Release: v0.27.0

* update

* update

* update

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -66,32 +66,32 @@ body:
         Questions on DiffusionPipeline (Saving, Loading, From pretrained, ...):
 
         Questions on pipelines:
-        - Stable Diffusion @yiyixuxu @DN6 @sayakpaul @patrickvonplaten
-        - Stable Diffusion XL @yiyixuxu @sayakpaul @DN6 @patrickvonplaten
-        - Kandinsky @yiyixuxu @patrickvonplaten
-        - ControlNet @sayakpaul @yiyixuxu @DN6 @patrickvonplaten
-        - T2I Adapter @sayakpaul @yiyixuxu @DN6 @patrickvonplaten
-        - IF @DN6 @patrickvonplaten
-        - Text-to-Video / Video-to-Video @DN6 @sayakpaul @patrickvonplaten
-        - Wuerstchen @DN6 @patrickvonplaten
+        - Stable Diffusion @yiyixuxu @DN6 @sayakpaul 
+        - Stable Diffusion XL @yiyixuxu @sayakpaul @DN6 
+        - Kandinsky @yiyixuxu 
+        - ControlNet @sayakpaul @yiyixuxu @DN6 
+        - T2I Adapter @sayakpaul @yiyixuxu @DN6 
+        - IF @DN6 
+        - Text-to-Video / Video-to-Video @DN6 @sayakpaul 
+        - Wuerstchen @DN6 
         - Other: @yiyixuxu @DN6
 
         Questions on models:
-        - UNet @DN6 @yiyixuxu @sayakpaul @patrickvonplaten
-        - VAE @sayakpaul @DN6 @yiyixuxu @patrickvonplaten
-        - Transformers/Attention @DN6 @yiyixuxu @sayakpaul @DN6 @patrickvonplaten
+        - UNet @DN6 @yiyixuxu @sayakpaul 
+        - VAE @sayakpaul @DN6 @yiyixuxu 
+        - Transformers/Attention @DN6 @yiyixuxu @sayakpaul @DN6 
 
-        Questions on Schedulers: @yiyixuxu @patrickvonplaten
+        Questions on Schedulers: @yiyixuxu 
 
-        Questions on LoRA: @sayakpaul @patrickvonplaten
+        Questions on LoRA: @sayakpaul 
 
-        Questions on Textual Inversion: @sayakpaul @patrickvonplaten
+        Questions on Textual Inversion: @sayakpaul 
 
         Questions on Training: 
-        - DreamBooth @sayakpaul @patrickvonplaten
-        - Text-to-Image Fine-tuning @sayakpaul @patrickvonplaten
-        - Textual Inversion @sayakpaul @patrickvonplaten
-        - ControlNet @sayakpaul @patrickvonplaten
+        - DreamBooth @sayakpaul 
+        - Text-to-Image Fine-tuning @sayakpaul 
+        - Textual Inversion @sayakpaul 
+        - ControlNet @sayakpaul 
 
         Questions on Tests: @DN6 @sayakpaul @yiyixuxu 
 
@@ -99,7 +99,7 @@ body:
 
         Questions on JAX- and MPS-related things: @pcuenca
 
-        Questions on audio pipelines: @DN6 @patrickvonplaten
+        Questions on audio pipelines: @DN6 
         
 
         

.github/PULL_REQUEST_TEMPLATE.md

@@ -38,13 +38,13 @@ members/contributors who may be interested in your PR.
 
 Core library:
 
-- Schedulers: @yiyixuxu and @patrickvonplaten
-- Pipelines:  @patrickvonplaten and @sayakpaul
-- Training examples: @sayakpaul and @patrickvonplaten
-- Docs: @stevhliu and @yiyixuxu
+- Schedulers: @yiyixuxu 
+- Pipelines:  @sayakpaul @yiyixuxu @DN6
+- Training examples: @sayakpaul 
+- Docs: @stevhliu and @sayakpaul
 - JAX and MPS: @pcuenca
 - Audio: @sanchit-gandhi
-- General functionalities: @patrickvonplaten and @sayakpaul
+- General functionalities: @sayakpaul @yiyixuxu @DN6
 
 Integrations:
 

.github/workflows/benchmark.yml

@@ -1,6 +1,7 @@
 name: Benchmarking tests
 
 on:
+  workflow_dispatch:
   schedule:
     - cron: "30 1 1,15 * *" # every 2 weeks on the 1st and the 15th of every month at 1:30 AM
 
@@ -31,8 +32,9 @@ jobs:
       - name: Install dependencies
         run: |
           apt-get update && apt-get install libsndfile1-dev libgl1 -y
-          python -m pip install -e .[quality,test]
-          python -m pip install pandas peft
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+          python -m uv pip install -e [quality,test]
+          python -m uv pip install pandas peft
       - name: Environment
         run: |
           python utils/print_env.py

.github/workflows/build_docker_images.yml

@@ -1,21 +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
@@ -50,3 +87,27 @@ jobs:
           context: ./docker/${{ matrix.image-name }}
           push: true
           tags: ${{ env.REGISTRY }}/${{ matrix.image-name }}:latest
+
+      - name: Post to a Slack channel
+        id: slack
+        uses: slackapi/slack-github-action@6c661ce58804a1a20f6dc5fbee7f0381b469e001
+        with:
+          # Slack channel id, channel name, or user id to post message.
+          # See also: https://api.slack.com/methods/chat.postMessage#channels
+          channel-id: ${{ env.CI_SLACK_CHANNEL }}
+          # For posting a rich message using Block Kit
+          payload: |
+            {
+              "text": "${{ matrix.image-name }} Docker Image build result: ${{ job.status }}\n${{ github.event.head_commit.url }}",
+              "blocks": [
+                {
+                  "type": "section",
+                  "text": {
+                    "type": "mrkdwn",
+                    "text": "${{ matrix.image-name }} Docker Image build result: ${{ job.status }}\n${{ github.event.head_commit.url }}"
+                  }
+                }
+              ]
+            }
+        env:
+          SLACK_BOT_TOKEN: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

.github/workflows/build_documentation.yml

@@ -7,6 +7,10 @@ on:
       - doc-builder*
       - v*-release
       - v*-patch
+    paths:
+      - "src/diffusers/**.py"
+      - "examples/**"
+      - "docs/**"
 
 jobs:
   build:

.github/workflows/build_pr_documentation.yml

@@ -2,6 +2,10 @@ name: Build PR Documentation
 
 on:
   pull_request:
+    paths:
+      - "src/diffusers/**.py"
+      - "examples/**"
+      - "docs/**"
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}

.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:
@@ -60,9 +61,11 @@ jobs:
 
       - name: Install dependencies
         run: |
-          python -m pip install -e .[quality,test]
-          python -m pip install -U git+https://github.com/huggingface/transformers
-          python -m pip install git+https://github.com/huggingface/accelerate
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+          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: |
@@ -73,19 +76,23 @@ jobs:
         env:
           HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
         run: |
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
           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' }}
         env:
           HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
         run: |
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
           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
@@ -93,9 +100,11 @@ jobs:
         env:
           HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
         run: |
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
           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
@@ -108,6 +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
@@ -132,10 +147,11 @@ jobs:
       - name: Install dependencies
         shell: arch -arch arm64 bash {0}
         run: |
-          ${CONDA_RUN} python -m pip install --upgrade pip
-          ${CONDA_RUN} python -m pip install -e .[quality,test]
-          ${CONDA_RUN} python -m pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cpu
-          ${CONDA_RUN} python -m pip install git+https://github.com/huggingface/accelerate
+          ${CONDA_RUN} python -m pip install --upgrade pip uv
+          ${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}
@@ -148,7 +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() }}
@@ -160,3 +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/pr_dependency_test.yml

@@ -4,6 +4,8 @@ on:
   pull_request:
     branches:
       - main
+    paths:
+      - "src/diffusers/**.py"
   push:
     branches:
       - main
@@ -23,10 +25,12 @@ jobs:
           python-version: "3.8"
       - name: Install dependencies
         run: |
-          python -m pip install --upgrade pip
-          pip install -e .
-          pip install pytest
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+          python -m pip install --upgrade pip uv
+          python -m uv pip install -e .
+          python -m uv pip install pytest
       - name: Check for soft dependencies
         run: |
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
           pytest tests/others/test_dependencies.py
       

.github/workflows/pr_flax_dependency_test.yml

@@ -4,6 +4,8 @@ on:
   pull_request:
     branches:
       - main
+    paths:
+      - "src/diffusers/**.py"
   push:
     branches:
       - main
@@ -23,12 +25,14 @@ jobs:
           python-version: "3.8"
       - name: Install dependencies
         run: |
-          python -m pip install --upgrade pip
-          pip install -e .
-          pip install "jax[cpu]>=0.2.16,!=0.3.2"
-          pip install "flax>=0.4.1"
-          pip install "jaxlib>=0.1.65"
-          pip install pytest
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+          python -m pip install --upgrade pip uv
+          python -m uv pip install -e .
+          python -m uv pip install "jax[cpu]>=0.2.16,!=0.3.2"
+          python -m uv pip install "flax>=0.4.1"
+          python -m uv pip install "jaxlib>=0.1.65"
+          python -m uv pip install pytest
       - name: Check for soft dependencies
         run: |
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
           pytest tests/others/test_dependencies.py

.github/workflows/pr_quality.yml

@@ -1,49 +0,0 @@
-name: Run code quality checks
-
-on:
-  pull_request:
-    branches:
-      - main
-  push:
-    branches:
-      - main
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
-  cancel-in-progress: true
-
-jobs:
-  check_code_quality:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v3
-      - name: Set up Python
-        uses: actions/setup-python@v4
-        with:
-          python-version: "3.8"
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install .[quality]
-      - name: Check quality
-        run: |
-          ruff check examples tests src utils scripts
-          ruff format examples tests src utils scripts --check
-
-  check_repository_consistency:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v3
-      - name: Set up Python
-        uses: actions/setup-python@v4
-        with:
-          python-version: "3.8"
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install .[quality]
-      - name: Check quality
-        run: |
-          python utils/check_copies.py
-          python utils/check_dummies.py
-          make deps_table_check_updated

.github/workflows/pr_test_fetcher.yml

@@ -33,7 +33,8 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
     - name: Environment
       run: |
         python utils/print_env.py
@@ -89,15 +90,18 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m pip install -e [quality,test]
         python -m pip install accelerate
 
     - name: Environment
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python utils/print_env.py
 
     - name: Run all selected tests on CPU
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.modules }}_tests_cpu ${{ fromJson(needs.setup_pr_tests.outputs.test_map)[matrix.modules] }}
 
     - name: Failure short reports
@@ -144,15 +148,18 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m pip install -e [quality,test]
 
     - name: Environment
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python utils/print_env.py
 
     - name: Run Hub tests for models, schedulers, and pipelines on a staging env
       if: ${{ matrix.config.framework == 'hub_tests_pytorch' }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         HUGGINGFACE_CO_STAGING=true python -m pytest \
           -m "is_staging_test" \
           --make-reports=tests_${{ matrix.config.report }} \

.github/workflows/pr_test_peft_backend.yml

@@ -4,6 +4,9 @@ on:
   pull_request:
     branches:
       - main
+    paths:
+      - "src/diffusers/**.py"
+      - "tests/**.py"
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
@@ -16,7 +19,44 @@ env:
   PYTEST_TIMEOUT: 60
 
 jobs:
+  check_code_quality:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.8"
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install .[quality]
+      - name: Check quality
+        run: |
+          ruff check examples tests src utils scripts
+          ruff format examples tests src utils scripts --check
+
+  check_repository_consistency:
+    needs: check_code_quality
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.8"
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install .[quality]
+      - name: Check quality
+        run: |
+          python utils/check_copies.py
+          python utils/check_dummies.py
+          make deps_table_check_updated
+
   run_fast_tests:
+    needs: [check_code_quality, check_repository_consistency]
     strategy:
       fail-fast: false
       matrix:
@@ -44,21 +84,24 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
         if [ "${{ matrix.lib-versions }}" == "main" ]; then
-            python -m pip install -U git+https://github.com/huggingface/peft.git
-            python -m pip install -U git+https://github.com/huggingface/transformers.git
-            python -m pip install -U git+https://github.com/huggingface/accelerate.git
+            python -m uv pip install -U peft@git+https://github.com/huggingface/peft.git
+            python -m uv pip install -U transformers@git+https://github.com/huggingface/transformers.git
+            python -m uv pip install -U accelerate@git+https://github.com/huggingface/accelerate.git
         else
-            python -m pip install -U peft transformers accelerate
+            python -m uv pip install -U peft transformers accelerate
         fi
 
     - name: Environment
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python utils/print_env.py
 
     - name: Run fast PyTorch LoRA CPU tests with PEFT backend
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 1 --max-worker-restart=0 --dist=loadfile \
           -s -v \
           --make-reports=tests_${{ matrix.config.report }} \

.github/workflows/pr_tests.yml

@@ -4,6 +4,14 @@ on:
   pull_request:
     branches:
       - main
+    paths:
+      - "src/diffusers/**.py"
+      - "benchmarks/**.py"
+      - "examples/**.py"
+      - "scripts/**.py"
+      - "tests/**.py"
+      - ".github/**.yml"
+      - "utils/**.py"
   push:
     branches:
       - ci-*
@@ -19,7 +27,44 @@ env:
   PYTEST_TIMEOUT: 60
 
 jobs:
+  check_code_quality:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.8"
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install .[quality]
+      - name: Check quality
+        run: |
+          ruff check examples tests src utils scripts
+          ruff format examples tests src utils scripts --check
+
+  check_repository_consistency:
+    needs: check_code_quality
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.8"
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install .[quality]
+      - name: Check quality
+        run: |
+          python utils/check_copies.py
+          python utils/check_dummies.py
+          make deps_table_check_updated
+
   run_fast_tests:
+    needs: [check_code_quality, check_repository_consistency]
     strategy:
       fail-fast: false
       matrix:
@@ -66,16 +111,19 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
-        python -m pip install accelerate
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
+        python -m uv pip install accelerate
 
     - name: Environment
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python utils/print_env.py
 
     - name: Run fast PyTorch Pipeline CPU tests
       if: ${{ matrix.config.framework == 'pytorch_pipelines' }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 2 --max-worker-restart=0 --dist=loadfile \
           -s -v -k "not Flax and not Onnx" \
           --make-reports=tests_${{ matrix.config.report }} \
@@ -84,6 +132,7 @@ jobs:
     - name: Run fast PyTorch Model Scheduler CPU tests
       if: ${{ matrix.config.framework == 'pytorch_models' }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 2 --max-worker-restart=0 --dist=loadfile \
           -s -v -k "not Flax and not Onnx and not Dependency" \
           --make-reports=tests_${{ matrix.config.report }} \
@@ -92,6 +141,7 @@ jobs:
     - name: Run fast Flax TPU tests
       if: ${{ matrix.config.framework == 'flax' }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 2 --max-worker-restart=0 --dist=loadfile \
           -s -v -k "Flax" \
           --make-reports=tests_${{ matrix.config.report }} \
@@ -100,7 +150,8 @@ jobs:
     - name: Run example PyTorch CPU tests
       if: ${{ matrix.config.framework == 'pytorch_examples' }}
       run: |
-        python -m pip install peft
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install peft
         python -m pytest -n 2 --max-worker-restart=0 --dist=loadfile \
           --make-reports=tests_${{ matrix.config.report }} \
           examples
@@ -117,6 +168,7 @@ jobs:
         path: reports
 
   run_staging_tests:
+    needs: [check_code_quality, check_repository_consistency]
     strategy:
       fail-fast: false
       matrix:
@@ -148,15 +200,18 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
 
     - name: Environment
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python utils/print_env.py
 
     - name: Run Hub tests for models, schedulers, and pipelines on a staging env
       if: ${{ matrix.config.framework == 'hub_tests_pytorch' }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         HUGGINGFACE_CO_STAGING=true python -m pytest \
           -m "is_staging_test" \
           --make-reports=tests_${{ matrix.config.report }} \

.github/workflows/pr_torch_dependency_test.yml

@@ -4,6 +4,8 @@ on:
   pull_request:
     branches:
       - main
+    paths:
+      - "src/diffusers/**.py"
   push:
     branches:
       - main
@@ -23,10 +25,12 @@ jobs:
           python-version: "3.8"
       - name: Install dependencies
         run: |
-          python -m pip install --upgrade pip
-          pip install -e .
-          pip install torch torchvision torchaudio
-          pip install pytest
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+          python -m pip install --upgrade pip uv
+          python -m uv pip install -e .
+          python -m uv pip install torch torchvision torchaudio
+          python -m uv pip install pytest
       - name: Check for soft dependencies
         run: |
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
           pytest tests/others/test_dependencies.py

.github/workflows/push_tests.yml

@@ -4,7 +4,10 @@ on:
   push:
     branches:
       - main
-
+    paths:
+      - "src/diffusers/**.py"
+      - "examples/**.py"
+      - "tests/**.py"
 
 env:
   DIFFUSERS_IS_CI: yes
@@ -18,7 +21,7 @@ env:
 jobs:
   setup_torch_cuda_pipeline_matrix:
     name: Setup Torch Pipelines CUDA Slow Tests Matrix
-    runs-on: docker-gpu
+    runs-on: [single-gpu, nvidia-gpu, t4, ci]
     container:
       image: diffusers/diffusers-pytorch-cpu # this is a CPU image, but we need it to fetch the matrix
       options: --shm-size "16gb" --ipc host
@@ -32,8 +35,9 @@ jobs:
       - name: Install dependencies
         run: |
           apt-get update && apt-get install libsndfile1-dev libgl1 -y
-          python -m pip install -e .[quality,test]
-          python -m pip install git+https://github.com/huggingface/accelerate.git
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+          python -m uv pip install -e [quality,test]
+          python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate.git
 
       - name: Environment
         run: |
@@ -58,10 +62,9 @@ jobs:
     needs: setup_torch_cuda_pipeline_matrix
     strategy:
       fail-fast: false
-      max-parallel: 1
       matrix:
         module: ${{ fromJson(needs.setup_torch_cuda_pipeline_matrix.outputs.pipeline_test_matrix) }}
-    runs-on: docker-gpu
+    runs-on: [single-gpu, nvidia-gpu, t4, ci]
     container:
       image: diffusers/diffusers-pytorch-cuda
       options: --shm-size "16gb" --ipc host -v /mnt/hf_cache:/mnt/cache/ --gpus 0
@@ -76,8 +79,9 @@ jobs:
       - name: Install dependencies
         run: |
           apt-get update && apt-get install libsndfile1-dev libgl1 -y
-          python -m pip install -e .[quality,test]
-          python -m pip install git+https://github.com/huggingface/accelerate.git
+          python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+          python -m uv pip install -e [quality,test]
+          python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate.git
       - name: Environment
         run: |
           python utils/print_env.py
@@ -125,8 +129,9 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
-        python -m pip install git+https://github.com/huggingface/accelerate.git
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
+        python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate.git
 
     - name: Environment
       run: |
@@ -174,9 +179,10 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
-        python -m pip install git+https://github.com/huggingface/accelerate.git
-        python -m pip install git+https://github.com/huggingface/peft.git
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
+        python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate.git
+        python -m uv pip install peft@git+https://github.com/huggingface/peft.git
 
     - name: Environment
       run: |
@@ -224,8 +230,9 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
-        python -m pip install git+https://github.com/huggingface/accelerate.git
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
+        python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate.git
 
     - name: Environment
       run: |
@@ -271,8 +278,9 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
-        python -m pip install git+https://github.com/huggingface/accelerate.git
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
+        python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate.git
 
     - name: Environment
       run: |
@@ -320,7 +328,8 @@ jobs:
         nvidia-smi
     - name: Install dependencies
       run: |
-        python -m pip install -e .[quality,test,training]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test,training]
     - name: Environment
       run: |
         python utils/print_env.py
@@ -360,7 +369,8 @@ jobs:
         nvidia-smi
     - name: Install dependencies
       run: |
-        python -m pip install -e .[quality,test,training]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test,training]
     - name: Environment
       run: |
         python utils/print_env.py
@@ -401,16 +411,19 @@ jobs:
 
     - name: Install dependencies
       run: |
-        python -m pip install -e .[quality,test,training]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test,training]
 
