Run ruff format

* move defaultModel logic to modelsLoaded and update to work for key instead of name/base/type string

* lint fix

---------

Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>

.github/actions/install-frontend-deps/action.yml

@@ -0,0 +1,33 @@
+name: install frontend dependencies
+description: Installs frontend dependencies with pnpm, with caching
+runs:
+  using: 'composite'
+  steps:
+    - name: setup node 18
+      uses: actions/setup-node@v4
+      with:
+        node-version: '18'
+
+    - name: setup pnpm
+      uses: pnpm/action-setup@v2
+      with:
+        version: 8
+        run_install: false
+
+    - name: get pnpm store directory
+      shell: bash
+      run: |
+        echo "STORE_PATH=$(pnpm store path --silent)" >> $GITHUB_ENV
+
+    - name: setup cache
+      uses: actions/cache@v4
+      with:
+        path: ${{ env.STORE_PATH }}
+        key: ${{ runner.os }}-pnpm-store-${{ hashFiles('**/pnpm-lock.yaml') }}
+        restore-keys: |
+          ${{ runner.os }}-pnpm-store-
+
+    - name: install frontend dependencies
+      run: pnpm install --prefer-frozen-lockfile
+      shell: bash
+      working-directory: invokeai/frontend/web

.github/pr_labels.yml

@@ -1,59 +1,59 @@
-Root:
+root:
 - changed-files:
   - any-glob-to-any-file: '*'
 
-PythonDeps:
+python-deps:
 - changed-files:
   - any-glob-to-any-file: 'pyproject.toml'
 
-Python:
+python:
 - changed-files:
   - all-globs-to-any-file: 
     - 'invokeai/**'
     - '!invokeai/frontend/web/**'
 
-PythonTests:
+python-tests:
 - changed-files:
   - any-glob-to-any-file: 'tests/**'
 
-CICD:
+ci-cd:
 - changed-files:
   - any-glob-to-any-file: .github/**
 
-Docker:
+docker:
 - changed-files:
   - any-glob-to-any-file: docker/**
 
-Installer:
+installer:
 - changed-files:
   - any-glob-to-any-file: installer/**
 
-Documentation:
+docs:
 - changed-files:
   - any-glob-to-any-file: docs/**
 
-Invocations:
+invocations:
 - changed-files:
   - any-glob-to-any-file: 'invokeai/app/invocations/**'
 
-Backend:
+backend:
 - changed-files:
   - any-glob-to-any-file: 'invokeai/backend/**'
 
-Api:
+api:
 - changed-files:
   - any-glob-to-any-file: 'invokeai/app/api/**'
 
-Services:
+services:
 - changed-files:
   - any-glob-to-any-file: 'invokeai/app/services/**'
 
-FrontendDeps:
+frontend-deps:
 - changed-files:
   - any-glob-to-any-file:
     - '**/*/package.json'
     - '**/*/pnpm-lock.yaml'
 
-Frontend:
+frontend:
 - changed-files:
   - any-glob-to-any-file: 'invokeai/frontend/web/**'

.github/workflows/build-container.yml

@@ -11,7 +11,7 @@ on:
       - 'docker/docker-entrypoint.sh'
       - 'workflows/build-container.yml'
     tags:
-      - 'v*'
+      - 'v*.*.*'
   workflow_dispatch:
 
 permissions:

.github/workflows/build-installer.yml

@@ -0,0 +1,45 @@
+# Builds and uploads the installer and python build artifacts.
+
+name: build installer
+
+on:
+  workflow_dispatch:
+  workflow_call:
+
+jobs:
+  build-installer:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5 # expected run time: <2 min
+    steps:
+      - name: checkout
+        uses: actions/checkout@v4
+
+      - name: setup python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.10'
+          cache: pip
+          cache-dependency-path: pyproject.toml
+
+      - name: install pypa/build
+        run: pip install --upgrade build
+
+      - name: setup frontend
+        uses: ./.github/actions/install-frontend-deps
+
+      - name: create installer
+        id: create_installer
+        run: ./create_installer.sh
+        working-directory: installer
+
+      - name: upload python distribution artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: ${{ steps.create_installer.outputs.DIST_PATH }}
+
+      - name: upload installer artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ steps.create_installer.outputs.INSTALLER_FILENAME }}
+          path: ${{ steps.create_installer.outputs.INSTALLER_PATH }}

.github/workflows/frontend-checks.yml

@@ -0,0 +1,68 @@
+# Runs frontend code quality checks.
+#
+# Checks for changes to frontend files before running the checks.
+# When manually triggered or when called from another workflow, always runs the checks.
+
+name: 'frontend checks'
+
+on:
+  push:
+    branches:
+      - 'main'
+  pull_request:
+    types:
+      - 'ready_for_review'
+      - 'opened'
+      - 'synchronize'
+  merge_group:
+  workflow_dispatch:
+  workflow_call:
+
+defaults:
+  run:
+    working-directory: invokeai/frontend/web
+
+jobs:
+  frontend-checks:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10 # expected run time: <2 min
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: check for changed frontend files
+        if: ${{ github.event_name != 'workflow_dispatch' && github.event_name != 'workflow_call' }}
+        id: changed-files
+        uses: tj-actions/changed-files@v42
+        with:
+          files_yaml: |
+            frontend:
+              - 'invokeai/frontend/web/**'
+
+      - name: install dependencies
+        if: ${{ steps.changed-files.outputs.frontend_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        uses: ./.github/actions/install-frontend-deps
+
+      - name: tsc
+        if: ${{ steps.changed-files.outputs.frontend_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: 'pnpm lint:tsc'
+        shell: bash
+
+      - name: dpdm
+        if: ${{ steps.changed-files.outputs.frontend_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: 'pnpm lint:dpdm'
+        shell: bash
+
+      - name: eslint
+        if: ${{ steps.changed-files.outputs.frontend_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: 'pnpm lint:eslint'
+        shell: bash
+
+      - name: prettier
+        if: ${{ steps.changed-files.outputs.frontend_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: 'pnpm lint:prettier'
+        shell: bash
+
+      - name: knip
+        if: ${{ steps.changed-files.outputs.frontend_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: 'pnpm lint:knip'
+        shell: bash

.github/workflows/frontend-tests.yml

@@ -0,0 +1,48 @@
+# Runs frontend tests.
+#
+# Checks for changes to frontend files before running the tests.
+# When manually triggered or called from another workflow, always runs the tests.
+
+name: 'frontend tests'
+
+on:
+  push:
+    branches:
+      - 'main'
+  pull_request:
+    types:
+      - 'ready_for_review'
+      - 'opened'
+      - 'synchronize'
+  merge_group:
+  workflow_dispatch:
+  workflow_call:
+
+defaults:
+  run:
+    working-directory: invokeai/frontend/web
+
+jobs:
+  frontend-tests:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10 # expected run time: <2 min
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: check for changed frontend files
+        if: ${{ github.event_name != 'workflow_dispatch' && github.event_name != 'workflow_call' }}
+        id: changed-files
+        uses: tj-actions/changed-files@v42
+        with:
+          files_yaml: |
+            frontend:
+              - 'invokeai/frontend/web/**'
+
+      - name: install dependencies
+        if: ${{ steps.changed-files.outputs.frontend_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        uses: ./.github/actions/install-frontend-deps
+
+      - name: vitest
+        if: ${{ steps.changed-files.outputs.frontend_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: 'pnpm test:no-watch'
+        shell: bash

.github/workflows/label-pr.yml

@@ -1,6 +1,6 @@
-name: "Pull Request Labeler"
+name: 'label PRs'
 on:
-- pull_request_target
+  - pull_request_target
 
 jobs:
   labeler:
@@ -9,8 +9,10 @@ jobs:
       pull-requests: write
     runs-on: ubuntu-latest
     steps:
-      - name: Checkout
+      - name: checkout
         uses: actions/checkout@v4
-      - uses: actions/labeler@v5
+
+      - name: label PRs
+        uses: actions/labeler@v5
         with:
-          configuration-path: .github/pr_labels.yml
+          configuration-path: .github/pr_labels.yml

.github/workflows/lint-frontend.yml

@@ -1,45 +0,0 @@
-name: Lint frontend
-
-on:
-  pull_request:
-    types:
-      - 'ready_for_review'
-      - 'opened'
-      - 'synchronize'
-  push:
-    branches:
-      - 'main'
-  merge_group:
-  workflow_dispatch:
-
-defaults:
-  run:
-    working-directory: invokeai/frontend/web
-
-jobs:
-  lint-frontend:
-    if: github.event.pull_request.draft == false
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Setup Node 18
-        uses: actions/setup-node@v4
-        with:
-          node-version: '18'
-      - name: Checkout
-        uses: actions/checkout@v4
-      - name: Setup pnpm
-        uses: pnpm/action-setup@v2
-        with:
-          version: '8.12.1'
-      - name: Install dependencies
-        run: 'pnpm install --prefer-frozen-lockfile'
-      - name: Typescript
-        run: 'pnpm run lint:tsc'
-      - name: Madge
-        run: 'pnpm run lint:dpdm'
-      - name: ESLint
-        run: 'pnpm run lint:eslint'
-      - name: Prettier
-        run: 'pnpm run lint:prettier'
-      - name: Knip
-        run: 'pnpm run lint:knip'

.github/workflows/mkdocs-material.yml

@@ -1,51 +1,49 @@
-name: mkdocs-material
+# This is a mostly a copy-paste from https://github.com/squidfunk/mkdocs-material/blob/master/docs/publishing-your-site.md
+
+name: mkdocs
+
 on:
   push:
     branches:
-      - 'refs/heads/main'
+      - main
+  workflow_dispatch:
 
 permissions:
-    contents: write
+  contents: write
 
 jobs:
-  mkdocs-material:
+  deploy:
     if: github.event.pull_request.draft == false
     runs-on: ubuntu-latest
     env:
       REPO_URL: '${{ github.server_url }}/${{ github.repository }}'
       REPO_NAME: '${{ github.repository }}'
       SITE_URL: 'https://${{ github.repository_owner }}.github.io/InvokeAI'
+
     steps:
-      - name: checkout sources
-        uses: actions/checkout@v3
-        with:
-          fetch-depth: 0
+      - name: checkout
+        uses: actions/checkout@v4
 
       - name: setup python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@v5
         with:
           python-version: '3.10'
           cache: pip
           cache-dependency-path: pyproject.toml
 
-      - name: install requirements
-        env:
-          PIP_USE_PEP517: 1
-        run: |
-          python -m \
-            pip install ".[docs]"
+      - name: set cache id
+        run: echo "cache_id=$(date --utc '+%V')" >> $GITHUB_ENV
 
-      - name: confirm buildability
-        run: |
-          python -m \
-            mkdocs build \
-            --clean \
-            --verbose
+      - name: use cache
+        uses: actions/cache@v4
+        with:
+          key: mkdocs-material-${{ env.cache_id }}
+          path: .cache
+          restore-keys: |
+            mkdocs-material-
 
-      - name: deploy to gh-pages
-        if: ${{ github.ref == 'refs/heads/main' }}
-        run: |
-          python -m \
-            mkdocs gh-deploy \
-            --clean \
-            --force
+      - name: install dependencies
+        run: python -m pip install ".[docs]"
+
+      - name: build & deploy
+        run: mkdocs gh-deploy --force

.github/workflows/pypi-release.yml

@@ -1,67 +0,0 @@
-name: PyPI Release
-
-on:
-  workflow_dispatch:
-    inputs:
-      publish_package:
-        description: 'Publish build on PyPi? [true/false]'
-        required: true
-        default: 'false'
-
-jobs:
-  build-and-release:
-    if: github.repository == 'invoke-ai/InvokeAI'
-    runs-on: ubuntu-22.04
-    env:
-      TWINE_USERNAME: __token__
-      TWINE_PASSWORD: ${{ secrets.PYPI_API_TOKEN }}
-      TWINE_NON_INTERACTIVE: 1
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-
-      - name: Setup Node 18
-        uses: actions/setup-node@v4
-        with:
-          node-version: '18'
-
-      - name: Setup pnpm
-        uses: pnpm/action-setup@v2
-        with:
-          version: '8.12.1'
-
-      - name: Install frontend dependencies
-        run: pnpm install --prefer-frozen-lockfile
-        working-directory: invokeai/frontend/web
-
-      - name: Build frontend
-        run: pnpm run build
-        working-directory: invokeai/frontend/web
-
-      - name: Install python dependencies
-        run: pip install --upgrade build twine
-
-      - name: Build python package
-        run: python3 -m build
-
-      - name: Upload build as workflow artifact
-        uses: actions/upload-artifact@v4
-        with:
-          name: dist
-          path: dist
-
-      - name: Check distribution
-        run: twine check dist/*
-
-      - name: Check PyPI versions
-        if: github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/heads/release/')
-        run: |
-          pip install --upgrade requests
-          python -c "\
-          import scripts.pypi_helper; \
-          EXISTS=scripts.pypi_helper.local_on_pypi(); \
-          print(f'PACKAGE_EXISTS={EXISTS}')" >> $GITHUB_ENV
-
-      - name: Publish build on PyPi
-        if: env.PACKAGE_EXISTS == 'False' && env.TWINE_PASSWORD != '' && github.event.inputs.publish_package == 'true'
-        run: twine upload dist/*

.github/workflows/python-checks.yml

@@ -0,0 +1,64 @@
+# Runs python code quality checks.
+#
+# Checks for changes to python files before running the checks.
+# When manually triggered or called from another workflow, always runs the tests.
+#
+# TODO: Add mypy or pyright to the checks.
+
+name: 'python checks'
+
+on:
+  push:
+    branches:
+      - 'main'
+  pull_request:
+    types:
+      - 'ready_for_review'
+      - 'opened'
+      - 'synchronize'
+  merge_group:
+  workflow_dispatch:
+  workflow_call:
+
+jobs:
+  python-checks:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5 # expected run time: <1 min
+    steps:
+      - name: checkout
+        uses: actions/checkout@v4
+
+      - name: check for changed python files
+        if: ${{ github.event_name != 'workflow_dispatch' && github.event_name != 'workflow_call' }}
+        id: changed-files
+        uses: tj-actions/changed-files@v42
+        with:
+          files_yaml: |
+            python:
+              - 'pyproject.toml'
+              - 'invokeai/**'
+              - '!invokeai/frontend/web/**'
+              - 'tests/**'
+
+      - name: setup python
+        if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.10'
+          cache: pip
+          cache-dependency-path: pyproject.toml
+
+      - name: install ruff
+        if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: pip install ruff
+        shell: bash
+
+      - name: ruff check
+        if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: ruff check --output-format=github .
+        shell: bash
+
+      - name: ruff format
+        if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: ruff format --check .
+        shell: bash

.github/workflows/python-tests.yml

@@ -0,0 +1,94 @@
+# Runs python tests on a matrix of python versions and platforms.
+#
+# Checks for changes to python files before running the tests.
+# When manually triggered or called from another workflow, always runs the tests.
+
+name: 'python tests'
+
+on:
+  push:
+    branches:
+      - 'main'
+  pull_request:
+    types:
+      - 'ready_for_review'
+      - 'opened'
+      - 'synchronize'
+  merge_group:
+  workflow_dispatch:
+  workflow_call:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  matrix:
+    strategy:
+      matrix:
+        python-version:
+          - '3.10'
+          - '3.11'
+        platform:
+          - linux-cuda-11_7
+          - linux-rocm-5_2
+          - linux-cpu
+          - macos-default
+          - windows-cpu
+        include:
+          - platform: linux-cuda-11_7
+            os: ubuntu-22.04
+            github-env: $GITHUB_ENV
+          - platform: linux-rocm-5_2
+            os: ubuntu-22.04
+            extra-index-url: 'https://download.pytorch.org/whl/rocm5.2'
+            github-env: $GITHUB_ENV
+          - platform: linux-cpu
+            os: ubuntu-22.04
+            extra-index-url: 'https://download.pytorch.org/whl/cpu'
+            github-env: $GITHUB_ENV
+          - platform: macos-default
+            os: macOS-12
+            github-env: $GITHUB_ENV
+          - platform: windows-cpu
+            os: windows-2022
+            github-env: $env:GITHUB_ENV
+    name: 'py${{ matrix.python-version }}: ${{ matrix.platform }}'
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 15 # expected run time: 2-6 min, depending on platform
+    env:
+      PIP_USE_PEP517: '1'
+    steps:
+      - name: checkout
+        uses: actions/checkout@v4
+
+      - name: check for changed python files
+        if: ${{ github.event_name != 'workflow_dispatch' && github.event_name != 'workflow_call' }}
+        id: changed-files
+        uses: tj-actions/changed-files@v42
+        with:
+          files_yaml: |
+            python:
+              - 'pyproject.toml'
+              - 'invokeai/**'
+              - '!invokeai/frontend/web/**'
+              - 'tests/**'
+
+      - name: setup python
+        if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: pip
+          cache-dependency-path: pyproject.toml
+
+      - name: install dependencies
+        if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        env:
+          PIP_EXTRA_INDEX_URL: ${{ matrix.extra-index-url }}
+        run: >
+          pip3 install --editable=".[test]"
+
+      - name: run pytest
+        if: ${{ steps.changed-files.outputs.python_any_changed == 'true' || github.event_name == 'workflow_dispatch' || github.event_name == 'workflow_call' }}
+        run: pytest

.github/workflows/release.yml

@@ -0,0 +1,96 @@
+# Main release workflow. Triggered on tag push or manual trigger.
+#
+# - Runs all code checks and tests
+# - Verifies the app version matches the tag version.
+# - Builds the installer and build, uploading them as artifacts.
+# - Publishes to TestPyPI and PyPI. Both are conditional on the previous steps passing and require a manual approval.
+#
+# See docs/RELEASE.md for more information on the release process.
+
+name: release
+
+on:
+  push:
+    tags:
+      - 'v*'
+  workflow_dispatch:
+
+jobs:
+  check-version:
+    runs-on: ubuntu-latest
+    steps:
+      - name: checkout
+        uses: actions/checkout@v4
+
+      - name: check python version
+        uses: samuelcolvin/check-python-version@v4
+        id: check-python-version
+        with:
+          version_file_path: invokeai/version/invokeai_version.py
+
+  frontend-checks:
+    uses: ./.github/workflows/frontend-checks.yml
+
+  frontend-tests:
+    uses: ./.github/workflows/frontend-tests.yml
+
+  python-checks:
+    uses: ./.github/workflows/python-checks.yml
+
+  python-tests:
+    uses: ./.github/workflows/python-tests.yml
+
+  build:
+    uses: ./.github/workflows/build-installer.yml
+
+  publish-testpypi:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5 # expected run time: <1 min
+    needs:
+      [
+        check-version,
+        frontend-checks,
+        frontend-tests,
+        python-checks,
+        python-tests,
+        build,
+      ]
+    environment:
+      name: testpypi
+      url: https://test.pypi.org/p/invokeai
+    steps:
+      - name: download distribution from build job
+        uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+      - name: publish distribution to TestPyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          repository-url: https://test.pypi.org/legacy/
+
+  publish-pypi:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5 # expected run time: <1 min
+    needs:
+      [
+        check-version,
+        frontend-checks,
+        frontend-tests,
+        python-checks,
+        python-tests,
+        build,
+      ]
+    environment:
+      name: pypi
+      url: https://pypi.org/p/invokeai
+    steps:
+      - name: download distribution from build job
+        uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+      - name: publish distribution to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1

.github/workflows/style-checks.yml

@@ -1,24 +0,0 @@
-name: style checks
-
-on:
-  pull_request:
-  push:
-    branches: main
-
-jobs:
-  ruff:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v3
-
-      - name: Setup Python
-        uses: actions/setup-python@v4
-        with:
-          python-version: '3.10'
-
-      - name: Install dependencies with pip
-        run: |
-          pip install ruff
-
-      - run: ruff check --output-format=github .
-      - run: ruff format --check .

.github/workflows/test-invoke-pip.yml

@@ -1,129 +0,0 @@
-name: Test invoke.py pip
-on:
-  push:
-    branches:
-      - 'main'
-  pull_request:
-    types:
-      - 'ready_for_review'
-      - 'opened'
-      - 'synchronize'
-  merge_group:
-  workflow_dispatch:
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
-  cancel-in-progress: true
-
-jobs:
-  matrix:
-    if: github.event.pull_request.draft == false
-    strategy:
-      matrix:
-        python-version:
-          # - '3.9'
-          - '3.10'
-        pytorch:
-          - linux-cuda-11_7
-          - linux-rocm-5_2
-          - linux-cpu
-          - macos-default
-          - windows-cpu
-        include:
-          - pytorch: linux-cuda-11_7
-            os: ubuntu-22.04
-            github-env: $GITHUB_ENV
-          - pytorch: linux-rocm-5_2
-            os: ubuntu-22.04
-            extra-index-url: 'https://download.pytorch.org/whl/rocm5.2'
-            github-env: $GITHUB_ENV
-          - pytorch: linux-cpu
-            os: ubuntu-22.04
-            extra-index-url: 'https://download.pytorch.org/whl/cpu'
-            github-env: $GITHUB_ENV
-          - pytorch: macos-default
-            os: macOS-12
-            github-env: $GITHUB_ENV
-          - pytorch: windows-cpu
-            os: windows-2022
-            github-env: $env:GITHUB_ENV
-    name: ${{ matrix.pytorch }} on ${{ matrix.python-version }}
-    runs-on: ${{ matrix.os }}
-    env:
-      PIP_USE_PEP517: '1'
-    steps:
-      - name: Checkout sources
-        id: checkout-sources
-        uses: actions/checkout@v3
-
-      - name: Check for changed python files
-        id: changed-files
-        uses: tj-actions/changed-files@v41
-        with:
-          files_yaml: |
-            python:
-              - 'pyproject.toml'
-              - 'invokeai/**'
-              - '!invokeai/frontend/web/**'
-              - 'tests/**'
-
-      - name: set test prompt to main branch validation
-        if: steps.changed-files.outputs.python_any_changed == 'true'
-        run: echo "TEST_PROMPTS=tests/validate_pr_prompt.txt" >> ${{ matrix.github-env }}
-
-      - name: setup python
-        if: steps.changed-files.outputs.python_any_changed == 'true'
-        uses: actions/setup-python@v4
-        with:
-          python-version: ${{ matrix.python-version }}
-          cache: pip
-          cache-dependency-path: pyproject.toml
-
-      - name: install invokeai
-        if: steps.changed-files.outputs.python_any_changed == 'true'
-        env:
-          PIP_EXTRA_INDEX_URL: ${{ matrix.extra-index-url }}
-        run: >
-          pip3 install
-          --editable=".[test]"
-
-      - name: run pytest
-        if: steps.changed-files.outputs.python_any_changed == 'true'
-        id: run-pytest
-        run: pytest
-
-      # - name: run invokeai-configure
-      #   env:
-      #     HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGINGFACE_TOKEN }}
-      #   run: >
-      #     invokeai-configure
-      #     --yes
-      #     --default_only
-      #     --full-precision
-      #   # can't use fp16 weights without a GPU
-
-      # - name: run invokeai
-      #   id: run-invokeai
-      #   env:
-      #     # Set offline mode to make sure configure preloaded successfully.
-      #     HF_HUB_OFFLINE: 1
-      #     HF_DATASETS_OFFLINE: 1
-      #     TRANSFORMERS_OFFLINE: 1
-      #     INVOKEAI_OUTDIR: ${{ github.workspace }}/results
-      #   run: >
-      #     invokeai
-      #     --no-patchmatch
-      #     --no-nsfw_checker
-      #     --precision=float32
-      #     --always_use_cpu
-      #     --use_memory_db
-      #     --outdir ${{ env.INVOKEAI_OUTDIR }}/${{ matrix.python-version }}/${{ matrix.pytorch }}
-      #     --from_file ${{ env.TEST_PROMPTS }}
-
-      # - name: Archive results
-      #   env:
-      #     INVOKEAI_OUTDIR: ${{ github.workspace }}/results
-      #   uses: actions/upload-artifact@v3
-      #   with:
-      #     name: results
-      #     path: ${{ env.INVOKEAI_OUTDIR }}

.prettierrc.yaml

@@ -7,7 +7,7 @@ embeddedLanguageFormatting: auto
 overrides:
   - files: '*.md'
     options:
-      proseWrap: always
+      proseWrap: preserve
       printWidth: 80
       parser: markdown
       cursorOffset: -1

docs/RELEASE.md

@@ -0,0 +1,142 @@
+# Release Process
+
+The app is published in twice, in different build formats.
+
+- A [PyPI] distribution. This includes both a source distribution and built distribution (a wheel). Users install with `pip install invokeai`. The updater uses this build.
+- An installer on the [InvokeAI Releases Page]. This is a zip file with install scripts and a wheel. This is only used for new installs.
+
+## General Prep
+
+Make a developer call-out for PRs to merge. Merge and test things out.
+
+While the release workflow does not include end-to-end tests, it does pause before publishing so you can download and test the final build.
+
+## Release Workflow
+
+The `release.yml` workflow runs a number of jobs to handle code checks, tests, build and publish on PyPI.
+
+It is triggered on **tag push**, when the tag matches `v*`. It doesn't matter if you've prepped a release branch like `release/v3.5.0` or are releasing from `main` - it works the same.
+
+> Because commits are reference-counted, it is safe to create a release branch, tag it, let the workflow run, then delete the branch. So long as the tag exists, that commit will exist.
+
+### Triggering the Workflow
+
+Run `make tag-release` to tag the current commit and kick off the workflow.
+
+The release may also be dispatched [manually].
+
+### Workflow Jobs and Process
+
+The workflow consists of a number of concurrently-run jobs, and two final publish jobs.
+
+The publish jobs require manual approval and are only run if the other jobs succeed.
+
+#### `check-version` Job
+
+This job checks that the git ref matches the app version. It matches the ref against the `__version__` variable in `invokeai/version/invokeai_version.py`.
+
+When the workflow is triggered by tag push, the ref is the tag. If the workflow is run manually, the ref is the target selected from the **Use workflow from** dropdown.
+
+This job uses [samuelcolvin/check-python-version].
+
+> Any valid [version specifier] works, so long as the tag matches the version. The release workflow works exactly the same for `RC`, `post`, `dev`, etc.
+
+#### Check and Test Jobs
+
+- **`python-tests`**: runs `pytest` on matrix of platforms
+- **`python-checks`**: runs `ruff` (format and lint)
+- **`frontend-tests`**: runs `vitest`
+- **`frontend-checks`**: runs `prettier` (format), `eslint` (lint), `dpdm` (circular refs), `tsc` (static type check) and `knip` (unused imports)
+
+> **TODO** We should add `mypy` or `pyright` to the **`check-python`** job.
+
+> **TODO** We should add an end-to-end test job that generates an image.
+
+#### `build-installer` Job
+
+This sets up both python and frontend dependencies and builds the python package. Internally, this runs `installer/create_installer.sh` and uploads two artifacts:
+
+- **`dist`**: the python distribution, to be published on PyPI
+- **`InvokeAI-installer-${VERSION}.zip`**: the installer to be included in the GitHub release
+
+#### Sanity Check & Smoke Test
+
+At this point, the release workflow pauses as the remaining publish jobs require approval.
+
+A maintainer should go to the **Summary** tab of the workflow, download the installer and test it. Ensure the app loads and generates.
+
+> The same wheel file is bundled in the installer and in the `dist` artifact, which is uploaded to PyPI. You should end up with the exactly the same installation of the `invokeai` package from any of these methods.
+
+#### PyPI Publish Jobs
+
+The publish jobs will run if any of the previous jobs fail.
+
+They use [GitHub environments], which are configured as [trusted publishers] on PyPI.
+
+Both jobs require a maintainer to approve them from the workflow's **Summary** tab.
+
+- Click the **Review deployments** button
+- Select the environment (either `testpypi` or `pypi`)
+- Click **Approve and deploy**
+
+> **If the version already exists on PyPI, the publish jobs will fail.** PyPI only allows a given version to be published once - you cannot change it. If version published on PyPI has a problem, you'll need to "fail forward" by bumping the app version and publishing a followup release.
+
+#### `publish-testpypi` Job
+
+Publishes the distribution on the [Test PyPI] index, using the `testpypi` GitHub environment.
+
+This job is not required for the production PyPI publish, but included just in case you want to test the PyPI release.
+
+If approved and successful, you could try out the test release like this:
+
+```sh
+# Create a new virtual environment
+python -m venv ~/.test-invokeai-dist --prompt test-invokeai-dist
+# Install the distribution from Test PyPI
+pip install --index-url https://test.pypi.org/simple/ invokeai
+# Run and test the app
+invokeai-web
+# Cleanup
+deactivate
+rm -rf ~/.test-invokeai-dist
+```
+
+#### `publish-pypi` Job
+
+Publishes the distribution on the production PyPI index, using the `pypi` GitHub environment.
+
+## Publish the GitHub Release with installer
+
+Once the release is published to PyPI, it's time to publish the GitHub release.
+
+1. [Draft a new release] on GitHub, choosing the tag that triggered the release.
+2. Write the release notes, describing important changes. The **Generate release notes** button automatically inserts the changelog and new contributors, and you can copy/paste the intro from previous releases.
+3. Upload the zip file created in **`build`** job into the Assets section of the release notes. You can also upload the zip into the body of the release notes, since it can be hard for users to find the Assets section.
+4. Check the **Set as a pre-release** and **Create a discussion for this release** checkboxes at the bottom of the release page.
+5. Publish the pre-release.
+6. Announce the pre-release in Discord.
+
+> **TODO** Workflows can create a GitHub release from a template and upload release assets. One popular action to handle this is [ncipollo/release-action]. A future enhancement to the release process could set this up.
+
+## Manual Build
+
+The `build installer` workflow can be dispatched manually. This is useful to test the installer for a given branch or tag.
+
+No checks are run, it just builds.
+
+## Manual Release
+
+The `release` workflow can be dispatched manually. You must dispatch the workflow from the right tag, else it will fail the version check.
+
+This functionality is available as a fallback in case something goes wonky. Typically, releases should be triggered via tag push as described above.
+
+[InvokeAI Releases Page]: https://github.com/invoke-ai/InvokeAI/releases
+[PyPI]: https://pypi.org/
+[Draft a new release]: https://github.com/invoke-ai/InvokeAI/releases/new
+[Test PyPI]: https://test.pypi.org/
+[version specifier]: https://packaging.python.org/en/latest/specifications/version-specifiers/
+[ncipollo/release-action]: https://github.com/ncipollo/release-action
+[GitHub environments]: https://docs.github.com/en/actions/deployment/targeting-different-environments/using-environments-for-deployment
+[trusted publishers]: https://docs.pypi.org/trusted-publishers/
+[samuelcolvin/check-python-version]: https://github.com/samuelcolvin/check-python-version
+[manually]: #manual-release

docs/contributing/MODEL_MANAGER.md

@@ -32,7 +32,6 @@ model. These are the:
   Responsible for loading a model from disk
   into RAM and VRAM and getting it ready for inference.
 
-  
 ## Location of the Code
 
 The four main services can be found in
@@ -63,23 +62,21 @@ provides the following fields:
 |----------------|-----------------|------------------|
 | `key`            | str           | Unique identifier for the model |
 | `name`           | str           | Name of the model (not unique) |
-| `model_type`     | ModelType     | The type of the model | 
-| `model_format`   | ModelFormat   | The format of the model (e.g. "diffusers"); also used as a Union discriminator | 
-| `base_model`     | BaseModelType | The base model that the model is compatible with | 
+| `model_type`     | ModelType     | The type of the model |
+| `model_format`   | ModelFormat   | The format of the model (e.g. "diffusers"); also used as a Union discriminator |
+| `base_model`     | BaseModelType | The base model that the model is compatible with |
 | `path`           | str           | Location of model on disk |
-| `original_hash`  | str           | Hash of the model when it was first installed |
-| `current_hash`   | str           | Most recent hash of the model's contents |
+| `hash`           | str           | Hash of the model |
 | `description`    | str           | Human-readable description of the model (optional) |
 | `source`         | str           | Model's source URL or repo id (optional) |
 
 The `key` is a unique 32-character random ID which was generated at
-install time. The `original_hash` field stores a hash of the model's
+install time. The `hash` field stores a hash of the model's
 contents at install time obtained by sampling several parts of the
 model's files using the `imohash` library. Over the course of the
 model's lifetime it may be transformed in various ways, such as
 changing its precision or converting it from a .safetensors to a
-diffusers model. When this happens, `original_hash` is unchanged, but
-`current_hash` is updated to indicate the current contents.
+diffusers model.
 
 `ModelType`, `ModelFormat` and `BaseModelType` are string enums that
 are defined in `invokeai.backend.model_manager.config`. They are also
@@ -94,7 +91,6 @@ The `path` field can be absolute or relative. If relative, it is taken
 to be relative to the `models_dir` setting in the user's
 `invokeai.yaml` file.
 
-
 ### CheckpointConfig
 
 This adds support for checkpoint configurations, and adds the
@@ -174,7 +170,7 @@ store = context.services.model_manager.store
 or from elsewhere in the code by accessing
 `ApiDependencies.invoker.services.model_manager.store`.
 
-### Creating a `ModelRecordService` 
+### Creating a `ModelRecordService`
 
 To create a new `ModelRecordService` database or open an existing one,
 you can directly create either a `ModelRecordServiceSQL` or a
@@ -217,27 +213,27 @@ for use in the InvokeAI web server. Its signature is:
 ```
 def open(
        cls, 
-	   config: InvokeAIAppConfig, 
-	   conn: Optional[sqlite3.Connection] = None, 
-	   lock: Optional[threading.Lock] = None
+    config: InvokeAIAppConfig, 
+    conn: Optional[sqlite3.Connection] = None, 
+    lock: Optional[threading.Lock] = None
     ) -> Union[ModelRecordServiceSQL, ModelRecordServiceFile]:
 ```
 
 The way it works is as follows:
 
 1. Retrieve the value of the `model_config_db` option from the user's
-	`invokeai.yaml` config file.
+ `invokeai.yaml` config file.
 2. If `model_config_db` is `auto` (the default), then:
-   - Use the values of `conn` and `lock` to return a `ModelRecordServiceSQL` object
-	 opened on the passed connection and lock.
-   - Open up a new connection to `databases/invokeai.db` if `conn`
+   * Use the values of `conn` and `lock` to return a `ModelRecordServiceSQL` object
+  opened on the passed connection and lock.
+   * Open up a new connection to `databases/invokeai.db` if `conn`
      and/or `lock` are missing (see note below).
 3. If `model_config_db` is a Path, then use `from_db_file`
    to return the appropriate type of ModelRecordService.
 4. If `model_config_db` is None, then retrieve the legacy
    `conf_path` option from `invokeai.yaml` and use the Path
    indicated there. This will default to `configs/models.yaml`.
-   
+
 So a typical startup pattern would be:
 
 ```
@@ -255,7 +251,7 @@ store = ModelRecordServiceBase.open(config, db_conn, lock)
 
 Configurations can be retrieved in several ways.
 
-#### get_model(key) -> AnyModelConfig:
+#### get_model(key) -> AnyModelConfig
 
 The basic functionality is to call the record store object's
 `get_model()` method with the desired model's unique key. It returns
@@ -272,28 +268,28 @@ print(model_conf.path)
 If the key is unrecognized, this call raises an
 `UnknownModelException`.
 
-#### exists(key) -> AnyModelConfig:
+#### exists(key) -> AnyModelConfig
 
 Returns True if a model with the given key exists in the databsae.
 
-#### search_by_path(path) -> AnyModelConfig:
+#### search_by_path(path) -> AnyModelConfig
 
 Returns the configuration of the model whose path is `path`. The path
 is matched using a simple string comparison and won't correctly match
 models referred to by different paths (e.g. using symbolic links).
 
-#### search_by_name(name, base, type) -> List[AnyModelConfig]:
+#### search_by_name(name, base, type) -> List[AnyModelConfig]
 
 This method searches for models that match some combination of `name`,
 `BaseType` and `ModelType`. Calling without any arguments will return
 all the models in the database.
 
-#### all_models() -> List[AnyModelConfig]:
+#### all_models() -> List[AnyModelConfig]
 
 Return all the model configs in the database. Exactly equivalent to
 calling `search_by_name()` with no arguments.
 
-#### search_by_tag(tags) -> List[AnyModelConfig]:
+#### search_by_tag(tags) -> List[AnyModelConfig]
 
 `tags` is a list of strings. This method returns a list of model
 configs that contain all of the given tags. Examples:
@@ -312,11 +308,11 @@ commercializable_models = [x for x in store.all_models() \
                            if x.license.contains('allowCommercialUse=Sell')]
 ```
 
-#### version() -> str:
+#### version() -> str
 
 Returns the version of the database, currently at `3.2`
 
-#### model_info_by_name(name, base_model, model_type) -> ModelConfigBase:
+#### model_info_by_name(name, base_model, model_type) -> ModelConfigBase
 
 This method exists to ease the transition from the previous version of
 the model manager, in which `get_model()` took the three arguments
@@ -337,7 +333,7 @@ model and pass its key to `get_model()`.
 Several methods allow you to create and update stored model config
 records.
 
-#### add_model(key, config) -> AnyModelConfig:
+#### add_model(key, config) -> AnyModelConfig
 
 Given a key and a configuration, this will add the model's
 configuration record to the database. `config` can either be a subclass of
@@ -352,7 +348,7 @@ model with the same key is already in the database, or an
 `InvalidModelConfigException` if a dict was passed and Pydantic
 experienced a parse or validation error.
 
-### update_model(key, config) -> AnyModelConfig:
+### update_model(key, config) -> AnyModelConfig
 
 Given a key and a configuration, this will update the model
 configuration record in the database. `config` can be either a
@@ -370,31 +366,31 @@ The `ModelInstallService` class implements the
 shop for all your model install needs. It provides the following
 functionality:
 
-- Registering a model config record for a model already located on the
+* Registering a model config record for a model already located on the
   local filesystem, without moving it or changing its path.
   
-- Installing a model alreadiy located on the local filesystem, by
+* Installing a model alreadiy located on the local filesystem, by
   moving it into the InvokeAI root directory under the
   `models` folder (or wherever config parameter `models_dir`
   specifies).
-	
-- Probing of models to determine their type, base type and other key
+
+* Probing of models to determine their type, base type and other key
   information.
   
-- Interface with the InvokeAI event bus to provide status updates on
+* Interface with the InvokeAI event bus to provide status updates on
   the download, installation and registration process.
   
-- Downloading a model from an arbitrary URL and installing it in
+* Downloading a model from an arbitrary URL and installing it in
   `models_dir`.
   
-- Special handling for Civitai model URLs which allow the user to
+* Special handling for Civitai model URLs which allow the user to
   paste in a model page's URL or download link
   
-- Special handling for HuggingFace repo_ids to recursively download
+* Special handling for HuggingFace repo_ids to recursively download
   the contents of the repository, paying attention to alternative
   variants such as fp16.
   
-- Saving tags and other metadata about the model into the invokeai database
+* Saving tags and other metadata about the model into the invokeai database
   when fetching from a repo that provides that type of information,
   (currently only Civitai and HuggingFace).
   
