineOutput`] or `tuple`:
+                If `return_dict` is `True`, [`~pipelines.animatediff.pipeline_output.AnimateDiffPipelineOutput`] is
                 returned, otherwise a `tuple` is returned where the first element is a list with the generated frames.
         """
         # 0. Default height and width to unet
@@ -958,11 +965,10 @@ class AnimateDiffImgToVideoPipeline(
             return AnimateDiffPipelineOutput(frames=latents)
 
         # 10. Post-processing
-        video_tensor = self.decode_latents(latents)
-
-        if output_type == "pt":
-            video = video_tensor
+        if output_type == "latent":
+            video = latents
         else:
+            video_tensor = self.decode_latents(latents)
             video = tensor2vid(video_tensor, self.image_processor, output_type=output_type)
 
         # 11. Offload all models

examples/community/pipeline_prompt2prompt.py

@@ -15,18 +15,47 @@
 from __future__ import annotations
 
 import abc
+import inspect
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import numpy as np
 import torch
 import torch.nn.functional as F
-
-from diffusers.models.attention import Attention
-from diffusers.pipelines.stable_diffusion import (
-    StableDiffusionPipeline,
-    StableDiffusionPipelineOutput,
+from packaging import version
+from transformers import (
+    CLIPImageProcessor,
+    CLIPTextModel,
+    CLIPTokenizer,
+    CLIPVisionModelWithProjection,
 )
 
+from diffusers import AutoencoderKL, DiffusionPipeline, UNet2DConditionModel
+from diffusers.configuration_utils import FrozenDict, deprecate
+from diffusers.image_processor import VaeImageProcessor
+from diffusers.loaders import (
+    FromSingleFileMixin,
+    IPAdapterMixin,
+    LoraLoaderMixin,
+    TextualInversionLoaderMixin,
+)
+from diffusers.models.attention import Attention
+from diffusers.models.lora import adjust_lora_scale_text_encoder
+from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
+from diffusers.pipelines.stable_diffusion.safety_checker import (
+    StableDiffusionSafetyChecker,
+)
+from diffusers.schedulers import KarrasDiffusionSchedulers
+from diffusers.utils import (
+    USE_PEFT_BACKEND,
+    logging,
+    scale_lora_layers,
+    unscale_lora_layers,
+)
+from diffusers.utils.torch_utils import randn_tensor
+
+
+logger = logging.get_logger(__name__)
+
 
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
@@ -43,34 +72,486 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-class Prompt2PromptPipeline(StableDiffusionPipeline):
+class Prompt2PromptPipeline(
+    DiffusionPipeline,
+    TextualInversionLoaderMixin,
+    LoraLoaderMixin,
+    IPAdapterMixin,
+    FromSingleFileMixin,
+):
     r"""
+    Pipeline for text-to-image generation using Stable Diffusion.
+
+    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
+    implemented for all pipelines (downloading, saving, running on a particular device, etc.).
+
+    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:
-    Prompt-to-Prompt-Pipeline for text-to-image generation using Stable Diffusion. This model inherits from
-    [`StableDiffusionPipeline`]. Check the superclass documentation for the generic methods the library implements for
-    all the pipelines (such as downloading or saving, running on a particular device, etc.)
         vae ([`AutoencoderKL`]):
-            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
-        text_encoder ([`CLIPTextModel`]):
-            Frozen text-encoder. Stable Diffusion uses the text portion of
-            [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically
-            the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.
-        tokenizer (`CLIPTokenizer`):
-            Tokenizer of class
-            [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).
-        unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents. scheduler
-        ([`SchedulerMixin`]):
+            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.
+        scheduler ([`SchedulerMixin`]):
             A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of
             [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
         safety_checker ([`StableDiffusionSafetyChecker`]):
             Classification module that estimates whether generated images could be considered offensive or harmful.
-            Please, refer to the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5) for details.
-        feature_extractor ([`CLIPFeatureExtractor`]):
-            Model that extracts features from generated images to be used as inputs for the `safety_checker`.
+            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"
+    _exclude_from_cpu_offload = ["safety_checker"]
+    _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
     _optional_components = ["safety_checker", "feature_extractor"]
 
+    def __init__(
+        self,
+        vae: AutoencoderKL,
+        text_encoder: CLIPTextModel,
+        tokenizer: CLIPTokenizer,
+        unet: UNet2DConditionModel,
+        scheduler: KarrasDiffusionSchedulers,
+        safety_checker: StableDiffusionSafetyChecker,
+        feature_extractor: CLIPImageProcessor,
+        image_encoder: CLIPVisionModelWithProjection = None,
+        requires_safety_checker: bool = True,
+    ):
+        super().__init__()
+
+        if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1:
+            deprecation_message = (
+                f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`"
+                f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure "
+                "to update the config accordingly as leaving `steps_offset` might led to incorrect results"
+                " in future versions. If you have downloaded this checkpoint from the Hugging Face Hub,"
+                " it would be very nice if you could open a Pull request for the `scheduler/scheduler_config.json`"
+                " file"
+            )
+            deprecate("steps_offset!=1", "1.0.0", deprecation_message, standard_warn=False)
+            new_config = dict(scheduler.config)
+            new_config["steps_offset"] = 1
+            scheduler._internal_dict = FrozenDict(new_config)
+
+        if hasattr(scheduler.config, "clip_sample") and scheduler.config.clip_sample is True:
+            deprecation_message = (
+                f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`."
+                " `clip_sample` should be set to False in the configuration file. Please make sure to update the"
+                " config accordingly as not setting `clip_sample` in the config might lead to incorrect results in"
+                " future versions. If you have downloaded this checkpoint from the Hugging Face Hub, it would be very"
+                " nice if you could open a Pull request for the `scheduler/scheduler_config.json` file"
+            )
+            deprecate("clip_sample not set", "1.0.0", deprecation_message, standard_warn=False)
+            new_config = dict(scheduler.config)
+            new_config["clip_sample"] = False
+            scheduler._internal_dict = FrozenDict(new_config)
+
+        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."
+            )
+
+        is_unet_version_less_0_9_0 = hasattr(unet.config, "_diffusers_version") and version.parse(
+            version.parse(unet.config._diffusers_version).base_version
+        ) < version.parse("0.9.0.dev0")
+        is_unet_sample_size_less_64 = hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        if is_unet_version_less_0_9_0 and is_unet_sample_size_less_64:
+            deprecation_message = (
+                "The configuration file of the unet has set the default `sample_size` to smaller than"
+                " 64 which seems highly unlikely. If your checkpoint is a fine-tuned version of any of the"
+                " following: \n- CompVis/stable-diffusion-v1-4 \n- CompVis/stable-diffusion-v1-3 \n-"
+                " CompVis/stable-diffusion-v1-2 \n- CompVis/stable-diffusion-v1-1 \n- runwayml/stable-diffusion-v1-5"
+                " \n- runwayml/stable-diffusion-inpainting \n you should change 'sample_size' to 64 in the"
+                " configuration file. Please make sure to update the config accordingly as leaving `sample_size=32`"
+                " in the config might lead to incorrect results in future versions. If you have downloaded this"
+                " checkpoint from the Hugging Face Hub, it would be very nice if you could open a Pull request for"
+                " the `unet/config.json` file"
+            )
+            deprecate("sample_size<64", "1.0.0", deprecation_message, standard_warn=False)
+            new_config = dict(unet.config)
+            new_config["sample_size"] = 64
+            unet._internal_dict = FrozenDict(new_config)
+
+        self.register_modules(
+            vae=vae,
+            text_encoder=text_encoder,
+            tokenizer=tokenizer,
+            unet=unet,
+            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)
+        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.FloatTensor] = None,
+        negative_prompt_embeds: Optional[torch.FloatTensor] = 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.FloatTensor] = None,
+        negative_prompt_embeds: Optional[torch.FloatTensor] = 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.FloatTensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+            lora_scale (`float`, *optional*):
+                A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
+            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 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.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.prepare_extra_step_kwargs
+    def prepare_extra_step_kwargs(self, generator, eta):
+        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
+        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
+        # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
+        # and should be between [0, 1]
+
+        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        extra_step_kwargs = {}
+        if accepts_eta:
+            extra_step_kwargs["eta"] = eta
+
+        # check if the scheduler accepts generator
+        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        if accepts_generator:
+            extra_step_kwargs["generator"] = generator
+        return extra_step_kwargs
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.check_inputs
+    def check_inputs(
+        self,
+        prompt,
+        height,
+        width,
+        callback_steps,
+        negative_prompt=None,
+        prompt_embeds=None,
+        negative_prompt_embeds=None,
+        ip_adapter_image=None,
+        ip_adapter_image_embeds=None,
+        callback_on_step_end_tensor_inputs=None,
+    ):
+        if height % 8 != 0 or width % 8 != 0:
+            raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")
+
+        if callback_steps is 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}."
+                )
+
+        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."
+            )
+
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
+    def prepare_latents(
+        self,
+        batch_size,
+        num_channels_latents,
+        height,
+        width,
+        dtype,
+        device,
+        generator,
+        latents=None,
+    ):
+        shape = (
+            batch_size,
+            num_channels_latents,
+            height // self.vae_scale_factor,
+            width // self.vae_scale_factor,
+        )
+        if isinstance(generator, list) and len(generator) != batch_size:
+            raise ValueError(
+                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
+                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
+            )
+
+        if latents is None:
+            latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
+        else:
+            latents = latents.to(device)
+
+        # scale the initial noise by the standard deviation required by the scheduler
+        latents = latents * self.scheduler.init_noise_sigma
+        return latents
+
     @torch.no_grad()
     def __call__(
         self,

examples/community/pipeline_stable_diffusion_xl_instantid.py

@@ -452,7 +452,7 @@ class StableDiffusionXLInstantIDPipeline(StableDiffusionXLControlNetPipeline):
 
                 xformers_version = version.parse(xformers.__version__)
                 if xformers_version == version.parse("0.0.16"):
-                    logger.warn(
+                    logger.warning(
                         "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                     )
                 self.enable_xformers_memory_efficient_attention()

examples/community/rerender_a_video.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import sys
 from dataclasses import dataclass
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
@@ -21,6 +20,7 @@ import PIL.Image
 import torch
 import torch.nn.functional as F
 import torchvision.transforms as T
+from gmflow.gmflow import GMFlow
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 
 from diffusers.image_processor import VaeImageProcessor
@@ -34,13 +34,6 @@ from diffusers.utils import BaseOutput, deprecate, logging
 from diffusers.utils.torch_utils import is_compiled_module, randn_tensor
 
 
-gmflow_dir = "/path/to/gmflow"
-sys.path.insert(0, gmflow_dir)
-from gmflow.gmflow import GMFlow  # noqa: E402
-
-from utils.utils import InputPadder  # noqa: E402
-
-
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
@@ -119,11 +112,11 @@ def forward_backward_consistency_check(fwd_flow, bwd_flow, alpha=0.01, beta=0.5)
 
 
 @torch.no_grad()
-def get_warped_and_mask(flow_model, image1, image2, image3=None, pixel_consistency=False):
+def get_warped_and_mask(flow_model, image1, image2, image3=None, pixel_consistency=False, device=None):
     if image3 is None:
         image3 = image1
     padder = InputPadder(image1.shape, padding_factor=8)
-    image1, image2 = padder.pad(image1[None].cuda(), image2[None].cuda())
+    image1, image2 = padder.pad(image1[None].to(device), image2[None].to(device))
     results_dict = flow_model(
         image1, image2, attn_splits_list=[2], corr_radius_list=[-1], prop_radius_list=[-1], pred_bidir_flow=True
     )
@@ -307,6 +300,7 @@ class RerenderAVideoPipeline(StableDiffusionControlNetImg2ImgPipeline):
         feature_extractor: CLIPImageProcessor,
         image_encoder=None,
         requires_safety_checker: bool = True,
+        device=None,
     ):
         super().__init__(
             vae,
@@ -320,6 +314,7 @@ class RerenderAVideoPipeline(StableDiffusionControlNetImg2ImgPipeline):
             image_encoder,
             requires_safety_checker,
         )
+        self.to(device)
 
         if safety_checker is None and requires_safety_checker:
             logger.warning(
@@ -374,7 +369,7 @@ class RerenderAVideoPipeline(StableDiffusionControlNetImg2ImgPipeline):
             attention_type="swin",
             ffn_dim_expansion=4,
             num_transformer_layers=6,
-        ).to("cuda")
+        ).to(self.device)
 
         checkpoint = torch.utils.model_zoo.load_url(
             "https://huggingface.co/Anonymous-sub/Rerender/resolve/main/models/gmflow_sintel-0c07dcb3.pth",
@@ -928,13 +923,13 @@ class RerenderAVideoPipeline(StableDiffusionControlNetImg2ImgPipeline):
             prev_image = self.image_processor.preprocess(prev_image).to(dtype=torch.float32)
 
             warped_0, bwd_occ_0, bwd_flow_0 = get_warped_and_mask(
-                self.flow_model, first_image, image[0], first_result, False
+                self.flow_model, first_image, image[0], first_result, False, self.device
             )
             blend_mask_0 = blur(F.max_pool2d(bwd_occ_0, kernel_size=9, stride=1, padding=4))
             blend_mask_0 = torch.clamp(blend_mask_0 + bwd_occ_0, 0, 1)
 
             warped_pre, bwd_occ_pre, bwd_flow_pre = get_warped_and_mask(
-                self.flow_model, prev_image[0], image[0], prev_result, False
+                self.flow_model, prev_image[0], image[0], prev_result, False, self.device
             )
             blend_mask_pre = blur(F.max_pool2d(bwd_occ_pre, kernel_size=9, stride=1, padding=4))
             blend_mask_pre = torch.clamp(blend_mask_pre + bwd_occ_pre, 0, 1)
@@ -1176,3 +1171,24 @@ class RerenderAVideoPipeline(StableDiffusionControlNetImg2ImgPipeline):
             return output_frames
 
         return TextToVideoSDPipelineOutput(frames=output_frames)
+
+
+class InputPadder:
+    """Pads images such that dimensions are divisible by 8"""
+
+    def __init__(self, dims, mode="sintel", padding_factor=8):
+        self.ht, self.wd = dims[-2:]
+        pad_ht = (((self.ht // padding_factor) + 1) * padding_factor - self.ht) % padding_factor
+        pad_wd = (((self.wd // padding_factor) + 1) * padding_factor - self.wd) % padding_factor
+        if mode == "sintel":
+            self._pad = [pad_wd // 2, pad_wd - pad_wd // 2, pad_ht // 2, pad_ht - pad_ht // 2]
+        else:
+            self._pad = [pad_wd // 2, pad_wd - pad_wd // 2, 0, pad_ht]
+
+    def pad(self, *inputs):
+        return [F.pad(x, self._pad, mode="replicate") for x in inputs]
+
+    def unpad(self, x):
+        ht, wd = x.shape[-2:]
+        c = [self._pad[2], ht - self._pad[3], self._pad[0], wd - self._pad[1]]
+        return x[..., c[0] : c[1], c[2] : c[3]]

examples/community/scheduling_ufogen.py

@@ -171,9 +171,7 @@ class UFOGenScheduler(SchedulerMixin, ConfigMixin):
             The way the timesteps should be scaled. Refer to Table 2 of the [Common Diffusion Noise Schedules and
             Sample Steps are Flawed](https://huggingface.co/papers/2305.08891) for more information.
         steps_offset (`int`, defaults to 0):
-            An offset added to the inference steps. You can use a combination of `offset=1` and
-            `set_alpha_to_one=False` to make the last step use step 0 for the previous alpha product like in Stable
-            Diffusion.
+            An offset added to the inference steps, as required by some model families.
         rescale_betas_zero_snr (`bool`, defaults to `False`):
             Whether to rescale the betas to have zero terminal SNR. This enables the model to generate very bright and
             dark samples instead of limiting it to samples with medium brightness. Loosely related to

examples/consistency_distillation/train_lcm_distill_lora_sd_wds.py

@@ -72,7 +72,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -308,7 +308,7 @@ def log_validation(vae, unet, args, accelerator, weight_dtype, step):
 
             tracker.log({"validation": formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -1068,7 +1068,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/consistency_distillation/train_lcm_distill_lora_sdxl.py

@@ -65,7 +65,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -180,7 +180,7 @@ def log_validation(vae, args, accelerator, weight_dtype, step, unet=None, is_fin
             logger_name = "test" if is_final_validation else "validation"
             tracker.log({logger_name: formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -928,7 +928,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/consistency_distillation/train_lcm_distill_lora_sdxl_wds.py