     - name: Environment
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python utils/print_env.py
 
     - name: Run example tests on GPU
       env:
         HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 1 --max-worker-restart=0 --dist=loadfile -s -v --make-reports=examples_torch_cuda examples/
 
     - name: Failure short reports

.github/workflows/push_tests_fast.yml

@@ -4,6 +4,10 @@ on:
   push:
     branches:
       - main
+    paths:
+      - "src/diffusers/**.py"
+      - "examples/**.py"
+      - "tests/**.py"
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
@@ -65,15 +69,18 @@ jobs:
     - name: Install dependencies
       run: |
         apt-get update && apt-get install libsndfile1-dev libgl1 -y
-        python -m pip install -e .[quality,test]
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install -e [quality,test]
 
     - name: Environment
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python utils/print_env.py
 
     - name: Run fast PyTorch CPU tests
       if: ${{ matrix.config.framework == 'pytorch' }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 2 --max-worker-restart=0 --dist=loadfile \
           -s -v -k "not Flax and not Onnx" \
           --make-reports=tests_${{ matrix.config.report }} \
@@ -82,6 +89,7 @@ jobs:
     - name: Run fast Flax TPU tests
       if: ${{ matrix.config.framework == 'flax' }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 2 --max-worker-restart=0 --dist=loadfile \
           -s -v -k "Flax" \
           --make-reports=tests_${{ matrix.config.report }} \
@@ -90,6 +98,7 @@ jobs:
     - name: Run fast ONNXRuntime CPU tests
       if: ${{ matrix.config.framework == 'onnxruntime' }}
       run: |
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
         python -m pytest -n 2 --max-worker-restart=0 --dist=loadfile \
           -s -v -k "Onnx" \
           --make-reports=tests_${{ matrix.config.report }} \
@@ -98,7 +107,8 @@ jobs:
     - name: Run example PyTorch CPU tests
       if: ${{ matrix.config.framework == 'pytorch_examples' }}
       run: |
-        python -m pip install peft
+        python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
+        python -m uv pip install peft
         python -m pytest -n 2 --max-worker-restart=0 --dist=loadfile \
           --make-reports=tests_${{ matrix.config.report }} \
           examples

.github/workflows/push_tests_mps.yml

@@ -4,6 +4,9 @@ on:
   push:
     branches:
       - main
+    paths:
+      - "src/diffusers/**.py"
+      - "tests/**.py"
 
 env:
   DIFFUSERS_IS_CI: yes
@@ -41,11 +44,11 @@ jobs:
     - name: Install dependencies
       shell: arch -arch arm64 bash {0}
       run: |
-        ${CONDA_RUN} python -m pip install --upgrade pip
-        ${CONDA_RUN} python -m pip install -e .[quality,test]
-        ${CONDA_RUN} python -m pip install torch torchvision torchaudio
-        ${CONDA_RUN} python -m pip install git+https://github.com/huggingface/accelerate.git
-        ${CONDA_RUN} python -m pip install transformers --upgrade
+        ${CONDA_RUN} python -m pip install --upgrade pip uv
+        ${CONDA_RUN} python -m uv pip install -e [quality,test]
+        ${CONDA_RUN} python -m uv pip install torch torchvision torchaudio
+        ${CONDA_RUN} python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate.git
+        ${CONDA_RUN} python -m uv pip install transformers --upgrade
 
     - name: Environment
       shell: arch -arch arm64 bash {0}

.github/workflows/pypi_publish.yaml

@@ -0,0 +1,79 @@
+# 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
+      
+      - 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

benchmarks/base_classes.py

@@ -141,6 +141,7 @@ class LCMLoRATextToImageBenchmark(TextToImageBenchmark):
         super().__init__(args)
         self.pipe.load_lora_weights(self.lora_id)
         self.pipe.fuse_lora()
+        self.pipe.unload_lora_weights()
         self.pipe.scheduler = LCMScheduler.from_config(self.pipe.scheduler.config)
 
     def get_result_filepath(self, args):
@@ -235,6 +236,35 @@ class InpaintingBenchmark(ImageToImageBenchmark):
         )
 
 
+class IPAdapterTextToImageBenchmark(TextToImageBenchmark):
+    url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/load_neg_embed.png"
+    image = load_image(url)
+
+    def __init__(self, args):
+        pipe = self.pipeline_class.from_pretrained(args.ckpt, torch_dtype=torch.float16).to("cuda")
+        pipe.load_ip_adapter(
+            args.ip_adapter_id[0],
+            subfolder="models" if "sdxl" not in args.ip_adapter_id[1] else "sdxl_models",
+            weight_name=args.ip_adapter_id[1],
+        )
+
+        if args.run_compile:
+            pipe.unet.to(memory_format=torch.channels_last)
+            print("Run torch compile")
+            pipe.unet = torch.compile(pipe.unet, mode="reduce-overhead", fullgraph=True)
+
+        pipe.set_progress_bar_config(disable=True)
+        self.pipe = pipe
+
+    def run_inference(self, pipe, args):
+        _ = pipe(
+            prompt=PROMPT,
+            ip_adapter_image=self.image,
+            num_inference_steps=args.num_inference_steps,
+            num_images_per_prompt=args.batch_size,
+        )
+
+
 class ControlNetBenchmark(TextToImageBenchmark):
     pipeline_class = StableDiffusionControlNetPipeline
     aux_network_class = ControlNetModel

benchmarks/benchmark_ip_adapters.py

@@ -0,0 +1,32 @@
+import argparse
+import sys
+
+
+sys.path.append(".")
+from base_classes import IPAdapterTextToImageBenchmark  # noqa: E402
+
+
+IP_ADAPTER_CKPTS = {
+    "runwayml/stable-diffusion-v1-5": ("h94/IP-Adapter", "ip-adapter_sd15.bin"),
+    "stabilityai/stable-diffusion-xl-base-1.0": ("h94/IP-Adapter", "ip-adapter_sdxl.bin"),
+}
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--ckpt",
+        type=str,
+        default="runwayml/stable-diffusion-v1-5",
+        choices=list(IP_ADAPTER_CKPTS.keys()),
+    )
+    parser.add_argument("--batch_size", type=int, default=1)
+    parser.add_argument("--num_inference_steps", type=int, default=50)
+    parser.add_argument("--model_cpu_offload", action="store_true")
+    parser.add_argument("--run_compile", action="store_true")
+    args = parser.parse_args()
+
+    args.ip_adapter_id = IP_ADAPTER_CKPTS[args.ckpt]
+    benchmark_pipe = IPAdapterTextToImageBenchmark(args)
+    args.ckpt = f"{args.ckpt} (IP-Adapter)"
+    benchmark_pipe.benchmark(args)

benchmarks/run_all.py

@@ -72,7 +72,7 @@ def main():
                 command += " --run_compile"
                 run_command(command.split())
 
-        elif file == "benchmark_sd_inpainting.py":
+        elif file in ["benchmark_sd_inpainting.py", "benchmark_ip_adapters.py"]:
             sdxl_ckpt = "stabilityai/stable-diffusion-xl-base-1.0"
             command = f"python {file} --ckpt {sdxl_ckpt}"
             run_command(command.split())

docker/diffusers-flax-cpu/Dockerfile

@@ -23,13 +23,13 @@ ENV PATH="/opt/venv/bin:$PATH"
 
 # pre-install the heavy dependencies (these can later be overridden by the deps from setup.py)
 # follow the instructions here: https://cloud.google.com/tpu/docs/run-in-container#train_a_jax_model_in_a_docker_container
-RUN python3 -m pip install --no-cache-dir --upgrade pip && \
-    python3 -m pip install --upgrade --no-cache-dir \
+RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
+    python3 -m uv pip install --upgrade --no-cache-dir \
         clu \
         "jax[cpu]>=0.2.16,!=0.3.2" \
         "flax>=0.4.1" \
         "jaxlib>=0.1.65" && \
-    python3 -m pip install --no-cache-dir \
+    python3 -m uv pip install --no-cache-dir \
         accelerate \
         datasets \
         hf-doc-builder \

docker/diffusers-flax-tpu/Dockerfile

@@ -23,15 +23,15 @@ ENV PATH="/opt/venv/bin:$PATH"
 
 # pre-install the heavy dependencies (these can later be overridden by the deps from setup.py)
 # follow the instructions here: https://cloud.google.com/tpu/docs/run-in-container#train_a_jax_model_in_a_docker_container
-RUN python3 -m pip install --no-cache-dir --upgrade pip && \
+RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
     python3 -m pip install --no-cache-dir \
         "jax[tpu]>=0.2.16,!=0.3.2" \
         -f https://storage.googleapis.com/jax-releases/libtpu_releases.html && \
-    python3 -m pip install --upgrade --no-cache-dir \
+    python3 -m uv pip install --upgrade --no-cache-dir \
         clu \
         "flax>=0.4.1" \
         "jaxlib>=0.1.65" && \
-    python3 -m pip install --no-cache-dir \
+    python3 -m uv pip install --no-cache-dir \
         accelerate \
         datasets \
         hf-doc-builder \

docker/diffusers-onnxruntime-cpu/Dockerfile

@@ -22,14 +22,14 @@ RUN python3 -m venv /opt/venv
 ENV PATH="/opt/venv/bin:$PATH"
 
 # pre-install the heavy dependencies (these can later be overridden by the deps from setup.py)
-RUN python3 -m pip install --no-cache-dir --upgrade pip && \
-    python3 -m pip install --no-cache-dir \
+RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
+    python3 -m uv pip install --no-cache-dir \
         torch==2.1.2 \
         torchvision==0.16.2 \
         torchaudio==2.1.2 \
         onnxruntime \
         --extra-index-url https://download.pytorch.org/whl/cpu && \
-    python3 -m pip install --no-cache-dir \
+    python3 -m uv pip install --no-cache-dir \
         accelerate \
         datasets \
         hf-doc-builder \

docker/diffusers-onnxruntime-cuda/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:11.6.2-cudnn8-devel-ubuntu20.04
+FROM nvidia/cuda:12.1.0-runtime-ubuntu20.04
 LABEL maintainer="Hugging Face"
 LABEL repository="diffusers"
 
@@ -22,14 +22,14 @@ RUN python3 -m venv /opt/venv
 ENV PATH="/opt/venv/bin:$PATH"
 
 # pre-install the heavy dependencies (these can later be overridden by the deps from setup.py)
-RUN python3 -m pip install --no-cache-dir --upgrade pip && \
-    python3 -m pip install --no-cache-dir \
-        torch==2.1.2 \
-        torchvision==0.16.2 \
-        torchaudio==2.1.2 \
+RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
+    python3 -m uv pip install --no-cache-dir \
+        torch \
+        torchvision \
+        torchaudio \
         "onnxruntime-gpu>=1.13.1" \
         --extra-index-url https://download.pytorch.org/whl/cu117 && \
-    python3 -m pip install --no-cache-dir \
+    python3 -m uv pip install --no-cache-dir \
         accelerate \
         datasets \
         hf-doc-builder \

docker/diffusers-pytorch-compile-cuda/Dockerfile

@@ -24,8 +24,8 @@ RUN python3.9 -m venv /opt/venv
 ENV PATH="/opt/venv/bin:$PATH"
 
 # pre-install the heavy dependencies (these can later be overridden by the deps from setup.py)
-RUN python3.9 -m pip install --no-cache-dir --upgrade pip && \
-    python3.9 -m pip install --no-cache-dir \
+RUN python3.9 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
+    python3.9 -m uv pip install --no-cache-dir \
     torch \
     torchvision \
     torchaudio \

docker/diffusers-pytorch-cpu/Dockerfile

@@ -23,14 +23,14 @@ RUN python3 -m venv /opt/venv
 ENV PATH="/opt/venv/bin:$PATH"
 
 # pre-install the heavy dependencies (these can later be overridden by the deps from setup.py)
-RUN python3 -m pip install --no-cache-dir --upgrade pip && \
-    python3 -m pip install --no-cache-dir \
+RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
+    python3 -m uv pip install --no-cache-dir \
         torch \
         torchvision \
         torchaudio \
         invisible_watermark \
         --extra-index-url https://download.pytorch.org/whl/cpu && \
-    python3 -m pip install --no-cache-dir \
+    python3 -m uv pip install --no-cache-dir \
         accelerate \
         datasets \
         hf-doc-builder \
@@ -40,6 +40,6 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip && \
         numpy \
         scipy \
         tensorboard \
-        transformers
+        transformers matplotlib
 
 CMD ["/bin/bash"]

docker/diffusers-pytorch-cuda/Dockerfile

@@ -23,8 +23,8 @@ RUN python3 -m venv /opt/venv
 ENV PATH="/opt/venv/bin:$PATH"
 
 # pre-install the heavy dependencies (these can later be overridden by the deps from setup.py)
-RUN python3 -m pip install --no-cache-dir --upgrade pip && \
-    python3 -m pip install --no-cache-dir \
+RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
+    python3 -m uv pip install --no-cache-dir \
     torch \
     torchvision \
     torchaudio \

docker/diffusers-pytorch-xformers-cuda/Dockerfile

@@ -23,13 +23,13 @@ RUN python3 -m venv /opt/venv
 ENV PATH="/opt/venv/bin:$PATH"
 
 # pre-install the heavy dependencies (these can later be overridden by the deps from setup.py)
-RUN python3 -m pip install --no-cache-dir --upgrade pip && \
+RUN python3 -m pip install --no-cache-dir --upgrade pip uv==0.1.11 && \
     python3 -m pip install --no-cache-dir \
         torch \
         torchvision \
         torchaudio \
         invisible_watermark && \
-    python3 -m pip install --no-cache-dir \
+    python3 -m uv pip install --no-cache-dir \
         accelerate \
         datasets \
         hf-doc-builder \

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
@@ -52,6 +52,8 @@
       title: Image-to-image
     - local: using-diffusers/inpaint
       title: Inpainting
+    - local: using-diffusers/text-img2vid
+      title: Text or image-to-video
     - local: using-diffusers/depth2img
       title: Depth-to-image
     title: Tasks
@@ -60,6 +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
@@ -100,6 +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
@@ -300,6 +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
@@ -316,6 +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
@@ -323,6 +333,8 @@
         title: Text-to-image
       - local: api/pipelines/stable_diffusion/img2img
         title: Image-to-image
+      - local: api/pipelines/stable_diffusion/svd
+        title: Image-to-video
       - local: api/pipelines/stable_diffusion/inpaint
         title: Inpainting
       - local: api/pipelines/stable_diffusion/depth2img
@@ -392,6 +404,10 @@
       title: EulerAncestralDiscreteScheduler
     - local: api/schedulers/euler
       title: EulerDiscreteScheduler
+    - local: api/schedulers/edm_euler
+      title: EDMEulerScheduler
+    - local: api/schedulers/edm_multistep_dpm_solver
+      title: EDMDPMSolverMultistepScheduler
     - local: api/schedulers/heun
       title: HeunDiscreteScheduler
     - local: api/schedulers/ipndm
@@ -414,6 +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/attnprocessor.md

@@ -20,14 +20,14 @@ An attention processor is a class for applying different types of attention mech
 ## AttnProcessor2_0
 [[autodoc]] models.attention_processor.AttnProcessor2_0
 
-## FusedAttnProcessor2_0
-[[autodoc]] models.attention_processor.FusedAttnProcessor2_0
+## AttnAddedKVProcessor
+[[autodoc]] models.attention_processor.AttnAddedKVProcessor
 
-## LoRAAttnProcessor
-[[autodoc]] models.attention_processor.LoRAAttnProcessor
+## AttnAddedKVProcessor2_0
+[[autodoc]] models.attention_processor.AttnAddedKVProcessor2_0
 
-## LoRAAttnProcessor2_0
-[[autodoc]] models.attention_processor.LoRAAttnProcessor2_0
+## CrossFrameAttnProcessor
+[[autodoc]] pipelines.text_to_video_synthesis.pipeline_text_to_video_zero.CrossFrameAttnProcessor
 
 ## CustomDiffusionAttnProcessor
 [[autodoc]] models.attention_processor.CustomDiffusionAttnProcessor
@@ -35,26 +35,23 @@ An attention processor is a class for applying different types of attention mech
 ## CustomDiffusionAttnProcessor2_0
 [[autodoc]] models.attention_processor.CustomDiffusionAttnProcessor2_0
 
-## AttnAddedKVProcessor
-[[autodoc]] models.attention_processor.AttnAddedKVProcessor
+## CustomDiffusionXFormersAttnProcessor
+[[autodoc]] models.attention_processor.CustomDiffusionXFormersAttnProcessor
 
-## AttnAddedKVProcessor2_0
-[[autodoc]] models.attention_processor.AttnAddedKVProcessor2_0
+## FusedAttnProcessor2_0
+[[autodoc]] models.attention_processor.FusedAttnProcessor2_0
 
 ## LoRAAttnAddedKVProcessor
 [[autodoc]] models.attention_processor.LoRAAttnAddedKVProcessor
 
-## XFormersAttnProcessor
-[[autodoc]] models.attention_processor.XFormersAttnProcessor
-
 ## LoRAXFormersAttnProcessor
 [[autodoc]] models.attention_processor.LoRAXFormersAttnProcessor
 
-## CustomDiffusionXFormersAttnProcessor
-[[autodoc]] models.attention_processor.CustomDiffusionXFormersAttnProcessor
-
 ## SlicedAttnProcessor
 [[autodoc]] models.attention_processor.SlicedAttnProcessor
 
 ## SlicedAttnAddedKVProcessor
 [[autodoc]] models.attention_processor.SlicedAttnAddedKVProcessor
+
+## XFormersAttnProcessor
+[[autodoc]] models.attention_processor.XFormersAttnProcessor

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/models/consistency_decoder_vae.md

@@ -1,6 +1,18 @@
+<!--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.
+-->
+
 # Consistency Decoder
 
-Consistency decoder can be used to decode the latents from the denoising UNet in the [`StableDiffusionPipeline`]. This decoder was introduced in the [DALL-E 3 technical report](https://openai.com/dall-e-3). 
+Consistency decoder can be used to decode the latents from the denoising UNet in the [`StableDiffusionPipeline`]. This decoder was introduced in the [DALL-E 3 technical report](https://openai.com/dall-e-3).
 
 The original codebase can be found at [openai/consistencydecoder](https://github.com/openai/consistencydecoder).
 