@@ -427,8 +423,8 @@ queue.start()
 
 installer = ModelInstallService(app_config=config, 
                                 record_store=record_store,
-						        download_queue=queue
-							    )
+              download_queue=queue
+           )
 installer.start()
 ```
 
@@ -443,7 +439,6 @@ required parameters:
 | `metadata_store`   | Optional[ModelMetadataStore]  | Metadata storage object |
 |`session`           | Optional[requests.Session]    | Swap in a different Session object (usually for debugging) |
 
-
 Once initialized, the installer will provide the following methods:
 
 #### install_job = installer.heuristic_import(source, [config], [access_token])
@@ -457,15 +452,15 @@ The `source` is a string that can be any of these forms
 1. A path on the local filesystem (`C:\\users\\fred\\model.safetensors`)
 2. A Url pointing to a single downloadable model file (`https://civitai.com/models/58390/detail-tweaker-lora-lora`)
 3. A HuggingFace repo_id with any of the following formats:
-   - `model/name` -- entire model
-   - `model/name:fp32` -- entire model, using the fp32 variant
-   - `model/name:fp16:vae` -- vae submodel, using the fp16 variant
-   - `model/name::vae` -- vae submodel, using default precision
-   - `model/name:fp16:path/to/model.safetensors` -- an individual model file, fp16 variant
-   - `model/name::path/to/model.safetensors` -- an individual model file, default variant
+   * `model/name` -- entire model
+   * `model/name:fp32` -- entire model, using the fp32 variant
+   * `model/name:fp16:vae` -- vae submodel, using the fp16 variant
+   * `model/name::vae` -- vae submodel, using default precision
+   * `model/name:fp16:path/to/model.safetensors` -- an individual model file, fp16 variant
+   * `model/name::path/to/model.safetensors` -- an individual model file, default variant
 
 Note that by specifying a relative path to the top of the HuggingFace
-repo, you can download and install arbitrary models files. 
+repo, you can download and install arbitrary models files.
 
 The variant, if not provided, will be automatically filled in with
 `fp32` if the user has requested full precision, and `fp16`
@@ -491,9 +486,9 @@ following illustrates basic usage:
 
 ```
 from invokeai.app.services.model_install import (
-	LocalModelSource,
-	HFModelSource,
-	URLModelSource,
+ LocalModelSource,
+ HFModelSource,
+ URLModelSource,
 )
 
 source1 = LocalModelSource(path='/opt/models/sushi.safetensors')   # a local safetensors file
@@ -513,13 +508,13 @@ for source in [source1, source2, source3, source4, source5, source6, source7]:
 source2job = installer.wait_for_installs(timeout=120)
 for source in sources:
     job = source2job[source]
-	if job.complete:
-		model_config = job.config_out
-		model_key = model_config.key
-		print(f"{source} installed as {model_key}")
-	elif job.errored:
-	    print(f"{source}: {job.error_type}.\nStack trace:\n{job.error}")
-	
+ if job.complete:
+  model_config = job.config_out
+  model_key = model_config.key
+  print(f"{source} installed as {model_key}")
+ elif job.errored:
+     print(f"{source}: {job.error_type}.\nStack trace:\n{job.error}")
+ 
 ```
 
 As shown here, the `import_model()` method accepts a variety of
@@ -528,7 +523,7 @@ HuggingFace repo_ids with and without a subfolder designation,
 Civitai model URLs and arbitrary URLs that point to checkpoint files
 (but not to folders).
 
-Each call to `import_model()` return a `ModelInstallJob` job, 
+Each call to `import_model()` return a `ModelInstallJob` job,
 an object which tracks the progress of the install.
 
 If a remote model is requested, the model's files are downloaded in
@@ -555,7 +550,7 @@ The full list of arguments to `import_model()` is as follows:
 | `config`         | Dict[str, Any]               | None        | Override all or a portion of model's probed attributes |
 
 The next few sections describe the various types of ModelSource that
-can be passed to `import_model()`. 
+can be passed to `import_model()`.
 
 `config` can be used to override all or a portion of the configuration
 attributes returned by the model prober. See the section below for
@@ -566,7 +561,6 @@ details.
 This is used for a model that is located on a locally-accessible Posix
 filesystem, such as a local disk or networked fileshare.
 
-
 | **Argument**     | **Type**                     | **Default** | **Description**                           |
 |------------------|------------------------------|-------------|-------------------------------------------|
 | `path`           | str | Path                   | None        | Path to the model file or directory |
@@ -625,7 +619,6 @@ HuggingFace has the most complicated `ModelSource` structure:
 | `subfolder`      | Path                         | None        | Look for the model in a subfolder of the repo. |
 | `access_token`   | str                          | None        | An access token needed to gain access to a subscriber's-only model. |
 
-
 The `repo_id` is the repository ID, such as `stabilityai/sdxl-turbo`.
 
 The `variant` is one of the various diffusers formats that HuggingFace
@@ -661,7 +654,6 @@ in. To download these files, you must provide an
 `HfFolder.get_token()` will be called to fill it in with the cached
 one.
 
-
 #### Monitoring the install job process
 
 When you create an install job with `import_model()`, it launches the
@@ -675,14 +667,13 @@ The `ModelInstallJob` class has the following structure:
 | `id`           | `int`           | Integer ID for this job |
 | `status`       | `InstallStatus`  | An enum of [`waiting`, `downloading`, `running`, `completed`, `error` and `cancelled`]|
 | `config_in`    | `dict`          | Overriding configuration values provided by the caller |
-| `config_out`   | `AnyModelConfig`| After successful completion, contains the configuration record written to the database | 
-| `inplace`      | `boolean`       | True if the caller asked to install the model in place using its local path | 
-| `source`       | `ModelSource`   | The local path, remote URL or repo_id of the model to be installed | 
+| `config_out`   | `AnyModelConfig`| After successful completion, contains the configuration record written to the database |
+| `inplace`      | `boolean`       | True if the caller asked to install the model in place using its local path |
+| `source`       | `ModelSource`   | The local path, remote URL or repo_id of the model to be installed |
 | `local_path`   | `Path`          | If a remote model, holds the path of the model after it is downloaded; if a local model, same as `source` |
 | `error_type`   | `str`           | Name of the exception that led to an error status |
 | `error`        | `str`           | Traceback of the error |
 
-
 If the `event_bus` argument was provided, events will also be
 broadcast to the InvokeAI event bus. The events will appear on the bus
 as an event of type `EventServiceBase.model_event`, a timestamp and
@@ -702,14 +693,13 @@ following keys:
 | `total_bytes`  | int       | Total size of all the files that make up the model |
 | `parts`        | List[Dict]| Information on the progress of the individual files that make up the model |
 
-
 The parts is a list of dictionaries that give information on each of
 the components pieces of the download. The dictionary's keys are
 `source`, `local_path`, `bytes` and `total_bytes`, and correspond to
 the like-named keys in the main event.
 
 Note that downloading events will not be issued for local models, and
-that downloading events occur *before* the running event.
+that downloading events occur _before_ the running event.
 
 ##### `model_install_running`
 
@@ -752,14 +742,13 @@ properties: `waiting`, `downloading`, `running`, `complete`, `errored`
 and `cancelled`, as well as `in_terminal_state`. The last will return
 True if the job is in the complete, errored or cancelled states.
 
-
 #### Model configuration and probing
 
 The install service uses the `invokeai.backend.model_manager.probe`
 module during import to determine the model's type, base type, and
 other configuration parameters. Among other things, it assigns a
 default name and description for the model based on probed
-fields. 
+fields.
 
 When downloading remote models is implemented, additional
 configuration information, such as list of trigger terms, will be
@@ -774,11 +763,11 @@ attributes. Here is an example of setting the
 ```
 install_job = installer.import_model(
                source=HFModelSource(repo_id='stabilityai/stable-diffusion-2-1',variant='fp32'),
-			   config=dict(
-			         prediction_type=SchedulerPredictionType('v_prediction')
-					 name='stable diffusion 2 base model',
-	           )
-	      )
+      config=dict(
+            prediction_type=SchedulerPredictionType('v_prediction')
+      name='stable diffusion 2 base model',
+            )
+       )
 ```
 
 ### Other installer methods
@@ -862,7 +851,6 @@ This method is similar to `unregister()`, but also unconditionally
 deletes the corresponding model weights file(s), regardless of whether
 they are inside or outside the InvokeAI models hierarchy.
 
-
 #### path = installer.download_and_cache(remote_source, [access_token], [timeout])
 
 This utility routine will download the model file located at source,
@@ -953,7 +941,7 @@ following fields:
 
 When you create a job, you can assign it a `priority`. If multiple
 jobs are queued, the job with the lowest priority runs first. (Don't
-blame me! The Unix developers came up with this convention.) 
+blame me! The Unix developers came up with this convention.)
 
 Every job has a `source` and a `destination`. `source` is a string in
 the base class, but subclassses redefine it more specifically.
@@ -974,7 +962,7 @@ is in its lifecycle. Values are defined in the string enum
 `DownloadJobStatus`, a symbol available from
 `invokeai.app.services.download_manager`. Possible values are:
 
-| **Value**    |   **String Value**  | ** Description ** |
+| **Value**    |   **String Value**  | **Description** |
 |--------------|---------------------|-------------------|
 | `IDLE`         | idle              | Job created, but not submitted to the queue |
 | `ENQUEUED`     | enqueued          | Job is patiently waiting on the queue       |
@@ -991,7 +979,7 @@ debugging and performance testing.
 
 In case of an error, the Exception that caused the error will be
 placed in the `error` field, and the job's status will be set to
-`DownloadJobStatus.ERROR`. 
+`DownloadJobStatus.ERROR`.
 
 After an error occurs, any partially downloaded files will be deleted
 from disk, unless `preserve_partial_downloads` was set to True at job
@@ -1040,11 +1028,11 @@ While a job is being downloaded, the queue will emit events at
 periodic intervals. A typical series of events during a successful
 download session will look like this:
 
-- enqueued
-- running
-- running
-- running
-- completed
+* enqueued
+* running
+* running
+* running
+* completed
 
 There will be a single enqueued event, followed by one or more running
 events, and finally one `completed`, `error` or `cancelled`
@@ -1053,12 +1041,12 @@ events.
 It is possible for a caller to pause download temporarily, in which
 case the events may look something like this:
 
-- enqueued
-- running
-- running
-- paused
-- running
-- completed
+* enqueued
+* running
+* running
+* paused
+* running
+* completed
 
 The download queue logs when downloads start and end (unless `quiet`
 is set to True at initialization time) but doesn't log any progress
@@ -1120,11 +1108,11 @@ A typical initialization sequence will look like:
 from invokeai.app.services.download_manager import DownloadQueueService
 
 def log_download_event(job: DownloadJobBase):
-	logger.info(f'job={job.id}: status={job.status}')
+ logger.info(f'job={job.id}: status={job.status}')
 
 queue = DownloadQueueService(
-					  event_handlers=[log_download_event]
-						  )
+       event_handlers=[log_download_event]
+        )
 ```
 
 Event handlers can be provided to the queue at initialization time as
@@ -1155,9 +1143,9 @@ To use the former method, follow this example:
 ```
 job = DownloadJobRemoteSource(
          source='http://www.civitai.com/models/13456',
-		 destination='/tmp/models/',
-		 event_handlers=[my_handler1, my_handler2], # if desired
-		 )
+   destination='/tmp/models/',
+   event_handlers=[my_handler1, my_handler2], # if desired
+   )
 queue.submit_download_job(job, start=True)
 ```
 
@@ -1172,13 +1160,13 @@ To have the queue create the job for you, follow this example instead:
 ```
 job = queue.create_download_job(
          source='http://www.civitai.com/models/13456',
-		 destdir='/tmp/models/',
-		 filename='my_model.safetensors',
-		 event_handlers=[my_handler1, my_handler2], # if desired
-		 start=True,
-	)
+   destdir='/tmp/models/',
+   filename='my_model.safetensors',
+   event_handlers=[my_handler1, my_handler2], # if desired
+   start=True,
+ )
 ```
- 
+
 The `filename` argument forces the downloader to use the specified
 name for the file rather than the name provided by the remote source,
 and is equivalent to manually specifying a destination of
@@ -1187,7 +1175,6 @@ and is equivalent to manually specifying a destination of
 Here is the full list of arguments that can be provided to
 `create_download_job()`:
 
-
 | **Argument**     | **Type**                     | **Default** | **Description**                           |
 |------------------|------------------------------|-------------|-------------------------------------------|
 | `source`         | Union[str, Path, AnyHttpUrl] |             | Download remote or local source           |
@@ -1200,7 +1187,7 @@ Here is the full list of arguments that can be provided to
 
 Internally, `create_download_job()` has a little bit of internal logic
 that looks at the type of the source and selects the right subclass of
-`DownloadJobBase` to create and enqueue. 
+`DownloadJobBase` to create and enqueue.
 
 **TODO**: move this logic into its own method for overriding in
 subclasses.
@@ -1275,7 +1262,7 @@ for getting the model to run. For example "author" is metadata, while
 "type", "base" and "format" are not. The latter fields are part of the
 model's config, as defined in `invokeai.backend.model_manager.config`.
 
-### Example Usage:
+### Example Usage
 
 ```
 from invokeai.backend.model_manager.metadata import (
@@ -1328,7 +1315,6 @@ This is the common base class for metadata:
 | `author`      | str           | Model's author |
 | `tags`        | Set[str]      | Model tags |
 
-
 Note that the model config record also has a `name` field. It is
 intended that the config record version be locally customizable, while
 the metadata version is read-only. However, enforcing this is expected
@@ -1348,7 +1334,6 @@ This descends from `ModelMetadataBase` and adds the following fields:
 | `last_modified`| datetime         | Date of last commit of this model to the repo |
 | `files`        | List[Path]       | List of the files in the model repo |
 
-
 #### `CivitaiMetadata`
 
 This descends from `ModelMetadataBase` and adds the following fields:
@@ -1415,7 +1400,6 @@ testing suite to avoid hitting the internet.
 The HuggingFace and Civitai fetcher subclasses add additional
 repo-specific fetching methods:
 
-
 #### HuggingFaceMetadataFetch
 
 This overrides its base class `from_json()` method to return a
@@ -1434,13 +1418,12 @@ retrieves its metadata. Functionally equivalent to `from_id()`, the
 only difference is that it returna a `CivitaiMetadata` object rather
 than an `AnyModelRepoMetadata`.
 
-
 ### Metadata Storage
 
 The `ModelMetadataStore` provides a simple facility to store model
 metadata in the `invokeai.db` database. The data is stored as a JSON
 blob, with a few common fields (`name`, `author`, `tags`) broken out
-to be searchable. 
+to be searchable.
 
 When a metadata object is saved to the database, it is identified
 using the model key, _and this key must correspond to an existing
@@ -1535,16 +1518,16 @@ from invokeai.app.services.model_load import ModelLoadService, ModelLoaderRegist
 
 config = InvokeAIAppConfig.get_config()
 ram_cache = ModelCache(
-	max_cache_size=config.ram_cache_size, max_vram_cache_size=config.vram_cache_size, logger=logger
+ max_cache_size=config.ram_cache_size, max_vram_cache_size=config.vram_cache_size, logger=logger
 )
 convert_cache = ModelConvertCache(
-	cache_path=config.models_convert_cache_path, max_size=config.convert_cache_size
+ cache_path=config.models_convert_cache_path, max_size=config.convert_cache_size
 )
 loader = ModelLoadService(
-	app_config=config,
-	ram_cache=ram_cache,
-	convert_cache=convert_cache,
-	registry=ModelLoaderRegistry
+ app_config=config,
+ ram_cache=ram_cache,
+ convert_cache=convert_cache,
+ registry=ModelLoaderRegistry
 )
 ```
 
@@ -1567,7 +1550,6 @@ The returned `LoadedModel` object contains a copy of the configuration
 record returned by the model record `get_model()` method, as well as
 the in-memory loaded model:
 
-
 | **Attribute Name** | **Type**        |  **Description** |
 |----------------|-----------------|------------------|
 | `config`       | AnyModelConfig         | A copy of the model's configuration record for retrieving base type, etc. |
@@ -1581,7 +1563,6 @@ return `AnyModel`, a Union `ModelMixin`, `torch.nn.Module`,
 models, `EmbeddingModelRaw` is used for LoRA and TextualInversion
 models. The others are obvious.
 
-
 `LoadedModel` acts as a context manager. The context loads the model
 into the execution device (e.g. VRAM on CUDA systems), locks the model
 in the execution device for the duration of the context, and returns
@@ -1590,14 +1571,14 @@ the model. Use it like this:
 ```
 model_info = loader.get_model_by_key('f13dd932c0c35c22dcb8d6cda4203764', SubModelType('vae'))
 with model_info as vae:
-	image = vae.decode(latents)[0]
+ image = vae.decode(latents)[0]
 ```
 
 `get_model_by_key()` may raise any of the following exceptions:
 
-- `UnknownModelException`   -- key not in database
-- `ModelNotFoundException`  -- key in database but model not found at path
-- `NotImplementedException` -- the loader doesn't know how to load this type of model
+* `UnknownModelException`   -- key not in database
+* `ModelNotFoundException`  -- key in database but model not found at path
+* `NotImplementedException` -- the loader doesn't know how to load this type of model
   
 ### Emitting model loading events
 
@@ -1609,15 +1590,15 @@ following payload:
 
 ```
 payload=dict(
-	queue_id=queue_id,
-	queue_item_id=queue_item_id,
-	queue_batch_id=queue_batch_id,
-	graph_execution_state_id=graph_execution_state_id,
-	model_key=model_key,
-	submodel_type=submodel,
-	hash=model_info.hash,
-	location=str(model_info.location),
-	precision=str(model_info.precision),
+ queue_id=queue_id,
+ queue_item_id=queue_item_id,
+ queue_batch_id=queue_batch_id,
+ graph_execution_state_id=graph_execution_state_id,
+ model_key=model_key,
+ submodel_type=submodel,
+ hash=model_info.hash,
+ location=str(model_info.location),
+ precision=str(model_info.precision),
 )
 ```
 
@@ -1724,6 +1705,7 @@ object, or in `context.services.model_manager` from within an
 invocation.
 
 In the examples below, we have retrieved the manager using:
+
 ```
 mm = ApiDependencies.invoker.services.model_manager
 ```

docs/nodes/INVOCATION_API.md

@@ -0,0 +1,45 @@
+# Invocation API
+
+Each invocation's `invoke` method is provided a single arg - the Invocation
+Context.
+
+This object provides access to various methods, used to interact with the
+application. Loading and saving images, logging messages, etc.
+
+!!! warning ""
+
+    This API may shift slightly until the release of v4.0.0 as we work through a few final updates to the Model Manager.
+
+```py
+class MyInvocation(BaseInvocation):
+  ...
+  def invoke(self, context: InvocationContext) -> ImageOutput:
+      image_pil = context.images.get_pil(image_name)
+      # Do something to the image
+      image_dto = context.images.save(image_pil)
+      # Log a message
+      context.logger.info(f"Did something cool, image saved!")
+      ...
+```
+
+<!-- prettier-ignore-start -->
+::: invokeai.app.services.shared.invocation_context.InvocationContext
+    options:
+        members: false
+
+::: invokeai.app.services.shared.invocation_context.ImagesInterface
+
+::: invokeai.app.services.shared.invocation_context.TensorsInterface
+
+::: invokeai.app.services.shared.invocation_context.ConditioningInterface
+
+::: invokeai.app.services.shared.invocation_context.ModelsInterface
+
+::: invokeai.app.services.shared.invocation_context.LoggerInterface
+
+::: invokeai.app.services.shared.invocation_context.ConfigInterface
+
+::: invokeai.app.services.shared.invocation_context.UtilInterface
+
+::: invokeai.app.services.shared.invocation_context.BoardsInterface
+<!-- prettier-ignore-end -->

docs/nodes/NODES_MIGRATION_V3_V4.md

@@ -0,0 +1,148 @@
+# Invoke v4.0.0 Nodes API Migration guide
+
+Invoke v4.0.0 is versioned as such due to breaking changes to the API utilized
+by nodes, both core and custom.
+
+## Motivation
+
+Prior to v4.0.0, the `invokeai` python package has not be set up to be utilized
+as a library. That is to say, it didn't have any explicitly public API, and node
+authors had to work with the unstable internal application API.
+
+v4.0.0 introduces a stable public API for nodes.
+
+## Changes
+
+There are two node-author-facing changes:
+
+1. Import Paths
+1. Invocation Context API
+
+### Import Paths
+
+All public objects are now exported from `invokeai.invocation_api`:
+
+```py
+# Old
+from invokeai.app.invocations.baseinvocation import (
+    BaseInvocation,
+    InputField,
+    InvocationContext,
+    invocation,
+)
+from invokeai.app.invocations.primitives import ImageField
+
+# New
+from invokeai.invocation_api import (
+    BaseInvocation,
+    ImageField,
+    InputField,
+    InvocationContext,
+    invocation,
+)
+```
+
+It's possible that we've missed some classes you need in your node. Please let
+us know if that's the case.
+
+### Invocation Context API
+
+Most nodes utilize the Invocation Context, an object that is passed to the
+`invoke` that provides access to data and services a node may need.
+
+Until now, that object and the services it exposed were internal. Exposing them
+to nodes means that changes to our internal implementation could break nodes.
+The methods on the services are also often fairly complicated and allowed nodes
+to footgun.
+
+In v4.0.0, this object has been refactored to be much simpler.
+
+See [INVOCATION_API](./INVOCATION_API.md) for full details of the API.
+
+!!! warning ""
+
+    This API may shift slightly until the release of v4.0.0 as we work through a few final updates to the Model Manager.
+
+#### Improved Service Methods
+
+The biggest offender was the image save method:
+
+```py
+# Old
+image_dto = context.services.images.create(
+    image=image,
+    image_origin=ResourceOrigin.INTERNAL,
+    image_category=ImageCategory.GENERAL,
+    node_id=self.id,
+    session_id=context.graph_execution_state_id,
+    is_intermediate=self.is_intermediate,
+    metadata=self.metadata,
+    workflow=context.workflow,
+)
+
+# New
+image_dto = context.images.save(image=image)
+```
+
+Other methods are simplified, or enhanced with additional functionality:
+
+```py
+# Old
+image = context.services.images.get_pil_image(image_name)
+
+# New
+image = context.images.get_pil(image_name)
+image_cmyk = context.images.get_pil(image_name, "CMYK")
+```
+
+We also had some typing issues around tensors:
+
+```py
+# Old
+# `latents` typed as `torch.Tensor`, but could be `ConditioningFieldData`
+latents = context.services.latents.get(self.latents.latents_name)
+# `data` typed as `torch.Tenssor,` but could be `ConditioningFieldData`
+context.services.latents.save(latents_name, data)
+
+# New - separate methods for tensors and conditioning data w/ correct typing
+# Also, the service generates the names
+tensor_name = context.tensors.save(tensor)
+tensor = context.tensors.load(tensor_name)
+# For conditioning
+cond_name = context.conditioning.save(cond_data)
+cond_data = context.conditioning.load(cond_name)
+```
+
+#### Output Construction
+
+Core Outputs have builder functions right on them - no need to manually
+construct these objects, or use an extra utility:
+
+```py
+# Old
+image_output = ImageOutput(
+    image=ImageField(image_name=image_dto.image_name),
+    width=image_dto.width,
+    height=image_dto.height,
+)
+latents_output = build_latents_output(latents_name=name, latents=latents, seed=None)
+noise_output = NoiseOutput(
+    noise=LatentsField(latents_name=latents_name, seed=seed),
+    width=latents.size()[3] * 8,
+    height=latents.size()[2] * 8,
+)
+cond_output = ConditioningOutput(
+    conditioning=ConditioningField(
+        conditioning_name=conditioning_name,
+    ),
+)
+
+# New
+image_output = ImageOutput.build(image_dto)
+latents_output = LatentsOutput.build(latents_name=name, latents=noise, seed=self.seed)
+noise_output = NoiseOutput.build(latents_name=name, latents=noise, seed=self.seed)
+cond_output = ConditioningOutput.build(conditioning_name)
+```
+
+You can still create the objects using constructors if you want, but we suggest
+using the builder methods.

docs/nodes/communityNodes.md

@@ -32,6 +32,7 @@ To use a community workflow, download the the `.json` node graph file and load i
     + [Image to Character Art Image Nodes](#image-to-character-art-image-nodes)
     + [Image Picker](#image-picker)
     + [Image Resize Plus](#image-resize-plus)
+    + [Latent Upscale](#latent-upscale)
     + [Load Video Frame](#load-video-frame)
     + [Make 3D](#make-3d)
     + [Mask Operations](#mask-operations)
@@ -290,6 +291,13 @@ View:
 </br><img src="https://raw.githubusercontent.com/VeyDlin/image-resize-plus-node/master/.readme/node.png" width="500" />
 
 
+--------------------------------
+### Latent Upscale
+
+**Description:** This node uses a small (~2.4mb) model to upscale the latents used in a Stable Diffusion 1.5 or Stable Diffusion XL image generation, rather than the typical interpolation method, avoiding the traditional downsides of the latent upscale technique.
+
+**Node Link:** [https://github.com/gogurtenjoyer/latent-upscale](https://github.com/gogurtenjoyer/latent-upscale)
+
 --------------------------------
 ### Load Video Frame
 
@@ -346,12 +354,21 @@ See full docs here: https://github.com/skunkworxdark/Prompt-tools-nodes/edit/mai
 
 **Description:** A set of nodes for Metadata. Collect Metadata from within an `iterate` node & extract metadata from an image.
 
-- `Metadata Item Linked` - Allows collecting of metadata while within an iterate node with no need for a collect node or conversion to metadata node.
-- `Metadata From Image` - Provides Metadata from an image.
-- `Metadata To String` - Extracts a String value of a label from metadata.
-- `Metadata To Integer` - Extracts an Integer value of a label from metadata.
-- `Metadata To Float` - Extracts a Float value of a label from metadata.
-- `Metadata To Scheduler` - Extracts a Scheduler value of a label from metadata.
+- `Metadata Item Linked` - Allows collecting of metadata while within an iterate node with no need for a collect node or conversion to metadata node
+- `Metadata From Image` - Provides Metadata from an image
+- `Metadata To String` - Extracts a String value of a label from metadata
+- `Metadata To Integer` - Extracts an Integer value of a label from metadata
+- `Metadata To Float` - Extracts a Float value of a label from metadata
+- `Metadata To Scheduler` - Extracts a Scheduler value of a label from metadata
+- `Metadata To Bool` - Extracts Bool types from metadata
+- `Metadata To Model` - Extracts model types from metadata
+- `Metadata To SDXL Model` - Extracts SDXL model types from metadata
+- `Metadata To LoRAs` - Extracts Loras from metadata. 
+- `Metadata To SDXL LoRAs` - Extracts SDXL Loras from metadata
+- `Metadata To ControlNets` - Extracts ControNets from metadata
+- `Metadata To IP-Adapters` - Extracts IP-Adapters from metadata
+- `Metadata To T2I-Adapters` - Extracts T2I-Adapters from metadata
+- `Denoise Latents + Metadata` - This is an inherited version of the existing `Denoise Latents` node but with a metadata input and output. 
 
 **Node Link:** https://github.com/skunkworxdark/metadata-linked-nodes
 

docs/nodes/defaultNodes.md

@@ -19,6 +19,8 @@ their descriptions.
 | Conditioning Primitive                                        | A conditioning tensor primitive value                                                                                                                |
 | Content Shuffle Processor                                     | Applies content shuffle processing to image                                                                                                          |
 | ControlNet                                                    | Collects ControlNet info to pass to other nodes                                                                                                      |
+| Create Denoise Mask                                           | Converts a greyscale or transparency image into a mask for denoising.                                                                                |
+| Create Gradient Mask                                          | Creates a mask for Gradient ("soft", "differential") inpainting that gradually expands during denoising. Improves edge coherence.                    |
 | Denoise Latents                                               | Denoises noisy latents to decodable images                                                                                                           |
 | Divide Integers                                               | Divides two numbers                                                                                                                                  |
 | Dynamic Prompt                                                | Parses a prompt using adieyal/dynamicprompts' random or combinatorial generator                                                                      |

docs/requirements-mkdocs.txt

@@ -1,5 +0,0 @@
-mkdocs
-mkdocs-material>=8, <9
-mkdocs-git-revision-date-localized-plugin
-mkdocs-redirects==1.2.0
-

docs/stylesheets/extra.css

@@ -1,5 +0,0 @@
-:root {
-    --md-primary-fg-color:        #35A4DB;
-    --md-primary-fg-color--light: #35A4DB;
-    --md-primary-fg-color--dark:  #35A4DB;
-  }

installer/create_installer.sh

@@ -2,22 +2,18 @@
 
 set -e
 
-BCYAN="\e[1;36m"
-BYELLOW="\e[1;33m"
-BGREEN="\e[1;32m"
-BRED="\e[1;31m"
-RED="\e[31m"
-RESET="\e[0m"
-
-function is_bin_in_path {
-    builtin type -P "$1" &>/dev/null
-}
+BCYAN="\033[1;36m"
+BYELLOW="\033[1;33m"
+BGREEN="\033[1;32m"
+BRED="\033[1;31m"
+RED="\033[31m"
+RESET="\033[0m"
 
 function git_show {
     git show -s --format=oneline --abbrev-commit "$1" | cat
 }
 
-if [[ -v "VIRTUAL_ENV" ]]; then
+if [[ ! -z "${VIRTUAL_ENV}" ]]; then
     # we can't just call 'deactivate' because this function is not exported
     # to the environment of this script from the bash process that runs the script
     echo -e "${BRED}A virtual environment is activated. Please deactivate it before proceeding.${RESET}"
@@ -26,31 +22,63 @@ fi
 
 cd "$(dirname "$0")"
 
-echo
-echo -e "${BYELLOW}This script must be run from the installer directory!${RESET}"
-echo "The current working directory is $(pwd)"
-read -p "If that looks right, press any key to proceed, or CTRL-C to exit..."
-echo
-
-# Some machines only have `python3` in PATH, others have `python` - make an alias.
-# We can use a function to approximate an alias within a non-interactive shell.
-if ! is_bin_in_path python && is_bin_in_path python3; then
-    function python {
-        python3 "$@"
-    }
-fi
-
 VERSION=$(
     cd ..
-    python -c "from invokeai.version import __version__ as version; print(version)"
+    python3 -c "from invokeai.version import __version__ as version; print(version)"
 )
-PATCH=""
-VERSION="v${VERSION}${PATCH}"
+VERSION="v${VERSION}"
+
+if [[ ! -z ${CI} ]]; then
+    echo
+    echo -e "${BCYAN}CI environment detected${RESET}"
+    echo
+else
+    echo
+    echo -e "${BYELLOW}This script must be run from the installer directory!${RESET}"
+    echo "The current working directory is $(pwd)"
+    read -p "If that looks right, press any key to proceed, or CTRL-C to exit..."
+    echo
+fi
 
 echo -e "${BGREEN}HEAD${RESET}:"
 git_show HEAD
 echo
 
+# ---------------------- FRONTEND ----------------------
+
+pushd ../invokeai/frontend/web >/dev/null
+echo "Installing frontend dependencies..."
+echo
+pnpm i --frozen-lockfile
+echo
+if [[ ! -z ${CI} ]]; then
+    echo "Building frontend without checks..."
+    # In CI, we have already done the frontend checks and can just build
+    pnpm vite build
+else
+    echo "Running checks and building frontend..."
+    # This runs all the frontend checks and builds
+    pnpm build
+fi
+echo
+popd
+
+# ---------------------- BACKEND ----------------------
+
+echo
+echo "Building wheel..."
+echo
+
+# install the 'build' package in the user site packages, if needed
+# could be improved by using a temporary venv, but it's tiny and harmless
+if [[ $(python3 -c 'from importlib.util import find_spec; print(find_spec("build") is None)') == "True" ]]; then
+    pip install --user build
+fi
+
+rm -rf ../build
+
+python3 -m build --outdir dist/ ../.
+
 # ----------------------
 
 echo
@@ -78,10 +106,28 @@ chmod a+x InvokeAI-Installer/install.sh
 cp install.bat.in InvokeAI-Installer/install.bat
 cp WinLongPathsEnabled.reg InvokeAI-Installer/
 
-# Zip everything up
-zip -r InvokeAI-installer-$VERSION.zip InvokeAI-Installer
+FILENAME=InvokeAI-installer-$VERSION.zip
 
-# clean up
-rm -rf InvokeAI-Installer tmp dist ../invokeai/frontend/web/dist/
+# Zip everything up
+zip -r ${FILENAME} InvokeAI-Installer
+
+echo
+echo -e "${BGREEN}Built installer: ./${FILENAME}${RESET}"
+echo -e "${BGREEN}Built PyPi distribution: ./dist${RESET}"
+
+# clean up, but only if we are not in a github action
+if [[ -z ${CI} ]]; then
+    echo
+    echo "Cleaning up intermediate build files..."
+    rm -rf InvokeAI-Installer tmp ../invokeai/frontend/web/dist/
+fi
+
+if [[ ! -z ${CI} ]]; then
+    echo
+    echo "Setting GitHub action outputs..."
+    echo "INSTALLER_FILENAME=${FILENAME}" >>$GITHUB_OUTPUT
+    echo "INSTALLER_PATH=installer/${FILENAME}" >>$GITHUB_OUTPUT
+    echo "DIST_PATH=installer/dist/" >>$GITHUB_OUTPUT
+fi
 
 exit 0

installer/tag_release.sh

@@ -2,12 +2,12 @@
 
 set -e
 
-BCYAN="\e[1;36m"
-BYELLOW="\e[1;33m"
-BGREEN="\e[1;32m"
-BRED="\e[1;31m"
-RED="\e[31m"
-RESET="\e[0m"
+BCYAN="\033[1;36m"
+BYELLOW="\033[1;33m"
+BGREEN="\033[1;32m"
+BRED="\033[1;31m"
+RED="\033[31m"
+RESET="\033[0m"
 
 function does_tag_exist {
     git rev-parse --quiet --verify "refs/tags/$1" >/dev/null
@@ -23,49 +23,40 @@ function git_show {
 
 VERSION=$(
     cd ..
-    python -c "from invokeai.version import __version__ as version; print(version)"
+    python3 -c "from invokeai.version import __version__ as version; print(version)"
 )
 PATCH=""
-MAJOR_VERSION=$(echo $VERSION | sed 's/\..*$//')
 VERSION="v${VERSION}${PATCH}"
-LATEST_TAG="v${MAJOR_VERSION}-latest"
 
 if does_tag_exist $VERSION; then
     echo -e "${BCYAN}${VERSION}${RESET} already exists:"
     git_show_ref tags/$VERSION
     echo
 fi
-if does_tag_exist $LATEST_TAG; then
-    echo -e "${BCYAN}${LATEST_TAG}${RESET} already exists:"
-    git_show_ref tags/$LATEST_TAG
-    echo
-fi
 
 echo -e "${BGREEN}HEAD${RESET}:"
 git_show
 echo
 
-echo -e -n "Create tags ${BCYAN}${VERSION}${RESET} and ${BCYAN}${LATEST_TAG}${RESET} @ ${BGREEN}HEAD${RESET}, ${RED}deleting existing tags on remote${RESET}? "
+echo -e "${BGREEN}git remote -v${RESET}:"
+git remote -v
+echo
+
+echo -e -n "Create tags ${BCYAN}${VERSION}${RESET} @ ${BGREEN}HEAD${RESET}, ${RED}deleting existing tags on origin remote${RESET}? "
 read -e -p 'y/n [n]: ' input
 RESPONSE=${input:='n'}
 if [ "$RESPONSE" == 'y' ]; then
     echo
-    echo -e "Deleting ${BCYAN}${VERSION}${RESET} tag on remote..."
-    git push --delete origin $VERSION
+    echo -e "Deleting ${BCYAN}${VERSION}${RESET} tag on origin remote..."
+    git push origin :refs/tags/$VERSION
 
-    echo -e "Tagging ${BGREEN}HEAD${RESET} with ${BCYAN}${VERSION}${RESET} locally..."
+    echo -e "Tagging ${BGREEN}HEAD${RESET} with ${BCYAN}${VERSION}${RESET} on locally..."
     if ! git tag -fa $VERSION; then
         echo "Existing/invalid tag"
         exit -1
     fi
 
-    echo -e "Deleting ${BCYAN}${LATEST_TAG}${RESET} tag on remote..."
-    git push --delete origin $LATEST_TAG
-
-    echo -e "Tagging ${BGREEN}HEAD${RESET} with ${BCYAN}${LATEST_TAG}${RESET} locally..."
-    git tag -fa $LATEST_TAG
-
-    echo -e "Pushing updated tags to remote..."
+    echo -e "Pushing updated tags to origin remote..."
     git push origin --tags
 fi
 exit 0

invokeai/app/api/dependencies.py

@@ -26,7 +26,6 @@ from ..services.invocation_services import InvocationServices
 from ..services.invocation_stats.invocation_stats_default import InvocationStatsService
 from ..services.invoker import Invoker
 from ..services.model_manager.model_manager_default import ModelManagerService
-from ..services.model_metadata import ModelMetadataStoreSQL
 from ..services.model_records import ModelRecordServiceSQL
 from ..services.names.names_default import SimpleNameService
 from ..services.session_processor.session_processor_default import DefaultSessionProcessor
@@ -93,10 +92,9 @@ class ApiDependencies:
             ObjectSerializerDisk[ConditioningFieldData](output_folder / "conditioning", ephemeral=True)
         )
         download_queue_service = DownloadQueueService(event_bus=events)
-        model_metadata_service = ModelMetadataStoreSQL(db=db)
         model_manager = ModelManagerService.build_model_manager(
             app_config=configuration,
-            model_record_service=ModelRecordServiceSQL(db=db, metadata_store=model_metadata_service),
+            model_record_service=ModelRecordServiceSQL(db=db),
             download_queue=download_queue_service,
             events=events,
         )

invokeai/app/api/routers/model_manager.py

@@ -3,10 +3,7 @@
 
 import pathlib
 import shutil
-from hashlib import sha1
-from random import randbytes
-import traceback
-from typing import Any, Dict, List, Optional, Set
+from typing import Any, Dict, List, Optional
 
 from fastapi import Body, Path, Query, Response
 from fastapi.routing import APIRouter
@@ -15,15 +12,11 @@ from starlette.exceptions import HTTPException
 from typing_extensions import Annotated
 
 from invokeai.app.services.model_install import ModelInstallJob
-from invokeai.app.services.model_metadata.metadata_store_base import ModelMetadataChanges
 from invokeai.app.services.model_records import (
-    DuplicateModelException,
     InvalidModelException,
-    ModelRecordOrderBy,
-    ModelSummary,
     UnknownModelException,
 )
-from invokeai.app.services.shared.pagination import PaginatedResults
+from invokeai.app.services.model_records.model_records_base import DuplicateModelException, ModelRecordChanges
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     BaseModelType,
@@ -32,9 +25,6 @@ from invokeai.backend.model_manager.config import (
     ModelType,
     SubModelType,
 )
-from invokeai.backend.model_manager.merge import MergeInterpolationMethod, ModelMerger
-from invokeai.backend.model_manager.metadata import AnyModelRepoMetadata
-from invokeai.backend.model_manager.metadata.metadata_base import BaseMetadata
 from invokeai.backend.model_manager.search import ModelSearch
 
 from ..dependencies import ApiDependencies
@@ -50,15 +40,6 @@ class ModelsList(BaseModel):
     model_config = ConfigDict(use_enum_values=True)
 
 
-class ModelTagSet(BaseModel):
-    """Return tags for a set of models."""
-
-    key: str
-    name: str
-    author: str
-    tags: Set[str]
-
-
 ##############################################################################
 # These are example inputs and outputs that are used in places where Swagger
 # is unable to generate a correct example.
@@ -69,19 +50,16 @@ example_model_config = {
     "base": "sd-1",
     "type": "main",
     "format": "checkpoint",
-    "config": "string",
+    "config_path": "string",
     "key": "string",
-    "original_hash": "string",
-    "current_hash": "string",
+    "hash": "string",
     "description": "string",
     "source": "string",
-    "last_modified": 0,
-    "vae": "string",
+    "converted_at": 0,
     "variant": "normal",
     "prediction_type": "epsilon",
     "repo_variant": "fp16",
     "upcast_attention": False,
-    "ztsnr_training": False,
 }
 
 example_model_input = {
@@ -90,50 +68,12 @@ example_model_input = {
     "base": "sd-1",
     "type": "main",
     "format": "checkpoint",
-    "config": "configs/stable-diffusion/v1-inference.yaml",
+    "config_path": "configs/stable-diffusion/v1-inference.yaml",
     "description": "Model description",
     "vae": None,
     "variant": "normal",
 }
 
-example_model_metadata = {
-    "name": "ip_adapter_sd_image_encoder",
-    "author": "InvokeAI",
-    "tags": [
-        "transformers",
-        "safetensors",
-        "clip_vision_model",
-        "endpoints_compatible",
-        "region:us",
-        "has_space",
-        "license:apache-2.0",
-    ],
-    "files": [
-        {
-            "url": "https://huggingface.co/InvokeAI/ip_adapter_sd_image_encoder/resolve/main/README.md",
-            "path": "ip_adapter_sd_image_encoder/README.md",
-            "size": 628,
-            "sha256": None,
-        },
-        {
-            "url": "https://huggingface.co/InvokeAI/ip_adapter_sd_image_encoder/resolve/main/config.json",
-            "path": "ip_adapter_sd_image_encoder/config.json",
-            "size": 560,
-            "sha256": None,
-        },
-        {
-            "url": "https://huggingface.co/InvokeAI/ip_adapter_sd_image_encoder/resolve/main/model.safetensors",
-            "path": "ip_adapter_sd_image_encoder/model.safetensors",
-            "size": 2528373448,
-            "sha256": "6ca9667da1ca9e0b0f75e46bb030f7e011f44f86cbfb8d5a36590fcd7507b030",
-        },
-    ],
-    "type": "huggingface",
-    "id": "InvokeAI/ip_adapter_sd_image_encoder",
-    "tag_dict": {"license": "apache-2.0"},
-    "last_modified": "2023-09-23T17:33:25Z",
-}
-
 ##############################################################################
 # ROUTES
 ##############################################################################
@@ -213,89 +153,16 @@ async def get_model_record(
         raise HTTPException(status_code=404, detail=str(e))
 
 
-@model_manager_router.get("/summary", operation_id="list_model_summary")
-async def list_model_summary(
-    page: int = Query(default=0, description="The page to get"),
-    per_page: int = Query(default=10, description="The number of models per page"),
-    order_by: ModelRecordOrderBy = Query(default=ModelRecordOrderBy.Default, description="The attribute to order by"),
-) -> PaginatedResults[ModelSummary]:
-    """Gets a page of model summary data."""
-    record_store = ApiDependencies.invoker.services.model_manager.store
-    results: PaginatedResults[ModelSummary] = record_store.list_models(page=page, per_page=per_page, order_by=order_by)
-    return results
-
-
-@model_manager_router.get(
-    "/i/{key}/metadata",
-    operation_id="get_model_metadata",
-    responses={
-        200: {
-            "description": "The model metadata was retrieved successfully",
-            "content": {"application/json": {"example": example_model_metadata}},
-        },
-        400: {"description": "Bad request"},
-    },
-)
-async def get_model_metadata(
-    key: str = Path(description="Key of the model repo metadata to fetch."),
-) -> Optional[AnyModelRepoMetadata]:
-    """Get a model metadata object."""
-    record_store = ApiDependencies.invoker.services.model_manager.store
-    result: Optional[AnyModelRepoMetadata] = record_store.get_metadata(key)
-
-    return result
-
-@model_manager_router.patch(
-    "/i/{key}/metadata",
-    operation_id="update_model_metadata",
-    responses={
-        201: {
-            "description": "The model metadata was updated successfully",
-            "content": {"application/json": {"example": example_model_metadata}},
-        },
-        400: {"description": "Bad request"},
-    },
-)
-async def update_model_metadata(
-    key: str = Path(description="Key of the model repo metadata to fetch."),
-    changes: ModelMetadataChanges = Body(description="The changes")
-) -> Optional[AnyModelRepoMetadata]:
-    """Updates or creates a model metadata object."""
-    record_store = ApiDependencies.invoker.services.model_manager.store
-    metadata_store = ApiDependencies.invoker.services.model_manager.store.metadata_store
-    
-    try:
-        original_metadata = record_store.get_metadata(key)
-        if original_metadata:
-            if changes.trigger_phrases:
-                original_metadata.trigger_phrases = changes.trigger_phrases
-
-            if changes.default_settings:
-                original_metadata.default_settings = changes.default_settings
-
-            metadata_store.update_metadata(key, original_metadata)
-        else:
-            metadata_store.add_metadata(key, BaseMetadata(name="", author="",trigger_phrases=changes.trigger_phrases, default_settings=changes.default_settings))
-    except Exception as e:
-        raise HTTPException(
-            status_code=500,
-            detail=f"An error occurred while updating the model metadata: {e}",
-        )
-
-    result: Optional[AnyModelRepoMetadata] = record_store.get_metadata(key)
-
-    return result
-
-
-@model_manager_router.get(
-    "/tags",
-    operation_id="list_tags",
-)
-async def list_tags() -> Set[str]:
-    """Get a unique set of all the model tags."""
-    record_store = ApiDependencies.invoker.services.model_manager.store
-    result: Set[str] = record_store.list_tags()
-    return result
+# @model_manager_router.get("/summary", operation_id="list_model_summary")
+# async def list_model_summary(
+#     page: int = Query(default=0, description="The page to get"),
+#     per_page: int = Query(default=10, description="The number of models per page"),
+#     order_by: ModelRecordOrderBy = Query(default=ModelRecordOrderBy.Default, description="The attribute to order by"),
+# ) -> PaginatedResults[ModelSummary]:
+#     """Gets a page of model summary data."""
+#     record_store = ApiDependencies.invoker.services.model_manager.store
+#     results: PaginatedResults[ModelSummary] = record_store.list_models(page=page, per_page=per_page, order_by=order_by)
+#     return results
 
 
 class FoundModel(BaseModel):
@@ -367,19 +234,6 @@ async def scan_for_models(
     return scan_results
 
 
-@model_manager_router.get(
-    "/tags/search",
-    operation_id="search_by_metadata_tags",
-)
-async def search_by_metadata_tags(
-    tags: Set[str] = Query(default=None, description="Tags to search for"),
-) -> ModelsList:
-    """Get a list of models."""
-    record_store = ApiDependencies.invoker.services.model_manager.store
-    results = record_store.search_by_metadata_tag(tags)
-    return ModelsList(models=results)
-
-
 @model_manager_router.patch(
     "/i/{key}",
     operation_id="update_model_record",
@@ -396,15 +250,13 @@ async def search_by_metadata_tags(
 )
 async def update_model_record(
     key: Annotated[str, Path(description="Unique key of model")],
-    info: Annotated[
-        AnyModelConfig, Body(description="Model config", discriminator="type", example=example_model_input)
-    ],
+    changes: Annotated[ModelRecordChanges, Body(description="Model config", example=example_model_input)],
 ) -> AnyModelConfig:
-    """Update model contents with a new config. If the model name or base fields are changed, then the model is renamed."""
+    """Update a model's config."""
     logger = ApiDependencies.invoker.services.logger
     record_store = ApiDependencies.invoker.services.model_manager.store
     try:
-        model_response: AnyModelConfig = record_store.update_model(key, config=info)
+        model_response: AnyModelConfig = record_store.update_model(key, changes=changes)
         logger.info(f"Updated model: {key}")
     except UnknownModelException as e:
         raise HTTPException(status_code=404, detail=str(e))
@@ -416,14 +268,14 @@ async def update_model_record(
 
 @model_manager_router.delete(
     "/i/{key}",
-    operation_id="del_model_record",
+    operation_id="delete_model",
     responses={
         204: {"description": "Model deleted successfully"},
         404: {"description": "Model not found"},
     },
     status_code=204,
 )
-async def del_model_record(
+async def delete_model(
     key: str = Path(description="Unique key of model to remove from model registry."),
 ) -> Response:
     """
@@ -444,42 +296,39 @@ async def del_model_record(
         raise HTTPException(status_code=404, detail=str(e))
 
 
-@model_manager_router.post(
-    "/i/",
-    operation_id="add_model_record",
-    responses={
-        201: {
-            "description": "The model added successfully",
-            "content": {"application/json": {"example": example_model_config}},
-        },
-        409: {"description": "There is already a model corresponding to this path or repo_id"},
-        415: {"description": "Unrecognized file/folder format"},
-    },
-    status_code=201,
-)
-async def add_model_record(
-    config: Annotated[
-        AnyModelConfig, Body(description="Model config", discriminator="type", example=example_model_input)
-    ],
-) -> AnyModelConfig:
-    """Add a model using the configuration information appropriate for its type."""
-    logger = ApiDependencies.invoker.services.logger
-    record_store = ApiDependencies.invoker.services.model_manager.store
-    if config.key == "<NOKEY>":
-        config.key = sha1(randbytes(100)).hexdigest()
-        logger.info(f"Created model {config.key} for {config.name}")
-    try:
-        record_store.add_model(config.key, config)
-    except DuplicateModelException as e:
-        logger.error(str(e))
-        raise HTTPException(status_code=409, detail=str(e))
-    except InvalidModelException as e:
-        logger.error(str(e))
-        raise HTTPException(status_code=415)
+# @model_manager_router.post(
+#     "/i/",
+#     operation_id="add_model_record",
+#     responses={
+#         201: {
+#             "description": "The model added successfully",
+#             "content": {"application/json": {"example": example_model_config}},
+#         },
+#         409: {"description": "There is already a model corresponding to this path or repo_id"},
+#         415: {"description": "Unrecognized file/folder format"},
+#     },
+#     status_code=201,
+# )
+# async def add_model_record(
+#     config: Annotated[
+#         AnyModelConfig, Body(description="Model config", discriminator="type", example=example_model_input)
+#     ],
+# ) -> AnyModelConfig:
+#     """Add a model using the configuration information appropriate for its type."""
+#     logger = ApiDependencies.invoker.services.logger
+#     record_store = ApiDependencies.invoker.services.model_manager.store
+#     try:
+#         record_store.add_model(config)
+#     except DuplicateModelException as e:
+#         logger.error(str(e))
+#         raise HTTPException(status_code=409, detail=str(e))
+#     except InvalidModelException as e:
+#         logger.error(str(e))
+#         raise HTTPException(status_code=415)
 
-    # now fetch it out
-    result: AnyModelConfig = record_store.get_model(config.key)
-    return result
+#     # now fetch it out
+#     result: AnyModelConfig = record_store.get_model(config.key)
+#     return result
 
 
 @model_manager_router.post(
@@ -495,6 +344,7 @@ async def add_model_record(
 )
 async def install_model(
     source: str = Query(description="Model source to install, can be a local path, repo_id, or remote URL"),
+    inplace: Optional[bool] = Query(description="Whether or not to install a local model in place", default=False),
     # TODO(MM2): Can we type this?
     config: Optional[Dict[str, Any]] = Body(
         description="Dict of fields that override auto-probed values in the model config record, such as name, description and prediction_type ",
@@ -537,6 +387,7 @@ async def install_model(
             source=source,
             config=config,
             access_token=access_token,
+            inplace=bool(inplace),
         )
         logger.info(f"Started installation of {source}")
     except UnknownModelException as e:
@@ -552,10 +403,10 @@ async def install_model(
 
 
 @model_manager_router.get(
-    "/import",
-    operation_id="list_model_install_jobs",
+    "/install",
+    operation_id="list_model_installs",
 )
-async def list_model_install_jobs() -> List[ModelInstallJob]:
+async def list_model_installs() -> List[ModelInstallJob]:
     """Return the list of model install jobs.
 