@@ -78,7 +78,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -325,7 +325,7 @@ def log_validation(vae, unet, args, accelerator, weight_dtype, step):
 
             tracker.log({"validation": formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -1083,7 +1083,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/consistency_distillation/train_lcm_distill_sd_wds.py

@@ -71,7 +71,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -285,7 +285,7 @@ def log_validation(vae, unet, args, accelerator, weight_dtype, step, name="targe
 
             tracker.log({f"validation/{name}": formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -1023,7 +1023,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/consistency_distillation/train_lcm_distill_sdxl_wds.py

@@ -77,7 +77,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -303,7 +303,7 @@ def log_validation(vae, unet, args, accelerator, weight_dtype, step, name="targe
 
             tracker.log({f"validation/{name}": formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -1083,7 +1083,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/controlnet/train_controlnet.py

@@ -60,7 +60,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -178,7 +178,7 @@ def log_validation(
 
             tracker.log({tracker_key: formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -242,6 +242,7 @@ These are controlnet weights trained on {base_model} with new type of conditioni
         "text-to-image",
         "diffusers",
         "controlnet",
+        "diffusers-training",
     ]
     model_card = populate_model_card(model_card, tags=tags)
 
@@ -860,7 +861,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/controlnet/train_controlnet_flax.py

@@ -60,7 +60,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = logging.getLogger(__name__)
 
@@ -128,7 +128,7 @@ def log_validation(pipeline, pipeline_params, controlnet_params, tokenizer, args
 
         wandb.log({"validation": formatted_images})
     else:
-        logger.warn(f"image logging not implemented for {args.report_to}")
+        logger.warning(f"image logging not implemented for {args.report_to}")
 
     return image_logs
 
@@ -169,6 +169,7 @@ These are controlnet weights trained on {base_model} with new type of conditioni
         "diffusers",
         "controlnet",
         "jax-diffusers-event",
+        "diffusers-training",
     ]
     model_card = populate_model_card(model_card, tags=tags)
 

examples/controlnet/train_controlnet_sdxl.py

@@ -61,7 +61,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -178,7 +178,7 @@ def log_validation(vae, unet, controlnet, args, accelerator, weight_dtype, step,
 
             tracker.log({tracker_key: formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -243,6 +243,7 @@ These are controlnet weights trained on {base_model} with new type of conditioni
         "text-to-image",
         "diffusers",
         "controlnet",
+        "diffusers-training",
     ]
     model_card = populate_model_card(model_card, tags=tags)
 
@@ -928,7 +929,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/custom_diffusion/train_custom_diffusion.py

@@ -63,7 +63,7 @@ from diffusers.utils.import_utils import is_xformers_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -97,7 +97,14 @@ These are Custom Diffusion adaption weights for {base_model}. The weights were t
         inference=True,
     )
 
-    tags = ["text-to-image", "diffusers", "stable-diffusion", "stable-diffusion-diffusers", "custom-diffusion"]
+    tags = [
+        "text-to-image",
+        "diffusers",
+        "stable-diffusion",
+        "stable-diffusion-diffusers",
+        "custom-diffusion",
+        "diffusers-training",
+    ]
     model_card = populate_model_card(model_card, tags=tags)
 
     model_card.save(os.path.join(repo_folder, "README.md"))
@@ -897,7 +904,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             attention_class = CustomDiffusionXFormersAttnProcessor
@@ -1171,7 +1178,7 @@ def main(args):
                         grads_text_encoder = text_encoder.get_input_embeddings().weight.grad
                     # Get the index for tokens that we want to zero the grads for
                     index_grads_to_zero = torch.arange(len(tokenizer)) != modifier_token_id[0]
-                    for i in range(len(modifier_token_id[1:])):
+                    for i in range(1, len(modifier_token_id)):
                         index_grads_to_zero = index_grads_to_zero & (
                             torch.arange(len(tokenizer)) != modifier_token_id[i]
                         )

examples/dreambooth/README_sdxl.md

@@ -243,3 +243,29 @@ accelerate launch train_dreambooth_lora_sdxl.py \
 
 > [!CAUTION]
 > Min-SNR gamma is not supported with the EDM-style training yet. When training with the PlaygroundAI model, it's recommended to not pass any "variant".
+
+### DoRA training 
+The script now supports DoRA training too!
+> Proposed in [DoRA: Weight-Decomposed Low-Rank Adaptation](https://arxiv.org/abs/2402.09353), 
+**DoRA** is very similar to LoRA, except it decomposes the pre-trained weight into two components, **magnitude** and **direction** and employs LoRA for _directional_ updates to efficiently minimize the number of trainable parameters. 
+The authors found that by using DoRA, both the learning capacity and training stability of LoRA are enhanced without any additional overhead during inference. 
+
+> [!NOTE]
+> 💡DoRA training is still _experimental_  
+> and is likely to require different hyperparameter values to perform best compared to a LoRA.
+> Specifically, we've noticed 2 differences to take into account your training: 
+> 1. **LoRA seem to converge faster than DoRA** (so a set of parameters that may lead to overfitting when training a LoRA may be working well for a DoRA)
+> 2. **DoRA quality superior to LoRA especially in lower ranks** the difference in quality of DoRA of rank 8 and LoRA of rank 8 appears to be more significant than when training ranks of 32 or 64 for example.  
+> This is also aligned with some of the quantitative analysis shown in the paper. 
+
+**Usage**
+1. To use DoRA you need to install `peft` from main: 
+```bash
+pip install git+https://github.com/huggingface/peft.git
+```
+2. Enable DoRA training by adding this flag
+```bash
+--use_dora
+```
+**Inference** 
+The inference is the same as if you train a regular LoRA 🤗

examples/dreambooth/train_dreambooth.py

@@ -63,7 +63,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -102,7 +102,7 @@ DreamBooth for the text encoder was enabled: {train_text_encoder}.
         inference=True,
     )
 
-    tags = ["text-to-image", "dreambooth"]
+    tags = ["text-to-image", "dreambooth", "diffusers-training"]
     if isinstance(pipeline, StableDiffusionPipeline):
         tags.extend(["stable-diffusion", "stable-diffusion-diffusers"])
     else:
@@ -987,7 +987,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/dreambooth/train_dreambooth_flax.py

@@ -35,7 +35,7 @@ from diffusers.utils import check_min_version
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 # Cache compiled models across invocations of this script.
 cc.initialize_cache(os.path.expanduser("~/.cache/jax/compilation_cache"))

examples/dreambooth/train_dreambooth_lora.py

@@ -70,7 +70,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -106,7 +106,7 @@ LoRA for the text encoder was enabled: {train_text_encoder}.
         model_description=model_description,
         inference=True,
     )
-    tags = ["text-to-image", "diffusers", "lora"]
+    tags = ["text-to-image", "diffusers", "lora", "diffusers-training"]
     if isinstance(pipeline, StableDiffusionPipeline):
         tags.extend(["stable-diffusion", "stable-diffusion-diffusers"])
     else:
@@ -895,7 +895,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/dreambooth/train_dreambooth_lora_sdxl.py

@@ -75,7 +75,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -96,6 +96,7 @@ def determine_scheduler_type(pretrained_model_name_or_path, revision):
 
 def save_model_card(
     repo_id: str,
+    use_dora: bool,
     images=None,
     base_model: str = None,
     train_text_encoder=False,
@@ -113,7 +114,7 @@ def save_model_card(
             )
 
     model_description = f"""
-# {'SDXL' if 'playgroundai' not in base_model else 'Playground'} LoRA DreamBooth - {repo_id}
+# {'SDXL' if 'playground' not in base_model else 'Playground'} LoRA DreamBooth - {repo_id}
 
 <Gallery />
 
@@ -138,7 +139,7 @@ Weights for this model are available in Safetensors format.
 [Download]({repo_id}/tree/main) them in the Files & versions tab.
 
 """
-    if "playgroundai" in args.pretrained_model_name_or_path:
+    if "playground" in base_model:
         model_description += """\n
 ## License
 
@@ -147,7 +148,7 @@ Please adhere to the licensing terms as described [here](https://huggingface.co/
     model_card = load_or_create_model_card(
         repo_id_or_path=repo_id,
         from_training=True,
-        license="openrail++" if "playgroundai" not in base_model else "playground-v2dot5-community",
+        license="openrail++" if "playground" not in base_model else "playground-v2dot5-community",
         base_model=base_model,
         prompt=instance_prompt,
         model_description=model_description,
@@ -156,11 +157,12 @@ Please adhere to the licensing terms as described [here](https://huggingface.co/
     tags = [
         "text-to-image",
         "text-to-image",
+        "diffusers-training",
         "diffusers",
-        "lora",
+        "lora" if not use_dora else "dora",
         "template:sd-lora",
     ]
-    if "playgroundai" in base_model:
+    if "playground" in base_model:
         tags.extend(["playground", "playground-diffusers"])
     else:
         tags.extend(["stable-diffusion-xl", "stable-diffusion-xl-diffusers"])
@@ -204,7 +206,7 @@ def log_validation(
     # Currently the context determination is a bit hand-wavy. We can improve it in the future if there's a better
     # way to condition it. Reference: https://github.com/huggingface/diffusers/pull/7126#issuecomment-1968523051
     inference_ctx = (
-        contextlib.nullcontext() if "playgroundai" in args.pretrained_model_name_or_path else torch.cuda.amp.autocast()
+        contextlib.nullcontext() if "playground" in args.pretrained_model_name_or_path else torch.cuda.amp.autocast()
     )
 
     with inference_ctx:
@@ -645,6 +647,15 @@ def parse_args(input_args=None):
         default=4,
         help=("The dimension of the LoRA update matrices."),
     )
+    parser.add_argument(
+        "--use_dora",
+        action="store_true",
+        default=False,
+        help=(
+            "Wether to train a DoRA as proposed in- DoRA: Weight-Decomposed Low-Rank Adaptation https://arxiv.org/abs/2402.09353. "
+            "Note: to use DoRA you need to install peft from main, `pip install git+https://github.com/huggingface/peft.git`"
+        ),
+    )
 
     if input_args is not None:
         args = parser.parse_args(input_args)
@@ -866,6 +877,8 @@ def collate_fn(examples, with_prior_preservation=False):
     if with_prior_preservation:
         pixel_values += [example["class_images"] for example in examples]
         prompts += [example["class_prompt"] for example in examples]
+        original_sizes += [example["original_size"] for example in examples]
+        crop_top_lefts += [example["crop_top_left"] for example in examples]
 
     pixel_values = torch.stack(pixel_values)
     pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
@@ -1128,7 +1141,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, "
                     "please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
@@ -1145,6 +1158,7 @@ def main(args):
     # now we will add new LoRA weights to the attention layers
     unet_lora_config = LoraConfig(
         r=args.rank,
+        use_dora=args.use_dora,
         lora_alpha=args.rank,
         init_lora_weights="gaussian",
         target_modules=["to_k", "to_q", "to_v", "to_out.0"],
@@ -1156,6 +1170,7 @@ def main(args):
     if args.train_text_encoder:
         text_lora_config = LoraConfig(
             r=args.rank,
+            use_dora=args.use_dora,
             lora_alpha=args.rank,
             init_lora_weights="gaussian",
             target_modules=["q_proj", "k_proj", "v_proj", "out_proj"],
@@ -1302,14 +1317,14 @@ def main(args):
 
     # Optimizer creation
     if not (args.optimizer.lower() == "prodigy" or args.optimizer.lower() == "adamw"):
-        logger.warn(
+        logger.warning(
             f"Unsupported choice of optimizer: {args.optimizer}.Supported optimizers include [adamW, prodigy]."
             "Defaulting to adamW"
         )
         args.optimizer = "adamw"
 
     if args.use_8bit_adam and not args.optimizer.lower() == "adamw":
-        logger.warn(
+        logger.warning(
             f"use_8bit_adam is ignored when optimizer is not set to 'AdamW'. Optimizer was "
             f"set to {args.optimizer.lower()}"
         )
@@ -1343,11 +1358,11 @@ def main(args):
         optimizer_class = prodigyopt.Prodigy
 
         if args.learning_rate <= 0.1:
-            logger.warn(
+            logger.warning(
                 "Learning rate is too low. When using prodigy, it's generally better to set learning rate around 1.0"
             )
         if args.train_text_encoder and args.text_encoder_lr:
-            logger.warn(
+            logger.warning(
                 f"Learning rates were provided both for the unet and the text encoder- e.g. text_encoder_lr:"
                 f" {args.text_encoder_lr} and learning_rate: {args.learning_rate}. "
                 f"When using prodigy only learning_rate is used as the initial learning rate."
@@ -1494,7 +1509,7 @@ def main(args):
     if accelerator.is_main_process:
         tracker_name = (
             "dreambooth-lora-sd-xl"
-            if "playgroundai" not in args.pretrained_model_name_or_path
+            if "playground" not in args.pretrained_model_name_or_path
             else "dreambooth-lora-playground"
         )
         accelerator.init_trackers(tracker_name, config=vars(args))
@@ -1912,6 +1927,7 @@ def main(args):
         if args.push_to_hub:
             save_model_card(
                 repo_id,
+                use_dora=args.use_dora,
                 images=images,
                 base_model=args.pretrained_model_name_or_path,
                 train_text_encoder=args.train_text_encoder,

examples/instruct_pix2pix/train_instruct_pix2pix.py

@@ -53,7 +53,7 @@ from diffusers.utils.torch_utils import is_compiled_module
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -488,7 +488,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/instruct_pix2pix/train_instruct_pix2pix_sdxl.py

@@ -59,7 +59,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -580,7 +580,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/kandinsky2_2/text_to_image/train_text_to_image_decoder.py

@@ -52,7 +52,7 @@ if is_wandb_available():
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -81,6 +81,7 @@ tags:
 - kandinsky
 - text-to-image
 - diffusers
+- diffusers-training
 inference: true
 ---
     """
@@ -176,7 +177,7 @@ def log_validation(vae, image_encoder, image_processor, unet, args, accelerator,
                 }
             )
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
     del pipeline
     torch.cuda.empty_cache()
@@ -533,7 +534,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/kandinsky2_2/text_to_image/train_text_to_image_lora_decoder.py

@@ -46,7 +46,7 @@ from diffusers.utils import check_min_version, is_wandb_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -65,6 +65,7 @@ tags:
 - kandinsky
 - text-to-image
 - diffusers
+- diffusers-training
 - lora
 inference: true
 ---

examples/kandinsky2_2/text_to_image/train_text_to_image_lora_prior.py

@@ -46,7 +46,7 @@ from diffusers.utils import check_min_version, is_wandb_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -65,6 +65,7 @@ tags:
 - kandinsky
 - text-to-image
 - diffusers
+- diffusers-training
 - lora
 inference: true
 ---

examples/kandinsky2_2/text_to_image/train_text_to_image_prior.py

@@ -51,7 +51,7 @@ if is_wandb_available():
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -82,6 +82,7 @@ tags:
 - kandinsky
 - text-to-image
 - diffusers
+- diffusers-training
 inference: true
 ---
     """
@@ -179,7 +180,7 @@ def log_validation(
                 }
             )
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
     del pipeline
     torch.cuda.empty_cache()

examples/research_projects/consistency_training/train_cm_ct_unconditional.py

@@ -219,7 +219,7 @@ def log_validation(unet, scheduler, args, accelerator, weight_dtype, step, name=
         if args.num_classes is not None:
             class_labels = list(range(args.num_classes))
         else:
-            logger.warn(
+            logger.warning(
                 "The model is class-conditional but the number of classes is not set. The generated images will be"
                 " unconditional rather than class-conditional."
             )
@@ -266,7 +266,7 @@ def log_validation(unet, scheduler, args, accelerator, weight_dtype, step, name=
 
             tracker.log({f"validation/{name}": formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
     del pipeline
     gc.collect()
@@ -863,14 +863,14 @@ def main(args):
     elif args.model_config_name_or_path is None:
         # TODO: use default architectures from iCT paper
         if not args.class_conditional and (args.num_classes is not None or args.class_embed_type is not None):
-            logger.warn(
+            logger.warning(
                 f"`--class_conditional` is set to `False` but `--num_classes` is set to {args.num_classes} and"
                 f" `--class_embed_type` is set to {args.class_embed_type}. These values will be overridden to `None`."
             )
             args.num_classes = None
             args.class_embed_type = None
         elif args.class_conditional and args.num_classes is None and args.class_embed_type is None:
-            logger.warn(
+            logger.warning(
                 "`--class_conditional` is set to `True` but neither `--num_classes` nor `--class_embed_type` is set."
                 "`class_conditional` will be overridden to `False`."
             )
@@ -996,7 +996,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/research_projects/controlnet/train_controlnet_webdataset.py

