diff --git a/.dockerignore b/.dockerignore
index cfdc7fc735..3f674f978d 100644
--- a/.dockerignore
+++ b/.dockerignore
@@ -1,25 +1,9 @@
-# use this file as a whitelist
*
!invokeai
-!ldm
!pyproject.toml
+!docker/docker-entrypoint.sh
+!LICENSE
-# ignore frontend/web but whitelist dist
-invokeai/frontend/web/
-!invokeai/frontend/web/dist/
-
-# ignore invokeai/assets but whitelist invokeai/assets/web
-invokeai/assets/
-!invokeai/assets/web/
-
-# Guard against pulling in any models that might exist in the directory tree
-**/*.pt*
-**/*.ckpt
-
-# Byte-compiled / optimized / DLL files
-**/__pycache__/
-**/*.py[cod]
-
-# Distribution / packaging
-**/*.egg-info/
-**/*.egg
+**/node_modules
+**/__pycache__
+**/*.egg-info
\ No newline at end of file
diff --git a/.github/workflows/build-container.yml b/.github/workflows/build-container.yml
index 23d7c82fe3..60eba4a297 100644
--- a/.github/workflows/build-container.yml
+++ b/.github/workflows/build-container.yml
@@ -3,17 +3,15 @@ on:
push:
branches:
- 'main'
- - 'update/ci/docker/*'
- - 'update/docker/*'
- - 'dev/ci/docker/*'
- - 'dev/docker/*'
paths:
- 'pyproject.toml'
- '.dockerignore'
- 'invokeai/**'
- 'docker/Dockerfile'
+ - 'docker/docker-entrypoint.sh'
+ - 'workflows/build-container.yml'
tags:
- - 'v*.*.*'
+ - 'v*'
workflow_dispatch:
permissions:
@@ -26,23 +24,27 @@ jobs:
strategy:
fail-fast: false
matrix:
- flavor:
- - rocm
- - cuda
- - cpu
- include:
- - flavor: rocm
- pip-extra-index-url: 'https://download.pytorch.org/whl/rocm5.2'
- - flavor: cuda
- pip-extra-index-url: ''
- - flavor: cpu
- pip-extra-index-url: 'https://download.pytorch.org/whl/cpu'
+ gpu-driver:
+ - cuda
+ - cpu
+ - rocm
runs-on: ubuntu-latest
- name: ${{ matrix.flavor }}
+ name: ${{ matrix.gpu-driver }}
env:
- PLATFORMS: 'linux/amd64,linux/arm64'
- DOCKERFILE: 'docker/Dockerfile'
+ # torch/arm64 does not support GPU currently, so arm64 builds
+ # would not be GPU-accelerated.
+ # re-enable arm64 if there is sufficient demand.
+ # PLATFORMS: 'linux/amd64,linux/arm64'
+ PLATFORMS: 'linux/amd64'
steps:
+ - name: Free up more disk space on the runner
+ # https://github.com/actions/runner-images/issues/2840#issuecomment-1284059930
+ run: |
+ sudo rm -rf /usr/share/dotnet
+ sudo rm -rf "$AGENT_TOOLSDIRECTORY"
+ sudo swapoff /mnt/swapfile
+ sudo rm -rf /mnt/swapfile
+
- name: Checkout
uses: actions/checkout@v3
@@ -53,7 +55,7 @@ jobs:
github-token: ${{ secrets.GITHUB_TOKEN }}
images: |
ghcr.io/${{ github.repository }}
- ${{ vars.DOCKERHUB_REPOSITORY }}
+ ${{ env.DOCKERHUB_REPOSITORY }}
tags: |
type=ref,event=branch
type=ref,event=tag
@@ -62,8 +64,8 @@ jobs:
type=pep440,pattern={{major}}
type=sha,enable=true,prefix=sha-,format=short
flavor: |
- latest=${{ matrix.flavor == 'cuda' && github.ref == 'refs/heads/main' }}
- suffix=-${{ matrix.flavor }},onlatest=false
+ latest=${{ matrix.gpu-driver == 'cuda' && github.ref == 'refs/heads/main' }}
+ suffix=-${{ matrix.gpu-driver }},onlatest=false
- name: Set up QEMU
uses: docker/setup-qemu-action@v2
@@ -81,34 +83,33 @@ jobs:
username: ${{ github.repository_owner }}
password: ${{ secrets.GITHUB_TOKEN }}
- - name: Login to Docker Hub
- if: github.event_name != 'pull_request' && vars.DOCKERHUB_REPOSITORY != ''
- uses: docker/login-action@v2
- with:
- username: ${{ secrets.DOCKERHUB_USERNAME }}
- password: ${{ secrets.DOCKERHUB_TOKEN }}
+ # - name: Login to Docker Hub
+ # if: github.event_name != 'pull_request' && vars.DOCKERHUB_REPOSITORY != ''
+ # uses: docker/login-action@v2
+ # with:
+ # username: ${{ secrets.DOCKERHUB_USERNAME }}
+ # password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Build container
id: docker_build
uses: docker/build-push-action@v4
with:
context: .
- file: ${{ env.DOCKERFILE }}
+ file: docker/Dockerfile
platforms: ${{ env.PLATFORMS }}
push: ${{ github.ref == 'refs/heads/main' || github.ref_type == 'tag' }}
tags: ${{ steps.meta.outputs.tags }}
labels: ${{ steps.meta.outputs.labels }}
- build-args: PIP_EXTRA_INDEX_URL=${{ matrix.pip-extra-index-url }}
cache-from: |
- type=gha,scope=${{ github.ref_name }}-${{ matrix.flavor }}
- type=gha,scope=main-${{ matrix.flavor }}
- cache-to: type=gha,mode=max,scope=${{ github.ref_name }}-${{ matrix.flavor }}
+ type=gha,scope=${{ github.ref_name }}-${{ matrix.gpu-driver }}
+ type=gha,scope=main-${{ matrix.gpu-driver }}
+ cache-to: type=gha,mode=max,scope=${{ github.ref_name }}-${{ matrix.gpu-driver }}
- - name: Docker Hub Description
- if: github.ref == 'refs/heads/main' || github.ref == 'refs/tags/*' && vars.DOCKERHUB_REPOSITORY != ''
- uses: peter-evans/dockerhub-description@v3
- with:
- username: ${{ secrets.DOCKERHUB_USERNAME }}
- password: ${{ secrets.DOCKERHUB_TOKEN }}
- repository: ${{ vars.DOCKERHUB_REPOSITORY }}
- short-description: ${{ github.event.repository.description }}
+ # - name: Docker Hub Description
+ # if: github.ref == 'refs/heads/main' || github.ref == 'refs/tags/*' && vars.DOCKERHUB_REPOSITORY != ''
+ # uses: peter-evans/dockerhub-description@v3
+ # with:
+ # username: ${{ secrets.DOCKERHUB_USERNAME }}
+ # password: ${{ secrets.DOCKERHUB_TOKEN }}
+ # repository: ${{ vars.DOCKERHUB_REPOSITORY }}
+ # short-description: ${{ github.event.repository.description }}
diff --git a/.github/workflows/test-invoke-pip-skip.yml b/.github/workflows/test-invoke-pip-skip.yml
index d4c9d9fc00..004b46d5a8 100644
--- a/.github/workflows/test-invoke-pip-skip.yml
+++ b/.github/workflows/test-invoke-pip-skip.yml
@@ -1,10 +1,16 @@
name: Test invoke.py pip
+
+# This is a dummy stand-in for the actual tests
+# we don't need to run python tests on non-Python changes
+# But PRs require passing tests to be mergeable
+
on:
pull_request:
paths:
- '**'
- '!pyproject.toml'
- '!invokeai/**'
+ - '!tests/**'
- 'invokeai/frontend/web/**'
merge_group:
workflow_dispatch:
@@ -19,48 +25,26 @@ jobs:
strategy:
matrix:
python-version:
- # - '3.9'
- '3.10'
pytorch:
- # - linux-cuda-11_6
- linux-cuda-11_7
- linux-rocm-5_2
- linux-cpu
- macos-default
- windows-cpu
- # - windows-cuda-11_6
- # - windows-cuda-11_7
include:
- # - pytorch: linux-cuda-11_6
- # os: ubuntu-22.04
- # extra-index-url: 'https://download.pytorch.org/whl/cu116'
- # github-env: $GITHUB_ENV
- 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
- # - pytorch: windows-cuda-11_6
- # os: windows-2022
- # extra-index-url: 'https://download.pytorch.org/whl/cu116'
- # github-env: $env:GITHUB_ENV
- # - pytorch: windows-cuda-11_7
- # os: windows-2022
- # extra-index-url: 'https://download.pytorch.org/whl/cu117'
- # github-env: $env:GITHUB_ENV
name: ${{ matrix.pytorch }} on ${{ matrix.python-version }}
runs-on: ${{ matrix.os }}
steps:
- - run: 'echo "No build required"'
+ - name: skip
+ run: echo "no build required"
diff --git a/.github/workflows/test-invoke-pip.yml b/.github/workflows/test-invoke-pip.yml
index 071232e06e..40be0a529e 100644
--- a/.github/workflows/test-invoke-pip.yml
+++ b/.github/workflows/test-invoke-pip.yml
@@ -11,6 +11,7 @@ on:
paths:
- 'pyproject.toml'
- 'invokeai/**'
+ - 'tests/**'
- '!invokeai/frontend/web/**'
types:
- 'ready_for_review'
@@ -32,19 +33,12 @@ jobs:
# - '3.9'
- '3.10'
pytorch:
- # - linux-cuda-11_6
- linux-cuda-11_7
- linux-rocm-5_2
- linux-cpu
- macos-default
- windows-cpu
- # - windows-cuda-11_6
- # - windows-cuda-11_7
include:
- # - pytorch: linux-cuda-11_6
- # os: ubuntu-22.04
- # extra-index-url: 'https://download.pytorch.org/whl/cu116'
- # github-env: $GITHUB_ENV
- pytorch: linux-cuda-11_7
os: ubuntu-22.04
github-env: $GITHUB_ENV
@@ -62,14 +56,6 @@ jobs:
- pytorch: windows-cpu
os: windows-2022
github-env: $env:GITHUB_ENV
- # - pytorch: windows-cuda-11_6
- # os: windows-2022
- # extra-index-url: 'https://download.pytorch.org/whl/cu116'
- # github-env: $env:GITHUB_ENV
- # - pytorch: windows-cuda-11_7
- # os: windows-2022
- # extra-index-url: 'https://download.pytorch.org/whl/cu117'
- # github-env: $env:GITHUB_ENV
name: ${{ matrix.pytorch }} on ${{ matrix.python-version }}
runs-on: ${{ matrix.os }}
env:
@@ -100,40 +86,38 @@ jobs:
id: run-pytest
run: pytest
- - name: run invokeai-configure
- id: run-preload-models
- 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-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: 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
- id: archive-results
- env:
- INVOKEAI_OUTDIR: ${{ github.workspace }}/results
- uses: actions/upload-artifact@v3
- with:
- name: results
- path: ${{ env.INVOKEAI_OUTDIR }}
+ # - name: Archive results
+ # env:
+ # INVOKEAI_OUTDIR: ${{ github.workspace }}/results
+ # uses: actions/upload-artifact@v3
+ # with:
+ # name: results
+ # path: ${{ env.INVOKEAI_OUTDIR }}
diff --git a/.gitignore b/.gitignore
index 7f3b1278df..b7b5c1ae75 100644
--- a/.gitignore
+++ b/.gitignore
@@ -34,7 +34,7 @@ __pycache__/
.Python
build/
develop-eggs/
-dist/
+# dist/
downloads/
eggs/
.eggs/
@@ -79,6 +79,7 @@ cov.xml
.pytest.ini
cover/
junit/
+notes/
# Translations
*.mo
@@ -201,7 +202,8 @@ checkpoints
# If it's a Mac
.DS_Store
-invokeai/frontend/web/dist/*
+invokeai/frontend/yarn.lock
+invokeai/frontend/node_modules
# Let the frontend manage its own gitignore
!invokeai/frontend/web/*
diff --git a/LICENSE b/LICENSE
index cb9631b9f6..fac28ea6b9 100644
--- a/LICENSE
+++ b/LICENSE
@@ -1,21 +1,176 @@
-MIT License
+ Apache License
+ Version 2.0, January 2004
+ http://www.apache.org/licenses/
-Copyright (c) 2022 InvokeAI Team
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
+ 1. Definitions.
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
+ "License" shall mean the terms and conditions for use, reproduction,
+ and distribution as defined by Sections 1 through 9 of this document.
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+ "Licensor" shall mean the copyright owner or entity authorized by
+ the copyright owner that is granting the License.