     Install jobs have a numeric `id`, a `status`, and other fields that provide information on
@@ -569,9 +420,8 @@ async def list_model_install_jobs() -> List[ModelInstallJob]:
     * "cancelled" -- Job was cancelled before completion.
 
     Once completed, information about the model such as its size, base
-    model, type, and metadata can be retrieved from the `config_out`
-    field. For multi-file models such as diffusers, information on individual files
-    can be retrieved from `download_parts`.
+    model and type can be retrieved from the `config_out` field. For multi-file models such as diffusers,
+    information on individual files can be retrieved from `download_parts`.
 
     See the example and schema below for more information.
     """
@@ -580,7 +430,7 @@ async def list_model_install_jobs() -> List[ModelInstallJob]:
 
 
 @model_manager_router.get(
-    "/import/{id}",
+    "/install/{id}",
     operation_id="get_model_install_job",
     responses={
         200: {"description": "Success"},
@@ -600,7 +450,7 @@ async def get_model_install_job(id: int = Path(description="Model install id"))
 
 
 @model_manager_router.delete(
-    "/import/{id}",
+    "/install/{id}",
     operation_id="cancel_model_install_job",
     responses={
         201: {"description": "The job was cancelled successfully"},
@@ -618,8 +468,8 @@ async def cancel_model_install_job(id: int = Path(description="Model install job
     installer.cancel_job(job)
 
 
-@model_manager_router.patch(
-    "/import",
+@model_manager_router.delete(
+    "/install",
     operation_id="prune_model_install_jobs",
     responses={
         204: {"description": "All completed and errored jobs have been pruned"},
@@ -698,7 +548,8 @@ async def convert_model(
     # temporarily rename the original safetensors file so that there is no naming conflict
     original_name = model_config.name
     model_config.name = f"{original_name}.DELETE"
-    store.update_model(key, config=model_config)
+    changes = ModelRecordChanges(name=model_config.name)
+    store.update_model(key, changes=changes)
 
     # install the diffusers
     try:
@@ -707,7 +558,7 @@ async def convert_model(
             config={
                 "name": original_name,
                 "description": model_config.description,
-                "original_hash": model_config.original_hash,
+                "hash": model_config.hash,
                 "source": model_config.source,
             },
         )
@@ -715,10 +566,6 @@ async def convert_model(
         logger.error(str(e))
         raise HTTPException(status_code=409, detail=str(e))
 
-    # get the original metadata
-    if orig_metadata := store.get_metadata(key):
-        store.metadata_store.add_metadata(new_key, orig_metadata)
-
     # delete the original safetensors file
     installer.delete(key)
 
@@ -730,66 +577,66 @@ async def convert_model(
     return new_config
 
 
-@model_manager_router.put(
-    "/merge",
-    operation_id="merge",
-    responses={
-        200: {
-            "description": "Model converted successfully",
-            "content": {"application/json": {"example": example_model_config}},
-        },
-        400: {"description": "Bad request"},
-        404: {"description": "Model not found"},
-        409: {"description": "There is already a model registered at this location"},
-    },
-)
-async def merge(
-    keys: List[str] = Body(description="Keys for two to three models to merge", min_length=2, max_length=3),
-    merged_model_name: Optional[str] = Body(description="Name of destination model", default=None),
-    alpha: float = Body(description="Alpha weighting strength to apply to 2d and 3d models", default=0.5),
-    force: bool = Body(
-        description="Force merging of models created with different versions of diffusers",
-        default=False,
-    ),
-    interp: Optional[MergeInterpolationMethod] = Body(description="Interpolation method", default=None),
-    merge_dest_directory: Optional[str] = Body(
-        description="Save the merged model to the designated directory (with 'merged_model_name' appended)",
-        default=None,
-    ),
-) -> AnyModelConfig:
-    """
-    Merge diffusers models. The process is controlled by a set parameters provided in the body of the request.
-    ```
-    Argument                Description [default]
-    --------               ----------------------
-    keys                   List of 2-3 model keys to merge together. All models must use the same base type.
-    merged_model_name      Name for the merged model [Concat model names]
-    alpha                  Alpha value (0.0-1.0). Higher values give more weight to the second model [0.5]
-    force                  If true, force the merge even if the models were generated by different versions of the diffusers library [False]
-    interp                 Interpolation method. One of "weighted_sum", "sigmoid", "inv_sigmoid" or "add_difference" [weighted_sum]
-    merge_dest_directory   Specify a directory to store the merged model in [models directory]
-    ```
-    """
-    logger = ApiDependencies.invoker.services.logger
-    try:
-        logger.info(f"Merging models: {keys} into {merge_dest_directory or '<MODELS>'}/{merged_model_name}")
-        dest = pathlib.Path(merge_dest_directory) if merge_dest_directory else None
-        installer = ApiDependencies.invoker.services.model_manager.install
-        merger = ModelMerger(installer)
-        model_names = [installer.record_store.get_model(x).name for x in keys]
-        response = merger.merge_diffusion_models_and_save(
-            model_keys=keys,
-            merged_model_name=merged_model_name or "+".join(model_names),
-            alpha=alpha,
-            interp=interp,
-            force=force,
-            merge_dest_directory=dest,
-        )
-    except UnknownModelException:
-        raise HTTPException(
-            status_code=404,
-            detail=f"One or more of the models '{keys}' not found",
-        )
-    except ValueError as e:
-        raise HTTPException(status_code=400, detail=str(e))
-    return response
+# @model_manager_router.put(
+#     "/merge",
+#     operation_id="merge",
+#     responses={
+#         200: {
+#             "description": "Model converted successfully",
+#             "content": {"application/json": {"example": example_model_config}},
+#         },
+#         400: {"description": "Bad request"},
+#         404: {"description": "Model not found"},
+#         409: {"description": "There is already a model registered at this location"},
+#     },
+# )
+# async def merge(
+#     keys: List[str] = Body(description="Keys for two to three models to merge", min_length=2, max_length=3),
+#     merged_model_name: Optional[str] = Body(description="Name of destination model", default=None),
+#     alpha: float = Body(description="Alpha weighting strength to apply to 2d and 3d models", default=0.5),
+#     force: bool = Body(
+#         description="Force merging of models created with different versions of diffusers",
+#         default=False,
+#     ),
+#     interp: Optional[MergeInterpolationMethod] = Body(description="Interpolation method", default=None),
+#     merge_dest_directory: Optional[str] = Body(
+#         description="Save the merged model to the designated directory (with 'merged_model_name' appended)",
+#         default=None,
+#     ),
+# ) -> AnyModelConfig:
+#     """
+#     Merge diffusers models. The process is controlled by a set parameters provided in the body of the request.
+#     ```
+#     Argument                Description [default]
+#     --------               ----------------------
+#     keys                   List of 2-3 model keys to merge together. All models must use the same base type.
+#     merged_model_name      Name for the merged model [Concat model names]
+#     alpha                  Alpha value (0.0-1.0). Higher values give more weight to the second model [0.5]
+#     force                  If true, force the merge even if the models were generated by different versions of the diffusers library [False]
+#     interp                 Interpolation method. One of "weighted_sum", "sigmoid", "inv_sigmoid" or "add_difference" [weighted_sum]
+#     merge_dest_directory   Specify a directory to store the merged model in [models directory]
+#     ```
+#     """
+#     logger = ApiDependencies.invoker.services.logger
+#     try:
+#         logger.info(f"Merging models: {keys} into {merge_dest_directory or '<MODELS>'}/{merged_model_name}")
+#         dest = pathlib.Path(merge_dest_directory) if merge_dest_directory else None
+#         installer = ApiDependencies.invoker.services.model_manager.install
+#         merger = ModelMerger(installer)
+#         model_names = [installer.record_store.get_model(x).name for x in keys]
+#         response = merger.merge_diffusion_models_and_save(
+#             model_keys=keys,
+#             merged_model_name=merged_model_name or "+".join(model_names),
+#             alpha=alpha,
+#             interp=interp,
+#             force=force,
+#             merge_dest_directory=dest,
+#         )
+#     except UnknownModelException:
+#         raise HTTPException(
+#             status_code=404,
+#             detail=f"One or more of the models '{keys}' not found",
+#         )
+#     except ValueError as e:
+#         raise HTTPException(status_code=400, detail=str(e))
+#     return response

invokeai/app/api_app.py

@@ -2,6 +2,7 @@
 # which are imported/used before parse_args() is called will get the default config values instead of the
 # values from the command line or config file.
 import sys
+from contextlib import asynccontextmanager
 
 from invokeai.app.api.no_cache_staticfiles import NoCacheStaticFiles
 from invokeai.version.invokeai_version import __version__
@@ -71,9 +72,25 @@ logger = InvokeAILogger.get_logger(config=app_config)
 mimetypes.add_type("application/javascript", ".js")
 mimetypes.add_type("text/css", ".css")
 
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    # Add startup event to load dependencies
+    ApiDependencies.initialize(config=app_config, event_handler_id=event_handler_id, logger=logger)
+    yield
+    # Shut down threads
+    ApiDependencies.shutdown()
+
+
 # Create the app
 # TODO: create this all in a method so configuration/etc. can be passed in?
-app = FastAPI(title="Invoke - Community Edition", docs_url=None, redoc_url=None, separate_input_output_schemas=False)
+app = FastAPI(
+    title="Invoke - Community Edition",
+    docs_url=None,
+    redoc_url=None,
+    separate_input_output_schemas=False,
+    lifespan=lifespan,
+)
 
 # Add event handler
 event_handler_id: int = id(app)
@@ -96,18 +113,6 @@ app.add_middleware(
 app.add_middleware(GZipMiddleware, minimum_size=1000)
 
 
-# Add startup event to load dependencies
-@app.on_event("startup")
-async def startup_event() -> None:
-    ApiDependencies.initialize(config=app_config, event_handler_id=event_handler_id, logger=logger)
-
-
-# Shut down threads
-@app.on_event("shutdown")
-async def shutdown_event() -> None:
-    ApiDependencies.shutdown()
-
-
 # Include all routers
 app.include_router(utilities.utilities_router, prefix="/api")
 app.include_router(model_manager.model_manager_router, prefix="/api")

invokeai/app/invocations/compel.py

@@ -1,24 +1,15 @@
-from typing import Iterator, List, Optional, Tuple, Union
+from typing import Iterator, List, Optional, Tuple, Union, cast
 
 import torch
 from compel import Compel, ReturnedEmbeddingsType
 from compel.prompt_parser import Blend, Conjunction, CrossAttentionControlSubstitute, FlattenedPrompt, Fragment
-from transformers import CLIPTokenizer
+from transformers import CLIPTextModel, CLIPTextModelWithProjection, CLIPTokenizer
 
-import invokeai.backend.util.logging as logger
-from invokeai.app.invocations.fields import (
-    FieldDescriptions,
-    Input,
-    InputField,
-    OutputField,
-    UIComponent,
-)
+from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIComponent
 from invokeai.app.invocations.primitives import ConditioningOutput
-from invokeai.app.services.model_records import UnknownModelException
 from invokeai.app.services.shared.invocation_context import InvocationContext
-from invokeai.app.util.ti_utils import extract_ti_triggers_from_prompt
+from invokeai.app.util.ti_utils import generate_ti_list
 from invokeai.backend.lora import LoRAModelRaw
-from invokeai.backend.model_manager import ModelType
 from invokeai.backend.model_patcher import ModelPatcher
 from invokeai.backend.stable_diffusion.diffusion.conditioning_data import (
     BasicConditioningInfo,
@@ -26,15 +17,9 @@ from invokeai.backend.stable_diffusion.diffusion.conditioning_data import (
     ExtraConditioningInfo,
     SDXLConditioningInfo,
 )
-from invokeai.backend.textual_inversion import TextualInversionModelRaw
 from invokeai.backend.util.devices import torch_dtype
 
-from .baseinvocation import (
-    BaseInvocation,
-    BaseInvocationOutput,
-    invocation,
-    invocation_output,
-)
+from .baseinvocation import BaseInvocation, BaseInvocationOutput, invocation, invocation_output
 from .model import ClipField
 
 # unconditioned: Optional[torch.Tensor]
@@ -70,7 +55,11 @@ class CompelInvocation(BaseInvocation):
     @torch.no_grad()
     def invoke(self, context: InvocationContext) -> ConditioningOutput:
         tokenizer_info = context.models.load(**self.clip.tokenizer.model_dump())
+        tokenizer_model = tokenizer_info.model
+        assert isinstance(tokenizer_model, CLIPTokenizer)
         text_encoder_info = context.models.load(**self.clip.text_encoder.model_dump())
+        text_encoder_model = text_encoder_info.model
+        assert isinstance(text_encoder_model, CLIPTextModel)
 
         def _lora_loader() -> Iterator[Tuple[LoRAModelRaw, float]]:
             for lora in self.clip.loras:
@@ -82,21 +71,10 @@ class CompelInvocation(BaseInvocation):
 
         # loras = [(context.models.get(**lora.dict(exclude={"weight"})).context.model, lora.weight) for lora in self.clip.loras]
 
-        ti_list = []
-        for trigger in extract_ti_triggers_from_prompt(self.prompt):
-            name = trigger[1:-1]
-            try:
-                loaded_model = context.models.load(key=name).model
-                assert isinstance(loaded_model, TextualInversionModelRaw)
-                ti_list.append((name, loaded_model))
-            except UnknownModelException:
-                # print(e)
-                # import traceback
-                # print(traceback.format_exc())
-                print(f'Warn: trigger: "{trigger}" not found')
+        ti_list = generate_ti_list(self.prompt, text_encoder_info.config.base, context)
 
         with (
-            ModelPatcher.apply_ti(tokenizer_info.model, text_encoder_info.model, ti_list) as (
+            ModelPatcher.apply_ti(tokenizer_model, text_encoder_model, ti_list) as (
                 tokenizer,
                 ti_manager,
             ),
@@ -104,8 +82,9 @@ class CompelInvocation(BaseInvocation):
             # Apply the LoRA after text_encoder has been moved to its target device for faster patching.
             ModelPatcher.apply_lora_text_encoder(text_encoder, _lora_loader()),
             # Apply CLIP Skip after LoRA to prevent LoRA application from failing on skipped layers.
-            ModelPatcher.apply_clip_skip(text_encoder_info.model, self.clip.skipped_layers),
+            ModelPatcher.apply_clip_skip(text_encoder_model, self.clip.skipped_layers),
         ):
+            assert isinstance(text_encoder, CLIPTextModel)
             compel = Compel(
                 tokenizer=tokenizer,
                 text_encoder=text_encoder,
@@ -155,7 +134,11 @@ class SDXLPromptInvocationBase:
         zero_on_empty: bool,
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[ExtraConditioningInfo]]:
         tokenizer_info = context.models.load(**clip_field.tokenizer.model_dump())
+        tokenizer_model = tokenizer_info.model
+        assert isinstance(tokenizer_model, CLIPTokenizer)
         text_encoder_info = context.models.load(**clip_field.text_encoder.model_dump())
+        text_encoder_model = text_encoder_info.model
+        assert isinstance(text_encoder_model, (CLIPTextModel, CLIPTextModelWithProjection))
 
         # return zero on empty
         if prompt == "" and zero_on_empty:
@@ -189,25 +172,10 @@ class SDXLPromptInvocationBase:
 
         # loras = [(context.models.get(**lora.dict(exclude={"weight"})).context.model, lora.weight) for lora in self.clip.loras]
 
-        ti_list = []
-        for trigger in extract_ti_triggers_from_prompt(prompt):
-            name = trigger[1:-1]
-            try:
-                ti_model = context.models.load_by_attrs(
-                    model_name=name, base_model=text_encoder_info.config.base, model_type=ModelType.TextualInversion
-                ).model
-                assert isinstance(ti_model, TextualInversionModelRaw)
-                ti_list.append((name, ti_model))
-            except UnknownModelException:
-                # print(e)
-                # import traceback
-                # print(traceback.format_exc())
-                logger.warning(f'trigger: "{trigger}" not found')
-            except ValueError:
-                logger.warning(f'trigger: "{trigger}" more than one similarly-named textual inversion models')
+        ti_list = generate_ti_list(prompt, text_encoder_info.config.base, context)
 
         with (
-            ModelPatcher.apply_ti(tokenizer_info.model, text_encoder_info.model, ti_list) as (
+            ModelPatcher.apply_ti(tokenizer_model, text_encoder_model, ti_list) as (
                 tokenizer,
                 ti_manager,
             ),
@@ -215,8 +183,10 @@ class SDXLPromptInvocationBase:
             # Apply the LoRA after text_encoder has been moved to its target device for faster patching.
             ModelPatcher.apply_lora(text_encoder, _lora_loader(), lora_prefix),
             # Apply CLIP Skip after LoRA to prevent LoRA application from failing on skipped layers.
-            ModelPatcher.apply_clip_skip(text_encoder_info.model, clip_field.skipped_layers),
+            ModelPatcher.apply_clip_skip(text_encoder_model, clip_field.skipped_layers),
         ):
+            assert isinstance(text_encoder, (CLIPTextModel, CLIPTextModelWithProjection))
+            text_encoder = cast(CLIPTextModel, text_encoder)
             compel = Compel(
                 tokenizer=tokenizer,
                 text_encoder=text_encoder,

invokeai/app/invocations/ip_adapter.py

@@ -93,7 +93,7 @@ class IPAdapterInvocation(BaseInvocation):
         image_encoder_model_id = ip_adapter_info.image_encoder_model_id
         image_encoder_model_name = image_encoder_model_id.split("/")[-1].strip()
         image_encoder_models = context.models.search_by_attrs(
-            model_name=image_encoder_model_name, base_model=BaseModelType.Any, model_type=ModelType.CLIPVision
+            name=image_encoder_model_name, base=BaseModelType.Any, type=ModelType.CLIPVision
         )
         assert len(image_encoder_models) == 1
         image_encoder_model = CLIPVisionModelField(key=image_encoder_models[0].key)

invokeai/app/invocations/latent.py

@@ -173,6 +173,16 @@ class CreateDenoiseMaskInvocation(BaseInvocation):
         )
 
 
+@invocation_output("gradient_mask_output")
+class GradientMaskOutput(BaseInvocationOutput):
+    """Outputs a denoise mask and an image representing the total gradient of the mask."""
+
+    denoise_mask: DenoiseMaskField = OutputField(description="Mask for denoise model run")
+    expanded_mask_area: ImageField = OutputField(
+        description="Image representing the total gradient area of the mask. For paste-back purposes."
+    )
+
+
 @invocation(
     "create_gradient_mask",
     title="Create Gradient Mask",
@@ -193,38 +203,42 @@ class CreateGradientMaskInvocation(BaseInvocation):
     )
 
     @torch.no_grad()
-    def invoke(self, context: InvocationContext) -> DenoiseMaskOutput:
+    def invoke(self, context: InvocationContext) -> GradientMaskOutput:
         mask_image = context.images.get_pil(self.mask.image_name, mode="L")
-        if self.coherence_mode == "Box Blur":
-            blur_mask = mask_image.filter(ImageFilter.BoxBlur(self.edge_radius))
-        else:  # Gaussian Blur OR Staged
-            # Gaussian Blur uses standard deviation. 1/2 radius is a good approximation
-            blur_mask = mask_image.filter(ImageFilter.GaussianBlur(self.edge_radius / 2))
+        if self.edge_radius > 0:
+            if self.coherence_mode == "Box Blur":
+                blur_mask = mask_image.filter(ImageFilter.BoxBlur(self.edge_radius))
+            else:  # Gaussian Blur OR Staged
+                # Gaussian Blur uses standard deviation. 1/2 radius is a good approximation
+                blur_mask = mask_image.filter(ImageFilter.GaussianBlur(self.edge_radius / 2))
 
-        mask_tensor: torch.Tensor = image_resized_to_grid_as_tensor(mask_image, normalize=False)
-        blur_tensor: torch.Tensor = image_resized_to_grid_as_tensor(blur_mask, normalize=False)
+            blur_tensor: torch.Tensor = image_resized_to_grid_as_tensor(blur_mask, normalize=False)
 
-        # redistribute blur so that the edges are 0 and blur out to 1
-        blur_tensor = (blur_tensor - 0.5) * 2
+            # redistribute blur so that the original edges are 0 and blur outwards to 1
+            blur_tensor = (blur_tensor - 0.5) * 2
 
-        threshold = 1 - self.minimum_denoise
+            threshold = 1 - self.minimum_denoise
+
+            if self.coherence_mode == "Staged":
+                # wherever the blur_tensor is less than fully masked, convert it to threshold
+                blur_tensor = torch.where((blur_tensor < 1) & (blur_tensor > 0), threshold, blur_tensor)
+            else:
+                # wherever the blur_tensor is above threshold but less than 1, drop it to threshold
+                blur_tensor = torch.where((blur_tensor > threshold) & (blur_tensor < 1), threshold, blur_tensor)
 
-        if self.coherence_mode == "Staged":
-            # wherever the blur_tensor is masked to any degree, convert it to threshold
-            blur_tensor = torch.where((blur_tensor < 1), threshold, blur_tensor)
         else:
-            # wherever the blur_tensor is above threshold but less than 1, drop it to threshold
-            blur_tensor = torch.where((blur_tensor > threshold) & (blur_tensor < 1), threshold, blur_tensor)
-
-        # multiply original mask to force actually masked regions to 0
-        blur_tensor = mask_tensor * blur_tensor
+            blur_tensor: torch.Tensor = image_resized_to_grid_as_tensor(mask_image, normalize=False)
 
         mask_name = context.tensors.save(tensor=blur_tensor.unsqueeze(1))
 
-        return DenoiseMaskOutput.build(
-            mask_name=mask_name,
-            masked_latents_name=None,
-            gradient=True,
+        # compute a [0, 1] mask from the blur_tensor
+        expanded_mask = torch.where((blur_tensor < 1), 0, 1)
+        expanded_mask_image = Image.fromarray((expanded_mask.squeeze(0).numpy() * 255).astype(np.uint8), mode="L")
+        expanded_image_dto = context.images.save(expanded_mask_image)
+
+        return GradientMaskOutput(
+            denoise_mask=DenoiseMaskField(mask_name=mask_name, masked_latents_name=None, gradient=True),
+            expanded_mask_area=ImageField(image_name=expanded_image_dto.image_name),
         )
 
 
@@ -360,7 +374,6 @@ class DenoiseLatentsInvocation(BaseInvocation):
     ) -> ConditioningData:
         positive_cond_data = context.conditioning.load(self.positive_conditioning.conditioning_name)
         c = positive_cond_data.conditionings[0].to(device=unet.device, dtype=unet.dtype)
-        extra_conditioning_info = c.extra_conditioning
 
         negative_cond_data = context.conditioning.load(self.negative_conditioning.conditioning_name)
         uc = negative_cond_data.conditionings[0].to(device=unet.device, dtype=unet.dtype)
@@ -370,7 +383,6 @@ class DenoiseLatentsInvocation(BaseInvocation):
             text_embeddings=c,
             guidance_scale=self.cfg_scale,
             guidance_rescale_multiplier=self.cfg_rescale_multiplier,
-            extra=extra_conditioning_info,
             postprocessing_settings=PostprocessingSettings(
                 threshold=0.0,  # threshold,
                 warmup=0.2,  # warmup,
@@ -777,10 +789,7 @@ class DenoiseLatentsInvocation(BaseInvocation):
                     denoising_end=self.denoising_end,
                 )
 
-                (
-                    result_latents,
-                    result_attention_map_saver,
-                ) = pipeline.latents_from_embeddings(
+                result_latents = pipeline.latents_from_embeddings(
                     latents=latents,
                     timesteps=timesteps,
                     init_timestep=init_timestep,

invokeai/app/invocations/model.py

@@ -133,7 +133,7 @@ class MainModelLoaderInvocation(BaseInvocation):
             vae=VaeField(
                 vae=ModelInfo(
                     key=key,
-                    submodel_type=SubModelType.Vae,
+                    submodel_type=SubModelType.VAE,
                 ),
             ),
         )

invokeai/app/invocations/sdxl.py

@@ -85,7 +85,7 @@ class SDXLModelLoaderInvocation(BaseInvocation):
             vae=VaeField(
                 vae=ModelInfo(
                     key=model_key,
-                    submodel_type=SubModelType.Vae,
+                    submodel_type=SubModelType.VAE,
                 ),
             ),
         )
@@ -142,7 +142,7 @@ class SDXLRefinerModelLoaderInvocation(BaseInvocation):
             vae=VaeField(
                 vae=ModelInfo(
                     key=model_key,
-                    submodel_type=SubModelType.Vae,
+                    submodel_type=SubModelType.VAE,
                 ),
             ),
         )

invokeai/app/services/config/config_default.py

@@ -166,6 +166,7 @@ two configs are kept in separate sections of the config file:
      ...
 
 """
+
 from __future__ import annotations
 
 import os
@@ -255,6 +256,7 @@ class InvokeAIAppConfig(InvokeAISettings):
     profile_graphs      : bool = Field(default=False, description="Enable graph profiling", json_schema_extra=Categories.Development)
     profile_prefix      : Optional[str] = Field(default=None, description="An optional prefix for profile output files.", json_schema_extra=Categories.Development)
     profiles_dir        : Path = Field(default=Path('profiles'), description="Directory for graph profiles", json_schema_extra=Categories.Development)
+    skip_model_hash     : bool = Field(default=False, description="Skip model hashing, instead assigning a UUID to models. Useful when using a memory db to reduce startup time.", json_schema_extra=Categories.Development)
 
     version             : bool = Field(default=False, description="Show InvokeAI version and exit", json_schema_extra=Categories.Other)
 

invokeai/app/services/download/__init__.py

@@ -1,4 +1,5 @@
 """Init file for download queue."""
+
 from .download_base import DownloadJob, DownloadJobStatus, DownloadQueueServiceBase, UnknownJobIDException
 from .download_default import DownloadQueueService, TqdmProgress
 

invokeai/app/services/download/download_default.py

@@ -224,7 +224,6 @@ class DownloadQueueService(DownloadQueueServiceBase):
                 job.job_started = get_iso_timestamp()
                 self._do_download(job)
                 self._signal_job_complete(job)
-
             except (OSError, HTTPError) as excp:
                 job.error_type = excp.__class__.__name__ + f"({str(excp)})"
                 job.error = traceback.format_exc()

invokeai/app/services/model_install/model_install_base.py

@@ -18,16 +18,16 @@ from invokeai.app.services.events.events_base import EventServiceBase
 from invokeai.app.services.invoker import Invoker
 from invokeai.app.services.model_records import ModelRecordServiceBase
 from invokeai.backend.model_manager import AnyModelConfig, ModelRepoVariant
+from invokeai.backend.model_manager.config import ModelSourceType
 from invokeai.backend.model_manager.metadata import AnyModelRepoMetadata
 
-from ..model_metadata import ModelMetadataStoreBase
-
 
 class InstallStatus(str, Enum):
     """State of an install job running in the background."""
 
     WAITING = "waiting"  # waiting to be dequeued
     DOWNLOADING = "downloading"  # downloading of model files in process
+    DOWNLOADS_DONE = "downloads_done"  # downloading done, waiting to run
     RUNNING = "running"  # being processed
     COMPLETED = "completed"  # finished running
     ERROR = "error"  # terminated with an error message
@@ -150,6 +150,13 @@ ModelSource = Annotated[
     Union[LocalModelSource, HFModelSource, CivitaiModelSource, URLModelSource], Field(discriminator="type")
 ]
 
+MODEL_SOURCE_TO_TYPE_MAP = {
+    URLModelSource: ModelSourceType.Url,
+    HFModelSource: ModelSourceType.HFRepoID,
+    CivitaiModelSource: ModelSourceType.CivitAI,
+    LocalModelSource: ModelSourceType.Path,
+}
+
 
 class ModelInstallJob(BaseModel):
     """Object that tracks the current status of an install request."""
@@ -229,6 +236,11 @@ class ModelInstallJob(BaseModel):
         """Return true if job is downloading."""
         return self.status == InstallStatus.DOWNLOADING
 
+    @property
+    def downloads_done(self) -> bool:
+        """Return true if job's downloads ae done."""
+        return self.status == InstallStatus.DOWNLOADS_DONE
+
     @property
     def running(self) -> bool:
         """Return true if job is running."""
@@ -254,7 +266,6 @@ class ModelInstallServiceBase(ABC):
         app_config: InvokeAIAppConfig,
         record_store: ModelRecordServiceBase,
         download_queue: DownloadQueueServiceBase,
-        metadata_store: ModelMetadataStoreBase,
         event_bus: Optional["EventServiceBase"] = None,
     ):
         """
@@ -341,6 +352,7 @@ class ModelInstallServiceBase(ABC):
         source: str,
         config: Optional[Dict[str, Any]] = None,
         access_token: Optional[str] = None,
+        inplace: Optional[bool] = False,
     ) -> ModelInstallJob:
         r"""Install the indicated model using heuristics to interpret user intentions.
 
@@ -386,7 +398,7 @@ class ModelInstallServiceBase(ABC):
          will override corresponding autoassigned probe fields in the
          model's config record. Use it to override
          `name`, `description`, `base_type`, `model_type`, `format`,
-         `prediction_type`, `image_size`, and/or `ztsnr_training`.
+         `prediction_type`, and/or `image_size`.
 
         This will download the model located at `source`,
         probe it, and install it into the models directory.

invokeai/app/services/model_install/model_install_default.py

@@ -7,7 +7,6 @@ import time
 from hashlib import sha256
 from pathlib import Path
 from queue import Empty, Queue
-from random import randbytes
 from shutil import copyfile, copytree, move, rmtree
 from tempfile import mkdtemp
 from typing import Any, Dict, List, Optional, Set, Union
@@ -21,14 +20,17 @@ from invokeai.app.services.download import DownloadJob, DownloadQueueServiceBase
 from invokeai.app.services.events.events_base import EventServiceBase
 from invokeai.app.services.invoker import Invoker
 from invokeai.app.services.model_records import DuplicateModelException, ModelRecordServiceBase
+from invokeai.app.services.model_records.model_records_base import ModelRecordChanges
+from invokeai.app.util.misc import uuid_string
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     BaseModelType,
+    CheckpointConfigBase,
     InvalidModelConfigException,
     ModelRepoVariant,
+    ModelSourceType,
     ModelType,
 )
-from invokeai.backend.model_manager.hash import FastModelHash
 from invokeai.backend.model_manager.metadata import (
     AnyModelRepoMetadata,
     CivitaiMetadataFetch,
@@ -36,12 +38,14 @@ from invokeai.backend.model_manager.metadata import (
     ModelMetadataWithFiles,
     RemoteModelFile,
 )
+from invokeai.backend.model_manager.metadata.metadata_base import CivitaiMetadata, HuggingFaceMetadata
 from invokeai.backend.model_manager.probe import ModelProbe
 from invokeai.backend.model_manager.search import ModelSearch
 from invokeai.backend.util import Chdir, InvokeAILogger
 from invokeai.backend.util.devices import choose_precision, choose_torch_device
 
 from .model_install_base import (
+    MODEL_SOURCE_TO_TYPE_MAP,
     CivitaiModelSource,
     HFModelSource,
     InstallStatus,
@@ -91,7 +95,6 @@ class ModelInstallService(ModelInstallServiceBase):
         self._running = False
         self._session = session
         self._next_job_id = 0
-        self._metadata_store = record_store.metadata_store  # for convenience
 
     @property
     def app_config(self) -> InvokeAIAppConfig:  # noqa D102
@@ -140,6 +143,7 @@ class ModelInstallService(ModelInstallServiceBase):
         config = config or {}
         if not config.get("source"):
             config["source"] = model_path.resolve().as_posix()
+        config["source_type"] = ModelSourceType.Path
         return self._register(model_path, config)
 
     def install_path(
@@ -149,11 +153,11 @@ class ModelInstallService(ModelInstallServiceBase):
     ) -> str:  # noqa D102
         model_path = Path(model_path)
         config = config or {}
-        if not config.get("source"):
-            config["source"] = model_path.resolve().as_posix()
 
-        info: AnyModelConfig = self._probe_model(Path(model_path), config)
-        old_hash = info.current_hash
+        if self._app_config.skip_model_hash:
+            config["hash"] = uuid_string()
+
+        info: AnyModelConfig = ModelProbe.probe(Path(model_path), config)
 
         if preferred_name := config.get("name"):
             preferred_name = Path(preferred_name).with_suffix(model_path.suffix)
@@ -167,8 +171,6 @@ class ModelInstallService(ModelInstallServiceBase):
             raise DuplicateModelException(
                 f"A model named {model_path.name} is already installed at {dest_path.as_posix()}"
             ) from excp
-        new_hash = FastModelHash.hash(new_path)
-        assert new_hash == old_hash, f"{model_path}: Model hash changed during installation, possibly corrupted."
 
         return self._register(
             new_path,
@@ -181,13 +183,14 @@ class ModelInstallService(ModelInstallServiceBase):
         source: str,
         config: Optional[Dict[str, Any]] = None,
         access_token: Optional[str] = None,
+        inplace: Optional[bool] = False,
     ) -> ModelInstallJob:
         variants = "|".join(ModelRepoVariant.__members__.values())
         hf_repoid_re = f"^([^/:]+/[^/:]+)(?::({variants})?(?::/?([^:]+))?)?$"
         source_obj: Optional[StringLikeSource] = None
 
         if Path(source).exists():  # A local file or directory
-            source_obj = LocalModelSource(path=Path(source))
+            source_obj = LocalModelSource(path=Path(source), inplace=inplace)
         elif match := re.match(hf_repoid_re, source):
             source_obj = HFModelSource(
                 repo_id=match.group(1),
@@ -282,9 +285,9 @@ class ModelInstallService(ModelInstallServiceBase):
         self._logger.info("Model installer (re)initialized")
 
     def scan_directory(self, scan_dir: Path, install: bool = False) -> List[str]:  # noqa D102
-        self._cached_model_paths = {Path(x.path) for x in self.record_store.all_models()}
+        self._cached_model_paths = {Path(x.path).absolute() for x in self.record_store.all_models()}
         callback = self._scan_install if install else self._scan_register
-        search = ModelSearch(on_model_found=callback)
+        search = ModelSearch(on_model_found=callback, config=self._app_config)
         self._models_installed.clear()
         search.search(scan_dir)
         return list(self._models_installed)
@@ -370,21 +373,24 @@ class ModelInstallService(ModelInstallServiceBase):
                     self._signal_job_errored(job)
 
                 elif (
-                    job.waiting or job.downloading
+                    job.waiting or job.downloads_done
                 ):  # local jobs will be in waiting state, remote jobs will be downloading state
                     job.total_bytes = self._stat_size(job.local_path)
                     job.bytes = job.total_bytes
                     self._signal_job_running(job)
                     job.config_in["source"] = str(job.source)
+                    job.config_in["source_type"] = MODEL_SOURCE_TO_TYPE_MAP[job.source.__class__]
+                    # enter the metadata, if there is any
+                    if isinstance(job.source_metadata, (CivitaiMetadata, HuggingFaceMetadata)):
+                        job.config_in["source_api_response"] = job.source_metadata.api_response
+                    if isinstance(job.source_metadata, CivitaiMetadata) and job.source_metadata.trigger_phrases:
+                        job.config_in["trigger_phrases"] = job.source_metadata.trigger_phrases
+
                     if job.inplace:
                         key = self.register_path(job.local_path, job.config_in)
                     else:
                         key = self.install_path(job.local_path, job.config_in)
                     job.config_out = self.record_store.get_model(key)
-
-                    # enter the metadata, if there is any
-                    if job.source_metadata:
-                        self._metadata_store.add_metadata(key, job.source_metadata)
                     self._signal_job_completed(job)
 
             except InvalidModelConfigException as excp:
@@ -448,7 +454,7 @@ class ModelInstallService(ModelInstallServiceBase):
                     installed.update(self.scan_directory(models_dir))
             self._logger.info(f"{len(installed)} new models registered; {len(defunct_models)} unregistered")
 
-    def _sync_model_path(self, key: str, ignore_hash_change: bool = False) -> AnyModelConfig:
+    def _sync_model_path(self, key: str) -> AnyModelConfig:
         """
         Move model into the location indicated by its basetype, type and name.
 