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

@@ -408,6 +408,91 @@ Make sure to check out the Schedulers [guide](../../using-diffusers/schedulers)
 
 </Tip>
 
+## Using AnimateLCM
+
+[AnimateLCM](https://animatelcm.github.io/) is a motion module checkpoint and an [LCM LoRA](https://huggingface.co/docs/diffusers/using-diffusers/inference_with_lcm_lora) that have been created using a consistency learning strategy that decouples the distillation of the image generation priors and the motion generation priors.
+
+```python
+import torch
+from diffusers import AnimateDiffPipeline, LCMScheduler, MotionAdapter
+from diffusers.utils import export_to_gif
+
+adapter = MotionAdapter.from_pretrained("wangfuyun/AnimateLCM")
+pipe = AnimateDiffPipeline.from_pretrained("emilianJR/epiCRealism", motion_adapter=adapter)
+pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config, beta_schedule="linear")
+
+pipe.load_lora_weights("wangfuyun/AnimateLCM", weight_name="sd15_lora_beta.safetensors", adapter_name="lcm-lora")
+
+pipe.enable_vae_slicing()
+pipe.enable_model_cpu_offload()
+
+output = pipe(
+    prompt="A space rocket with trails of smoke behind it launching into space from the desert, 4k, high resolution",
+    negative_prompt="bad quality, worse quality, low resolution",
+    num_frames=16,
+    guidance_scale=1.5,
+    num_inference_steps=6,
+    generator=torch.Generator("cpu").manual_seed(0),
+)
+frames = output.frames[0]
+export_to_gif(frames, "animatelcm.gif")
+```
+
+<table>
+    <tr>
+        <td><center>
+        A space rocket, 4K.
+        <br>
+        <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatelcm-output.gif"
+            alt="A space rocket, 4K"
+            style="width: 300px;" />
+        </center></td>
+    </tr>
+</table>
+
+AnimateLCM is also compatible with existing [Motion LoRAs](https://huggingface.co/collections/dn6/animatediff-motion-loras-654cb8ad732b9e3cf4d3c17e).
+
+```python
+import torch
+from diffusers import AnimateDiffPipeline, LCMScheduler, MotionAdapter
+from diffusers.utils import export_to_gif
+
+adapter = MotionAdapter.from_pretrained("wangfuyun/AnimateLCM")
+pipe = AnimateDiffPipeline.from_pretrained("emilianJR/epiCRealism", motion_adapter=adapter)
+pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config, beta_schedule="linear")
+
+pipe.load_lora_weights("wangfuyun/AnimateLCM", weight_name="sd15_lora_beta.safetensors", adapter_name="lcm-lora")
+pipe.load_lora_weights("guoyww/animatediff-motion-lora-tilt-up", adapter_name="tilt-up")
+
+pipe.set_adapters(["lcm-lora", "tilt-up"], [1.0, 0.8])
+pipe.enable_vae_slicing()
+pipe.enable_model_cpu_offload()
+
+output = pipe(
+    prompt="A space rocket with trails of smoke behind it launching into space from the desert, 4k, high resolution",
+    negative_prompt="bad quality, worse quality, low resolution",
+    num_frames=16,
+    guidance_scale=1.5,
+    num_inference_steps=6,
+    generator=torch.Generator("cpu").manual_seed(0),
+)
+frames = output.frames[0]
+export_to_gif(frames, "animatelcm-motion-lora.gif")
+```
+
+<table>
+    <tr>
+        <td><center>
+        A space rocket, 4K.
+        <br>
+        <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatelcm-motion-lora.gif"
+            alt="A space rocket, 4K"
+            style="width: 300px;" />
+        </center></td>
+    </tr>
+</table>
+
+
 ## AnimateDiffPipeline
 
 [[autodoc]] AnimateDiffPipeline

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/sdxl_turbo.md

@@ -21,7 +21,7 @@ The abstract from the paper is:
 ## Tips
 
 - SDXL Turbo uses the exact same architecture as [SDXL](./stable_diffusion_xl), which means it also has the same API. Please refer to the [SDXL](./stable_diffusion_xl) API reference for more details.
-- SDXL Turbo should disable guidance scale by setting `guidance_scale=0.0`
+- SDXL Turbo should disable guidance scale by setting `guidance_scale=0.0`.
 - SDXL Turbo should use `timestep_spacing='trailing'` for the scheduler and use between 1 and 4 steps.
 - SDXL Turbo has been trained to generate images of size 512x512.
 - SDXL Turbo is open-access, but not open-source meaning that one might have to buy a model license in order to use it for commercial applications. Make sure to read the [official model card](https://huggingface.co/stabilityai/sdxl-turbo) to learn more.

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

@@ -0,0 +1,43 @@
+<!--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 Video Diffusion
+
+Stable Video Diffusion was proposed in [Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets](https://hf.co/papers/2311.15127) by Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram Voleti, Adam Letts, Varun Jampani, Robin Rombach.
+
+The abstract from the paper is:
+
+*We present Stable Video Diffusion - a latent video diffusion model for high-resolution, state-of-the-art text-to-video and image-to-video generation. Recently, latent diffusion models trained for 2D image synthesis have been turned into generative video models by inserting temporal layers and finetuning them on small, high-quality video datasets. However, training methods in the literature vary widely, and the field has yet to agree on a unified strategy for curating video data. In this paper, we identify and evaluate three different stages for successful training of video LDMs: text-to-image pretraining, video pretraining, and high-quality video finetuning. Furthermore, we demonstrate the necessity of a well-curated pretraining dataset for generating high-quality videos and present a systematic curation process to train a strong base model, including captioning and filtering strategies. We then explore the impact of finetuning our base model on high-quality data and train a text-to-video model that is competitive with closed-source video generation. We also show that our base model provides a powerful motion representation for downstream tasks such as image-to-video generation and adaptability to camera motion-specific LoRA modules. Finally, we demonstrate that our model provides a strong multi-view 3D-prior and can serve as a base to finetune a multi-view diffusion model that jointly generates multiple views of objects in a feedforward fashion, outperforming image-based methods at a fraction of their compute budget. We release code and model weights at this https URL.*
+
+<Tip>
+
+To learn how to use Stable Video Diffusion, take a look at the [Stable Video Diffusion](../../../using-diffusers/svd) guide.
+
+<br>
+
+Check out the [Stability AI](https://huggingface.co/stabilityai) Hub organization for the [base](https://huggingface.co/stabilityai/stable-video-diffusion-img2vid) and [extended frame](https://huggingface.co/stabilityai/stable-video-diffusion-img2vid-xt) checkpoints!
+
+</Tip>
+
+## Tips
+
+Video generation is memory-intensive and one way to reduce your memory usage is to set `enable_forward_chunking` on the pipeline's UNet so you don't run the entire feedforward layer at once. Breaking it up into chunks in a loop is more efficient.
+
+Check out the [Text or image-to-video](text-img2vid) guide for more details about how certain parameters can affect video generation and how to optimize inference by reducing memory usage.
+
+## StableVideoDiffusionPipeline
+
+[[autodoc]] StableVideoDiffusionPipeline
+
+## StableVideoDiffusionPipelineOutput
+
+[[autodoc]] pipelines.stable_video_diffusion.StableVideoDiffusionPipelineOutput

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

@@ -167,6 +167,12 @@ Here are some sample outputs:
     </tr>
 </table>
 
+## Tips
+
+Video generation is memory-intensive and one way to reduce your memory usage is to set `enable_forward_chunking` on the pipeline's UNet so you don't run the entire feedforward layer at once. Breaking it up into chunks in a loop is more efficient.
+
+Check out the [Text or image-to-video](text-img2vid) guide for more details about how certain parameters can affect video generation and how to optimize inference by reducing memory usage.
+
 <Tip>
 
 Make sure to check out the Schedulers [guide](../../using-diffusers/schedulers) to learn how to explore the tradeoff between scheduler speed and quality, and see the [reuse components across pipelines](../../using-diffusers/loading#reuse-components-across-pipelines) section to learn how to efficiently load the same components into multiple pipelines.

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

@@ -1,9 +1,21 @@
+<!--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.
+-->
+
 # ConsistencyDecoderScheduler
 
-This scheduler is a part of the [`ConsistencyDecoderPipeline`] and was introduced in [DALL-E 3](https://openai.com/dall-e-3). 
+This scheduler is a part of the [`ConsistencyDecoderPipeline`] and was introduced in [DALL-E 3](https://openai.com/dall-e-3).
 
 The original codebase can be found at [openai/consistency_models](https://github.com/openai/consistency_models).
 
 
 ## ConsistencyDecoderScheduler
-[[autodoc]] schedulers.scheduling_consistency_decoder.ConsistencyDecoderScheduler
+[[autodoc]] schedulers.scheduling_consistency_decoder.ConsistencyDecoderScheduler

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/optimization/fp16.md

@@ -66,3 +66,9 @@ image = pipe(prompt).images[0]
 Don't use [`torch.autocast`](https://pytorch.org/docs/stable/amp.html#torch.autocast) in any of the pipelines as it can lead to black images and is always slower than pure float16 precision.
 
 </Tip>
+
+## Distilled model
+
+You could also use a distilled Stable Diffusion model and autoencoder to speed up inference. During distillation, many of the UNet's residual and attention blocks are shed to reduce the model size. The distilled model is faster and uses less memory while generating images of comparable quality to the full Stable Diffusion model.
+
+Learn more about in the [Distilled Stable Diffusion inference](../using-diffusers/distilled_sd) guide!

docs/source/en/optimization/torch2.0.md

@@ -75,6 +75,9 @@ Compilation requires some time to complete, so it is best suited for situations
 
 For more information and different options about `torch.compile`, refer to the [`torch_compile`](https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html) tutorial.
 
+> [!TIP]
+> Learn more about other ways PyTorch 2.0 can help optimize your model in the [Accelerate inference of text-to-image diffusion models](../tutorials/fast_diffusion) tutorial.
+
 ## Benchmark
 
 We conducted a comprehensive benchmark with PyTorch 2.0's efficient attention implementation and `torch.compile` across different GPUs and batch sizes for five of our most used pipelines. The code is benchmarked on 🤗 Diffusers v0.17.0.dev0 to optimize `torch.compile` usage (see [here](https://github.com/huggingface/diffusers/pull/3313) for more details).

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()
@@ -113,36 +113,50 @@ The dataset preprocessing code and training loop are found in the [`main()`](htt
 
 As with the script parameters, a walkthrough of the training script is provided in the [Text-to-image](text2image#training-script) training guide. Instead, this guide takes a look at the LoRA relevant parts of the script.
 
-The script begins by adding the [new LoRA weights](https://github.com/huggingface/diffusers/blob/dd9a5caf61f04d11c0fa9f3947b69ab0010c9a0f/examples/text_to_image/train_text_to_image_lora.py#L447) to the attention layers. This involves correctly configuring the weight size for each block in the UNet. You'll see the `rank` parameter is used to create the [`~models.attention_processor.LoRAAttnProcessor`]:
+<hfoptions id="lora">
+<hfoption id="UNet">
+
+Diffusers uses [`~peft.LoraConfig`] from the [PEFT](https://hf.co/docs/peft) library to set up the parameters of the LoRA adapter such as the rank, alpha, and which modules to insert the LoRA weights into. The adapter is added to the UNet, and only the LoRA layers are filtered for optimization in `lora_layers`.
 
 ```py
-lora_attn_procs = {}
-for name in unet.attn_processors.keys():
-    cross_attention_dim = None if name.endswith("attn1.processor") else unet.config.cross_attention_dim
-    if name.startswith("mid_block"):
-        hidden_size = unet.config.block_out_channels[-1]
-    elif name.startswith("up_blocks"):
-        block_id = int(name[len("up_blocks.")])
-        hidden_size = list(reversed(unet.config.block_out_channels))[block_id]
-    elif name.startswith("down_blocks"):
-        block_id = int(name[len("down_blocks.")])
-        hidden_size = unet.config.block_out_channels[block_id]
+unet_lora_config = LoraConfig(
+    r=args.rank,
+    lora_alpha=args.rank,
+    init_lora_weights="gaussian",
+    target_modules=["to_k", "to_q", "to_v", "to_out.0"],
+)
 
-    lora_attn_procs[name] = LoRAAttnProcessor(
-        hidden_size=hidden_size,
-        cross_attention_dim=cross_attention_dim,
-        rank=args.rank,
-    )
-
-unet.set_attn_processor(lora_attn_procs)
-lora_layers = AttnProcsLayers(unet.attn_processors)
+unet.add_adapter(unet_lora_config)
+lora_layers = filter(lambda p: p.requires_grad, unet.parameters())
 ```
 
-The [optimizer](https://github.com/huggingface/diffusers/blob/dd9a5caf61f04d11c0fa9f3947b69ab0010c9a0f/examples/text_to_image/train_text_to_image_lora.py#L519) is initialized with the `lora_layers` because these are the only weights that'll be optimized:
+</hfoption>
+<hfoption id="text encoder">
+
+Diffusers also supports finetuning the text encoder with LoRA from the [PEFT](https://hf.co/docs/peft) library when necessary such as finetuning Stable Diffusion XL (SDXL). The [`~peft.LoraConfig`] is used to configure the parameters of the LoRA adapter which are then added to the text encoder, and only the LoRA layers are filtered for training.
+
+```py
+text_lora_config = LoraConfig(
+    r=args.rank,
+    lora_alpha=args.rank,
+    init_lora_weights="gaussian",
+    target_modules=["q_proj", "k_proj", "v_proj", "out_proj"],
+)
+
+text_encoder_one.add_adapter(text_lora_config)
+text_encoder_two.add_adapter(text_lora_config)
+text_lora_parameters_one = list(filter(lambda p: p.requires_grad, text_encoder_one.parameters()))
+text_lora_parameters_two = list(filter(lambda p: p.requires_grad, text_encoder_two.parameters()))
+```
+
+</hfoption>
+</hfoptions>
+
+The [optimizer](https://github.com/huggingface/diffusers/blob/e4b8f173b97731686e290b2eb98e7f5df2b1b322/examples/text_to_image/train_text_to_image_lora.py#L529) is initialized with the `lora_layers` because these are the only weights that'll be optimized:
 
 ```py
 optimizer = optimizer_cls(
-    lora_layers.parameters(),
+    lora_layers,
     lr=args.learning_rate,
     betas=(args.adam_beta1, args.adam_beta2),
     weight_decay=args.adam_weight_decay,
@@ -156,7 +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()
@@ -164,74 +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.
-
-```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()
-pipe.unload_lora_weights()
-
-pipe = torch.compile(pipe)
-
-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]
-```
-
-## 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">
 
@@ -48,10 +51,10 @@ Create a text prompt and load an image prompt before passing them to the pipelin
 image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/ip_adapter_diner.png")
 generator = torch.Generator(device="cpu").manual_seed(0)
 images = pipeline(
-    prompt="a polar bear sitting in a chair drinking a milkshake", 
+    prompt="a polar bear sitting in a chair drinking a milkshake",
     ip_adapter_image=image,
     negative_prompt="deformed, ugly, wrong proportion, low res, bad anatomy, worst quality, low quality",
-    num_inference_steps=100, 
+    num_inference_steps=100,
     generator=generator,
 ).images
 images[0]
@@ -231,8 +234,127 @@ export_to_gif(frames, "gummy_bear.gif")
 </hfoption>
 </hfoptions>
 
+## Configure parameters
+
+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]
-> While calling `load_ip_adapter()`, pass `low_cpu_mem_usage=True` to speed up the loading time.
+> 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(
+    ip_adapter_image=image,
+    ip_adapter_image_embeds=None,
+    device="cuda",
+    num_images_per_prompt=1,
+    do_classifier_free_guidance=True,
+)
+
+torch.save(image_embeds, "image_embeds.ipadpt")
+```
+
+Now load the image embeddings by passing them to the `ip_adapter_image_embeds` parameter.
+
+```py
+image_embeds = torch.load("image_embeds.ipadpt")
+images = pipeline(
+    prompt="a polar bear sitting in a chair drinking a milkshake",
+    ip_adapter_image_embeds=image_embeds,
+    negative_prompt="deformed, ugly, wrong proportion, low res, bad anatomy, worst quality, low quality",
+    num_inference_steps=100,
+    generator=generator,
+).images
+```
+
+### 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
 
@@ -245,7 +367,8 @@ Generating accurate faces is challenging because they are complex and nuanced. D
 * [ip-adapter-full-face_sd15.safetensors](https://huggingface.co/h94/IP-Adapter/blob/main/models/ip-adapter-full-face_sd15.safetensors) is conditioned with images of cropped faces and removed backgrounds
 * [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]
+> [!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.
@@ -270,7 +393,7 @@ generator = torch.Generator(device="cpu").manual_seed(26)
 image = pipeline(
     prompt="A photo of Einstein as a chef, wearing an apron, cooking in a French restaurant",
     ip_adapter_image=image,
-    negative_prompt="lowres, bad anatomy, worst quality, low quality", 
+    negative_prompt="lowres, bad anatomy, worst quality, low quality",
     num_inference_steps=100,
     generator=generator,
 ).images[0]
@@ -304,7 +427,7 @@ from transformers import CLIPVisionModelWithProjection
 from diffusers.utils import load_image
 
 image_encoder = CLIPVisionModelWithProjection.from_pretrained(
-    "h94/IP-Adapter", 
+    "h94/IP-Adapter",
     subfolder="models/image_encoder",
     torch_dtype=torch.float16,
 )
@@ -323,8 +446,8 @@ pipeline = AutoPipelineForText2Image.from_pretrained(
 )
 pipeline.scheduler = DDIMScheduler.from_config(pipeline.scheduler.config)
 pipeline.load_ip_adapter(
-  "h94/IP-Adapter", 
-  subfolder="sdxl_models", 
+  "h94/IP-Adapter",
+  subfolder="sdxl_models",
   weight_name=["ip-adapter-plus_sdxl_vit-h.safetensors", "ip-adapter-plus-face_sdxl_vit-h.safetensors"]
 )
 pipeline.set_ip_adapter_scale([0.7, 0.3])
@@ -336,7 +459,7 @@ Load an image prompt and a folder containing images of a certain style you want
 ```py
 face_image = load_image("https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/women_input.png")
 style_folder = "https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/style_ziggy"
-style_images =  [load_image(f"{style_folder}/img{i}.png") for i in range(10)]
+style_images = [load_image(f"{style_folder}/img{i}.png") for i in range(10)]
 ```
 
 <div class="flex flex-row gap-4">
@@ -358,10 +481,11 @@ generator = torch.Generator(device="cpu").manual_seed(0)
 image = pipeline(
     prompt="wonderwoman",
     ip_adapter_image=[style_images, face_image],
-    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality", 
+    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality",
     num_inference_steps=50, num_images_per_prompt=1,
     generator=generator,
 ).images[0]
+image
 ```
 
 <div class="flex justify-center">
@@ -379,14 +503,14 @@ from diffusers import DiffusionPipeline, LCMScheduler
 import torch
 from diffusers.utils import load_image
 
-model_id =  "sd-dreambooth-library/herge-style"
+model_id = "sd-dreambooth-library/herge-style"
 lcm_lora_id = "latent-consistency/lcm-lora-sdv1-5"
 
 pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
 
 pipeline.load_ip_adapter("h94/IP-Adapter", subfolder="models", weight_name="ip-adapter_sd15.bin")
 pipeline.load_lora_weights(lcm_lora_id)
-pipeline.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
+pipeline.scheduler = LCMScheduler.from_config(pipeline.scheduler.config)
 pipeline.enable_model_cpu_offload()
 ```
 
@@ -455,13 +579,13 @@ Pass the depth map and IP-Adapter image to the pipeline to generate an image.
 ```py
 generator = torch.Generator(device="cpu").manual_seed(33)
 image = pipeline(
-    prompt="best quality, high quality", 
+    prompt="best quality, high quality",
     image=depth_map,
     ip_adapter_image=ip_adapter_image,
-    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality", 
+    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality",
     num_inference_steps=50,
     generator=generator,
-).image[0]
+).images[0]
 image
 ```
 

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.
@@ -340,9 +259,9 @@ Once loaded, you can use the pipeline with an image and text prompt to guide the
 image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/load_neg_embed.png")
 generator = torch.Generator(device="cpu").manual_seed(33)
 images = pipeline(
-    prompt='best quality, high quality, wearing sunglasses', 
+    prompt='best quality, high quality, wearing sunglasses',
     ip_adapter_image=image,
-    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality", 
+    negative_prompt="monochrome, lowres, bad anatomy, worst quality, low quality",
     num_inference_steps=50,
     generator=generator,
 ).images[0]
@@ -355,11 +274,13 @@ images
 
 ### IP-Adapter Plus
 
-IP-Adapter relies on an image encoder to generate image features. If the IP-Adapter repository contains a `image_encoder` subfolder, the image encoder is automatically loaded and registed to the pipeline. Otherwise, you'll need to explicitly load the image encoder with a [`~transformers.CLIPVisionModelWithProjection`] model and pass it to the pipeline.
+IP-Adapter relies on an image encoder to generate image features. If the IP-Adapter repository contains an `image_encoder` subfolder, the image encoder is automatically loaded and registered to the pipeline. Otherwise, you'll need to explicitly load the image encoder with a [`~transformers.CLIPVisionModelWithProjection`] model and pass it to the pipeline.
 
 This is the case for *IP-Adapter Plus* checkpoints which use the ViT-H image encoder.
 