@@ -407,7 +407,7 @@ def log_validation(vae, unet, controlnet, args, accelerator, weight_dtype, step)
 
             tracker.log({"validation": formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -460,6 +460,8 @@ tags:
 - text-to-image
 - diffusers
 - controlnet
+- diffusers-training
+- webdataset
 inference: true
 ---
     """
@@ -1055,7 +1057,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/research_projects/diffusion_dpo/README.md

@@ -61,6 +61,34 @@ accelerate launch train_diffusion_dpo_sdxl.py \
   --push_to_hub
 ```
 
+## SDXL Turbo training command
+
+```bash
+accelerate launch train_diffusion_dpo_sdxl.py \
+  --pretrained_model_name_or_path=stabilityai/sdxl-turbo \
+  --pretrained_vae_model_name_or_path=madebyollin/sdxl-vae-fp16-fix \
+  --output_dir="diffusion-sdxl-turbo-dpo" \
+  --mixed_precision="fp16" \
+  --dataset_name=kashif/pickascore \
+  --train_batch_size=8 \
+  --gradient_accumulation_steps=2 \
+  --gradient_checkpointing \
+  --use_8bit_adam \
+  --rank=8 \
+  --learning_rate=1e-5 \
+  --report_to="wandb" \
+  --lr_scheduler="constant" \
+  --lr_warmup_steps=0 \
+  --max_train_steps=2000 \
+  --checkpointing_steps=500 \
+  --run_validation --validation_steps=50 \
+  --seed="0" \
+  --report_to="wandb" \
+  --is_turbo --resolution 512 \
+  --push_to_hub
+```
+
+
 ## Acknowledgements
 
 This is based on the amazing work done by [Bram](https://github.com/bram-w) here for Diffusion DPO: https://github.com/bram-w/trl/blob/dpo/. 

examples/research_projects/diffusion_dpo/train_diffusion_dpo.py

@@ -574,7 +574,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/research_projects/diffusion_dpo/train_diffusion_dpo_sdxl.py

@@ -118,9 +118,16 @@ def log_validation(args, unet, vae, accelerator, weight_dtype, epoch, is_final_v
     images = []
     context = contextlib.nullcontext() if is_final_validation else torch.cuda.amp.autocast()
 
+    guidance_scale = 5.0
+    num_inference_steps = 25
+    if args.is_turbo:
+        guidance_scale = 0.0
+        num_inference_steps = 4
     for prompt in VALIDATION_PROMPTS:
         with context:
-            image = pipeline(prompt, num_inference_steps=25, generator=generator).images[0]
+            image = pipeline(
+                prompt, num_inference_steps=num_inference_steps, guidance_scale=guidance_scale, generator=generator
+            ).images[0]
             images.append(image)
 
     tracker_key = "test" if is_final_validation else "validation"
@@ -141,7 +148,10 @@ def log_validation(args, unet, vae, accelerator, weight_dtype, epoch, is_final_v
     if is_final_validation:
         pipeline.disable_lora()
         no_lora_images = [
-            pipeline(prompt, num_inference_steps=25, generator=generator).images[0] for prompt in VALIDATION_PROMPTS
+            pipeline(
+                prompt, num_inference_steps=num_inference_steps, guidance_scale=guidance_scale, generator=generator
+            ).images[0]
+            for prompt in VALIDATION_PROMPTS
         ]
 
         for tracker in accelerator.trackers:
@@ -423,6 +433,11 @@ def parse_args(input_args=None):
     parser.add_argument(
         "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
     )
+    parser.add_argument(
+        "--is_turbo",
+        action="store_true",
+        help=("Use if tuning SDXL Turbo instead of SDXL"),
+    )
     parser.add_argument(
         "--rank",
         type=int,
@@ -444,6 +459,9 @@ def parse_args(input_args=None):
     if args.dataset_name is None:
         raise ValueError("Must provide a `dataset_name`.")
 
+    if args.is_turbo:
+        assert "turbo" in args.pretrained_model_name_or_path
+
     env_local_rank = int(os.environ.get("LOCAL_RANK", -1))
     if env_local_rank != -1 and env_local_rank != args.local_rank:
         args.local_rank = env_local_rank
@@ -560,6 +578,36 @@ def main(args):
 
     # Load scheduler and models
     noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+
+    def enforce_zero_terminal_snr(scheduler):
+        # Modified from https://github.com/huggingface/diffusers/blob/main/src/diffusers/schedulers/scheduling_ddpm.py#L93
+        # Original implementation https://arxiv.org/pdf/2305.08891.pdf
+        # Turbo needs zero terminal SNR
+        # Turbo: https://static1.squarespace.com/static/6213c340453c3f502425776e/t/65663480a92fba51d0e1023f/1701197769659/adversarial_diffusion_distillation.pdf
+        # Convert betas to alphas_bar_sqrt
+        alphas = 1 - scheduler.betas
+        alphas_bar = alphas.cumprod(0)
+        alphas_bar_sqrt = alphas_bar.sqrt()
+
+        # Store old values.
+        alphas_bar_sqrt_0 = alphas_bar_sqrt[0].clone()
+        alphas_bar_sqrt_T = alphas_bar_sqrt[-1].clone()
+        # Shift so last timestep is zero.
+        alphas_bar_sqrt -= alphas_bar_sqrt_T
+        # Scale so first timestep is back to old value.
+        alphas_bar_sqrt *= alphas_bar_sqrt_0 / (alphas_bar_sqrt_0 - alphas_bar_sqrt_T)
+
+        alphas_bar = alphas_bar_sqrt**2
+        alphas = alphas_bar[1:] / alphas_bar[:-1]
+        alphas = torch.cat([alphas_bar[0:1], alphas])
+
+        alphas_cumprod = torch.cumprod(alphas, dim=0)
+        scheduler.alphas_cumprod = alphas_cumprod
+        return
+
+    if args.is_turbo:
+        enforce_zero_terminal_snr(noise_scheduler)
+
     text_encoder_one = text_encoder_cls_one.from_pretrained(
         args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant
     )
@@ -624,7 +672,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()
@@ -909,6 +957,10 @@ def main(args):
                 timesteps = torch.randint(
                     0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device, dtype=torch.long
                 ).repeat(2)
+                if args.is_turbo:
+                    # Learn a 4 timestep schedule
+                    timesteps_0_to_3 = timesteps % 4
+                    timesteps = 250 * timesteps_0_to_3 + 249
 
                 # Add noise to the model input according to the noise magnitude at each timestep
                 # (this is the forward diffusion process)

examples/research_projects/instructpix2pix_lora/train_instruct_pix2pix_lora.py

@@ -516,7 +516,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/research_projects/intel_opts/textual_inversion/textual_inversion_bf16.py

@@ -608,7 +608,7 @@ def main():
     # Create the pipeline using using the trained modules and save it.
     if accelerator.is_main_process:
         if args.push_to_hub and args.only_save_embeds:
-            logger.warn("Enabling full model saving because --push_to_hub=True was specified.")
+            logger.warning("Enabling full model saving because --push_to_hub=True was specified.")
             save_full_model = True
         else:
             save_full_model = not args.only_save_embeds

examples/research_projects/lora/train_text_to_image_lora.py

@@ -69,6 +69,7 @@ tags:
 - stable-diffusion-diffusers
 - text-to-image
 - diffusers
+- diffusers-training
 - lora
 inference: true
 ---
@@ -540,7 +541,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/research_projects/multi_token_textual_inversion/textual_inversion.py

@@ -645,7 +645,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()
@@ -901,7 +901,7 @@ def main():
     accelerator.wait_for_everyone()
     if accelerator.is_main_process:
         if args.push_to_hub and args.only_save_embeds:
-            logger.warn("Enabling full model saving because --push_to_hub=True was specified.")
+            logger.warning("Enabling full model saving because --push_to_hub=True was specified.")
             save_full_model = True
         else:
             save_full_model = not args.only_save_embeds

examples/research_projects/onnxruntime/text_to_image/train_text_to_image.py

@@ -108,7 +108,7 @@ def log_validation(vae, text_encoder, tokenizer, unet, args, accelerator, weight
                 }
             )
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
     del pipeline
     torch.cuda.empty_cache()
@@ -523,7 +523,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/research_projects/onnxruntime/textual_inversion/textual_inversion.py

@@ -100,6 +100,8 @@ tags:
 - text-to-image
 - diffusers
 - textual_inversion
+- diffusers-training
+- onxruntime
 inference: true
 ---
     """
@@ -685,7 +687,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()
@@ -914,7 +916,7 @@ def main():
     accelerator.wait_for_everyone()
     if accelerator.is_main_process:
         if args.push_to_hub and not args.save_as_full_pipeline:
-            logger.warn("Enabling full model saving because --push_to_hub=True was specified.")
+            logger.warning("Enabling full model saving because --push_to_hub=True was specified.")
             save_full_model = True
         else:
             save_full_model = args.save_as_full_pipeline

examples/research_projects/onnxruntime/unconditional_image_generation/train_unconditional.py

@@ -410,7 +410,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             model.enable_xformers_memory_efficient_attention()
@@ -637,7 +637,7 @@ def main(args):
                     generator=generator,
                     batch_size=args.eval_batch_size,
                     num_inference_steps=args.ddpm_num_inference_steps,
-                    output_type="numpy",
+                    output_type="np",
                 ).images
 
                 if args.use_ema:

examples/research_projects/promptdiffusion/README.md

@@ -0,0 +1,50 @@
+# PromptDiffusion Pipeline
+
+From the project [page](https://zhendong-wang.github.io/prompt-diffusion.github.io/)
+
+"With a prompt consisting of a task-specific example pair of images and text guidance, and a new query image, Prompt Diffusion can comprehend the desired task and generate the corresponding output image on both seen (trained) and unseen (new) task types."
+
+For any usage questions, please refer to the [paper](https://arxiv.org/abs/2305.01115).
+
+Prepare models by converting them from the [checkpoint](https://huggingface.co/zhendongw/prompt-diffusion)
+
+To convert the controlnet, use cldm_v15.yaml from the [repository](https://github.com/Zhendong-Wang/Prompt-Diffusion/tree/main/models/):
+
+```bash
+python convert_original_promptdiffusion_to_diffusers.py --checkpoint_path path-to-network-step04999.ckpt --original_config_file path-to-cldm_v15.yaml --dump_path path-to-output-directory
+```
+
+To learn about how to convert the fine-tuned stable diffusion model, see the [Load different Stable Diffusion formats guide](https://huggingface.co/docs/diffusers/main/en/using-diffusers/other-formats).
+
+
+```py
+import torch
+from diffusers import UniPCMultistepScheduler
+from diffusers.utils import load_image
+from promptdiffusioncontrolnet import PromptDiffusionControlNetModel
+from pipeline_prompt_diffusion import PromptDiffusionPipeline
+
+
+from PIL import ImageOps
+
+image_a = ImageOps.invert(load_image("https://github.com/Zhendong-Wang/Prompt-Diffusion/blob/main/images_to_try/house_line.png?raw=true"))
+
+image_b = load_image("https://github.com/Zhendong-Wang/Prompt-Diffusion/blob/main/images_to_try/house.png?raw=true")
+query = ImageOps.invert(load_image("https://github.com/Zhendong-Wang/Prompt-Diffusion/blob/main/images_to_try/new_01.png?raw=true"))
+
+# load prompt diffusion controlnet and prompt diffusion
+
+controlnet = PromptDiffusionControlNetModel.from_pretrained("iczaw/prompt-diffusion-diffusers", subfolder="controlnet", torch_dtype=torch.float16)
+model_id = "path-to-model"
+pipe = PromptDiffusionPipeline.from_pretrained("iczaw/prompt-diffusion-diffusers", subfolder="base", controlnet=controlnet, torch_dtype=torch.float16, variant="fp16")
+
+# 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("a tortoise", num_inference_steps=20, generator=generator, image_pair=[image_a,image_b], image=query).images[0]
+
+```

examples/research_projects/promptdiffusion/promptdiffusioncontrolnet.py

@@ -0,0 +1,385 @@
+# Copyright 2023 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import Any, Dict, Optional, Tuple, Union
+
+import torch
+
+from diffusers.configuration_utils import register_to_config
+from diffusers.models.controlnet import (
+    ControlNetConditioningEmbedding,
+    ControlNetModel,
+    ControlNetOutput,
+)
+from diffusers.utils import logging
+
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+class PromptDiffusionControlNetModel(ControlNetModel):
+    """
+    A PromptDiffusionControlNet model.
+
+    Args:
+        in_channels (`int`, defaults to 4):
+            The number of channels in the input sample.
+        flip_sin_to_cos (`bool`, defaults to `True`):
+            Whether to flip the sin to cos in the time embedding.
+        freq_shift (`int`, defaults to 0):
+            The frequency shift to apply to the time embedding.
+        down_block_types (`tuple[str]`, defaults to `("CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D")`):
+            The tuple of downsample blocks to use.
+        only_cross_attention (`Union[bool, Tuple[bool]]`, defaults to `False`):
+        block_out_channels (`tuple[int]`, defaults to `(320, 640, 1280, 1280)`):
+            The tuple of output channels for each block.
+        layers_per_block (`int`, defaults to 2):
+            The number of layers per block.
+        downsample_padding (`int`, defaults to 1):
+            The padding to use for the downsampling convolution.
+        mid_block_scale_factor (`float`, defaults to 1):
+            The scale factor to use for the mid block.
+        act_fn (`str`, defaults to "silu"):
+            The activation function to use.
+        norm_num_groups (`int`, *optional*, defaults to 32):
+            The number of groups to use for the normalization. If None, normalization and activation layers is skipped
+            in post-processing.
+        norm_eps (`float`, defaults to 1e-5):
+            The epsilon to use for the normalization.
+        cross_attention_dim (`int`, defaults to 1280):
+            The dimension of the cross attention features.
+        transformer_layers_per_block (`int` or `Tuple[int]`, *optional*, defaults to 1):
+            The number of transformer blocks of type [`~models.attention.BasicTransformerBlock`]. Only relevant for
+            [`~models.unet_2d_blocks.CrossAttnDownBlock2D`], [`~models.unet_2d_blocks.CrossAttnUpBlock2D`],
+            [`~models.unet_2d_blocks.UNetMidBlock2DCrossAttn`].
+        encoder_hid_dim (`int`, *optional*, defaults to None):
+            If `encoder_hid_dim_type` is defined, `encoder_hidden_states` will be projected from `encoder_hid_dim`
+            dimension to `cross_attention_dim`.
+        encoder_hid_dim_type (`str`, *optional*, defaults to `None`):
+            If given, the `encoder_hidden_states` and potentially other embeddings are down-projected to text
+            embeddings of dimension `cross_attention` according to `encoder_hid_dim_type`.
+        attention_head_dim (`Union[int, Tuple[int]]`, defaults to 8):
+            The dimension of the attention heads.
+        use_linear_projection (`bool`, defaults to `False`):
+        class_embed_type (`str`, *optional*, defaults to `None`):
+            The type of class embedding to use which is ultimately summed with the time embeddings. Choose from None,
+            `"timestep"`, `"identity"`, `"projection"`, or `"simple_projection"`.
+        addition_embed_type (`str`, *optional*, defaults to `None`):
+            Configures an optional embedding which will be summed with the time embeddings. Choose from `None` or
+            "text". "text" will use the `TextTimeEmbedding` layer.
+        num_class_embeds (`int`, *optional*, defaults to 0):
+            Input dimension of the learnable embedding matrix to be projected to `time_embed_dim`, when performing
+            class conditioning with `class_embed_type` equal to `None`.
+        upcast_attention (`bool`, defaults to `False`):
+        resnet_time_scale_shift (`str`, defaults to `"default"`):
+            Time scale shift config for ResNet blocks (see `ResnetBlock2D`). Choose from `default` or `scale_shift`.
+        projection_class_embeddings_input_dim (`int`, *optional*, defaults to `None`):
+            The dimension of the `class_labels` input when `class_embed_type="projection"`. Required when
+            `class_embed_type="projection"`.
+        controlnet_conditioning_channel_order (`str`, defaults to `"rgb"`):
+            The channel order of conditional image. Will convert to `rgb` if it's `bgr`.
+        conditioning_embedding_out_channels (`tuple[int]`, *optional*, defaults to `(16, 32, 96, 256)`):
+            The tuple of output channel for each block in the `conditioning_embedding` layer.
+        global_pool_conditions (`bool`, defaults to `False`):
+            TODO(Patrick) - unused parameter.
+        addition_embed_type_num_heads (`int`, defaults to 64):
+            The number of heads to use for the `TextTimeEmbedding` layer.
+    """
+
+    _supports_gradient_checkpointing = True