@@ -469,15 +475,8 @@ class ModelInstallService(ModelInstallServiceBase):
         new_path = models_dir / model.base.value / model.type.value / model.name
         self._logger.info(f"Moving {model.name} to {new_path}.")
         new_path = self._move_model(old_path, new_path)
-        new_hash = FastModelHash.hash(new_path)
         model.path = new_path.relative_to(models_dir).as_posix()
-        if model.current_hash != new_hash:
-            assert (
-                ignore_hash_change
-            ), f"{model.name}: Model hash changed during installation, model is possibly corrupted"
-            model.current_hash = new_hash
-            self._logger.info(f"Model has new hash {model.current_hash}, but will continue to be identified by {key}")
-        self.record_store.update_model(key, model)
+        self.record_store.update_model(key, ModelRecordChanges(path=model.path))
         return model
 
     def _scan_register(self, model: Path) -> bool:
@@ -529,22 +528,14 @@ class ModelInstallService(ModelInstallServiceBase):
         move(old_path, new_path)
         return new_path
 
-    def _probe_model(self, model_path: Path, config: Optional[Dict[str, Any]] = None) -> AnyModelConfig:
-        info: AnyModelConfig = ModelProbe.probe(Path(model_path))
-        if config:  # used to override probe fields
-            for key, value in config.items():
-                setattr(info, key, value)
-        return info
-
-    def _create_key(self) -> str:
-        return sha256(randbytes(100)).hexdigest()[0:32]
-
     def _register(
         self, model_path: Path, config: Optional[Dict[str, Any]] = None, info: Optional[AnyModelConfig] = None
     ) -> str:
-        key = self._create_key()
-        if config and not config.get("key", None):
-            config["key"] = key
+        config = config or {}
+
+        if self._app_config.skip_model_hash:
+            config["hash"] = uuid_string()
+
         info = info or ModelProbe.probe(model_path, config)
 
         model_path = model_path.absolute()
@@ -554,11 +545,11 @@ class ModelInstallService(ModelInstallServiceBase):
         info.path = model_path.as_posix()
 
         # add 'main' specific fields
-        if hasattr(info, "config"):
+        if isinstance(info, CheckpointConfigBase):
             # make config relative to our root
-            legacy_conf = (self.app_config.root_dir / self.app_config.legacy_conf_dir / info.config).resolve()
-            info.config = legacy_conf.relative_to(self.app_config.root_dir).as_posix()
-        self.record_store.add_model(info.key, info)
+            legacy_conf = (self.app_config.root_dir / self.app_config.legacy_conf_dir / info.config_path).resolve()
+            info.config_path = legacy_conf.relative_to(self.app_config.root_dir).as_posix()
+        self.record_store.add_model(info)
         return info.key
 
     def _next_id(self) -> int:
@@ -579,13 +570,15 @@ class ModelInstallService(ModelInstallServiceBase):
             source=source,
             config_in=config or {},
             local_path=Path(source.path),
-            inplace=source.inplace,
+            inplace=source.inplace or False,
         )
 
     def _import_from_civitai(self, source: CivitaiModelSource, config: Optional[Dict[str, Any]]) -> ModelInstallJob:
         if not source.access_token:
             self._logger.info("No Civitai access token provided; some models may not be downloadable.")
-        metadata = CivitaiMetadataFetch(self._session).from_id(str(source.version_id))
+        metadata = CivitaiMetadataFetch(self._session, self.app_config.get_config().civitai_api_key).from_id(
+            str(source.version_id)
+        )
         assert isinstance(metadata, ModelMetadataWithFiles)
         remote_files = metadata.download_urls(session=self._session)
         return self._import_remote_model(source=source, config=config, metadata=metadata, remote_files=remote_files)
@@ -613,15 +606,17 @@ class ModelInstallService(ModelInstallServiceBase):
 
     def _import_from_url(self, source: URLModelSource, config: Optional[Dict[str, Any]]) -> ModelInstallJob:
         # URLs from Civitai or HuggingFace will be handled specially
-        url_patterns = {
-            r"^https?://civitai.com/": CivitaiMetadataFetch,
-            r"^https?://huggingface.co/[^/]+/[^/]+$": HuggingFaceMetadataFetch,
-        }
         metadata = None
-        for pattern, fetcher in url_patterns.items():
-            if re.match(pattern, str(source.url), re.IGNORECASE):
-                metadata = fetcher(self._session).from_url(source.url)
-                break
+        fetcher = None
+        try:
+            fetcher = self.get_fetcher_from_url(str(source.url))
+        except ValueError:
+            pass
+        kwargs: dict[str, Any] = {"session": self._session}
+        if fetcher is CivitaiMetadataFetch:
+            kwargs["api_key"] = self._app_config.get_config().civitai_api_key
+        if fetcher is not None:
+            metadata = fetcher(**kwargs).from_url(source.url)
         self._logger.debug(f"metadata={metadata}")
         if metadata and isinstance(metadata, ModelMetadataWithFiles):
             remote_files = metadata.download_urls(session=self._session)
@@ -636,7 +631,7 @@ class ModelInstallService(ModelInstallServiceBase):
 
     def _import_remote_model(
         self,
-        source: ModelSource,
+        source: HFModelSource | CivitaiModelSource | URLModelSource,
         remote_files: List[RemoteModelFile],
         metadata: Optional[AnyModelRepoMetadata],
         config: Optional[Dict[str, Any]],
@@ -664,7 +659,7 @@ class ModelInstallService(ModelInstallServiceBase):
         # In the event that there is a subfolder specified in the source,
         # we need to remove it from the destination path in order to avoid
         # creating unwanted subfolders
-        if hasattr(source, "subfolder") and source.subfolder:
+        if isinstance(source, HFModelSource) and source.subfolder:
             root = Path(remote_files[0].path.parts[0])
             subfolder = root / source.subfolder
         else:
@@ -749,8 +744,8 @@ class ModelInstallService(ModelInstallServiceBase):
             self._download_cache.pop(download_job.source, None)
 
             # are there any more active jobs left in this task?
-            if all(x.complete for x in install_job.download_parts):
-                #  now enqueue job for actual installation into the models directory
+            if install_job.downloading and all(x.complete for x in install_job.download_parts):
+                install_job.status = InstallStatus.DOWNLOADS_DONE
                 self._install_queue.put(install_job)
 
             # Let other threads know that the number of downloads has changed
@@ -851,3 +846,11 @@ class ModelInstallService(ModelInstallServiceBase):
         self._logger.info(f"{job.source}: model installation was cancelled")
         if self._event_bus:
             self._event_bus.emit_model_install_cancelled(str(job.source))
+
+    @staticmethod
+    def get_fetcher_from_url(url: str):
+        if re.match(r"^https?://civitai.com/", url.lower()):
+            return CivitaiMetadataFetch
+        elif re.match(r"^https?://huggingface.co/[^/]+/[^/]+$", url.lower()):
+            return HuggingFaceMetadataFetch
+        raise ValueError(f"Unsupported model source: '{url}'")

invokeai/app/services/model_metadata/__init__.py

@@ -1,9 +0,0 @@
-"""Init file for ModelMetadataStoreService module."""
-
-from .metadata_store_base import ModelMetadataStoreBase
-from .metadata_store_sql import ModelMetadataStoreSQL
-
-__all__ = [
-    "ModelMetadataStoreBase",
-    "ModelMetadataStoreSQL",
-]

invokeai/app/services/model_metadata/metadata_store_base.py

@@ -1,83 +0,0 @@
-# Copyright (c) 2023 Lincoln D. Stein and the InvokeAI Development Team
-"""
-Storage for Model Metadata
-"""
-
-from abc import ABC, abstractmethod
-from typing import List, Optional, Set, Tuple
-
-from pydantic import Field
-from invokeai.app.util.model_exclude_null import BaseModelExcludeNull
-
-from invokeai.backend.model_manager.metadata import AnyModelRepoMetadata
-from invokeai.backend.model_manager.metadata.metadata_base import ModelDefaultSettings
-
-class ModelMetadataChanges(BaseModelExcludeNull, extra="allow"):
-    """A set of changes to apply to model metadata.
-
-    Only limited changes are valid:
-      - `trigger_phrases`: the list of trigger phrases for this model
-      - `default_settings`: the user-configured default settings for this model
-    """
-
-    trigger_phrases: Optional[List[str]] = Field(default=None, description="The model's list of trigger phrases")
-    """The model's list of trigger phrases"""
-
-    default_settings: Optional[ModelDefaultSettings] = Field(default=None, description="The user-configured default settings for this model")
-    """The user-configured default settings for this model"""
-
-
-class ModelMetadataStoreBase(ABC):
-    """Store, search and fetch model metadata retrieved from remote repositories."""
-
-    @abstractmethod
-    def add_metadata(self, model_key: str, metadata: AnyModelRepoMetadata) -> None:
-        """
-        Add a block of repo metadata to a model record.
-
-        The model record config must already exist in the database with the
-        same key. Otherwise a FOREIGN KEY constraint exception will be raised.
-
-        :param model_key: Existing model key in the `model_config` table
-        :param metadata: ModelRepoMetadata object to store
-        """
-
-    @abstractmethod
-    def get_metadata(self, model_key: str) -> AnyModelRepoMetadata:
-        """Retrieve the ModelRepoMetadata corresponding to model key."""
-
-    @abstractmethod
-    def list_all_metadata(self) -> List[Tuple[str, AnyModelRepoMetadata]]:  # key, metadata
-        """Dump out all the metadata."""
-
-    @abstractmethod
-    def update_metadata(self, model_key: str, metadata: AnyModelRepoMetadata) -> AnyModelRepoMetadata:
-        """
-        Update metadata corresponding to the model with the indicated key.
-
-        :param model_key: Existing model key in the `model_config` table
-        :param metadata: ModelRepoMetadata object to update
-        """
-
-    @abstractmethod
-    def list_tags(self) -> Set[str]:
-        """Return all tags in the tags table."""
-
-    @abstractmethod
-    def search_by_tag(self, tags: Set[str]) -> Set[str]:
-        """Return the keys of models containing all of the listed tags."""
-
-    @abstractmethod
-    def search_by_author(self, author: str) -> Set[str]:
-        """Return the keys of models authored by the indicated author."""
-
-    @abstractmethod
-    def search_by_name(self, name: str) -> Set[str]:
-        """
-        Return the keys of models with the indicated name.
-
-        Note that this is the name of the model given to it by
-        the remote source. The user may have changed the local
-        name. The local name will be located in the model config
-        record object.
-        """

invokeai/app/services/model_metadata/metadata_store_sql.py

@@ -1,225 +0,0 @@
-# Copyright (c) 2023 Lincoln D. Stein and the InvokeAI Development Team
-"""
-SQL Storage for Model Metadata
-"""
-
-import sqlite3
-from typing import List, Optional, Set, Tuple
-
-from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
-from invokeai.backend.model_manager.metadata import AnyModelRepoMetadata, UnknownMetadataException
-from invokeai.backend.model_manager.metadata.fetch import ModelMetadataFetchBase
-
-from .metadata_store_base import ModelMetadataStoreBase
-
-
-class ModelMetadataStoreSQL(ModelMetadataStoreBase):
-    """Store, search and fetch model metadata retrieved from remote repositories."""
-
-    def __init__(self, db: SqliteDatabase):
-        """
-        Initialize a new object from preexisting sqlite3 connection and threading lock objects.
-
-        :param conn: sqlite3 connection object
-        :param lock: threading Lock object
-        """
-        super().__init__()
-        self._db = db
-        self._cursor = self._db.conn.cursor()
-
-    def add_metadata(self, model_key: str, metadata: AnyModelRepoMetadata) -> None:
-        """
-        Add a block of repo metadata to a model record.
-
-        The model record config must already exist in the database with the
-        same key. Otherwise a FOREIGN KEY constraint exception will be raised.
-
-        :param model_key: Existing model key in the `model_config` table
-        :param metadata: ModelRepoMetadata object to store
-        """
-        json_serialized = metadata.model_dump_json()
-        with self._db.lock:
-            try:
-                self._cursor.execute(
-                    """--sql
-                    INSERT INTO model_metadata(
-                       id,
-                       metadata
-                    )
-                    VALUES (?,?);
-                    """,
-                    (
-                        model_key,
-                        json_serialized,
-                    ),
-                )
-                self._update_tags(model_key, metadata.tags)
-                self._db.conn.commit()
-            except sqlite3.IntegrityError as excp:  # FOREIGN KEY error: the key was not in model_config table
-                self._db.conn.rollback()
-                raise UnknownMetadataException from excp
-            except sqlite3.Error as excp:
-                self._db.conn.rollback()
-                raise excp
-
-    def get_metadata(self, model_key: str) -> AnyModelRepoMetadata:
-        """Retrieve the ModelRepoMetadata corresponding to model key."""
-        with self._db.lock:
-            self._cursor.execute(
-                """--sql
-                SELECT metadata FROM model_metadata
-                WHERE id=?;
-                """,
-                (model_key,),
-            )
-            rows = self._cursor.fetchone()
-            if not rows:
-                raise UnknownMetadataException("model metadata not found")
-            return ModelMetadataFetchBase.from_json(rows[0])
-
-    def list_all_metadata(self) -> List[Tuple[str, AnyModelRepoMetadata]]:  # key, metadata
-        """Dump out all the metadata."""
-        with self._db.lock:
-            self._cursor.execute(
-                """--sql
-                SELECT id,metadata FROM model_metadata;
-                """,
-                (),
-            )
-            rows = self._cursor.fetchall()
-        return [(x[0], ModelMetadataFetchBase.from_json(x[1])) for x in rows]
-
-    def update_metadata(self, model_key: str, metadata: AnyModelRepoMetadata) -> AnyModelRepoMetadata:
-        """
-        Update metadata corresponding to the model with the indicated key.
-
-        :param model_key: Existing model key in the `model_config` table
-        :param metadata: ModelRepoMetadata object to update
-        """
-        json_serialized = metadata.model_dump_json()  # turn it into a json string.
-        with self._db.lock:
-            try:
-                self._cursor.execute(
-                    """--sql
-                    UPDATE model_metadata
-                    SET
-                        metadata=?
-                    WHERE id=?;
-                    """,
-                    (json_serialized, model_key),
-                )
-                if self._cursor.rowcount == 0:
-                    raise UnknownMetadataException("model metadata not found")
-                self._update_tags(model_key, metadata.tags)
-                self._db.conn.commit()
-            except sqlite3.Error as e:
-                self._db.conn.rollback()
-                raise e
-            except Exception as e:
-                raise e
-
-        return self.get_metadata(model_key)
-
-    def list_tags(self) -> Set[str]:
-        """Return all tags in the tags table."""
-        self._cursor.execute(
-            """--sql
-            select tag_text from tags;
-            """
-        )
-        return {x[0] for x in self._cursor.fetchall()}
-
-    def search_by_tag(self, tags: Set[str]) -> Set[str]:
-        """Return the keys of models containing all of the listed tags."""
-        with self._db.lock:
-            try:
-                matches: Optional[Set[str]] = None
-                for tag in tags:
-                    self._cursor.execute(
-                        """--sql
-                        SELECT a.model_id FROM model_tags AS a,
-                                                     tags AS b
-                        WHERE a.tag_id=b.tag_id
-                          AND b.tag_text=?;
-                        """,
-                        (tag,),
-                    )
-                    model_keys = {x[0] for x in self._cursor.fetchall()}
-                    if matches is None:
-                        matches = model_keys
-                    matches = matches.intersection(model_keys)
-            except sqlite3.Error as e:
-                raise e
-        return matches if matches else set()
-
-    def search_by_author(self, author: str) -> Set[str]:
-        """Return the keys of models authored by the indicated author."""
-        self._cursor.execute(
-            """--sql
-            SELECT id FROM model_metadata
-            WHERE author=?;
-            """,
-            (author,),
-        )
-        return {x[0] for x in self._cursor.fetchall()}
-
-    def search_by_name(self, name: str) -> Set[str]:
-        """
-        Return the keys of models with the indicated name.
-
-        Note that this is the name of the model given to it by
-        the remote source. The user may have changed the local
-        name. The local name will be located in the model config
-        record object.
-        """
-        self._cursor.execute(
-            """--sql
-            SELECT id FROM model_metadata
-            WHERE name=?;
-            """,
-            (name,),
-        )
-        return {x[0] for x in self._cursor.fetchall()}
-
-    def _update_tags(self, model_key: str, tags: Optional[Set[str]]) -> None:
-        """Update tags for the model referenced by model_key."""
-        if tags:
-        # remove previous tags from this model
-            self._cursor.execute(
-                """--sql
-                DELETE FROM model_tags
-                WHERE model_id=?;
-                """,
-                (model_key,),
-            )
-
-            for tag in tags:
-                self._cursor.execute(
-                    """--sql
-                    INSERT OR IGNORE INTO tags (
-                    tag_text
-                    )
-                    VALUES (?);
-                    """,
-                    (tag,),
-                )
-                self._cursor.execute(
-                    """--sql
-                    SELECT tag_id
-                    FROM tags
-                    WHERE tag_text = ?
-                    LIMIT 1;
-                    """,
-                    (tag,),
-                )
-                tag_id = self._cursor.fetchone()[0]
-                self._cursor.execute(
-                    """--sql
-                    INSERT OR IGNORE INTO model_tags (
-                    model_id,
-                    tag_id
-                    )
-                    VALUES (?,?);
-                    """,
-                    (model_key, tag_id),
-                )

invokeai/app/services/model_records/__init__.py

@@ -1,4 +1,5 @@
 """Init file for model record services."""
+
 from .model_records_base import (  # noqa F401
     DuplicateModelException,
     InvalidModelException,

invokeai/app/services/model_records/model_records_base.py

@@ -6,20 +6,19 @@ Abstract base class for storing and retrieving model configuration records.
 from abc import ABC, abstractmethod
 from enum import Enum
 from pathlib import Path
-from typing import Any, Dict, List, Optional, Set, Tuple, Union
+from typing import List, Optional, Set, Union
 
 from pydantic import BaseModel, Field
 
 from invokeai.app.services.shared.pagination import PaginatedResults
+from invokeai.app.util.model_exclude_null import BaseModelExcludeNull
 from invokeai.backend.model_manager import (
     AnyModelConfig,
     BaseModelType,
     ModelFormat,
     ModelType,
 )
-from invokeai.backend.model_manager.metadata import AnyModelRepoMetadata
-
-from ..model_metadata import ModelMetadataStoreBase
+from invokeai.backend.model_manager.config import ModelDefaultSettings, ModelVariantType, SchedulerPredictionType
 
 
 class DuplicateModelException(Exception):
@@ -60,11 +59,33 @@ class ModelSummary(BaseModel):
     tags: Set[str] = Field(description="tags associated with model")
 
 
+class ModelRecordChanges(BaseModelExcludeNull):
+    """A set of changes to apply to a model."""
+
+    # Changes applicable to all models
+    name: Optional[str] = Field(description="Name of the model.", default=None)
+    path: Optional[str] = Field(description="Path to the model.", default=None)
+    description: Optional[str] = Field(description="Model description", default=None)
+    base: Optional[BaseModelType] = Field(description="The base model.", default=None)
+    trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
+    default_settings: Optional[ModelDefaultSettings] = Field(
+        description="Default settings for this model", default=None
+    )
+
+    # Checkpoint-specific changes
+    # TODO(MM2): Should we expose these? Feels footgun-y...
+    variant: Optional[ModelVariantType] = Field(description="The variant of the model.", default=None)
+    prediction_type: Optional[SchedulerPredictionType] = Field(
+        description="The prediction type of the model.", default=None
+    )
+    upcast_attention: Optional[bool] = Field(description="Whether to upcast attention.", default=None)
+
+
 class ModelRecordServiceBase(ABC):
     """Abstract base class for storage and retrieval of model configs."""
 
     @abstractmethod
-    def add_model(self, key: str, config: Union[Dict[str, Any], AnyModelConfig]) -> AnyModelConfig:
+    def add_model(self, config: AnyModelConfig) -> AnyModelConfig:
         """
         Add a model to the database.
 
@@ -88,13 +109,12 @@ class ModelRecordServiceBase(ABC):
         pass
 
     @abstractmethod
-    def update_model(self, key: str, config: Union[Dict[str, Any], AnyModelConfig]) -> AnyModelConfig:
+    def update_model(self, key: str, changes: ModelRecordChanges) -> AnyModelConfig:
         """
         Update the model, returning the updated version.
 
-        :param key: Unique key for the model to be updated
-        :param config: Model configuration record. Either a dict with the
-         required fields, or a ModelConfigBase instance.
+        :param key: Unique key for the model to be updated.
+        :param changes: A set of changes to apply to this model. Changes are validated before being written.
         """
         pass
 
@@ -109,40 +129,6 @@ class ModelRecordServiceBase(ABC):
         """
         pass
 
-    @property
-    @abstractmethod
-    def metadata_store(self) -> ModelMetadataStoreBase:
-        """Return a ModelMetadataStore initialized on the same database."""
-        pass
-
-    @abstractmethod
-    def get_metadata(self, key: str) -> Optional[AnyModelRepoMetadata]:
-        """
-        Retrieve metadata (if any) from when model was downloaded from a repo.
-
-        :param key: Model key
-        """
-        pass
-
-    @abstractmethod
-    def list_all_metadata(self) -> List[Tuple[str, AnyModelRepoMetadata]]:
-        """List metadata for all models that have it."""
-        pass
-
-    @abstractmethod
-    def search_by_metadata_tag(self, tags: Set[str]) -> List[AnyModelConfig]:
-        """
-        Search model metadata for ones with all listed tags and return their corresponding configs.
-
-        :param tags: Set of tags to search for. All tags must be present.
-        """
-        pass
-
-    @abstractmethod
-    def list_tags(self) -> Set[str]:
-        """Return a unique set of all the model tags in the metadata database."""
-        pass
-
     @abstractmethod
     def list_models(
         self, page: int = 0, per_page: int = 10, order_by: ModelRecordOrderBy = ModelRecordOrderBy.Default
@@ -217,21 +203,3 @@ class ModelRecordServiceBase(ABC):
                 f"More than one model matched the search criteria: base_model='{base_model}', model_type='{model_type}', model_name='{model_name}'."
             )
         return model_configs[0]
-
-    def rename_model(
-        self,
-        key: str,
-        new_name: str,
-    ) -> AnyModelConfig:
-        """
-        Rename the indicated model. Just a special case of update_model().
-
-        In some implementations, renaming the model may involve changing where
-        it is stored on the filesystem. So this is broken out.
-
-        :param key: Model key
-        :param new_name: New name for model
-        """
-        config = self.get_model(key)
-        config.name = new_name
-        return self.update_model(key, config)

invokeai/app/services/model_records/model_records_sql.py

@@ -39,12 +39,11 @@ Typical usage:
   configs = store.search_by_attr(base_model='sd-2', model_type='main')
 """
 
-
 import json
 import sqlite3
 from math import ceil
 from pathlib import Path
-from typing import Any, Dict, List, Optional, Set, Tuple, Union
+from typing import List, Optional, Union
 
 from invokeai.app.services.shared.pagination import PaginatedResults
 from invokeai.backend.model_manager.config import (
@@ -54,12 +53,11 @@ from invokeai.backend.model_manager.config import (
     ModelFormat,
     ModelType,
 )
-from invokeai.backend.model_manager.metadata import AnyModelRepoMetadata, UnknownMetadataException
 
-from ..model_metadata import ModelMetadataStoreBase, ModelMetadataStoreSQL
 from ..shared.sqlite.sqlite_database import SqliteDatabase
 from .model_records_base import (
     DuplicateModelException,
+    ModelRecordChanges,
     ModelRecordOrderBy,
     ModelRecordServiceBase,
     ModelSummary,
@@ -70,7 +68,7 @@ from .model_records_base import (
 class ModelRecordServiceSQL(ModelRecordServiceBase):
     """Implementation of the ModelConfigStore ABC using a SQL database."""
 
-    def __init__(self, db: SqliteDatabase, metadata_store: ModelMetadataStoreBase):
+    def __init__(self, db: SqliteDatabase):
         """
         Initialize a new object from preexisting sqlite3 connection and threading lock objects.
 
@@ -79,14 +77,13 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         super().__init__()
         self._db = db
         self._cursor = db.conn.cursor()
-        self._metadata_store = metadata_store
 
     @property
     def db(self) -> SqliteDatabase:
         """Return the underlying database."""
         return self._db
 
-    def add_model(self, key: str, config: Union[Dict[str, Any], AnyModelConfig]) -> AnyModelConfig:
+    def add_model(self, config: AnyModelConfig) -> AnyModelConfig:
         """
         Add a model to the database.
 
@@ -96,23 +93,19 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
 
         Can raise DuplicateModelException and InvalidModelConfigException exceptions.
         """
-        record = ModelConfigFactory.make_config(config, key=key)  # ensure it is a valid config obect.
-        json_serialized = record.model_dump_json()  # and turn it into a json string.
         with self._db.lock:
             try:
                 self._cursor.execute(
                     """--sql
-                    INSERT INTO model_config (
+                    INSERT INTO models (
                        id,
-                       original_hash,
                        config
                       )
-                    VALUES (?,?,?);
+                    VALUES (?,?);
                     """,
                     (
-                        key,
-                        record.original_hash,
-                        json_serialized,
+                        config.key,
+                        config.model_dump_json(),
                     ),
                 )
                 self._db.conn.commit()
@@ -120,12 +113,12 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
             except sqlite3.IntegrityError as e:
                 self._db.conn.rollback()
                 if "UNIQUE constraint failed" in str(e):
-                    if "model_config.path" in str(e):
-                        msg = f"A model with path '{record.path}' is already installed"
-                    elif "model_config.name" in str(e):
-                        msg = f"A model with name='{record.name}', type='{record.type}', base='{record.base}' is already installed"
+                    if "models.path" in str(e):
+                        msg = f"A model with path '{config.path}' is already installed"
+                    elif "models.name" in str(e):
+                        msg = f"A model with name='{config.name}', type='{config.type}', base='{config.base}' is already installed"
                     else:
-                        msg = f"A model with key '{key}' is already installed"
+                        msg = f"A model with key '{config.key}' is already installed"
                     raise DuplicateModelException(msg) from e
                 else:
                     raise e
@@ -133,7 +126,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
                 self._db.conn.rollback()
                 raise e
 
-        return self.get_model(key)
+        return self.get_model(config.key)
 
     def del_model(self, key: str) -> None:
         """
@@ -147,7 +140,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
             try:
                 self._cursor.execute(
                     """--sql
-                    DELETE FROM model_config
+                    DELETE FROM models
                     WHERE id=?;
                     """,
                     (key,),
@@ -159,21 +152,20 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
                 self._db.conn.rollback()
                 raise e
 
-    def update_model(self, key: str, config: Union[Dict[str, Any], AnyModelConfig]) -> AnyModelConfig:
-        """
-        Update the model, returning the updated version.
+    def update_model(self, key: str, changes: ModelRecordChanges) -> AnyModelConfig:
+        record = self.get_model(key)
+
+        # Model configs use pydantic's `validate_assignment`, so each change is validated by pydantic.
+        for field_name in changes.model_fields_set:
+            setattr(record, field_name, getattr(changes, field_name))
+
+        json_serialized = record.model_dump_json()
 
-        :param key: Unique key for the model to be updated
-        :param config: Model configuration record. Either a dict with the
-         required fields, or a ModelConfigBase instance.
-        """
-        record = ModelConfigFactory.make_config(config, key=key)  # ensure it is a valid config obect
-        json_serialized = record.model_dump_json()  # and turn it into a json string.
         with self._db.lock:
             try:
                 self._cursor.execute(
                     """--sql
-                    UPDATE model_config
+                    UPDATE models
                     SET
                         config=?
                     WHERE id=?;
@@ -200,7 +192,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         with self._db.lock:
             self._cursor.execute(
                 """--sql
-                SELECT config, strftime('%s',updated_at) FROM model_config
+                SELECT config, strftime('%s',updated_at) FROM models
                 WHERE id=?;
                 """,
                 (key,),
@@ -221,7 +213,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         with self._db.lock:
             self._cursor.execute(
                 """--sql
-                select count(*) FROM model_config
+                select count(*) FROM models
                 WHERE id=?;
                 """,
                 (key,),
@@ -247,9 +239,8 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         If none of the optional filters are passed, will return all
         models in the database.
         """
-        results = []
-        where_clause = []
-        bindings = []
+        where_clause: list[str] = []
+        bindings: list[str] = []
         if model_name:
             where_clause.append("name=?")
             bindings.append(model_name)
@@ -266,14 +257,13 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         with self._db.lock:
             self._cursor.execute(
                 f"""--sql
-                select config, strftime('%s',updated_at) FROM model_config
+                SELECT config, strftime('%s',updated_at) FROM models
                 {where};
                 """,
                 tuple(bindings),
             )
-            results = [
-                ModelConfigFactory.make_config(json.loads(x[0]), timestamp=x[1]) for x in self._cursor.fetchall()
-            ]
+            result = self._cursor.fetchall()
+            results = [ModelConfigFactory.make_config(json.loads(x[0]), timestamp=x[1]) for x in result]
         return results
 
     def search_by_path(self, path: Union[str, Path]) -> List[AnyModelConfig]:
@@ -282,7 +272,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         with self._db.lock:
             self._cursor.execute(
                 """--sql
-                SELECT config, strftime('%s',updated_at) FROM model_config
+                SELECT config, strftime('%s',updated_at) FROM models
                 WHERE path=?;
                 """,
                 (str(path),),
@@ -293,13 +283,13 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
         return results
 
     def search_by_hash(self, hash: str) -> List[AnyModelConfig]:
-        """Return models with the indicated original_hash."""
+        """Return models with the indicated hash."""
         results = []
         with self._db.lock:
             self._cursor.execute(
                 """--sql
-                SELECT config, strftime('%s',updated_at) FROM model_config
-                WHERE original_hash=?;
+                SELECT config, strftime('%s',updated_at) FROM models
+                WHERE hash=?;
                 """,
                 (hash,),
             )
@@ -308,83 +298,35 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
             ]
         return results
 
-    @property
-    def metadata_store(self) -> ModelMetadataStoreBase:
-        """Return a ModelMetadataStore initialized on the same database."""
-        return self._metadata_store
-
-    def get_metadata(self, key: str) -> Optional[AnyModelRepoMetadata]:
-        """
-        Retrieve metadata (if any) from when model was downloaded from a repo.
-
-        :param key: Model key
-        """
-        store = self.metadata_store
-        try:
-            metadata = store.get_metadata(key)
-            return metadata
-        except UnknownMetadataException:
-            return None
-
-    def search_by_metadata_tag(self, tags: Set[str]) -> List[AnyModelConfig]:
-        """
-        Search model metadata for ones with all listed tags and return their corresponding configs.
-
-        :param tags: Set of tags to search for. All tags must be present.
-        """
-        store = ModelMetadataStoreSQL(self._db)
-        keys = store.search_by_tag(tags)
-        return [self.get_model(x) for x in keys]
-
-    def list_tags(self) -> Set[str]:
-        """Return a unique set of all the model tags in the metadata database."""
-        store = ModelMetadataStoreSQL(self._db)
-        return store.list_tags()
-
-    def list_all_metadata(self) -> List[Tuple[str, AnyModelRepoMetadata]]:
-        """List metadata for all models that have it."""
-        store = ModelMetadataStoreSQL(self._db)
-        return store.list_all_metadata()
-
     def list_models(
         self, page: int = 0, per_page: int = 10, order_by: ModelRecordOrderBy = ModelRecordOrderBy.Default
     ) -> PaginatedResults[ModelSummary]:
         """Return a paginated summary listing of each model in the database."""
+        assert isinstance(order_by, ModelRecordOrderBy)
         ordering = {
-            ModelRecordOrderBy.Default: "a.type, a.base, a.format, a.name",
-            ModelRecordOrderBy.Type: "a.type",
-            ModelRecordOrderBy.Base: "a.base",
-            ModelRecordOrderBy.Name: "a.name",
-            ModelRecordOrderBy.Format: "a.format",
+            ModelRecordOrderBy.Default: "type, base, format, name",
+            ModelRecordOrderBy.Type: "type",
+            ModelRecordOrderBy.Base: "base",
+            ModelRecordOrderBy.Name: "name",
+            ModelRecordOrderBy.Format: "format",
         }
 
-        def _fixup(summary: Dict[str, str]) -> Dict[str, Union[str, int, Set[str]]]:
-            """Fix up results so that there are no null values."""
-            result: Dict[str, Union[str, int, Set[str]]] = {}
-            for key, item in summary.items():
-                result[key] = item or ""
-            result["tags"] = set(json.loads(summary["tags"] or "[]"))
-            return result
-
         # Lock so that the database isn't updated while we're doing the two queries.
         with self._db.lock:
             # query1: get the total number of model configs
             self._cursor.execute(
                 """--sql
-                select count(*) from model_config;
+                select count(*) from models;
                 """,
                 (),
             )
             total = int(self._cursor.fetchone()[0])
 
-            # query2: fetch key fields from the join of model_config and model_metadata
+            # query2: fetch key fields
             self._cursor.execute(
                 f"""--sql
-                SELECT a.id as key, a.type, a.base, a.format, a.name,
-                       json_extract(a.config, '$.description') as description,
-                       json_extract(b.metadata, '$.tags') as tags
-                FROM model_config AS a
-                LEFT JOIN model_metadata AS b on a.id=b.id
+                SELECT config
+                FROM models
                 ORDER BY {ordering[order_by]} -- using ? to bind doesn't work here for some reason
                 LIMIT ?
                 OFFSET ?;
@@ -395,7 +337,7 @@ class ModelRecordServiceSQL(ModelRecordServiceBase):
                 ),
             )
             rows = self._cursor.fetchall()
-            items = [ModelSummary.model_validate(_fixup(dict(x))) for x in rows]
+            items = [ModelSummary.model_validate(dict(x)) for x in rows]
             return PaginatedResults(
                 page=page, pages=ceil(total / per_page), per_page=per_page, total=total, items=items
             )

invokeai/app/services/session_processor/session_processor_base.py

@@ -1,6 +1,35 @@
 from abc import ABC, abstractmethod
+from threading import Event
 
+from invokeai.app.services.invocation_services import InvocationServices
 from invokeai.app.services.session_processor.session_processor_common import SessionProcessorStatus
+from invokeai.app.services.session_queue.session_queue_common import SessionQueueItem
+
+
+class SessionRunnerBase(ABC):
+    """
+    Base class for session runner.
+    """
+
+    @abstractmethod
+    def start(self, services: InvocationServices, cancel_event: Event) -> None:
+        """Starts the session runner"""
+        pass
+
+    @abstractmethod
+    def run(self, queue_item: SessionQueueItem) -> None:
+        """Runs the session"""
+        pass
+
+    @abstractmethod
+    def complete(self, queue_item: SessionQueueItem) -> None:
+        """Completes the session"""
+        pass
+
+    @abstractmethod
+    def run_node(self, node_id: str, queue_item: SessionQueueItem) -> None:
+        """Runs an already prepared node on the session"""
+        pass
 
 
 class SessionProcessorBase(ABC):

invokeai/app/services/session_processor/session_processor_default.py

@@ -2,13 +2,14 @@ import traceback
 from contextlib import suppress
 from threading import BoundedSemaphore, Thread
 from threading import Event as ThreadEvent
-from typing import Optional
+from typing import Callable, Optional, Union
 
 from fastapi_events.handlers.local import local_handler
 from fastapi_events.typing import Event as FastAPIEvent
 
 from invokeai.app.invocations.baseinvocation import BaseInvocation
 from invokeai.app.services.events.events_base import EventServiceBase
+from invokeai.app.services.invocation_services import InvocationServices
 from invokeai.app.services.invocation_stats.invocation_stats_common import GESStatsNotFoundError
 from invokeai.app.services.session_processor.session_processor_common import CanceledException
 from invokeai.app.services.session_queue.session_queue_common import SessionQueueItem
@@ -16,15 +17,164 @@ from invokeai.app.services.shared.invocation_context import InvocationContextDat
 from invokeai.app.util.profiler import Profiler
 
 from ..invoker import Invoker
-from .session_processor_base import SessionProcessorBase
+from .session_processor_base import SessionProcessorBase, SessionRunnerBase
 from .session_processor_common import SessionProcessorStatus
 
 
+class DefaultSessionRunner(SessionRunnerBase):
+    """Processes a single session's invocations"""
+
+    def __init__(
+        self,
+        on_before_run_node: Union[Callable[[BaseInvocation, SessionQueueItem], bool], None] = None,
+        on_after_run_node: Union[Callable[[BaseInvocation, SessionQueueItem], bool], None] = None,
+    ):
+        self.on_before_run_node = on_before_run_node
+        self.on_after_run_node = on_after_run_node
+
+    def start(self, services: InvocationServices, cancel_event: ThreadEvent):
+        """Start the session runner"""
+        self.services = services
+        self.cancel_event = cancel_event
+
+    def run(self, queue_item: SessionQueueItem):
+        """Run the graph"""
+        if not queue_item.session:
+            raise ValueError("Queue item has no session")
+        # Loop over invocations until the session is complete or canceled
+        while not (queue_item.session.is_complete() or self.cancel_event.is_set()):
+            # Prepare the next node
+            invocation = queue_item.session.next()
+            if invocation is None:
+                # If there are no more invocations, complete the graph
+                break
+            # Build invocation context (the node-facing API
+            self.run_node(invocation.id, queue_item)
+        self.complete(queue_item)
+
+    def complete(self, queue_item: SessionQueueItem):
+        """Complete the graph"""
+        self.services.events.emit_graph_execution_complete(
+            queue_batch_id=queue_item.batch_id,
+            queue_item_id=queue_item.item_id,
+            queue_id=queue_item.queue_id,
+            graph_execution_state_id=queue_item.session.id,
+        )
+
+    def _on_before_run_node(self, invocation: BaseInvocation, queue_item: SessionQueueItem):
+        """Run before a node is executed"""
+        # Send starting event
+        self.services.events.emit_invocation_started(
+            queue_batch_id=queue_item.batch_id,
+            queue_item_id=queue_item.item_id,
+            queue_id=queue_item.queue_id,
+            graph_execution_state_id=queue_item.session_id,
+            node=invocation.model_dump(),
+            source_node_id=queue_item.session.prepared_source_mapping[invocation.id],
+        )
+        if self.on_before_run_node is not None:
+            self.on_before_run_node(invocation, queue_item)
+
+    def _on_after_run_node(self, invocation: BaseInvocation, queue_item: SessionQueueItem):
+        """Run after a node is executed"""
+        if self.on_after_run_node is not None:
+            self.on_after_run_node(invocation, queue_item)
+
+    def run_node(self, node_id: str, queue_item: SessionQueueItem):
+        """Run a single node in the graph"""
+        # If this error raises a NodeNotFoundError that's handled by the processor
+        invocation = queue_item.session.execution_graph.get_node(node_id)
+        try:
+            self._on_before_run_node(invocation, queue_item)
+            data = InvocationContextData(
+                invocation=invocation,
+                source_invocation_id=queue_item.session.prepared_source_mapping[invocation.id],
+                queue_item=queue_item,
+            )
+
+            # Innermost processor try block; any unhandled exception is an invocation error & will fail the graph
+            with self.services.performance_statistics.collect_stats(invocation, queue_item.session_id):
+                context = build_invocation_context(
+                    data=data,
+                    services=self.services,
+                    cancel_event=self.cancel_event,
+                )
+
+                # Invoke the node
+                outputs = invocation.invoke_internal(context=context, services=self.services)
+
+                # Save outputs and history
+                queue_item.session.complete(invocation.id, outputs)
+
+            self._on_after_run_node(invocation, queue_item)
+            # Send complete event on successful runs
+            self.services.events.emit_invocation_complete(
+                queue_batch_id=queue_item.batch_id,
+                queue_item_id=queue_item.item_id,
+                queue_id=queue_item.queue_id,
+                graph_execution_state_id=queue_item.session.id,
+                node=invocation.model_dump(),
+                source_node_id=data.source_invocation_id,
+                result=outputs.model_dump(),
+            )
+        except KeyboardInterrupt:
+            # TODO(MM2): Create an event for this
+            pass
+        except CanceledException:
+            # When the user cancels the graph, we first set the cancel event. The event is checked
+            # between invocations, in this loop. Some invocations are long-running, and we need to
+            # be able to cancel them mid-execution.
+            #
+            # For example, denoising is a long-running invocation with many steps. A step callback
+            # is executed after each step. This step callback checks if the canceled event is set,
+            # then raises a CanceledException to stop execution immediately.
+            #
+            # When we get a CanceledException, we don't need to do anything - just pass and let the
+            # loop go to its next iteration, and the cancel event will be handled correctly.
+            pass
+        except Exception as e:
+            error = traceback.format_exc()
+
+            # Save error
+            queue_item.session.set_node_error(invocation.id, error)
+            self.services.logger.error(
+                f"Error while invoking session {queue_item.session_id}, invocation {invocation.id} ({invocation.get_type()}):\n{e}"
+            )
+            self.services.logger.error(error)
+
+            # Send error event
+            self.services.events.emit_invocation_error(
+                queue_batch_id=queue_item.session_id,
+                queue_item_id=queue_item.item_id,
+                queue_id=queue_item.queue_id,
+                graph_execution_state_id=queue_item.session.id,
+                node=invocation.model_dump(),
+                source_node_id=queue_item.session.prepared_source_mapping[invocation.id],
+                error_type=e.__class__.__name__,
+                error=error,
+            )
+
+
 class DefaultSessionProcessor(SessionProcessorBase):
-    def start(self, invoker: Invoker, thread_limit: int = 1, polling_interval: int = 1) -> None:
+    """Processes sessions from the session queue"""
+
+    def __init__(self, session_runner: Union[SessionRunnerBase, None] = None) -> None:
+        super().__init__()
+        self.session_runner = session_runner if session_runner else DefaultSessionRunner()
+
+    def start(
+        self,
+        invoker: Invoker,
+        thread_limit: int = 1,
+        polling_interval: int = 1,
+        on_before_run_session: Union[Callable[[SessionQueueItem], bool], None] = None,
+        on_after_run_session: Union[Callable[[SessionQueueItem], bool], None] = None,
+    ) -> None:
         self._invoker: Invoker = invoker
         self._queue_item: Optional[SessionQueueItem] = None
         self._invocation: Optional[BaseInvocation] = None
+        self.on_before_run_session = on_before_run_session
+        self.on_after_run_session = on_after_run_session
 
         self._resume_event = ThreadEvent()
         self._stop_event = ThreadEvent()
@@ -59,6 +209,7 @@ class DefaultSessionProcessor(SessionProcessorBase):
                 "cancel_event": self._cancel_event,
             },
         )
+        self.session_runner.start(services=invoker.services, cancel_event=self._cancel_event)
         self._thread.start()
 
     def stop(self, *args, **kwargs) -> None:
@@ -117,130 +268,34 @@ class DefaultSessionProcessor(SessionProcessorBase):
                         self._invoker.services.logger.debug(f"Executing queue item {self._queue_item.item_id}")
                         cancel_event.clear()
 
+                        # If we have a on_before_run_session callback, call it
+                        if self.on_before_run_session is not None:
+                            self.on_before_run_session(self._queue_item)
+
                         # If profiling is enabled, start the profiler
                         if self._profiler is not None:
                             self._profiler.start(profile_id=self._queue_item.session_id)
 
-                        # Prepare invocations and take the first
-                        self._invocation = self._queue_item.session.next()
+                        # Run the graph
+                        self.session_runner.run(queue_item=self._queue_item)
 
-                        # Loop over invocations until the session is complete or canceled
-                        while self._invocation is not None and not cancel_event.is_set():
-                            # get the source node id to provide to clients (the prepared node id is not as useful)
-                            source_invocation_id = self._queue_item.session.prepared_source_mapping[self._invocation.id]
-
-                            # Send starting event
-                            self._invoker.services.events.emit_invocation_started(
-                                queue_batch_id=self._queue_item.batch_id,
-                                queue_item_id=self._queue_item.item_id,
-                                queue_id=self._queue_item.queue_id,
-                                graph_execution_state_id=self._queue_item.session_id,
-                                node=self._invocation.model_dump(),
-                                source_node_id=source_invocation_id,
+                        # If we are profiling, stop the profiler and dump the profile & stats
+                        if self._profiler:
+                            profile_path = self._profiler.stop()
+                            stats_path = profile_path.with_suffix(".json")
+                            self._invoker.services.performance_statistics.dump_stats(
+                                graph_execution_state_id=self._queue_item.session.id, output_path=stats_path
                             )
 
-                            # Innermost processor try block; any unhandled exception is an invocation error & will fail the graph
-                            try:
-                                with self._invoker.services.performance_statistics.collect_stats(
-                                    self._invocation, self._queue_item.session.id
-                                ):
-                                    # Build invocation context (the node-facing API)
-                                    data = InvocationContextData(
-                                        invocation=self._invocation,
-                                        source_invocation_id=source_invocation_id,
-                                        queue_item=self._queue_item,
-                                    )
-                                    context = build_invocation_context(
-                                        data=data,
-                                        services=self._invoker.services,
-                                        cancel_event=self._cancel_event,
-                                    )
+                        # We'll get a GESStatsNotFoundError if we try to log stats for an untracked graph, but in the processor
+                        # we don't care about that - suppress the error.
+                        with suppress(GESStatsNotFoundError):
+                            self._invoker.services.performance_statistics.log_stats(self._queue_item.session.id)
+                            self._invoker.services.performance_statistics.reset_stats()
 
-                                    # Invoke the node
-                                    outputs = self._invocation.invoke_internal(
-                                        context=context, services=self._invoker.services
-                                    )
-
-                                    # Save outputs and history
-                                    self._queue_item.session.complete(self._invocation.id, outputs)
-
-                                    # Send complete event
-                                    self._invoker.services.events.emit_invocation_complete(
-                                        queue_batch_id=self._queue_item.batch_id,
-                                        queue_item_id=self._queue_item.item_id,
-                                        queue_id=self._queue_item.queue_id,
-                                        graph_execution_state_id=self._queue_item.session.id,
-                                        node=self._invocation.model_dump(),
-                                        source_node_id=source_invocation_id,
-                                        result=outputs.model_dump(),
-                                    )
-
-                            except KeyboardInterrupt:
-                                # TODO(MM2): Create an event for this
-                                pass
-
-                            except CanceledException:
-                                # When the user cancels the graph, we first set the cancel event. The event is checked
-                                # between invocations, in this loop. Some invocations are long-running, and we need to
-                                # be able to cancel them mid-execution.
-                                #
-                                # For example, denoising is a long-running invocation with many steps. A step callback
-                                # is executed after each step. This step callback checks if the canceled event is set,
-                                # then raises a CanceledException to stop execution immediately.
-                                #
-                                # When we get a CanceledException, we don't need to do anything - just pass and let the
-                                # loop go to its next iteration, and the cancel event will be handled correctly.
-                                pass
-
-                            except Exception as e:
-                                error = traceback.format_exc()
-
-                                # Save error
-                                self._queue_item.session.set_node_error(self._invocation.id, error)
-                                self._invoker.services.logger.error(
-                                    f"Error while invoking session {self._queue_item.session_id}, invocation {self._invocation.id} ({self._invocation.get_type()}):\n{e}"
-                                )
-
-                                # Send error event
-                                self._invoker.services.events.emit_invocation_error(
-                                    queue_batch_id=self._queue_item.session_id,
-                                    queue_item_id=self._queue_item.item_id,
-                                    queue_id=self._queue_item.queue_id,
-                                    graph_execution_state_id=self._queue_item.session.id,
-                                    node=self._invocation.model_dump(),
-                                    source_node_id=source_invocation_id,
-                                    error_type=e.__class__.__name__,
-                                    error=error,
-                                )
-                                pass
-
-                            # The session is complete if the all invocations are complete or there was an error
-                            if self._queue_item.session.is_complete() or cancel_event.is_set():
-                                # Send complete event
-                                self._invoker.services.events.emit_graph_execution_complete(
-                                    queue_batch_id=self._queue_item.batch_id,
-                                    queue_item_id=self._queue_item.item_id,
-                                    queue_id=self._queue_item.queue_id,
-                                    graph_execution_state_id=self._queue_item.session.id,
-                                )
-                                # If we are profiling, stop the profiler and dump the profile & stats
-                                if self._profiler:
-                                    profile_path = self._profiler.stop()
-                                    stats_path = profile_path.with_suffix(".json")
-                                    self._invoker.services.performance_statistics.dump_stats(
-                                        graph_execution_state_id=self._queue_item.session.id, output_path=stats_path
-                                    )
-                                # We'll get a GESStatsNotFoundError if we try to log stats for an untracked graph, but in the processor
-                                # we don't care about that - suppress the error.
-                                with suppress(GESStatsNotFoundError):
-                                    self._invoker.services.performance_statistics.log_stats(self._queue_item.session.id)
-                                    self._invoker.services.performance_statistics.reset_stats()
-
-                                # Set the invocation to None to prepare for the next session
-                                self._invocation = None
-                            else:
-                                # Prepare the next invocation
-                                self._invocation = self._queue_item.session.next()
+                        # If we have a on_after_run_session callback, call it
+                        if self.on_after_run_session is not None:
+                            self.on_after_run_session(self._queue_item)
 
                         # The session is complete, immediately poll for next session
                         self._queue_item = None
@@ -274,3 +329,4 @@ class DefaultSessionProcessor(SessionProcessorBase):
             poll_now_event.clear()
             self._queue_item = None
             self._thread_semaphore.release()
+            self._invoker.services.logger.debug("Session processor stopped")

invokeai/app/services/shared/invocation_context.py

@@ -65,75 +65,86 @@ class InvocationContextInterface:
 
 class BoardsInterface(InvocationContextInterface):
     def create(self, board_name: str) -> BoardDTO:
-        """
-        Creates a board.
+        """Creates a board.
 