+
+ "Legal Entity" shall mean the union of the acting entity and all
+ other entities that control, are controlled by, or are under common
+ control with that entity. For the purposes of this definition,
+ "control" means (i) the power, direct or indirect, to cause the
+ direction or management of such entity, whether by contract or
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
+ outstanding shares, or (iii) beneficial ownership of such entity.
+
+ "You" (or "Your") shall mean an individual or Legal Entity
+ exercising permissions granted by this License.
+
+ "Source" form shall mean the preferred form for making modifications,
+ including but not limited to software source code, documentation
+ source, and configuration files.
+
+ "Object" form shall mean any form resulting from mechanical
+ transformation or translation of a Source form, including but
+ not limited to compiled object code, generated documentation,
+ and conversions to other media types.
+
+ "Work" shall mean the work of authorship, whether in Source or
+ Object form, made available under the License, as indicated by a
+ copyright notice that is included in or attached to the work
+ (an example is provided in the Appendix below).
+
+ "Derivative Works" shall mean any work, whether in Source or Object
+ form, that is based on (or derived from) the Work and for which the
+ editorial revisions, annotations, elaborations, or other modifications
+ represent, as a whole, an original work of authorship. For the purposes
+ of this License, Derivative Works shall not include works that remain
+ separable from, or merely link (or bind by name) to the interfaces of,
+ the Work and Derivative Works thereof.
+
+ "Contribution" shall mean any work of authorship, including
+ the original version of the Work and any modifications or additions
+ to that Work or Derivative Works thereof, that is intentionally
+ submitted to Licensor for inclusion in the Work by the copyright owner
+ or by an individual or Legal Entity authorized to submit on behalf of
+ the copyright owner. For the purposes of this definition, "submitted"
+ means any form of electronic, verbal, or written communication sent
+ to the Licensor or its representatives, including but not limited to
+ communication on electronic mailing lists, source code control systems,
+ and issue tracking systems that are managed by, or on behalf of, the
+ Licensor for the purpose of discussing and improving the Work, but
+ excluding communication that is conspicuously marked or otherwise
+ designated in writing by the copyright owner as "Not a Contribution."
+
+ "Contributor" shall mean Licensor and any individual or Legal Entity
+ on behalf of whom a Contribution has been received by Licensor and
+ subsequently incorporated within the Work.
+
+ 2. Grant of Copyright License. Subject to the terms and conditions of
+ this License, each Contributor hereby grants to You a perpetual,
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+ copyright license to reproduce, prepare Derivative Works of,
+ publicly display, publicly perform, sublicense, and distribute the
+ Work and such Derivative Works in Source or Object form.
+
+ 3. Grant of Patent License. Subject to the terms and conditions of
+ this License, each Contributor hereby grants to You a perpetual,
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+ (except as stated in this section) patent license to make, have made,
+ use, offer to sell, sell, import, and otherwise transfer the Work,
+ where such license applies only to those patent claims licensable
+ by such Contributor that are necessarily infringed by their
+ Contribution(s) alone or by combination of their Contribution(s)
+ with the Work to which such Contribution(s) was submitted. If You
+ institute patent litigation against any entity (including a
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
+ or a Contribution incorporated within the Work constitutes direct
+ or contributory patent infringement, then any patent licenses
+ granted to You under this License for that Work shall terminate
+ as of the date such litigation is filed.
+
+ 4. Redistribution. You may reproduce and distribute copies of the
+ Work or Derivative Works thereof in any medium, with or without
+ modifications, and in Source or Object form, provided that You
+ meet the following conditions:
+
+ (a) You must give any other recipients of the Work or
+ Derivative Works a copy of this License; and
+
+ (b) You must cause any modified files to carry prominent notices
+ stating that You changed the files; and
+
+ (c) You must retain, in the Source form of any Derivative Works
+ that You distribute, all copyright, patent, trademark, and
+ attribution notices from the Source form of the Work,
+ excluding those notices that do not pertain to any part of
+ the Derivative Works; and
+
+ (d) If the Work includes a "NOTICE" text file as part of its
+ distribution, then any Derivative Works that You distribute must
+ include a readable copy of the attribution notices contained
+ within such NOTICE file, excluding those notices that do not
+ pertain to any part of the Derivative Works, in at least one
+ of the following places: within a NOTICE text file distributed
+ as part of the Derivative Works; within the Source form or
+ documentation, if provided along with the Derivative Works; or,
+ within a display generated by the Derivative Works, if and
+ wherever such third-party notices normally appear. The contents
+ of the NOTICE file are for informational purposes only and
+ do not modify the License. You may add Your own attribution
+ notices within Derivative Works that You distribute, alongside
+ or as an addendum to the NOTICE text from the Work, provided
+ that such additional attribution notices cannot be construed
+ as modifying the License.
+
+ You may add Your own copyright statement to Your modifications and
+ may provide additional or different license terms and conditions
+ for use, reproduction, or distribution of Your modifications, or
+ for any such Derivative Works as a whole, provided Your use,
+ reproduction, and distribution of the Work otherwise complies with
+ the conditions stated in this License.
+
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
+ any Contribution intentionally submitted for inclusion in the Work
+ by You to the Licensor shall be under the terms and conditions of
+ this License, without any additional terms or conditions.
+ Notwithstanding the above, nothing herein shall supersede or modify
+ the terms of any separate license agreement you may have executed
+ with Licensor regarding such Contributions.
+
+ 6. Trademarks. This License does not grant permission to use the trade
+ names, trademarks, service marks, or product names of the Licensor,
+ except as required for reasonable and customary use in describing the
+ origin of the Work and reproducing the content of the NOTICE file.
+
+ 7. Disclaimer of Warranty. Unless required by applicable law or
+ agreed to in writing, Licensor provides the Work (and each
+ Contributor provides its Contributions) on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+ implied, including, without limitation, any warranties or conditions
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+ PARTICULAR PURPOSE. You are solely responsible for determining the
+ appropriateness of using or redistributing the Work and assume any
+ risks associated with Your exercise of permissions under this License.
+
+ 8. Limitation of Liability. In no event and under no legal theory,
+ whether in tort (including negligence), contract, or otherwise,
+ unless required by applicable law (such as deliberate and grossly
+ negligent acts) or agreed to in writing, shall any Contributor be
+ liable to You for damages, including any direct, indirect, special,
+ incidental, or consequential damages of any character arising as a
+ result of this License or out of the use or inability to use the
+ Work (including but not limited to damages for loss of goodwill,
+ work stoppage, computer failure or malfunction, or any and all
+ other commercial damages or losses), even if such Contributor
+ has been advised of the possibility of such damages.
+
+ 9. Accepting Warranty or Additional Liability. While redistributing
+ the Work or Derivative Works thereof, You may choose to offer,
+ and charge a fee for, acceptance of support, warranty, indemnity,
+ or other liability obligations and/or rights consistent with this
+ License. However, in accepting such obligations, You may act only
+ on Your own behalf and on Your sole responsibility, not on behalf
+ of any other Contributor, and only if You agree to indemnify,
+ defend, and hold each Contributor harmless for any liability
+ incurred by, or claims asserted against, such Contributor by reason
+ of your accepting any such warranty or additional liability.
+
+
diff --git a/README.md b/README.md
index 9471e60404..5ad2b47509 100644
--- a/README.md
+++ b/README.md
@@ -1,8 +1,11 @@
-
+
+
+# Invoke AI - Generative AI for Professional Creatives
+## Professional Creative Tools for Stable Diffusion, Custom-Trained Models, and more.
+ To learn more about Invoke AI, get started instantly, or implement our Business solutions, visit [invoke.ai](https://invoke.ai)
-# InvokeAI: A Stable Diffusion Toolkit
[![discord badge]][discord link]
@@ -33,32 +36,32 @@
-_**Note: The UI is not fully functional on `main`. If you need a stable UI based on `main`, use the `pre-nodes` tag while we [migrate to a new backend](https://github.com/invoke-ai/InvokeAI/discussions/3246).**_
+_**Note: This is an alpha release. Bugs are expected and not all
+features are fully implemented. Please use the GitHub [Issues
+pages](https://github.com/invoke-ai/InvokeAI/issues?q=is%3Aissue+is%3Aopen)
+to report unexpected problems. Also note that InvokeAI root directory
+which contains models, outputs and configuration files, has changed
+between the 2.x and 3.x release. If you wish to use your v2.3 root
+directory with v3.0, please follow the directions in [Migrating a 2.3
+root directory to 3.0](#migrating-to-3).**_
-InvokeAI is a leading creative engine built to empower professionals and enthusiasts alike. Generate and create stunning visual media using the latest AI-driven technologies. InvokeAI offers an industry leading Web Interface, interactive Command Line Interface, and also serves as the foundation for multiple commercial products.
+InvokeAI is a leading creative engine built to empower professionals
+and enthusiasts alike. Generate and create stunning visual media using
+the latest AI-driven technologies. InvokeAI offers an industry leading
+Web Interface, interactive Command Line Interface, and also serves as
+the foundation for multiple commercial products.
-**Quick links**: [[How to Install](https://invoke-ai.github.io/InvokeAI/#installation)] [Discord Server] [Documentation and Tutorials] [Code and Downloads] [Bug Reports] [Discussion, Ideas & Q&A]
-
-_Note: InvokeAI is rapidly evolving. Please use the
-[Issues](https://github.com/invoke-ai/InvokeAI/issues) tab to report bugs and make feature
-requests. Be sure to use the provided templates. They will help us diagnose issues faster._
-
-## FOR DEVELOPERS - MIGRATING TO THE 3.0.0 MODELS FORMAT
-
-The models directory and models.yaml have changed. To migrate to the
-new layout, please follow this recipe:
-
-1. Run `python scripts/migrate_models_to_3.0.py
-
-2. This will create a new models directory named `models-3.0` and a
- new config directory named `models.yaml-3.0`, both in the current
- working directory. If you prefer to name them something else, pass
- the `--dest-directory` and/or `--dest-yaml` arguments.
-
-3. Check that the new models directory and yaml file look ok.
-
-4. Replace the existing directory and file, keeping backup copies just in
-case.
+**Quick links**: [[How to
+ Install](https://invoke-ai.github.io/InvokeAI/#installation)] [Discord Server] [Documentation and
+ Tutorials] [Code and
+ Downloads] [Bug Reports]
+ [Discussion,
+ Ideas & Q&A]
-
-
-!!! tip
-
- When designing prompts, think about how the images scraped from the internet were
- captioned. Very few photographs will be labeled "photograph" or "photorealistic."
- They will, however, be captioned with the publication, photographer, camera model,
- or film settings.
-
-If the initial image contains transparent regions, then Stable Diffusion will
-only draw within the transparent regions, a process called
-[`inpainting`](./INPAINTING.md#creating-transparent-regions-for-inpainting).
-However, for this to work correctly, the color information underneath the
-transparent needs to be preserved, not erased.
-
-!!! warning "**IMPORTANT ISSUE** "
-
- `img2img` does not work properly on initial images smaller
- than 512x512. Please scale your image to at least 512x512 before using it.
- Larger images are not a problem, but may run out of VRAM on your GPU card. To
- fix this, use the --fit option, which downscales the initial image to fit within
- the box specified by width x height:
-
- ```
- tree on a hill with a river, national geographic -I./test-pictures/big-sketch.png -H512 -W512 --fit
- ```
-
-## How does it actually work, though?
+For a walkthrough of using Image-to-Image in the Web UI, see [InvokeAI
+Web Server](./WEB.md#image-to-image).
The main difference between `img2img` and `prompt2img` is the starting point.
While `prompt2img` always starts with pure gaussian noise and progressively
@@ -99,10 +26,6 @@ seed `1592514025` develops something like this:
!!! example ""
- ```bash
- invoke> "fire" -s10 -W384 -H384 -S1592514025
- ```
-
{ width=720 }
@@ -157,17 +80,8 @@ Diffusion has less chance to refine itself, so the result ends up inheriting all
the problems of my bad drawing.
If you want to try this out yourself, all of these are using a seed of
-`1592514025` with a width/height of `384`, step count `10`, the default sampler
-(`k_lms`), and the single-word prompt `"fire"`:
-
-```bash
-invoke> "fire" -s10 -W384 -H384 -S1592514025 -I /tmp/fire-drawing.png --strength 0.7
-```
-
-The code for rendering intermediates is on my (damian0815's) branch
-[document-img2img](https://github.com/damian0815/InvokeAI/tree/document-img2img) -
-run `invoke.py` and check your `outputs/img-samples/intermediates` folder while
-generating an image.
+`1592514025` with a width/height of `384`, step count `10`, the
+`k_lms` sampler, and the single-word prompt `"fire"`.
### Compensating for the reduced step count
@@ -180,10 +94,6 @@ give each generation 20 steps.