+
+    @register_to_config
+    def __init__(
+        self,
+        in_channels: int = 4,
+        conditioning_channels: int = 3,
+        flip_sin_to_cos: bool = True,
+        freq_shift: int = 0,
+        down_block_types: Tuple[str, ...] = (
+            "CrossAttnDownBlock2D",
+            "CrossAttnDownBlock2D",
+            "CrossAttnDownBlock2D",
+            "DownBlock2D",
+        ),
+        mid_block_type: Optional[str] = "UNetMidBlock2DCrossAttn",
+        only_cross_attention: Union[bool, Tuple[bool]] = False,
+        block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
+        layers_per_block: int = 2,
+        downsample_padding: int = 1,
+        mid_block_scale_factor: float = 1,
+        act_fn: str = "silu",
+        norm_num_groups: Optional[int] = 32,
+        norm_eps: float = 1e-5,
+        cross_attention_dim: int = 1280,
+        transformer_layers_per_block: Union[int, Tuple[int, ...]] = 1,
+        encoder_hid_dim: Optional[int] = None,
+        encoder_hid_dim_type: Optional[str] = None,
+        attention_head_dim: Union[int, Tuple[int, ...]] = 8,
+        num_attention_heads: Optional[Union[int, Tuple[int, ...]]] = None,
+        use_linear_projection: bool = False,
+        class_embed_type: Optional[str] = None,
+        addition_embed_type: Optional[str] = None,
+        addition_time_embed_dim: Optional[int] = None,
+        num_class_embeds: Optional[int] = None,
+        upcast_attention: bool = False,
+        resnet_time_scale_shift: str = "default",
+        projection_class_embeddings_input_dim: Optional[int] = None,
+        controlnet_conditioning_channel_order: str = "rgb",
+        conditioning_embedding_out_channels: Optional[Tuple[int, ...]] = (16, 32, 96, 256),
+        global_pool_conditions: bool = False,
+        addition_embed_type_num_heads: int = 64,
+    ):
+        super().__init__(
+            in_channels,
+            conditioning_channels,
+            flip_sin_to_cos,
+            freq_shift,
+            down_block_types,
+            mid_block_type,
+            only_cross_attention,
+            block_out_channels,
+            layers_per_block,
+            downsample_padding,
+            mid_block_scale_factor,
+            act_fn,
+            norm_num_groups,
+            norm_eps,
+            cross_attention_dim,
+            transformer_layers_per_block,
+            encoder_hid_dim,
+            encoder_hid_dim_type,
+            attention_head_dim,
+            num_attention_heads,
+            use_linear_projection,
+            class_embed_type,
+            addition_embed_type,
+            addition_time_embed_dim,
+            num_class_embeds,
+            upcast_attention,
+            resnet_time_scale_shift,
+            projection_class_embeddings_input_dim,
+            controlnet_conditioning_channel_order,
+            conditioning_embedding_out_channels,
+            global_pool_conditions,
+            addition_embed_type_num_heads,
+        )
+        self.controlnet_query_cond_embedding = ControlNetConditioningEmbedding(
+            conditioning_embedding_channels=block_out_channels[0],
+            block_out_channels=conditioning_embedding_out_channels,
+            conditioning_channels=3,
+        )
+
+    def forward(
+        self,
+        sample: torch.FloatTensor,
+        timestep: Union[torch.Tensor, float, int],
+        encoder_hidden_states: torch.Tensor,
+        controlnet_cond: torch.FloatTensor,
+        controlnet_query_cond: torch.FloatTensor,
+        conditioning_scale: float = 1.0,
+        class_labels: Optional[torch.Tensor] = None,
+        timestep_cond: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        added_cond_kwargs: Optional[Dict[str, torch.Tensor]] = None,
+        cross_attention_kwargs: Optional[Dict[str, Any]] = None,
+        guess_mode: bool = False,
+        return_dict: bool = True,
+    ) -> Union[ControlNetOutput, Tuple[Tuple[torch.FloatTensor, ...], torch.FloatTensor]]:
+        """
+        The [`~PromptDiffusionControlNetModel`] forward method.
+
+        Args:
+            sample (`torch.FloatTensor`):
+                The noisy input tensor.
+            timestep (`Union[torch.Tensor, float, int]`):
+                The number of timesteps to denoise an input.
+            encoder_hidden_states (`torch.Tensor`):
+                The encoder hidden states.
+            controlnet_cond (`torch.FloatTensor`):
+                The conditional input tensor of shape `(batch_size, sequence_length, hidden_size)`.
+            controlnet_query_cond (`torch.FloatTensor`):
+                The conditional input tensor of shape `(batch_size, sequence_length, hidden_size)`.
+            conditioning_scale (`float`, defaults to `1.0`):
+                The scale factor for ControlNet outputs.
+            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`):
+                Additional conditional embeddings for timestep. If provided, the embeddings will be summed with the
+                timestep_embedding passed through the `self.time_embedding` layer to obtain the final 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.
+            added_cond_kwargs (`dict`):
+                Additional conditions for the Stable Diffusion XL UNet.
+            cross_attention_kwargs (`dict[str]`, *optional*, defaults to `None`):
+                A kwargs dictionary that if specified is passed along to the `AttnProcessor`.
+            guess_mode (`bool`, defaults to `False`):
+                In this mode, the ControlNet encoder tries its best to recognize the input content of the input even if
+                you remove all prompts. A `guidance_scale` between 3.0 and 5.0 is recommended.
+            return_dict (`bool`, defaults to `True`):
+                Whether or not to return a [`~models.controlnet.ControlNetOutput`] instead of a plain tuple.
+
+        Returns:
+            [`~models.controlnet.ControlNetOutput`] **or** `tuple`:
+                If `return_dict` is `True`, a [`~models.controlnet.ControlNetOutput`] is returned, otherwise a tuple is
+                returned where the first element is the sample tensor.
+        """
+        # check channel order
+        channel_order = self.config.controlnet_conditioning_channel_order
+
+        if channel_order == "rgb":
+            # in rgb order by default
+            ...
+        elif channel_order == "bgr":
+            controlnet_cond = torch.flip(controlnet_cond, dims=[1])
+        else:
+            raise ValueError(f"unknown `controlnet_conditioning_channel_order`: {channel_order}")
+
+        # prepare attention_mask
+        if attention_mask is not None:
+            attention_mask = (1 - attention_mask.to(sample.dtype)) * -10000.0
+            attention_mask = attention_mask.unsqueeze(1)
+
+        # 1. time
+        timesteps = timestep
+        if not torch.is_tensor(timesteps):
+            # TODO: this requires sync between CPU and GPU. So try to pass timesteps as tensors if you can
+            # This would be a good case for the `match` statement (Python 3.10+)
+            is_mps = sample.device.type == "mps"
+            if isinstance(timestep, float):
+                dtype = torch.float32 if is_mps else torch.float64
+            else:
+                dtype = torch.int32 if is_mps else torch.int64
+            timesteps = torch.tensor([timesteps], dtype=dtype, device=sample.device)
+        elif len(timesteps.shape) == 0:
+            timesteps = timesteps[None].to(sample.device)
+
+        # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
+        timesteps = timesteps.expand(sample.shape[0])
+
+        t_emb = self.time_proj(timesteps)
+
+        # timesteps does not contain any weights and will always return f32 tensors
+        # but time_embedding might actually be running in fp16. so we need to cast here.
+        # there might be better ways to encapsulate this.
+        t_emb = t_emb.to(dtype=sample.dtype)
+
+        emb = self.time_embedding(t_emb, timestep_cond)
+        aug_emb = None
+
+        if self.class_embedding is not None:
+            if class_labels is None:
+                raise ValueError("class_labels should be provided when num_class_embeds > 0")
+
+            if self.config.class_embed_type == "timestep":
+                class_labels = self.time_proj(class_labels)
+
+            class_emb = self.class_embedding(class_labels).to(dtype=self.dtype)
+            emb = emb + class_emb
+
+        if self.config.addition_embed_type is not None:
+            if self.config.addition_embed_type == "text":
+                aug_emb = self.add_embedding(encoder_hidden_states)
+
+            elif self.config.addition_embed_type == "text_time":
+                if "text_embeds" not in added_cond_kwargs:
+                    raise ValueError(
+                        f"{self.__class__} has the config param `addition_embed_type` set to 'text_time' which requires the keyword argument `text_embeds` to be passed in `added_cond_kwargs`"
+                    )
+                text_embeds = added_cond_kwargs.get("text_embeds")
+                if "time_ids" not in added_cond_kwargs:
+                    raise ValueError(
+                        f"{self.__class__} has the config param `addition_embed_type` set to 'text_time' which requires the keyword argument `time_ids` to be passed in `added_cond_kwargs`"
+                    )
+                time_ids = added_cond_kwargs.get("time_ids")
+                time_embeds = self.add_time_proj(time_ids.flatten())
+                time_embeds = time_embeds.reshape((text_embeds.shape[0], -1))
+
+                add_embeds = torch.concat([text_embeds, time_embeds], dim=-1)
+                add_embeds = add_embeds.to(emb.dtype)
+                aug_emb = self.add_embedding(add_embeds)
+
+        emb = emb + aug_emb if aug_emb is not None else emb
+
+        # 2. pre-process
+        sample = self.conv_in(sample)
+
+        controlnet_cond = self.controlnet_cond_embedding(controlnet_cond)
+        controlnet_query_cond = self.controlnet_query_cond_embedding(controlnet_query_cond)
+        sample = sample + controlnet_cond + controlnet_query_cond
+
+        # 3. down
+        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 = downsample_block(
+                    hidden_states=sample,
+                    temb=emb,
+                    encoder_hidden_states=encoder_hidden_states,
+                    attention_mask=attention_mask,
+                    cross_attention_kwargs=cross_attention_kwargs,
+                )
+            else:
+                sample, res_samples = downsample_block(hidden_states=sample, temb=emb)
+
+            down_block_res_samples += 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,
+                )
+            else:
+                sample = self.mid_block(sample, emb)
+
+        # 5. Control net blocks
+
+        controlnet_down_block_res_samples = ()
+
+        for down_block_res_sample, controlnet_block in zip(down_block_res_samples, self.controlnet_down_blocks):
+            down_block_res_sample = controlnet_block(down_block_res_sample)
+            controlnet_down_block_res_samples = controlnet_down_block_res_samples + (down_block_res_sample,)
+
+        down_block_res_samples = controlnet_down_block_res_samples
+
+        mid_block_res_sample = self.controlnet_mid_block(sample)
+
+        # 6. scaling
+        if guess_mode and not self.config.global_pool_conditions:
+            scales = torch.logspace(-1, 0, len(down_block_res_samples) + 1, device=sample.device)  # 0.1 to 1.0
+            scales = scales * conditioning_scale
+            down_block_res_samples = [sample * scale for sample, scale in zip(down_block_res_samples, scales)]
+            mid_block_res_sample = mid_block_res_sample * scales[-1]  # last one
+        else:
+            down_block_res_samples = [sample * conditioning_scale for sample in down_block_res_samples]
+            mid_block_res_sample = mid_block_res_sample * conditioning_scale
+
+        if self.config.global_pool_conditions:
+            down_block_res_samples = [
+                torch.mean(sample, dim=(2, 3), keepdim=True) for sample in down_block_res_samples
+            ]
+            mid_block_res_sample = torch.mean(mid_block_res_sample, dim=(2, 3), keepdim=True)
+
+        if not return_dict:
+            return (down_block_res_samples, mid_block_res_sample)
+
+        return ControlNetOutput(
+            down_block_res_samples=down_block_res_samples, mid_block_res_sample=mid_block_res_sample
+        )

examples/research_projects/realfill/train_realfill.py

@@ -87,6 +87,7 @@ tags:
 - text-to-image
 - diffusers
 - realfill
+- diffusers-training
 inference: true
 ---
     """
@@ -628,7 +629,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/t2i_adapter/train_t2i_adapter_sdxl.py

@@ -60,7 +60,7 @@ if is_wandb_available():
     import wandb
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -167,7 +167,7 @@ def log_validation(vae, unet, adapter, args, accelerator, weight_dtype, step):
 
             tracker.log({"validation": formatted_images})
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
         del pipeline
         gc.collect()
@@ -225,7 +225,14 @@ These are t2iadapter weights trained on {base_model} with new type of conditioni
         inference=True,
     )
 
-    tags = ["stable-diffusion-xl", "stable-diffusion-xl-diffusers", "text-to-image", "diffusers", "t2iadapter"]
+    tags = [
+        "stable-diffusion-xl",
+        "stable-diffusion-xl-diffusers",
+        "text-to-image",
+        "diffusers",
+        "t2iadapter",
+        "diffusers-training",
+    ]
     model_card = populate_model_card(model_card, tags=tags)
 
     model_card.save(os.path.join(repo_folder, "README.md"))
@@ -925,7 +932,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/text_to_image/train_text_to_image.py

@@ -56,7 +56,7 @@ if is_wandb_available():
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -131,7 +131,7 @@ More information on all the CLI arguments and the environment are available on y
         inference=True,
     )
 
-    tags = ["stable-diffusion", "stable-diffusion-diffusers", "text-to-image", "diffusers"]
+    tags = ["stable-diffusion", "stable-diffusion-diffusers", "text-to-image", "diffusers", "diffusers-training"]
     model_card = populate_model_card(model_card, tags=tags)
 
     model_card.save(os.path.join(repo_folder, "README.md"))
@@ -183,7 +183,7 @@ def log_validation(vae, text_encoder, tokenizer, unet, args, accelerator, weight
                 }
             )
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
     del pipeline
     torch.cuda.empty_cache()
@@ -608,7 +608,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/text_to_image/train_text_to_image_flax.py

@@ -49,7 +49,7 @@ from diffusers.utils import check_min_version
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = logging.getLogger(__name__)
 

examples/text_to_image/train_text_to_image_lora.py

@@ -52,7 +52,7 @@ from diffusers.utils.torch_utils import is_compiled_module
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -90,6 +90,7 @@ These are LoRA adaption weights for {base_model}. The weights were fine-tuned on
         "stable-diffusion-diffusers",
         "text-to-image",
         "diffusers",
+        "diffusers-training",
         "lora",
     ]
     model_card = populate_model_card(model_card, tags=tags)
@@ -496,7 +497,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/text_to_image/train_text_to_image_lora_sdxl.py

@@ -64,7 +64,7 @@ from diffusers.utils.torch_utils import is_compiled_module
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -103,7 +103,14 @@ Special VAE used for training: {vae_path}.
         inference=True,
     )
 
-    tags = ["stable-diffusion-xl", "stable-diffusion-xl-diffusers", "text-to-image", "diffusers", "lora"]
+    tags = [
+        "stable-diffusion-xl",
+        "stable-diffusion-xl-diffusers",
+        "text-to-image",
+        "diffusers",
+        "diffusers-training",
+        "lora",
+    ]
     model_card = populate_model_card(model_card, tags=tags)
 
     model_card.save(os.path.join(repo_folder, "README.md"))
@@ -418,6 +425,11 @@ def parse_args(input_args=None):
         default=4,
         help=("The dimension of the LoRA update matrices."),
     )
+    parser.add_argument(
+        "--debug_loss",
+        action="store_true",
+        help="debug loss for each image, if filenames are awailable in the dataset",
+    )
 
     if input_args is not None:
         args = parser.parse_args(input_args)
@@ -596,6 +608,7 @@ def main(args):
     # Move unet, vae and text_encoder to device and cast to weight_dtype
     # The VAE is in float32 to avoid NaN losses.
     unet.to(accelerator.device, dtype=weight_dtype)
+
     if args.pretrained_vae_model_name_or_path is None:
         vae.to(accelerator.device, dtype=torch.float32)
     else:
@@ -609,7 +622,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()
@@ -883,13 +896,17 @@ def main(args):
         tokens_one, tokens_two = tokenize_captions(examples)
         examples["input_ids_one"] = tokens_one
         examples["input_ids_two"] = tokens_two
+        if args.debug_loss:
+            fnames = [os.path.basename(image.filename) for image in examples[image_column] if image.filename]
+            if fnames:
+                examples["filenames"] = fnames
         return examples
 
     with accelerator.main_process_first():
         if args.max_train_samples is not None:
             dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
         # Set the training transforms
-        train_dataset = dataset["train"].with_transform(preprocess_train)
+        train_dataset = dataset["train"].with_transform(preprocess_train, output_all_columns=True)
 
     def collate_fn(examples):
         pixel_values = torch.stack([example["pixel_values"] for example in examples])
@@ -898,7 +915,7 @@ def main(args):
         crop_top_lefts = [example["crop_top_lefts"] for example in examples]
         input_ids_one = torch.stack([example["input_ids_one"] for example in examples])
         input_ids_two = torch.stack([example["input_ids_two"] for example in examples])
-        return {
+        result = {
             "pixel_values": pixel_values,
             "input_ids_one": input_ids_one,
             "input_ids_two": input_ids_two,
@@ -906,6 +923,11 @@ def main(args):
             "crop_top_lefts": crop_top_lefts,
         }
 
+        filenames = [example["filenames"] for example in examples if "filenames" in example]
+        if filenames:
+            result["filenames"] = filenames
+        return result
+
     # DataLoaders creation:
     train_dataloader = torch.utils.data.DataLoader(
         train_dataset,
@@ -1098,7 +1120,9 @@ def main(args):
                     loss = F.mse_loss(model_pred.float(), target.float(), reduction="none")
                     loss = loss.mean(dim=list(range(1, len(loss.shape)))) * mse_loss_weights
                     loss = loss.mean()
-
+                if args.debug_loss and "filenames" in batch:
+                    for fname in batch["filenames"]:
+                        accelerator.log({"loss_for_" + fname: loss}, step=global_step)
                 # Gather the losses across all processes for logging (if we use distributed training).
                 avg_loss = accelerator.gather(loss.repeat(args.train_batch_size)).mean()
                 train_loss += avg_loss.item() / args.gradient_accumulation_steps

examples/text_to_image/train_text_to_image_sdxl.py

@@ -54,7 +54,7 @@ from diffusers.utils.torch_utils import is_compiled_module
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -101,6 +101,7 @@ Special VAE used for training: {vae_path}.
         "stable-diffusion-xl",
         "stable-diffusion-xl-diffusers",
         "text-to-image",
+        "diffusers-training",
         "diffusers",
     ]
     model_card = populate_model_card(model_card, tags=tags)
@@ -711,7 +712,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()

examples/textual_inversion/textual_inversion.py

@@ -80,7 +80,7 @@ else:
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -105,7 +105,14 @@ These are textual inversion adaption weights for {base_model}. You can find some
         inference=True,
     )
 
-    tags = ["stable-diffusion", "stable-diffusion-diffusers", "text-to-image", "diffusers", "textual_inversion"]
+    tags = [
+        "stable-diffusion",
+        "stable-diffusion-diffusers",
+        "text-to-image",
+        "diffusers",
+        "textual_inversion",
+        "diffusers-training",
+    ]
     model_card = populate_model_card(model_card, tags=tags)
 
     model_card.save(os.path.join(repo_folder, "README.md"))
@@ -701,7 +708,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()
@@ -959,7 +966,7 @@ def main():
     accelerator.wait_for_everyone()
     if accelerator.is_main_process:
         if args.push_to_hub and not args.save_as_full_pipeline:
-            logger.warn("Enabling full model saving because --push_to_hub=True was specified.")
+            logger.warning("Enabling full model saving because --push_to_hub=True was specified.")
             save_full_model = True
         else:
             save_full_model = args.save_as_full_pipeline

examples/textual_inversion/textual_inversion_flax.py

@@ -56,7 +56,7 @@ else:
 # ------------------------------------------------------------------------------
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = logging.getLogger(__name__)
 

examples/textual_inversion/textual_inversion_sdxl.py

@@ -76,7 +76,7 @@ else:
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -106,6 +106,7 @@ These are textual inversion adaption weights for {base_model}. You can find some
         "stable-diffusion-xl-diffusers",
         "text-to-image",
         "diffusers",
+        "diffusers-training",
         "textual_inversion",
     ]
 
@@ -710,7 +711,7 @@ def main():
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             unet.enable_xformers_memory_efficient_attention()
@@ -1021,7 +1022,7 @@ def main():
             )
 
         if args.push_to_hub and not args.save_as_full_pipeline:
-            logger.warn("Enabling full model saving because --push_to_hub=True was specified.")
+            logger.warning("Enabling full model saving because --push_to_hub=True was specified.")
             save_full_model = True
         else:
             save_full_model = args.save_as_full_pipeline

examples/unconditional_image_generation/train_unconditional.py

@@ -29,7 +29,7 @@ from diffusers.utils.import_utils import is_xformers_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -408,7 +408,7 @@ def main(args):
 
             xformers_version = version.parse(xformers.__version__)
             if xformers_version == version.parse("0.0.16"):
-                logger.warn(
+                logger.warning(
                     "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                 )
             model.enable_xformers_memory_efficient_attention()
@@ -648,7 +648,7 @@ def main(args):
                     generator=generator,
                     batch_size=args.eval_batch_size,
                     num_inference_steps=args.ddpm_num_inference_steps,
-                    output_type="numpy",
+                    output_type="np",
                 ).images
 
                 if args.use_ema:

examples/wuerstchen/text_to_image/train_text_to_image_lora_prior.py

@@ -50,7 +50,7 @@ if is_wandb_available():
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -81,6 +81,7 @@ tags:
 - wuerstchen
 - text-to-image
 - diffusers
+- diffusers-training
 - lora
 inference: true
 ---
@@ -183,7 +184,7 @@ def log_validation(text_encoder, tokenizer, prior, args, accelerator, weight_dty
                 }
             )
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
     del pipeline
     torch.cuda.empty_cache()

examples/wuerstchen/text_to_image/train_text_to_image_prior.py

@@ -51,7 +51,7 @@ if is_wandb_available():
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.27.0.dev0")
+check_min_version("0.28.0.dev0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -82,6 +82,7 @@ tags:
 - wuerstchen
 - text-to-image
 - diffusers
+- diffusers-training
 inference: true
 ---
     """
@@ -181,7 +182,7 @@ def log_validation(text_encoder, tokenizer, prior, args, accelerator, weight_dty
                 }
             )
         else:
-            logger.warn(f"image logging not implemented for {tracker.name}")
+            logger.warning(f"image logging not implemented for {tracker.name}")
 
     del pipeline
     torch.cuda.empty_cache()

scripts/convert_stable_cascade.py

@@ -0,0 +1,218 @@
+# Run this script to convert the Stable Cascade model weights to a diffusers pipeline.
+import argparse
+from contextlib import nullcontext
+
+import torch
+from safetensors.torch import load_file
+from transformers import (
+    AutoTokenizer,
+    CLIPConfig,
+    CLIPImageProcessor,
+    CLIPTextModelWithProjection,
+    CLIPVisionModelWithProjection,
+)
+
+from diffusers import (
+    DDPMWuerstchenScheduler,
+    StableCascadeCombinedPipeline,
+    StableCascadeDecoderPipeline,
+    StableCascadePriorPipeline,
+)
+from diffusers.loaders.single_file_utils import convert_stable_cascade_unet_single_file_to_diffusers
+from diffusers.models import StableCascadeUNet
+from diffusers.models.modeling_utils import load_model_dict_into_meta
+from diffusers.pipelines.wuerstchen import PaellaVQModel
+from diffusers.utils import is_accelerate_available
+
+
+if is_accelerate_available():
+    from accelerate import init_empty_weights
+
+parser = argparse.ArgumentParser(description="Convert Stable Cascade model weights to a diffusers pipeline")
+parser.add_argument("--model_path", type=str, help="Location of Stable Cascade weights")
+parser.add_argument("--stage_c_name", type=str, default="stage_c.safetensors", help="Name of stage c checkpoint file")
+parser.add_argument("--stage_b_name", type=str, default="stage_b.safetensors", help="Name of stage b checkpoint file")
+parser.add_argument("--skip_stage_c", action="store_true", help="Skip converting stage c")
+parser.add_argument("--skip_stage_b", action="store_true", help="Skip converting stage b")
+parser.add_argument("--use_safetensors", action="store_true", help="Use SafeTensors for conversion")
+parser.add_argument(
+    "--prior_output_path", default="stable-cascade-prior", type=str, help="Hub organization to save the pipelines to"
+)
+parser.add_argument(
+    "--decoder_output_path",
+    type=str,
+    default="stable-cascade-decoder",
+    help="Hub organization to save the pipelines to",
+)
+parser.add_argument(
+    "--combined_output_path",
+    type=str,
+    default="stable-cascade-combined",
+    help="Hub organization to save the pipelines to",
+)
+parser.add_argument("--save_combined", action="store_true")
+parser.add_argument("--push_to_hub", action="store_true", help="Push to hub")
+parser.add_argument("--variant", type=str, help="Set to bf16 to save bfloat16 weights")
+
+args = parser.parse_args()
+
+if args.skip_stage_b and args.skip_stage_c:
+    raise ValueError("At least one stage should be converted")
+if (args.skip_stage_b or args.skip_stage_c) and args.save_combined:
+    raise ValueError("Cannot skip stages when creating a combined pipeline")
+
+model_path = args.model_path
+
+device = "cpu"
+if args.variant == "bf16":
+    dtype = torch.bfloat16
+else:
+    dtype = torch.float32
+
+# set paths to model weights
+prior_checkpoint_path = f"{model_path}/{args.stage_c_name}"
+decoder_checkpoint_path = f"{model_path}/{args.stage_b_name}"
+
+# Clip Text encoder and tokenizer
+config = CLIPConfig.from_pretrained("laion/CLIP-ViT-bigG-14-laion2B-39B-b160k")
+config.text_config.projection_dim = config.projection_dim
+text_encoder = CLIPTextModelWithProjection.from_pretrained(
+    "laion/CLIP-ViT-bigG-14-laion2B-39B-b160k", config=config.text_config
+)
+tokenizer = AutoTokenizer.from_pretrained("laion/CLIP-ViT-bigG-14-laion2B-39B-b160k")
+
+# image processor
+feature_extractor = CLIPImageProcessor()
+image_encoder = CLIPVisionModelWithProjection.from_pretrained("openai/clip-vit-large-patch14")
+
+# scheduler for prior and decoder
+scheduler = DDPMWuerstchenScheduler()
+ctx = init_empty_weights if is_accelerate_available() else nullcontext
+
+if not args.skip_stage_c:
+    # Prior
+    if args.use_safetensors:
+        prior_orig_state_dict = load_file(prior_checkpoint_path, device=device)
+    else:
+        prior_orig_state_dict = torch.load(prior_checkpoint_path, map_location=device)
+
+    prior_state_dict = convert_stable_cascade_unet_single_file_to_diffusers(prior_orig_state_dict)
+
+    with ctx():
+        prior_model = StableCascadeUNet(
+            in_channels=16,
+            out_channels=16,
+            timestep_ratio_embedding_dim=64,
+            patch_size=1,
+            conditioning_dim=2048,
+            block_out_channels=[2048, 2048],
+            num_attention_heads=[32, 32],
+            down_num_layers_per_block=[8, 24],
+            up_num_layers_per_block=[24, 8],
+            down_blocks_repeat_mappers=[1, 1],
+            up_blocks_repeat_mappers=[1, 1],
+            block_types_per_layer=[
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock", "SDCascadeAttnBlock"],
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock", "SDCascadeAttnBlock"],
+            ],
+            clip_text_in_channels=1280,
+            clip_text_pooled_in_channels=1280,
+            clip_image_in_channels=768,
+            clip_seq=4,
+            kernel_size=3,
+            dropout=[0.1, 0.1],
+            self_attn=True,
+            timestep_conditioning_type=["sca", "crp"],
+            switch_level=[False],
+        )
+    if is_accelerate_available():
+        load_model_dict_into_meta(prior_model, prior_state_dict)
+    else:
+        prior_model.load_state_dict(prior_state_dict)
+
+    # Prior pipeline
+    prior_pipeline = StableCascadePriorPipeline(
+        prior=prior_model,
+        tokenizer=tokenizer,
+        text_encoder=text_encoder,
+        image_encoder=image_encoder,
+        scheduler=scheduler,
+        feature_extractor=feature_extractor,
+    )
+    prior_pipeline.to(dtype).save_pretrained(
+        args.prior_output_path, push_to_hub=args.push_to_hub, variant=args.variant
+    )
+
+if not args.skip_stage_b:
+    # Decoder
+    if args.use_safetensors:
+        decoder_orig_state_dict = load_file(decoder_checkpoint_path, device=device)
+    else:
+        decoder_orig_state_dict = torch.load(decoder_checkpoint_path, map_location=device)
+
+    decoder_state_dict = convert_stable_cascade_unet_single_file_to_diffusers(decoder_orig_state_dict)
+    with ctx():
+        decoder = StableCascadeUNet(
+            in_channels=4,
+            out_channels=4,
+            timestep_ratio_embedding_dim=64,
+            patch_size=2,
+            conditioning_dim=1280,
+            block_out_channels=[320, 640, 1280, 1280],
+            down_num_layers_per_block=[2, 6, 28, 6],
+            up_num_layers_per_block=[6, 28, 6, 2],
+            down_blocks_repeat_mappers=[1, 1, 1, 1],
+            up_blocks_repeat_mappers=[3, 3, 2, 2],
+            num_attention_heads=[0, 0, 20, 20],
+            block_types_per_layer=[
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock"],
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock"],
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock", "SDCascadeAttnBlock"],
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock", "SDCascadeAttnBlock"],
+            ],
+            clip_text_pooled_in_channels=1280,
+            clip_seq=4,
+            effnet_in_channels=16,
+            pixel_mapper_in_channels=3,
+            kernel_size=3,
+            dropout=[0, 0, 0.1, 0.1],
+            self_attn=True,
+            timestep_conditioning_type=["sca"],
+        )
+
+    if is_accelerate_available():
+        load_model_dict_into_meta(decoder, decoder_state_dict)
+    else:
+        decoder.load_state_dict(decoder_state_dict)
+
+    # VQGAN from Wuerstchen-V2
+    vqmodel = PaellaVQModel.from_pretrained("warp-ai/wuerstchen", subfolder="vqgan")
+
+    # Decoder pipeline
+    decoder_pipeline = StableCascadeDecoderPipeline(
+        decoder=decoder, text_encoder=text_encoder, tokenizer=tokenizer, vqgan=vqmodel, scheduler=scheduler
+    )
+    decoder_pipeline.to(dtype).save_pretrained(
+        args.decoder_output_path, push_to_hub=args.push_to_hub, variant=args.variant
+    )
+
+if args.save_combined:
+    # Stable Cascade combined pipeline
+    stable_cascade_pipeline = StableCascadeCombinedPipeline(
+        # Decoder
+        text_encoder=text_encoder,
+        tokenizer=tokenizer,
+        decoder=decoder,
+        scheduler=scheduler,
+        vqgan=vqmodel,
+        # Prior
+        prior_text_encoder=text_encoder,
+        prior_tokenizer=tokenizer,
+        prior_prior=prior_model,
+        prior_scheduler=scheduler,
+        prior_image_encoder=image_encoder,
+        prior_feature_extractor=feature_extractor,
+    )
+    stable_cascade_pipeline.to(dtype).save_pretrained(
+        args.combined_output_path, push_to_hub=args.push_to_hub, variant=args.variant
+    )

scripts/convert_stable_cascade_lite.py

@@ -0,0 +1,226 @@
+# Run this script to convert the Stable Cascade model weights to a diffusers pipeline.
+import argparse
+from contextlib import nullcontext
+
+import torch
+from safetensors.torch import load_file
+from transformers import (
+    AutoTokenizer,
+    CLIPConfig,
+    CLIPImageProcessor,
+    CLIPTextModelWithProjection,
+    CLIPVisionModelWithProjection,
+)
+
+from diffusers import (
+    DDPMWuerstchenScheduler,
+    StableCascadeCombinedPipeline,
+    StableCascadeDecoderPipeline,
+    StableCascadePriorPipeline,
+)
+from diffusers.loaders.single_file_utils import convert_stable_cascade_unet_single_file_to_diffusers