-        :param board_name: The name of the board to create.
+        Args:
+            board_name: The name of the board to create.
+
+        Returns:
+            The created board DTO.
         """
         return self._services.boards.create(board_name)
 
     def get_dto(self, board_id: str) -> BoardDTO:
-        """
-        Gets a board DTO.
+        """Gets a board DTO.
 
-        :param board_id: The ID of the board to get.
+        Args:
+            board_id: The ID of the board to get.
+
+        Returns:
+            The board DTO.
         """
         return self._services.boards.get_dto(board_id)
 
     def get_all(self) -> list[BoardDTO]:
-        """
-        Gets all boards.
+        """Gets all boards.
+
+        Returns:
+            A list of all boards.
         """
         return self._services.boards.get_all()
 
     def add_image_to_board(self, board_id: str, image_name: str) -> None:
-        """
-        Adds an image to a board.
+        """Adds an image to a board.
 
-        :param board_id: The ID of the board to add the image to.
-        :param image_name: The name of the image to add to the board.
+        Args:
+            board_id: The ID of the board to add the image to.
+            image_name: The name of the image to add to the board.
         """
         return self._services.board_images.add_image_to_board(board_id, image_name)
 
     def get_all_image_names_for_board(self, board_id: str) -> list[str]:
-        """
-        Gets all image names for a board.
+        """Gets all image names for a board.
 
-        :param board_id: The ID of the board to get the image names for.
+        Args:
+            board_id: The ID of the board to get the image names for.
+
+        Returns:
+            A list of all image names for the board.
         """
         return self._services.board_images.get_all_board_image_names_for_board(board_id)
 
 
 class LoggerInterface(InvocationContextInterface):
     def debug(self, message: str) -> None:
-        """
-        Logs a debug message.
+        """Logs a debug message.
 
-        :param message: The message to log.
+        Args:
+            message: The message to log.
         """
         self._services.logger.debug(message)
 
     def info(self, message: str) -> None:
-        """
-        Logs an info message.
+        """Logs an info message.
 
-        :param message: The message to log.
+        Args:
+            message: The message to log.
         """
         self._services.logger.info(message)
 
     def warning(self, message: str) -> None:
-        """
-        Logs a warning message.
+        """Logs a warning message.
 
-        :param message: The message to log.
+        Args:
+            message: The message to log.
         """
         self._services.logger.warning(message)
 
     def error(self, message: str) -> None:
-        """
-        Logs an error message.
+        """Logs an error message.
 
-        :param message: The message to log.
+        Args:
+            message: The message to log.
         """
         self._services.logger.error(message)
 
@@ -146,20 +157,23 @@ class ImagesInterface(InvocationContextInterface):
         image_category: ImageCategory = ImageCategory.GENERAL,
         metadata: Optional[MetadataField] = None,
     ) -> ImageDTO:
-        """
-        Saves an image, returning its DTO.
+        """Saves an image, returning its DTO.
 
         If the current queue item has a workflow or metadata, it is automatically saved with the image.
 
-        :param image: The image to save, as a PIL image.
-        :param board_id: The board ID to add the image to, if it should be added. It the invocation \
+        Args:
+            image: The image to save, as a PIL image.
+            board_id: The board ID to add the image to, if it should be added. It the invocation \
             inherits from `WithBoard`, that board will be used automatically. **Use this only if \
             you want to override or provide a board manually!**
-        :param image_category: The category of the image. Only the GENERAL category is added \
+            image_category: The category of the image. Only the GENERAL category is added \
             to the gallery.
-        :param metadata: The metadata to save with the image, if it should have any. If the \
+            metadata: The metadata to save with the image, if it should have any. If the \
             invocation inherits from `WithMetadata`, that metadata will be used automatically. \
             **Use this only if you want to override or provide metadata manually!**
+
+        Returns:
+            The saved image DTO.
         """
 
         # If `metadata` is provided directly, use that. Else, use the metadata provided by `WithMetadata`, falling back to None.
@@ -189,11 +203,14 @@ class ImagesInterface(InvocationContextInterface):
         )
 
     def get_pil(self, image_name: str, mode: IMAGE_MODES | None = None) -> Image:
-        """
-        Gets an image as a PIL Image object.
+        """Gets an image as a PIL Image object.
 
-        :param image_name: The name of the image to get.
-        :param mode: The color mode to convert the image to. If None, the original mode is used.
+        Args:
+            image_name: The name of the image to get.
+            mode: The color mode to convert the image to. If None, the original mode is used.
+
+        Returns:
+            The image as a PIL Image object.
         """
         image = self._services.images.get_pil_image(image_name)
         if mode and mode != image.mode:
@@ -206,58 +223,76 @@ class ImagesInterface(InvocationContextInterface):
         return image
 
     def get_metadata(self, image_name: str) -> Optional[MetadataField]:
-        """
-        Gets an image's metadata, if it has any.
+        """Gets an image's metadata, if it has any.
 
-        :param image_name: The name of the image to get the metadata for.
+        Args:
+            image_name: The name of the image to get the metadata for.
+
+        Returns:
+            The image's metadata, if it has any.
         """
         return self._services.images.get_metadata(image_name)
 
     def get_dto(self, image_name: str) -> ImageDTO:
-        """
-        Gets an image as an ImageDTO object.
+        """Gets an image as an ImageDTO object.
 
-        :param image_name: The name of the image to get.
+        Args:
+            image_name: The name of the image to get.
+
+        Returns:
+            The image as an ImageDTO object.
         """
         return self._services.images.get_dto(image_name)
 
 
 class TensorsInterface(InvocationContextInterface):
     def save(self, tensor: Tensor) -> str:
-        """
-        Saves a tensor, returning its name.
+        """Saves a tensor, returning its name.
 
-        :param tensor: The tensor to save.
+        Args:
+            tensor: The tensor to save.
+
+        Returns:
+            The name of the saved tensor.
         """
 
         name = self._services.tensors.save(obj=tensor)
         return name
 
     def load(self, name: str) -> Tensor:
-        """
-        Loads a tensor by name.
+        """Loads a tensor by name.
 
-        :param name: The name of the tensor to load.
+        Args:
+            name: The name of the tensor to load.
+
+        Returns:
+            The loaded tensor.
         """
         return self._services.tensors.load(name)
 
 
 class ConditioningInterface(InvocationContextInterface):
     def save(self, conditioning_data: ConditioningFieldData) -> str:
-        """
-        Saves a conditioning data object, returning its name.
+        """Saves a conditioning data object, returning its name.
 
-        :param conditioning_data: The conditioning data to save.
+        Args:
+            conditioning_data: The conditioning data to save.
+
+        Returns:
+            The name of the saved conditioning data.
         """
 
         name = self._services.conditioning.save(obj=conditioning_data)
         return name
 
     def load(self, name: str) -> ConditioningFieldData:
-        """
-        Loads conditioning data by name.
+        """Loads conditioning data by name.
 
-        :param name: The name of the conditioning data to load.
+        Args:
+            name: The name of the conditioning data to load.
+
+        Returns:
+            The loaded conditioning data.
         """
 
         return self._services.conditioning.load(name)
@@ -265,20 +300,25 @@ class ConditioningInterface(InvocationContextInterface):
 
 class ModelsInterface(InvocationContextInterface):
     def exists(self, key: str) -> bool:
-        """
-        Checks if a model exists.
+        """Checks if a model exists.
 
-        :param key: The key of the model.
+        Args:
+            key: The key of the model.
+
+        Returns:
+            True if the model exists, False if not.
         """
         return self._services.model_manager.store.exists(key)
 
     def load(self, key: str, submodel_type: Optional[SubModelType] = None) -> LoadedModel:
-        """
-        Loads a model.
+        """Loads a model.
 
-        :param key: The key of the model.
-        :param submodel_type: The submodel of the model to get.
-        :returns: An object representing the loaded model.
+        Args:
+            key: The key of the model.
+            submodel_type: The submodel of the model to get.
+
+        Returns:
+            An object representing the loaded model.
         """
 
         # The model manager emits events as it loads the model. It needs the context data to build
@@ -289,75 +329,95 @@ class ModelsInterface(InvocationContextInterface):
         )
 
     def load_by_attrs(
-        self, model_name: str, base_model: BaseModelType, model_type: ModelType, submodel: Optional[SubModelType] = None
+        self, name: str, base: BaseModelType, type: ModelType, submodel_type: Optional[SubModelType] = None
     ) -> LoadedModel:
-        """
-        Loads a model by its attributes.
+        """Loads a model by its attributes.
 
-        :param model_name: Name of to be fetched.
-        :param base_model: Base model
-        :param model_type: Type of the model
-        :param submodel: For main (pipeline models), the submodel to fetch
+        Args:
+            name: Name of the model.
+            base: The models' base type, e.g. `BaseModelType.StableDiffusion1`, `BaseModelType.StableDiffusionXL`, etc.
+            type: Type of the model, e.g. `ModelType.Main`, `ModelType.Vae`, etc.
+            submodel_type: The type of submodel to load, e.g. `SubModelType.UNet`, `SubModelType.TextEncoder`, etc. Only main
+            models have submodels.
+
+        Returns:
+            An object representing the loaded model.
         """
         return self._services.model_manager.load_model_by_attr(
-            model_name=model_name,
-            base_model=base_model,
-            model_type=model_type,
-            submodel=submodel,
+            model_name=name,
+            base_model=base,
+            model_type=type,
+            submodel=submodel_type,
             context_data=self._data,
         )
 
     def get_config(self, key: str) -> AnyModelConfig:
-        """
-        Gets a model's info, an dict-like object.
+        """Gets a model's config.
 
-        :param key: The key of the model.
+        Args:
+            key: The key of the model.
+
+        Returns:
+            The model's config.
         """
         return self._services.model_manager.store.get_model(key=key)
 
     def get_metadata(self, key: str) -> Optional[AnyModelRepoMetadata]:
-        """
-        Gets a model's metadata, if it has any.
+        """Gets a model's metadata, if it has any.
 
-        :param key: The key of the model.
+        Args:
+            key: The key of the model.
+
+        Returns:
+            The model's metadata, if it has any.
         """
         return self._services.model_manager.store.get_metadata(key=key)
 
     def search_by_path(self, path: Path) -> list[AnyModelConfig]:
-        """
-        Searches for models by path.
+        """Searches for models by path.
 
-        :param path: The path to search for.
+        Args:
+            path: The path to search for.
+
+        Returns:
+            A list of models that match the path.
         """
         return self._services.model_manager.store.search_by_path(path)
 
     def search_by_attrs(
         self,
-        model_name: Optional[str] = None,
-        base_model: Optional[BaseModelType] = None,
-        model_type: Optional[ModelType] = None,
-        model_format: Optional[ModelFormat] = None,
+        name: Optional[str] = None,
+        base: Optional[BaseModelType] = None,
+        type: Optional[ModelType] = None,
+        format: Optional[ModelFormat] = None,
     ) -> list[AnyModelConfig]:
-        """
-        Searches for models by attributes.
+        """Searches for models by attributes.
 
-        :param model_name: Name of to be fetched.
-        :param base_model: Base model
-        :param model_type: Type of the model
-        :param submodel: For main (pipeline models), the submodel to fetch
+        Args:
+            name: The name to search for (exact match).
+            base: The base to search for, e.g. `BaseModelType.StableDiffusion1`, `BaseModelType.StableDiffusionXL`, etc.
+            type: Type type of model to search for, e.g. `ModelType.Main`, `ModelType.Vae`, etc.
+            format: The format of model to search for, e.g. `ModelFormat.Checkpoint`, `ModelFormat.Diffusers`, etc.
+
+        Returns:
+            A list of models that match the attributes.
         """
 
         return self._services.model_manager.store.search_by_attr(
-            model_name=model_name,
-            base_model=base_model,
-            model_type=model_type,
-            model_format=model_format,
+            model_name=name,
+            base_model=base,
+            model_type=type,
+            model_format=format,
         )
 
 
 class ConfigInterface(InvocationContextInterface):
     def get(self) -> InvokeAIAppConfig:
-        """Gets the app's config."""
+        """Gets the app's config.
+
+        Returns:
+            The app's config.
+        """
 
         return self._services.configuration.get_config()
 
@@ -370,7 +430,11 @@ class UtilInterface(InvocationContextInterface):
         self._cancel_event = cancel_event
 
     def is_canceled(self) -> bool:
-        """Checks if the current invocation has been canceled."""
+        """Checks if the current session has been canceled.
+
+        Returns:
+            True if the current session has been canceled, False if not.
+        """
         return self._cancel_event.is_set()
 
     def sd_step_callback(self, intermediate_state: PipelineIntermediateState, base_model: BaseModelType) -> None:
@@ -380,8 +444,9 @@ class UtilInterface(InvocationContextInterface):
 
         This should be called after each denoising step.
 
-        :param intermediate_state: The intermediate state of the diffusion pipeline.
-        :param base_model: The base model for the current denoising step.
+        Args:
+            intermediate_state: The intermediate state of the diffusion pipeline.
+            base_model: The base model for the current denoising step.
         """
 
         stable_diffusion_step_callback(
@@ -394,8 +459,17 @@ class UtilInterface(InvocationContextInterface):
 
 
 class InvocationContext:
-    """
-    The `InvocationContext` provides access to various services and data for the current invocation.
+    """Provides access to various services and data for the current invocation.
+
+    Attributes:
+        images (ImagesInterface): Methods to save, get and update images and their metadata.
+        tensors (TensorsInterface): Methods to save and get tensors, including image, noise, masks, and masked images.
+        conditioning (ConditioningInterface): Methods to save and get conditioning data.
+        models (ModelsInterface): Methods to check if a model exists, get a model, and get a model's info.
+        logger (LoggerInterface): The app logger.
+        config (ConfigInterface): The app config.
+        util (UtilInterface): Utility methods, including a method to check if an invocation was canceled and step callbacks.
+        boards (BoardsInterface): Methods to interact with boards.
     """
 
     def __init__(
@@ -438,11 +512,14 @@ def build_invocation_context(
     data: InvocationContextData,
     cancel_event: threading.Event,
 ) -> InvocationContext:
-    """
-    Builds the invocation context for a specific invocation execution.
+    """Builds the invocation context for a specific invocation execution.
 
-    :param services: The invocation services to wrap.
-    :param data: The invocation context data.
+    Args:
+        services: The invocation services to wrap.
+        data: The invocation context data.
+
+    Returns:
+        The invocation context.
     """
 
     logger = LoggerInterface(services=services, data=data)

invokeai/app/services/shared/sqlite/sqlite_util.py

@@ -9,6 +9,7 @@ from invokeai.app.services.shared.sqlite_migrator.migrations.migration_3 import
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_4 import build_migration_4
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_5 import build_migration_5
 from invokeai.app.services.shared.sqlite_migrator.migrations.migration_6 import build_migration_6
+from invokeai.app.services.shared.sqlite_migrator.migrations.migration_7 import build_migration_7
 from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_impl import SqliteMigrator
 
 
@@ -35,6 +36,7 @@ def init_db(config: InvokeAIAppConfig, logger: Logger, image_files: ImageFileSto
     migrator.register_migration(build_migration_4())
     migrator.register_migration(build_migration_5())
     migrator.register_migration(build_migration_6())
+    migrator.register_migration(build_migration_7())
     migrator.run_migrations()
 
     return db

invokeai/app/services/shared/sqlite_migrator/migrations/migration_7.py

@@ -0,0 +1,88 @@
+import sqlite3
+
+from invokeai.app.services.shared.sqlite_migrator.sqlite_migrator_common import Migration
+
+
+class Migration7Callback:
+    def __call__(self, cursor: sqlite3.Cursor) -> None:
+        self._create_models_table(cursor)
+        self._drop_old_models_tables(cursor)
+
+    def _drop_old_models_tables(self, cursor: sqlite3.Cursor) -> None:
+        """Drops the old model_records, model_metadata, model_tags and tags tables."""
+
+        tables = ["model_records", "model_metadata", "model_tags", "tags"]
+
+        for table in tables:
+            cursor.execute(f"DROP TABLE IF EXISTS {table};")
+
+    def _create_models_table(self, cursor: sqlite3.Cursor) -> None:
+        """Creates the v4.0.0 models table."""
+
+        tables = [
+            """--sql
+            CREATE TABLE IF NOT EXISTS models (
+                id TEXT NOT NULL PRIMARY KEY,
+                hash TEXT GENERATED ALWAYS as (json_extract(config, '$.hash')) VIRTUAL NOT NULL,
+                base TEXT GENERATED ALWAYS as (json_extract(config, '$.base')) VIRTUAL NOT NULL,
+                type TEXT GENERATED ALWAYS as (json_extract(config, '$.type')) VIRTUAL NOT NULL,
+                path TEXT GENERATED ALWAYS as (json_extract(config, '$.path')) VIRTUAL NOT NULL,
+                format TEXT GENERATED ALWAYS as (json_extract(config, '$.format')) VIRTUAL NOT NULL,
+                name TEXT GENERATED ALWAYS as (json_extract(config, '$.name')) VIRTUAL NOT NULL,
+                description TEXT GENERATED ALWAYS as (json_extract(config, '$.description')) VIRTUAL,
+                source TEXT GENERATED ALWAYS as (json_extract(config, '$.source')) VIRTUAL NOT NULL,
+                source_type TEXT GENERATED ALWAYS as (json_extract(config, '$.source_type')) VIRTUAL NOT NULL,
+                source_api_response TEXT GENERATED ALWAYS as (json_extract(config, '$.source_api_response')) VIRTUAL,
+                trigger_phrases TEXT GENERATED ALWAYS as (json_extract(config, '$.trigger_phrases')) VIRTUAL,
+                -- Serialized JSON representation of the whole config object, which will contain additional fields from subclasses
+                config TEXT NOT NULL,
+                created_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- Updated via trigger
+                updated_at DATETIME NOT NULL DEFAULT(STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')),
+                -- unique constraint on combo of name, base and type
+                UNIQUE(name, base, type)
+            );
+            """
+        ]
+
+        # Add trigger for `updated_at`.
+        triggers = [
+            """--sql
+            CREATE TRIGGER IF NOT EXISTS models_updated_at
+            AFTER UPDATE
+            ON models FOR EACH ROW
+            BEGIN
+                UPDATE models SET updated_at = STRFTIME('%Y-%m-%d %H:%M:%f', 'NOW')
+                    WHERE id = old.id;
+            END;
+            """
+        ]
+
+        # Add indexes for searchable fields
+        indices = [
+            "CREATE INDEX IF NOT EXISTS base_index ON models(base);",
+            "CREATE INDEX IF NOT EXISTS type_index ON models(type);",
+            "CREATE INDEX IF NOT EXISTS name_index ON models(name);",
+            "CREATE UNIQUE INDEX IF NOT EXISTS path_index ON models(path);",
+        ]
+
+        for stmt in tables + indices + triggers:
+            cursor.execute(stmt)
+
+
+def build_migration_7() -> Migration:
+    """
+    Build the migration from database version 6 to 7.
+
+    This migration does the following:
+    - Adds the new models table
+    - Drops the old model_records, model_metadata, model_tags and tags tables.
+    - TODO(MM2): Migrates model names and descriptions from `models.yaml` to the new table (?).
+    """
+    migration_7 = Migration(
+        from_version=6,
+        to_version=7,
+        callback=Migration7Callback(),
+    )
+
+    return migration_7

invokeai/app/services/shared/sqlite_migrator/migrations/util/migrate_yaml_config_1.py

@@ -3,7 +3,6 @@
 
 import json
 import sqlite3
-from hashlib import sha1
 from logging import Logger
 from pathlib import Path
 from typing import Optional
@@ -22,7 +21,7 @@ from invokeai.backend.model_manager.config import (
     ModelConfigFactory,
     ModelType,
 )
-from invokeai.backend.model_manager.hash import FastModelHash
+from invokeai.backend.model_manager.hash import ModelHash
 
 ModelsValidator = TypeAdapter(AnyModelConfig)
 
@@ -73,19 +72,27 @@ class MigrateModelYamlToDb1:
 
             base_type, model_type, model_name = str(model_key).split("/")
             try:
-                hash = FastModelHash.hash(self.config.models_path / stanza.path)
+                hash = ModelHash().hash(self.config.models_path / stanza.path)
             except OSError:
                 self.logger.warning(f"The model at {stanza.path} is not a valid file or directory. Skipping migration.")
                 continue
 
-            assert isinstance(model_key, str)
-            new_key = sha1(model_key.encode("utf-8")).hexdigest()
-
             stanza["base"] = BaseModelType(base_type)
             stanza["type"] = ModelType(model_type)
             stanza["name"] = model_name
             stanza["original_hash"] = hash
             stanza["current_hash"] = hash
+            new_key = hash  # deterministic key assignment
+
+            # special case for ip adapters, which need the new `image_encoder_model_id` field
+            if stanza["type"] == ModelType.IPAdapter:
+                try:
+                    stanza["image_encoder_model_id"] = self._get_image_encoder_model_id(
+                        self.config.models_path / stanza.path
+                    )
+                except OSError:
+                    self.logger.warning(f"Could not determine image encoder for {stanza.path}. Skipping.")
+                    continue
 
             new_config: AnyModelConfig = ModelsValidator.validate_python(stanza)  # type: ignore # see https://github.com/pydantic/pydantic/discussions/7094
 
@@ -95,7 +102,7 @@ class MigrateModelYamlToDb1:
                     self.logger.info(f"Updating model {model_name} with information from models.yaml using key {key}")
                     self._update_model(key, new_config)
                 else:
-                    self.logger.info(f"Adding model {model_name} with key {model_key}")
+                    self.logger.info(f"Adding model {model_name} with key {new_key}")
                     self._add_model(new_key, new_config)
             except DuplicateModelException:
                 self.logger.warning(f"Model {model_name} is already in the database")
@@ -143,9 +150,14 @@ class MigrateModelYamlToDb1:
                 """,
                 (
                     key,
-                    record.original_hash,
+                    record.hash,
                     json_serialized,
                 ),
             )
         except sqlite3.IntegrityError as exc:
             raise DuplicateModelException(f"{record.name}: model is already in database") from exc
+
+    def _get_image_encoder_model_id(self, model_path: Path) -> str:
+        with open(model_path / "image_encoder.txt") as f:
+            encoder = f.read()
+        return encoder.strip()

invokeai/app/util/metadata.py

@@ -1,55 +0,0 @@
-import json
-from typing import Optional
-
-from pydantic import ValidationError
-
-from invokeai.app.services.shared.graph import Edge
-
-
-def get_metadata_graph_from_raw_session(session_raw: str) -> Optional[dict]:
-    """
-    Parses raw session string, returning a dict of the graph.
-
-    Only the general graph shape is validated; none of the fields are validated.
-
-    Any `metadata_accumulator` nodes and edges are removed.
-
-    Any validation failure will return None.
-    """
-
-    graph = json.loads(session_raw).get("graph", None)
-
-    # sanity check make sure the graph is at least reasonably shaped
-    if (
-        not isinstance(graph, dict)
-        or "nodes" not in graph
-        or not isinstance(graph["nodes"], dict)
-        or "edges" not in graph
-        or not isinstance(graph["edges"], list)
-    ):
-        # something has gone terribly awry, return an empty dict
-        return None
-
-    try:
-        # delete the `metadata_accumulator` node
-        del graph["nodes"]["metadata_accumulator"]
-    except KeyError:
-        # no accumulator node, all good
-        pass
-
-    # delete any edges to or from it
-    for i, edge in enumerate(graph["edges"]):
-        try:
-            # try to parse the edge
-            Edge(**edge)
-        except ValidationError:
-            # something has gone terribly awry, return an empty dict
-            return None
-
-        if (
-            edge["source"]["node_id"] == "metadata_accumulator"
-            or edge["destination"]["node_id"] == "metadata_accumulator"
-        ):
-            del graph["edges"][i]
-
-    return graph

invokeai/app/util/ti_utils.py

@@ -1,8 +1,47 @@
 import re
+from typing import List, Tuple
+
+import invokeai.backend.util.logging as logger
+from invokeai.app.services.model_records import UnknownModelException
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.model_manager.config import BaseModelType, ModelType
+from invokeai.backend.textual_inversion import TextualInversionModelRaw
 
 
-def extract_ti_triggers_from_prompt(prompt: str) -> list[str]:
-    ti_triggers = []
+def extract_ti_triggers_from_prompt(prompt: str) -> List[str]:
+    ti_triggers: List[str] = []
     for trigger in re.findall(r"<[a-zA-Z0-9., _-]+>", prompt):
-        ti_triggers.append(trigger)
+        ti_triggers.append(str(trigger))
     return ti_triggers
+
+
+def generate_ti_list(
+    prompt: str, base: BaseModelType, context: InvocationContext
+) -> List[Tuple[str, TextualInversionModelRaw]]:
+    ti_list: List[Tuple[str, TextualInversionModelRaw]] = []
+    for trigger in extract_ti_triggers_from_prompt(prompt):
+        name_or_key = trigger[1:-1]
+        try:
+            loaded_model = context.models.load(key=name_or_key)
+            model = loaded_model.model
+            assert isinstance(model, TextualInversionModelRaw)
+            assert loaded_model.config.base == base
+            ti_list.append((name_or_key, model))
+        except UnknownModelException:
+            try:
+                loaded_model = context.models.load_by_attrs(
+                    name=name_or_key, base=base, type=ModelType.TextualInversion
+                )
+                model = loaded_model.model
+                assert isinstance(model, TextualInversionModelRaw)
+                assert loaded_model.config.base == base
+                ti_list.append((name_or_key, model))
+            except UnknownModelException:
+                pass
+        except ValueError:
+            logger.warning(f'trigger: "{trigger}" more than one similarly-named textual inversion models')
+        except AssertionError:
+            logger.warning(f'trigger: "{trigger}" not a valid textual inversion model for this graph')
+        except Exception:
+            logger.warning(f'Failed to load TI model for trigger: "{trigger}"')
+    return ti_list

invokeai/backend/image_util/__init__.py

@@ -1,6 +1,7 @@
 """
 Initialization file for invokeai.backend.image_util methods.
 """
+
 from .patchmatch import PatchMatch  # noqa: F401
 from .pngwriter import PngWriter, PromptFormatter, retrieve_metadata, write_metadata  # noqa: F401
 from .seamless import configure_model_padding  # noqa: F401

invokeai/backend/image_util/invisible_watermark.py

@@ -3,6 +3,7 @@ This module defines a singleton object, "invisible_watermark" that
 wraps the invisible watermark model. It respects the global "invisible_watermark"
 configuration variable, that allows the watermarking to be supressed.
 """
+
 import cv2
 import numpy as np
 from imwatermark import WatermarkEncoder

invokeai/backend/image_util/patchmatch.py

@@ -4,6 +4,7 @@ wraps the actual patchmatch object. It respects the global
 "try_patchmatch" attribute, so that patchmatch loading can
 be suppressed or deferred
 """
+
 import numpy as np
 
 import invokeai.backend.util.logging as logger

invokeai/backend/image_util/pngwriter.py

@@ -6,6 +6,7 @@ PngWriter -- Converts Images generated by T2I into PNGs, finds
 
 Exports function retrieve_metadata(path)
 """
+
 import json
 import os
 import re

invokeai/backend/image_util/safety_checker.py

@@ -3,6 +3,7 @@ This module defines a singleton object, "safety_checker" that
 wraps the safety_checker model. It respects the global "nsfw_checker"
 configuration variable, that allows the checker to be supressed.
 """
+
 import numpy as np
 from PIL import Image
 

invokeai/backend/install/check_root.py

@@ -1,6 +1,7 @@
 """
 Check that the invokeai_root is correctly configured and exit if not.
 """
+
 import sys
 
 from invokeai.app.services.config import InvokeAIAppConfig

invokeai/backend/install/install_helper.py

@@ -1,4 +1,5 @@
 """Utility (backend) functions used by model_install.py"""
+
 from logging import Logger
 from pathlib import Path
 from typing import Any, Dict, List, Optional

invokeai/backend/model_manager/__init__.py

@@ -1,4 +1,5 @@
 """Re-export frequently-used symbols from the Model Manager backend."""
+
 from .config import (
     AnyModel,
     AnyModelConfig,

invokeai/backend/model_manager/config.py

@@ -19,15 +19,19 @@ Typical usage:
 Validation errors will raise an InvalidModelConfigException error.
 
 """
+
 import time
 from enum import Enum
 from typing import Literal, Optional, Type, Union
 
 import torch
-from diffusers import ModelMixin
-from pydantic import BaseModel, ConfigDict, Field, TypeAdapter
+from diffusers.models.modeling_utils import ModelMixin
+from pydantic import BaseModel, ConfigDict, Discriminator, Field, Tag, TypeAdapter
 from typing_extensions import Annotated, Any, Dict
 
+from invokeai.app.invocations.constants import SCHEDULER_NAME_VALUES
+from invokeai.app.util.misc import uuid_string
+
 from ..raw_model import RawModel
 
 # ModelMixin is the base class for all diffusers and transformers models
@@ -55,8 +59,8 @@ class ModelType(str, Enum):
 
     ONNX = "onnx"
     Main = "main"
-    Vae = "vae"
-    Lora = "lora"
+    VAE = "vae"
+    LoRA = "lora"
     ControlNet = "controlnet"  # used by model_probe
     TextualInversion = "embedding"
     IPAdapter = "ip_adapter"
@@ -72,9 +76,9 @@ class SubModelType(str, Enum):
     TextEncoder2 = "text_encoder_2"
     Tokenizer = "tokenizer"
     Tokenizer2 = "tokenizer_2"
-    Vae = "vae"
-    VaeDecoder = "vae_decoder"
-    VaeEncoder = "vae_encoder"
+    VAE = "vae"
+    VAEDecoder = "vae_decoder"
+    VAEEncoder = "vae_encoder"
     Scheduler = "scheduler"
     SafetyChecker = "safety_checker"
 
@@ -92,8 +96,8 @@ class ModelFormat(str, Enum):
 
     Diffusers = "diffusers"
     Checkpoint = "checkpoint"
-    Lycoris = "lycoris"
-    Onnx = "onnx"
+    LyCORIS = "lycoris"
+    ONNX = "onnx"
     Olive = "olive"
     EmbeddingFile = "embedding_file"
     EmbeddingFolder = "embedding_folder"
@@ -111,128 +115,187 @@ class SchedulerPredictionType(str, Enum):
 class ModelRepoVariant(str, Enum):
     """Various hugging face variants on the diffusers format."""
 
-    DEFAULT = ""  # model files without "fp16" or other qualifier - empty str
+    Default = ""  # model files without "fp16" or other qualifier - empty str
     FP16 = "fp16"
     FP32 = "fp32"
     ONNX = "onnx"
-    OPENVINO = "openvino"
-    FLAX = "flax"
+    OpenVINO = "openvino"
+    Flax = "flax"
+
+
+class ModelSourceType(str, Enum):
+    """Model source type."""
+
+    Path = "path"
+    Url = "url"
+    HFRepoID = "hf_repo_id"
+    CivitAI = "civitai"
+
+
+class ModelDefaultSettings(BaseModel):
+    vae: str | None
+    vae_precision: str | None
+    scheduler: SCHEDULER_NAME_VALUES | None
+    steps: int | None
+    cfg_scale: float | None
+    cfg_rescale_multiplier: float | None
 
 
 class ModelConfigBase(BaseModel):
     """Base class for model configuration information."""
 
-    path: str = Field(description="filesystem path to the model file or directory")
-    name: str = Field(description="model name")
-    base: BaseModelType = Field(description="base model")
-    type: ModelType = Field(description="type of the model")
-    format: ModelFormat = Field(description="model format")
-    key: str = Field(description="unique key for model", default="<NOKEY>")
-    original_hash: Optional[str] = Field(
-        description="original fasthash of model contents", default=None
-    )  # this is assigned at install time and will not change
-    current_hash: Optional[str] = Field(
-        description="current fasthash of model contents", default=None
-    )  # if model is converted or otherwise modified, this will hold updated hash
-    description: Optional[str] = Field(description="human readable description of the model", default=None)
-    source: Optional[str] = Field(description="model original source (path, URL or repo_id)", default=None)
-    last_modified: Optional[float] = Field(description="timestamp for modification time", default_factory=time.time)
+    key: str = Field(description="A unique key for this model.", default_factory=uuid_string)
+    hash: str = Field(description="The hash of the model file(s).")
+    path: str = Field(
+        description="Path to the model on the filesystem. Relative paths are relative to the Invoke root directory."
+    )
+    name: str = Field(description="Name of the model.")
+    base: BaseModelType = Field(description="The base model.")
+    description: Optional[str] = Field(description="Model description", default=None)
+    source: str = Field(description="The original source of the model (path, URL or repo_id).")
+    source_type: ModelSourceType = Field(description="The type of source")
+    source_api_response: Optional[str] = Field(
+        description="The original API response from the source, as stringified JSON.", default=None
+    )
+    trigger_phrases: Optional[set[str]] = Field(description="Set of trigger phrases for this model", default=None)
+    default_settings: Optional[ModelDefaultSettings] = Field(
+        description="Default settings for this model", default=None
+    )
 
     @staticmethod
     def json_schema_extra(schema: dict[str, Any], model_class: Type[BaseModel]) -> None:
-        schema["required"].extend(
-            ["key", "base", "type", "format", "original_hash", "current_hash", "source", "last_modified"]
-        )
+        schema["required"].extend(["key", "type", "format"])
 
-    model_config = ConfigDict(
-        use_enum_values=False,
-        validate_assignment=True,
-        json_schema_extra=json_schema_extra,
-    )
-
-    def update(self, attributes: Dict[str, Any]) -> None:
-        """Update the object with fields in dict."""
-        for key, value in attributes.items():
-            setattr(self, key, value)  # may raise a validation error
+    model_config = ConfigDict(validate_assignment=True, json_schema_extra=json_schema_extra)
 
 
-class _CheckpointConfig(ModelConfigBase):
+class CheckpointConfigBase(ModelConfigBase):
     """Model config for checkpoint-style models."""
 
     format: Literal[ModelFormat.Checkpoint] = ModelFormat.Checkpoint
-    config: str = Field(description="path to the checkpoint model config file")
+    config_path: str = Field(description="path to the checkpoint model config file")
+    converted_at: Optional[float] = Field(
+        description="When this model was last converted to diffusers", default_factory=time.time
+    )
 
 
-class _DiffusersConfig(ModelConfigBase):
+class DiffusersConfigBase(ModelConfigBase):
     """Model config for diffusers-style models."""
 
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
-    repo_variant: Optional[ModelRepoVariant] = ModelRepoVariant.DEFAULT
+    repo_variant: Optional[ModelRepoVariant] = ModelRepoVariant.Default
 
 
-class LoRAConfig(ModelConfigBase):
+class LoRALyCORISConfig(ModelConfigBase):
     """Model config for LoRA/Lycoris models."""
 
-    type: Literal[ModelType.Lora] = ModelType.Lora
-    format: Literal[ModelFormat.Lycoris, ModelFormat.Diffusers]
+    type: Literal[ModelType.LoRA] = ModelType.LoRA
+    format: Literal[ModelFormat.LyCORIS] = ModelFormat.LyCORIS
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.LoRA.value}.{ModelFormat.LyCORIS.value}")
 
 
-class VaeCheckpointConfig(ModelConfigBase):
+class LoRADiffusersConfig(ModelConfigBase):
+    """Model config for LoRA/Diffusers models."""
+
+    type: Literal[ModelType.LoRA] = ModelType.LoRA
+    format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.LoRA.value}.{ModelFormat.Diffusers.value}")
+
+
+class VAECheckpointConfig(CheckpointConfigBase):
     """Model config for standalone VAE models."""
 
-    type: Literal[ModelType.Vae] = ModelType.Vae
+    type: Literal[ModelType.VAE] = ModelType.VAE
     format: Literal[ModelFormat.Checkpoint] = ModelFormat.Checkpoint
 
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.VAE.value}.{ModelFormat.Checkpoint.value}")
 
-class VaeDiffusersConfig(ModelConfigBase):
+
+class VAEDiffusersConfig(ModelConfigBase):
     """Model config for standalone VAE models (diffusers version)."""
 
-    type: Literal[ModelType.Vae] = ModelType.Vae
+    type: Literal[ModelType.VAE] = ModelType.VAE
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.VAE.value}.{ModelFormat.Diffusers.value}")
 
-class ControlNetDiffusersConfig(_DiffusersConfig):
+
+class ControlNetDiffusersConfig(DiffusersConfigBase):
     """Model config for ControlNet models (diffusers version)."""
 
     type: Literal[ModelType.ControlNet] = ModelType.ControlNet
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
 
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.ControlNet.value}.{ModelFormat.Diffusers.value}")
 
-class ControlNetCheckpointConfig(_CheckpointConfig):
+
+class ControlNetCheckpointConfig(CheckpointConfigBase):
     """Model config for ControlNet models (diffusers version)."""
 
     type: Literal[ModelType.ControlNet] = ModelType.ControlNet
     format: Literal[ModelFormat.Checkpoint] = ModelFormat.Checkpoint
 
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.ControlNet.value}.{ModelFormat.Checkpoint.value}")
 
-class TextualInversionConfig(ModelConfigBase):
+
+class TextualInversionFileConfig(ModelConfigBase):
     """Model config for textual inversion embeddings."""
 
     type: Literal[ModelType.TextualInversion] = ModelType.TextualInversion
-    format: Literal[ModelFormat.EmbeddingFile, ModelFormat.EmbeddingFolder]
+    format: Literal[ModelFormat.EmbeddingFile] = ModelFormat.EmbeddingFile
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.TextualInversion.value}.{ModelFormat.EmbeddingFile.value}")
 
 
-class _MainConfig(ModelConfigBase):
-    """Model config for main models."""
+class TextualInversionFolderConfig(ModelConfigBase):
+    """Model config for textual inversion embeddings."""
 
-    vae: Optional[str] = Field(default=None)
-    variant: ModelVariantType = ModelVariantType.Normal
-    prediction_type: SchedulerPredictionType = SchedulerPredictionType.Epsilon
-    upcast_attention: bool = False
-    ztsnr_training: bool = False
+    type: Literal[ModelType.TextualInversion] = ModelType.TextualInversion
+    format: Literal[ModelFormat.EmbeddingFolder] = ModelFormat.EmbeddingFolder
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.TextualInversion.value}.{ModelFormat.EmbeddingFolder.value}")
 
 
-class MainCheckpointConfig(_CheckpointConfig, _MainConfig):
+class MainCheckpointConfig(CheckpointConfigBase):
     """Model config for main checkpoint models."""
 
     type: Literal[ModelType.Main] = ModelType.Main
+    variant: ModelVariantType = ModelVariantType.Normal
+    prediction_type: SchedulerPredictionType = SchedulerPredictionType.Epsilon
+    upcast_attention: bool = False
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.Main.value}.{ModelFormat.Checkpoint.value}")
 
 
-class MainDiffusersConfig(_DiffusersConfig, _MainConfig):
+class MainDiffusersConfig(DiffusersConfigBase):
     """Model config for main diffusers models."""
 
     type: Literal[ModelType.Main] = ModelType.Main
 
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.Main.value}.{ModelFormat.Diffusers.value}")
+
 
 class IPAdapterConfig(ModelConfigBase):
     """Model config for IP Adaptor format models."""
@@ -241,6 +304,10 @@ class IPAdapterConfig(ModelConfigBase):
     image_encoder_model_id: str
     format: Literal[ModelFormat.InvokeAI]
 
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.IPAdapter.value}.{ModelFormat.InvokeAI.value}")
+
 
 class CLIPVisionDiffusersConfig(ModelConfigBase):
     """Model config for ClipVision."""
@@ -248,58 +315,65 @@ class CLIPVisionDiffusersConfig(ModelConfigBase):
     type: Literal[ModelType.CLIPVision] = ModelType.CLIPVision
     format: Literal[ModelFormat.Diffusers]
 
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.CLIPVision.value}.{ModelFormat.Diffusers.value}")
 
-class T2IConfig(ModelConfigBase):
+
+class T2IAdapterConfig(ModelConfigBase):
     """Model config for T2I."""
 
     type: Literal[ModelType.T2IAdapter] = ModelType.T2IAdapter
     format: Literal[ModelFormat.Diffusers]
 
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.T2IAdapter.value}.{ModelFormat.Diffusers.value}")
 
-_ControlNetConfig = Annotated[
-    Union[ControlNetDiffusersConfig, ControlNetCheckpointConfig],
-    Field(discriminator="format"),
-]
-_VaeConfig = Annotated[Union[VaeDiffusersConfig, VaeCheckpointConfig], Field(discriminator="format")]
-_MainModelConfig = Annotated[Union[MainDiffusersConfig, MainCheckpointConfig], Field(discriminator="format")]
 
-AnyModelConfig = Union[
-    _MainModelConfig,
-    _VaeConfig,
-    _ControlNetConfig,
-    # ModelConfigBase,
-    LoRAConfig,
-    TextualInversionConfig,
-    IPAdapterConfig,
-    CLIPVisionDiffusersConfig,
-    T2IConfig,
+def get_model_discriminator_value(v: Any) -> str:
+    """
+    Computes the discriminator value for a model config.
+    https://docs.pydantic.dev/latest/concepts/unions/#discriminated-unions-with-callable-discriminator
+    """
+    format_ = None
+    type_ = None
+    if isinstance(v, dict):
+        format_ = v.get("format")
+        if isinstance(format_, Enum):
+            format_ = format_.value
+        type_ = v.get("type")
+        if isinstance(type_, Enum):
+            type_ = type_.value
+    else:
+        format_ = v.format.value
+        type_ = v.type.value
+    v = f"{type_}.{format_}"
+    return v
+
+
+AnyModelConfig = Annotated[
+    Union[
+        Annotated[MainDiffusersConfig, MainDiffusersConfig.get_tag()],
+        Annotated[MainCheckpointConfig, MainCheckpointConfig.get_tag()],
+        Annotated[VAEDiffusersConfig, VAEDiffusersConfig.get_tag()],
+        Annotated[VAECheckpointConfig, VAECheckpointConfig.get_tag()],
+        Annotated[ControlNetDiffusersConfig, ControlNetDiffusersConfig.get_tag()],
+        Annotated[ControlNetCheckpointConfig, ControlNetCheckpointConfig.get_tag()],
+        Annotated[LoRALyCORISConfig, LoRALyCORISConfig.get_tag()],
+        Annotated[LoRADiffusersConfig, LoRADiffusersConfig.get_tag()],
+        Annotated[TextualInversionFileConfig, TextualInversionFileConfig.get_tag()],
+        Annotated[TextualInversionFolderConfig, TextualInversionFolderConfig.get_tag()],
+        Annotated[IPAdapterConfig, IPAdapterConfig.get_tag()],
+        Annotated[T2IAdapterConfig, T2IAdapterConfig.get_tag()],
+        Annotated[CLIPVisionDiffusersConfig, CLIPVisionDiffusersConfig.get_tag()],
+    ],
+    Discriminator(get_model_discriminator_value),
 ]
 
 AnyModelConfigValidator = TypeAdapter(AnyModelConfig)
 
 
-# IMPLEMENTATION NOTE:
-# The preferred alternative to the above is a discriminated Union as shown
-# below. However, it breaks FastAPI when used as the input Body parameter in a route.
-# This is a known issue. Please see:
-#   https://github.com/tiangolo/fastapi/discussions/9761 and
-#   https://github.com/tiangolo/fastapi/discussions/9287
-# AnyModelConfig = Annotated[
-#     Union[
-#         _MainModelConfig,
-#         _ONNXConfig,
-#         _VaeConfig,
-#         _ControlNetConfig,
-#         LoRAConfig,
-#         TextualInversionConfig,
-#         IPAdapterConfig,
-#         CLIPVisionDiffusersConfig,
-#         T2IConfig,
-#     ],
-#     Field(discriminator="type"),
-# ]
-
-
 class ModelConfigFactory(object):
     """Class for parsing config dicts into StableDiffusion Config obects."""
 