Here's strength `0.4` (note step count `50`, which is `20 ÷ 0.4` to make sure SD
does `20` steps from my image):
-```bash
-invoke> "fire" -s50 -W384 -H384 -S1592514025 -I /tmp/fire-drawing.png -f 0.4
-```
-
@@ -191,10 +101,6 @@ invoke> "fire" -s50 -W384 -H384 -S1592514025 -I /tmp/fire-drawing.png -f 0.4
and here is strength `0.7` (note step count `30`, which is roughly `20 ÷ 0.7` to
make sure SD does `20` steps from my image):
-```commandline
-invoke> "fire" -s30 -W384 -H384 -S1592514025 -I /tmp/fire-drawing.png -f 0.7
-```
-
diff --git a/docs/features/MODEL_MERGING.md b/docs/features/MODEL_MERGING.md
index 7157e035ae..36e15ad0e4 100644
--- a/docs/features/MODEL_MERGING.md
+++ b/docs/features/MODEL_MERGING.md
@@ -71,6 +71,3 @@ under the selected name and register it with InvokeAI.
use InvokeAI conventions - only alphanumeric letters and the
characters ".+-".
-## Caveats
-
-This is a new script and may contain bugs.
diff --git a/docs/features/NSFW.md b/docs/features/NSFW.md
index 06d382f47d..f0539cc6a9 100644
--- a/docs/features/NSFW.md
+++ b/docs/features/NSFW.md
@@ -31,10 +31,22 @@ turned on and off on the command line using `--nsfw_checker` and
At installation time, InvokeAI will ask whether the checker should be
activated by default (neither argument given on the command line). The
-response is stored in the InvokeAI initialization file (usually
-`invokeai.init` in your home directory). You can change the default at any
-time by opening this file in a text editor and commenting or
-uncommenting the line `--nsfw_checker`.
+response is stored in the InvokeAI initialization file
+(`invokeai.yaml` in the InvokeAI root directory). You can change the
+default at any time by opening this file in a text editor and
+changing the line `nsfw_checker:` from true to false or vice-versa:
+
+
+```
+...
+ Features:
+ esrgan: true
+ internet_available: true
+ log_tokenization: false
+ nsfw_checker: true
+ patchmatch: true
+ restore: true
+```
## Caveats
@@ -79,11 +91,3 @@ generates. However, it does write metadata into the PNG data area,
including the prompt used to generate the image and relevant parameter
settings. These fields can be examined using the `sd-metadata.py`
script that comes with the InvokeAI package.
-
-Note that several other Stable Diffusion distributions offer
-wavelet-based "invisible" watermarking. We have experimented with the
-library used to generate these watermarks and have reached the
-conclusion that while the watermarking library may be adding
-watermarks to PNG images, the currently available version is unable to
-retrieve them successfully. If and when a functioning version of the
-library becomes available, we will offer this feature as well.
diff --git a/docs/features/OTHER.md b/docs/features/OTHER.md
index 2ba76e912a..2401f644ba 100644
--- a/docs/features/OTHER.md
+++ b/docs/features/OTHER.md
@@ -18,43 +18,16 @@ Output Example:
## **Seamless Tiling**
-The seamless tiling mode causes generated images to seamlessly tile with itself. To use it, add the
-`--seamless` option when starting the script which will result in all generated images to tile, or
-for each `invoke>` prompt as shown here:
+The seamless tiling mode causes generated images to seamlessly tile
+with itself creating repetitive wallpaper-like patterns. To use it,
+activate the Seamless Tiling option in the Web GUI and then select
+whether to tile on the X (horizontal) and/or Y (vertical) axes. Tiling
+will then be active for the next set of generations.
+
+A nice prompt to test seamless tiling with is:
-```python
-invoke> "pond garden with lotus by claude monet" --seamless -s100 -n4
```
-
-By default this will tile on both the X and Y axes. However, you can also specify specific axes to tile on with `--seamless_axes`.
-Possible values are `x`, `y`, and `x,y`:
-```python
-invoke> "pond garden with lotus by claude monet" --seamless --seamless_axes=x -s100 -n4
-```
-
----
-
-## **Shortcuts: Reusing Seeds**
-
-Since it is so common to reuse seeds while refining a prompt, there is now a shortcut as of version
-1.11. Provide a `-S` (or `--seed`) switch of `-1` to use the seed of the most recent image
-generated. If you produced multiple images with the `-n` switch, then you can go back further
-using `-2`, `-3`, etc. up to the first image generated by the previous command. Sorry, but you can't go
-back further than one command.
-
-Here's an example of using this to do a quick refinement. It also illustrates using the new `-G`
-switch to turn on upscaling and face enhancement (see previous section):
-
-```bash
-invoke> a cute child playing hopscotch -G0.5
-[...]
-outputs/img-samples/000039.3498014304.png: "a cute child playing hopscotch" -s50 -W512 -H512 -C7.5 -mk_lms -S3498014304
-
-# I wonder what it will look like if I bump up the steps and set facial enhancement to full strength?
-invoke> a cute child playing hopscotch -G1.0 -s100 -S -1
-reusing previous seed 3498014304
-[...]
-outputs/img-samples/000040.3498014304.png: "a cute child playing hopscotch" -G1.0 -s100 -W512 -H512 -C7.5 -mk_lms -S3498014304
+pond garden with lotus by claude monet"
```
---
@@ -73,66 +46,27 @@ This will tell the sampler to invest 25% of its effort on the tabby cat aspect o
on the white duck aspect (surprisingly, this example actually works). The prompt weights can use any
combination of integers and floating point numbers, and they do not need to add up to 1.
----
-
-## **Filename Format**
-
-The argument `--fnformat` allows to specify the filename of the
- image. Supported wildcards are all arguments what can be set such as
- `perlin`, `seed`, `threshold`, `height`, `width`, `gfpgan_strength`,
- `sampler_name`, `steps`, `model`, `upscale`, `prompt`, `cfg_scale`,
- `prefix`.
-
-The following prompt
-```bash
-dream> a red car --steps 25 -C 9.8 --perlin 0.1 --fnformat {prompt}_steps.{steps}_cfg.{cfg_scale}_perlin.{perlin}.png
-```
-
-generates a file with the name: `outputs/img-samples/a red car_steps.25_cfg.9.8_perlin.0.1.png`
-
----
-
## **Thresholding and Perlin Noise Initialization Options**
-Two new options are the thresholding (`--threshold`) and the perlin noise initialization (`--perlin`) options. Thresholding limits the range of the latent values during optimization, which helps combat oversaturation with higher CFG scale values. Perlin noise initialization starts with a percentage (a value ranging from 0 to 1) of perlin noise mixed into the initial noise. Both features allow for more variations and options in the course of generating images.
+Under the Noise section of the Web UI, you will find two options named
+Perlin Noise and Noise Threshold. [Perlin
+noise](https://en.wikipedia.org/wiki/Perlin_noise) is a type of
+structured noise used to simulate terrain and other natural
+textures. The slider controls the percentage of perlin noise that will
+be mixed into the image at the beginning of generation. Adding a little
+perlin noise to a generation will alter the image substantially.
+
+The noise threshold limits the range of the latent values during
+sampling and helps combat the oversharpening seem with higher CFG
+scale values.
For better intuition into what these options do in practice:
-In generating this graphic, perlin noise at initialization was programmatically varied going across on the diagram by values 0.0, 0.1, 0.2, 0.4, 0.5, 0.6, 0.8, 0.9, 1.0; and the threshold was varied going down from
-0, 1, 2, 3, 4, 5, 10, 20, 100. The other options are fixed, so the initial prompt is as follows (no thresholding or perlin noise):
-
-```bash
-invoke> "a portrait of a beautiful young lady" -S 1950357039 -s 100 -C 20 -A k_euler_a --threshold 0 --perlin 0
-```
-
-Here's an example of another prompt used when setting the threshold to 5 and perlin noise to 0.2:
-
-```bash
-invoke> "a portrait of a beautiful young lady" -S 1950357039 -s 100 -C 20 -A k_euler_a --threshold 5 --perlin 0.2
-```
-
-!!! note
-
- currently the thresholding feature is only implemented for the k-diffusion style samplers, and empirically appears to work best with `k_euler_a` and `k_dpm_2_a`. Using 0 disables thresholding. Using 0 for perlin noise disables using perlin noise for initialization. Finally, using 1 for perlin noise uses only perlin noise for initialization.
-
----
-
-## **Simplified API**
-
-For programmers who wish to incorporate stable-diffusion into other products, this repository
-includes a simplified API for text to image generation, which lets you create images from a prompt
-in just three lines of code:
-
-```bash
-from ldm.generate import Generate
-g = Generate()
-outputs = g.txt2img("a unicorn in manhattan")
-```
-
-Outputs is a list of lists in the format [filename1,seed1],[filename2,seed2]...].
-
-Please see the documentation in ldm/generate.py for more information.
-
----
+In generating this graphic, perlin noise at initialization was
+programmatically varied going across on the diagram by values 0.0,
+0.1, 0.2, 0.4, 0.5, 0.6, 0.8, 0.9, 1.0; and the threshold was varied
+going down from 0, 1, 2, 3, 4, 5, 10, 20, 100. The other options are
+fixed using the prompt "a portrait of a beautiful young lady" a CFG of
+20, 100 steps, and a seed of 1950357039.
diff --git a/docs/features/POSTPROCESS.md b/docs/features/POSTPROCESS.md
index c0233951da..22ef51e109 100644
--- a/docs/features/POSTPROCESS.md
+++ b/docs/features/POSTPROCESS.md
@@ -8,12 +8,6 @@ title: Postprocessing
This extension provides the ability to restore faces and upscale images.
-Face restoration and upscaling can be applied at the time you generate the
-images, or at any later time against a previously-generated PNG file, using the
-[!fix](#fixing-previously-generated-images) command.
-[Outpainting and outcropping](OUTPAINTING.md) can only be applied after the
-fact.
-
## Face Fixing
The default face restoration module is GFPGAN. The default upscale is
@@ -23,8 +17,7 @@ Real-ESRGAN. For an alternative face restoration module, see
As of version 1.14, environment.yaml will install the Real-ESRGAN package into
the standard install location for python packages, and will put GFPGAN into a
subdirectory of "src" in the InvokeAI directory. Upscaling with Real-ESRGAN
-should "just work" without further intervention. Simply pass the `--upscale`
-(`-U`) option on the `invoke>` command line, or indicate the desired scale on
+should "just work" without further intervention. Simply indicate the desired scale on
the popup in the Web GUI.
**GFPGAN** requires a series of downloadable model files to work. These are
@@ -41,48 +34,75 @@ reconstruction.
### Upscaling
-`-U : `
+Open the upscaling dialog by clicking on the "expand" icon located
+above the image display area in the Web UI:
-The upscaling prompt argument takes two values. The first value is a scaling
-factor and should be set to either `2` or `4` only. This will either scale the
-image 2x or 4x respectively using different models.
+
+
+
-You can set the scaling stength between `0` and `1.0` to control intensity of
-the of the scaling. This is handy because AI upscalers generally tend to smooth
-out texture details. If you wish to retain some of those for natural looking
-results, we recommend using values between `0.5 to 0.8`.
+There are three different upscaling parameters that you can
+adjust. The first is the scale itself, either 2x or 4x.
-If you do not explicitly specify an upscaling_strength, it will default to 0.75.
+The second is the "Denoising Strength." Higher values will smooth out
+the image and remove digital chatter, but may lose fine detail at
+higher values.
+
+Third, "Upscale Strength" allows you to adjust how the You can set the
+scaling stength between `0` and `1.0` to control the intensity of the
+scaling. AI upscalers generally tend to smooth out texture details. If
+you wish to retain some of those for natural looking results, we
+recommend using values between `0.5 to 0.8`.
+
+[This figure](../assets/features/upscaling-montage.png) illustrates
+the effects of denoising and strength. The original image was 512x512,
+4x scaled to 2048x2048. The "original" version on the upper left was
+scaled using simple pixel averaging. The remainder use the ESRGAN
+upscaling algorithm at different levels of denoising and strength.
+
+
+{ width=720 }
+
+
+Both denoising and strength default to 0.75.
### Face Restoration
-`-G : `
+InvokeAI offers alternative two face restoration algorithms,
+[GFPGAN](https://github.com/TencentARC/GFPGAN) and
+[CodeFormer](https://huggingface.co/spaces/sczhou/CodeFormer). These
+algorithms improve the appearance of faces, particularly eyes and
+mouths. Issues with faces are less common with the latest set of
+Stable Diffusion models than with the original 1.4 release, but the
+restoration algorithms can still make a noticeable improvement in
+certain cases. You can also apply restoration to old photographs you
+upload.
-This prompt argument controls the strength of the face restoration that is being
-applied. Similar to upscaling, values between `0.5 to 0.8` are recommended.
+To access face restoration, click the "smiley face" icon in the
+toolbar above the InvokeAI image panel. You will be presented with a
+dialog that offers a choice between the two algorithm and sliders that
+allow you to adjust their parameters. Alternatively, you may open the
+left-hand accordion panel labeled "Face Restoration" and have the
+restoration algorithm of your choice applied to generated images
+automatically.