+from diffusers.models import StableCascadeUNet
+from diffusers.models.modeling_utils import load_model_dict_into_meta
+from diffusers.pipelines.wuerstchen import PaellaVQModel
+from diffusers.utils import is_accelerate_available
+
+
+if is_accelerate_available():
+    from accelerate import init_empty_weights
+
+parser = argparse.ArgumentParser(description="Convert Stable Cascade model weights to a diffusers pipeline")
+parser.add_argument("--model_path", type=str, help="Location of Stable Cascade weights")
+parser.add_argument(
+    "--stage_c_name", type=str, default="stage_c_lite.safetensors", help="Name of stage c checkpoint file"
+)
+parser.add_argument(
+    "--stage_b_name", type=str, default="stage_b_lite.safetensors", help="Name of stage b checkpoint file"
+)
+parser.add_argument("--skip_stage_c", action="store_true", help="Skip converting stage c")
+parser.add_argument("--skip_stage_b", action="store_true", help="Skip converting stage b")
+parser.add_argument("--use_safetensors", action="store_true", help="Use SafeTensors for conversion")
+parser.add_argument(
+    "--prior_output_path",
+    default="stable-cascade-prior-lite",
+    type=str,
+    help="Hub organization to save the pipelines to",
+)
+parser.add_argument(
+    "--decoder_output_path",
+    type=str,
+    default="stable-cascade-decoder-lite",
+    help="Hub organization to save the pipelines to",
+)
+parser.add_argument(
+    "--combined_output_path",
+    type=str,
+    default="stable-cascade-combined-lite",
+    help="Hub organization to save the pipelines to",
+)
+parser.add_argument("--save_combined", action="store_true")
+parser.add_argument("--push_to_hub", action="store_true", help="Push to hub")
+parser.add_argument("--variant", type=str, help="Set to bf16 to save bfloat16 weights")
+
+args = parser.parse_args()
+
+if args.skip_stage_b and args.skip_stage_c:
+    raise ValueError("At least one stage should be converted")
+if (args.skip_stage_b or args.skip_stage_c) and args.save_combined:
+    raise ValueError("Cannot skip stages when creating a combined pipeline")
+
+model_path = args.model_path
+
+device = "cpu"
+if args.variant == "bf16":
+    dtype = torch.bfloat16
+else:
+    dtype = torch.float32
+
+# set paths to model weights
+prior_checkpoint_path = f"{model_path}/{args.stage_c_name}"
+decoder_checkpoint_path = f"{model_path}/{args.stage_b_name}"
+
+# Clip Text encoder and tokenizer
+config = CLIPConfig.from_pretrained("laion/CLIP-ViT-bigG-14-laion2B-39B-b160k")
+config.text_config.projection_dim = config.projection_dim
+text_encoder = CLIPTextModelWithProjection.from_pretrained(
+    "laion/CLIP-ViT-bigG-14-laion2B-39B-b160k", config=config.text_config
+)
+tokenizer = AutoTokenizer.from_pretrained("laion/CLIP-ViT-bigG-14-laion2B-39B-b160k")
+
+# image processor
+feature_extractor = CLIPImageProcessor()
+image_encoder = CLIPVisionModelWithProjection.from_pretrained("openai/clip-vit-large-patch14")
+# scheduler for prior and decoder
+scheduler = DDPMWuerstchenScheduler()
+
+ctx = init_empty_weights if is_accelerate_available() else nullcontext
+
+if not args.skip_stage_c:
+    # Prior
+    if args.use_safetensors:
+        prior_orig_state_dict = load_file(prior_checkpoint_path, device=device)
+    else:
+        prior_orig_state_dict = torch.load(prior_checkpoint_path, map_location=device)
+
+    prior_state_dict = convert_stable_cascade_unet_single_file_to_diffusers(prior_orig_state_dict)
+    with ctx():
+        prior_model = StableCascadeUNet(
+            in_channels=16,
+            out_channels=16,
+            timestep_ratio_embedding_dim=64,
+            patch_size=1,
+            conditioning_dim=1536,
+            block_out_channels=[1536, 1536],
+            num_attention_heads=[24, 24],
+            down_num_layers_per_block=[4, 12],
+            up_num_layers_per_block=[12, 4],
+            down_blocks_repeat_mappers=[1, 1],
+            up_blocks_repeat_mappers=[1, 1],
+            block_types_per_layer=[
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock", "SDCascadeAttnBlock"],
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock", "SDCascadeAttnBlock"],
+            ],
+            clip_text_in_channels=1280,
+            clip_text_pooled_in_channels=1280,
+            clip_image_in_channels=768,
+            clip_seq=4,
+            kernel_size=3,
+            dropout=[0.1, 0.1],
+            self_attn=True,
+            timestep_conditioning_type=["sca", "crp"],
+            switch_level=[False],
+        )
+
+    if is_accelerate_available():
+        load_model_dict_into_meta(prior_model, prior_state_dict)
+    else:
+        prior_model.load_state_dict(prior_state_dict)
+
+    # Prior pipeline
+    prior_pipeline = StableCascadePriorPipeline(
+        prior=prior_model,
+        tokenizer=tokenizer,
+        text_encoder=text_encoder,
+        image_encoder=image_encoder,
+        scheduler=scheduler,
+        feature_extractor=feature_extractor,
+    )
+    prior_pipeline.to(dtype).save_pretrained(
+        args.prior_output_path, push_to_hub=args.push_to_hub, variant=args.variant
+    )
+
+if not args.skip_stage_b:
+    # Decoder
+    if args.use_safetensors:
+        decoder_orig_state_dict = load_file(decoder_checkpoint_path, device=device)
+    else:
+        decoder_orig_state_dict = torch.load(decoder_checkpoint_path, map_location=device)
+
+    decoder_state_dict = convert_stable_cascade_unet_single_file_to_diffusers(decoder_orig_state_dict)
+
+    with ctx():
+        decoder = StableCascadeUNet(
+            in_channels=4,
+            out_channels=4,
+            timestep_ratio_embedding_dim=64,
+            patch_size=2,
+            conditioning_dim=1280,
+            block_out_channels=[320, 576, 1152, 1152],
+            down_num_layers_per_block=[2, 4, 14, 4],
+            up_num_layers_per_block=[4, 14, 4, 2],
+            down_blocks_repeat_mappers=[1, 1, 1, 1],
+            up_blocks_repeat_mappers=[2, 2, 2, 2],
+            num_attention_heads=[0, 9, 18, 18],
+            block_types_per_layer=[
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock"],
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock"],
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock", "SDCascadeAttnBlock"],
+                ["SDCascadeResBlock", "SDCascadeTimestepBlock", "SDCascadeAttnBlock"],
+            ],
+            clip_text_pooled_in_channels=1280,
+            clip_seq=4,
+            effnet_in_channels=16,
+            pixel_mapper_in_channels=3,
+            kernel_size=3,
+            dropout=[0, 0, 0.1, 0.1],
+            self_attn=True,
+            timestep_conditioning_type=["sca"],
+        )
+
+    if is_accelerate_available():
+        load_model_dict_into_meta(decoder, decoder_state_dict)
+    else:
+        decoder.load_state_dict(decoder_state_dict)
+
+    # VQGAN from Wuerstchen-V2
+    vqmodel = PaellaVQModel.from_pretrained("warp-ai/wuerstchen", subfolder="vqgan")
+
+    # Decoder pipeline
+    decoder_pipeline = StableCascadeDecoderPipeline(
+        decoder=decoder, text_encoder=text_encoder, tokenizer=tokenizer, vqgan=vqmodel, scheduler=scheduler
+    )
+    decoder_pipeline.to(dtype).save_pretrained(
+        args.decoder_output_path, push_to_hub=args.push_to_hub, variant=args.variant
+    )
+
+if args.save_combined:
+    # Stable Cascade combined pipeline
+    stable_cascade_pipeline = StableCascadeCombinedPipeline(
+        # Decoder
+        text_encoder=text_encoder,
+        tokenizer=tokenizer,
+        decoder=decoder,
+        scheduler=scheduler,
+        vqgan=vqmodel,
+        # Prior
+        prior_text_encoder=text_encoder,
+        prior_tokenizer=tokenizer,
+        prior_prior=prior_model,
+        prior_scheduler=scheduler,
+        prior_image_encoder=image_encoder,
+        prior_feature_extractor=feature_extractor,
+    )
+    stable_cascade_pipeline.to(dtype).save_pretrained(
+        args.combined_output_path, push_to_hub=args.push_to_hub, variant=args.variant
+    )

scripts/log_reports.py

@@ -0,0 +1,139 @@
+import argparse
+import json
+import os
+from datetime import date
+from pathlib import Path
+
+from slack_sdk import WebClient
+from tabulate import tabulate
+
+
+MAX_LEN_MESSAGE = 2900  # slack endpoint has a limit of 3001 characters
+
+parser = argparse.ArgumentParser()
+parser.add_argument("--slack_channel_name", default="diffusers-ci-nightly")
+
+
+def main(slack_channel_name=None):
+    failed = []
+    passed = []
+
+    group_info = []
+
+    total_num_failed = 0
+    empty_file = False or len(list(Path().glob("*.log"))) == 0
+
+    total_empty_files = []
+
+    for log in Path().glob("*.log"):
+        section_num_failed = 0
+        i = 0
+        with open(log) as f:
+            for line in f:
+                line = json.loads(line)
+                i += 1
+                if line.get("nodeid", "") != "":
+                    test = line["nodeid"]
+                    if line.get("duration", None) is not None:
+                        duration = f'{line["duration"]:.4f}'
+                        if line.get("outcome", "") == "failed":
+                            section_num_failed += 1
+                            failed.append([test, duration, log.name.split("_")[0]])
+                            total_num_failed += 1
+                        else:
+                            passed.append([test, duration, log.name.split("_")[0]])
+            empty_file = i == 0
+        group_info.append([str(log), section_num_failed, failed])
+        total_empty_files.append(empty_file)
+        os.remove(log)
+        failed = []
+    text = (
+        "🌞 There were no failures!"
+        if not any(total_empty_files)
+        else "Something went wrong there is at least one empty file - please check GH action results."
+    )
+    no_error_payload = {
+        "type": "section",
+        "text": {
+            "type": "plain_text",
+            "text": text,
+            "emoji": True,
+        },
+    }
+
+    message = ""
+    payload = [
+        {
+            "type": "header",
+            "text": {
+                "type": "plain_text",
+                "text": "🤗 Results of the Diffusers scheduled nightly tests.",
+            },
+        },
+    ]
+    if total_num_failed > 0:
+        for i, (name, num_failed, failed_tests) in enumerate(group_info):
+            if num_failed > 0:
+                if num_failed == 1:
+                    message += f"*{name}: {num_failed} failed test*\n"
+                else:
+                    message += f"*{name}: {num_failed} failed tests*\n"
+                failed_table = []
+                for test in failed_tests:
+                    failed_table.append(test[0].split("::"))
+                failed_table = tabulate(
+                    failed_table,
+                    headers=["Test Location", "Test Case", "Test Name"],
+                    showindex="always",
+                    tablefmt="grid",
+                    maxcolwidths=[12, 12, 12],
+                )
+                message += "\n```\n" + failed_table + "\n```"
+
+            if total_empty_files[i]:
+                message += f"\n*{name}: Warning! Empty file - please check the GitHub action job *\n"
+        print(f"### {message}")
+    else:
+        payload.append(no_error_payload)
+
+    if len(message) > MAX_LEN_MESSAGE:
+        print(f"Truncating long message from {len(message)} to {MAX_LEN_MESSAGE}")
+        message = message[:MAX_LEN_MESSAGE] + "..."
+
+    if len(message) != 0:
+        md_report = {
+            "type": "section",
+            "text": {"type": "mrkdwn", "text": message},
+        }
+        payload.append(md_report)
+        action_button = {
+            "type": "section",
+            "text": {"type": "mrkdwn", "text": "*For more details:*"},
+            "accessory": {
+                "type": "button",
+                "text": {"type": "plain_text", "text": "Check Action results", "emoji": True},
+                "url": f"https://github.com/huggingface/diffusers/actions/runs/{os.environ['GITHUB_RUN_ID']}",
+            },
+        }
+        payload.append(action_button)
+
+    date_report = {
+        "type": "context",
+        "elements": [
+            {
+                "type": "plain_text",
+                "text": f"Nightly test results for {date.today()}",
+            },
+        ],
+    }
+    payload.append(date_report)
+
+    print(payload)
+
+    client = WebClient(token=os.environ.get("SLACK_API_TOKEN"))
+    client.chat_postMessage(channel=f"#{slack_channel_name}", text=message, blocks=payload)
+
+
+if __name__ == "__main__":
+    args = parser.parse_args()
+    main(args.slack_channel_name)

setup.py

@@ -249,7 +249,7 @@ version_range_max = max(sys.version_info[1], 10) + 1
 
 setup(
     name="diffusers",
-    version="0.27.0.dev0",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
+    version="0.28.0.dev0",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
     description="State-of-the-art diffusion in PyTorch and JAX.",
     long_description=open("README.md", "r", encoding="utf-8").read(),
     long_description_content_type="text/markdown",

src/diffusers/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.27.0.dev0"
+__version__ = "0.28.0.dev0"
 
 from typing import TYPE_CHECKING
 
@@ -86,6 +86,7 @@ else:
             "MotionAdapter",
             "MultiAdapter",
             "PriorTransformer",
+            "StableCascadeUNet",
             "T2IAdapter",
             "T5FilmDecoder",
             "Transformer2DModel",
@@ -160,6 +161,7 @@ else:
             "SASolverScheduler",
             "SchedulerMixin",
             "ScoreSdeVeScheduler",
+            "TCDScheduler",
             "UnCLIPScheduler",
             "UniPCMultistepScheduler",
             "VQDiffusionScheduler",
@@ -251,6 +253,8 @@ else:
             "LatentConsistencyModelImg2ImgPipeline",
             "LatentConsistencyModelPipeline",
             "LDMTextToImagePipeline",
+            "LEditsPPPipelineStableDiffusion",
+            "LEditsPPPipelineStableDiffusionXL",
             "MusicLDMPipeline",
             "PaintByExamplePipeline",
             "PIAPipeline",
@@ -258,6 +262,9 @@ else:
             "SemanticStableDiffusionPipeline",
             "ShapEImg2ImgPipeline",
             "ShapEPipeline",
+            "StableCascadeCombinedPipeline",
+            "StableCascadeDecoderPipeline",
+            "StableCascadePriorPipeline",
             "StableDiffusionAdapterPipeline",
             "StableDiffusionAttendAndExcitePipeline",
             "StableDiffusionControlNetImg2ImgPipeline",
@@ -545,6 +552,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             SASolverScheduler,
             SchedulerMixin,
             ScoreSdeVeScheduler,
+            TCDScheduler,
             UnCLIPScheduler,
             UniPCMultistepScheduler,
             VQDiffusionScheduler,
@@ -617,6 +625,8 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             LatentConsistencyModelImg2ImgPipeline,
             LatentConsistencyModelPipeline,
             LDMTextToImagePipeline,
+            LEditsPPPipelineStableDiffusion,
+            LEditsPPPipelineStableDiffusionXL,
             MusicLDMPipeline,
             PaintByExamplePipeline,
             PIAPipeline,
@@ -624,6 +634,9 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             SemanticStableDiffusionPipeline,
             ShapEImg2ImgPipeline,
             ShapEPipeline,
+            StableCascadeCombinedPipeline,
+            StableCascadeDecoderPipeline,
+            StableCascadePriorPipeline,
             StableDiffusionAdapterPipeline,
             StableDiffusionAttendAndExcitePipeline,
             StableDiffusionControlNetImg2ImgPipeline,

src/diffusers/configuration_utils.py

@@ -127,7 +127,7 @@ class ConfigMixin:
         """The only reason we overwrite `getattr` here is to gracefully deprecate accessing
         config attributes directly. See https://github.com/huggingface/diffusers/pull/3129
 
-        Tihs funtion is mostly copied from PyTorch's __getattr__ overwrite:
+        This function is mostly copied from PyTorch's __getattr__ overwrite:
         https://pytorch.org/docs/stable/_modules/torch/nn/modules/module.html#Module
         """
 
@@ -259,6 +259,10 @@ class ConfigMixin:
         model = cls(**init_dict)
 
         # make sure to also save config parameters that might be used for compatible classes
+        # update _class_name
+        if "_class_name" in hidden_dict:
+            hidden_dict["_class_name"] = cls.__name__
+
         model.register_to_config(**hidden_dict)
 
         # add hidden kwargs of compatible classes to unused_kwargs
@@ -529,7 +533,7 @@ class ConfigMixin:
                 f"{cls.config_name} configuration file."
             )
 
-        # 5. Give nice info if config attributes are initiliazed to default because they have not been passed