 ```py
+from transformers import CLIPVisionModelWithProjection
+
 image_encoder = CLIPVisionModelWithProjection.from_pretrained(
     "h94/IP-Adapter",
     subfolder="models/image_encoder",

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/sdxl.md

@@ -63,11 +63,12 @@ from diffusers import StableDiffusionXLPipeline, StableDiffusionXLImg2ImgPipelin
 import torch
 
 pipeline = StableDiffusionXLPipeline.from_single_file(
-    "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/sd_xl_base_1.0.safetensors", torch_dtype=torch.float16, variant="fp16", use_safetensors=True
+    "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/sd_xl_base_1.0.safetensors", 
+    torch_dtype=torch.float16
 ).to("cuda")
 
 refiner = StableDiffusionXLImg2ImgPipeline.from_single_file(
-    "https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/blob/main/sd_xl_refiner_1.0.safetensors", torch_dtype=torch.float16, use_safetensors=True, variant="fp16"
+    "https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/blob/main/sd_xl_refiner_1.0.safetensors", torch_dtype=torch.float16
 ).to("cuda")
 ```
 

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

@@ -31,29 +31,31 @@ Before you begin, make sure you have the following libraries installed:
 Model weights may be stored in separate subfolders on the Hub or locally, in which case, you should use the [`~StableDiffusionXLPipeline.from_pretrained`] method:
 
 ```py
-from diffusers import AutoPipelineForText2Image, AutoPipelineForImage2Image
+from diffusers import AutoPipelineForText2Image
 import torch
 
 pipeline = AutoPipelineForText2Image.from_pretrained("stabilityai/sdxl-turbo", torch_dtype=torch.float16, variant="fp16")
 pipeline = pipeline.to("cuda")
 ```
 
-You can also use the [`~StableDiffusionXLPipeline.from_single_file`] method to load a model checkpoint stored in a single file format (`.ckpt` or `.safetensors`) from the Hub or locally:
+You can also use the [`~StableDiffusionXLPipeline.from_single_file`] method to load a model checkpoint stored in a single file format (`.ckpt` or `.safetensors`) from the Hub or locally. For this loading method, you need to set `timestep_spacing="trailing"` (feel free to experiment with the other scheduler config values to get better results):
 
 ```py
-from diffusers import StableDiffusionXLPipeline
+from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
 import torch
 
 pipeline = StableDiffusionXLPipeline.from_single_file(
-    "https://huggingface.co/stabilityai/sdxl-turbo/blob/main/sd_xl_turbo_1.0_fp16.safetensors", torch_dtype=torch.float16)
+    "https://huggingface.co/stabilityai/sdxl-turbo/blob/main/sd_xl_turbo_1.0_fp16.safetensors",
+    torch_dtype=torch.float16, variant="fp16")
 pipeline = pipeline.to("cuda")
+pipeline.scheduler = EulerAncestralDiscreteScheduler.from_config(pipeline.scheduler.config, timestep_spacing="trailing")
 ```
 
 ## Text-to-image
 
 For text-to-image, pass a text prompt. By default, SDXL Turbo generates a 512x512 image, and that resolution gives the best results. You can try setting the `height` and `width` parameters to 768x768 or 1024x1024, but you should expect quality degradations when doing so.
 
-Make sure to set `guidance_scale` to 0.0 to disable, as the model was trained without it. A single inference step is enough to generate high quality images. 
+Make sure to set `guidance_scale` to 0.0 to disable, as the model was trained without it. A single inference step is enough to generate high quality images.
 Increasing the number of steps to 2, 3 or 4 should improve image quality.
 
 ```py
@@ -75,7 +77,7 @@ image
 
 ## Image-to-image
 
-For image-to-image generation, make sure that `num_inference_steps * strength` is larger or equal to 1. 
+For image-to-image generation, make sure that `num_inference_steps * strength` is larger or equal to 1.
 The image-to-image pipeline will run for `int(num_inference_steps * strength)` steps, e.g. `0.5 * 2.0 = 1` step in
 our example below.
 
@@ -84,14 +86,14 @@ from diffusers import AutoPipelineForImage2Image
 from diffusers.utils import load_image, make_image_grid
 
 # use from_pipe to avoid consuming additional memory when loading a checkpoint
-pipeline = AutoPipelineForImage2Image.from_pipe(pipeline_text2image).to("cuda")
+pipeline_image2image = AutoPipelineForImage2Image.from_pipe(pipeline_text2image).to("cuda")
 
 init_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png")
 init_image = init_image.resize((512, 512))
 
 prompt = "cat wizard, gandalf, lord of the rings, detailed, fantasy, cute, adorable, Pixar, Disney, 8k"
 
-image = pipeline(prompt, image=init_image, strength=0.5, guidance_scale=0.0, num_inference_steps=2).images[0]
+image = pipeline_image2image(prompt, image=init_image, strength=0.5, guidance_scale=0.0, num_inference_steps=2).images[0]
 make_image_grid([init_image, image], rows=1, cols=2)
 ```
 
@@ -101,7 +103,7 @@ make_image_grid([init_image, image], rows=1, cols=2)
 
 ## Speed-up SDXL Turbo even more
 
-- Compile the UNet if you are using PyTorch version 2 or better. The first inference run will be very slow, but subsequent ones will be much faster.
+- Compile the UNet if you are using PyTorch version 2.0 or higher. The first inference run will be very slow, but subsequent ones will be much faster.
 
 ```py
 pipe.unet = torch.compile(pipe.unet, mode="reduce-overhead", fullgraph=True)

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

@@ -217,3 +217,9 @@ Check your image dimensions to see if they're correct:
 images.shape
 # (8, 1, 512, 512, 3)
 ```
+
+## Resources
+
+To learn more about how JAX works with Stable Diffusion, you may be interested in reading:
+
+* [Accelerating Stable Diffusion XL Inference with JAX on Cloud TPU v5e](https://hf.co/blog/sdxl_jax)

docs/source/en/using-diffusers/text-img2vid.md

@@ -0,0 +1,497 @@
+<!--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.
+-->
+
+# Text or image-to-video
+
+Driven by the success of text-to-image diffusion models, generative video models are able to generate short clips of video from a text prompt or an initial image. These models extend a pretrained diffusion model to generate videos by adding some type of temporal and/or spatial convolution layer to the architecture. A mixed dataset of images and videos are used to train the model which learns to output a series of video frames based on the text or image conditioning.
+
+This guide will show you how to generate videos, how to configure video model parameters, and how to control video generation.
+
+## Popular models
+
+> [!TIP]
+> Discover other cool and trending video generation models on the Hub [here](https://huggingface.co/models?pipeline_tag=text-to-video&sort=trending)!
+
+[Stable Video Diffusions (SVD)](https://huggingface.co/stabilityai/stable-video-diffusion-img2vid), [I2VGen-XL](https://huggingface.co/ali-vilab/i2vgen-xl/), [AnimateDiff](https://huggingface.co/guoyww/animatediff), and [ModelScopeT2V](https://huggingface.co/ali-vilab/text-to-video-ms-1.7b) are popular models used for video diffusion. Each model is distinct. For example, AnimateDiff inserts a motion modeling module into a frozen text-to-image model to generate personalized animated images, whereas SVD is entirely pretrained from scratch with a three-stage training process to generate short high-quality videos.
+
+### Stable Video Diffusion
+
+[SVD](../api/pipelines/svd) is based on the Stable Diffusion 2.1 model and it is trained on images, then low-resolution videos, and finally a smaller dataset of high-resolution videos. This model generates a short 2-4 second video from an initial image. You can learn more details about model, like micro-conditioning, in the [Stable Video Diffusion](../using-diffusers/svd) guide.
+
+Begin by loading the [`StableVideoDiffusionPipeline`] and passing an initial image to generate a video from.
+
+```py
+import torch
+from diffusers import StableVideoDiffusionPipeline
+from diffusers.utils import load_image, export_to_video
+
+pipeline = StableVideoDiffusionPipeline.from_pretrained(
+    "stabilityai/stable-video-diffusion-img2vid-xt", torch_dtype=torch.float16, variant="fp16"
+)
+pipeline.enable_model_cpu_offload()
+
+image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/svd/rocket.png")
+image = image.resize((1024, 576))
+
+generator = torch.manual_seed(42)
+frames = pipeline(image, decode_chunk_size=8, generator=generator).frames[0]
+export_to_video(frames, "generated.mp4", fps=7)
+```
+
+<div class="flex gap-4">
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/svd/rocket.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">initial image</figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/svd/output_rocket.gif"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">generated video</figcaption>
+  </div>
+</div>
+
+### I2VGen-XL
+
+[I2VGen-XL](../api/pipelines/i2vgenxl) is a diffusion model that can generate higher resolution videos than SVD and it is also capable of accepting text prompts in addition to images. The model is trained with two hierarchical encoders (detail and global encoder) to better capture low and high-level details in images. These learned details are used to train a video diffusion model which refines the video resolution and details in the generated video.
+
+You can use I2VGen-XL by loading the [`I2VGenXLPipeline`], and passing a text and image prompt to generate a video.
+
+```py
+import torch
+from diffusers import I2VGenXLPipeline
+from diffusers.utils import export_to_gif, load_image
+
+pipeline = I2VGenXLPipeline.from_pretrained("ali-vilab/i2vgen-xl", torch_dtype=torch.float16, variant="fp16")
+pipeline.enable_model_cpu_offload()
+
+image_url = "https://huggingface.co/datasets/diffusers/docs-images/resolve/main/i2vgen_xl_images/img_0009.png"
+image = load_image(image_url).convert("RGB")
+
+prompt = "Papers were floating in the air on a table in the library"
+negative_prompt = "Distorted, discontinuous, Ugly, blurry, low resolution, motionless, static, disfigured, disconnected limbs, Ugly faces, incomplete arms"
+generator = torch.manual_seed(8888)
+
+frames = pipeline(
+    prompt=prompt,
+    image=image,
+    num_inference_steps=50,
+    negative_prompt=negative_prompt,
+    guidance_scale=9.0,
+    generator=generator
+).frames[0]
+export_to_gif(frames, "i2v.gif")
+```
+
+<div class="flex gap-4">
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/diffusers/docs-images/resolve/main/i2vgen_xl_images/img_0009.png"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">initial image</figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/i2vgen-xl-example.gif"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">generated video</figcaption>
+  </div>
+</div>
+
+### AnimateDiff
+
+[AnimateDiff](../api/pipelines/animatediff) is an adapter model that inserts a motion module into a pretrained diffusion model to animate an image. The adapter is trained on video clips to learn motion which is used to condition the generation process to create a video. It is faster and easier to only train the adapter and it can be loaded into most diffusion models, effectively turning them into "video models".
+
+Start by loading a [`MotionAdapter`].
+
+```py
+import torch
+from diffusers import AnimateDiffPipeline, DDIMScheduler, MotionAdapter
+from diffusers.utils import export_to_gif
+
+adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5-2", torch_dtype=torch.float16)
+```
+
+Then load a finetuned Stable Diffusion model with the [`AnimateDiffPipeline`].
+
+```py
+pipeline = AnimateDiffPipeline.from_pretrained("emilianJR/epiCRealism", motion_adapter=adapter, torch_dtype=torch.float16)
+scheduler = DDIMScheduler.from_pretrained(
+    "emilianJR/epiCRealism",
+    subfolder="scheduler",
+    clip_sample=False,
+    timestep_spacing="linspace",
+    beta_schedule="linear",
+    steps_offset=1,
+)
+pipeline.scheduler = scheduler
+pipeline.enable_vae_slicing()
+pipeline.enable_model_cpu_offload()
+```
+
+Create a prompt and generate the video.
+
+```py
+output = pipeline(
+    prompt="A space rocket with trails of smoke behind it launching into space from the desert, 4k, high resolution",
+    negative_prompt="bad quality, worse quality, low resolution",
+    num_frames=16,
+    guidance_scale=7.5,
+    num_inference_steps=50,
+    generator=torch.Generator("cpu").manual_seed(49),
+)
+frames = output.frames[0]
+export_to_gif(frames, "animation.gif")
+```
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff.gif"/>
+</div>
+
+### ModelscopeT2V
+
+[ModelscopeT2V](../api/pipelines/text_to_video) adds spatial and temporal convolutions and attention to a UNet, and it is trained on image-text and video-text datasets to enhance what it learns during training. The model takes a prompt, encodes it and creates text embeddings which are denoised by the UNet, and then decoded by a VQGAN into a video.
+
+<Tip>
+
+ModelScopeT2V generates watermarked videos due to the datasets it was trained on. To use a watermark-free model, try the [cerspense/zeroscope_v2_76w](https://huggingface.co/cerspense/zeroscope_v2_576w) model with the [`TextToVideoSDPipeline`] first, and then upscale it's output with the [cerspense/zeroscope_v2_XL](https://huggingface.co/cerspense/zeroscope_v2_XL) checkpoint using the [`VideoToVideoSDPipeline`].
+
+</Tip>
+
+Load a ModelScopeT2V checkpoint into the [`DiffusionPipeline`] along with a prompt to generate a video.
+
+```py
+import torch
+from diffusers import DiffusionPipeline
+from diffusers.utils import export_to_video
+
+pipeline = DiffusionPipeline.from_pretrained("damo-vilab/text-to-video-ms-1.7b", torch_dtype=torch.float16, variant="fp16")
+pipeline.enable_model_cpu_offload()
+pipeline.enable_vae_slicing()
+
+prompt = "Confident teddy bear surfer rides the wave in the tropics"
+video_frames = pipeline(prompt).frames[0]
+export_to_video(video_frames, "modelscopet2v.mp4", fps=10)
+```
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/modelscopet2v.gif" />
+</div>
+
+## Configure model parameters
+
+There are a few important parameters you can configure in the pipeline that'll affect the video generation process and quality. Let's take a closer look at what these parameters do and how changing them affects the output.
+
+### Number of frames
+
+The `num_frames` parameter determines how many video frames are generated per second. A frame is an image that is played in a sequence of other frames to create motion or a video. This affects video length because the pipeline generates a certain number of frames per second (check a pipeline's API reference for the default value). To increase the video duration, you'll need to increase the `num_frames` parameter.
+
+```py
+import torch
+from diffusers import StableVideoDiffusionPipeline
+from diffusers.utils import load_image, export_to_video
+
+pipeline = StableVideoDiffusionPipeline.from_pretrained(
+    "stabilityai/stable-video-diffusion-img2vid", torch_dtype=torch.float16, variant="fp16"
+)
+pipeline.enable_model_cpu_offload()
+
+image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/svd/rocket.png")
+image = image.resize((1024, 576))
+
+generator = torch.manual_seed(42)
+frames = pipeline(image, decode_chunk_size=8, generator=generator, num_frames=25).frames[0]
+export_to_video(frames, "generated.mp4", fps=7)
+```
+
+<div class="flex gap-4">
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/num_frames_14.gif"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">num_frames=14</figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/num_frames_25.gif"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">num_frames=25</figcaption>
+  </div>
+</div>
+
+### Guidance scale
+
+The `guidance_scale` parameter controls how closely aligned the generated video and text prompt or initial image is. A higher `guidance_scale` value means your generated video is more aligned with the text prompt or initial image, while a lower `guidance_scale` value means your generated video is less aligned which could give the model more "creativity" to interpret the conditioning input.
+
+<Tip>
+
+SVD uses the `min_guidance_scale` and `max_guidance_scale` parameters for applying guidance to the first and last frames respectively.
+
+</Tip>
+
+```py
+import torch
+from diffusers import I2VGenXLPipeline
+from diffusers.utils import export_to_gif, load_image
+
+pipeline = I2VGenXLPipeline.from_pretrained("ali-vilab/i2vgen-xl", torch_dtype=torch.float16, variant="fp16")
+pipeline.enable_model_cpu_offload()
+
+image_url = "https://huggingface.co/datasets/diffusers/docs-images/resolve/main/i2vgen_xl_images/img_0009.png"
+image = load_image(image_url).convert("RGB")
+
+prompt = "Papers were floating in the air on a table in the library"
+negative_prompt = "Distorted, discontinuous, Ugly, blurry, low resolution, motionless, static, disfigured, disconnected limbs, Ugly faces, incomplete arms"
+generator = torch.manual_seed(0)
+
+frames = pipeline(
+    prompt=prompt,
+    image=image,
+    num_inference_steps=50,
+    negative_prompt=negative_prompt,
+    guidance_scale=1.0,
+    generator=generator
+).frames[0]
+export_to_gif(frames, "i2v.gif")
+```
+
+<div class="flex gap-4">
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/i2vgen-xl-example.gif"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">guidance_scale=9.0</figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/guidance_scale_1.0.gif"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">guidance_scale=1.0</figcaption>
+  </div>
+</div>
+
+### Negative prompt
+
+A negative prompt deters the model from generating things you don’t want it to. This parameter is commonly used to improve overall generation quality by removing poor or bad features such as “low resolution” or “bad details”.
+
+```py
+import torch
+from diffusers import AnimateDiffPipeline, DDIMScheduler, MotionAdapter
+from diffusers.utils import export_to_gif
+
+adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5-2", torch_dtype=torch.float16)
+
+pipeline = AnimateDiffPipeline.from_pretrained("emilianJR/epiCRealism", motion_adapter=adapter, torch_dtype=torch.float16)
+scheduler = DDIMScheduler.from_pretrained(
+    "emilianJR/epiCRealism",
+    subfolder="scheduler",
+    clip_sample=False,
+    timestep_spacing="linspace",
+    beta_schedule="linear",
+    steps_offset=1,
+)
+pipeline.scheduler = scheduler
+pipeline.enable_vae_slicing()
+pipeline.enable_model_cpu_offload()
+
+output = pipeline(
+    prompt="360 camera shot of a sushi roll in a restaurant",
+    negative_prompt="Distorted, discontinuous, ugly, blurry, low resolution, motionless, static",
+    num_frames=16,
+    guidance_scale=7.5,
+    num_inference_steps=50,
+    generator=torch.Generator("cpu").manual_seed(0),
+)
+frames = output.frames[0]
+export_to_gif(frames, "animation.gif")
+```
+
+<div class="flex gap-4">
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff_no_neg.gif"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">no negative prompt</figcaption>
+  </div>
+  <div>
+    <img class="rounded-xl" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff_neg.gif"/>
+    <figcaption class="mt-2 text-center text-sm text-gray-500">negative prompt applied</figcaption>
+  </div>
+</div>
+
+### Model-specific parameters
+
+There are some pipeline parameters that are unique to each model such as adjusting the motion in a video or adding noise to the initial image.
+
+<hfoptions id="special-parameters">
+<hfoption id="Stable Video Diffusion">
+
+Stable Video Diffusion provides additional micro-conditioning for the frame rate with the `fps` parameter and for motion with the `motion_bucket_id` parameter. Together, these parameters allow for adjusting the amount of motion in the generated video.
+
+There is also a `noise_aug_strength` parameter that increases the amount of noise added to the initial image. Varying this parameter affects how similar the generated video and initial image are. A higher `noise_aug_strength` also increases the amount of motion. To learn more, read the [Micro-conditioning](../using-diffusers/svd#micro-conditioning) guide.
+
+</hfoption>
+<hfoption id="Text2Video-Zero">
+
+Text2Video-Zero computes the amount of motion to apply to each frame from randomly sampled latents. You can use the `motion_field_strength_x` and `motion_field_strength_y` parameters to control the amount of motion to apply to the x and y-axes of the video. The parameters `t0` and `t1` are the timesteps to apply motion to the latents.
+
+</hfoption>
+</hfoptions>
+
+## Control video generation
+
+Video generation can be controlled similar to how text-to-image, image-to-image, and inpainting can be controlled with a [`ControlNetModel`]. The only difference is you need to use the [`~pipelines.text_to_video_synthesis.pipeline_text_to_video_zero.CrossFrameAttnProcessor`] so each frame attends to the first frame.
+
+### Text2Video-Zero
+
+Text2Video-Zero video generation can be conditioned on pose and edge images for even greater control over a subject's motion in the generated video or to preserve the identity of a subject/object in the video. You can also use Text2Video-Zero with [InstructPix2Pix](../api/pipelines/pix2pix) for editing videos with text.
+
+<hfoptions id="t2v-zero">
+<hfoption id="pose control">
+
+Start by downloading a video and extracting the pose images from it.
+
+```py
+from huggingface_hub import hf_hub_download
+from PIL import Image
+import imageio
+
+filename = "__assets__/poses_skeleton_gifs/dance1_corr.mp4"
+repo_id = "PAIR/Text2Video-Zero"
+video_path = hf_hub_download(repo_type="space", repo_id=repo_id, filename=filename)
+
+reader = imageio.get_reader(video_path, "ffmpeg")
+frame_count = 8
+pose_images = [Image.fromarray(reader.get_data(i)) for i in range(frame_count)]
+```
+
+Load a [`ControlNetModel`] for pose estimation and a checkpoint into the [`StableDiffusionControlNetPipeline`]. Then you'll use the [`~pipelines.text_to_video_synthesis.pipeline_text_to_video_zero.CrossFrameAttnProcessor`] for the UNet and ControlNet.
+
+```py
+import torch
+from diffusers import StableDiffusionControlNetPipeline, ControlNetModel
+from diffusers.pipelines.text_to_video_synthesis.pipeline_text_to_video_zero import CrossFrameAttnProcessor
+
+model_id = "runwayml/stable-diffusion-v1-5"
+controlnet = ControlNetModel.from_pretrained("lllyasviel/sd-controlnet-openpose", torch_dtype=torch.float16)
+pipeline = StableDiffusionControlNetPipeline.from_pretrained(
+    model_id, controlnet=controlnet, torch_dtype=torch.float16
+).to("cuda")
+
+pipeline.unet.set_attn_processor(CrossFrameAttnProcessor(batch_size=2))
+pipeline.controlnet.set_attn_processor(CrossFrameAttnProcessor(batch_size=2))
+```
+
+Fix the latents for all the frames, and then pass your prompt and extracted pose images to the model to generate a video.
+
+```py
+latents = torch.randn((1, 4, 64, 64), device="cuda", dtype=torch.float16).repeat(len(pose_images), 1, 1, 1)
+
+prompt = "Darth Vader dancing in a desert"
+result = pipeline(prompt=[prompt] * len(pose_images), image=pose_images, latents=latents).images
+imageio.mimsave("video.mp4", result, fps=4)
+```
+
+</hfoption>
+<hfoption id="edge control">
+
+Download a video and extract the edges from it.
+
+```py
+from huggingface_hub import hf_hub_download
+from PIL import Image
+import imageio
+
+filename = "__assets__/poses_skeleton_gifs/dance1_corr.mp4"
+repo_id = "PAIR/Text2Video-Zero"
+video_path = hf_hub_download(repo_type="space", repo_id=repo_id, filename=filename)
+
+reader = imageio.get_reader(video_path, "ffmpeg")
+frame_count = 8
+pose_images = [Image.fromarray(reader.get_data(i)) for i in range(frame_count)]
+```
+
+Load a [`ControlNetModel`] for canny edge and a checkpoint into the [`StableDiffusionControlNetPipeline`]. Then you'll use the [`~pipelines.text_to_video_synthesis.pipeline_text_to_video_zero.CrossFrameAttnProcessor`] for the UNet and ControlNet.
+
+```py
+import torch
+from diffusers import StableDiffusionControlNetPipeline, ControlNetModel
+from diffusers.pipelines.text_to_video_synthesis.pipeline_text_to_video_zero import CrossFrameAttnProcessor
+
+model_id = "runwayml/stable-diffusion-v1-5"
+controlnet = ControlNetModel.from_pretrained("lllyasviel/sd-controlnet-canny", torch_dtype=torch.float16)
+pipeline = StableDiffusionControlNetPipeline.from_pretrained(
+    model_id, controlnet=controlnet, torch_dtype=torch.float16
+).to("cuda")
+
+pipeline.unet.set_attn_processor(CrossFrameAttnProcessor(batch_size=2))
+pipeline.controlnet.set_attn_processor(CrossFrameAttnProcessor(batch_size=2))
+```
+
+Fix the latents for all the frames, and then pass your prompt and extracted edge images to the model to generate a video.
+
+```py
+latents = torch.randn((1, 4, 64, 64), device="cuda", dtype=torch.float16).repeat(len(pose_images), 1, 1, 1)
+
+prompt = "Darth Vader dancing in a desert"
+result = pipeline(prompt=[prompt] * len(pose_images), image=pose_images, latents=latents).images
+imageio.mimsave("video.mp4", result, fps=4)
+```
+
+</hfoption>
+<hfoption id="InstructPix2Pix">
+
+InstructPix2Pix allows you to use text to describe the changes you want to make to the video. Start by downloading and reading a video.
+
+```py
+from huggingface_hub import hf_hub_download
+from PIL import Image
+import imageio
+
+filename = "__assets__/pix2pix video/camel.mp4"
+repo_id = "PAIR/Text2Video-Zero"
+video_path = hf_hub_download(repo_type="space", repo_id=repo_id, filename=filename)
+
+reader = imageio.get_reader(video_path, "ffmpeg")
+frame_count = 8
+video = [Image.fromarray(reader.get_data(i)) for i in range(frame_count)]
+```
+
+Load the [`StableDiffusionInstructPix2PixPipeline`] and set the [`~pipelines.text_to_video_synthesis.pipeline_text_to_video_zero.CrossFrameAttnProcessor`] for the UNet.
+
+```py
+import torch
+from diffusers import StableDiffusionInstructPix2PixPipeline
+from diffusers.pipelines.text_to_video_synthesis.pipeline_text_to_video_zero import CrossFrameAttnProcessor
+
+pipeline = StableDiffusionInstructPix2PixPipeline.from_pretrained("timbrooks/instruct-pix2pix", torch_dtype=torch.float16).to("cuda")
+pipeline.unet.set_attn_processor(CrossFrameAttnProcessor(batch_size=3))
+```
+
+Pass a prompt describing the change you want to apply to the video.
+
+```py
+prompt = "make it Van Gogh Starry Night style"
+result = pipeline(prompt=[prompt] * len(video), image=video).images
+imageio.mimsave("edited_video.mp4", result, fps=4)
+```
+
+</hfoption>
+</hfoptions>
+
+## Optimize
+
+Video generation requires a lot of memory because you're generating many video frames at once. You can reduce your memory requirements at the expense of some inference speed. Try:
+
+1. offloading pipeline components that are no longer needed to the CPU
+2. feed-forward chunking runs the feed-forward layer in a loop instead of all at once
+3. break up the number of frames the VAE has to decode into chunks instead of decoding them all at once
+
+```diff
+- pipeline.enable_model_cpu_offload()
+- frames = pipeline(image, decode_chunk_size=8, generator=generator).frames[0]
++ pipeline.enable_model_cpu_offload()
++ pipeline.unet.enable_forward_chunking()
++ frames = pipeline(image, decode_chunk_size=2, generator=generator, num_frames=25).frames[0]
+```
+
+If memory is not an issue and you want to optimize for speed, try wrapping the UNet with [`torch.compile`](../optimization/torch2.0#torchcompile).
+
+```diff
+- pipeline.enable_model_cpu_offload()
++ pipeline.to("cuda")
++ pipeline.unet = torch.compile(pipeline.unet, mode="reduce-overhead", fullgraph=True)
+```

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()
->>> images = (image * 255).round().astype("uint8")
 >>> image = Image.fromarray(image)
 >>> image
 ```