-You can use either one or both without any conflicts. In cases where you use
-both, the image will be first upscaled and then the face restoration process
-will be executed to ensure you get the highest quality facial features.
-`--save_orig`
+Like upscaling, there are a number of parameters that adjust the face
+restoration output. GFPGAN has a single parameter, `strength`, which
+controls how much the algorithm is allowed to adjust the
+image. CodeFormer has two parameters, `strength`, and `fidelity`,
+which together control the quality of the output image as described in
+the [CodeFormer project
+page](https://shangchenzhou.com/projects/CodeFormer/). Default values
+are 0.75 for both parameters, which achieves a reasonable balance
+between changing the image too much and not enough.
-When you use either `-U` or `-G`, the final result you get is upscaled or face
-modified. If you want to save the original Stable Diffusion generation, you can
-use the `-save_orig` prompt argument to save the original unaffected version
-too.
+[This figure](../assets/features/restoration-montage.png) illustrates
+the effects of adjusting GFPGAN and CodeFormer parameters.
-### Example Usage
-
-```bash
-invoke> "superman dancing with a panda bear" -U 2 0.6 -G 0.4
-```
-
-This also works with img2img:
-
-```bash
-invoke> "a man wearing a pineapple hat" -I path/to/your/file.png -U 2 0.5 -G 0.6
-```
+
+{ width=720 }
+
!!! note
@@ -95,69 +115,8 @@ invoke> "a man wearing a pineapple hat" -I path/to/your/file.png -U 2 0.5 -G 0.6
process is complete. While the image generation is taking place, you will still be able to preview
the base images.
-If you wish to stop during the image generation but want to upscale or face
-restore a particular generated image, pass it again with the same prompt and
-generated seed along with the `-U` and `-G` prompt arguments to perform those
-actions.
-
-## CodeFormer Support
-
-This repo also allows you to perform face restoration using
-[CodeFormer](https://github.com/sczhou/CodeFormer).
-
-In order to setup CodeFormer to work, you need to download the models like with
-GFPGAN. You can do this either by running `invokeai-configure` or by manually
-downloading the
-[model file](https://github.com/sczhou/CodeFormer/releases/download/v0.1.0/codeformer.pth)
-and saving it to `ldm/invoke/restoration/codeformer/weights` folder.
-
-You can use `-ft` prompt argument to swap between CodeFormer and the default
-GFPGAN. The above mentioned `-G` prompt argument will allow you to control the
-strength of the restoration effect.
-
-### CodeFormer Usage
-
-The following command will perform face restoration with CodeFormer instead of
-the default gfpgan.
-
-` -G 0.8 -ft codeformer`
-
-### Other Options
-
-- `-cf` - cf or CodeFormer Fidelity takes values between `0` and `1`. 0 produces
- high quality results but low accuracy and 1 produces lower quality results but
- higher accuacy to your original face.
-
-The following command will perform face restoration with CodeFormer. CodeFormer
-will output a result that is closely matching to the input face.
-
-` -G 1.0 -ft codeformer -cf 0.9`
-
-The following command will perform face restoration with CodeFormer. CodeFormer
-will output a result that is the best restoration possible. This may deviate
-slightly from the original face. This is an excellent option to use in
-situations when there is very little facial data to work with.
-
-` -G 1.0 -ft codeformer -cf 0.1`
-
-## Fixing Previously-Generated Images
-
-It is easy to apply face restoration and/or upscaling to any
-previously-generated file. Just use the syntax
-`!fix path/to/file.png `. For example, to apply GFPGAN at strength 0.8
-and upscale 2X for a file named `./outputs/img-samples/000044.2945021133.png`,
-just run:
-
-```bash
-invoke> !fix ./outputs/img-samples/000044.2945021133.png -G 0.8 -U 2
-```
-
-A new file named `000044.2945021133.fixed.png` will be created in the output
-directory. Note that the `!fix` command does not replace the original file,
-unlike the behavior at generate time.
-
## How to disable
If, for some reason, you do not wish to load the GFPGAN and/or ESRGAN libraries,
you can disable them on the invoke.py command line with the `--no_restore` and
-`--no_upscale` options, respectively.
+`--no_esrgan` options, respectively.
diff --git a/docs/features/PROMPTS.md b/docs/features/PROMPTS.md
index 045e0d658a..1fd4550493 100644
--- a/docs/features/PROMPTS.md
+++ b/docs/features/PROMPTS.md
@@ -4,77 +4,12 @@ title: Prompting-Features
# :octicons-command-palette-24: Prompting-Features
-## **Reading Prompts from a File**
-
-You can automate `invoke.py` by providing a text file with the prompts you want
-to run, one line per prompt. The text file must be composed with a text editor
-(e.g. Notepad) and not a word processor. Each line should look like what you
-would type at the invoke> prompt:
-
-```bash
-"a beautiful sunny day in the park, children playing" -n4 -C10
-"stormy weather on a mountain top, goats grazing" -s100
-"innovative packaging for a squid's dinner" -S137038382
-```
-
-Then pass this file's name to `invoke.py` when you invoke it:
-
-```bash
-python scripts/invoke.py --from_file "/path/to/prompts.txt"
-```
-
-You may also read a series of prompts from standard input by providing
-a filename of `-`. For example, here is a python script that creates a
-matrix of prompts, each one varying slightly:
-
-```bash
-#!/usr/bin/env python
-
-adjectives = ['sunny','rainy','overcast']
-samplers = ['k_lms','k_euler_a','k_heun']
-cfg = [7.5, 9, 11]
-
-for adj in adjectives:
- for samp in samplers:
- for cg in cfg:
- print(f'a {adj} day -A{samp} -C{cg}')
-```
-
-Its output looks like this (abbreviated):
-
-```bash
-a sunny day -Aklms -C7.5
-a sunny day -Aklms -C9
-a sunny day -Aklms -C11
-a sunny day -Ak_euler_a -C7.5
-a sunny day -Ak_euler_a -C9
-...
-a overcast day -Ak_heun -C9
-a overcast day -Ak_heun -C11
-```
-
-To feed it to invoke.py, pass the filename of "-"
-
-```bash
-python matrix.py | python scripts/invoke.py --from_file -
-```
-
-When the script is finished, each of the 27 combinations
-of adjective, sampler and CFG will be executed.
-
-The command-line interface provides `!fetch` and `!replay` commands
-which allow you to read the prompts from a single previously-generated
-image or a whole directory of them, write the prompts to a file, and
-then replay them. Or you can create your own file of prompts and feed
-them to the command-line client from within an interactive session.
-See [Command-Line Interface](CLI.md) for details.
-
----
-
## **Negative and Unconditioned Prompts**
-Any words between a pair of square brackets will instruct Stable Diffusion to
-attempt to ban the concept from the generated image.
+Any words between a pair of square brackets will instruct Stable
+Diffusion to attempt to ban the concept from the generated image. The
+same effect is achieved by placing words in the "Negative Prompts"
+textbox in the Web UI.
```text
this is a test prompt [not really] to make you understand [cool] how this works.
@@ -87,7 +22,9 @@ Here's a prompt that depicts what it does.
original prompt:
-`#!bash "A fantastical translucent pony made of water and foam, ethereal, radiant, hyperalism, scottish folklore, digital painting, artstation, concept art, smooth, 8 k frostbite 3 engine, ultra detailed, art by artgerm and greg rutkowski and magali villeneuve" -s 20 -W 512 -H 768 -C 7.5 -A k_euler_a -S 1654590180`
+`#!bash "A fantastical translucent pony made of water and foam, ethereal, radiant, hyperalism, scottish folklore, digital painting, artstation, concept art, smooth, 8 k frostbite 3 engine, ultra detailed, art by artgerm and greg rutkowski and magali villeneuve"`
+
+`#!bash parameters: steps=20, dimensions=512x768, CFG=7.5, Scheduler=k_euler_a, seed=1654590180`
@@ -99,7 +36,8 @@ That image has a woman, so if we want the horse without a rider, we can
influence the image not to have a woman by putting [woman] in the prompt, like
this:
-`#!bash "A fantastical translucent poney made of water and foam, ethereal, radiant, hyperalism, scottish folklore, digital painting, artstation, concept art, smooth, 8 k frostbite 3 engine, ultra detailed, art by artgerm and greg rutkowski and magali villeneuve [woman]" -s 20 -W 512 -H 768 -C 7.5 -A k_euler_a -S 1654590180`
+`#!bash "A fantastical translucent poney made of water and foam, ethereal, radiant, hyperalism, scottish folklore, digital painting, artstation, concept art, smooth, 8 k frostbite 3 engine, ultra detailed, art by artgerm and greg rutkowski and magali villeneuve [woman]"`
+(same parameters as above)
@@ -110,7 +48,8 @@ this:
That's nice - but say we also don't want the image to be quite so blue. We can
add "blue" to the list of negative prompts, so it's now [woman blue]:
-`#!bash "A fantastical translucent poney made of water and foam, ethereal, radiant, hyperalism, scottish folklore, digital painting, artstation, concept art, smooth, 8 k frostbite 3 engine, ultra detailed, art by artgerm and greg rutkowski and magali villeneuve [woman blue]" -s 20 -W 512 -H 768 -C 7.5 -A k_euler_a -S 1654590180`
+`#!bash "A fantastical translucent poney made of water and foam, ethereal, radiant, hyperalism, scottish folklore, digital painting, artstation, concept art, smooth, 8 k frostbite 3 engine, ultra detailed, art by artgerm and greg rutkowski and magali villeneuve [woman blue]"`
+(same parameters as above)
@@ -121,7 +60,8 @@ add "blue" to the list of negative prompts, so it's now [woman blue]:
Getting close - but there's no sense in having a saddle when our horse doesn't
have a rider, so we'll add one more negative prompt: [woman blue saddle].
-`#!bash "A fantastical translucent poney made of water and foam, ethereal, radiant, hyperalism, scottish folklore, digital painting, artstation, concept art, smooth, 8 k frostbite 3 engine, ultra detailed, art by artgerm and greg rutkowski and magali villeneuve [woman blue saddle]" -s 20 -W 512 -H 768 -C 7.5 -A k_euler_a -S 1654590180`
+`#!bash "A fantastical translucent poney made of water and foam, ethereal, radiant, hyperalism, scottish folklore, digital painting, artstation, concept art, smooth, 8 k frostbite 3 engine, ultra detailed, art by artgerm and greg rutkowski and magali villeneuve [woman blue saddle]"`
+(same parameters as above)
@@ -261,19 +201,6 @@ Prompt2prompt `.swap()` is not compatible with xformers, which will be temporari
The `prompt2prompt` code is based off
[bloc97's colab](https://github.com/bloc97/CrossAttentionControl).
-Note that `prompt2prompt` is not currently working with the runwayML inpainting
-model, and may never work due to the way this model is set up. If you attempt to
-use `prompt2prompt` you will get the original image back. However, since this
-model is so good at inpainting, a good substitute is to use the `clipseg` text
-masking option:
-
-```bash
-invoke> a fluffy cat eating a hotdog
-Outputs:
-[1010] outputs/000025.2182095108.png: a fluffy cat eating a hotdog
-invoke> a smiling dog eating a hotdog -I 000025.2182095108.png -tm cat
-```
-
### Escaping parantheses () and speech marks ""
If the model you are using has parentheses () or speech marks "" as part of its
@@ -374,6 +301,48 @@ summoning up the concept of some sort of scifi creature? Let's find out.
Indeed, removing the word "hybrid" produces an image that is more like what we'd
expect.
-In conclusion, prompt blending is great for exploring creative space, but can be
-difficult to direct. A forthcoming release of InvokeAI will feature more
-deterministic prompt weighting.
+## Dynamic Prompts
+
+Dynamic Prompts are a powerful feature designed to produce a variety of prompts based on user-defined options. Using a special syntax, you can construct a prompt with multiple possibilities, and the system will automatically generate a series of permutations based on your settings. This is extremely beneficial for ideation, exploring various scenarios, or testing different concepts swiftly and efficiently.
+
+### Structure of a Dynamic Prompt
+
+A Dynamic Prompt comprises of regular text, supplemented with alternatives enclosed within curly braces {} and separated by a vertical bar |. For example: {option1|option2|option3}. The system will then select one of the options to include in the final prompt. This flexible system allows for options to be placed throughout the text as needed.
+
+Furthermore, Dynamic Prompts can designate multiple selections from a single group of options. This feature is triggered by prefixing the options with a numerical value followed by $$. For example, in {2$$option1|option2|option3}, the system will select two distinct options from the set.
+### Creating Dynamic Prompts
+
+To create a Dynamic Prompt, follow these steps:
+
+ Draft your sentence or phrase, identifying words or phrases with multiple possible options.
+ Encapsulate the different options within curly braces {}.