+        # 5. Give nice info if config attributes are initialized to default because they have not been passed
         passed_keys = set(init_dict.keys())
         if len(expected_keys - passed_keys) > 0:
             logger.info(

src/diffusers/image_processor.py

@@ -332,7 +332,7 @@ class VaeImageProcessor(ConfigMixin):
         image: Union[PIL.Image.Image, np.ndarray, torch.Tensor],
         height: int,
         width: int,
-        resize_mode: str = "default",  # "defalt", "fill", "crop"
+        resize_mode: str = "default",  # "default", "fill", "crop"
     ) -> Union[PIL.Image.Image, np.ndarray, torch.Tensor]:
         """
         Resize image.
@@ -448,7 +448,7 @@ class VaeImageProcessor(ConfigMixin):
         image: PipelineImageInput,
         height: Optional[int] = None,
         width: Optional[int] = None,
-        resize_mode: str = "default",  # "defalt", "fill", "crop"
+        resize_mode: str = "default",  # "default", "fill", "crop"
         crops_coords: Optional[Tuple[int, int, int, int]] = None,
     ) -> torch.Tensor:
         """
@@ -479,7 +479,7 @@ class VaeImageProcessor(ConfigMixin):
             if isinstance(image, torch.Tensor):
                 # if image is a pytorch tensor could have 2 possible shapes:
                 #    1. batch x height x width: we should insert the channel dimension at position 1
-                #    2. channnel x height x width: we should insert batch dimension at position 0,
+                #    2. channel x height x width: we should insert batch dimension at position 0,
                 #       however, since both channel and batch dimension has same size 1, it is same to insert at position 1
                 #    for simplicity, we insert a dimension of size 1 at position 1 for both cases
                 image = image.unsqueeze(1)

src/diffusers/loaders/ip_adapter.py

@@ -215,7 +215,7 @@ class IPAdapterMixin:
                 else:
                     logger.warning(
                         "image_encoder is not loaded since `image_encoder_folder=None` passed. You will not be able to use `ip_adapter_image` when calling the pipeline with IP-Adapter."
-                        "Use `ip_adapter_image_embedding` to pass pre-geneated image embedding instead."
+                        "Use `ip_adapter_image_embeds` to pass pre-generated image embedding instead."
                     )
 
             # create feature extractor if it has not been registered to the pipeline yet

src/diffusers/loaders/lora.py

@@ -430,7 +430,7 @@ class LoraLoaderMixin:
             # contain the module names of the `unet` as its keys WITHOUT any prefix.
             if not USE_PEFT_BACKEND:
                 warn_message = "You have saved the LoRA weights using the old format. To convert the old LoRA weights to the new format, you can first load them in a dictionary and then create a new dictionary like the following: `new_state_dict = {f'unet.{module_name}': params for module_name, params in old_state_dict.items()}`."
-                logger.warn(warn_message)
+                logger.warning(warn_message)
 
         if len(state_dict.keys()) > 0:
             if adapter_name in getattr(unet, "peft_config", {}):
@@ -882,7 +882,7 @@ class LoraLoaderMixin:
         if fuse_unet or fuse_text_encoder:
             self.num_fused_loras += 1
             if self.num_fused_loras > 1:
-                logger.warn(
+                logger.warning(
                     "The current API is supported for operating with a single LoRA file. You are trying to load and fuse more than one LoRA which is not well-supported.",
                 )
 

src/diffusers/loaders/single_file.py

@@ -56,6 +56,8 @@ def build_sub_model_components(
 
     if component_name == "unet":
         num_in_channels = kwargs.pop("num_in_channels", None)
+        upcast_attention = kwargs.pop("upcast_attention", None)
+
         unet_components = create_diffusers_unet_model_from_ldm(
             pipeline_class_name,
             original_config,
@@ -63,13 +65,21 @@ def build_sub_model_components(
             num_in_channels=num_in_channels,
             image_size=image_size,
             torch_dtype=torch_dtype,
+            model_type=model_type,
+            upcast_attention=upcast_attention,
         )
         return unet_components
 
     if component_name == "vae":
         scaling_factor = kwargs.get("scaling_factor", None)
         vae_components = create_diffusers_vae_model_from_ldm(
-            pipeline_class_name, original_config, checkpoint, image_size, scaling_factor, torch_dtype
+            pipeline_class_name,
+            original_config,
+            checkpoint,
+            image_size,
+            scaling_factor,
+            torch_dtype,
+            model_type=model_type,
         )
         return vae_components
 
@@ -124,11 +134,12 @@ def build_sub_model_components(
 def set_additional_components(
     pipeline_class_name,
     original_config,
+    checkpoint=None,
     model_type=None,
 ):
     components = {}
     if pipeline_class_name in REFINER_PIPELINES:
-        model_type = infer_model_type(original_config, model_type=model_type)
+        model_type = infer_model_type(original_config, checkpoint=checkpoint, model_type=model_type)
         is_refiner = model_type == "SDXL-Refiner"
         components.update(
             {
@@ -181,6 +192,30 @@ class FromSingleFileMixin:
             revision (`str`, *optional*, defaults to `"main"`):
                 The specific model version to use. It can be a branch name, a tag name, a commit id, or any identifier
                 allowed by Git.
+            original_config_file (`str`, *optional*):
+                The path to the original config file that was used to train the model. If not provided, the config file
+                will be inferred from the checkpoint file.
+            model_type (`str`, *optional*):
+                The type of model to load. If not provided, the model type will be inferred from the checkpoint file.
+            image_size (`int`, *optional*):
+                The size of the image output. It's used to configure the `sample_size` parameter of the UNet and VAE model.
+            load_safety_checker (`bool`, *optional*, defaults to `False`):
+                Whether to load the safety checker model or not. By default, the safety checker is not loaded unless a `safety_checker` component is passed to the `kwargs`.
+            num_in_channels (`int`, *optional*):
+                Specify the number of input channels for the UNet model. Read more about how to configure UNet model with this parameter
+                [here](https://huggingface.co/docs/diffusers/training/adapt_a_model#configure-unet2dconditionmodel-parameters).
+            scaling_factor (`float`, *optional*):
+                The scaling factor to use for the VAE model. If not provided, it is inferred from the config file first.
+                If the scaling factor is not found in the config file, the default value 0.18215 is used.
+            scheduler_type (`str`, *optional*):
+                The type of scheduler to load. If not provided, the scheduler type will be inferred from the checkpoint file.
+            prediction_type (`str`, *optional*):
+                The type of prediction to load. If not provided, the prediction type will be inferred from the checkpoint file.
+            kwargs (remaining dictionary of keyword arguments, *optional*):
+                Can be used to overwrite load and saveable variables (the pipeline components of the specific pipeline
+                class). The overwritten components are passed directly to the pipelines `__init__` method. See example
+                below for more information.
+
         Examples:
 
         ```py
@@ -268,7 +303,9 @@ class FromSingleFileMixin:
                     continue
                 init_kwargs.update(components)
 
-        additional_components = set_additional_components(class_name, original_config, model_type=model_type)
+        additional_components = set_additional_components(
+            class_name, original_config, checkpoint=checkpoint, model_type=model_type
+        )
         if additional_components:
             init_kwargs.update(additional_components)
 

src/diffusers/loaders/single_file_utils.py

@@ -28,6 +28,7 @@ from ..schedulers import (
     DDIMScheduler,
     DDPMScheduler,
     DPMSolverMultistepScheduler,
+    EDMDPMSolverMultistepScheduler,
     EulerAncestralDiscreteScheduler,
     EulerDiscreteScheduler,
     HeunDiscreteScheduler,
@@ -80,6 +81,87 @@ SCHEDULER_DEFAULT_CONFIG = {
     "timestep_spacing": "leading",
 }
 
+
+STABLE_CASCADE_DEFAULT_CONFIGS = {
+    "stage_c": {"pretrained_model_name_or_path": "diffusers/stable-cascade-configs", "subfolder": "prior"},
+    "stage_c_lite": {"pretrained_model_name_or_path": "diffusers/stable-cascade-configs", "subfolder": "prior_lite"},
+    "stage_b": {"pretrained_model_name_or_path": "diffusers/stable-cascade-configs", "subfolder": "decoder"},
+    "stage_b_lite": {"pretrained_model_name_or_path": "diffusers/stable-cascade-configs", "subfolder": "decoder_lite"},
+}
+
+
+def convert_stable_cascade_unet_single_file_to_diffusers(original_state_dict):
+    is_stage_c = "clip_txt_mapper.weight" in original_state_dict
+
+    if is_stage_c:
+        state_dict = {}
+        for key in original_state_dict.keys():
+            if key.endswith("in_proj_weight"):
+                weights = original_state_dict[key].chunk(3, 0)
+                state_dict[key.replace("attn.in_proj_weight", "to_q.weight")] = weights[0]
+                state_dict[key.replace("attn.in_proj_weight", "to_k.weight")] = weights[1]
+                state_dict[key.replace("attn.in_proj_weight", "to_v.weight")] = weights[2]
+            elif key.endswith("in_proj_bias"):
+                weights = original_state_dict[key].chunk(3, 0)
+                state_dict[key.replace("attn.in_proj_bias", "to_q.bias")] = weights[0]
+                state_dict[key.replace("attn.in_proj_bias", "to_k.bias")] = weights[1]
+                state_dict[key.replace("attn.in_proj_bias", "to_v.bias")] = weights[2]
+            elif key.endswith("out_proj.weight"):
+                weights = original_state_dict[key]
+                state_dict[key.replace("attn.out_proj.weight", "to_out.0.weight")] = weights
+            elif key.endswith("out_proj.bias"):
+                weights = original_state_dict[key]
+                state_dict[key.replace("attn.out_proj.bias", "to_out.0.bias")] = weights
+            else:
+                state_dict[key] = original_state_dict[key]
+    else:
+        state_dict = {}
+        for key in original_state_dict.keys():
+            if key.endswith("in_proj_weight"):
+                weights = original_state_dict[key].chunk(3, 0)
+                state_dict[key.replace("attn.in_proj_weight", "to_q.weight")] = weights[0]
+                state_dict[key.replace("attn.in_proj_weight", "to_k.weight")] = weights[1]
+                state_dict[key.replace("attn.in_proj_weight", "to_v.weight")] = weights[2]
+            elif key.endswith("in_proj_bias"):
+                weights = original_state_dict[key].chunk(3, 0)
+                state_dict[key.replace("attn.in_proj_bias", "to_q.bias")] = weights[0]
+                state_dict[key.replace("attn.in_proj_bias", "to_k.bias")] = weights[1]
+                state_dict[key.replace("attn.in_proj_bias", "to_v.bias")] = weights[2]
+            elif key.endswith("out_proj.weight"):
+                weights = original_state_dict[key]
+                state_dict[key.replace("attn.out_proj.weight", "to_out.0.weight")] = weights
+            elif key.endswith("out_proj.bias"):
+                weights = original_state_dict[key]
+                state_dict[key.replace("attn.out_proj.bias", "to_out.0.bias")] = weights
+            # rename clip_mapper to clip_txt_pooled_mapper
+            elif key.endswith("clip_mapper.weight"):
+                weights = original_state_dict[key]
+                state_dict[key.replace("clip_mapper.weight", "clip_txt_pooled_mapper.weight")] = weights
+            elif key.endswith("clip_mapper.bias"):
+                weights = original_state_dict[key]
+                state_dict[key.replace("clip_mapper.bias", "clip_txt_pooled_mapper.bias")] = weights
+            else:
+                state_dict[key] = original_state_dict[key]
+
+    return state_dict
+
+
+def infer_stable_cascade_single_file_config(checkpoint):
+    is_stage_c = "clip_txt_mapper.weight" in checkpoint
+    is_stage_b = "down_blocks.1.0.channelwise.0.weight" in checkpoint
+
+    if is_stage_c and (checkpoint["clip_txt_mapper.weight"].shape[0] == 1536):
+        config_type = "stage_c_lite"
+    elif is_stage_c and (checkpoint["clip_txt_mapper.weight"].shape[0] == 2048):
+        config_type = "stage_c"
+    elif is_stage_b and checkpoint["down_blocks.1.0.channelwise.0.weight"].shape[-1] == 576:
+        config_type = "stage_b_lite"
+    elif is_stage_b and checkpoint["down_blocks.1.0.channelwise.0.weight"].shape[-1] == 640:
+        config_type = "stage_b"
+
+    return STABLE_CASCADE_DEFAULT_CONFIGS[config_type]
+
+
 DIFFUSERS_TO_LDM_MAPPING = {
     "unet": {
         "layers": {
@@ -175,6 +257,7 @@ DIFFUSERS_TO_LDM_MAPPING = {
 
 LDM_VAE_KEY = "first_stage_model."
 LDM_VAE_DEFAULT_SCALING_FACTOR = 0.18215
+PLAYGROUND_VAE_SCALING_FACTOR = 0.5
 LDM_UNET_KEY = "model.diffusion_model."
 LDM_CONTROLNET_KEY = "control_model."
 LDM_CLIP_PREFIX_TO_REMOVE = ["cond_stage_model.transformer.", "conditioner.embedders.0.transformer."]
@@ -227,10 +310,34 @@ def fetch_ldm_config_and_checkpoint(
     cache_dir=None,
     local_files_only=None,
     revision=None,
+):
+    checkpoint = load_single_file_model_checkpoint(
+        pretrained_model_link_or_path,
+        resume_download=resume_download,
+        force_download=force_download,
+        proxies=proxies,
+        token=token,
+        cache_dir=cache_dir,
+        local_files_only=local_files_only,
+        revision=revision,
+    )
+    original_config = fetch_original_config(class_name, checkpoint, original_config_file)
+
+    return original_config, checkpoint
+
+
+def load_single_file_model_checkpoint(
+    pretrained_model_link_or_path,
+    resume_download=False,
+    force_download=False,
+    proxies=None,
+    token=None,
+    cache_dir=None,
+    local_files_only=None,
+    revision=None,
 ):
     if os.path.isfile(pretrained_model_link_or_path):
         checkpoint = load_state_dict(pretrained_model_link_or_path)
-
     else:
         repo_id, weights_name = _extract_repo_id_and_weights_name(pretrained_model_link_or_path)
         checkpoint_path = _get_model_file(
@@ -250,9 +357,7 @@ def fetch_ldm_config_and_checkpoint(
     while "state_dict" in checkpoint:
         checkpoint = checkpoint["state_dict"]
 
-    original_config = fetch_original_config(class_name, checkpoint, original_config_file)
-
-    return original_config, checkpoint
+    return checkpoint
 
 
 def infer_original_config_file(class_name, checkpoint):
@@ -305,7 +410,7 @@ def fetch_original_config(pipeline_class_name, checkpoint, original_config_file=
     return original_config
 
 
-def infer_model_type(original_config, model_type=None):
+def infer_model_type(original_config, checkpoint, model_type=None):
     if model_type is not None:
         return model_type
 
@@ -323,7 +428,9 @@ def infer_model_type(original_config, model_type=None):
 
     elif has_network_config:
         context_dim = original_config["model"]["params"]["network_config"]["params"]["context_dim"]
-        if context_dim == 2048:
+        if "edm_mean" in checkpoint and "edm_std" in checkpoint:
+            model_type = "Playground"
+        elif context_dim == 2048:
             model_type = "SDXL"
         else:
             model_type = "SDXL-Refiner"
@@ -344,13 +451,13 @@ def set_image_size(pipeline_class_name, original_config, checkpoint, image_size=
         return image_size
 
     global_step = checkpoint["global_step"] if "global_step" in checkpoint else None
-    model_type = infer_model_type(original_config, model_type)
+    model_type = infer_model_type(original_config, checkpoint, model_type)
 
     if pipeline_class_name == "StableDiffusionUpscalePipeline":
         image_size = original_config["model"]["params"]["unet_config"]["params"]["image_size"]
         return image_size
 
-    elif model_type in ["SDXL", "SDXL-Refiner"]:
+    elif model_type in ["SDXL", "SDXL-Refiner", "Playground"]:
         image_size = 1024
         return image_size
 
@@ -458,8 +565,8 @@ def create_unet_diffusers_config(original_config, image_size: int):
     config = {
         "sample_size": image_size // vae_scale_factor,
         "in_channels": unet_params["in_channels"],
-        "down_block_types": tuple(down_block_types),
-        "block_out_channels": tuple(block_out_channels),
+        "down_block_types": down_block_types,
+        "block_out_channels": block_out_channels,