@@ -331,6 +405,6 @@ class ModelConfigFactory(object):
         assert model is not None
         if key:
             model.key = key
-        if timestamp:
-            model.last_modified = timestamp
+        if isinstance(model, CheckpointConfigBase) and timestamp is not None:
+            model.converted_at = timestamp
         return model  # type: ignore

invokeai/backend/model_manager/convert_ckpt_to_diffusers.py

@@ -15,7 +15,7 @@
 #
 # Adapted for use in InvokeAI by Lincoln Stein, July 2023
 #
-""" Conversion script for the Stable Diffusion checkpoints."""
+"""Conversion script for the Stable Diffusion checkpoints."""
 
 import re
 from contextlib import nullcontext

invokeai/backend/model_manager/hash.py

@@ -11,56 +11,175 @@ from invokeai.backend.model_managre.model_hash import FastModelHash
 import hashlib
 import os
 from pathlib import Path
-from typing import Dict, Union
+from typing import Callable, Literal, Optional, Union
 
-from imohash import hashfile
+from blake3 import blake3
+
+MODEL_FILE_EXTENSIONS = (".ckpt", ".safetensors", ".bin", ".pt", ".pth")
+
+ALGORITHM = Literal[
+    "md5",
+    "sha1",
+    "sha224",
+    "sha256",
+    "sha384",
+    "sha512",
+    "blake2b",
+    "blake2s",
+    "sha3_224",
+    "sha3_256",
+    "sha3_384",
+    "sha3_512",
+    "shake_128",
+    "shake_256",
+    "blake3",
+]
 
 
-class FastModelHash(object):
-    """FastModelHash obect provides one public class method, hash()."""
+class ModelHash:
+    """
+    Creates a hash of a model using a specified algorithm.
 
-    @classmethod
-    def hash(cls, model_location: Union[str, Path]) -> str:
-        """
-        Return hexdigest string for model located at model_location.
+    Args:
+        algorithm: Hashing algorithm to use. Defaults to BLAKE3.
+        file_filter: A function that takes a file name and returns True if the file should be included in the hash.
 
-        :param model_location: Path to the model
-        """
-        model_location = Path(model_location)
-        if model_location.is_file():
-            return cls._hash_file(model_location)
-        elif model_location.is_dir():
-            return cls._hash_dir(model_location)
+    If the model is a single file, it is hashed directly using the provided algorithm.
+
+    If the model is a directory, each model weights file in the directory is hashed using the provided algorithm.
+
+    Only files with the following extensions are hashed: .ckpt, .safetensors, .bin, .pt, .pth
+
+    The final hash is computed by hashing the hashes of all model files in the directory using BLAKE3, ensuring
+    that directory hashes are never weaker than the file hashes.
+
+    Usage:
+        ```py
+        # BLAKE3 hash
+        ModelHash().hash("path/to/some/model.safetensors")
+        # MD5
+        ModelHash("md5").hash("path/to/model/dir/")
+        ```
+    """
+
+    def __init__(self, algorithm: ALGORITHM = "blake3", file_filter: Optional[Callable[[str], bool]] = None) -> None:
+        if algorithm == "blake3":
+            self._hash_file = self._blake3
+        elif algorithm in hashlib.algorithms_available:
+            self._hash_file = self._get_hashlib(algorithm)
         else:
-            raise OSError(f"Not a valid file or directory: {model_location}")
+            raise ValueError(f"Algorithm {algorithm} not available")
 
-    @classmethod
-    def _hash_file(cls, model_location: Union[str, Path]) -> str:
+        self._file_filter = file_filter or self._default_file_filter
+
+    def hash(self, model_path: Union[str, Path]) -> str:
         """
-        Fasthash a single file and return its hexdigest.
+        Return hexdigest of hash of model located at model_path using the algorithm provided at class instantiation.
 
-        :param model_location: Path to the model file
+        If model_path is a directory, the hash is computed by hashing the hashes of all model files in the
+        directory. The final composite hash is always computed using BLAKE3.
+
+        Args:
+            model_path: Path to the model
+
+        Returns:
+            str: Hexdigest of the hash of the model
         """
-        # we return md5 hash of the filehash to make it shorter
-        # cryptographic security not needed here
-        return hashlib.md5(hashfile(model_location)).hexdigest()
 
-    @classmethod
-    def _hash_dir(cls, model_location: Union[str, Path]) -> str:
-        components: Dict[str, str] = {}
+        model_path = Path(model_path)
+        if model_path.is_file():
+            return self._hash_file(model_path)
+        elif model_path.is_dir():
+            return self._hash_dir(model_path)
+        else:
+            raise OSError(f"Not a valid file or directory: {model_path}")
 
-        for root, _dirs, files in os.walk(model_location):
-            for file in files:
-                # only tally tensor files because diffusers config files change slightly
-                # depending on how the model was downloaded/converted.
-                if not file.endswith((".ckpt", ".safetensors", ".bin", ".pt", ".pth")):
-                    continue
-                path = (Path(root) / file).as_posix()
-                fast_hash = cls._hash_file(path)
-                components.update({path: fast_hash})
+    def _hash_dir(self, dir: Path) -> str:
+        """Compute the hash for all files in a directory and return a hexdigest.
 
-        # hash all the model hashes together, using alphabetic file order
-        md5 = hashlib.md5()
-        for _path, fast_hash in sorted(components.items()):
-            md5.update(fast_hash.encode("utf-8"))
-        return md5.hexdigest()
+        Args:
+            dir: Path to the directory
+
+        Returns:
+            str: Hexdigest of the hash of the directory
+        """
+        model_component_paths = self._get_file_paths(dir, self._file_filter)
+
+        component_hashes: list[str] = []
+        for component in sorted(model_component_paths):
+            component_hashes.append(self._hash_file(component))
+
+        # BLAKE3 is cryptographically secure. We may as well fall back on a secure algorithm
+        # for the composite hash
+        composite_hasher = blake3()
+        for h in component_hashes:
+            composite_hasher.update(h.encode("utf-8"))
+        return composite_hasher.hexdigest()
+
+    @staticmethod
+    def _get_file_paths(model_path: Path, file_filter: Callable[[str], bool]) -> list[Path]:
+        """Return a list of all model files in the directory.
+
+        Args:
+            model_path: Path to the model
+            file_filter: Function that takes a file name and returns True if the file should be included in the list.
+
+        Returns:
+            List of all model files in the directory
+        """
+
+        files: list[Path] = []
+        for root, _dirs, _files in os.walk(model_path):
+            for file in _files:
+                if file_filter(file):
+                    files.append(Path(root, file))
+        return files
+
+    @staticmethod
+    def _blake3(file_path: Path) -> str:
+        """Hashes a file using BLAKE3
+
+        Args:
+            file_path: Path to the file to hash
+
+        Returns:
+            Hexdigest of the hash of the file
+        """
+        file_hasher = blake3(max_threads=blake3.AUTO)
+        file_hasher.update_mmap(file_path)
+        return file_hasher.hexdigest()
+
+    @staticmethod
+    def _get_hashlib(algorithm: ALGORITHM) -> Callable[[Path], str]:
+        """Factory function that returns a function to hash a file with the given algorithm.
+
+        Args:
+            algorithm: Hashing algorithm to use
+
+        Returns:
+            A function that hashes a file using the given algorithm
+        """
+
+        def hashlib_hasher(file_path: Path) -> str:
+            """Hashes a file using a hashlib algorithm. Uses `memoryview` to avoid reading the entire file into memory."""
+            hasher = hashlib.new(algorithm)
+            buffer = bytearray(128 * 1024)
+            mv = memoryview(buffer)
+            with open(file_path, "rb", buffering=0) as f:
+                while n := f.readinto(mv):
+                    hasher.update(mv[:n])
+            return hasher.hexdigest()
+
+        return hashlib_hasher
+
+    @staticmethod
+    def _default_file_filter(file_path: str) -> bool:
+        """A default file filter that only includes files with the following extensions: .ckpt, .safetensors, .bin, .pt, .pth
+
+        Args:
+            file_path: Path to the file
+
+        Returns:
+            True if the file matches the given extensions, otherwise False
+        """
+        return file_path.endswith(MODEL_FILE_EXTENSIONS)

invokeai/backend/model_manager/load/__init__.py

@@ -2,6 +2,7 @@
 """
 Init file for the model loader.
 """
+
 from importlib import import_module
 from pathlib import Path
 

invokeai/backend/model_manager/load/convert_cache/convert_cache_base.py

@@ -1,6 +1,7 @@
 """
 Disk-based converted model cache.
 """
+
 from abc import ABC, abstractmethod
 from pathlib import Path
 

invokeai/backend/model_manager/load/load_default.py

@@ -13,6 +13,7 @@ from invokeai.backend.model_manager import (
     ModelRepoVariant,
     SubModelType,
 )
+from invokeai.backend.model_manager.config import DiffusersConfigBase, ModelType
 from invokeai.backend.model_manager.load.convert_cache import ModelConvertCacheBase
 from invokeai.backend.model_manager.load.load_base import LoadedModel, ModelLoaderBase
 from invokeai.backend.model_manager.load.model_cache.model_cache_base import ModelCacheBase, ModelLockerBase
@@ -50,7 +51,7 @@ class ModelLoader(ModelLoaderBase):
         :param submodel_type: an ModelType enum indicating the portion of
                the model to retrieve (e.g. ModelType.Vae)
         """
-        if model_config.type == "main" and not submodel_type:
+        if model_config.type is ModelType.Main and not submodel_type:
             raise InvalidModelConfigException("submodel_type is required when loading a main model")
 
         model_path, model_config, submodel_type = self._get_model_path(model_config, submodel_type)
@@ -80,7 +81,7 @@ class ModelLoader(ModelLoaderBase):
         self._convert_cache.make_room(self.get_size_fs(config, model_path, submodel_type))
         return self._convert_model(config, model_path, cache_path)
 
-    def _needs_conversion(self, config: AnyModelConfig, model_path: Path, cache_path: Path) -> bool:
+    def _needs_conversion(self, config: AnyModelConfig, model_path: Path, dest_path: Path) -> bool:
         return False
 
     def _load_if_needed(
@@ -119,7 +120,7 @@ class ModelLoader(ModelLoaderBase):
         return calc_model_size_by_fs(
             model_path=model_path,
             subfolder=submodel_type.value if submodel_type else None,
-            variant=config.repo_variant if hasattr(config, "repo_variant") else None,
+            variant=config.repo_variant if isinstance(config, DiffusersConfigBase) else None,
         )
 
     # This needs to be implemented in subclasses that handle checkpoints

invokeai/backend/model_manager/load/model_loader_registry.py

@@ -14,10 +14,9 @@ Use like this:
     ).load_model(model_config, submodel_type)
 
 """
-import hashlib
+
 from abc import ABC, abstractmethod
-from pathlib import Path
-from typing import Callable, Dict, Optional, Tuple, Type
+from typing import Callable, Dict, Optional, Tuple, Type, TypeVar
 
 from ..config import (
     AnyModelConfig,
@@ -26,8 +25,6 @@ from ..config import (
     ModelFormat,
     ModelType,
     SubModelType,
-    VaeCheckpointConfig,
-    VaeDiffusersConfig,
 )
 from . import ModelLoaderBase
 
@@ -60,6 +57,9 @@ class ModelLoaderRegistryBase(ABC):
         """
 
 
+TModelLoader = TypeVar("TModelLoader", bound=ModelLoaderBase)
+
+
 class ModelLoaderRegistry:
     """
     This class allows model loaders to register their type, base and format.
@@ -70,10 +70,10 @@ class ModelLoaderRegistry:
     @classmethod
     def register(
         cls, type: ModelType, format: ModelFormat, base: BaseModelType = BaseModelType.Any
-    ) -> Callable[[Type[ModelLoaderBase]], Type[ModelLoaderBase]]:
+    ) -> Callable[[Type[TModelLoader]], Type[TModelLoader]]:
         """Define a decorator which registers the subclass of loader."""
 
-        def decorator(subclass: Type[ModelLoaderBase]) -> Type[ModelLoaderBase]:
+        def decorator(subclass: Type[TModelLoader]) -> Type[TModelLoader]:
             key = cls._to_registry_key(base, type, format)
             if key in cls._registry:
                 raise Exception(
@@ -89,33 +89,15 @@ class ModelLoaderRegistry:
         cls, config: AnyModelConfig, submodel_type: Optional[SubModelType]
     ) -> Tuple[Type[ModelLoaderBase], ModelConfigBase, Optional[SubModelType]]:
         """Get subclass of ModelLoaderBase registered to handle base and type."""
-        # We have to handle VAE overrides here because this will change the model type and the corresponding implementation returned
-        conf2, submodel_type = cls._handle_subtype_overrides(config, submodel_type)
 
-        key1 = cls._to_registry_key(conf2.base, conf2.type, conf2.format)  # for a specific base type
-        key2 = cls._to_registry_key(BaseModelType.Any, conf2.type, conf2.format)  # with wildcard Any
+        key1 = cls._to_registry_key(config.base, config.type, config.format)  # for a specific base type
+        key2 = cls._to_registry_key(BaseModelType.Any, config.type, config.format)  # with wildcard Any
         implementation = cls._registry.get(key1) or cls._registry.get(key2)
         if not implementation:
             raise NotImplementedError(
                 f"No subclass of LoadedModel is registered for base={config.base}, type={config.type}, format={config.format}"
             )
-        return implementation, conf2, submodel_type
-
-    @classmethod
-    def _handle_subtype_overrides(
-        cls, config: AnyModelConfig, submodel_type: Optional[SubModelType]
-    ) -> Tuple[ModelConfigBase, Optional[SubModelType]]:
-        if submodel_type == SubModelType.Vae and hasattr(config, "vae") and config.vae is not None:
-            model_path = Path(config.vae)
-            config_class = (
-                VaeCheckpointConfig if model_path.suffix in [".pt", ".safetensors", ".ckpt"] else VaeDiffusersConfig
-            )
-            hash = hashlib.md5(model_path.as_posix().encode("utf-8")).hexdigest()
-            new_conf = config_class(path=model_path.as_posix(), name=model_path.stem, base=config.base, key=hash)
-            submodel_type = None
-        else:
-            new_conf = config
-        return new_conf, submodel_type
+        return implementation, config, submodel_type
 
     @staticmethod
     def _to_registry_key(base: BaseModelType, type: ModelType, format: ModelFormat) -> str:

invokeai/backend/model_manager/load/model_loaders/controlnet.py

@@ -3,8 +3,8 @@
 
 from pathlib import Path
 
-import safetensors
 import torch
+from safetensors.torch import load_file as safetensors_load_file
 
 from invokeai.backend.model_manager import (
     AnyModelConfig,
@@ -12,6 +12,7 @@ from invokeai.backend.model_manager import (
     ModelFormat,
     ModelType,
 )
+from invokeai.backend.model_manager.config import CheckpointConfigBase
 from invokeai.backend.model_manager.convert_ckpt_to_diffusers import convert_controlnet_to_diffusers
 
 from .. import ModelLoaderRegistry
@@ -20,15 +21,15 @@ from .generic_diffusers import GenericDiffusersLoader
 
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.ControlNet, format=ModelFormat.Diffusers)
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.ControlNet, format=ModelFormat.Checkpoint)
-class ControlnetLoader(GenericDiffusersLoader):
+class ControlNetLoader(GenericDiffusersLoader):
     """Class to load ControlNet models."""
 
     def _needs_conversion(self, config: AnyModelConfig, model_path: Path, dest_path: Path) -> bool:
-        if config.format != ModelFormat.Checkpoint:
+        if not isinstance(config, CheckpointConfigBase):
             return False
         elif (
             dest_path.exists()
-            and (dest_path / "config.json").stat().st_mtime >= (config.last_modified or 0.0)
+            and (dest_path / "config.json").stat().st_mtime >= (config.converted_at or 0.0)
             and (dest_path / "config.json").stat().st_mtime >= model_path.stat().st_mtime
         ):
             return False
@@ -37,13 +38,13 @@ class ControlnetLoader(GenericDiffusersLoader):
 
     def _convert_model(self, config: AnyModelConfig, model_path: Path, output_path: Path) -> Path:
         if config.base not in {BaseModelType.StableDiffusion1, BaseModelType.StableDiffusion2}:
-            raise Exception(f"Vae conversion not supported for model type: {config.base}")
+            raise Exception(f"ControlNet conversion not supported for model type: {config.base}")
         else:
-            assert hasattr(config, "config")
-            config_file = config.config
+            assert isinstance(config, CheckpointConfigBase)
+            config_file = config.config_path
 
         if model_path.suffix == ".safetensors":
-            checkpoint = safetensors.torch.load_file(model_path, device="cpu")
+            checkpoint = safetensors_load_file(model_path, device="cpu")
         else:
             checkpoint = torch.load(model_path, map_location="cpu")
 

invokeai/backend/model_manager/load/model_loaders/generic_diffusers.py

@@ -3,9 +3,10 @@
 
 import sys
 from pathlib import Path
-from typing import Any, Dict, Optional
+from typing import Any, Optional
 
-from diffusers import ConfigMixin, ModelMixin
+from diffusers.configuration_utils import ConfigMixin
+from diffusers.models.modeling_utils import ModelMixin
 
 from invokeai.backend.model_manager import (
     AnyModel,
@@ -41,6 +42,7 @@ class GenericDiffusersLoader(ModelLoader):
     # TO DO: Add exception handling
     def get_hf_load_class(self, model_path: Path, submodel_type: Optional[SubModelType] = None) -> ModelMixin:
         """Given the model path and submodel, returns the diffusers ModelMixin subclass needed to load."""
+        result = None
         if submodel_type:
             try:
                 config = self._load_diffusers_config(model_path, config_name="model_index.json")
@@ -64,6 +66,7 @@ class GenericDiffusersLoader(ModelLoader):
                     raise InvalidModelConfigException("Unable to decifer Load Class based on given config.json")
             except KeyError as e:
                 raise InvalidModelConfigException("An expected config.json file is missing from this model.") from e
+        assert result is not None
         return result
 
     # TO DO: Add exception handling
@@ -75,7 +78,7 @@ class GenericDiffusersLoader(ModelLoader):
         result: ModelMixin = getattr(res_type, class_name)
         return result
 
-    def _load_diffusers_config(self, model_path: Path, config_name: str = "config.json") -> Dict[str, Any]:
+    def _load_diffusers_config(self, model_path: Path, config_name: str = "config.json") -> dict[str, Any]:
         return ConfigLoader.load_config(model_path, config_name=config_name)
 
 
@@ -83,8 +86,8 @@ class ConfigLoader(ConfigMixin):
     """Subclass of ConfigMixin for loading diffusers configuration files."""
 
     @classmethod
-    def load_config(cls, *args: Any, **kwargs: Any) -> Dict[str, Any]:
+    def load_config(cls, *args: Any, **kwargs: Any) -> dict[str, Any]:  # pyright: ignore [reportIncompatibleMethodOverride]
         """Load a diffusrs ConfigMixin configuration."""
         cls.config_name = kwargs.pop("config_name")
-        # Diffusers doesn't provide typing info
+        # TODO(psyche): the types on this diffusers method are not correct
         return super().load_config(*args, **kwargs)  # type: ignore

invokeai/backend/model_manager/load/model_loaders/ip_adapter.py

@@ -31,7 +31,7 @@ class IPAdapterInvokeAILoader(ModelLoader):
         if submodel_type is not None:
             raise ValueError("There are no submodels in an IP-Adapter model.")
         model = build_ip_adapter(
-            ip_adapter_ckpt_path=model_path / "ip_adapter.bin",
+            ip_adapter_ckpt_path=str(model_path / "ip_adapter.bin"),
             device=torch.device("cpu"),
             dtype=self._torch_dtype,
         )

invokeai/backend/model_manager/load/model_loaders/lora.py

@@ -1,7 +1,6 @@
 # Copyright (c) 2024, Lincoln D. Stein and the InvokeAI Development Team
 """Class for LoRA model loading in InvokeAI."""
 
-
 from logging import Logger
 from pathlib import Path
 from typing import Optional, Tuple
@@ -23,8 +22,8 @@ from invokeai.backend.model_manager.load.model_cache.model_cache_base import Mod
 from .. import ModelLoader, ModelLoaderRegistry
 
 
-@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.Lora, format=ModelFormat.Diffusers)
-@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.Lora, format=ModelFormat.Lycoris)
+@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.LoRA, format=ModelFormat.Diffusers)
+@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.LoRA, format=ModelFormat.LyCORIS)
 class LoraLoader(ModelLoader):
     """Class to load LoRA models."""
 

invokeai/backend/model_manager/load/model_loaders/onnx.py

@@ -18,7 +18,7 @@ from .. import ModelLoaderRegistry
 from .generic_diffusers import GenericDiffusersLoader
 
 
-@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.ONNX, format=ModelFormat.Onnx)
+@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.ONNX, format=ModelFormat.ONNX)
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.ONNX, format=ModelFormat.Olive)
 class OnnyxDiffusersModel(GenericDiffusersLoader):
     """Class to load onnx models."""

invokeai/backend/model_manager/load/model_loaders/stable_diffusion.py

@@ -1,11 +1,11 @@
 # Copyright (c) 2024, Lincoln D. Stein and the InvokeAI Development Team
 """Class for StableDiffusion model loading in InvokeAI."""
 
-
 from pathlib import Path
 from typing import Optional
 
-from diffusers import StableDiffusionInpaintPipeline, StableDiffusionPipeline
+from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion import StableDiffusionPipeline
+from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_inpaint import StableDiffusionInpaintPipeline
 
 from invokeai.backend.model_manager import (
     AnyModel,
@@ -17,7 +17,7 @@ from invokeai.backend.model_manager import (
     ModelVariantType,
     SubModelType,
 )
-from invokeai.backend.model_manager.config import MainCheckpointConfig
+from invokeai.backend.model_manager.config import CheckpointConfigBase, MainCheckpointConfig
 from invokeai.backend.model_manager.convert_ckpt_to_diffusers import convert_ckpt_to_diffusers
 
 from .. import ModelLoaderRegistry
@@ -55,11 +55,11 @@ class StableDiffusionDiffusersModel(GenericDiffusersLoader):
         return result
 
     def _needs_conversion(self, config: AnyModelConfig, model_path: Path, dest_path: Path) -> bool:
-        if config.format != ModelFormat.Checkpoint:
+        if not isinstance(config, CheckpointConfigBase):
             return False
         elif (
             dest_path.exists()
-            and (dest_path / "model_index.json").stat().st_mtime >= (config.last_modified or 0.0)
+            and (dest_path / "model_index.json").stat().st_mtime >= (config.converted_at or 0.0)
             and (dest_path / "model_index.json").stat().st_mtime >= model_path.stat().st_mtime
         ):
             return False
@@ -74,7 +74,7 @@ class StableDiffusionDiffusersModel(GenericDiffusersLoader):
             StableDiffusionInpaintPipeline if variant == ModelVariantType.Inpaint else StableDiffusionPipeline
         )
 
-        config_file = config.config
+        config_file = config.config_path
 
         self._logger.info(f"Converting {model_path} to diffusers format")
         convert_ckpt_to_diffusers(

invokeai/backend/model_manager/load/model_loaders/textual_inversion.py

@@ -1,7 +1,6 @@
 # Copyright (c) 2024, Lincoln D. Stein and the InvokeAI Development Team
 """Class for TI model loading in InvokeAI."""
 
-
 from pathlib import Path
 from typing import Optional, Tuple
 

invokeai/backend/model_manager/load/model_loaders/vae.py

@@ -3,9 +3,9 @@
 
 from pathlib import Path
 
-import safetensors
 import torch
 from omegaconf import DictConfig, OmegaConf
+from safetensors.torch import load_file as safetensors_load_file
 
 from invokeai.backend.model_manager import (
     AnyModelConfig,
@@ -13,24 +13,25 @@ from invokeai.backend.model_manager import (
     ModelFormat,
     ModelType,
 )
+from invokeai.backend.model_manager.config import CheckpointConfigBase
 from invokeai.backend.model_manager.convert_ckpt_to_diffusers import convert_ldm_vae_to_diffusers
 
 from .. import ModelLoaderRegistry
 from .generic_diffusers import GenericDiffusersLoader
 
 
-@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.Vae, format=ModelFormat.Diffusers)
-@ModelLoaderRegistry.register(base=BaseModelType.StableDiffusion1, type=ModelType.Vae, format=ModelFormat.Checkpoint)
-@ModelLoaderRegistry.register(base=BaseModelType.StableDiffusion2, type=ModelType.Vae, format=ModelFormat.Checkpoint)
+@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.VAE, format=ModelFormat.Diffusers)
+@ModelLoaderRegistry.register(base=BaseModelType.StableDiffusion1, type=ModelType.VAE, format=ModelFormat.Checkpoint)
+@ModelLoaderRegistry.register(base=BaseModelType.StableDiffusion2, type=ModelType.VAE, format=ModelFormat.Checkpoint)
 class VaeLoader(GenericDiffusersLoader):
     """Class to load VAE models."""
 
     def _needs_conversion(self, config: AnyModelConfig, model_path: Path, dest_path: Path) -> bool:
-        if config.format != ModelFormat.Checkpoint:
+        if not isinstance(config, CheckpointConfigBase):
             return False
         elif (
             dest_path.exists()
-            and (dest_path / "config.json").stat().st_mtime >= (config.last_modified or 0.0)
+            and (dest_path / "config.json").stat().st_mtime >= (config.converted_at or 0.0)
             and (dest_path / "config.json").stat().st_mtime >= model_path.stat().st_mtime
         ):
             return False
@@ -38,16 +39,15 @@ class VaeLoader(GenericDiffusersLoader):
             return True
 
     def _convert_model(self, config: AnyModelConfig, model_path: Path, output_path: Path) -> Path:
-        # TO DO: check whether sdxl VAE models convert.
+        # TODO(MM2): check whether sdxl VAE models convert.
         if config.base not in {BaseModelType.StableDiffusion1, BaseModelType.StableDiffusion2}:
-            raise Exception(f"Vae conversion not supported for model type: {config.base}")
+            raise Exception(f"VAE conversion not supported for model type: {config.base}")
         else:
-            config_file = (
-                "v1-inference.yaml" if config.base == BaseModelType.StableDiffusion1 else "v2-inference-v.yaml"
-            )
+            assert isinstance(config, CheckpointConfigBase)
+            config_file = config.config_path
 
         if model_path.suffix == ".safetensors":
-            checkpoint = safetensors.torch.load_file(model_path, device="cpu")
+            checkpoint = safetensors_load_file(model_path, device="cpu")
         else:
             checkpoint = torch.load(model_path, map_location="cpu")
 
@@ -55,7 +55,7 @@ class VaeLoader(GenericDiffusersLoader):
         if "state_dict" in checkpoint:
             checkpoint = checkpoint["state_dict"]
 
-        ckpt_config = OmegaConf.load(self._app_config.legacy_conf_path / config_file)
+        ckpt_config = OmegaConf.load(self._app_config.root_path / config_file)
         assert isinstance(ckpt_config, DictConfig)
 
         vae_model = convert_ldm_vae_to_diffusers(

invokeai/backend/model_manager/merge.py

@@ -16,6 +16,7 @@ from diffusers import AutoPipelineForText2Image
 from diffusers.utils import logging as dlogging
 
 from invokeai.app.services.model_install import ModelInstallServiceBase
+from invokeai.app.services.model_records.model_records_base import ModelRecordChanges
 from invokeai.backend.util.devices import choose_torch_device, torch_dtype
 
 from . import (
@@ -117,7 +118,6 @@ class ModelMerger(object):
         config = self._installer.app_config
         store = self._installer.record_store
         base_models: Set[BaseModelType] = set()
-        vae = None
         variant = None if self._installer.app_config.full_precision else "fp16"
 
         assert (
@@ -134,10 +134,6 @@ class ModelMerger(object):
                 "normal"
             ), f"{info.name} ({info.key}) is a {info.variant} model, which cannot currently be merged"
 
-            # pick up the first model's vae
-            if key == model_keys[0]:
-                vae = info.vae
-
             # tally base models used
             base_models.add(info.base)
             model_paths.extend([config.models_path / info.path])
@@ -163,12 +159,10 @@ class ModelMerger(object):
 
         # update model's config
         model_config = self._installer.record_store.get_model(key)
-        model_config.update(
-            {
-                "name": merged_model_name,
-                "description": f"Merge of models {', '.join(model_names)}",
-                "vae": vae,
-            }
+        model_config.name = merged_model_name
+        model_config.description = f"Merge of models {', '.join(model_names)}"
+
+        self._installer.record_store.update_model(
+            key, ModelRecordChanges(name=model_config.name, description=model_config.description)
         )
-        self._installer.record_store.update_model(key, model_config)
         return model_config

invokeai/backend/model_manager/metadata/__init__.py

@@ -18,15 +18,14 @@ assert isinstance(data, CivitaiMetadata)
 if data.allow_commercial_use:
    print("Commercial use of this model is allowed")
 """
+
 from .fetch import CivitaiMetadataFetch, HuggingFaceMetadataFetch, ModelMetadataFetchBase
 from .metadata_base import (
     AnyModelRepoMetadata,
     AnyModelRepoMetadataValidator,
     BaseMetadata,
     CivitaiMetadata,
-    CommercialUsage,
     HuggingFaceMetadata,
-    LicenseRestrictions,
     ModelMetadataWithFiles,
     RemoteModelFile,
     UnknownMetadataException,
@@ -37,10 +36,8 @@ __all__ = [
     "AnyModelRepoMetadataValidator",
     "CivitaiMetadata",
     "CivitaiMetadataFetch",
-    "CommercialUsage",
     "HuggingFaceMetadata",
     "HuggingFaceMetadataFetch",
-    "LicenseRestrictions",
     "ModelMetadataFetchBase",
     "BaseMetadata",
     "ModelMetadataWithFiles",

invokeai/backend/model_manager/metadata/fetch/civitai.py

@@ -23,22 +23,21 @@ metadata = fetcher.from_url("https://civitai.com/models/206883/split")
 print(metadata.trained_words)
 """
 
+import json
 import re
-from datetime import datetime
 from pathlib import Path
-from typing import Any, Dict, Optional
+from typing import Any, Optional
 
 import requests
+from pydantic import TypeAdapter, ValidationError
 from pydantic.networks import AnyHttpUrl
 from requests.sessions import Session
 
-from invokeai.backend.model_manager import ModelRepoVariant
+from invokeai.backend.model_manager.config import ModelRepoVariant
 
 from ..metadata_base import (
     AnyModelRepoMetadata,
     CivitaiMetadata,
-    CommercialUsage,
-    LicenseRestrictions,
     RemoteModelFile,
     UnknownMetadataException,
 )
@@ -52,10 +51,13 @@ CIVITAI_VERSION_ENDPOINT = "https://civitai.com/api/v1/model-versions/"
 CIVITAI_MODEL_ENDPOINT = "https://civitai.com/api/v1/models/"
 
 
+StringSetAdapter = TypeAdapter(set[str])
+
+
 class CivitaiMetadataFetch(ModelMetadataFetchBase):
     """Fetch model metadata from Civitai."""
 
-    def __init__(self, session: Optional[Session] = None):
+    def __init__(self, session: Optional[Session] = None, api_key: Optional[str] = None):
         """
         Initialize the fetcher with an optional requests.sessions.Session object.
 
@@ -63,6 +65,7 @@ class CivitaiMetadataFetch(ModelMetadataFetchBase):
         this module without an internet connection.
         """
         self._requests = session or requests.Session()
+        self._api_key = api_key
 
     def from_url(self, url: AnyHttpUrl) -> AnyModelRepoMetadata:
         """
@@ -102,22 +105,21 @@ class CivitaiMetadataFetch(ModelMetadataFetchBase):
         May raise an `UnknownMetadataException`.
         """
         model_url = CIVITAI_MODEL_ENDPOINT + str(model_id)
-        model_json = self._requests.get(model_url).json()
-        return self._from_model_json(model_json)
+        model_json = self._requests.get(self._get_url_with_api_key(model_url)).json()
+        return self._from_api_response(model_json)
 
-    def _from_model_json(self, model_json: Dict[str, Any], version_id: Optional[int] = None) -> CivitaiMetadata:
+    def _from_api_response(self, api_response: dict[str, Any], version_id: Optional[int] = None) -> CivitaiMetadata:
         try:
-            version_id = version_id or model_json["modelVersions"][0]["id"]
+            version_id = version_id or api_response["modelVersions"][0]["id"]
         except TypeError as excp:
             raise UnknownMetadataException from excp
 
         # loop till we find the section containing the version requested
-        version_sections = [x for x in model_json["modelVersions"] if x["id"] == version_id]
+        version_sections = [x for x in api_response["modelVersions"] if x["id"] == version_id]
         if not version_sections:
             raise UnknownMetadataException(f"Version {version_id} not found in model metadata")
 
         version_json = version_sections[0]
-        safe_thumbnails = [x["url"] for x in version_json["images"] if x["nsfw"] == "None"]
 
         # Civitai has one "primary" file plus others such as VAEs. We only fetch the primary.
         primary = [x for x in version_json["files"] if x.get("primary")]
@@ -134,37 +136,23 @@ class CivitaiMetadataFetch(ModelMetadataFetchBase):
             url = url + f"?type={primary_file['type']}{metadata_string}"
         model_files = [
             RemoteModelFile(
-                url=url,
+                url=self._get_url_with_api_key(url),
                 path=Path(primary_file["name"]),
                 size=int(primary_file["sizeKB"] * 1024),
                 sha256=primary_file["hashes"]["SHA256"],
             )
         ]
+
+        try:
+            trigger_phrases = StringSetAdapter.validate_python(version_json.get("trainedWords"))
+        except ValidationError:
+            trigger_phrases: set[str] = set()
+
         return CivitaiMetadata(
-            id=model_json["id"],
             name=version_json["name"],
-            version_id=version_json["id"],
-            version_name=version_json["name"],
-            created=datetime.fromisoformat(_fix_timezone(version_json["createdAt"])),
-            updated=datetime.fromisoformat(_fix_timezone(version_json["updatedAt"])),
-            published=datetime.fromisoformat(_fix_timezone(version_json["publishedAt"])),
-            base_model_trained_on=version_json["baseModel"],  # note - need a dictionary to turn into a BaseModelType
             files=model_files,
-            download_url=version_json["downloadUrl"],
-            thumbnail_url=safe_thumbnails[0] if safe_thumbnails else None,
-            author=model_json["creator"]["username"],
-            description=model_json["description"],
-            version_description=version_json["description"] or "",
-            tags=model_json["tags"],
-            trained_words=version_json["trainedWords"],
-            nsfw=model_json["nsfw"],
-            restrictions=LicenseRestrictions(
-                AllowNoCredit=model_json["allowNoCredit"],
-                AllowCommercialUse=CommercialUsage(model_json["allowCommercialUse"]),
-                AllowDerivatives=model_json["allowDerivatives"],
-                AllowDifferentLicense=model_json["allowDifferentLicense"],
-            ),
-            trigger_phrases=version_json["trainedWords"],
+            trigger_phrases=trigger_phrases,
+            api_response=json.dumps(version_json),
         )
 
     def from_civitai_versionid(self, version_id: int, model_id: Optional[int] = None) -> CivitaiMetadata:
@@ -175,14 +163,14 @@ class CivitaiMetadataFetch(ModelMetadataFetchBase):
         """
         if model_id is None:
             version_url = CIVITAI_VERSION_ENDPOINT + str(version_id)
-            version = self._requests.get(version_url).json()
+            version = self._requests.get(self._get_url_with_api_key(version_url)).json()
             if error := version.get("error"):
                 raise UnknownMetadataException(error)
             model_id = version["modelId"]
 
         model_url = CIVITAI_MODEL_ENDPOINT + str(model_id)
-        model_json = self._requests.get(model_url).json()
-        return self._from_model_json(model_json, version_id)
+        model_json = self._requests.get(self._get_url_with_api_key(model_url)).json()
+        return self._from_api_response(model_json, version_id)
 
     @classmethod
     def from_json(cls, json: str) -> CivitaiMetadata:
@@ -190,6 +178,11 @@ class CivitaiMetadataFetch(ModelMetadataFetchBase):
         metadata = CivitaiMetadata.model_validate_json(json)
         return metadata
 
+    def _get_url_with_api_key(self, url: str) -> str:
+        if not self._api_key:
+            return url
 
-def _fix_timezone(date: str) -> str:
-    return re.sub(r"Z$", "+00:00", date)
+        if "?" in url:
+            return f"{url}&token={self._api_key}"
+
+        return f"{url}?token={self._api_key}"

invokeai/backend/model_manager/metadata/fetch/huggingface.py

@@ -13,6 +13,7 @@ metadata = fetcher.from_url("https://huggingface.co/stabilityai/sdxl-turbo")
 print(metadata.tags)
 """
 
+import json
 import re
 from pathlib import Path
 from typing import Optional
@@ -23,7 +24,7 @@ from huggingface_hub.utils._errors import RepositoryNotFoundError, RevisionNotFo
 from pydantic.networks import AnyHttpUrl
 from requests.sessions import Session
 
-from invokeai.backend.model_manager import ModelRepoVariant
+from invokeai.backend.model_manager.config import ModelRepoVariant
 
 from ..metadata_base import (
     AnyModelRepoMetadata,
@@ -60,6 +61,7 @@ class HuggingFaceMetadataFetch(ModelMetadataFetchBase):
         # Little loop which tries fetching a revision corresponding to the selected variant.
         # If not available, then set variant to None and get the default.
         # If this too fails, raise exception.
+
         model_info = None
         while not model_info:
             try:
@@ -72,23 +74,24 @@ class HuggingFaceMetadataFetch(ModelMetadataFetchBase):
                 else:
                     variant = None
 
+        files: list[RemoteModelFile] = []
+
         _, name = id.split("/")
-        return HuggingFaceMetadata(
-            id=model_info.id,
-            author=model_info.author,
-            name=name,
-            last_modified=model_info.last_modified,
-            tag_dict=model_info.card_data.to_dict() if model_info.card_data else {},
-            tags=model_info.tags,
-            files=[
+
+        for s in model_info.siblings or []:
+            assert s.rfilename is not None
+            assert s.size is not None
+            files.append(
                 RemoteModelFile(
-                    url=hf_hub_url(id, x.rfilename, revision=variant),
-                    path=Path(name, x.rfilename),
-                    size=x.size,
-                    sha256=x.lfs.get("sha256") if x.lfs else None,
+                    url=hf_hub_url(id, s.rfilename, revision=variant),
+                    path=Path(name, s.rfilename),
+                    size=s.size,
+                    sha256=s.lfs.get("sha256") if s.lfs else None,
                 )
-                for x in model_info.siblings
-            ],
+            )
+
+        return HuggingFaceMetadata(
+            id=model_info.id, name=name, files=files, api_response=json.dumps(model_info.__dict__, default=str)
         )
 
     def from_url(self, url: AnyHttpUrl) -> AnyModelRepoMetadata:

invokeai/backend/model_manager/metadata/metadata_base.py

@@ -14,17 +14,14 @@ versions of these fields are intended to be kept in sync with the
 remote repo.
 """
 
-from datetime import datetime
-from enum import Enum
 from pathlib import Path
-from typing import Any, Dict, List, Literal, Optional, Set, Tuple, Union
+from typing import List, Literal, Optional, Union
 
 from huggingface_hub import configure_http_backend, hf_hub_url
 from pydantic import BaseModel, Field, TypeAdapter
 from pydantic.networks import AnyHttpUrl
 from requests.sessions import Session
 from typing_extensions import Annotated
-from invokeai.app.invocations.constants import SCHEDULER_NAME_VALUES
 
 from invokeai.backend.model_manager import ModelRepoVariant
 
@@ -35,31 +32,6 @@ class UnknownMetadataException(Exception):
     """Raised when no metadata is available for a model."""
 
 
-class CommercialUsage(str, Enum):
-    """Type of commercial usage allowed."""
-
-    No = "None"
-    Image = "Image"
-    Rent = "Rent"
-    RentCivit = "RentCivit"
-    Sell = "Sell"
-
-
-class LicenseRestrictions(BaseModel):
-    """Broad categories of licensing restrictions."""
-
-    AllowNoCredit: bool = Field(
-        description="if true, model can be redistributed without crediting author", default=False
-    )
-    AllowDerivatives: bool = Field(description="if true, derivatives of this model can be redistributed", default=False)
-    AllowDifferentLicense: bool = Field(
-        description="if true, derivatives of this model be redistributed under a different license", default=False
-    )
-    AllowCommercialUse: Optional[CommercialUsage] = Field(
-        description="Type of commercial use allowed or 'No' if no commercial use is allowed.", default=None
-    )
-
-
 class RemoteModelFile(BaseModel):
     """Information about a downloadable file that forms part of a model."""
 