+ Within the braces, separate each option using a vertical bar |.
+ If you want to include multiple options from a single group, prefix with the desired number and $$.
+
+For instance: A {house|apartment|lodge|cottage} in {summer|winter|autumn|spring} designed in {2$$style1|style2|style3}.
+### How Dynamic Prompts Work
+
+Once a Dynamic Prompt is configured, the system generates an array of combinations using the options provided. Each group of options in curly braces is treated independently, with the system selecting one option from each group. For a prefixed set (e.g., 2$$), the system will select two distinct options.
+
+For example, the following prompts could be generated from the above Dynamic Prompt:
+
+ A house in summer designed in style1, style2
+ A lodge in autumn designed in style3, style1
+ A cottage in winter designed in style2, style3
+ And many more!
+
+When the `Combinatorial` setting is on, Invoke will disable the "Images" selection, and generate every combination up until the setting for Max Prompts is reached.
+When the `Combinatorial` setting is off, Invoke will randomly generate combinations up until the setting for Images has been reached.
+
+
+
+### Tips and Tricks for Using Dynamic Prompts
+
+Below are some useful strategies for creating Dynamic Prompts:
+
+ Utilize Dynamic Prompts to generate a wide spectrum of prompts, perfect for brainstorming and exploring diverse ideas.
+ Ensure that the options within a group are contextually relevant to the part of the sentence where they are used. For instance, group building types together, and seasons together.
+ Apply the 2$$ prefix when you want to incorporate more than one option from a single group. This becomes quite handy when mixing and matching different elements.
+ Experiment with different quantities for the prefix. For example, 3$$ will select three distinct options.
+ Be aware of coherence in your prompts. Although the system can generate all possible combinations, not all may semantically make sense. Therefore, carefully choose the options for each group.
+ Always review and fine-tune the generated prompts as needed. While Dynamic Prompts can help you generate a multitude of combinations, the final polishing and refining remain in your hands.
diff --git a/docs/features/TEXTUAL_INVERSION.md b/docs/features/TRAINING.md
similarity index 93%
rename from docs/features/TEXTUAL_INVERSION.md
rename to docs/features/TRAINING.md
index e90bd1a7f4..41197a334f 100644
--- a/docs/features/TEXTUAL_INVERSION.md
+++ b/docs/features/TRAINING.md
@@ -1,9 +1,10 @@
---
-title: Textual-Inversion
+title: Training
---
-# :material-file-document: Textual Inversion
+# :material-file-document: Training
+# Textual Inversion Training
## **Personalizing Text-to-Image Generation**
You may personalize the generated images to provide your own styles or objects
@@ -46,11 +47,19 @@ start the front end by selecting choice (3):
```sh
Do you want to generate images using the
-1. command-line
-2. browser-based UI
-3. textual inversion training
-4. open the developer console
-Please enter 1, 2, 3, or 4: [1] 3
+1: Browser-based UI
+2: Command-line interface
+3: Run textual inversion training
+4: Merge models (diffusers type only)
+5: Download and install models
+6: Change InvokeAI startup options
+7: Re-run the configure script to fix a broken install
+8: Open the developer console
+9: Update InvokeAI
+10: Command-line help
+Q: Quit
+
+Please enter 1-10, Q: [1]
```
From the command line, with the InvokeAI virtual environment active,
@@ -250,16 +259,6 @@ invokeai-ti \
--only_save_embeds
```
-## Using Embeddings
-
-After training completes, the resultant embeddings will be saved into your `$INVOKEAI_ROOT/embeddings//learned_embeds.bin`.
-
-These will be automatically loaded when you start InvokeAI.
-
-Add the trigger word, surrounded by angle brackets, to use that embedding. For example, if your trigger word was `terence`, use `` in prompts. This is the same syntax used by the HuggingFace concepts library.
-
-**Note:** `.pt` embeddings do not require the angle brackets.
-
## Troubleshooting
### `Cannot load embedding for . It was trained on a model with token dimension 1024, but the current model has token dimension 768`
diff --git a/docs/features/VARIATIONS.md b/docs/features/VARIATIONS.md
index aeeb57d52c..e6e21490c8 100644
--- a/docs/features/VARIATIONS.md
+++ b/docs/features/VARIATIONS.md
@@ -6,9 +6,7 @@ title: Variations
## Intro
-Release 1.13 of SD-Dream adds support for image variations.
-
-You are able to do the following:
+InvokeAI's support for variations enables you to do the following:
1. Generate a series of systematic variations of an image, given a prompt. The
amount of variation from one image to the next can be controlled.
@@ -30,19 +28,7 @@ The prompt we will use throughout is:
This will be indicated as `#!bash "prompt"` in the examples below.
First we let SD create a series of images in the usual way, in this case
-requesting six iterations:
-
-```bash
-invoke> lucy lawless as xena, warrior princess, character portrait, high resolution -n6
-...
-Outputs:
-./outputs/Xena/000001.1579445059.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -S1579445059
-./outputs/Xena/000001.1880768722.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -S1880768722
-./outputs/Xena/000001.332057179.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -S332057179
-./outputs/Xena/000001.2224800325.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -S2224800325
-./outputs/Xena/000001.465250761.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -S465250761
-./outputs/Xena/000001.3357757885.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -S3357757885
-```
+requesting six iterations.
@@ -53,22 +39,16 @@ Outputs:
## Step 2 - Generating Variations
-Let's try to generate some variations. Using the same seed, we pass the argument
-`-v0.1` (or --variant_amount), which generates a series of variations each
-differing by a variation amount of 0.2. This number ranges from `0` to `1.0`,
-with higher numbers being larger amounts of variation.
+Let's try to generate some variations on this image. We select the "*"
+symbol in the line of icons above the image in order to fix the prompt
+and seed. Then we open up the "Variations" section of the generation
+panel and use the slider to set the variation amount to 0.2. The
+higher this value, the more each generated image will differ from the
+previous one.
-```bash
-invoke> "prompt" -n6 -S3357757885 -v0.2
-...
-Outputs:
-./outputs/Xena/000002.784039624.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -V 784039624:0.2 -S3357757885
-./outputs/Xena/000002.3647897225.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -V 3647897225:0.2 -S3357757885
-./outputs/Xena/000002.917731034.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -V 917731034:0.2 -S3357757885
-./outputs/Xena/000002.4116285959.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -V 4116285959:0.2 -S3357757885
-./outputs/Xena/000002.1614299449.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -V 1614299449:0.2 -S3357757885
-./outputs/Xena/000002.1335553075.png: "prompt" -s50 -W512 -H512 -C7.5 -Ak_lms -V 1335553075:0.2 -S3357757885
-```
+Now we run the prompt a second time, requesting six iterations. You
+will see six images that are thematically related to each other. Try
+increasing and decreasing the variation amount and see what happens.
### **Variation Sub Seeding**
diff --git a/docs/features/WEB.md b/docs/features/WEB.md
index b904f147b4..ef541736d9 100644
--- a/docs/features/WEB.md
+++ b/docs/features/WEB.md
@@ -299,14 +299,6 @@ initial image" icons are located.
See the [Unified Canvas Guide](UNIFIED_CANVAS.md)
-## Parting remarks
-
-This concludes the walkthrough, but there are several more features that you can
-explore. Please check out the [Command Line Interface](CLI.md) documentation for
-further explanation of the advanced features that were not covered here.
-
-The WebUI is only rapid development. Check back regularly for updates!
-
## Reference
### Additional Options
@@ -349,11 +341,9 @@ the settings configured in the toolbar.
See below for additional documentation related to each feature:
-- [Core Prompt Settings](./CLI.md)
- [Variations](./VARIATIONS.md)
- [Upscaling](./POSTPROCESS.md#upscaling)
- [Image to Image](./IMG2IMG.md)
-- [Inpainting](./INPAINTING.md)
- [Other](./OTHER.md)
#### Invocation Gallery
diff --git a/docs/features/index.md b/docs/features/index.md
index 53d380f3fb..12539ca98f 100644
--- a/docs/features/index.md
+++ b/docs/features/index.md
@@ -13,28 +13,16 @@ Build complex scenes by combine and modifying multiple images in a stepwise
fashion. This feature combines img2img, inpainting and outpainting in
a single convenient digital artist-optimized user interface.
-### * The [Command Line Interface (CLI)](CLI.md)
-Scriptable access to InvokeAI's features.
-
## Image Generation
### * [Prompt Engineering](PROMPTS.md)
Get the images you want with the InvokeAI prompt engineering language.
-## * [Post-Processing](POSTPROCESS.md)
-Restore mangled faces and make images larger with upscaling. Also see the [Embiggen Upscaling Guide](EMBIGGEN.md).
-
## * The [Concepts Library](CONCEPTS.md)
Add custom subjects and styles using HuggingFace's repository of embeddings.
-### * [Image-to-Image Guide for the CLI](IMG2IMG.md)
+### * [Image-to-Image Guide](IMG2IMG.md)
Use a seed image to build new creations in the CLI.
-### * [Inpainting Guide for the CLI](INPAINTING.md)
-Selectively erase and replace portions of an existing image in the CLI.
-
-### * [Outpainting Guide for the CLI](OUTPAINTING.md)
-Extend the borders of the image with an "outcrop" function within the CLI.
-
### * [Generating Variations](VARIATIONS.md)
Have an image you like and want to generate many more like it? Variations
are the ticket.
diff --git a/docs/index.md b/docs/index.md
index 0aa99a1747..3af8571df9 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -13,6 +13,7 @@ title: Home
+
[](https://github.com/invoke-ai/InvokeAI)
[![discord badge]][discord link]
@@ -131,17 +132,13 @@ This method is recommended for those familiar with running Docker containers
- [WebUI overview](features/WEB.md)
- [WebUI hotkey reference guide](features/WEBUIHOTKEYS.md)
- [WebUI Unified Canvas for Img2Img, inpainting and outpainting](features/UNIFIED_CANVAS.md)
+
-### The InvokeAI Command Line Interface
-- [Command Line Interace Reference Guide](features/CLI.md)
-
+
### Image Management
- [Image2Image](features/IMG2IMG.md)
-- [Inpainting](features/INPAINTING.md)
-- [Outpainting](features/OUTPAINTING.md)
- [Adding custom styles and subjects](features/CONCEPTS.md)
- [Upscaling and Face Reconstruction](features/POSTPROCESS.md)
-- [Embiggen upscaling](features/EMBIGGEN.md)
- [Other Features](features/OTHER.md)
@@ -156,83 +153,60 @@ This method is recommended for those familiar with running Docker containers
- [Prompt Syntax](features/PROMPTS.md)
- [Generating Variations](features/VARIATIONS.md)
-## :octicons-log-16: Latest Changes
+## :octicons-log-16: Important Changes Since Version 2.3
-### v2.3.0 (9 February 2023)
+### Nodes
-#### Migration to Stable Diffusion `diffusers` models
+Behind the scenes, InvokeAI has been completely rewritten to support
+"nodes," small unitary operations that can be combined into graphs to
+form arbitrary workflows. For example, there is a prompt node that
+processes the prompt string and feeds it to a text2latent node that
+generates a latent image. The latents are then fed to a latent2image
+node that translates the latent image into a PNG.
-Previous versions of InvokeAI supported the original model file format introduced with Stable Diffusion 1.4. In the original format, known variously as "checkpoint", or "legacy" format, there is a single large weights file ending with `.ckpt` or `.safetensors`. Though this format has served the community well, it has a number of disadvantages, including file size, slow loading times, and a variety of non-standard variants that require special-case code to handle. In addition, because checkpoint files are actually a bundle of multiple machine learning sub-models, it is hard to swap different sub-models in and out, or to share common sub-models. A new format, introduced by the StabilityAI company in collaboration with HuggingFace, is called `diffusers` and consists of a directory of individual models. The most immediate benefit of `diffusers` is that they load from disk very quickly. A longer term benefit is that in the near future `diffusers` models will be able to share common sub-models, dramatically reducing disk space when you have multiple fine-tune models derived from the same base.
+The WebGUI has a node editor that allows you to graphically design and
+execute custom node graphs. The ability to save and load graphs is
+still a work in progress, but coming soon.
-When you perform a new install of version 2.3.0, you will be offered the option to install the `diffusers` versions of a number of popular SD models, including Stable Diffusion versions 1.5 and 2.1 (including the 768x768 pixel version of 2.1). These will act and work just like the checkpoint versions. Do not be concerned if you already have a lot of ".ckpt" or ".safetensors" models on disk! InvokeAI 2.3.0 can still load these and generate images from them without any extra intervention on your part.
+### Command-Line Interface Retired
-To take advantage of the optimized loading times of `diffusers` models, InvokeAI offers options to convert legacy checkpoint models into optimized `diffusers` models. If you use the `invokeai` command line interface, the relevant commands are:
+The original "invokeai" command-line interface has been retired. The
+`invokeai` command will now launch a new command-line client that can
+be used by developers to create and test nodes. It is not intended to
+be used for routine image generation or manipulation.