         "layers_per_block": unet_params["num_res_blocks"],
         "cross_attention_dim": context_dim,
         "attention_head_dim": head_dim,
@@ -478,7 +585,7 @@ def create_unet_diffusers_config(original_config, image_size: int):
         config["num_class_embeds"] = unet_params["num_classes"]
 
     config["out_channels"] = unet_params["out_channels"]
-    config["up_block_types"] = tuple(up_block_types)
+    config["up_block_types"] = up_block_types
 
     return config
 
@@ -506,12 +613,14 @@ def create_controlnet_diffusers_config(original_config, image_size: int):
     return controlnet_config
 
 
-def create_vae_diffusers_config(original_config, image_size, scaling_factor=None):
+def create_vae_diffusers_config(original_config, image_size, scaling_factor=None, latents_mean=None, latents_std=None):
     """
     Creates a config for the diffusers based on the config of the LDM model.
     """
     vae_params = original_config["model"]["params"]["first_stage_config"]["params"]["ddconfig"]
-    if scaling_factor is None and "scale_factor" in original_config["model"]["params"]:
+    if (scaling_factor is None) and (latents_mean is not None) and (latents_std is not None):
+        scaling_factor = PLAYGROUND_VAE_SCALING_FACTOR
+    elif (scaling_factor is None) and ("scale_factor" in original_config["model"]["params"]):
         scaling_factor = original_config["model"]["params"]["scale_factor"]
     elif scaling_factor is None:
         scaling_factor = LDM_VAE_DEFAULT_SCALING_FACTOR
@@ -524,13 +633,15 @@ def create_vae_diffusers_config(original_config, image_size, scaling_factor=None
         "sample_size": image_size,
         "in_channels": vae_params["in_channels"],
         "out_channels": vae_params["out_ch"],
-        "down_block_types": tuple(down_block_types),
-        "up_block_types": tuple(up_block_types),
-        "block_out_channels": tuple(block_out_channels),
+        "down_block_types": down_block_types,
+        "up_block_types": up_block_types,
+        "block_out_channels": block_out_channels,
         "latent_channels": vae_params["z_channels"],
         "layers_per_block": vae_params["num_res_blocks"],
         "scaling_factor": scaling_factor,
     }
+    if latents_mean is not None and latents_std is not None:
+        config.update({"latents_mean": latents_mean, "latents_std": latents_std})
 
     return config
 
@@ -876,7 +987,7 @@ def create_diffusers_controlnet_model_from_ldm(
                 unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
 
         if len(unexpected_keys) > 0:
-            logger.warn(
+            logger.warning(
                 f"Some weights of the model checkpoint were not used when initializing {controlnet.__name__}: \n {[', '.join(unexpected_keys)]}"
             )
     else:
@@ -1052,7 +1163,7 @@ def create_text_encoder_from_ldm_clip_checkpoint(config_name, checkpoint, local_
                 unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
 
         if len(unexpected_keys) > 0:
-            logger.warn(
+            logger.warning(
                 f"Some weights of the model checkpoint were not used when initializing {text_model.__class__.__name__}: \n {[', '.join(unexpected_keys)]}"
             )
     else:
@@ -1147,7 +1258,7 @@ def create_text_encoder_from_open_clip_checkpoint(
                 unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
 
         if len(unexpected_keys) > 0:
-            logger.warn(
+            logger.warning(
                 f"Some weights of the model checkpoint were not used when initializing {text_model.__class__.__name__}: \n {[', '.join(unexpected_keys)]}"
             )
 
@@ -1168,10 +1279,11 @@ def create_diffusers_unet_model_from_ldm(
     original_config,
     checkpoint,
     num_in_channels=None,
-    upcast_attention=False,
+    upcast_attention=None,
     extract_ema=False,
     image_size=None,
     torch_dtype=None,
+    model_type=None,
 ):
     from ..models import UNet2DConditionModel
 
@@ -1190,10 +1302,13 @@ def create_diffusers_unet_model_from_ldm(
         else:
             num_in_channels = 4
 
-    image_size = set_image_size(pipeline_class_name, original_config, checkpoint, image_size=image_size)
+    image_size = set_image_size(
+        pipeline_class_name, original_config, checkpoint, image_size=image_size, model_type=model_type
+    )
     unet_config = create_unet_diffusers_config(original_config, image_size=image_size)
     unet_config["in_channels"] = num_in_channels
-    unet_config["upcast_attention"] = upcast_attention
+    if upcast_attention is not None:
+        unet_config["upcast_attention"] = upcast_attention
 
     diffusers_format_unet_checkpoint = convert_ldm_unet_checkpoint(checkpoint, unet_config, extract_ema=extract_ema)
     ctx = init_empty_weights if is_accelerate_available() else nullcontext
@@ -1210,7 +1325,7 @@ def create_diffusers_unet_model_from_ldm(
                 unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
 
         if len(unexpected_keys) > 0:
-            logger.warn(
+            logger.warning(
                 f"Some weights of the model checkpoint were not used when initializing {unet.__name__}: \n {[', '.join(unexpected_keys)]}"
             )
     else:
@@ -1223,14 +1338,40 @@ def create_diffusers_unet_model_from_ldm(
 
 
 def create_diffusers_vae_model_from_ldm(
-    pipeline_class_name, original_config, checkpoint, image_size=None, scaling_factor=None, torch_dtype=None
+    pipeline_class_name,
+    original_config,
+    checkpoint,
+    image_size=None,
+    scaling_factor=None,
+    torch_dtype=None,
+    model_type=None,
 ):
     # import here to avoid circular imports
     from ..models import AutoencoderKL
 
-    image_size = set_image_size(pipeline_class_name, original_config, checkpoint, image_size=image_size)
+    image_size = set_image_size(
+        pipeline_class_name, original_config, checkpoint, image_size=image_size, model_type=model_type
+    )
+    model_type = infer_model_type(original_config, checkpoint, model_type)
 
-    vae_config = create_vae_diffusers_config(original_config, image_size=image_size, scaling_factor=scaling_factor)
+    if model_type == "Playground":
+        edm_mean = (
+            checkpoint["edm_mean"].to(dtype=torch_dtype).tolist() if torch_dtype else checkpoint["edm_mean"].tolist()
+        )
+        edm_std = (
+            checkpoint["edm_std"].to(dtype=torch_dtype).tolist() if torch_dtype else checkpoint["edm_std"].tolist()
+        )
+    else:
+        edm_mean = None
+        edm_std = None
+
+    vae_config = create_vae_diffusers_config(
+        original_config,
+        image_size=image_size,
+        scaling_factor=scaling_factor,
+        latents_mean=edm_mean,
+        latents_std=edm_std,
+    )
     diffusers_format_vae_checkpoint = convert_ldm_vae_checkpoint(checkpoint, vae_config)
     ctx = init_empty_weights if is_accelerate_available() else nullcontext
 
@@ -1246,7 +1387,7 @@ def create_diffusers_vae_model_from_ldm(
                 unexpected_keys = [k for k in unexpected_keys if re.search(pat, k) is None]
 
         if len(unexpected_keys) > 0:
-            logger.warn(
+            logger.warning(
                 f"Some weights of the model checkpoint were not used when initializing {vae.__name__}: \n {[', '.join(unexpected_keys)]}"
             )
     else:
@@ -1265,7 +1406,7 @@ def create_text_encoders_and_tokenizers_from_ldm(
     local_files_only=False,
     torch_dtype=None,
 ):
-    model_type = infer_model_type(original_config, model_type=model_type)
+    model_type = infer_model_type(original_config, checkpoint=checkpoint, model_type=model_type)
 
     if model_type == "FrozenOpenCLIPEmbedder":
         config_name = "stabilityai/stable-diffusion-2"
@@ -1332,7 +1473,7 @@ def create_text_encoders_and_tokenizers_from_ldm(
                 "text_encoder_2": text_encoder_2,
             }
 
-    elif model_type == "SDXL":
+    elif model_type in ["SDXL", "Playground"]:
         try:
             config_name = "openai/clip-vit-large-patch14"
             tokenizer = CLIPTokenizer.from_pretrained(config_name, local_files_only=local_files_only)
@@ -1383,7 +1524,7 @@ def create_scheduler_from_ldm(
     model_type=None,
 ):
     scheduler_config = get_default_scheduler_config()
-    model_type = infer_model_type(original_config, model_type=model_type)
+    model_type = infer_model_type(original_config, checkpoint=checkpoint, model_type=model_type)
 
     global_step = checkpoint["global_step"] if "global_step" in checkpoint else None
 
@@ -1406,7 +1547,8 @@ def create_scheduler_from_ldm(
 
     if model_type in ["SDXL", "SDXL-Refiner"]:
         scheduler_type = "euler"
-
+    elif model_type == "Playground":
+        scheduler_type = "edm_dpm_solver_multistep"
     else:
         beta_start = original_config["model"]["params"].get("linear_start", 0.02)
         beta_end = original_config["model"]["params"].get("linear_end", 0.085)
@@ -1438,6 +1580,26 @@ def create_scheduler_from_ldm(
     elif scheduler_type == "ddim":
         scheduler = DDIMScheduler.from_config(scheduler_config)
 
+    elif scheduler_type == "edm_dpm_solver_multistep":
+        scheduler_config = {
+            "algorithm_type": "dpmsolver++",
+            "dynamic_thresholding_ratio": 0.995,
+            "euler_at_final": False,
+            "final_sigmas_type": "zero",
+            "lower_order_final": True,
+            "num_train_timesteps": 1000,
+            "prediction_type": "epsilon",
+            "rho": 7.0,
+            "sample_max_value": 1.0,
+            "sigma_data": 0.5,
+            "sigma_max": 80.0,
+            "sigma_min": 0.002,
+            "solver_order": 2,
+            "solver_type": "midpoint",
+            "thresholding": False,
+        }
+        scheduler = EDMDPMSolverMultistepScheduler(**scheduler_config)
+
     else:
         raise ValueError(f"Scheduler of type {scheduler_type} doesn't exist!")
 

src/diffusers/loaders/unet.py

@@ -42,6 +42,11 @@ from ..utils import (
     set_adapter_layers,
     set_weights_and_activate_adapters,
 )
+from .single_file_utils import (
+    convert_stable_cascade_unet_single_file_to_diffusers,
+    infer_stable_cascade_single_file_config,
+    load_single_file_model_checkpoint,
+)
 from .utils import AttnProcsLayers
 
 
@@ -345,7 +350,7 @@ class UNet2DConditionLoadersMixin:
         is_model_cpu_offload = False
         is_sequential_cpu_offload = False
 
-        # For PEFT backend the Unet is already offloaded at this stage as it is handled inside `lora_lora_weights_into_unet`
+        # For PEFT backend the Unet is already offloaded at this stage as it is handled inside `load_lora_weights_into_unet`
         if not USE_PEFT_BACKEND:
             if _pipeline is not None:
                 for _, component in _pipeline.components.items():
@@ -384,7 +389,7 @@ class UNet2DConditionLoadersMixin:
             is_text_encoder_present = any(key.startswith(self.text_encoder_name) for key in state_dict.keys())
             if is_text_encoder_present:
                 warn_message = "The state_dict contains LoRA params corresponding to the text encoder which are not being used here. To use both UNet and text encoder related LoRA params, use [`pipe.load_lora_weights()`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.LoraLoaderMixin.load_lora_weights)."
-                logger.warn(warn_message)
+                logger.warning(warn_message)
             unet_keys = [k for k in state_dict.keys() if k.startswith(self.unet_name)]
             state_dict = {k.replace(f"{self.unet_name}.", ""): v for k, v in state_dict.items() if k in unet_keys}
 
@@ -896,3 +901,103 @@ class UNet2DConditionLoadersMixin:
         self.config.encoder_hid_dim_type = "ip_image_proj"
 
         self.to(dtype=self.dtype, device=self.device)
+
+
+class FromOriginalUNetMixin:
+    """
+    Load pretrained UNet model weights saved in the `.ckpt` or `.safetensors` format into a [`StableCascadeUNet`].
+    """
+
+    @classmethod
+    @validate_hf_hub_args
+    def from_single_file(cls, pretrained_model_link_or_path, **kwargs):
+        r"""
+        Instantiate a [`StableCascadeUNet`] from pretrained StableCascadeUNet weights saved in the original `.ckpt` or
+        `.safetensors` format. The pipeline is set in evaluation mode (`model.eval()`) by default.
+
+        Parameters:
+            pretrained_model_link_or_path (`str` or `os.PathLike`, *optional*):
+                Can be either:
+                    - A link to the `.ckpt` file (for example
+                      `"https://huggingface.co/<repo_id>/blob/main/<path_to_file>.ckpt"`) on the Hub.
+                    - A path to a *file* containing all pipeline weights.
+            config: (`dict`, *optional*):
+                Dictionary containing the configuration of the model:
+            torch_dtype (`str` or `torch.dtype`, *optional*):
+                Override the default `torch.dtype` and load the model with another dtype. If `"auto"` is passed, the
+                dtype is automatically derived from the model's weights.
+            force_download (`bool`, *optional*, defaults to `False`):
+                Whether or not to force the (re-)download of the model weights and configuration files, overriding the
+                cached versions if they exist.
+            cache_dir (`Union[str, os.PathLike]`, *optional*):
+                Path to a directory where a downloaded pretrained model configuration is cached if the standard cache
+                is not used.
+            resume_download (`bool`, *optional*, defaults to `False`):
+                Whether or not to resume downloading the model weights and configuration files. If set to `False`, any
+                incompletely downloaded files are deleted.
+            proxies (`Dict[str, str]`, *optional*):
+                A dictionary of proxy servers to use by protocol or endpoint, for example, `{'http': 'foo.bar:3128',
+                'http://hostname': 'foo.bar:4012'}`. The proxies are used on each request.
+            local_files_only (`bool`, *optional*, defaults to `False`):
+                Whether to only load local model weights and configuration files or not. If set to True, the model
+                won't be downloaded from the Hub.
+            token (`str` or *bool*, *optional*):
+                The token to use as HTTP bearer authorization for remote files. If `True`, the token generated from
+                `diffusers-cli login` (stored in `~/.huggingface`) is used.
+            revision (`str`, *optional*, defaults to `"main"`):
+                The specific model version to use. It can be a branch name, a tag name, a commit id, or any identifier
+                allowed by Git.
+            kwargs (remaining dictionary of keyword arguments, *optional*):
+                Can be used to overwrite load and saveable variables of the model.
+
+        """
+        class_name = cls.__name__
+        if class_name != "StableCascadeUNet":
+            raise ValueError("FromOriginalUNetMixin is currently only compatible with StableCascadeUNet")
+
+        config = kwargs.pop("config", None)
+        resume_download = kwargs.pop("resume_download", False)
+        force_download = kwargs.pop("force_download", False)
+        proxies = kwargs.pop("proxies", None)
+        token = kwargs.pop("token", None)
+        cache_dir = kwargs.pop("cache_dir", None)
+        local_files_only = kwargs.pop("local_files_only", None)
+        revision = kwargs.pop("revision", None)
+        torch_dtype = kwargs.pop("torch_dtype", None)
+
+        checkpoint = load_single_file_model_checkpoint(
+            pretrained_model_link_or_path,
+            resume_download=resume_download,
+            force_download=force_download,
+            proxies=proxies,
+            token=token,
+            cache_dir=cache_dir,
+            local_files_only=local_files_only,
+            revision=revision,
+        )
+
+        if config is None:
+            config = infer_stable_cascade_single_file_config(checkpoint)
+            model_config = cls.load_config(**config, **kwargs)
+        else:
+            model_config = config
+
+        ctx = init_empty_weights if is_accelerate_available() else nullcontext
+        with ctx():
+            model = cls.from_config(model_config, **kwargs)
+
+        diffusers_format_checkpoint = convert_stable_cascade_unet_single_file_to_diffusers(checkpoint)
+        if is_accelerate_available():
+            unexpected_keys = load_model_dict_into_meta(model, diffusers_format_checkpoint, dtype=torch_dtype)
+            if len(unexpected_keys) > 0:
+                logger.warn(
+                    f"Some weights of the model checkpoint were not used when initializing {cls.__name__}: \n {[', '.join(unexpected_keys)]}"
+                )
+
+        else:
+            model.load_state_dict(diffusers_format_checkpoint)
+
+        if torch_dtype is not None:
+            model.to(torch_dtype)
+
+        return model