@@ -69,22 +41,10 @@ class RemoteModelFile(BaseModel):
     sha256: Optional[str] = Field(description="SHA256 hash of this model (not always available)", default=None)
 
 
-class ModelDefaultSettings(BaseModel):
-    vae: str | None
-    vae_precision: str | None
-    scheduler: SCHEDULER_NAME_VALUES | None
-    steps: int | None
-    cfg_scale: float | None
-    cfg_rescale_multiplier: float | None
-
 class ModelMetadataBase(BaseModel):
     """Base class for model metadata information."""
 
     name: str = Field(description="model's name")
-    author: str = Field(description="model's author")
-    tags: Optional[Set[str]] = Field(description="tags provided by model source", default=None)
-    trigger_phrases: Optional[List[str]] = Field(description="trigger phrases for this model", default=None)
-    default_settings: Optional[ModelDefaultSettings] = Field(description="default settings for this model", default=None)
 
 
 class BaseMetadata(ModelMetadataBase):
@@ -122,57 +82,16 @@ class CivitaiMetadata(ModelMetadataWithFiles):
     """Extended metadata fields provided by Civitai."""
 
     type: Literal["civitai"] = "civitai"
-    id: int = Field(description="Civitai version identifier")
-    version_name: str = Field(description="Version identifier, such as 'V2-alpha'")
-    version_id: int = Field(description="Civitai model version identifier")
-    created: datetime = Field(description="date the model was created")
-    updated: datetime = Field(description="date the model was last modified")
-    published: datetime = Field(description="date the model was published to Civitai")
-    description: str = Field(description="text description of model; may contain HTML")
-    version_description: str = Field(
-        description="text description of the model's reversion; usually change history; may contain HTML"
-    )
-    nsfw: bool = Field(description="whether the model tends to generate NSFW content", default=False)
-    restrictions: LicenseRestrictions = Field(description="license terms", default_factory=LicenseRestrictions)
-    trained_words: Set[str] = Field(description="words to trigger the model", default_factory=set)
-    download_url: AnyHttpUrl = Field(description="download URL for this model")
-    base_model_trained_on: str = Field(description="base model on which this model was trained (currently not an enum)")
-    thumbnail_url: Optional[AnyHttpUrl] = Field(description="a thumbnail image for this model", default=None)
-    weight_minmax: Tuple[float, float] = Field(
-        description="minimum and maximum slider values for a LoRA or other secondary model", default=(-1.0, +2.0)
-    )  # note: For future use
-
-    @property
-    def credit_required(self) -> bool:
-        """Return True if you must give credit for derivatives of this model and images generated from it."""
-        return not self.restrictions.AllowNoCredit
-
-    @property
-    def allow_commercial_use(self) -> bool:
-        """Return True if commercial use is allowed."""
-        if self.restrictions.AllowCommercialUse is None:
-            return False
-        else:
-            return self.restrictions.AllowCommercialUse != CommercialUsage("None")
-
-    @property
-    def allow_derivatives(self) -> bool:
-        """Return True if derivatives of this model can be redistributed."""
-        return self.restrictions.AllowDerivatives
-
-    @property
-    def allow_different_license(self) -> bool:
-        """Return true if derivatives of this model can use a different license."""
-        return self.restrictions.AllowDifferentLicense
+    trigger_phrases: set[str] = Field(description="Trigger phrases extracted from the API response")
+    api_response: Optional[str] = Field(description="Response from the Civitai API as stringified JSON", default=None)
 
 
 class HuggingFaceMetadata(ModelMetadataWithFiles):
     """Extended metadata fields provided by HuggingFace."""
 
     type: Literal["huggingface"] = "huggingface"
-    id: str = Field(description="huggingface model id")
-    tag_dict: Dict[str, Any]
-    last_modified: datetime = Field(description="date of last commit to repo")
+    id: str = Field(description="The HF model id")
+    api_response: Optional[str] = Field(description="Response from the HF API as stringified JSON", default=None)
 
     def download_urls(
         self,
@@ -201,7 +120,7 @@ class HuggingFaceMetadata(ModelMetadataWithFiles):
         # the next step reads model_index.json to determine which subdirectories belong
         # to the model
         if Path(f"{prefix}model_index.json") in paths:
-            url = hf_hub_url(self.id, filename="model_index.json", subfolder=subfolder)
+            url = hf_hub_url(self.id, filename="model_index.json", subfolder=str(subfolder) if subfolder else None)
             resp = session.get(url)
             resp.raise_for_status()
             submodels = resp.json()

invokeai/backend/model_manager/metadata/metadata_store.py

@@ -1,221 +0,0 @@
-# Copyright (c) 2023 Lincoln D. Stein and the InvokeAI Development Team
-"""
-SQL Storage for Model Metadata
-"""
-
-import sqlite3
-from typing import List, Optional, Set, Tuple
-
-from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
-
-from .fetch import ModelMetadataFetchBase
-from .metadata_base import AnyModelRepoMetadata, UnknownMetadataException
-
-
-class ModelMetadataStore:
-    """Store, search and fetch model metadata retrieved from remote repositories."""
-
-    def __init__(self, db: SqliteDatabase):
-        """
-        Initialize a new object from preexisting sqlite3 connection and threading lock objects.
-
-        :param conn: sqlite3 connection object
-        :param lock: threading Lock object
-        """
-        super().__init__()
-        self._db = db
-        self._cursor = self._db.conn.cursor()
-
-    def add_metadata(self, model_key: str, metadata: AnyModelRepoMetadata) -> None:
-        """
-        Add a block of repo metadata to a model record.
-
-        The model record config must already exist in the database with the
-        same key. Otherwise a FOREIGN KEY constraint exception will be raised.
-
-        :param model_key: Existing model key in the `model_config` table
-        :param metadata: ModelRepoMetadata object to store
-        """
-        json_serialized = metadata.model_dump_json()
-        with self._db.lock:
-            try:
-                self._cursor.execute(
-                    """--sql
-                    INSERT INTO model_metadata(
-                       id,
-                       metadata
-                    )
-                    VALUES (?,?);
-                    """,
-                    (
-                        model_key,
-                        json_serialized,
-                    ),
-                )
-                self._update_tags(model_key, metadata.tags)
-                self._db.conn.commit()
-            except sqlite3.IntegrityError as excp:  # FOREIGN KEY error: the key was not in model_config table
-                self._db.conn.rollback()
-                raise UnknownMetadataException from excp
-            except sqlite3.Error as excp:
-                self._db.conn.rollback()
-                raise excp
-
-    def get_metadata(self, model_key: str) -> AnyModelRepoMetadata:
-        """Retrieve the ModelRepoMetadata corresponding to model key."""
-        with self._db.lock:
-            self._cursor.execute(
-                """--sql
-                SELECT metadata FROM model_metadata
-                WHERE id=?;
-                """,
-                (model_key,),
-            )
-            rows = self._cursor.fetchone()
-            if not rows:
-                raise UnknownMetadataException("model metadata not found")
-            return ModelMetadataFetchBase.from_json(rows[0])
-
-    def list_all_metadata(self) -> List[Tuple[str, AnyModelRepoMetadata]]:  # key, metadata
-        """Dump out all the metadata."""
-        with self._db.lock:
-            self._cursor.execute(
-                """--sql
-                SELECT id,metadata FROM model_metadata;
-                """,
-                (),
-            )
-            rows = self._cursor.fetchall()
-        return [(x[0], ModelMetadataFetchBase.from_json(x[1])) for x in rows]
-
-    def update_metadata(self, model_key: str, metadata: AnyModelRepoMetadata) -> AnyModelRepoMetadata:
-        """
-        Update metadata corresponding to the model with the indicated key.
-
-        :param model_key: Existing model key in the `model_config` table
-        :param metadata: ModelRepoMetadata object to update
-        """
-        json_serialized = metadata.model_dump_json()  # turn it into a json string.
-        with self._db.lock:
-            try:
-                self._cursor.execute(
-                    """--sql
-                    UPDATE model_metadata
-                    SET
-                        metadata=?
-                    WHERE id=?;
-                    """,
-                    (json_serialized, model_key),
-                )
-                if self._cursor.rowcount == 0:
-                    raise UnknownMetadataException("model metadata not found")
-                self._update_tags(model_key, metadata.tags)
-                self._db.conn.commit()
-            except sqlite3.Error as e:
-                self._db.conn.rollback()
-                raise e
-
-        return self.get_metadata(model_key)
-
-    def list_tags(self) -> Set[str]:
-        """Return all tags in the tags table."""
-        self._cursor.execute(
-            """--sql
-            select tag_text from tags;
-            """
-        )
-        return {x[0] for x in self._cursor.fetchall()}
-
-    def search_by_tag(self, tags: Set[str]) -> Set[str]:
-        """Return the keys of models containing all of the listed tags."""
-        with self._db.lock:
-            try:
-                matches: Optional[Set[str]] = None
-                for tag in tags:
-                    self._cursor.execute(
-                        """--sql
-                        SELECT a.model_id FROM model_tags AS a,
-                                                     tags AS b
-                        WHERE a.tag_id=b.tag_id
-                          AND b.tag_text=?;
-                        """,
-                        (tag,),
-                    )
-                    model_keys = {x[0] for x in self._cursor.fetchall()}
-                    if matches is None:
-                        matches = model_keys
-                    matches = matches.intersection(model_keys)
-            except sqlite3.Error as e:
-                raise e
-        return matches if matches else set()
-
-    def search_by_author(self, author: str) -> Set[str]:
-        """Return the keys of models authored by the indicated author."""
-        self._cursor.execute(
-            """--sql
-            SELECT id FROM model_metadata
-            WHERE author=?;
-            """,
-            (author,),
-        )
-        return {x[0] for x in self._cursor.fetchall()}
-
-    def search_by_name(self, name: str) -> Set[str]:
-        """
-        Return the keys of models with the indicated name.
-
-        Note that this is the name of the model given to it by
-        the remote source. The user may have changed the local
-        name. The local name will be located in the model config
-        record object.
-        """
-        self._cursor.execute(
-            """--sql
-            SELECT id FROM model_metadata
-            WHERE name=?;
-            """,
-            (name,),
-        )
-        return {x[0] for x in self._cursor.fetchall()}
-
-    def _update_tags(self, model_key: str, tags: Set[str]) -> None:
-        """Update tags for the model referenced by model_key."""
-        # remove previous tags from this model
-        self._cursor.execute(
-            """--sql
-            DELETE FROM model_tags
-            WHERE model_id=?;
-            """,
-            (model_key,),
-        )
-
-        for tag in tags:
-            self._cursor.execute(
-                """--sql
-                INSERT OR IGNORE INTO tags (
-                  tag_text
-                  )
-                VALUES (?);
-                """,
-                (tag,),
-            )
-            self._cursor.execute(
-                """--sql
-                SELECT tag_id
-                FROM tags
-                WHERE tag_text = ?
-                LIMIT 1;
-                """,
-                (tag,),
-            )
-            tag_id = self._cursor.fetchone()[0]
-            self._cursor.execute(
-                """--sql
-                INSERT OR IGNORE INTO model_tags (
-                   model_id,
-                   tag_id
-                  )
-                VALUES (?,?);
-                """,
-                (model_key, tag_id),
-            )

invokeai/backend/model_manager/probe.py

@@ -8,6 +8,7 @@ import torch
 from picklescan.scanner import scan_file_path
 
 import invokeai.backend.util.logging as logger
+from invokeai.app.util.misc import uuid_string
 from invokeai.backend.util.util import SilenceWarnings
 
 from .config import (
@@ -17,11 +18,12 @@ from .config import (
     ModelConfigFactory,
     ModelFormat,
     ModelRepoVariant,
+    ModelSourceType,
     ModelType,
     ModelVariantType,
     SchedulerPredictionType,
 )
-from .hash import FastModelHash
+from .hash import ModelHash
 from .util.model_util import lora_token_vector_length, read_checkpoint_meta
 
 CkptType = Dict[str, Any]
@@ -95,8 +97,8 @@ class ModelProbe(object):
         "StableDiffusionXLImg2ImgPipeline": ModelType.Main,
         "StableDiffusionXLInpaintPipeline": ModelType.Main,
         "LatentConsistencyModelPipeline": ModelType.Main,
-        "AutoencoderKL": ModelType.Vae,
-        "AutoencoderTiny": ModelType.Vae,
+        "AutoencoderKL": ModelType.VAE,
+        "AutoencoderTiny": ModelType.VAE,
         "ControlNetModel": ModelType.ControlNet,
         "CLIPVisionModelWithProjection": ModelType.CLIPVision,
         "T2IAdapter": ModelType.T2IAdapter,
@@ -108,14 +110,6 @@ class ModelProbe(object):
     ) -> None:
         cls.PROBES[format][model_type] = probe_class
 
-    @classmethod
-    def heuristic_probe(
-        cls,
-        model_path: Path,
-        fields: Optional[Dict[str, Any]] = None,
-    ) -> AnyModelConfig:
-        return cls.probe(model_path, fields)
-
     @classmethod
     def probe(
         cls,
@@ -137,19 +131,21 @@ class ModelProbe(object):
         format_type = ModelFormat.Diffusers if model_path.is_dir() else ModelFormat.Checkpoint
         model_info = None
         model_type = None
-        if format_type == "diffusers":
+        if format_type is ModelFormat.Diffusers:
             model_type = cls.get_model_type_from_folder(model_path)
         else:
             model_type = cls.get_model_type_from_checkpoint(model_path)
-        format_type = ModelFormat.Onnx if model_type == ModelType.ONNX else format_type
+        format_type = ModelFormat.ONNX if model_type == ModelType.ONNX else format_type
 
         probe_class = cls.PROBES[format_type].get(model_type)
         if not probe_class:
             raise InvalidModelConfigException(f"Unhandled combination of {format_type} and {model_type}")
 
-        hash = FastModelHash.hash(model_path)
         probe = probe_class(model_path)
 
+        fields["source_type"] = fields.get("source_type") or ModelSourceType.Path
+        fields["source"] = fields.get("source") or model_path.as_posix()
+        fields["key"] = fields.get("key", uuid_string())
         fields["path"] = model_path.as_posix()
         fields["type"] = fields.get("type") or model_type
         fields["base"] = fields.get("base") or probe.get_base_type()
@@ -161,15 +157,17 @@ class ModelProbe(object):
             fields.get("description") or f"{fields['base'].value} {fields['type'].value} model {fields['name']}"
         )
         fields["format"] = fields.get("format") or probe.get_format()
-        fields["original_hash"] = fields.get("original_hash") or hash
-        fields["current_hash"] = fields.get("current_hash") or hash
+        fields["hash"] = fields.get("hash") or ModelHash().hash(model_path)
 
-        if format_type == ModelFormat.Diffusers and hasattr(probe, "get_repo_variant"):
+        if format_type == ModelFormat.Diffusers and isinstance(probe, FolderProbeBase):
             fields["repo_variant"] = fields.get("repo_variant") or probe.get_repo_variant()
 
         # additional fields needed for main and controlnet models
-        if fields["type"] in [ModelType.Main, ModelType.ControlNet] and fields["format"] == ModelFormat.Checkpoint:
-            fields["config"] = cls._get_checkpoint_config_path(
+        if (
+            fields["type"] in [ModelType.Main, ModelType.ControlNet, ModelType.VAE]
+            and fields["format"] is ModelFormat.Checkpoint
+        ):
+            fields["config_path"] = cls._get_checkpoint_config_path(
                 model_path,
                 model_type=fields["type"],
                 base_type=fields["base"],
@@ -179,7 +177,7 @@ class ModelProbe(object):
 
         # additional fields needed for main non-checkpoint models
         elif fields["type"] == ModelType.Main and fields["format"] in [
-            ModelFormat.Onnx,
+            ModelFormat.ONNX,
             ModelFormat.Olive,
             ModelFormat.Diffusers,
         ]:
@@ -188,7 +186,7 @@ class ModelProbe(object):
                 and fields["prediction_type"] == SchedulerPredictionType.VPrediction
             )
 
-        model_info = ModelConfigFactory.make_config(fields, key=fields.get("key", None))
+        model_info = ModelConfigFactory.make_config(fields)  # , key=fields.get("key", None))
         return model_info
 
     @classmethod
@@ -213,11 +211,11 @@ class ModelProbe(object):
             if any(key.startswith(v) for v in {"cond_stage_model.", "first_stage_model.", "model.diffusion_model."}):
                 return ModelType.Main
             elif any(key.startswith(v) for v in {"encoder.conv_in", "decoder.conv_in"}):
-                return ModelType.Vae
+                return ModelType.VAE
             elif any(key.startswith(v) for v in {"lora_te_", "lora_unet_"}):
-                return ModelType.Lora
+                return ModelType.LoRA
             elif any(key.endswith(v) for v in {"to_k_lora.up.weight", "to_q_lora.down.weight"}):
-                return ModelType.Lora
+                return ModelType.LoRA
             elif any(key.startswith(v) for v in {"control_model", "input_blocks"}):
                 return ModelType.ControlNet
             elif key in {"emb_params", "string_to_param"}:
@@ -239,7 +237,7 @@ class ModelProbe(object):
             if (folder_path / f"learned_embeds.{suffix}").exists():
                 return ModelType.TextualInversion
             if (folder_path / f"pytorch_lora_weights.{suffix}").exists():
-                return ModelType.Lora
+                return ModelType.LoRA
         if (folder_path / "unet/model.onnx").exists():
             return ModelType.ONNX
         if (folder_path / "image_encoder.txt").exists():
@@ -285,13 +283,21 @@ class ModelProbe(object):
         if possible_conf.exists():
             return possible_conf.absolute()
 
-        if model_type == ModelType.Main:
+        if model_type is ModelType.Main:
             config_file = LEGACY_CONFIGS[base_type][variant_type]
             if isinstance(config_file, dict):  # need another tier for sd-2.x models
                 config_file = config_file[prediction_type]
-        elif model_type == ModelType.ControlNet:
+        elif model_type is ModelType.ControlNet:
             config_file = (
-                "../controlnet/cldm_v15.yaml" if base_type == BaseModelType("sd-1") else "../controlnet/cldm_v21.yaml"
+                "../controlnet/cldm_v15.yaml"
+                if base_type is BaseModelType.StableDiffusion1
+                else "../controlnet/cldm_v21.yaml"
+            )
+        elif model_type is ModelType.VAE:
+            config_file = (
+                "../stable-diffusion/v1-inference.yaml"
+                if base_type is BaseModelType.StableDiffusion1
+                else "../stable-diffusion/v2-inference.yaml"
             )
         else:
             raise InvalidModelConfigException(
@@ -497,12 +503,12 @@ class FolderProbeBase(ProbeBase):
             if ".fp16" in x.suffixes:
                 return ModelRepoVariant.FP16
             if "openvino_model" in x.name:
-                return ModelRepoVariant.OPENVINO
+                return ModelRepoVariant.OpenVINO
             if "flax_model" in x.name:
-                return ModelRepoVariant.FLAX
+                return ModelRepoVariant.Flax
             if x.suffix == ".onnx":
                 return ModelRepoVariant.ONNX
-        return ModelRepoVariant.DEFAULT
+        return ModelRepoVariant.Default
 
 
 class PipelineFolderProbe(FolderProbeBase):
@@ -708,8 +714,8 @@ class T2IAdapterFolderProbe(FolderProbeBase):
 
 ############## register probe classes ######
 ModelProbe.register_probe("diffusers", ModelType.Main, PipelineFolderProbe)
-ModelProbe.register_probe("diffusers", ModelType.Vae, VaeFolderProbe)
-ModelProbe.register_probe("diffusers", ModelType.Lora, LoRAFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.VAE, VaeFolderProbe)
+ModelProbe.register_probe("diffusers", ModelType.LoRA, LoRAFolderProbe)
 ModelProbe.register_probe("diffusers", ModelType.TextualInversion, TextualInversionFolderProbe)
 ModelProbe.register_probe("diffusers", ModelType.ControlNet, ControlNetFolderProbe)
 ModelProbe.register_probe("diffusers", ModelType.IPAdapter, IPAdapterFolderProbe)
@@ -717,8 +723,8 @@ ModelProbe.register_probe("diffusers", ModelType.CLIPVision, CLIPVisionFolderPro
 ModelProbe.register_probe("diffusers", ModelType.T2IAdapter, T2IAdapterFolderProbe)
 
 ModelProbe.register_probe("checkpoint", ModelType.Main, PipelineCheckpointProbe)
-ModelProbe.register_probe("checkpoint", ModelType.Vae, VaeCheckpointProbe)
-ModelProbe.register_probe("checkpoint", ModelType.Lora, LoRACheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.VAE, VaeCheckpointProbe)
+ModelProbe.register_probe("checkpoint", ModelType.LoRA, LoRACheckpointProbe)
 ModelProbe.register_probe("checkpoint", ModelType.TextualInversion, TextualInversionCheckpointProbe)
 ModelProbe.register_probe("checkpoint", ModelType.ControlNet, ControlNetCheckpointProbe)
 ModelProbe.register_probe("checkpoint", ModelType.IPAdapter, IPAdapterCheckpointProbe)

invokeai/backend/model_manager/search.py

@@ -28,6 +28,7 @@ from typing import Callable, Optional, Set, Union
 
 from pydantic import BaseModel, Field
 
+from invokeai.app.services.config import InvokeAIAppConfig
 from invokeai.backend.util.logging import InvokeAILogger
 
 default_logger: Logger = InvokeAILogger.get_logger()
@@ -117,13 +118,10 @@ class ModelSearch(ModelSearchBase):
     """
 
     models_found: Set[Path] = Field(default_factory=set)
-    scanned_dirs: Set[Path] = Field(default_factory=set)
-    pruned_paths: Set[Path] = Field(default_factory=set)
+    config: InvokeAIAppConfig = InvokeAIAppConfig.get_config()
 
     def search_started(self) -> None:
         self.models_found = set()
-        self.scanned_dirs = set()
-        self.pruned_paths = set()
         if self.on_search_started:
             self.on_search_started(self._directory)
 
@@ -139,53 +137,53 @@ class ModelSearch(ModelSearchBase):
 
     def search(self, directory: Union[Path, str]) -> Set[Path]:
         self._directory = Path(directory)
+        if not self._directory.is_absolute():
+            self._directory = self.config.models_path / self._directory
         self.stats = SearchStats()  # zero out
         self.search_started()  # This will initialize _models_found to empty
-        self._walk_directory(directory)
+        self._walk_directory(self._directory)
         self.search_completed()
         return self.models_found
 
-    def _walk_directory(self, path: Union[Path, str]) -> None:
-        for root, dirs, files in os.walk(path, followlinks=True):
-            # don't descend into directories that start with a "."
-            # to avoid the Mac .DS_STORE issue.
-            if str(Path(root).name).startswith("."):
-                self.pruned_paths.add(Path(root))
-            if any(Path(root).is_relative_to(x) for x in self.pruned_paths):
-                continue
+    def _walk_directory(self, path: Union[Path, str], max_depth: int = 20) -> None:
+        absolute_path = Path(path)
+        if (
+            len(absolute_path.parts) - len(self._directory.parts) > max_depth
+            or not absolute_path.exists()
+            or absolute_path.parent in self.models_found
+        ):
+            return
+        entries = os.scandir(absolute_path.as_posix())
+        entries = [entry for entry in entries if not entry.name.startswith(".")]
+        dirs = [entry for entry in entries if entry.is_dir()]
+        file_names = [entry.name for entry in entries if entry.is_file()]
+        if any(
+            x in file_names
+            for x in [
+                "config.json",
+                "model_index.json",
+                "learned_embeds.bin",
+                "pytorch_lora_weights.bin",
+                "image_encoder.txt",
+            ]
+        ):
+            try:
+                self.model_found(absolute_path)
+                return
+            except KeyboardInterrupt:
+                raise
+            except Exception as e:
+                self.logger.warning(str(e))
+                return
 
-            self.stats.items_scanned += len(dirs) + len(files)
-            for d in dirs:
-                path = Path(root) / d
-                if path.parent in self.scanned_dirs:
-                    self.scanned_dirs.add(path)
-                    continue
-                if any(
-                    (path / x).exists()
-                    for x in [
-                        "config.json",
-                        "model_index.json",
-                        "learned_embeds.bin",
-                        "pytorch_lora_weights.bin",
-                        "image_encoder.txt",
-                    ]
-                ):
-                    self.scanned_dirs.add(path)
-                    try:
-                        self.model_found(path)
-                    except KeyboardInterrupt:
-                        raise
-                    except Exception as e:
-                        self.logger.warning(str(e))
+        for n in file_names:
+            if n.endswith((".ckpt", ".bin", ".pth", ".safetensors", ".pt")):
+                try:
+                    self.model_found(absolute_path / n)
+                except KeyboardInterrupt:
+                    raise
+                except Exception as e:
+                    self.logger.warning(str(e))
 
-            for f in files:
-                path = Path(root) / f
-                if path.parent in self.scanned_dirs:
-                    continue
-                if path.suffix in {".ckpt", ".bin", ".pth", ".safetensors", ".pt"}:
-                    try:
-                        self.model_found(path)
-                    except KeyboardInterrupt:
-                        raise
-                    except Exception as e:
-                        self.logger.warning(str(e))
+        for d in dirs:
+            self._walk_directory(absolute_path / d)

invokeai/backend/model_manager/util/select_hf_files.py

@@ -13,6 +13,7 @@ files_to_download = select_hf_model_files(metadata.files, variant='onnx')
 """
 
 import re
+from dataclasses import dataclass
 from pathlib import Path
 from typing import Dict, List, Optional, Set
 
@@ -34,7 +35,7 @@ def filter_files(
     The file list can be obtained from the `files` field of HuggingFaceMetadata,
     as defined in `invokeai.backend.model_manager.metadata.metadata_base`.
     """
-    variant = variant or ModelRepoVariant.DEFAULT
+    variant = variant or ModelRepoVariant.Default
     paths: List[Path] = []
     root = files[0].parts[0]
 
@@ -73,64 +74,81 @@ def filter_files(
     return sorted(_filter_by_variant(paths, variant))
 
 
+@dataclass
+class SubfolderCandidate:
+    path: Path
+    score: int
+
+
 def _filter_by_variant(files: List[Path], variant: ModelRepoVariant) -> Set[Path]:
     """Select the proper variant files from a list of HuggingFace repo_id paths."""
-    result = set()
-    basenames: Dict[Path, Path] = {}
+    result: set[Path] = set()
+    subfolder_weights: dict[Path, list[SubfolderCandidate]] = {}
     for path in files:
         if path.suffix in [".onnx", ".pb", ".onnx_data"]:
             if variant == ModelRepoVariant.ONNX:
                 result.add(path)
 
         elif "openvino_model" in path.name:
-            if variant == ModelRepoVariant.OPENVINO:
+            if variant == ModelRepoVariant.OpenVINO:
                 result.add(path)
 
         elif "flax_model" in path.name:
-            if variant == ModelRepoVariant.FLAX:
+            if variant == ModelRepoVariant.Flax:
                 result.add(path)
 
         elif path.suffix in [".json", ".txt"]:
             result.add(path)
 
-        elif path.suffix in [".bin", ".safetensors", ".pt", ".ckpt"] and variant in [
+        elif variant in [
             ModelRepoVariant.FP16,
             ModelRepoVariant.FP32,
-            ModelRepoVariant.DEFAULT,
-        ]:
-            parent = path.parent
-            suffixes = path.suffixes
-            if len(suffixes) == 2:
-                variant_label, suffix = suffixes
-                basename = parent / Path(path.stem).stem
-            else:
-                variant_label = ""
-                suffix = suffixes[0]
-                basename = parent / path.stem
+            ModelRepoVariant.Default,
+        ] and path.suffix in [".bin", ".safetensors", ".pt", ".ckpt"]:
+            # For weights files, we want to select the best one for each subfolder. For example, we may have multiple
+            # text encoders:
+            #
+            # - text_encoder/model.fp16.safetensors
+            # - text_encoder/model.safetensors
+            # - text_encoder/pytorch_model.bin
+            # - text_encoder/pytorch_model.fp16.bin
+            #
+            # We prefer safetensors over other file formats and an exact variant match. We'll score each file based on
+            # variant and format and select the best one.
 
-            if previous := basenames.get(basename):
-                if (
-                    previous.suffix != ".safetensors" and suffix == ".safetensors"
-                ):  # replace non-safetensors with safetensors when available
-                    basenames[basename] = path
-                if variant_label == f".{variant}":
-                    basenames[basename] = path
-                elif not variant_label and variant in [ModelRepoVariant.FP32, ModelRepoVariant.DEFAULT]:
-                    basenames[basename] = path
-            else:
-                basenames[basename] = path
+            parent = path.parent
+            score = 0
+
+            if path.suffix == ".safetensors":
+                score += 1
+
+            candidate_variant_label = path.suffixes[0] if len(path.suffixes) == 2 else None
+
+            # Some special handling is needed here if there is not an exact match and if we cannot infer the variant
+            # from the file name. In this case, we only give this file a point if the requested variant is FP32 or DEFAULT.
+            if candidate_variant_label == f".{variant}" or (
+                not candidate_variant_label and variant in [ModelRepoVariant.FP32, ModelRepoVariant.Default]
+            ):
+                score += 1
+
+            if parent not in subfolder_weights:
+                subfolder_weights[parent] = []
+
+            subfolder_weights[parent].append(SubfolderCandidate(path=path, score=score))
 
         else:
             continue
 
-    for v in basenames.values():
-        result.add(v)
+    for candidate_list in subfolder_weights.values():
+        highest_score_candidate = max(candidate_list, key=lambda candidate: candidate.score)
+        if highest_score_candidate:
+            result.add(highest_score_candidate.path)
 
     # If one of the architecture-related variants was specified and no files matched other than
     # config and text files then we return an empty list
     if (
         variant
-        and variant in [ModelRepoVariant.ONNX, ModelRepoVariant.OPENVINO, ModelRepoVariant.FLAX]
+        and variant in [ModelRepoVariant.ONNX, ModelRepoVariant.OpenVINO, ModelRepoVariant.Flax]
         and not any(variant.value in x.name for x in result)
     ):
         return set()

invokeai/backend/model_patcher.py

@@ -1,15 +1,16 @@
 # Copyright (c) 2024 Ryan Dick, Lincoln D. Stein, and the InvokeAI Development Team
 """These classes implement model patching with LoRAs and Textual Inversions."""
+
 from __future__ import annotations
 
 import pickle
 from contextlib import contextmanager
-from typing import Any, Dict, Iterator, List, Optional, Tuple
+from typing import Any, Dict, Iterator, List, Optional, Tuple, Union
 
 import numpy as np
 import torch
 from diffusers import OnnxRuntimeModel, UNet2DConditionModel
-from transformers import CLIPTextModel, CLIPTokenizer
+from transformers import CLIPTextModel, CLIPTextModelWithProjection, CLIPTokenizer
 
 from invokeai.app.shared.models import FreeUConfig
 from invokeai.backend.model_manager import AnyModel
@@ -168,7 +169,7 @@ class ModelPatcher:
     def apply_ti(
         cls,
         tokenizer: CLIPTokenizer,
-        text_encoder: CLIPTextModel,
+        text_encoder: Union[CLIPTextModel, CLIPTextModelWithProjection],
         ti_list: List[Tuple[str, TextualInversionModelRaw]],
     ) -> Iterator[Tuple[CLIPTokenizer, TextualInversionManager]]:
         init_tokens_count = None
@@ -265,7 +266,7 @@ class ModelPatcher:
     @contextmanager
     def apply_clip_skip(
         cls,
-        text_encoder: CLIPTextModel,
+        text_encoder: Union[CLIPTextModel, CLIPTextModelWithProjection],
         clip_skip: int,
     ) -> None:
         skipped_layers = []

invokeai/backend/stable_diffusion/__init__.py

@@ -1,15 +1,14 @@
 """
 Initialization file for the invokeai.backend.stable_diffusion package
 """
+
 from .diffusers_pipeline import PipelineIntermediateState, StableDiffusionGeneratorPipeline  # noqa: F401
 from .diffusion import InvokeAIDiffuserComponent  # noqa: F401
-from .diffusion.cross_attention_map_saving import AttentionMapSaver  # noqa: F401
 from .seamless import set_seamless  # noqa: F401
 
 __all__ = [
     "PipelineIntermediateState",
     "StableDiffusionGeneratorPipeline",
     "InvokeAIDiffuserComponent",
-    "AttentionMapSaver",
     "set_seamless",
 ]

invokeai/backend/stable_diffusion/diffusers_pipeline.py

@@ -12,7 +12,6 @@ import torch
 import torchvision.transforms as T
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
 from diffusers.models.controlnet import ControlNetModel
-from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion import StableDiffusionPipeline
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import KarrasDiffusionSchedulers
@@ -26,9 +25,9 @@ from invokeai.app.services.config import InvokeAIAppConfig
 from invokeai.backend.ip_adapter.ip_adapter import IPAdapter
 from invokeai.backend.ip_adapter.unet_patcher import UNetPatcher
 from invokeai.backend.stable_diffusion.diffusion.conditioning_data import ConditioningData
+from invokeai.backend.stable_diffusion.diffusion.shared_invokeai_diffusion import InvokeAIDiffuserComponent
 
 from ..util import auto_detect_slice_size, normalize_device
-from .diffusion import AttentionMapSaver, InvokeAIDiffuserComponent
 
 
 @dataclass
@@ -39,7 +38,6 @@ class PipelineIntermediateState:
     timestep: int
     latents: torch.Tensor
     predicted_original: Optional[torch.Tensor] = None
-    attention_map_saver: Optional[AttentionMapSaver] = None
 
 
 @dataclass
@@ -190,19 +188,6 @@ class T2IAdapterData:
     end_step_percent: float = Field(default=1.0)
 
 
-@dataclass
-class InvokeAIStableDiffusionPipelineOutput(StableDiffusionPipelineOutput):
-    r"""
-    Output class for InvokeAI's Stable Diffusion pipeline.
-
-    Args:
-        attention_map_saver (`AttentionMapSaver`): Object containing attention maps that can be displayed to the user
-         after generation completes. Optional.
-    """
-
-    attention_map_saver: Optional[AttentionMapSaver]
-
-
 class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion.
@@ -343,9 +328,9 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
         masked_latents: Optional[torch.Tensor] = None,
         gradient_mask: Optional[bool] = False,
         seed: Optional[int] = None,
-    ) -> tuple[torch.Tensor, Optional[AttentionMapSaver]]:
+    ) -> torch.Tensor:
         if init_timestep.shape[0] == 0:
-            return latents, None
+            return latents
 
         if additional_guidance is None:
             additional_guidance = []
@@ -385,7 +370,7 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
                 additional_guidance.append(AddsMaskGuidance(mask, orig_latents, self.scheduler, noise, gradient_mask))
 
         try:
-            latents, attention_map_saver = self.generate_latents_from_embeddings(
+            latents = self.generate_latents_from_embeddings(
                 latents,
                 timesteps,
                 conditioning_data,
@@ -402,7 +387,7 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
         if mask is not None and not gradient_mask:
             latents = torch.lerp(orig_latents, latents.to(dtype=orig_latents.dtype), mask.to(dtype=orig_latents.dtype))
 
-        return latents, attention_map_saver
+        return latents
 
     def generate_latents_from_embeddings(
         self,
@@ -415,23 +400,22 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
         ip_adapter_data: Optional[list[IPAdapterData]] = None,
         t2i_adapter_data: Optional[list[T2IAdapterData]] = None,
         callback: Callable[[PipelineIntermediateState], None] = None,
-    ):
+    ) -> torch.Tensor:
         self._adjust_memory_efficient_attention(latents)
         if additional_guidance is None:
             additional_guidance = []
 
         batch_size = latents.shape[0]
-        attention_map_saver: Optional[AttentionMapSaver] = None
 
         if timesteps.shape[0] == 0:
-            return latents, attention_map_saver
+            return latents
 
         ip_adapter_unet_patcher = None
-        if conditioning_data.extra is not None and conditioning_data.extra.wants_cross_attention_control:
+        extra_conditioning_info = conditioning_data.text_embeddings.extra_conditioning
+        if extra_conditioning_info is not None and extra_conditioning_info.wants_cross_attention_control:
             attn_ctx = self.invokeai_diffuser.custom_attention_context(
                 self.invokeai_diffuser.model,
-                extra_conditioning_info=conditioning_data.extra,
-                step_count=len(self.scheduler.timesteps),
+                extra_conditioning_info=extra_conditioning_info,
             )
             self.use_ip_adapter = False
         elif ip_adapter_data is not None:
@@ -482,13 +466,6 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
 
                 predicted_original = getattr(step_output, "pred_original_sample", None)
 
-                # TODO resuscitate attention map saving
-                # if i == len(timesteps)-1 and extra_conditioning_info is not None:
-                #    eos_token_index = extra_conditioning_info.tokens_count_including_eos_bos - 1
-                #    attention_map_token_ids = range(1, eos_token_index)
-                #    attention_map_saver = AttentionMapSaver(token_ids=attention_map_token_ids, latents_shape=latents.shape[-2:])
-                #    self.invokeai_diffuser.setup_attention_map_saving(attention_map_saver)
-
                 if callback is not None:
                     callback(
                         PipelineIntermediateState(
@@ -498,11 +475,10 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
                             timestep=int(t),
                             latents=latents,
                             predicted_original=predicted_original,
-                            attention_map_saver=attention_map_saver,
                         )
                     )
 
-            return latents, attention_map_saver
+            return latents
 
     @torch.inference_mode()
     def step(
@@ -544,15 +520,9 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
                     # Otherwise, set the IP-Adapter's scale to 0, so it has no effect.
                     ip_adapter_unet_patcher.set_scale(i, 0.0)
 
-        # Handle ControlNet(s) and T2I-Adapter(s)
+        # Handle ControlNet(s)
         down_block_additional_residuals = None
         mid_block_additional_residual = None
-        down_intrablock_additional_residuals = None
-        # if control_data is not None and t2i_adapter_data is not None:
-        # TODO(ryand): This is a limitation of the UNet2DConditionModel API, not a fundamental incompatibility
-        # between ControlNets and T2I-Adapters. We will try to fix this upstream in diffusers.
-        #    raise Exception("ControlNet(s) and T2I-Adapter(s) cannot be used simultaneously (yet).")
-        # elif control_data is not None:
         if control_data is not None:
             down_block_additional_residuals, mid_block_additional_residual = self.invokeai_diffuser.do_controlnet_step(
                 control_data=control_data,
@@ -562,7 +532,9 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
                 total_step_count=total_step_count,
                 conditioning_data=conditioning_data,
             )
-        # elif t2i_adapter_data is not None:
+
+        # Handle T2I-Adapter(s)
+        down_intrablock_additional_residuals = None
         if t2i_adapter_data is not None:
             accum_adapter_state = None
             for single_t2i_adapter_data in t2i_adapter_data:
@@ -588,7 +560,6 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
                     for idx, value in enumerate(single_t2i_adapter_data.adapter_state):
                         accum_adapter_state[idx] += value * t2i_adapter_weight
 
-            # down_block_additional_residuals = accum_adapter_state
             down_intrablock_additional_residuals = accum_adapter_state
 
         uc_noise_pred, c_noise_pred = self.invokeai_diffuser.do_unet_step(
@@ -597,7 +568,6 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
             step_index=step_index,
             total_step_count=total_step_count,
             conditioning_data=conditioning_data,
-            # extra:
             down_block_additional_residuals=down_block_additional_residuals,  # for ControlNet
             mid_block_additional_residual=mid_block_additional_residual,  # for ControlNet
             down_intrablock_additional_residuals=down_intrablock_additional_residuals,  # for T2I-Adapter

invokeai/backend/stable_diffusion/diffusion/__init__.py

@@ -1,6 +1,5 @@
 """
 Initialization file for invokeai.models.diffusion
 """
-from .cross_attention_control import InvokeAICrossAttentionMixin  # noqa: F401
-from .cross_attention_map_saving import AttentionMapSaver  # noqa: F401
+
 from .shared_invokeai_diffusion import InvokeAIDiffuserComponent  # noqa: F401

invokeai/backend/stable_diffusion/diffusion/conditioning_data.py

@@ -21,11 +21,7 @@ class ExtraConditioningInfo:
 @dataclass
 class BasicConditioningInfo:
     embeds: torch.Tensor
-    # TODO(ryand): Right now we awkwardly copy the extra conditioning info from here up to `ConditioningData`. This
-    # should only be stored in one place.
     extra_conditioning: Optional[ExtraConditioningInfo]
-    # weight: float
-    # mode: ConditioningAlgo
 
     def to(self, device, dtype=None):
         self.embeds = self.embeds.to(device=device, dtype=dtype)
@@ -83,7 +79,6 @@ class ConditioningData:
      ref [Common Diffusion Noise Schedules and Sample Steps are Flawed](https://arxiv.org/pdf/2305.08891.pdf)
     """
     guidance_rescale_multiplier: float = 0
-    extra: Optional[ExtraConditioningInfo] = None
     scheduler_args: dict[str, Any] = field(default_factory=dict)
     """
     Additional arguments to pass to invokeai_diffuser.do_latent_postprocessing().

invokeai/backend/stable_diffusion/diffusion/cross_attention_control.py

@@ -3,19 +3,13 @@
 
 
 import enum
-import math
 from dataclasses import dataclass, field
-from typing import Callable, Optional
+from typing import Optional
 
-import diffusers
-import psutil
 import torch
 from compel.cross_attention_control import Arguments
-from diffusers.models.attention_processor import Attention, AttentionProcessor, AttnProcessor, SlicedAttnProcessor
+from diffusers.models.attention_processor import Attention, SlicedAttnProcessor
 from diffusers.models.unets.unet_2d_condition import UNet2DConditionModel
-from torch import nn
-
-import invokeai.backend.util.logging as logger
 
 from ...util import torch_dtype
 
@@ -25,72 +19,14 @@ class CrossAttentionType(enum.Enum):
     TOKENS = 2
 
 
-class Context:
-    cross_attention_mask: Optional[torch.Tensor]
-    cross_attention_index_map: Optional[torch.Tensor]
-
-    class Action(enum.Enum):
-        NONE = 0
-        SAVE = (1,)
-        APPLY = 2
-
-    def __init__(self, arguments: Arguments, step_count: int):
+class CrossAttnControlContext:
+    def __init__(self, arguments: Arguments):
         """
         :param arguments: Arguments for the cross-attention control process
-        :param step_count: The absolute total number of steps of diffusion (for img2img this is likely larger than the number of steps that will actually run)
         """
-        self.cross_attention_mask = None
-        self.cross_attention_index_map = None
-        self.self_cross_attention_action = Context.Action.NONE
-        self.tokens_cross_attention_action = Context.Action.NONE
+        self.cross_attention_mask: Optional[torch.Tensor] = None
+        self.cross_attention_index_map: Optional[torch.Tensor] = None
         self.arguments = arguments
-        self.step_count = step_count
-
-        self.self_cross_attention_module_identifiers = []
-        self.tokens_cross_attention_module_identifiers = []
-
-        self.saved_cross_attention_maps = {}
-
-        self.clear_requests(cleanup=True)
-
-    def register_cross_attention_modules(self, model):
-        for name, _module in get_cross_attention_modules(model, CrossAttentionType.SELF):
-            if name in self.self_cross_attention_module_identifiers:
-                raise AssertionError(f"name {name} cannot appear more than once")
-            self.self_cross_attention_module_identifiers.append(name)
-        for name, _module in get_cross_attention_modules(model, CrossAttentionType.TOKENS):
-            if name in self.tokens_cross_attention_module_identifiers:
-                raise AssertionError(f"name {name} cannot appear more than once")
-            self.tokens_cross_attention_module_identifiers.append(name)
-
-    def request_save_attention_maps(self, cross_attention_type: CrossAttentionType):
-        if cross_attention_type == CrossAttentionType.SELF:
-            self.self_cross_attention_action = Context.Action.SAVE
-        else:
-            self.tokens_cross_attention_action = Context.Action.SAVE
-
-    def request_apply_saved_attention_maps(self, cross_attention_type: CrossAttentionType):
-        if cross_attention_type == CrossAttentionType.SELF:
-            self.self_cross_attention_action = Context.Action.APPLY
-        else:
-            self.tokens_cross_attention_action = Context.Action.APPLY
-
-    def is_tokens_cross_attention(self, module_identifier) -> bool:
-        return module_identifier in self.tokens_cross_attention_module_identifiers
-
-    def get_should_save_maps(self, module_identifier: str) -> bool:
-        if module_identifier in self.self_cross_attention_module_identifiers:
-            return self.self_cross_attention_action == Context.Action.SAVE
-        elif module_identifier in self.tokens_cross_attention_module_identifiers:
-            return self.tokens_cross_attention_action == Context.Action.SAVE
-        return False
-
-    def get_should_apply_saved_maps(self, module_identifier: str) -> bool:
-        if module_identifier in self.self_cross_attention_module_identifiers:
-            return self.self_cross_attention_action == Context.Action.APPLY
-        elif module_identifier in self.tokens_cross_attention_module_identifiers:
-            return self.tokens_cross_attention_action == Context.Action.APPLY
-        return False
 
     def get_active_cross_attention_control_types_for_step(
         self, percent_through: float = None
@@ -111,219 +47,8 @@ class Context:
             to_control.append(CrossAttentionType.TOKENS)
         return to_control
 