-* `!convert_model` -- Take the path to a local checkpoint file or a URL that is pointing to one, convert it into a `diffusers` model, and import it into InvokeAI's models registry file.
-* `!optimize_model` -- If you already have a checkpoint model in your InvokeAI models file, this command will accept its short name and convert it into a like-named `diffusers` model, optionally deleting the original checkpoint file.
-* `!import_model` -- Take the local path of either a checkpoint file or a `diffusers` model directory and import it into InvokeAI's registry file. You may also provide the ID of any diffusers model that has been published on the [HuggingFace models repository](https://huggingface.co/models?pipeline_tag=text-to-image&sort=downloads) and it will be downloaded and installed automatically.
+To launch the Web GUI from the command-line, use the command
+`invokeai-web` rather than the traditional `invokeai --web`.
-The WebGUI offers similar functionality for model management.
+### ControlNet
-For advanced users, new command-line options provide additional functionality. Launching `invokeai` with the argument `--autoconvert ` takes the path to a directory of checkpoint files, automatically converts them into `diffusers` models and imports them. Each time the script is launched, the directory will be scanned for new checkpoint files to be loaded. Alternatively, the `--ckpt_convert` argument will cause any checkpoint or safetensors model that is already registered with InvokeAI to be converted into a `diffusers` model on the fly, allowing you to take advantage of future diffusers-only features without explicitly converting the model and saving it to disk.
+This version of InvokeAI features ControlNet, a system that allows you
+to achieve exact poses for human and animal figures by providing a
+model to follow. Full details are found in [ControlNet](features/CONTROLNET.md)
-Please see [INSTALLING MODELS](https://invoke-ai.github.io/InvokeAI/installation/050_INSTALLING_MODELS/) for more information on model management in both the command-line and Web interfaces.
+### New Schedulers
-#### Support for the `XFormers` Memory-Efficient Crossattention Package
+The list of schedulers has been completely revamped and brought up to date:
-On CUDA (Nvidia) systems, version 2.3.0 supports the `XFormers` library. Once installed, the`xformers` package dramatically reduces the memory footprint of loaded Stable Diffusion models files and modestly increases image generation speed. `xformers` will be installed and activated automatically if you specify a CUDA system at install time.
+| **Short Name** | **Scheduler** | **Notes** |
+|----------------|---------------------------------|-----------------------------|
+| **ddim** | DDIMScheduler | |
+| **ddpm** | DDPMScheduler | |
+| **deis** | DEISMultistepScheduler | |
+| **lms** | LMSDiscreteScheduler | |
+| **pndm** | PNDMScheduler | |
+| **heun** | HeunDiscreteScheduler | original noise schedule |
+| **heun_k** | HeunDiscreteScheduler | using karras noise schedule |
+| **euler** | EulerDiscreteScheduler | original noise schedule |
+| **euler_k** | EulerDiscreteScheduler | using karras noise schedule |
+| **kdpm_2** | KDPM2DiscreteScheduler | |
+| **kdpm_2_a** | KDPM2AncestralDiscreteScheduler | |
+| **dpmpp_2s** | DPMSolverSinglestepScheduler | |
+| **dpmpp_2m** | DPMSolverMultistepScheduler | original noise scnedule |
+| **dpmpp_2m_k** | DPMSolverMultistepScheduler | using karras noise schedule |
+| **unipc** | UniPCMultistepScheduler | CPU only |
-The caveat with using `xformers` is that it introduces slightly non-deterministic behavior, and images generated using the same seed and other settings will be subtly different between invocations. Generally the changes are unnoticeable unless you rapidly shift back and forth between images, but to disable `xformers` and restore fully deterministic behavior, you may launch InvokeAI using the `--no-xformers` option. This is most conveniently done by opening the file `invokeai/invokeai.init` with a text editor, and adding the line `--no-xformers` at the bottom.
-
-#### A Negative Prompt Box in the WebUI
-
-There is now a separate text input box for negative prompts in the WebUI. This is convenient for stashing frequently-used negative prompts ("mangled limbs, bad anatomy"). The `[negative prompt]` syntax continues to work in the main prompt box as well.
-
-To see exactly how your prompts are being parsed, launch `invokeai` with the `--log_tokenization` option. The console window will then display the tokenization process for both positive and negative prompts.
-
-#### Model Merging
-
-Version 2.3.0 offers an intuitive user interface for merging up to three Stable Diffusion models using an intuitive user interface. Model merging allows you to mix the behavior of models to achieve very interesting effects. To use this, each of the models must already be imported into InvokeAI and saved in `diffusers` format, then launch the merger using a new menu item in the InvokeAI launcher script (`invoke.sh`, `invoke.bat`) or directly from the command line with `invokeai-merge --gui`. You will be prompted to select the models to merge, the proportions in which to mix them, and the mixing algorithm. The script will create a new merged `diffusers` model and import it into InvokeAI for your use.
-
-See [MODEL MERGING](https://invoke-ai.github.io/InvokeAI/features/MODEL_MERGING/) for more details.
-
-#### Textual Inversion Training
-
-Textual Inversion (TI) is a technique for training a Stable Diffusion model to emit a particular subject or style when triggered by a keyword phrase. You can perform TI training by placing a small number of images of the subject or style in a directory, and choosing a distinctive trigger phrase, such as "pointillist-style". After successful training, The subject or style will be activated by including `` in your prompt.
-
-Previous versions of InvokeAI were able to perform TI, but it required using a command-line script with dozens of obscure command-line arguments. Version 2.3.0 features an intuitive TI frontend that will build a TI model on top of any `diffusers` model. To access training you can launch from a new item in the launcher script or from the command line using `invokeai-ti --gui`.
-
-See [TEXTUAL INVERSION](https://invoke-ai.github.io/InvokeAI/features/TEXTUAL_INVERSION/) for further details.
-
-#### A New Installer Experience
-
-The InvokeAI installer has been upgraded in order to provide a smoother and hopefully more glitch-free experience. In addition, InvokeAI is now packaged as a PyPi project, allowing developers and power-users to install InvokeAI with the command `pip install InvokeAI --use-pep517`. Please see [Installation](#installation) for details.
-
-Developers should be aware that the `pip` installation procedure has been simplified and that the `conda` method is no longer supported at all. Accordingly, the `environments_and_requirements` directory has been deleted from the repository.
-
-#### Command-line name changes
-
-All of InvokeAI's functionality, including the WebUI, command-line interface, textual inversion training and model merging, can all be accessed from the `invoke.sh` and `invoke.bat` launcher scripts. The menu of options has been expanded to add the new functionality. For the convenience of developers and power users, we have normalized the names of the InvokeAI command-line scripts:
-
-* `invokeai` -- Command-line client
-* `invokeai --web` -- Web GUI
-* `invokeai-merge --gui` -- Model merging script with graphical front end
-* `invokeai-ti --gui` -- Textual inversion script with graphical front end
-* `invokeai-configure` -- Configuration tool for initializing the `invokeai` directory and selecting popular starter models.
-
-For backward compatibility, the old command names are also recognized, including `invoke.py` and `configure-invokeai.py`. However, these are deprecated and will eventually be removed.
-
-Developers should be aware that the locations of the script's source code has been moved. The new locations are:
-* `invokeai` => `ldm/invoke/CLI.py`
-* `invokeai-configure` => `ldm/invoke/config/configure_invokeai.py`
-* `invokeai-ti`=> `ldm/invoke/training/textual_inversion.py`
-* `invokeai-merge` => `ldm/invoke/merge_diffusers`
-
-Developers are strongly encouraged to perform an "editable" install of InvokeAI using `pip install -e . --use-pep517` in the Git repository, and then to call the scripts using their 2.3.0 names, rather than executing the scripts directly. Developers should also be aware that the several important data files have been relocated into a new directory named `invokeai`. This includes the WebGUI's `frontend` and `backend` directories, and the `INITIAL_MODELS.yaml` files used by the installer to select starter models. Eventually all InvokeAI modules will be in subdirectories of `invokeai`.
-
-Please see [2.3.0 Release Notes](https://github.com/invoke-ai/InvokeAI/releases/tag/v2.3.0) for further details.
-For older changelogs, please visit the
-**[CHANGELOG](CHANGELOG/#v223-2-december-2022)**.
+Please see [3.0.0 Release Notes](https://github.com/invoke-ai/InvokeAI/releases/tag/v3.0.0) for further details.
## :material-target: Troubleshooting
@@ -268,8 +242,3 @@ free to send me an email if you use and like the script.
Original portions of the software are Copyright (c) 2022-23
by [The InvokeAI Team](https://github.com/invoke-ai).
-## :octicons-book-24: Further Reading
-
-Please see the original README for more information on this software and
-underlying algorithm, located in the file
-[README-CompViz.md](other/README-CompViz.md).
diff --git a/docs/installation/060_INSTALL_PATCHMATCH.md b/docs/installation/060_INSTALL_PATCHMATCH.md
index 6ac6e2814e..ccfd19d207 100644
--- a/docs/installation/060_INSTALL_PATCHMATCH.md
+++ b/docs/installation/060_INSTALL_PATCHMATCH.md
@@ -87,18 +87,18 @@ Prior to installing PyPatchMatch, you need to take the following steps:
sudo pacman -S --needed base-devel
```
-2. Install `opencv`:
+2. Install `opencv` and `blas`:
```sh
- sudo pacman -S opencv
+ sudo pacman -S opencv blas
```
or for CUDA support
```sh
- sudo pacman -S opencv-cuda
+ sudo pacman -S opencv-cuda blas
```
-
+
3. Fix the naming of the `opencv` package configuration file:
```sh
diff --git a/installer/install.bat.in b/installer/install.bat.in
index 924f2cd40d..8f729e758d 100644
--- a/installer/install.bat.in
+++ b/installer/install.bat.in
@@ -38,6 +38,7 @@ echo https://learn.microsoft.com/en-US/cpp/windows/latest-supported-vc-redist
echo.
echo See %INSTRUCTIONS% for more details.
echo.
+echo "For the best user experience we suggest enlarging or maximizing this window now."
pause
@rem ---------------------------- check Python version ---------------
diff --git a/installer/install.sh.in b/installer/install.sh.in
index c1014b2496..0063c0ebbf 100755
--- a/installer/install.sh.in
+++ b/installer/install.sh.in
@@ -25,7 +25,8 @@ done
if [ -z "$PYTHON" ]; then
echo "A suitable Python interpreter could not be found"
- echo "Please install Python 3.9 or higher before running this script. See instructions at $INSTRUCTIONS for help."
+ echo "Please install Python $MINIMUM_PYTHON_VERSION or higher (maximum $MAXIMUM_PYTHON_VERSION) before running this script. See instructions at $INSTRUCTIONS for help."
+ echo "For the best user experience we suggest enlarging or maximizing this window now."
read -p "Press any key to exit"
exit -1
fi
diff --git a/installer/lib/installer.py b/installer/lib/installer.py
index 96d8ce55e3..5904a71968 100644
--- a/installer/lib/installer.py
+++ b/installer/lib/installer.py
@@ -149,7 +149,7 @@ class Installer:
return venv_dir
- def install(self, root: str = "~/invokeai", version: str = "latest", yes_to_all=False, find_links: Path = None) -> None:
+ def install(self, root: str = "~/invokeai-3", version: str = "latest", yes_to_all=False, find_links: Path = None) -> None:
"""
Install the InvokeAI application into the given runtime path
@@ -248,6 +248,7 @@ class InvokeAiInstance:
"install",
"--require-virtualenv",
"torch~=2.0.0",
+ "torchmetrics==0.11.4",
"torchvision>=0.14.1",
"--force-reinstall",
"--find-links" if find_links is not None else None,
diff --git a/installer/lib/messages.py b/installer/lib/messages.py
index dfc09dd94a..196c0e1eb0 100644
--- a/installer/lib/messages.py
+++ b/installer/lib/messages.py
@@ -293,6 +293,8 @@ def introduction() -> None:
"3. Create initial configuration files.",
"",
"[i]At any point you may interrupt this program and resume later.",
+ "",
+ "[b]For the best user experience, please enlarge or maximize this window",
),
)
)
diff --git a/installer/templates/invoke.bat.in b/installer/templates/invoke.bat.in
index b5816164a3..7900eac1cf 100644
--- a/installer/templates/invoke.bat.in
+++ b/installer/templates/invoke.bat.in
@@ -14,13 +14,13 @@ echo 3. Run textual inversion training
echo 4. Merge models (diffusers type only)
echo 5. Download and install models
echo 6. Change InvokeAI startup options
-echo 7. Re-run the configure script to fix a broken install
+echo 7. Re-run the configure script to fix a broken install or to complete a major upgrade
echo 8. Open the developer console
echo 9. Update InvokeAI
echo 10. Command-line help
echo Q - Quit
set /P choice="Please enter 1-10, Q: [2] "
-if not defined choice set choice=2
+if not defined choice set choice=1
IF /I "%choice%" == "1" (
echo Starting the InvokeAI browser-based UI..
python .venv\Scripts\invokeai-web.exe %*
@@ -56,7 +56,7 @@ IF /I "%choice%" == "1" (
call cmd /k
) ELSE IF /I "%choice%" == "9" (
echo Running invokeai-update...