docs/source/ko/api/pipelines/stable_diffusion/stable_diffusion_xl.md

@@ -313,12 +313,12 @@ from diffusers import StableDiffusionXLPipeline, StableDiffusionXLImg2ImgPipelin
 import torch
 
 pipe = StableDiffusionXLPipeline.from_single_file(
-    "./sd_xl_base_1.0.safetensors", torch_dtype=torch.float16, variant="fp16", use_safetensors=True
+    "./sd_xl_base_1.0.safetensors", torch_dtype=torch.float16
 )
 pipe.to("cuda")
 
 refiner = StableDiffusionXLImg2ImgPipeline.from_single_file(
-    "./sd_xl_refiner_1.0.safetensors", torch_dtype=torch.float16, use_safetensors=True, variant="fp16"
+    "./sd_xl_refiner_1.0.safetensors", torch_dtype=torch.float16
 )
 refiner.to("cuda")
 ```

examples/advanced_diffusion_training/README.md

@@ -80,8 +80,7 @@ To do so, just specify `--train_text_encoder_ti` while launching training (for r
 Please keep the following points in mind:
 
 * SDXL has two text encoders. So, we fine-tune both using LoRA.
-* When not fine-tuning the text encoders, we ALWAYS precompute the text embeddings to save memoםהקרry.
-
+* When not fine-tuning the text encoders, we ALWAYS precompute the text embeddings to save memory.
 
 ### 3D icon example
 
@@ -234,6 +233,76 @@ In ComfyUI we will load a LoRA and a textual embedding at the same time.
 
 SDXL's VAE is known to suffer from numerical instability issues. This is why we also expose a CLI argument namely `--pretrained_vae_model_name_or_path` that lets you specify the location of a better VAE (such as [this one](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix)).
 
+### DoRA training 
+The advanced 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 🤗
+
+## 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("0.27.0")
 
 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}
@@ -651,6 +654,16 @@ def parse_args(input_args=None):
         default=4,
         help=("The dimension of the LoRA update matrices."),
     )
+    parser.add_argument(
+        "--use_dora",
+        type=bool,
+        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`"
+        ),
+    )
     parser.add_argument(
         "--cache_latents",
         action="store_true",
@@ -1202,7 +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."
                 )
@@ -1219,6 +1232,7 @@ def main(args):
     unet_lora_config = LoraConfig(
         r=args.rank,
         lora_alpha=args.rank,
+        use_dora=args.use_dora,
         init_lora_weights="gaussian",
         target_modules=["to_k", "to_q", "to_v", "to_out.0"],
     )
@@ -1230,6 +1244,7 @@ def main(args):
         text_lora_config = LoraConfig(
             r=args.rank,
             lora_alpha=args.rank,
+            use_dora=args.use_dora,
             init_lora_weights="gaussian",
             target_modules=["q_proj", "k_proj", "v_proj", "out_proj"],
         )
@@ -1351,14 +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()}"
         )
@@ -1392,11 +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."
@@ -1955,6 +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("0.27.0")
 
 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,
@@ -661,6 +687,15 @@ 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`"
+        ),
+    )
     parser.add_argument(
         "--cache_latents",
         action="store_true",
@@ -1105,6 +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)
 
@@ -1222,7 +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
     )
@@ -1240,7 +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
     )
@@ -1305,7 +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."
                 )
@@ -1323,6 +1377,7 @@ def main(args):
     unet_lora_config = LoraConfig(
         r=args.rank,
         lora_alpha=args.rank,
+        use_dora=args.use_dora,
         init_lora_weights="gaussian",
         target_modules=["to_k", "to_q", "to_v", "to_out.0"],
     )
@@ -1334,6 +1389,7 @@ def main(args):
         text_lora_config = LoraConfig(
             r=args.rank,
             lora_alpha=args.rank,
+            use_dora=args.use_dora,
             init_lora_weights="gaussian",
             target_modules=["q_proj", "k_proj", "v_proj", "out_proj"],
         )
@@ -1508,14 +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()}"
         )
@@ -1549,11 +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."
@@ -1776,6 +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
@@ -1827,9 +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)
@@ -1840,15 +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(
@@ -1874,7 +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,
@@ -1892,14 +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}")
 
@@ -1909,10 +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.
@@ -2035,17 +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)
@@ -2053,8 +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)
@@ -2130,15 +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)
@@ -2192,6 +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

@@ -57,12 +57,13 @@ If a community doesn't work as expected, please open an issue and ping the autho
 |   DemoFusion Pipeline                                                                                                    | Implementation of [DemoFusion: Democratising High-Resolution Image Generation With No $$$](https://arxiv.org/abs/2311.16973)                                                                                                                                                                                                                                                                                                                                                                                                                                      | [DemoFusion Pipeline](#DemoFusion)      | - |              [Ruoyi Du](https://github.com/RuoyiDu) |
 |   Instaflow Pipeline                                                                                                    | Implementation of [InstaFlow! One-Step Stable Diffusion with Rectified Flow](https://arxiv.org/abs/2309.06380)                                                                                                                                                                                                                                                                                                                                                                                                                                      | [Instaflow Pipeline](#instaflow-pipeline)      | - |              [Ayush Mangal](https://github.com/ayushtues) |
 |   Null-Text Inversion Pipeline  | Implement [Null-text Inversion for Editing Real Images using Guided Diffusion Models](https://arxiv.org/abs/2211.09794) as a pipeline.                                                                                                                                                                                                                                                                                                                                                                                                                                      | [Null-Text Inversion](https://github.com/google/prompt-to-prompt/)      | - |              [Junsheng Luan](https://github.com/Junsheng121) |
-|   Rerender A Video Pipeline                                                                                                    | Implementation of [[SIGGRAPH Asia 2023] Rerender A Video: Zero-Shot Text-Guided Video-to-Video Translation](https://arxiv.org/abs/2306.07954)                                                                                                                                                                                                                                                                                                                                                                                                                                      | [Rerender A Video Pipeline](#Rerender_A_Video)      | - |              [Yifan Zhou](https://github.com/SingleZombie) |
+|   Rerender A Video Pipeline                                                                                                    | Implementation of [[SIGGRAPH Asia 2023] Rerender A Video: Zero-Shot Text-Guided Video-to-Video Translation](https://arxiv.org/abs/2306.07954)                                                                                                                                                                                                                                                                                                                                                                                                                                      | [Rerender A Video Pipeline](#Rerender-A-Video)      | - |              [Yifan Zhou](https://github.com/SingleZombie) |
 | StyleAligned Pipeline                                                                                                    | Implementation of [Style Aligned Image Generation via Shared Attention](https://arxiv.org/abs/2312.02133)                                                                                                                                                                                                                                                                                                                                                                                                                                   | [StyleAligned Pipeline](#stylealigned-pipeline) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://drive.google.com/file/d/15X2E0jFPTajUIjS0FzX50OaHsCbP2lQ0/view?usp=sharing) | [Aryan V S](https://github.com/a-r-r-o-w) |
 | AnimateDiff Image-To-Video Pipeline | Experimental Image-To-Video support for AnimateDiff (open to improvements) | [AnimateDiff Image To Video Pipeline](#animatediff-image-to-video-pipeline) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://drive.google.com/file/d/1TvzCDPHhfFtdcJZe4RLloAwyoLKuttWK/view?usp=sharing) | [Aryan V S](https://github.com/a-r-r-o-w) |
 |   IP Adapter FaceID Stable Diffusion                                                                                               | Stable Diffusion Pipeline that supports IP Adapter Face ID                                                                                                                                                                                                                                                                                                                                                  |  [IP Adapter Face ID](#ip-adapter-face-id) | - | [Fabio Rigano](https://github.com/fabiorigano) |
 |   InstantID Pipeline                                                                                               | Stable Diffusion XL Pipeline that supports InstantID                                                                                                                                                                                                                                                                                                                                                 |  [InstantID Pipeline](#instantid-pipeline) | [![Hugging Face Space](https://img.shields.io/badge/🤗%20Hugging%20Face-Space-yellow)](https://huggingface.co/spaces/InstantX/InstantID) | [Haofan Wang](https://github.com/haofanwang) |
 |   UFOGen Scheduler                                                                                               | Scheduler for UFOGen Model (compatible with Stable Diffusion pipelines)                                                                                                                                                                                                                                                                                                                                                 |  [UFOGen Scheduler](#ufogen-scheduler) | - | [dg845](https://github.com/dg845) |
+| Stable Diffusion XL IPEX Pipeline | Accelerate Stable Diffusion XL inference pipeline with BF16/FP32 precision on Intel Xeon CPUs with [IPEX](https://github.com/intel/intel-extension-for-pytorch) | [Stable Diffusion XL on IPEX](#stable-diffusion-xl-on-ipex) | - | [Dan Li](https://github.com/ustcuna/) |
 
 To load a custom pipeline you just need to pass the `custom_pipeline` argument to `DiffusionPipeline`, as one of the files in `diffusers/examples/community`. Feel free to send a PR with your own pipelines, we will merge them quickly.
 
@@ -104,7 +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.
 )
 
@@ -749,7 +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)
 
@@ -781,9 +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")
 ```
@@ -1707,6 +1708,111 @@ print("Latency of StableDiffusionPipeline--fp32",latency)
 
 ```
 
+### Stable Diffusion XL on IPEX
+
+This diffusion pipeline aims to accelarate the inference of Stable-Diffusion XL on Intel Xeon CPUs with BF16/FP32 precision using [IPEX](https://github.com/intel/intel-extension-for-pytorch).
+
+To use this pipeline, you need to:
+1. Install [IPEX](https://github.com/intel/intel-extension-for-pytorch)
+
+**Note:** For each PyTorch release, there is a corresponding release of IPEX. Here is the mapping relationship. It is recommended to install Pytorch/IPEX2.0 to get the best performance.
+
+|PyTorch Version|IPEX Version|
+|--|--|
+|[v2.0.\*](https://github.com/pytorch/pytorch/tree/v2.0.1 "v2.0.1")|[v2.0.\*](https://github.com/intel/intel-extension-for-pytorch/tree/v2.0.100+cpu)|
+|[v1.13.\*](https://github.com/pytorch/pytorch/tree/v1.13.0 "v1.13.0")|[v1.13.\*](https://github.com/intel/intel-extension-for-pytorch/tree/v1.13.100+cpu)|
+
+You can simply use pip to install IPEX with the latest version.
+```python
+python -m pip install intel_extension_for_pytorch
+```
+**Note:** To install a specific version, run with the following command:
+```
+python -m pip install intel_extension_for_pytorch==<version_name> -f https://developer.intel.com/ipex-whl-stable-cpu
+```
+
+2. After pipeline initialization, `prepare_for_ipex()` should be called to enable IPEX accelaration. Supported inference datatypes are Float32 and BFloat16.
+
+**Note:** The values of `height` and `width` used during preparation with `prepare_for_ipex()` should be the same when running inference with the prepared pipeline.
+
+```python
+pipe = StableDiffusionXLPipelineIpex.from_pretrained("stabilityai/sdxl-turbo", low_cpu_mem_usage=True, use_safetensors=True)
+# value of image height/width should be consistent with the pipeline inference
+# For Float32
+pipe.prepare_for_ipex(torch.float32, prompt, height=512, width=512)
+# For BFloat16
+pipe.prepare_for_ipex(torch.bfloat16, prompt, height=512, width=512)
+```
+
+Then you can use the ipex pipeline in a similar way to the default stable diffusion xl pipeline.
+```python
+# value of image height/width should be consistent with 'prepare_for_ipex()'
+# For Float32
+image = pipe(prompt, num_inference_steps=num_inference_steps, height=512, width=512, guidance_scale=guidance_scale).images[0]
+# For BFloat16
+with torch.cpu.amp.autocast(enabled=True, dtype=torch.bfloat16):
+    image = pipe(prompt, num_inference_steps=num_inference_steps, height=512, width=512, guidance_scale=guidance_scale).images[0]
+```
+
+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,
+code-named Sapphire Rapids.
+
+```python
+import torch
+from diffusers import StableDiffusionXLPipeline
+from pipeline_stable_diffusion_xl_ipex import StableDiffusionXLPipelineIpex
+import time
+
+prompt = "sailing ship in storm by Rembrandt"
+model_id = "stabilityai/sdxl-turbo"
+steps = 4
+
+# Helper function for time evaluation
+def elapsed_time(pipeline, nb_pass=3, num_inference_steps=1):
+    # warmup
+    for _ in range(2):
+        images = pipeline(prompt, num_inference_steps=num_inference_steps, height=512, width=512, guidance_scale=0.0).images
+    #time evaluation
+    start = time.time()
+    for _ in range(nb_pass):
+        pipeline(prompt, num_inference_steps=num_inference_steps, height=512, width=512, guidance_scale=0.0)
+    end = time.time()
+    return (end - start) / nb_pass
+
+##############     bf16 inference performance    ###############
+
+# 1. IPEX Pipeline initialization
+pipe = StableDiffusionXLPipelineIpex.from_pretrained(model_id, low_cpu_mem_usage=True, use_safetensors=True)
+pipe.prepare_for_ipex(torch.bfloat16, prompt, height=512, width=512)
+
+# 2. Original Pipeline initialization
+pipe2 = StableDiffusionXLPipeline.from_pretrained(model_id, low_cpu_mem_usage=True, use_safetensors=True)
+
+# 3. Compare performance between Original Pipeline and IPEX Pipeline
+with torch.cpu.amp.autocast(enabled=True, dtype=torch.bfloat16):
+    latency = elapsed_time(pipe, num_inference_steps=steps)
+    print("Latency of StableDiffusionXLPipelineIpex--bf16", latency, "s for total", steps, "steps")
+    latency = elapsed_time(pipe2, num_inference_steps=steps)
+    print("Latency of StableDiffusionXLPipeline--bf16", latency, "s for total", steps, "steps")
+
+##############     fp32 inference performance    ###############
+
+# 1. IPEX Pipeline initialization
+pipe3 = StableDiffusionXLPipelineIpex.from_pretrained(model_id, low_cpu_mem_usage=True, use_safetensors=True)
+pipe3.prepare_for_ipex(torch.float32, prompt, height=512, width=512)
+
+# 2. Original Pipeline initialization
+pipe4 = StableDiffusionXLPipeline.from_pretrained(model_id, low_cpu_mem_usage=True, use_safetensors=True)
+
+# 3. Compare performance between Original Pipeline and IPEX Pipeline
+latency = elapsed_time(pipe3, num_inference_steps=steps)
+print("Latency of StableDiffusionXLPipelineIpex--fp32", latency, "s for total", steps, "steps")
+latency = elapsed_time(pipe4, num_inference_steps=steps)
+print("Latency of StableDiffusionXLPipeline--fp32",latency, "s for total", steps, "steps")
+
+```
+
 ### CLIP Guided Images Mixing With Stable Diffusion
 
 ![clip_guided_images_mixing_examples](https://huggingface.co/datasets/TheDenk/images_mixing/resolve/main/main.png)
@@ -1720,7 +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
@@ -3291,7 +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"
@@ -3306,18 +3412,15 @@ inverted_latent, uncond = pipeline.invert(input_image, invert_prompt, num_inner_
 pipeline(prompt, uncond, inverted_latent, guidance_scale=7.5, num_inference_steps=steps).images[0].save(input_image+".output.jpg")
 ```
 
-### Rerender_A_Video
+### 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
@@ -3388,7 +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)
@@ -3456,14 +3559,17 @@ pipe.disable_style_aligned()
 
 This pipeline adds experimental support for the image-to-video task using AnimateDiff. Refer to [this](https://github.com/huggingface/diffusers/pull/6328) PR for more examples and results.
 