-    def save_slice(
-        self,
-        identifier: str,
-        slice: torch.Tensor,
-        dim: Optional[int],
-        offset: int,
-        slice_size: Optional[int],
-    ):
-        if identifier not in self.saved_cross_attention_maps:
-            self.saved_cross_attention_maps[identifier] = {
-                "dim": dim,
-                "slice_size": slice_size,
-                "slices": {offset or 0: slice},
-            }
-        else:
-            self.saved_cross_attention_maps[identifier]["slices"][offset or 0] = slice
 
-    def get_slice(
-        self,
-        identifier: str,
-        requested_dim: Optional[int],
-        requested_offset: int,
-        slice_size: int,
-    ):
-        saved_attention_dict = self.saved_cross_attention_maps[identifier]
-        if requested_dim is None:
-            if saved_attention_dict["dim"] is not None:
-                raise RuntimeError(f"dim mismatch: expected dim=None, have {saved_attention_dict['dim']}")
-            return saved_attention_dict["slices"][0]
-
-        if saved_attention_dict["dim"] == requested_dim:
-            if slice_size != saved_attention_dict["slice_size"]:
-                raise RuntimeError(
-                    f"slice_size mismatch: expected slice_size={slice_size}, have {saved_attention_dict['slice_size']}"
-                )
-            return saved_attention_dict["slices"][requested_offset]
-
-        if saved_attention_dict["dim"] is None:
-            whole_saved_attention = saved_attention_dict["slices"][0]
-            if requested_dim == 0:
-                return whole_saved_attention[requested_offset : requested_offset + slice_size]
-            elif requested_dim == 1:
-                return whole_saved_attention[:, requested_offset : requested_offset + slice_size]
-
-        raise RuntimeError(f"Cannot convert dim {saved_attention_dict['dim']} to requested dim {requested_dim}")
-
-    def get_slicing_strategy(self, identifier: str) -> tuple[Optional[int], Optional[int]]:
-        saved_attention = self.saved_cross_attention_maps.get(identifier, None)
-        if saved_attention is None:
-            return None, None
-        return saved_attention["dim"], saved_attention["slice_size"]
-
-    def clear_requests(self, cleanup=True):
-        self.tokens_cross_attention_action = Context.Action.NONE
-        self.self_cross_attention_action = Context.Action.NONE
-        if cleanup:
-            self.saved_cross_attention_maps = {}
-
-    def offload_saved_attention_slices_to_cpu(self):
-        for _key, map_dict in self.saved_cross_attention_maps.items():
-            for offset, slice in map_dict["slices"].items():
-                map_dict[offset] = slice.to("cpu")
-
-
-class InvokeAICrossAttentionMixin:
-    """
-    Enable InvokeAI-flavoured Attention calculation, which does aggressive low-memory slicing and calls
-    through both to an attention_slice_wrangler and a slicing_strategy_getter for custom attention map wrangling
-    and dymamic slicing strategy selection.
-    """
-
-    def __init__(self):
-        self.mem_total_gb = psutil.virtual_memory().total // (1 << 30)
-        self.attention_slice_wrangler = None
-        self.slicing_strategy_getter = None
-        self.attention_slice_calculated_callback = None
-
-    def set_attention_slice_wrangler(
-        self,
-        wrangler: Optional[Callable[[nn.Module, torch.Tensor, int, int, int], torch.Tensor]],
-    ):
-        """
-        Set custom attention calculator to be called when attention is calculated
-        :param wrangler: Callback, with args (module, suggested_attention_slice, dim, offset, slice_size),
-        which returns either the suggested_attention_slice or an adjusted equivalent.
-            `module` is the current Attention module for which the callback is being invoked.
-            `suggested_attention_slice` is the default-calculated attention slice
-            `dim` is -1 if the attenion map has not been sliced, or 0 or 1 for dimension-0 or dimension-1 slicing.
-                If `dim` is >= 0, `offset` and `slice_size` specify the slice start and length.
-
-        Pass None to use the default attention calculation.
-        :return:
-        """
-        self.attention_slice_wrangler = wrangler
-
-    def set_slicing_strategy_getter(self, getter: Optional[Callable[[nn.Module], tuple[int, int]]]):
-        self.slicing_strategy_getter = getter
-
-    def set_attention_slice_calculated_callback(self, callback: Optional[Callable[[torch.Tensor], None]]):
-        self.attention_slice_calculated_callback = callback
-
-    def einsum_lowest_level(self, query, key, value, dim, offset, slice_size):
-        # calculate attention scores
-        # attention_scores = torch.einsum('b i d, b j d -> b i j', q, k)
-        attention_scores = torch.baddbmm(
-            torch.empty(
-                query.shape[0],
-                query.shape[1],
-                key.shape[1],
-                dtype=query.dtype,
-                device=query.device,
-            ),
-            query,
-            key.transpose(-1, -2),
-            beta=0,
-            alpha=self.scale,
-        )
-
-        # calculate attention slice by taking the best scores for each latent pixel
-        default_attention_slice = attention_scores.softmax(dim=-1, dtype=attention_scores.dtype)
-        attention_slice_wrangler = self.attention_slice_wrangler
-        if attention_slice_wrangler is not None:
-            attention_slice = attention_slice_wrangler(self, default_attention_slice, dim, offset, slice_size)
-        else:
-            attention_slice = default_attention_slice
-
-        if self.attention_slice_calculated_callback is not None:
-            self.attention_slice_calculated_callback(attention_slice, dim, offset, slice_size)
-
-        hidden_states = torch.bmm(attention_slice, value)
-        return hidden_states
-
-    def einsum_op_slice_dim0(self, q, k, v, slice_size):
-        r = torch.zeros(q.shape[0], q.shape[1], v.shape[2], device=q.device, dtype=q.dtype)
-        for i in range(0, q.shape[0], slice_size):
-            end = i + slice_size
-            r[i:end] = self.einsum_lowest_level(q[i:end], k[i:end], v[i:end], dim=0, offset=i, slice_size=slice_size)
-        return r
-
-    def einsum_op_slice_dim1(self, q, k, v, slice_size):
-        r = torch.zeros(q.shape[0], q.shape[1], v.shape[2], device=q.device, dtype=q.dtype)
-        for i in range(0, q.shape[1], slice_size):
-            end = i + slice_size
-            r[:, i:end] = self.einsum_lowest_level(q[:, i:end], k, v, dim=1, offset=i, slice_size=slice_size)
-        return r
-
-    def einsum_op_mps_v1(self, q, k, v):
-        if q.shape[1] <= 4096:  # (512x512) max q.shape[1]: 4096
-            return self.einsum_lowest_level(q, k, v, None, None, None)
-        else:
-            slice_size = math.floor(2**30 / (q.shape[0] * q.shape[1]))
-            return self.einsum_op_slice_dim1(q, k, v, slice_size)
-
-    def einsum_op_mps_v2(self, q, k, v):
-        if self.mem_total_gb > 8 and q.shape[1] <= 4096:
-            return self.einsum_lowest_level(q, k, v, None, None, None)
-        else:
-            return self.einsum_op_slice_dim0(q, k, v, 1)
-
-    def einsum_op_tensor_mem(self, q, k, v, max_tensor_mb):
-        size_mb = q.shape[0] * q.shape[1] * k.shape[1] * q.element_size() // (1 << 20)
-        if size_mb <= max_tensor_mb:
-            return self.einsum_lowest_level(q, k, v, None, None, None)
-        div = 1 << int((size_mb - 1) / max_tensor_mb).bit_length()
-        if div <= q.shape[0]:
-            return self.einsum_op_slice_dim0(q, k, v, q.shape[0] // div)
-        return self.einsum_op_slice_dim1(q, k, v, max(q.shape[1] // div, 1))
-
-    def einsum_op_cuda(self, q, k, v):
-        # check if we already have a slicing strategy (this should only happen during cross-attention controlled generation)
-        slicing_strategy_getter = self.slicing_strategy_getter
-        if slicing_strategy_getter is not None:
-            (dim, slice_size) = slicing_strategy_getter(self)
-            if dim is not None:
-                # print("using saved slicing strategy with dim", dim, "slice size", slice_size)
-                if dim == 0:
-                    return self.einsum_op_slice_dim0(q, k, v, slice_size)
-                elif dim == 1:
-                    return self.einsum_op_slice_dim1(q, k, v, slice_size)
-
-        # fallback for when there is no saved strategy, or saved strategy does not slice
-        mem_free_total = get_mem_free_total(q.device)
-        # Divide factor of safety as there's copying and fragmentation
-        return self.einsum_op_tensor_mem(q, k, v, mem_free_total / 3.3 / (1 << 20))
-
-    def get_invokeai_attention_mem_efficient(self, q, k, v):
-        if q.device.type == "cuda":
-            # print("in get_attention_mem_efficient with q shape", q.shape, ", k shape", k.shape, ", free memory is", get_mem_free_total(q.device))
-            return self.einsum_op_cuda(q, k, v)
-
-        if q.device.type == "mps" or q.device.type == "cpu":
-            if self.mem_total_gb >= 32:
-                return self.einsum_op_mps_v1(q, k, v)
-            return self.einsum_op_mps_v2(q, k, v)
-
-        # Smaller slices are faster due to L2/L3/SLC caches.
-        # Tested on i7 with 8MB L3 cache.
-        return self.einsum_op_tensor_mem(q, k, v, 32)
-
-
-def restore_default_cross_attention(
-    model,
-    is_running_diffusers: bool,
-    restore_attention_processor: Optional[AttentionProcessor] = None,
-):
-    if is_running_diffusers:
-        unet = model
-        unet.set_attn_processor(restore_attention_processor or AttnProcessor())
-    else:
-        remove_attention_function(model)
-
-
-def setup_cross_attention_control_attention_processors(unet: UNet2DConditionModel, context: Context):
+def setup_cross_attention_control_attention_processors(unet: UNet2DConditionModel, context: CrossAttnControlContext):
     """
     Inject attention parameters and functions into the passed in model to enable cross attention editing.
 
@@ -362,170 +87,6 @@ def setup_cross_attention_control_attention_processors(unet: UNet2DConditionMode
         unet.set_attn_processor(SlicedSwapCrossAttnProcesser(slice_size=slice_size))
 
 
-def get_cross_attention_modules(model, which: CrossAttentionType) -> list[tuple[str, InvokeAICrossAttentionMixin]]:
-    cross_attention_class: type = InvokeAIDiffusersCrossAttention
-    which_attn = "attn1" if which is CrossAttentionType.SELF else "attn2"
-    attention_module_tuples = [
-        (name, module)
-        for name, module in model.named_modules()
-        if isinstance(module, cross_attention_class) and which_attn in name
-    ]
-    cross_attention_modules_in_model_count = len(attention_module_tuples)
-    expected_count = 16
-    if cross_attention_modules_in_model_count != expected_count:
-        # non-fatal error but .swap() won't work.
-        logger.error(
-            f"Error! CrossAttentionControl found an unexpected number of {cross_attention_class} modules in the model "
-            f"(expected {expected_count}, found {cross_attention_modules_in_model_count}). Either monkey-patching "
-            "failed or some assumption has changed about the structure of the model itself. Please fix the "
-            f"monkey-patching, and/or update the {expected_count} above to an appropriate number, and/or find and "
-            "inform someone who knows what it means. This error is non-fatal, but it is likely that .swap() and "
-            "attention map display will not work properly until it is fixed."
-        )
-    return attention_module_tuples
-
-
-def inject_attention_function(unet, context: Context):
-    # ORIGINAL SOURCE CODE: https://github.com/huggingface/diffusers/blob/91ddd2a25b848df0fa1262d4f1cd98c7ccb87750/src/diffusers/models/attention.py#L276
-
-    def attention_slice_wrangler(module, suggested_attention_slice: torch.Tensor, dim, offset, slice_size):
-        # memory_usage = suggested_attention_slice.element_size() * suggested_attention_slice.nelement()
-
-        attention_slice = suggested_attention_slice
-
-        if context.get_should_save_maps(module.identifier):
-            # print(module.identifier, "saving suggested_attention_slice of shape",
-            #      suggested_attention_slice.shape, "dim", dim, "offset", offset)
-            slice_to_save = attention_slice.to("cpu") if dim is not None else attention_slice
-            context.save_slice(
-                module.identifier,
-                slice_to_save,
-                dim=dim,
-                offset=offset,
-                slice_size=slice_size,
-            )
-        elif context.get_should_apply_saved_maps(module.identifier):
-            # print(module.identifier, "applying saved attention slice for dim", dim, "offset", offset)
-            saved_attention_slice = context.get_slice(module.identifier, dim, offset, slice_size)
-
-            # slice may have been offloaded to CPU
-            saved_attention_slice = saved_attention_slice.to(suggested_attention_slice.device)
-
-            if context.is_tokens_cross_attention(module.identifier):
-                index_map = context.cross_attention_index_map
-                remapped_saved_attention_slice = torch.index_select(saved_attention_slice, -1, index_map)
-                this_attention_slice = suggested_attention_slice
-
-                mask = context.cross_attention_mask.to(torch_dtype(suggested_attention_slice.device))
-                saved_mask = mask
-                this_mask = 1 - mask
-                attention_slice = remapped_saved_attention_slice * saved_mask + this_attention_slice * this_mask
-            else:
-                # just use everything
-                attention_slice = saved_attention_slice
-
-        return attention_slice
-
-    cross_attention_modules = get_cross_attention_modules(
-        unet, CrossAttentionType.TOKENS
-    ) + get_cross_attention_modules(unet, CrossAttentionType.SELF)
-    for identifier, module in cross_attention_modules:
-        module.identifier = identifier
-        try:
-            module.set_attention_slice_wrangler(attention_slice_wrangler)
-            module.set_slicing_strategy_getter(lambda module: context.get_slicing_strategy(identifier))  # noqa: B023
-        except AttributeError as e:
-            if is_attribute_error_about(e, "set_attention_slice_wrangler"):
-                print(f"TODO: implement set_attention_slice_wrangler for {type(module)}")  # TODO
-            else:
-                raise
-
-
-def remove_attention_function(unet):
-    cross_attention_modules = get_cross_attention_modules(
-        unet, CrossAttentionType.TOKENS
-    ) + get_cross_attention_modules(unet, CrossAttentionType.SELF)
-    for _identifier, module in cross_attention_modules:
-        try:
-            # clear wrangler callback
-            module.set_attention_slice_wrangler(None)
-            module.set_slicing_strategy_getter(None)
-        except AttributeError as e:
-            if is_attribute_error_about(e, "set_attention_slice_wrangler"):
-                print(f"TODO: implement set_attention_slice_wrangler for {type(module)}")
-            else:
-                raise
-
-
-def is_attribute_error_about(error: AttributeError, attribute: str):
-    if hasattr(error, "name"):  # Python 3.10
-        return error.name == attribute
-    else:  # Python 3.9
-        return attribute in str(error)
-
-
-def get_mem_free_total(device):
-    # only on cuda
-    if not torch.cuda.is_available():
-        return None
-    stats = torch.cuda.memory_stats(device)
-    mem_active = stats["active_bytes.all.current"]
-    mem_reserved = stats["reserved_bytes.all.current"]
-    mem_free_cuda, _ = torch.cuda.mem_get_info(device)
-    mem_free_torch = mem_reserved - mem_active
-    mem_free_total = mem_free_cuda + mem_free_torch
-    return mem_free_total
-
-
-class InvokeAIDiffusersCrossAttention(diffusers.models.attention.Attention, InvokeAICrossAttentionMixin):
-    def __init__(self, **kwargs):
-        super().__init__(**kwargs)
-        InvokeAICrossAttentionMixin.__init__(self)
-
-    def _attention(self, query, key, value, attention_mask=None):
-        # default_result = super()._attention(query,  key, value)
-        if attention_mask is not None:
-            print(f"{type(self).__name__} ignoring passed-in attention_mask")
-        attention_result = self.get_invokeai_attention_mem_efficient(query, key, value)
-
-        hidden_states = self.reshape_batch_dim_to_heads(attention_result)
-        return hidden_states
-
-
-## 🧨diffusers implementation follows
-
-
-"""
-# base implementation
-
-class AttnProcessor:
-    def __call__(self, attn: Attention, hidden_states, encoder_hidden_states=None, attention_mask=None):
-        batch_size, sequence_length, _ = hidden_states.shape
-        attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length)
-
-        query = attn.to_q(hidden_states)
-        query = attn.head_to_batch_dim(query)
-
-        encoder_hidden_states = encoder_hidden_states if encoder_hidden_states is not None else hidden_states
-        key = attn.to_k(encoder_hidden_states)
-        value = attn.to_v(encoder_hidden_states)
-        key = attn.head_to_batch_dim(key)
-        value = attn.head_to_batch_dim(value)
-
-        attention_probs = attn.get_attention_scores(query, key, attention_mask)
-        hidden_states = torch.bmm(attention_probs, value)
-        hidden_states = attn.batch_to_head_dim(hidden_states)
-
-        # linear proj
-        hidden_states = attn.to_out[0](hidden_states)
-        # dropout
-        hidden_states = attn.to_out[1](hidden_states)
-
-        return hidden_states
-
-"""
-
-
 @dataclass
 class SwapCrossAttnContext:
     modified_text_embeddings: torch.Tensor
@@ -533,18 +94,6 @@ class SwapCrossAttnContext:
     mask: torch.Tensor  # in the target space of the index_map
     cross_attention_types_to_do: list[CrossAttentionType] = field(default_factory=list)
 
-    def __int__(
-        self,
-        cac_types_to_do: [CrossAttentionType],
-        modified_text_embeddings: torch.Tensor,
-        index_map: torch.Tensor,
-        mask: torch.Tensor,
-    ):
-        self.cross_attention_types_to_do = cac_types_to_do
-        self.modified_text_embeddings = modified_text_embeddings
-        self.index_map = index_map
-        self.mask = mask
-
     def wants_cross_attention_control(self, attn_type: CrossAttentionType) -> bool:
         return attn_type in self.cross_attention_types_to_do
 

invokeai/backend/stable_diffusion/diffusion/cross_attention_map_saving.py

@@ -1,100 +0,0 @@
-import math
-from typing import Optional
-
-import torch
-from PIL import Image
-from torchvision.transforms.functional import InterpolationMode
-from torchvision.transforms.functional import resize as tv_resize
-
-
-class AttentionMapSaver:
-    def __init__(self, token_ids: range, latents_shape: torch.Size):
-        self.token_ids = token_ids
-        self.latents_shape = latents_shape
-        # self.collated_maps = #torch.zeros([len(token_ids), latents_shape[0], latents_shape[1]])
-        self.collated_maps: dict[str, torch.Tensor] = {}
-
-    def clear_maps(self):
-        self.collated_maps = {}
-
-    def add_attention_maps(self, maps: torch.Tensor, key: str):
-        """
-        Accumulate the given attention maps and store by summing with existing maps at the passed-in key (if any).
-        :param maps: Attention maps to store. Expected shape [A, (H*W), N] where A is attention heads count, H and W are the map size (fixed per-key) and N is the number of tokens (typically 77).
-        :param key: Storage key. If a map already exists for this key it will be summed with the incoming data. In this case the maps sizes (H and W) should match.
-        :return: None
-        """
-        key_and_size = f"{key}_{maps.shape[1]}"
-
-        # extract desired tokens
-        maps = maps[:, :, self.token_ids]
-
-        # merge attention heads to a single map per token
-        maps = torch.sum(maps, 0)
-
-        # store
-        if key_and_size not in self.collated_maps:
-            self.collated_maps[key_and_size] = torch.zeros_like(maps, device="cpu")
-        self.collated_maps[key_and_size] += maps.cpu()
-
-    def write_maps_to_disk(self, path: str):
-        pil_image = self.get_stacked_maps_image()
-        if pil_image is not None:
-            pil_image.save(path, "PNG")
-
-    def get_stacked_maps_image(self) -> Optional[Image.Image]:
-        """
-        Scale all collected attention maps to the same size, blend them together and return as an image.
-        :return: An image containing a vertical stack of blended attention maps, one for each requested token.
-        """
-        num_tokens = len(self.token_ids)
-        if num_tokens == 0:
-            return None
-
-        latents_height = self.latents_shape[0]
-        latents_width = self.latents_shape[1]
-
-        merged = None
-
-        for _key, maps in self.collated_maps.items():
-            # maps has shape [(H*W), N] for N tokens
-            # but we want [N, H, W]
-            this_scale_factor = math.sqrt(maps.shape[0] / (latents_width * latents_height))
-            this_maps_height = int(float(latents_height) * this_scale_factor)
-            this_maps_width = int(float(latents_width) * this_scale_factor)
-            # and we need to do some dimension juggling
-            maps = torch.reshape(
-                torch.swapdims(maps, 0, 1),
-                [num_tokens, this_maps_height, this_maps_width],
-            )
-
-            # scale to output size if necessary
-            if this_scale_factor != 1:
-                maps = tv_resize(maps, [latents_height, latents_width], InterpolationMode.BICUBIC)
-
-            # normalize
-            maps_min = torch.min(maps)
-            maps_range = torch.max(maps) - maps_min
-            # print(f"map {key} size {[this_maps_width, this_maps_height]} range {[maps_min, maps_min + maps_range]}")
-            maps_normalized = (maps - maps_min) / maps_range
-            # expand to (-0.1, 1.1) and clamp
-            maps_normalized_expanded = maps_normalized * 1.1 - 0.05
-            maps_normalized_expanded_clamped = torch.clamp(maps_normalized_expanded, 0, 1)
-
-            # merge together, producing a vertical stack
-            maps_stacked = torch.reshape(
-                maps_normalized_expanded_clamped,
-                [num_tokens * latents_height, latents_width],
-            )
-
-            if merged is None:
-                merged = maps_stacked
-            else:
-                # screen blend
-                merged = 1 - (1 - maps_stacked) * (1 - merged)
-
-        if merged is None:
-            return None
-
-        merged_bytes = merged.mul(0xFF).byte()
-        return Image.fromarray(merged_bytes.numpy(), mode="L")

invokeai/backend/stable_diffusion/diffusion/shared_invokeai_diffusion.py

@@ -17,13 +17,11 @@ from invokeai.backend.stable_diffusion.diffusion.conditioning_data import (
 )
 
 from .cross_attention_control import (
-    Context,
     CrossAttentionType,
+    CrossAttnControlContext,
     SwapCrossAttnContext,
-    get_cross_attention_modules,
     setup_cross_attention_control_attention_processors,
 )
-from .cross_attention_map_saving import AttentionMapSaver
 
 ModelForwardCallback: TypeAlias = Union[
     # x, t, conditioning, Optional[cross-attention kwargs]
@@ -69,14 +67,12 @@ class InvokeAIDiffuserComponent:
         self,
         unet: UNet2DConditionModel,
         extra_conditioning_info: Optional[ExtraConditioningInfo],
-        step_count: int,
     ):
         old_attn_processors = unet.attn_processors
 
         try:
-            self.cross_attention_control_context = Context(
+            self.cross_attention_control_context = CrossAttnControlContext(
                 arguments=extra_conditioning_info.cross_attention_control_args,
-                step_count=step_count,
             )
             setup_cross_attention_control_attention_processors(
                 unet,
@@ -87,27 +83,6 @@ class InvokeAIDiffuserComponent:
         finally:
             self.cross_attention_control_context = None
             unet.set_attn_processor(old_attn_processors)
-            # TODO resuscitate attention map saving
-            # self.remove_attention_map_saving()
-
-    def setup_attention_map_saving(self, saver: AttentionMapSaver):
-        def callback(slice, dim, offset, slice_size, key):
-            if dim is not None:
-                # sliced tokens attention map saving is not implemented
-                return
-            saver.add_attention_maps(slice, key)
-
-        tokens_cross_attention_modules = get_cross_attention_modules(self.model, CrossAttentionType.TOKENS)
-        for identifier, module in tokens_cross_attention_modules:
-            key = "down" if identifier.startswith("down") else "up" if identifier.startswith("up") else "mid"
-            module.set_attention_slice_calculated_callback(
-                lambda slice, dim, offset, slice_size, key=key: callback(slice, dim, offset, slice_size, key)
-            )
-
-    def remove_attention_map_saving(self):
-        tokens_cross_attention_modules = get_cross_attention_modules(self.model, CrossAttentionType.TOKENS)
-        for _, module in tokens_cross_attention_modules:
-            module.set_attention_slice_calculated_callback(None)
 
     def do_controlnet_step(
         self,
@@ -224,51 +199,47 @@ class InvokeAIDiffuserComponent:
         self,
         sample: torch.Tensor,
         timestep: torch.Tensor,
-        conditioning_data,  # TODO: type
+        conditioning_data: ConditioningData,
         step_index: int,
         total_step_count: int,
-        **kwargs,
+        down_block_additional_residuals: Optional[torch.Tensor] = None,  # for ControlNet
+        mid_block_additional_residual: Optional[torch.Tensor] = None,  # for ControlNet
+        down_intrablock_additional_residuals: Optional[torch.Tensor] = None,  # for T2I-Adapter
     ):
         cross_attention_control_types_to_do = []
-        context: Context = self.cross_attention_control_context
         if self.cross_attention_control_context is not None:
             percent_through = step_index / total_step_count
-            cross_attention_control_types_to_do = context.get_active_cross_attention_control_types_for_step(
-                percent_through
+            cross_attention_control_types_to_do = (
+                self.cross_attention_control_context.get_active_cross_attention_control_types_for_step(percent_through)
             )
-
         wants_cross_attention_control = len(cross_attention_control_types_to_do) > 0
 
-        if wants_cross_attention_control:
-            (
-                unconditioned_next_x,
-                conditioned_next_x,
-            ) = self._apply_cross_attention_controlled_conditioning(
-                sample,
-                timestep,
-                conditioning_data,
-                cross_attention_control_types_to_do,
-                **kwargs,
-            )
-        elif self.sequential_guidance:
+        if wants_cross_attention_control or self.sequential_guidance:
+            # If wants_cross_attention_control is True, we force the sequential mode to be used, because cross-attention
+            # control is currently only supported in sequential mode.
             (
                 unconditioned_next_x,
                 conditioned_next_x,
             ) = self._apply_standard_conditioning_sequentially(
-                sample,
-                timestep,
-                conditioning_data,
-                **kwargs,
+                x=sample,
+                sigma=timestep,
+                conditioning_data=conditioning_data,
+                cross_attention_control_types_to_do=cross_attention_control_types_to_do,
+                down_block_additional_residuals=down_block_additional_residuals,
+                mid_block_additional_residual=mid_block_additional_residual,
+                down_intrablock_additional_residuals=down_intrablock_additional_residuals,
             )
         else:
             (
                 unconditioned_next_x,
                 conditioned_next_x,
             ) = self._apply_standard_conditioning(
-                sample,
-                timestep,
-                conditioning_data,
-                **kwargs,
+                x=sample,
+                sigma=timestep,
+                conditioning_data=conditioning_data,
+                down_block_additional_residuals=down_block_additional_residuals,
+                mid_block_additional_residual=mid_block_additional_residual,
+                down_intrablock_additional_residuals=down_intrablock_additional_residuals,
             )
 
         return unconditioned_next_x, conditioned_next_x
@@ -335,7 +306,15 @@ class InvokeAIDiffuserComponent:
 
     # methods below are called from do_diffusion_step and should be considered private to this class.
 
-    def _apply_standard_conditioning(self, x, sigma, conditioning_data: ConditioningData, **kwargs):
+    def _apply_standard_conditioning(
+        self,
+        x,
+        sigma,
+        conditioning_data: ConditioningData,
+        down_block_additional_residuals: Optional[torch.Tensor] = None,  # for ControlNet
+        mid_block_additional_residual: Optional[torch.Tensor] = None,  # for ControlNet
+        down_intrablock_additional_residuals: Optional[torch.Tensor] = None,  # for T2I-Adapter
+    ):
         """Runs the conditioned and unconditioned UNet forward passes in a single batch for faster inference speed at
         the cost of higher memory usage.
         """
@@ -383,8 +362,10 @@ class InvokeAIDiffuserComponent:
             both_conditionings,
             cross_attention_kwargs=cross_attention_kwargs,
             encoder_attention_mask=encoder_attention_mask,
+            down_block_additional_residuals=down_block_additional_residuals,
+            mid_block_additional_residual=mid_block_additional_residual,
+            down_intrablock_additional_residuals=down_intrablock_additional_residuals,
             added_cond_kwargs=added_cond_kwargs,
-            **kwargs,
         )
         unconditioned_next_x, conditioned_next_x = both_results.chunk(2)
         return unconditioned_next_x, conditioned_next_x
@@ -394,14 +375,17 @@ class InvokeAIDiffuserComponent:
         x: torch.Tensor,
         sigma,
         conditioning_data: ConditioningData,
-        **kwargs,
+        cross_attention_control_types_to_do: list[CrossAttentionType],
+        down_block_additional_residuals: Optional[torch.Tensor] = None,  # for ControlNet
+        mid_block_additional_residual: Optional[torch.Tensor] = None,  # for ControlNet
+        down_intrablock_additional_residuals: Optional[torch.Tensor] = None,  # for T2I-Adapter
     ):
         """Runs the conditioned and unconditioned UNet forward passes sequentially for lower memory usage at the cost of
         slower execution speed.
         """
-        # low-memory sequential path
+        # Since we are running the conditioned and unconditioned passes sequentially, we need to split the ControlNet
+        # and T2I-Adapter residuals into two chunks.
         uncond_down_block, cond_down_block = None, None
-        down_block_additional_residuals = kwargs.pop("down_block_additional_residuals", None)
         if down_block_additional_residuals is not None:
             uncond_down_block, cond_down_block = [], []
             for down_block in down_block_additional_residuals:
@@ -410,7 +394,6 @@ class InvokeAIDiffuserComponent:
                 cond_down_block.append(_cond_down)
 
         uncond_down_intrablock, cond_down_intrablock = None, None
-        down_intrablock_additional_residuals = kwargs.pop("down_intrablock_additional_residuals", None)
         if down_intrablock_additional_residuals is not None:
             uncond_down_intrablock, cond_down_intrablock = [], []
             for down_intrablock in down_intrablock_additional_residuals:
@@ -419,12 +402,29 @@ class InvokeAIDiffuserComponent:
                 cond_down_intrablock.append(_cond_down)
 
         uncond_mid_block, cond_mid_block = None, None
-        mid_block_additional_residual = kwargs.pop("mid_block_additional_residual", None)
         if mid_block_additional_residual is not None:
             uncond_mid_block, cond_mid_block = mid_block_additional_residual.chunk(2)
 
-        # Run unconditional UNet denoising.
+        # If cross-attention control is enabled, prepare the SwapCrossAttnContext.
+        cross_attn_processor_context = None
+        if self.cross_attention_control_context is not None:
+            # Note that the SwapCrossAttnContext is initialized with an empty list of cross_attention_types_to_do.
+            # This list is empty because cross-attention control is not applied in the unconditioned pass. This field
+            # will be populated before the conditioned pass.
+            cross_attn_processor_context = SwapCrossAttnContext(
+                modified_text_embeddings=self.cross_attention_control_context.arguments.edited_conditioning,
+                index_map=self.cross_attention_control_context.cross_attention_index_map,
+                mask=self.cross_attention_control_context.cross_attention_mask,
+                cross_attention_types_to_do=[],
+            )
+
+        #####################
+        # Unconditioned pass
+        #####################
+
         cross_attention_kwargs = None
+
+        # Prepare IP-Adapter cross-attention kwargs for the unconditioned pass.
         if conditioning_data.ip_adapter_conditioning is not None:
             # Note that we 'unsqueeze' to produce tensors of shape (batch_size=1, num_ip_images, seq_len, token_len).
             cross_attention_kwargs = {
@@ -434,6 +434,11 @@ class InvokeAIDiffuserComponent:
                 ]
             }
 
+        # Prepare cross-attention control kwargs for the unconditioned pass.
+        if cross_attn_processor_context is not None:
+            cross_attention_kwargs = {"swap_cross_attn_context": cross_attn_processor_context}
+
+        # Prepare SDXL conditioning kwargs for the unconditioned pass.
         added_cond_kwargs = None
         is_sdxl = type(conditioning_data.text_embeddings) is SDXLConditioningInfo
         if is_sdxl:
@@ -442,6 +447,7 @@ class InvokeAIDiffuserComponent:
                 "time_ids": conditioning_data.unconditioned_embeddings.add_time_ids,
             }
 
+        # Run unconditioned UNet denoising (i.e. negative prompt).
         unconditioned_next_x = self.model_forward_callback(
             x,
             sigma,
@@ -451,11 +457,15 @@ class InvokeAIDiffuserComponent:
             mid_block_additional_residual=uncond_mid_block,
             down_intrablock_additional_residuals=uncond_down_intrablock,
             added_cond_kwargs=added_cond_kwargs,
-            **kwargs,
         )
 
-        # Run conditional UNet denoising.
+        ###################
+        # Conditioned pass
+        ###################
+
         cross_attention_kwargs = None
+
+        # Prepare IP-Adapter cross-attention kwargs for the conditioned pass.
         if conditioning_data.ip_adapter_conditioning is not None:
             # Note that we 'unsqueeze' to produce tensors of shape (batch_size=1, num_ip_images, seq_len, token_len).
             cross_attention_kwargs = {
@@ -465,6 +475,12 @@ class InvokeAIDiffuserComponent:
                 ]
             }
 
+        # Prepare cross-attention control kwargs for the conditioned pass.
+        if cross_attn_processor_context is not None:
+            cross_attn_processor_context.cross_attention_types_to_do = cross_attention_control_types_to_do
+            cross_attention_kwargs = {"swap_cross_attn_context": cross_attn_processor_context}
+
+        # Prepare SDXL conditioning kwargs for the conditioned pass.
         added_cond_kwargs = None
         if is_sdxl:
             added_cond_kwargs = {
@@ -472,6 +488,7 @@ class InvokeAIDiffuserComponent:
                 "time_ids": conditioning_data.text_embeddings.add_time_ids,
             }
 
+        # Run conditioned UNet denoising (i.e. positive prompt).
         conditioned_next_x = self.model_forward_callback(
             x,
             sigma,
@@ -481,89 +498,6 @@ class InvokeAIDiffuserComponent:
             mid_block_additional_residual=cond_mid_block,
             down_intrablock_additional_residuals=cond_down_intrablock,
             added_cond_kwargs=added_cond_kwargs,
-            **kwargs,
-        )
-        return unconditioned_next_x, conditioned_next_x
-
-    def _apply_cross_attention_controlled_conditioning(
-        self,
-        x: torch.Tensor,
-        sigma,
-        conditioning_data,
-        cross_attention_control_types_to_do,
-        **kwargs,
-    ):
-        context: Context = self.cross_attention_control_context
-
-        uncond_down_block, cond_down_block = None, None
-        down_block_additional_residuals = kwargs.pop("down_block_additional_residuals", None)
-        if down_block_additional_residuals is not None:
-            uncond_down_block, cond_down_block = [], []
-            for down_block in down_block_additional_residuals:
-                _uncond_down, _cond_down = down_block.chunk(2)
-                uncond_down_block.append(_uncond_down)
-                cond_down_block.append(_cond_down)
-
-        uncond_down_intrablock, cond_down_intrablock = None, None
-        down_intrablock_additional_residuals = kwargs.pop("down_intrablock_additional_residuals", None)
-        if down_intrablock_additional_residuals is not None:
-            uncond_down_intrablock, cond_down_intrablock = [], []
-            for down_intrablock in down_intrablock_additional_residuals:
-                _uncond_down, _cond_down = down_intrablock.chunk(2)
-                uncond_down_intrablock.append(_uncond_down)
-                cond_down_intrablock.append(_cond_down)
-
-        uncond_mid_block, cond_mid_block = None, None
-        mid_block_additional_residual = kwargs.pop("mid_block_additional_residual", None)
-        if mid_block_additional_residual is not None:
-            uncond_mid_block, cond_mid_block = mid_block_additional_residual.chunk(2)
-
-        cross_attn_processor_context = SwapCrossAttnContext(
-            modified_text_embeddings=context.arguments.edited_conditioning,
-            index_map=context.cross_attention_index_map,
-            mask=context.cross_attention_mask,
-            cross_attention_types_to_do=[],
-        )
-
-        added_cond_kwargs = None
-        is_sdxl = type(conditioning_data.text_embeddings) is SDXLConditioningInfo
-        if is_sdxl:
-            added_cond_kwargs = {
-                "text_embeds": conditioning_data.unconditioned_embeddings.pooled_embeds,
-                "time_ids": conditioning_data.unconditioned_embeddings.add_time_ids,
-            }
-
-        # no cross attention for unconditioning (negative prompt)
-        unconditioned_next_x = self.model_forward_callback(
-            x,
-            sigma,
-            conditioning_data.unconditioned_embeddings.embeds,
-            {"swap_cross_attn_context": cross_attn_processor_context},
-            down_block_additional_residuals=uncond_down_block,
-            mid_block_additional_residual=uncond_mid_block,
-            down_intrablock_additional_residuals=uncond_down_intrablock,
-            added_cond_kwargs=added_cond_kwargs,
-            **kwargs,
-        )
-
-        if is_sdxl:
-            added_cond_kwargs = {
-                "text_embeds": conditioning_data.text_embeddings.pooled_embeds,
-                "time_ids": conditioning_data.text_embeddings.add_time_ids,
-            }
-
-        # do requested cross attention types for conditioning (positive prompt)
-        cross_attn_processor_context.cross_attention_types_to_do = cross_attention_control_types_to_do
-        conditioned_next_x = self.model_forward_callback(
-            x,
-            sigma,
-            conditioning_data.text_embeddings.embeds,
-            {"swap_cross_attn_context": cross_attn_processor_context},
-            down_block_additional_residuals=cond_down_block,
-            mid_block_additional_residual=cond_mid_block,
-            down_intrablock_additional_residuals=cond_down_intrablock,
-            added_cond_kwargs=added_cond_kwargs,
-            **kwargs,
         )
         return unconditioned_next_x, conditioned_next_x
 
@@ -633,54 +567,3 @@ class InvokeAIDiffuserComponent:
 
         self.last_percent_through = percent_through
         return latents.to(device=dev)
-
-    # todo: make this work
-    @classmethod
-    def apply_conjunction(cls, x, t, forward_func, uc, c_or_weighted_c_list, global_guidance_scale):
-        x_in = torch.cat([x] * 2)
-        t_in = torch.cat([t] * 2)  # aka sigmas
-
-        deltas = None
-        uncond_latents = None
-        weighted_cond_list = (
-            c_or_weighted_c_list if isinstance(c_or_weighted_c_list, list) else [(c_or_weighted_c_list, 1)]
-        )
-
-        # below is fugly omg
-        conditionings = [uc] + [c for c, weight in weighted_cond_list]
-        weights = [1] + [weight for c, weight in weighted_cond_list]
-        chunk_count = math.ceil(len(conditionings) / 2)
-        deltas = None
-        for chunk_index in range(chunk_count):
-            offset = chunk_index * 2
-            chunk_size = min(2, len(conditionings) - offset)
-
-            if chunk_size == 1:
-                c_in = conditionings[offset]
-                latents_a = forward_func(x_in[:-1], t_in[:-1], c_in)
-                latents_b = None
-            else:
-                c_in = torch.cat(conditionings[offset : offset + 2])
-                latents_a, latents_b = forward_func(x_in, t_in, c_in).chunk(2)
-
-            # first chunk is guaranteed to be 2 entries: uncond_latents + first conditioining
-            if chunk_index == 0:
-                uncond_latents = latents_a
-                deltas = latents_b - uncond_latents
-            else:
-                deltas = torch.cat((deltas, latents_a - uncond_latents))
-                if latents_b is not None:
-                    deltas = torch.cat((deltas, latents_b - uncond_latents))
-
-        # merge the weighted deltas together into a single merged delta
-        per_delta_weights = torch.tensor(weights[1:], dtype=deltas.dtype, device=deltas.device)
-        normalize = False
-        if normalize:
-            per_delta_weights /= torch.sum(per_delta_weights)
-        reshaped_weights = per_delta_weights.reshape(per_delta_weights.shape + (1, 1, 1))
-        deltas_merged = torch.sum(deltas * reshaped_weights, dim=0, keepdim=True)
-
-        # old_return_value = super().forward(x, sigma, uncond, cond, cond_scale)
-        # assert(0 == len(torch.nonzero(old_return_value - (uncond_latents + deltas_merged * cond_scale))))
-
-        return uncond_latents + deltas_merged * global_guidance_scale

invokeai/backend/training/__init__.py

@@ -1,4 +1,5 @@
 """
 Initialization file for invokeai.backend.training
 """
+
 from .textual_inversion_training import do_textual_inversion_training, parse_args  # noqa: F401

invokeai/backend/util/__init__.py

@@ -1,6 +1,7 @@
 """
 Initialization file for invokeai.backend.util
 """
+
 from .attention import auto_detect_slice_size  # noqa: F401
 from .devices import (  # noqa: F401
     CPU_DEVICE,

invokeai/backend/util/attention.py

@@ -3,6 +3,7 @@
 Utility routine used for autodetection of optimal slice size
 for attention mechanism.
 """
+
 import psutil
 import torch
 

invokeai/backend/util/silence_warnings.py

@@ -1,4 +1,5 @@
 """Context class to silence transformers and diffusers warnings."""
+
 import warnings
 from typing import Any
 

invokeai/backend/util/util.py

@@ -340,14 +340,17 @@ def download_with_resume(url: str, dest: Path, access_token: str = None) -> Path
             logger.error(f"ERROR DOWNLOADING {url}: {resp.text}")
             return None
 
-        with open(dest, open_mode) as file, tqdm(
-            desc=str(dest),
-            initial=exist_size,
-            total=content_length,
-            unit="iB",
-            unit_scale=True,
-            unit_divisor=1000,
-        ) as bar:
+        with (
+            open(dest, open_mode) as file,
+            tqdm(
+                desc=str(dest),
+                initial=exist_size,
+                total=content_length,
+                unit="iB",
+                unit_scale=True,
+                unit_divisor=1000,
+            ) as bar,
+        ):
             for data in resp.iter_content(chunk_size=1024):
                 size = file.write(data)
                 bar.update(size)

invokeai/frontend/CLI/__init__.py

@@ -1,4 +1,5 @@
 """
 Initialization file for invokeai.frontend.CLI
 """
+
 from .CLI import main as invokeai_command_line_interface  # noqa: F401