- python .venv\Scripts\invokeai-update.exe %*
+ python -m invokeai.frontend.install.invokeai_update
) ELSE IF /I "%choice%" == "10" (
echo Displaying command line help...
python .venv\Scripts\invokeai.exe --help %*
diff --git a/installer/templates/invoke.sh.in b/installer/templates/invoke.sh.in
index e457d57842..cb54074aa2 100644
--- a/installer/templates/invoke.sh.in
+++ b/installer/templates/invoke.sh.in
@@ -81,7 +81,7 @@ do_choice() {
;;
7)
clear
- printf "Re-run the configure script to fix a broken install\n"
+ printf "Re-run the configure script to fix a broken install or to complete a major upgrade\n"
invokeai-configure --root ${INVOKEAI_ROOT} --yes --default_only
;;
8)
@@ -93,7 +93,7 @@ do_choice() {
9)
clear
printf "Update InvokeAI\n"
- invokeai-update
+ python -m invokeai.frontend.install.invokeai_update
;;
10)
clear
@@ -118,12 +118,12 @@ do_choice() {
do_dialog() {
options=(
1 "Generate images with a browser-based interface"
- 2 "Generate images using a command-line interface"
+ 2 "Explore InvokeAI nodes using a command-line interface"
3 "Textual inversion training"
4 "Merge models (diffusers type only)"
5 "Download and install models"
6 "Change InvokeAI startup options"
- 7 "Re-run the configure script to fix a broken install"
+ 7 "Re-run the configure script to fix a broken install or to complete a major upgrade"
8 "Open the developer console"
9 "Update InvokeAI")
diff --git a/invokeai/app/api/dependencies.py b/invokeai/app/api/dependencies.py
index efeb778922..cb867354a5 100644
--- a/invokeai/app/api/dependencies.py
+++ b/invokeai/app/api/dependencies.py
@@ -13,10 +13,10 @@ from invokeai.app.services.board_record_storage import SqliteBoardRecordStorage
from invokeai.app.services.boards import BoardService, BoardServiceDependencies
from invokeai.app.services.image_record_storage import SqliteImageRecordStorage
from invokeai.app.services.images import ImageService, ImageServiceDependencies
-from invokeai.app.services.metadata import CoreMetadataService
from invokeai.app.services.resource_name import SimpleNameService
from invokeai.app.services.urls import LocalUrlService
from invokeai.backend.util.logging import InvokeAILogger
+from invokeai.version.invokeai_version import __version__
from ..services.default_graphs import create_system_graphs
from ..services.latent_storage import DiskLatentsStorage, ForwardCacheLatentsStorage
@@ -58,7 +58,8 @@ class ApiDependencies:
@staticmethod
def initialize(config, event_handler_id: int, logger: Logger = logger):
- logger.info(f"Internet connectivity is {config.internet_available}")
+ logger.debug(f'InvokeAI version {__version__}')
+ logger.debug(f"Internet connectivity is {config.internet_available}")
events = FastAPIEventService(event_handler_id)
@@ -73,7 +74,6 @@ class ApiDependencies:
)
urls = LocalUrlService()
- metadata = CoreMetadataService()
image_record_storage = SqliteImageRecordStorage(db_location)
image_file_storage = DiskImageFileStorage(f"{output_folder}/images")
names = SimpleNameService()
@@ -109,7 +109,6 @@ class ApiDependencies:
board_image_record_storage=board_image_record_storage,
image_record_storage=image_record_storage,
image_file_storage=image_file_storage,
- metadata=metadata,
url=urls,
logger=logger,
names=names,
diff --git a/invokeai/app/api/routers/app_info.py b/invokeai/app/api/routers/app_info.py
new file mode 100644
index 0000000000..8e2955c9aa
--- /dev/null
+++ b/invokeai/app/api/routers/app_info.py
@@ -0,0 +1,36 @@
+from fastapi.routing import APIRouter
+from pydantic import BaseModel, Field
+
+from invokeai.backend.image_util.patchmatch import PatchMatch
+from invokeai.version import __version__
+
+app_router = APIRouter(prefix="/v1/app", tags=["app"])
+
+
+class AppVersion(BaseModel):
+ """App Version Response"""
+
+ version: str = Field(description="App version")
+
+
+class AppConfig(BaseModel):
+ """App Config Response"""
+
+ infill_methods: list[str] = Field(description="List of available infill methods")
+
+
+@app_router.get(
+ "/version", operation_id="app_version", status_code=200, response_model=AppVersion
+)
+async def get_version() -> AppVersion:
+ return AppVersion(version=__version__)
+
+
+@app_router.get(
+ "/config", operation_id="get_config", status_code=200, response_model=AppConfig
+)
+async def get_config() -> AppConfig:
+ infill_methods = ['tile']
+ if PatchMatch.patchmatch_available():
+ infill_methods.append('patchmatch')
+ return AppConfig(infill_methods=infill_methods)
diff --git a/invokeai/app/api/routers/boards.py b/invokeai/app/api/routers/boards.py
index 55cd7c8ca2..94d8667ae4 100644
--- a/invokeai/app/api/routers/boards.py
+++ b/invokeai/app/api/routers/boards.py
@@ -5,6 +5,7 @@ from invokeai.app.services.board_record_storage import BoardChanges
from invokeai.app.services.image_record_storage import OffsetPaginatedResults
from invokeai.app.services.models.board_record import BoardDTO
+
from ..dependencies import ApiDependencies
boards_router = APIRouter(prefix="/v1/boards", tags=["boards"])
@@ -71,11 +72,19 @@ async def update_board(
@boards_router.delete("/{board_id}", operation_id="delete_board")
async def delete_board(
board_id: str = Path(description="The id of board to delete"),
+ include_images: Optional[bool] = Query(
+ description="Permanently delete all images on the board", default=False
+ ),
) -> None:
"""Deletes a board"""
-
try:
- ApiDependencies.invoker.services.boards.delete(board_id=board_id)
+ if include_images is True:
+ ApiDependencies.invoker.services.images.delete_images_on_board(
+ board_id=board_id
+ )
+ ApiDependencies.invoker.services.boards.delete(board_id=board_id)
+ else:
+ ApiDependencies.invoker.services.boards.delete(board_id=board_id)
except Exception as e:
# TODO: Does this need any exception handling at all?
pass
diff --git a/invokeai/app/api/routers/images.py b/invokeai/app/api/routers/images.py
index a8c84b81b9..a0428e772e 100644
--- a/invokeai/app/api/routers/images.py
+++ b/invokeai/app/api/routers/images.py
@@ -1,20 +1,19 @@
import io
from typing import Optional
-from fastapi import Body, HTTPException, Path, Query, Request, Response, UploadFile
-from fastapi.routing import APIRouter
+
+from fastapi import (Body, HTTPException, Path, Query, Request, Response,
+ UploadFile)
from fastapi.responses import FileResponse
+from fastapi.routing import APIRouter
from PIL import Image
-from invokeai.app.models.image import (
- ImageCategory,
- ResourceOrigin,
-)
+
+from invokeai.app.invocations.metadata import ImageMetadata
+from invokeai.app.models.image import ImageCategory, ResourceOrigin
from invokeai.app.services.image_record_storage import OffsetPaginatedResults
-from invokeai.app.services.models.image_record import (
- ImageDTO,
- ImageRecordChanges,
- ImageUrlsDTO,
-)
from invokeai.app.services.item_storage import PaginatedResults
+from invokeai.app.services.models.image_record import (ImageDTO,
+ ImageRecordChanges,
+ ImageUrlsDTO)
from ..dependencies import ApiDependencies
@@ -103,23 +102,38 @@ async def update_image(
@images_router.get(
- "/{image_name}/metadata",
- operation_id="get_image_metadata",
+ "/{image_name}",
+ operation_id="get_image_dto",
response_model=ImageDTO,
)
-async def get_image_metadata(
+async def get_image_dto(
image_name: str = Path(description="The name of image to get"),
) -> ImageDTO:
- """Gets an image's metadata"""
+ """Gets an image's DTO"""
try:
return ApiDependencies.invoker.services.images.get_dto(image_name)
except Exception as e:
raise HTTPException(status_code=404)
+@images_router.get(
+ "/{image_name}/metadata",
+ operation_id="get_image_metadata",
+ response_model=ImageMetadata,
+)
+async def get_image_metadata(
+ image_name: str = Path(description="The name of image to get"),
+) -> ImageMetadata:
+ """Gets an image's metadata"""
+
+ try:
+ return ApiDependencies.invoker.services.images.get_metadata(image_name)
+ except Exception as e:
+ raise HTTPException(status_code=404)
+
@images_router.get(
- "/{image_name}",
+ "/{image_name}/full",
operation_id="get_image_full",
response_class=Response,
responses={
@@ -208,10 +222,10 @@ async def get_image_urls(
@images_router.get(
"/",
- operation_id="list_images_with_metadata",
+ operation_id="list_image_dtos",
response_model=OffsetPaginatedResults[ImageDTO],
)
-async def list_images_with_metadata(
+async def list_image_dtos(
image_origin: Optional[ResourceOrigin] = Query(
default=None, description="The origin of images to list"
),
@@ -227,7 +241,7 @@ async def list_images_with_metadata(
offset: int = Query(default=0, description="The page offset"),
limit: int = Query(default=10, description="The number of images per page"),
) -> OffsetPaginatedResults[ImageDTO]:
- """Gets a list of images"""
+ """Gets a list of image DTOs"""
image_dtos = ApiDependencies.invoker.services.images.get_many(
offset,
diff --git a/invokeai/app/api/routers/models.py b/invokeai/app/api/routers/models.py
index 50d645eb57..8dbeaa3d05 100644
--- a/invokeai/app/api/routers/models.py
+++ b/invokeai/app/api/routers/models.py
@@ -1,69 +1,30 @@
-# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654) and 2023 Kent Keirsey (https://github.com/hipsterusername)
+# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654), 2023 Kent Keirsey (https://github.com/hipsterusername), 2024 Lincoln Stein
-from typing import Annotated, Literal, Optional, Union, Dict
-from fastapi import Query
-from fastapi.routing import APIRouter, HTTPException
-from pydantic import BaseModel, Field, parse_obj_as
-from ..dependencies import ApiDependencies
+from typing import Literal, List, Optional, Union
+
+from fastapi import Body, Path, Query, Response
+from fastapi.routing import APIRouter
+from pydantic import BaseModel, parse_obj_as
+from starlette.exceptions import HTTPException
+
from invokeai.backend import BaseModelType, ModelType
-from invokeai.backend.model_management.models import OPENAPI_MODEL_CONFIGS
-MODEL_CONFIGS = Union[tuple(OPENAPI_MODEL_CONFIGS)]
+from invokeai.backend.model_management.models import (
+ OPENAPI_MODEL_CONFIGS,
+ SchedulerPredictionType,
+)
+from invokeai.backend.model_management import MergeInterpolationMethod
+from ..dependencies import ApiDependencies
models_router = APIRouter(prefix="/v1/models", tags=["models"])
-
-class VaeRepo(BaseModel):
- repo_id: str = Field(description="The repo ID to use for this VAE")
- path: Optional[str] = Field(description="The path to the VAE")
- subfolder: Optional[str] = Field(description="The subfolder to use for this VAE")
-
-class ModelInfo(BaseModel):
- description: Optional[str] = Field(description="A description of the model")
- model_name: str = Field(description="The name of the model")
- model_type: str = Field(description="The type of the model")
-
-class DiffusersModelInfo(ModelInfo):
- format: Literal['folder'] = 'folder'
-
- vae: Optional[VaeRepo] = Field(description="The VAE repo to use for this model")
- repo_id: Optional[str] = Field(description="The repo ID to use for this model")
- path: Optional[str] = Field(description="The path to the model")
-
-class CkptModelInfo(ModelInfo):
- format: Literal['ckpt'] = 'ckpt'
-
- config: str = Field(description="The path to the model config")
- weights: str = Field(description="The path to the model weights")
- vae: str = Field(description="The path to the model VAE")
- width: Optional[int] = Field(description="The width of the model")
- height: Optional[int] = Field(description="The height of the model")
-
-class SafetensorsModelInfo(CkptModelInfo):
- format: Literal['safetensors'] = 'safetensors'
-
-class CreateModelRequest(BaseModel):
- name: str = Field(description="The name of the model")
- info: Union[CkptModelInfo, DiffusersModelInfo] = Field(discriminator="format", description="The model info")
-
-class CreateModelResponse(BaseModel):
- name: str = Field(description="The name of the new model")
- info: Union[CkptModelInfo, DiffusersModelInfo] = Field(discriminator="format", description="The model info")
- status: str = Field(description="The status of the API response")