+This pipeline relies on a "hack" discovered by the community that allows the generation of videos given an input image with AnimateDiff. It works by creating a copy of the image `num_frames` times and progressively adding more noise to the image based on the strength and latent interpolation method.
+
 ```py
 import torch
 from diffusers import MotionAdapter, DiffusionPipeline, DDIMScheduler
 from diffusers.utils import export_to_gif, load_image
 
+model_id = "SG161222/Realistic_Vision_V5.1_noVAE"
 adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5-2")
-pipe = DiffusionPipeline.from_pretrained("SG161222/Realistic_Vision_V5.1_noVAE", motion_adapter=adapter, custom_pipeline="pipeline_animatediff_img2video").to("cuda")
-pipe.scheduler = DDIMScheduler(beta_schedule="linear", steps_offset=1, clip_sample=False, timespace_spacing="linspace")
+pipe = DiffusionPipeline.from_pretrained(model_id, motion_adapter=adapter, custom_pipeline="pipeline_animatediff_img2video").to("cuda")
+pipe.scheduler = DDIMScheduler.from_pretrained(model_id, subfolder="scheduler", clip_sample=False, timestep_spacing="linspace", beta_schedule="linear", steps_offset=1)
 
 image = load_image("snail.png")
 output = pipe(
@@ -3528,8 +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/checkpoint_merger.py

@@ -81,6 +81,8 @@ class CheckpointMergerPipeline(DiffusionPipeline):
 
                 force - Whether to ignore mismatch in model_config.json for the current models. Defaults to False.
 
+                variant - which variant of a pretrained model to load, e.g. "fp16" (None)
+
         """
         # Default kwargs from DiffusionPipeline
         cache_dir = kwargs.pop("cache_dir", None)
@@ -89,6 +91,7 @@ class CheckpointMergerPipeline(DiffusionPipeline):
         proxies = kwargs.pop("proxies", None)
         local_files_only = kwargs.pop("local_files_only", False)
         token = kwargs.pop("token", None)
+        variant = kwargs.pop("variant", None)
         revision = kwargs.pop("revision", None)
         torch_dtype = kwargs.pop("torch_dtype", None)
         device_map = kwargs.pop("device_map", None)
@@ -173,7 +176,10 @@ class CheckpointMergerPipeline(DiffusionPipeline):
         # Step 3:-
         # Load the first checkpoint as a diffusion pipeline and modify its module state_dict in place
         final_pipe = DiffusionPipeline.from_pretrained(
-            cached_folders[0], torch_dtype=torch_dtype, device_map=device_map
+            cached_folders[0],
+            torch_dtype=torch_dtype,
+            device_map=device_map,
+            variant=variant,
         )
         final_pipe.to(self.device)
 

examples/community/clip_guided_images_mixing_stable_diffusion.py

@@ -12,12 +12,12 @@ from transformers import CLIPFeatureExtractor, CLIPModel, CLIPTextModel, CLIPTok
 from diffusers import (
     AutoencoderKL,
     DDIMScheduler,
-    DiffusionPipeline,
     DPMSolverMultistepScheduler,
     LMSDiscreteScheduler,
     PNDMScheduler,
     UNet2DConditionModel,
 )
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.utils import PIL_INTERPOLATION
 from diffusers.utils.torch_utils import randn_tensor
@@ -77,7 +77,7 @@ def set_requires_grad(model, value):
         param.requires_grad = value
 
 
-class CLIPGuidedImagesMixingStableDiffusion(DiffusionPipeline):
+class CLIPGuidedImagesMixingStableDiffusion(DiffusionPipeline, StableDiffusionMixin):
     def __init__(
         self,
         vae: AutoencoderKL,
@@ -113,16 +113,6 @@ class CLIPGuidedImagesMixingStableDiffusion(DiffusionPipeline):
         set_requires_grad(self.text_encoder, False)
         set_requires_grad(self.clip_model, False)
 
-    def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
-        if slice_size == "auto":
-            # half the attention head size is usually a good trade-off between
-            # speed and memory
-            slice_size = self.unet.config.attention_head_dim // 2
-        self.unet.set_attention_slice(slice_size)
-
-    def disable_attention_slicing(self):
-        self.enable_attention_slicing(None)
-
     def freeze_vae(self):
         set_requires_grad(self.vae, False)
 

examples/community/clip_guided_stable_diffusion.py

@@ -10,12 +10,12 @@ from transformers import CLIPImageProcessor, CLIPModel, CLIPTextModel, CLIPToken
 from diffusers import (
     AutoencoderKL,
     DDIMScheduler,
-    DiffusionPipeline,
     DPMSolverMultistepScheduler,
     LMSDiscreteScheduler,
     PNDMScheduler,
     UNet2DConditionModel,
 )
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion import StableDiffusionPipelineOutput
 
 
@@ -51,7 +51,7 @@ def set_requires_grad(model, value):
         param.requires_grad = value
 
 
-class CLIPGuidedStableDiffusion(DiffusionPipeline):
+class CLIPGuidedStableDiffusion(DiffusionPipeline, StableDiffusionMixin):
     """CLIP guided stable diffusion based on the amazing repo by @crowsonkb and @Jack000
     - https://github.com/Jack000/glid-3-xl
     - https://github.dev/crowsonkb/k-diffusion
@@ -89,16 +89,6 @@ class CLIPGuidedStableDiffusion(DiffusionPipeline):
         set_requires_grad(self.text_encoder, False)
         set_requires_grad(self.clip_model, False)
 
-    def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
-        if slice_size == "auto":
-            # half the attention head size is usually a good trade-off between
-            # speed and memory
-            slice_size = self.unet.config.attention_head_dim // 2
-        self.unet.set_attention_slice(slice_size)
-
-    def disable_attention_slicing(self):
-        self.enable_attention_slicing(None)
-
     def freeze_vae(self):
         set_requires_grad(self.vae, False)
 

examples/community/clip_guided_stable_diffusion_img2img.py

@@ -12,12 +12,12 @@ from transformers import CLIPFeatureExtractor, CLIPModel, CLIPTextModel, CLIPTok
 from diffusers import (
     AutoencoderKL,
     DDIMScheduler,
-    DiffusionPipeline,
     DPMSolverMultistepScheduler,
     LMSDiscreteScheduler,
     PNDMScheduler,
     UNet2DConditionModel,
 )
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.utils import PIL_INTERPOLATION, deprecate
 from diffusers.utils.torch_utils import randn_tensor
@@ -125,7 +125,7 @@ def set_requires_grad(model, value):
         param.requires_grad = value
 
 
-class CLIPGuidedStableDiffusion(DiffusionPipeline):
+class CLIPGuidedStableDiffusion(DiffusionPipeline, StableDiffusionMixin):
     """CLIP guided stable diffusion based on the amazing repo by @crowsonkb and @Jack000
     - https://github.com/Jack000/glid-3-xl
     - https://github.dev/crowsonkb/k-diffusion
@@ -163,16 +163,6 @@ class CLIPGuidedStableDiffusion(DiffusionPipeline):
         set_requires_grad(self.text_encoder, False)
         set_requires_grad(self.clip_model, False)
 
-    def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
-        if slice_size == "auto":
-            # half the attention head size is usually a good trade-off between
-            # speed and memory
-            slice_size = self.unet.config.attention_head_dim // 2
-        self.unet.set_attention_slice(slice_size)
-
-    def disable_attention_slicing(self):
-        self.enable_attention_slicing(None)
-
     def freeze_vae(self):
         set_requires_grad(self.vae, False)
 

examples/community/composable_stable_diffusion.py

@@ -22,6 +22,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 from diffusers import DiffusionPipeline
 from diffusers.configuration_utils import FrozenDict
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
+from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import (
@@ -32,13 +33,13 @@ from diffusers.schedulers import (
     LMSDiscreteScheduler,
     PNDMScheduler,
 )
-from diffusers.utils import deprecate, is_accelerate_available, logging
+from diffusers.utils import deprecate, logging
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-class ComposableStableDiffusionPipeline(DiffusionPipeline):
+class ComposableStableDiffusionPipeline(DiffusionPipeline, StableDiffusionMixin):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion.
 
@@ -164,62 +165,6 @@ class ComposableStableDiffusionPipeline(DiffusionPipeline):
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding.
-
-        When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several
-        steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously invoked, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    def enable_sequential_cpu_offload(self, gpu_id=0):
-        r"""
-        Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
-        text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
-        `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
-        """
-        if is_accelerate_available():
-            from accelerate import cpu_offload
-        else:
-            raise ImportError("Please install accelerate via `pip install accelerate`")
-
-        device = torch.device(f"cuda:{gpu_id}")
-
-        for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            if cpu_offloaded_model is not None:
-                cpu_offload(cpu_offloaded_model, device)
-
-        if self.safety_checker is not None:
-            # TODO(Patrick) - there is currently a bug with cpu offload of nn.Parameter in accelerate
-            # fix by only offloading self.safety_checker for now
-            cpu_offload(self.safety_checker.vision_model, device)
-
-    @property
-    def _execution_device(self):
-        r"""
-        Returns the device on which the pipeline's models will be executed. After calling
-        `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module
-        hooks.
-        """
-        if self.device != torch.device("meta") or not hasattr(self.unet, "_hf_hook"):
-            return self.device
-        for module in self.unet.modules():
-            if (
-                hasattr(module, "_hf_hook")
-                and hasattr(module._hf_hook, "execution_device")
-                and module._hf_hook.execution_device is not None
-            ):
-                return torch.device(module._hf_hook.execution_device)
-        return self.device
-
     def _encode_prompt(self, prompt, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt):
         r"""
         Encodes the prompt into text encoder hidden states.

examples/community/gluegen.py

@@ -10,6 +10,7 @@ from diffusers.image_processor import VaeImageProcessor
 from diffusers.loaders import LoraLoaderMixin
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
+from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import KarrasDiffusionSchedulers
@@ -193,7 +194,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-class GlueGenStableDiffusionPipeline(DiffusionPipeline, LoraLoaderMixin):
+class GlueGenStableDiffusionPipeline(DiffusionPipeline, StableDiffusionMixin, LoraLoaderMixin):
     def __init__(
         self,
         vae: AutoencoderKL,
@@ -241,35 +242,6 @@ class GlueGenStableDiffusionPipeline(DiffusionPipeline, LoraLoaderMixin):
         )
         self.language_adapter.load_state_dict(torch.load(model_path))
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     def _adapt_language(self, prompt_embeds: torch.FloatTensor):
         prompt_embeds = prompt_embeds / 3
         prompt_embeds = self.language_adapter(prompt_embeds) * (self.tensor_norm / 2)
@@ -544,32 +516,6 @@ class GlueGenStableDiffusionPipeline(DiffusionPipeline, LoraLoaderMixin):
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
     def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):
         """

examples/community/imagic_stable_diffusion.py

@@ -19,6 +19,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 
 from diffusers import DiffusionPipeline
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
+from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
@@ -56,7 +57,7 @@ def preprocess(image):
     return 2.0 * image - 1.0
 
 
-class ImagicStableDiffusionPipeline(DiffusionPipeline):
+class ImagicStableDiffusionPipeline(DiffusionPipeline, StableDiffusionMixin):
     r"""
     Pipeline for imagic image editing.
     See paper here: https://arxiv.org/pdf/2210.09276.pdf
@@ -105,31 +106,6 @@ class ImagicStableDiffusionPipeline(DiffusionPipeline):
             feature_extractor=feature_extractor,
         )
 
-    def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
-        r"""
-        Enable sliced attention computation.
-        When this option is enabled, the attention module will split the input tensor in slices, to compute attention
-        in several steps. This is useful to save some memory in exchange for a small speed decrease.
-        Args:
-            slice_size (`str` or `int`, *optional*, defaults to `"auto"`):
-                When `"auto"`, halves the input to the attention heads, so attention will be computed in two steps. If
-                a number is provided, uses as many slices as `attention_head_dim // slice_size`. In this case,
-                `attention_head_dim` must be a multiple of `slice_size`.
-        """
-        if slice_size == "auto":
-            # half the attention head size is usually a good trade-off between
-            # speed and memory
-            slice_size = self.unet.config.attention_head_dim // 2
-        self.unet.set_attention_slice(slice_size)
-
-    def disable_attention_slicing(self):
-        r"""
-        Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
-        back to computing attention in one step.
-        """
-        # set slice_size = `None` to disable `attention slicing`
-        self.enable_attention_slicing(None)
-
     def train(
         self,
         prompt: Union[str, List[str]],
@@ -346,8 +322,9 @@ class ImagicStableDiffusionPipeline(DiffusionPipeline):
         r"""
         Function invoked when calling the pipeline for generation.
         Args:
-            prompt (`str` or `List[str]`):
-                The prompt or prompts to guide the image generation.
+            alpha (`float`, *optional*, defaults to 1.2):
+                The interpolation factor between the original and optimized text embeddings. A value closer to 0
+                will resemble the original input image.
             height (`int`, *optional*, defaults to 512):
                 The height in pixels of the generated image.
             width (`int`, *optional*, defaults to 512):
@@ -361,22 +338,18 @@ class ImagicStableDiffusionPipeline(DiffusionPipeline):
                 Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
                 1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
                 usually at the expense of lower image quality.
-            eta (`float`, *optional*, defaults to 0.0):
-                Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
-                [`schedulers.DDIMScheduler`], will be ignored for others.
             generator (`torch.Generator`, *optional*):
                 A [torch generator](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make generation
                 deterministic.
-            latents (`torch.FloatTensor`, *optional*):
-                Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
-                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
-                tensor will ge generated by sampling using the supplied random `generator`.
             output_type (`str`, *optional*, defaults to `"pil"`):
                 The output format of the generate image. Choose between
                 [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `nd.array`.
             return_dict (`bool`, *optional*, defaults to `True`):
                 Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
                 plain tuple.
+            eta (`float`, *optional*, defaults to 0.0):
+                Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
+                [`schedulers.DDIMScheduler`], will be ignored for others.
         Returns:
             [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
             [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] if `return_dict` is True, otherwise a `tuple.

examples/community/img2img_inpainting.py

@@ -129,33 +129,6 @@ class ImageToImageInpaintingPipeline(DiffusionPipeline):
             feature_extractor=feature_extractor,
         )
 
-    def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
-        r"""
-        Enable sliced attention computation.
-
-        When this option is enabled, the attention module will split the input tensor in slices, to compute attention
-        in several steps. This is useful to save some memory in exchange for a small speed decrease.
-
-        Args:
-            slice_size (`str` or `int`, *optional*, defaults to `"auto"`):
-                When `"auto"`, halves the input to the attention heads, so attention will be computed in two steps. If
-                a number is provided, uses as many slices as `attention_head_dim // slice_size`. In this case,
-                `attention_head_dim` must be a multiple of `slice_size`.
-        """
-        if slice_size == "auto":
-            # half the attention head size is usually a good trade-off between
-            # speed and memory
-            slice_size = self.unet.config.attention_head_dim // 2
-        self.unet.set_attention_slice(slice_size)
-
-    def disable_attention_slicing(self):
-        r"""
-        Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
-        back to computing attention in one step.
-        """
-        # set slice_size = `None` to disable `attention slicing`
-        self.enable_attention_slicing(None)
-
     @torch.no_grad()
     def __call__(
         self,

examples/community/instaflow_one_step.py

@@ -24,7 +24,7 @@ from diffusers.image_processor import VaeImageProcessor
 from diffusers.loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import KarrasDiffusionSchedulers
@@ -52,7 +52,9 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-class InstaFlowPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin):
+class InstaFlowPipeline(
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
+):
     r"""
     Pipeline for text-to-image generation using Rectified Flow and Euler discretization.
     This customized pipeline is based on StableDiffusionPipeline from the official Diffusers library (0.21.4)
@@ -180,35 +182,6 @@ class InstaFlowPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoad
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     def _encode_prompt(
         self,
         prompt,

examples/community/interpolate_stable_diffusion.py

@@ -7,9 +7,9 @@ import numpy as np
 import torch
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 
-from diffusers import DiffusionPipeline
 from diffusers.configuration_utils import FrozenDict
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
@@ -46,7 +46,7 @@ def slerp(t, v0, v1, DOT_THRESHOLD=0.9995):
     return v2
 
 
-class StableDiffusionWalkPipeline(DiffusionPipeline):
+class StableDiffusionWalkPipeline(DiffusionPipeline, StableDiffusionMixin):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion.
 
@@ -120,33 +120,6 @@ class StableDiffusionWalkPipeline(DiffusionPipeline):
             feature_extractor=feature_extractor,
         )
 
-    def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
-        r"""
-        Enable sliced attention computation.
-
-        When this option is enabled, the attention module will split the input tensor in slices, to compute attention
-        in several steps. This is useful to save some memory in exchange for a small speed decrease.
-
-        Args:
-            slice_size (`str` or `int`, *optional*, defaults to `"auto"`):
-                When `"auto"`, halves the input to the attention heads, so attention will be computed in two steps. If
-                a number is provided, uses as many slices as `attention_head_dim // slice_size`. In this case,
-                `attention_head_dim` must be a multiple of `slice_size`.
-        """
-        if slice_size == "auto":
-            # half the attention head size is usually a good trade-off between
-            # speed and memory
-            slice_size = self.unet.config.attention_head_dim // 2
-        self.unet.set_attention_slice(slice_size)
-
-    def disable_attention_slicing(self):
-        r"""
-        Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
-        back to computing attention in one step.
-        """
-        # set slice_size = `None` to disable `attention slicing`
-        self.enable_attention_slicing(None)
-
     @torch.no_grad()
     def __call__(
         self,

examples/community/ip_adapter_face_id.py

@@ -26,9 +26,8 @@ from diffusers.configuration_utils import FrozenDict
 from diffusers.image_processor import VaeImageProcessor
 from diffusers.loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
-from diffusers.models.attention_processor import FusedAttnProcessor2_0
 from diffusers.models.lora import LoRALinearLayer, adjust_lora_scale_text_encoder
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import KarrasDiffusionSchedulers
@@ -415,7 +414,12 @@ def retrieve_timesteps(
 
 
 class IPAdapterFaceIDStableDiffusionPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
+    DiffusionPipeline,
+    StableDiffusionMixin,
+    TextualInversionLoaderMixin,
+    LoraLoaderMixin,
+    IPAdapterMixin,
+    FromSingleFileMixin,
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion.
@@ -727,35 +731,6 @@ class IPAdapterFaceIDStableDiffusionPipeline(
             if isinstance(attn_processor, (LoRAIPAdapterAttnProcessor, LoRAIPAdapterAttnProcessor2_0)):
                 attn_processor.scale = scale
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     def _encode_prompt(
         self,
         prompt,
@@ -1080,93 +1055,6 @@ class IPAdapterFaceIDStableDiffusionPipeline(
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.fuse_qkv_projections
-    def fuse_qkv_projections(self, unet: bool = True, vae: bool = True):
-        """
-        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query,
-        key, value) are fused. For cross-attention modules, key and value projection matrices are fused.
-
-        <Tip warning={true}>
-
-        This API is 🧪 experimental.
-
-        </Tip>
-
-        Args:
-            unet (`bool`, defaults to `True`): To apply fusion on the UNet.
-            vae (`bool`, defaults to `True`): To apply fusion on the VAE.
-        """
-        self.fusing_unet = False
-        self.fusing_vae = False
-
-        if unet:
-            self.fusing_unet = True
-            self.unet.fuse_qkv_projections()
-            self.unet.set_attn_processor(FusedAttnProcessor2_0())
-
-        if vae:
-            if not isinstance(self.vae, AutoencoderKL):
-                raise ValueError("`fuse_qkv_projections()` is only supported for the VAE of type `AutoencoderKL`.")
-
-            self.fusing_vae = True
-            self.vae.fuse_qkv_projections()
-            self.vae.set_attn_processor(FusedAttnProcessor2_0())
-
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.unfuse_qkv_projections
-    def unfuse_qkv_projections(self, unet: bool = True, vae: bool = True):
-        """Disable QKV projection fusion if enabled.
-
-        <Tip warning={true}>
-
-        This API is 🧪 experimental.
-
-        </Tip>
-
-        Args:
-            unet (`bool`, defaults to `True`): To apply fusion on the UNet.
-            vae (`bool`, defaults to `True`): To apply fusion on the VAE.
-
-        """
-        if unet:
-            if not self.fusing_unet:
-                logger.warning("The UNet was not initially fused for QKV projections. Doing nothing.")
-            else:
-                self.unet.unfuse_qkv_projections()
-                self.fusing_unet = False
-
-        if vae:
-            if not self.fusing_vae:
-                logger.warning("The VAE was not initially fused for QKV projections. Doing nothing.")
-            else:
-                self.vae.unfuse_qkv_projections()
-                self.fusing_vae = False
-
     # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
     def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):
         """

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_interpolate.py

@@ -9,7 +9,7 @@ from diffusers.image_processor import VaeImageProcessor
 from diffusers.loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput, StableDiffusionSafetyChecker
 from diffusers.schedulers import LCMScheduler
 from diffusers.utils import (
@@ -190,7 +190,7 @@ def slerp(
 
 
 class LatentConsistencyModelWalkPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for text-to-image generation using a latent consistency model.
@@ -273,67 +273,6 @@ class LatentConsistencyModelWalkPipeline(
         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.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
     def encode_prompt(
         self,

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/llm_grounded_diffusion.py

@@ -35,6 +35,7 @@ from diffusers.models.attention import Attention, GatedSelfAttentionDense
 from diffusers.models.attention_processor import AttnProcessor2_0
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.pipelines import DiffusionPipeline
+from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import KarrasDiffusionSchedulers
@@ -267,7 +268,12 @@ class AttnProcessorWithHook(AttnProcessor2_0):
 
 
 class LLMGroundedDiffusionPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
+    DiffusionPipeline,
+    StableDiffusionMixin,
+    TextualInversionLoaderMixin,
+    LoraLoaderMixin,
+    IPAdapterMixin,
+    FromSingleFileMixin,
 ):
     r"""
     Pipeline for layout-grounded text-to-image generation using LLM-grounded Diffusion (LMD+): https://arxiv.org/pdf/2305.13655.pdf.
@@ -1180,39 +1186,6 @@ class LLMGroundedDiffusionPipeline(
     # Below are methods copied from StableDiffusionPipeline
     # The design choice of not inheriting from StableDiffusionPipeline is discussed here: https://github.com/huggingface/diffusers/pull/5993#issuecomment-1834258517
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
     def _encode_prompt(
         self,
@@ -1522,34 +1495,6 @@ class LLMGroundedDiffusionPipeline(
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
     def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):
         """

examples/community/lpw_stable_diffusion.py

@@ -13,13 +13,12 @@ from diffusers.configuration_utils import FrozenDict
 from diffusers.image_processor import VaeImageProcessor
 from diffusers.loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
+from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput, StableDiffusionSafetyChecker
 from diffusers.schedulers import KarrasDiffusionSchedulers
 from diffusers.utils import (
     PIL_INTERPOLATION,
     deprecate,
-    is_accelerate_available,
-    is_accelerate_version,
     logging,
 )
 from diffusers.utils.torch_utils import randn_tensor
@@ -410,7 +409,7 @@ def preprocess_mask(mask, batch_size, scale_factor=8):
 
 
 class StableDiffusionLongPromptWeightingPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion without tokens length limit, and support parsing
@@ -534,112 +533,6 @@ class StableDiffusionLongPromptWeightingPipeline(
             requires_safety_checker=requires_safety_checker,
         )
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding.
-
-        When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several
-        steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously invoked, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding.
-
-        When this option is enabled, the VAE will split the input tensor into tiles to compute decoding and encoding in
-        several steps. This is useful to save a large amount of memory and to allow the processing of larger images.
-        """
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously invoked, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_sequential_cpu_offload
-    def enable_sequential_cpu_offload(self, gpu_id=0):
-        r"""
-        Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
-        text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
-        `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
-        Note that offloading happens on a submodule basis. Memory savings are higher than with
-        `enable_model_cpu_offload`, but performance is lower.
-        """
-        if is_accelerate_available() and is_accelerate_version(">=", "0.14.0"):
-            from accelerate import cpu_offload
-        else:
-            raise ImportError("`enable_sequential_cpu_offload` requires `accelerate v0.14.0` or higher")
-
-        device = torch.device(f"cuda:{gpu_id}")
-
-        if self.device.type != "cpu":
-            self.to("cpu", silence_dtype_warnings=True)
-            torch.cuda.empty_cache()  # otherwise we don't see the memory savings (but they probably exist)
-
-        for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            cpu_offload(cpu_offloaded_model, device)
-
-        if self.safety_checker is not None:
-            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_model_cpu_offload
-    def enable_model_cpu_offload(self, gpu_id=0):
-        r"""
-        Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
-        to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
-        method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
-        `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
-        """
-        if is_accelerate_available() and is_accelerate_version(">=", "0.17.0.dev0"):
-            from accelerate import cpu_offload_with_hook
-        else:
-            raise ImportError("`enable_model_cpu_offload` requires `accelerate v0.17.0` or higher.")
-
-        device = torch.device(f"cuda:{gpu_id}")
-
-        if self.device.type != "cpu":
-            self.to("cpu", silence_dtype_warnings=True)
-            torch.cuda.empty_cache()  # otherwise we don't see the memory savings (but they probably exist)
-
-        hook = None
-        for cpu_offloaded_model in [self.text_encoder, self.unet, self.vae]:
-            _, hook = cpu_offload_with_hook(cpu_offloaded_model, device, prev_module_hook=hook)
-
-        if self.safety_checker is not None:
-            _, hook = cpu_offload_with_hook(self.safety_checker, device, prev_module_hook=hook)
-
-        # We'll offload the last model manually.
-        self.final_offload_hook = hook
-
-    @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device
-    def _execution_device(self):
-        r"""
-        Returns the device on which the pipeline's models will be executed. After calling
-        `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module
-        hooks.
-        """
-        if not hasattr(self.unet, "_hf_hook"):
-            return self.device
-        for module in self.unet.modules():
-            if (
-                hasattr(module, "_hf_hook")
-                and hasattr(module._hf_hook, "execution_device")
-                and module._hf_hook.execution_device is not None
-            ):
-                return torch.device(module._hf_hook.execution_device)
-        return self.device
-
     def _encode_prompt(
         self,
         prompt,

examples/community/lpw_stable_diffusion_xl.py

@@ -26,11 +26,11 @@ from diffusers.loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMix
 from diffusers.models import AutoencoderKL, ImageProjection, UNet2DConditionModel
 from diffusers.models.attention_processor import (
     AttnProcessor2_0,
-    FusedAttnProcessor2_0,
     LoRAAttnProcessor2_0,
     LoRAXFormersAttnProcessor,
     XFormersAttnProcessor,
 )
+from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
 from diffusers.schedulers import KarrasDiffusionSchedulers
 from diffusers.utils import (
@@ -545,7 +545,12 @@ def retrieve_timesteps(
 
 
 class SDXLLongPromptWeightingPipeline(
-    DiffusionPipeline, FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
+    DiffusionPipeline,
+    StableDiffusionMixin,
+    FromSingleFileMixin,
+    IPAdapterMixin,
+    LoraLoaderMixin,
+    TextualInversionLoaderMixin,
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion XL.
@@ -649,39 +654,6 @@ class SDXLLongPromptWeightingPipeline(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     def enable_model_cpu_offload(self, gpu_id=0):
         r"""
         Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
@@ -1030,95 +1002,6 @@ class SDXLLongPromptWeightingPipeline(
                 "If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`."
             )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.fuse_qkv_projections
-    def fuse_qkv_projections(self, unet: bool = True, vae: bool = True):
-        """
-        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query,
-        key, value) are fused. For cross-attention modules, key and value projection matrices are fused.
-
-        <Tip warning={true}>
-
-        This API is 🧪 experimental.
-
-        </Tip>
-
-        Args:
-            unet (`bool`, defaults to `True`): To apply fusion on the UNet.
-            vae (`bool`, defaults to `True`): To apply fusion on the VAE.
-        """
-        self.fusing_unet = False
-        self.fusing_vae = False
-
-        if unet:
-            self.fusing_unet = True
-            self.unet.fuse_qkv_projections()
-            self.unet.set_attn_processor(FusedAttnProcessor2_0())
-
-        if vae:
-            if not isinstance(self.vae, AutoencoderKL):
-                raise ValueError("`fuse_qkv_projections()` is only supported for the VAE of type `AutoencoderKL`.")
-
-            self.fusing_vae = True
-            self.vae.fuse_qkv_projections()
-            self.vae.set_attn_processor(FusedAttnProcessor2_0())
-
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.unfuse_qkv_projections
-    def unfuse_qkv_projections(self, unet: bool = True, vae: bool = True):
-        """Disable QKV projection fusion if enabled.
-
-        <Tip warning={true}>
-
-        This API is 🧪 experimental.
-
-        </Tip>
-
-        Args:
-            unet (`bool`, defaults to `True`): To apply fusion on the UNet.
-            vae (`bool`, defaults to `True`): To apply fusion on the VAE.
-
-        """
-        if unet:
-            if not self.fusing_unet:
-                logger.warning("The UNet was not initially fused for QKV projections. Doing nothing.")
-            else:
-                self.unet.unfuse_qkv_projections()
-                self.fusing_unet = False
-
-        if vae:
-            if not self.fusing_vae:
-                logger.warning("The VAE was not initially fused for QKV projections. Doing nothing.")
-            else:
-                self.vae.unfuse_qkv_projections()
-                self.fusing_vae = False
-
     def get_timesteps(self, num_inference_steps, strength, device, denoising_start=None):
         # get the original timestep using init_timestep
         if denoising_start is None:
@@ -1766,7 +1649,7 @@ class SDXLLongPromptWeightingPipeline(
 
         # 4. Prepare timesteps
         def denoising_value_valid(dnv):
-            return isinstance(self.denoising_end, float) and 0 < dnv < 1
+            return isinstance(dnv, float) and 0 < dnv < 1
 
         timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, num_inference_steps, device, timesteps)
         if image is not None:
@@ -1774,7 +1657,7 @@ class SDXLLongPromptWeightingPipeline(
                 num_inference_steps,
                 strength,
                 device,
-                denoising_start=self.denoising_start if denoising_value_valid else None,
+                denoising_start=self.denoising_start if denoising_value_valid(self.denoising_start) else None,
             )
 
             # check that number of inference steps is not < 1 - as this doesn't make sense

examples/community/marigold_depth_estimation.py

@@ -40,8 +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("0.27.0")
 
 class MarigoldDepthOutput(BaseOutput):
     """
@@ -50,14 +49,14 @@ 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 +138,15 @@ 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 +233,15 @@ 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/mixture_canvas.py

@@ -12,7 +12,7 @@ from tqdm.auto import tqdm
 from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer
 
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionSafetyChecker
 from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
 
@@ -264,7 +264,7 @@ class MaskWeightsBuilder:
         return torch.tile(torch.tensor(weights), (self.nbatch, self.latent_space_dim, 1, 1))
 
 
-class StableDiffusionCanvasPipeline(DiffusionPipeline):
+class StableDiffusionCanvasPipeline(DiffusionPipeline, StableDiffusionMixin):
     """Stable Diffusion pipeline that mixes several diffusers in the same canvas"""
 
     def __init__(

examples/community/multilingual_stable_diffusion.py

@@ -11,9 +11,9 @@ from transformers import (
     pipeline,
 )
 
-from diffusers import DiffusionPipeline
 from diffusers.configuration_utils import FrozenDict
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
@@ -48,7 +48,7 @@ def translate_prompt(prompt, translation_tokenizer, translation_model, device):
     return en_trans[0]
 
 
-class MultilingualStableDiffusion(DiffusionPipeline):
+class MultilingualStableDiffusion(DiffusionPipeline, StableDiffusionMixin):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion in different languages.
 
@@ -135,33 +135,6 @@ class MultilingualStableDiffusion(DiffusionPipeline):
             feature_extractor=feature_extractor,
         )
 
-    def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
-        r"""
-        Enable sliced attention computation.
-
-        When this option is enabled, the attention module will split the input tensor in slices, to compute attention
-        in several steps. This is useful to save some memory in exchange for a small speed decrease.
-
-        Args:
-            slice_size (`str` or `int`, *optional*, defaults to `"auto"`):
-                When `"auto"`, halves the input to the attention heads, so attention will be computed in two steps. If
-                a number is provided, uses as many slices as `attention_head_dim // slice_size`. In this case,
-                `attention_head_dim` must be a multiple of `slice_size`.
-        """
-        if slice_size == "auto":
-            # half the attention head size is usually a good trade-off between
-            # speed and memory
-            slice_size = self.unet.config.attention_head_dim // 2
-        self.unet.set_attention_slice(slice_size)
-
-    def disable_attention_slicing(self):
-        r"""
-        Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
-        back to computing attention in one step.
-        """
-        # set slice_size = `None` to disable `attention slicing`
-        self.enable_attention_slicing(None)
-
     @torch.no_grad()
     def __call__(
         self,

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
@@ -24,11 +23,12 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPV
 
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
 from diffusers.loaders import IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
-from diffusers.models import AutoencoderKL, ControlNetModel, UNet2DConditionModel, UNetMotionModel
+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
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.schedulers import (
     DDIMScheduler,
     DPMSolverMultistepScheduler,
@@ -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):
@@ -103,15 +101,21 @@ 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, TextualInversionLoaderMixin, IPAdapterMixin, LoraLoaderMixin):
+class AnimateDiffControlNetPipeline(
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, IPAdapterMixin, LoraLoaderMixin
+):
     r"""
     Pipeline for text-to-video generation.
 
@@ -382,6 +386,41 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
         uncond_image_embeds = torch.zeros_like(image_embeds)
         return image_embeds, uncond_image_embeds
 
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
+
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
+
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
+
+                if self.do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
+
+                image_embeds.append(single_image_embeds)
+        else:
+            image_embeds = ip_adapter_image_embeds
+        return image_embeds
+
     # Copied from diffusers.pipelines.text_to_video_synthesis/pipeline_text_to_video_synth.TextToVideoSDPipeline.decode_latents
     def decode_latents(self, latents):
         latents = 1 / self.vae.config.scaling_factor * latents
@@ -406,67 +445,6 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     # 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
@@ -767,6 +745,7 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
         prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_prompt_embeds: Optional[torch.FloatTensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[PipelineImageInput] = None,
         conditioning_frames: Optional[List[PipelineImageInput]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
@@ -821,6 +800,11 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
             ip_adapter_image (`PipelineImageInput`, *optional*):
                 Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of IP-adapters.
+                Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. It should contain the negative image embedding
+                if `do_classifier_free_guidance` is set to `True`.
+                If not provided, embeddings are computed from the `ip_adapter_image` input argument.
             conditioning_frames (`List[PipelineImageInput]`, *optional*):
                 The ControlNet input condition to provide guidance to the `unet` for generation. If multiple ControlNets
                 are specified, images must be passed as a list such that each element of the list can be correctly
@@ -861,8 +845,8 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
         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.
         """
 
@@ -965,9 +949,9 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
             prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])
 
         if ip_adapter_image is not None:
-            image_embeds, negative_image_embeds = self.encode_image(ip_adapter_image, device, num_videos_per_prompt)
-            if self.do_classifier_free_guidance:
-                image_embeds = torch.cat([negative_image_embeds, image_embeds])
+            image_embeds = self.prepare_ip_adapter_image_embeds(
+                ip_adapter_image, ip_adapter_image_embeds, device, batch_size * num_videos_per_prompt
+            )
 
         if isinstance(controlnet, ControlNetModel):
             conditioning_frames = self.prepare_image(
@@ -1023,7 +1007,11 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
         extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
 
         # 7. Add image embeds for IP-Adapter
-        added_cond_kwargs = {"image_embeds": image_embeds} if ip_adapter_image is not None else None
+        added_cond_kwargs = (
+            {"image_embeds": image_embeds}
+            if ip_adapter_image is not None or ip_adapter_image_embeds is not None
+            else None
+        )
 
         # 7.1 Create tensor stating which controlnets to keep
         controlnet_keep = []
@@ -1034,7 +1022,7 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
             ]
             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):
@@ -1110,21 +1098,17 @@ class AnimateDiffControlNetPipeline(DiffusionPipeline, TextualInversionLoaderMix
                     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

@@ -11,9 +11,14 @@
 # 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.
+#
+# Note:
+# This pipeline relies on a "hack" discovered by the community that allows
+# the generation of videos given an input image with AnimateDiff. It works
+# by creating a copy of the image `num_frames` times and progressively adding
+# more noise to the image based on the strength and latent interpolation method.
 
 import inspect
-from dataclasses import dataclass
 from types import FunctionType
 from typing import Any, Callable, Dict, List, Optional, Union
 
@@ -26,7 +31,8 @@ from diffusers.loaders import IPAdapterMixin, LoraLoaderMixin, TextualInversionL
 from diffusers.models import AutoencoderKL, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.models.unet_motion_model import MotionAdapter
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.animatediff.pipeline_output import AnimateDiffPipelineOutput
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.schedulers import (
     DDIMScheduler,
     DPMSolverMultistepScheduler,
@@ -35,7 +41,7 @@ from diffusers.schedulers import (
     LMSDiscreteScheduler,
     PNDMScheduler,
 )
-from diffusers.utils import USE_PEFT_BACKEND, BaseOutput, logging, scale_lora_layers, unscale_lora_layers
+from diffusers.utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
 from diffusers.utils.torch_utils import randn_tensor
 
 
@@ -48,9 +54,10 @@ EXAMPLE_DOC_STRING = """
         >>> from diffusers import MotionAdapter, DiffusionPipeline, DDIMScheduler
         >>> from diffusers.utils import export_to_gif, load_image
 
+        >>> model_id = "SG161222/Realistic_Vision_V5.1_noVAE"
         >>> adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5-2")
         >>> pipe = DiffusionPipeline.from_pretrained("SG161222/Realistic_Vision_V5.1_noVAE", motion_adapter=adapter, custom_pipeline="pipeline_animatediff_img2video").to("cuda")
-        >>> pipe.scheduler = DDIMScheduler(beta_schedule="linear", steps_offset=1, clip_sample=False, timespace_spacing="linspace")
+        >>> pipe.scheduler = pipe.scheduler = DDIMScheduler.from_pretrained(model_id, subfolder="scheduler", clip_sample=False, timestep_spacing="linspace", beta_schedule="linear", steps_offset=1)
 
         >>> image = load_image("snail.png")
         >>> output = pipe(image=image, prompt="A snail moving on the ground", strength=0.8, latent_interpolation_method="slerp")
@@ -151,10 +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):
@@ -163,6 +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
 
 
@@ -225,14 +239,11 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-@dataclass
-class AnimateDiffImgToVideoPipelineOutput(BaseOutput):
-    frames: Union[torch.Tensor, np.ndarray]
-
-
-class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMixin, IPAdapterMixin, LoraLoaderMixin):
+class AnimateDiffImgToVideoPipeline(
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, IPAdapterMixin, LoraLoaderMixin
+):
     r"""
-    Pipeline for text-to-video generation.
+    Pipeline for image-to-video generation.
 
     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.).
@@ -503,6 +514,41 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
 
             return image_embeds, uncond_image_embeds
 
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
+    def prepare_ip_adapter_image_embeds(
+        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
+    ):
+        if ip_adapter_image_embeds is None:
+            if not isinstance(ip_adapter_image, list):
+                ip_adapter_image = [ip_adapter_image]
+
+            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
+                raise ValueError(
+                    f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
+                )
+
+            image_embeds = []
+            for single_ip_adapter_image, image_proj_layer in zip(
+                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
+            ):
+                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
+                single_image_embeds, single_negative_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1, output_hidden_state
+                )
+                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
+                single_negative_image_embeds = torch.stack(
+                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
+                )
+
+                if self.do_classifier_free_guidance:
+                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
+                    single_image_embeds = single_image_embeds.to(device)
+
+                image_embeds.append(single_image_embeds)
+        else:
+            image_embeds = ip_adapter_image_embeds
+        return image_embeds
+
     # Copied from diffusers.pipelines.text_to_video_synthesis/pipeline_text_to_video_synth.TextToVideoSDPipeline.decode_latents
     def decode_latents(self, latents):
         latents = 1 / self.vae.config.scaling_factor * latents
@@ -527,67 +573,6 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     # 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
@@ -765,6 +750,7 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
         prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_prompt_embeds: Optional[torch.FloatTensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
+        ip_adapter_image_embeds: Optional[PipelineImageInput] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
@@ -818,6 +804,11 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
             ip_adapter_image: (`PipelineImageInput`, *optional*):
                 Optional image input to work with IP Adapters.
+            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *optional*):
+                Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of IP-adapters.
+                Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. It should contain the negative image embedding
+                if `do_classifier_free_guidance` is set to `True`.
+                If not provided, embeddings are computed from the `ip_adapter_image` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
                 The output format of the generated video. Choose between `torch.FloatTensor`, `PIL.Image` or
                 `np.array`.
@@ -842,8 +833,8 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
         Examples:
 
         Returns:
-            [`AnimateDiffImgToVideoPipelineOutput`] or `tuple`:
-                If `return_dict` is `True`, [`AnimateDiffImgToVideoPipelineOutput`] 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
@@ -902,12 +893,9 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
             prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])
 
         if ip_adapter_image is not None:
-            output_hidden_state = False if isinstance(self.unet.encoder_hid_proj, ImageProjection) else True
-            image_embeds, negative_image_embeds = self.encode_image(
-                ip_adapter_image, device, num_videos_per_prompt, output_hidden_state
+            image_embeds = self.prepare_ip_adapter_image_embeds(
+                ip_adapter_image, ip_adapter_image_embeds, device, batch_size * num_videos_per_prompt
             )
-            if do_classifier_free_guidance:
-                image_embeds = torch.cat([negative_image_embeds, image_embeds])
 
         # 4. Preprocess image
         image = self.image_processor.preprocess(image, height=height, width=width)
@@ -936,7 +924,11 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
         extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
 
         # 8. Add image embeds for IP-Adapter
-        added_cond_kwargs = {"image_embeds": image_embeds} if ip_adapter_image is not None else None
+        added_cond_kwargs = (
+            {"image_embeds": image_embeds}
+            if ip_adapter_image is not None or ip_adapter_image_embeds is not None
+            else None
+        )
 
         # 9. Denoising loop
         num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
@@ -970,14 +962,13 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
                         callback(i, t, latents)
 
         if output_type == "latent":
-            return AnimateDiffImgToVideoPipelineOutput(frames=latents)
+            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
@@ -986,4 +977,4 @@ class AnimateDiffImgToVideoPipeline(DiffusionPipeline, TextualInversionLoaderMix
         if not return_dict:
             return (video,)
 
-        return AnimateDiffImgToVideoPipelineOutput(frames=video)
+        return AnimateDiffPipelineOutput(frames=video)

examples/community/pipeline_demofusion_sdxl.py

@@ -23,7 +23,7 @@ from diffusers.models.attention_processor import (
     XFormersAttnProcessor,
 )
 from diffusers.models.lora import adjust_lora_scale_text_encoder
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.schedulers import KarrasDiffusionSchedulers
 from diffusers.utils import (
     is_accelerate_available,
@@ -93,7 +93,9 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-class DemoFusionSDXLPipeline(DiffusionPipeline, FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin):
+class DemoFusionSDXLPipeline(
+    DiffusionPipeline, StableDiffusionMixin, FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
+):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion XL.
 
@@ -176,39 +178,6 @@ class DemoFusionSDXLPipeline(DiffusionPipeline, FromSingleFileMixin, LoraLoaderM
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     def encode_prompt(
         self,
         prompt: str,

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_sdxl_style_aligned.py

@@ -51,7 +51,7 @@ from diffusers.models.attention_processor import (
     XFormersAttnProcessor,
 )
 from diffusers.models.lora import adjust_lora_scale_text_encoder
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
 from diffusers.schedulers import KarrasDiffusionSchedulers
 from diffusers.utils import (
@@ -389,6 +389,7 @@ def retrieve_latents(
 
 class StyleAlignedSDXLPipeline(
     DiffusionPipeline,
+    StableDiffusionMixin,
     FromSingleFileMixin,
     StableDiffusionXLLoraLoaderMixin,
     TextualInversionLoaderMixin,
@@ -504,39 +505,6 @@ class StyleAlignedSDXLPipeline(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     def encode_prompt(
         self,
         prompt: str,
@@ -1187,34 +1155,6 @@ class StyleAlignedSDXLPipeline(
             self.vae.decoder.conv_in.to(dtype)
             self.vae.decoder.mid_block.to(dtype)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     def _enable_shared_attention_processors(
         self,
         share_attention: bool,
@@ -1361,65 +1301,6 @@ class StyleAlignedSDXLPipeline(
             self._style_aligned_norm_layers = None
             self._disable_shared_attention_processors()
 
-    def fuse_qkv_projections(self, unet: bool = True, vae: bool = True):
-        """
-        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query,
-        key, value) are fused. For cross-attention modules, key and value projection matrices are fused.
-
-        <Tip warning={true}>
-
-        This API is 🧪 experimental.
-
-        </Tip>
-
-        Args:
-            unet (`bool`, defaults to `True`): To apply fusion on the UNet.
-            vae (`bool`, defaults to `True`): To apply fusion on the VAE.
-        """
-        self.fusing_unet = False
-        self.fusing_vae = False
-
-        if unet:
-            self.fusing_unet = True
-            self.unet.fuse_qkv_projections()
-            self.unet.set_attn_processor(FusedAttnProcessor2_0())
-
-        if vae:
-            if not isinstance(self.vae, AutoencoderKL):
-                raise ValueError("`fuse_qkv_projections()` is only supported for the VAE of type `AutoencoderKL`.")
-
-            self.fusing_vae = True
-            self.vae.fuse_qkv_projections()
-            self.vae.set_attn_processor(FusedAttnProcessor2_0())
-
-    def unfuse_qkv_projections(self, unet: bool = True, vae: bool = True):
-        """Disable QKV projection fusion if enabled.
-
-        <Tip warning={true}>
-
-        This API is 🧪 experimental.
-
-        </Tip>
-
-        Args:
-            unet (`bool`, defaults to `True`): To apply fusion on the UNet.
-            vae (`bool`, defaults to `True`): To apply fusion on the VAE.
-
-        """
-        if unet:
-            if not self.fusing_unet:
-                logger.warning("The UNet was not initially fused for QKV projections. Doing nothing.")
-            else:
-                self.unet.unfuse_qkv_projections()
-                self.fusing_unet = False
-
-        if vae:
-            if not self.fusing_vae:
-                logger.warning("The VAE was not initially fused for QKV projections. Doing nothing.")
-            else:
-                self.vae.unfuse_qkv_projections()
-                self.fusing_vae = False
-
     # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
     def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):
         """
@@ -1769,7 +1650,7 @@ class StyleAlignedSDXLPipeline(
 
         # 4. Prepare timesteps
         def denoising_value_valid(dnv):
-            return isinstance(self.denoising_end, float) and 0 < dnv < 1
+            return isinstance(dnv, float) and 0 < dnv < 1
 
         timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, num_inference_steps, device, timesteps)
 
@@ -1778,7 +1659,7 @@ class StyleAlignedSDXLPipeline(
                 num_inference_steps,
                 strength,
                 device,
-                denoising_start=self.denoising_start if denoising_value_valid else None,
+                denoising_start=self.denoising_start if denoising_value_valid(self.denoising_start) else None,
             )
 
             # check that number of inference steps is not < 1 - as this doesn't make sense

examples/community/pipeline_stable_diffusion_xl_controlnet_adapter.py

@@ -33,7 +33,7 @@ from diffusers.models.attention_processor import (
 )
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.pipelines.controlnet.multicontrolnet import MultiControlNetModel
-from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
 from diffusers.schedulers import KarrasDiffusionSchedulers
 from diffusers.utils import (
@@ -158,7 +158,11 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
 
 
 class StableDiffusionXLControlNetAdapterPipeline(
-    DiffusionPipeline, FromSingleFileMixin, StableDiffusionXLLoraLoaderMixin, TextualInversionLoaderMixin
+    DiffusionPipeline,
+    StableDiffusionMixin,
+    FromSingleFileMixin,
+    StableDiffusionXLLoraLoaderMixin,
+    TextualInversionLoaderMixin,
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion augmented with T2I-Adapter
@@ -234,39 +238,6 @@ class StableDiffusionXLControlNetAdapterPipeline(
         )
         self.default_sample_size = self.unet.config.sample_size
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
     def encode_prompt(
         self,
@@ -863,34 +834,6 @@ class StableDiffusionXLControlNetAdapterPipeline(
 
         return height, width
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     def prepare_control_image(
         self,
         image,

examples/community/pipeline_stable_diffusion_xl_controlnet_adapter_inpaint.py

@@ -52,6 +52,7 @@ from diffusers.models.attention_processor import (
 )
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.pipelines.controlnet.multicontrolnet import MultiControlNetModel
+from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
 from diffusers.schedulers import KarrasDiffusionSchedulers
 from diffusers.utils import (
@@ -303,7 +304,9 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-class StableDiffusionXLControlNetAdapterInpaintPipeline(DiffusionPipeline, FromSingleFileMixin, LoraLoaderMixin):
+class StableDiffusionXLControlNetAdapterInpaintPipeline(
+    DiffusionPipeline, StableDiffusionMixin, FromSingleFileMixin, LoraLoaderMixin
+):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion augmented with T2I-Adapter
     https://arxiv.org/abs/2302.08453
@@ -383,39 +386,6 @@ class StableDiffusionXLControlNetAdapterInpaintPipeline(DiffusionPipeline, FromS
         )
         self.default_sample_size = self.unet.config.sample_size
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
     def encode_prompt(
         self,
@@ -1207,34 +1177,6 @@ class StableDiffusionXLControlNetAdapterInpaintPipeline(DiffusionPipeline, FromS
 
         return height, width
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     def prepare_control_image(
         self,
         image,
@@ -1563,14 +1505,14 @@ class StableDiffusionXLControlNetAdapterInpaintPipeline(DiffusionPipeline, FromS
 
         # 4. set timesteps
         def denoising_value_valid(dnv):
-            return isinstance(denoising_end, float) and 0 < dnv < 1
+            return isinstance(dnv, float) and 0 < dnv < 1
 
         self.scheduler.set_timesteps(num_inference_steps, device=device)
         timesteps, num_inference_steps = self.get_timesteps(
             num_inference_steps,
             strength,
             device,
-            denoising_start=denoising_start if denoising_value_valid else None,
+            denoising_start=denoising_start if denoising_value_valid(denoising_start) else None,
         )
         # check that number of inference steps is not < 1 - as this doesn't make sense
         if num_inference_steps < 1:

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/pipeline_zero1to3.py

@@ -22,18 +22,16 @@ from transformers import CLIPFeatureExtractor, CLIPVisionModelWithProjection
 #     randn_tensor,
 #     replace_example_docstring,
 # )
-# from ..pipeline_utils import DiffusionPipeline
+# from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 # from . import StableDiffusionPipelineOutput
 # from .safety_checker import StableDiffusionSafetyChecker
-from diffusers import AutoencoderKL, DiffusionPipeline, UNet2DConditionModel
+from diffusers import AutoencoderKL, DiffusionPipeline, StableDiffusionMixin, UNet2DConditionModel
 from diffusers.configuration_utils import ConfigMixin, FrozenDict
 from diffusers.models.modeling_utils import ModelMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput, StableDiffusionSafetyChecker
 from diffusers.schedulers import KarrasDiffusionSchedulers
 from diffusers.utils import (
     deprecate,
-    is_accelerate_available,
-    is_accelerate_version,
     logging,
     replace_example_docstring,
 )
@@ -68,7 +66,7 @@ class CCProjection(ModelMixin, ConfigMixin):
         return self.projection(x)
 
 
-class Zero1to3StableDiffusionPipeline(DiffusionPipeline):
+class Zero1to3StableDiffusionPipeline(DiffusionPipeline, StableDiffusionMixin):
     r"""
     Pipeline for single view conditioned novel view generation using Zero1to3.
 
@@ -187,109 +185,6 @@ class Zero1to3StableDiffusionPipeline(DiffusionPipeline):
         self.register_to_config(requires_safety_checker=requires_safety_checker)
         # self.model_mode = None
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding.
-
-        When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several
-        steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously invoked, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding.
-
-        When this option is enabled, the VAE will split the input tensor into tiles to compute decoding and encoding in
-        several steps. This is useful to save a large amount of memory and to allow the processing of larger images.
-        """
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously invoked, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
-    def enable_sequential_cpu_offload(self, gpu_id=0):
-        r"""
-        Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
-        text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
-        `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
-        Note that offloading happens on a submodule basis. Memory savings are higher than with
-        `enable_model_cpu_offload`, but performance is lower.
-        """
-        if is_accelerate_available() and is_accelerate_version(">=", "0.14.0"):
-            from accelerate import cpu_offload
-        else:
-            raise ImportError("`enable_sequential_cpu_offload` requires `accelerate v0.14.0` or higher")
-
-        device = torch.device(f"cuda:{gpu_id}")
-
-        if self.device.type != "cpu":
-            self.to("cpu", silence_dtype_warnings=True)
-            torch.cuda.empty_cache()  # otherwise we don't see the memory savings (but they probably exist)
-
-        for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]:
-            cpu_offload(cpu_offloaded_model, device)
-
-        if self.safety_checker is not None:
-            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)
-
-    def enable_model_cpu_offload(self, gpu_id=0):
-        r"""
-        Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
-        to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
-        method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
-        `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
-        """
-        if is_accelerate_available() and is_accelerate_version(">=", "0.17.0.dev0"):
-            from accelerate import cpu_offload_with_hook
-        else:
-            raise ImportError("`enable_model_offload` requires `accelerate v0.17.0` or higher.")
-
-        device = torch.device(f"cuda:{gpu_id}")
-
-        if self.device.type != "cpu":
-            self.to("cpu", silence_dtype_warnings=True)
-            torch.cuda.empty_cache()  # otherwise we don't see the memory savings (but they probably exist)
-
-        hook = None
-        for cpu_offloaded_model in [self.text_encoder, self.unet, self.vae]:
-            _, hook = cpu_offload_with_hook(cpu_offloaded_model, device, prev_module_hook=hook)
-
-        if self.safety_checker is not None:
-            _, hook = cpu_offload_with_hook(self.safety_checker, device, prev_module_hook=hook)
-
-        # We'll offload the last model manually.
-        self.final_offload_hook = hook
-
-    @property
-    def _execution_device(self):
-        r"""
-        Returns the device on which the pipeline's models will be executed. After calling
-        `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module
-        hooks.
-        """
-        if not hasattr(self.unet, "_hf_hook"):
-            return self.device
-        for module in self.unet.modules():
-            if (
-                hasattr(module, "_hf_hook")
-                and hasattr(module._hf_hook, "execution_device")
-                and module._hf_hook.execution_device is not None
-            ):
-                return torch.device(module._hf_hook.execution_device)
-        return self.device
-
     def _encode_prompt(
         self,
         prompt,

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/community/sd_text2img_k_diffusion.py

@@ -19,9 +19,9 @@ from typing import Callable, List, Optional, Union
 import torch
 from k_diffusion.external import CompVisDenoiser, CompVisVDenoiser
 
-from diffusers import DiffusionPipeline, LMSDiscreteScheduler
+from diffusers import DiffusionPipeline, LMSDiscreteScheduler, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
-from diffusers.utils import is_accelerate_available, logging
+from diffusers.utils import logging
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -41,7 +41,7 @@ class ModelWrapper:
         return self.model(*args, encoder_hidden_states=encoder_hidden_states, **kwargs).sample
 
 
-class StableDiffusionPipeline(DiffusionPipeline):
+class StableDiffusionPipeline(DiffusionPipeline, StableDiffusionMixin):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion.
 
@@ -120,68 +120,6 @@ class StableDiffusionPipeline(DiffusionPipeline):
         sampling = getattr(library, "sampling")
         self.sampler = getattr(sampling, scheduler_type)
 
-    def enable_attention_slicing(self, slice_size: Optional[Union[str, int]] = "auto"):
-        r"""
-        Enable sliced attention computation.
-
-        When this option is enabled, the attention module will split the input tensor in slices, to compute attention
-        in several steps. This is useful to save some memory in exchange for a small speed decrease.
-
-        Args:
-            slice_size (`str` or `int`, *optional*, defaults to `"auto"`):
-                When `"auto"`, halves the input to the attention heads, so attention will be computed in two steps. If
-                a number is provided, uses as many slices as `attention_head_dim // slice_size`. In this case,
-                `attention_head_dim` must be a multiple of `slice_size`.
-        """
-        if slice_size == "auto":
-            # half the attention head size is usually a good trade-off between
-            # speed and memory
-            slice_size = self.unet.config.attention_head_dim // 2
-        self.unet.set_attention_slice(slice_size)
-
-    def disable_attention_slicing(self):
-        r"""
-        Disable sliced attention computation. If `enable_attention_slicing` was previously invoked, this method will go
-        back to computing attention in one step.
-        """
-        # set slice_size = `None` to disable `attention slicing`
-        self.enable_attention_slicing(None)
-
-    def enable_sequential_cpu_offload(self, gpu_id=0):
-        r"""
-        Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
-        text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
-        `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
-        """
-        if is_accelerate_available():
-            from accelerate import cpu_offload
-        else:
-            raise ImportError("Please install accelerate via `pip install accelerate`")
-
-        device = torch.device(f"cuda:{gpu_id}")
-
-        for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae, self.safety_checker]:
-            if cpu_offloaded_model is not None:
-                cpu_offload(cpu_offloaded_model, device)
-
-    @property
-    def _execution_device(self):
-        r"""
-        Returns the device on which the pipeline's models will be executed. After calling
-        `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module
-        hooks.
-        """
-        if self.device != torch.device("meta") or not hasattr(self.unet, "_hf_hook"):
-            return self.device
-        for module in self.unet.modules():
-            if (
-                hasattr(module, "_hf_hook")
-                and hasattr(module._hf_hook, "execution_device")
-                and module._hf_hook.execution_device is not None
-            ):
-                return torch.device(module._hf_hook.execution_device)
-        return self.device
-
     def _encode_prompt(self, prompt, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt):
         r"""
         Encodes the prompt into text encoder hidden states.