-
-class ConversionRequest(BaseModel):
- name: str = Field(description="The name of the new model")
- info: CkptModelInfo = Field(description="The converted model info")
- save_location: str = Field(description="The path to save the converted model weights")
-
-
-class ConvertedModelResponse(BaseModel):
- name: str = Field(description="The name of the new model")
- info: DiffusersModelInfo = Field(description="The converted model info")
+UpdateModelResponse = Union[tuple(OPENAPI_MODEL_CONFIGS)]
+ImportModelResponse = Union[tuple(OPENAPI_MODEL_CONFIGS)]
+ConvertModelResponse = Union[tuple(OPENAPI_MODEL_CONFIGS)]
+MergeModelResponse = Union[tuple(OPENAPI_MODEL_CONFIGS)]
class ModelsList(BaseModel):
- models: list[MODEL_CONFIGS]
-
+ models: list[Union[tuple(OPENAPI_MODEL_CONFIGS)]]
@models_router.get(
"/",
@@ -71,43 +32,103 @@ class ModelsList(BaseModel):
responses={200: {"model": ModelsList }},
)
async def list_models(
- base_model: Optional[BaseModelType] = Query(
- default=None, description="Base model"
- ),
- model_type: Optional[ModelType] = Query(
- default=None, description="The type of model to get"
- ),
+ base_model: Optional[BaseModelType] = Query(default=None, description="Base model"),
+ model_type: Optional[ModelType] = Query(default=None, description="The type of model to get"),
) -> ModelsList:
"""Gets a list of models"""
models_raw = ApiDependencies.invoker.services.model_manager.list_models(base_model, model_type)
models = parse_obj_as(ModelsList, { "models": models_raw })
return models
-
-@models_router.post(
- "/",
+@models_router.patch(
+ "/{base_model}/{model_type}/{model_name}",
operation_id="update_model",
- responses={200: {"status": "success"}},
+ responses={200: {"description" : "The model was updated successfully"},
+ 404: {"description" : "The model could not be found"},
+ 400: {"description" : "Bad request"}
+ },
+ status_code = 200,
+ response_model = UpdateModelResponse,
)
async def update_model(
- model_request: CreateModelRequest
-) -> CreateModelResponse:
+ base_model: BaseModelType = Path(description="Base model"),
+ model_type: ModelType = Path(description="The type of model"),
+ model_name: str = Path(description="model name"),
+ info: Union[tuple(OPENAPI_MODEL_CONFIGS)] = Body(description="Model configuration"),
+) -> UpdateModelResponse:
""" Add Model """
- model_request_info = model_request.info
- info_dict = model_request_info.dict()
- model_response = CreateModelResponse(name=model_request.name, info=model_request.info, status="success")
-
- ApiDependencies.invoker.services.model_manager.add_model(
- model_name=model_request.name,
- model_attributes=info_dict,
- clobber=True,
- )
+ try:
+ ApiDependencies.invoker.services.model_manager.update_model(
+ model_name=model_name,
+ base_model=base_model,
+ model_type=model_type,
+ model_attributes=info.dict()
+ )
+ model_raw = ApiDependencies.invoker.services.model_manager.list_model(
+ model_name=model_name,
+ base_model=base_model,
+ model_type=model_type,
+ )
+ model_response = parse_obj_as(UpdateModelResponse, model_raw)
+ except KeyError as e:
+ raise HTTPException(status_code=404, detail=str(e))
+ except ValueError as e:
+ raise HTTPException(status_code=400, detail=str(e))
return model_response
+@models_router.post(
+ "/",
+ operation_id="import_model",
+ responses= {
+ 201: {"description" : "The model imported successfully"},
+ 404: {"description" : "The model could not be found"},
+ 424: {"description" : "The model appeared to import successfully, but could not be found in the model manager"},
+ 409: {"description" : "There is already a model corresponding to this path or repo_id"},
+ },
+ status_code=201,
+ response_model=ImportModelResponse
+)
+async def import_model(
+ location: str = Body(description="A model path, repo_id or URL to import"),
+ prediction_type: Optional[Literal['v_prediction','epsilon','sample']] = \
+ Body(description='Prediction type for SDv2 checkpoint files', default="v_prediction"),
+) -> ImportModelResponse:
+ """ Add a model using its local path, repo_id, or remote URL """
+
+ items_to_import = {location}
+ prediction_types = { x.value: x for x in SchedulerPredictionType }
+ logger = ApiDependencies.invoker.services.logger
+
+ try:
+ installed_models = ApiDependencies.invoker.services.model_manager.heuristic_import(
+ items_to_import = items_to_import,
+ prediction_type_helper = lambda x: prediction_types.get(prediction_type)
+ )
+ info = installed_models.get(location)
+
+ if not info:
+ logger.error("Import failed")
+ raise HTTPException(status_code=424)
+
+ logger.info(f'Successfully imported {location}, got {info}')
+ model_raw = ApiDependencies.invoker.services.model_manager.list_model(
+ model_name=info.name,
+ base_model=info.base_model,
+ model_type=info.model_type
+ )
+ return parse_obj_as(ImportModelResponse, model_raw)
+
+ except KeyError as e:
+ logger.error(str(e))
+ raise HTTPException(status_code=404, detail=str(e))
+ except ValueError as e:
+ logger.error(str(e))
+ raise HTTPException(status_code=409, detail=str(e))
+
@models_router.delete(
- "/{model_name}",
+ "/{base_model}/{model_type}/{model_name}",
operation_id="del_model",
responses={
204: {
@@ -118,144 +139,95 @@ async def update_model(
}
},
)
-async def delete_model(model_name: str) -> None:
+async def delete_model(
+ base_model: BaseModelType = Path(description="Base model"),
+ model_type: ModelType = Path(description="The type of model"),
+ model_name: str = Path(description="model name"),
+) -> Response:
"""Delete Model"""
- model_names = ApiDependencies.invoker.services.model_manager.model_names()
logger = ApiDependencies.invoker.services.logger
- model_exists = model_name in model_names
-
- # check if model exists
- logger.info(f"Checking for model {model_name}...")
-
- if model_exists:
- logger.info(f"Deleting Model: {model_name}")
- ApiDependencies.invoker.services.model_manager.del_model(model_name, delete_files=True)
- logger.info(f"Model Deleted: {model_name}")
- raise HTTPException(status_code=204, detail=f"Model '{model_name}' deleted successfully")
- else:
- logger.error("Model not found")
+ try:
+ ApiDependencies.invoker.services.model_manager.del_model(model_name,
+ base_model = base_model,
+ model_type = model_type
+ )
+ logger.info(f"Deleted model: {model_name}")
+ return Response(status_code=204)
+ except KeyError:
+ logger.error(f"Model not found: {model_name}")
raise HTTPException(status_code=404, detail=f"Model '{model_name}' not found")
-
- # @socketio.on("convertToDiffusers")
- # def convert_to_diffusers(model_to_convert: dict):
- # try:
- # if model_info := self.generate.model_manager.model_info(
- # model_name=model_to_convert["model_name"]
- # ):
- # if "weights" in model_info:
- # ckpt_path = Path(model_info["weights"])
- # original_config_file = Path(model_info["config"])
- # model_name = model_to_convert["model_name"]
- # model_description = model_info["description"]
- # else:
- # self.socketio.emit(
- # "error", {"message": "Model is not a valid checkpoint file"}
- # )
- # else:
- # self.socketio.emit(
- # "error", {"message": "Could not retrieve model info."}
- # )
-
- # if not ckpt_path.is_absolute():
- # ckpt_path = Path(Globals.root, ckpt_path)
-
- # if original_config_file and not original_config_file.is_absolute():
- # original_config_file = Path(Globals.root, original_config_file)
-
- # diffusers_path = Path(
- # ckpt_path.parent.absolute(), f"{model_name}_diffusers"
- # )
-
- # if model_to_convert["save_location"] == "root":
- # diffusers_path = Path(
- # global_converted_ckpts_dir(), f"{model_name}_diffusers"
- # )
-
- # if (
- # model_to_convert["save_location"] == "custom"
- # and model_to_convert["custom_location"] is not None
- # ):
- # diffusers_path = Path(
- # model_to_convert["custom_location"], f"{model_name}_diffusers"
- # )
-
- # if diffusers_path.exists():
- # shutil.rmtree(diffusers_path)
-
- # self.generate.model_manager.convert_and_import(
- # ckpt_path,
- # diffusers_path,
- # model_name=model_name,
- # model_description=model_description,
- # vae=None,
- # original_config_file=original_config_file,
- # commit_to_conf=opt.conf,
- # )
-
- # new_model_list = self.generate.model_manager.list_models()
- # socketio.emit(
- # "modelConverted",
- # {
- # "new_model_name": model_name,
- # "model_list": new_model_list,
- # "update": True,
- # },
- # )
- # print(f">> Model Converted: {model_name}")
- # except Exception as e:
- # self.handle_exceptions(e)
-
- # @socketio.on("mergeDiffusersModels")
- # def merge_diffusers_models(model_merge_info: dict):
- # try:
- # models_to_merge = model_merge_info["models_to_merge"]
- # model_ids_or_paths = [
- # self.generate.model_manager.model_name_or_path(x)
- # for x in models_to_merge
- # ]
- # merged_pipe = merge_diffusion_models(
- # model_ids_or_paths,
- # model_merge_info["alpha"],
- # model_merge_info["interp"],
- # model_merge_info["force"],
- # )
-
- # dump_path = global_models_dir() / "merged_models"
- # if model_merge_info["model_merge_save_path"] is not None:
- # dump_path = Path(model_merge_info["model_merge_save_path"])
-
- # os.makedirs(dump_path, exist_ok=True)
- # dump_path = dump_path / model_merge_info["merged_model_name"]
- # merged_pipe.save_pretrained(dump_path, safe_serialization=1)
-
- # merged_model_config = dict(
- # model_name=model_merge_info["merged_model_name"],
- # description=f'Merge of models {", ".join(models_to_merge)}',
- # commit_to_conf=opt.conf,
- # )
-
- # if vae := self.generate.model_manager.config[models_to_merge[0]].get(
- # "vae", None
- # ):
- # print(f">> Using configured VAE assigned to {models_to_merge[0]}")
- # merged_model_config.update(vae=vae)
-
- # self.generate.model_manager.import_diffuser_model(
- # dump_path, **merged_model_config
- # )
- # new_model_list = self.generate.model_manager.list_models()
-
- # socketio.emit(
- # "modelsMerged",
- # {
- # "merged_models": models_to_merge,
- # "merged_model_name": model_merge_info["merged_model_name"],
- # "model_list": new_model_list,
- # "update": True,
- # },
- # )
- # print(f">> Models Merged: {models_to_merge}")
- # print(f">> New Model Added: {model_merge_info['merged_model_name']}")
- # except Exception as e:
+@models_router.put(
+ "/convert/{base_model}/{model_type}/{model_name}",
+ operation_id="convert_model",
+ responses={
+ 200: { "description": "Model converted successfully" },
+ 400: {"description" : "Bad request" },
+ 404: { "description": "Model not found" },
+ },
+ status_code = 200,
+ response_model = ConvertModelResponse,
+)
+async def convert_model(
+ base_model: BaseModelType = Path(description="Base model"),
+ model_type: ModelType = Path(description="The type of model"),
+ model_name: str = Path(description="model name"),
+) -> ConvertModelResponse:
+ """Convert a checkpoint model into a diffusers model"""
+ logger = ApiDependencies.invoker.services.logger
+ try:
+ logger.info(f"Converting model: {model_name}")
+ ApiDependencies.invoker.services.model_manager.convert_model(model_name,
+ base_model = base_model,
+ model_type = model_type
+ )
+ model_raw = ApiDependencies.invoker.services.model_manager.list_model(model_name,
+ base_model = base_model,
+ model_type = model_type)
+ response = parse_obj_as(ConvertModelResponse, model_raw)
+ except KeyError:
+ raise HTTPException(status_code=404, detail=f"Model '{model_name}' not found")
+ except ValueError as e:
+ raise HTTPException(status_code=400, detail=str(e))
+ return response
+
+@models_router.put(
+ "/merge/{base_model}",
+ operation_id="merge_models",
+ responses={
+ 200: { "description": "Model converted successfully" },
+ 400: { "description": "Incompatible models" },
+ 404: { "description": "One or more models not found" },
+ },
+ status_code = 200,
+ response_model = MergeModelResponse,
+)
+async def merge_models(
+ base_model: BaseModelType = Path(description="Base model"),
+ model_names: List[str] = Body(description="model name", min_items=2, max_items=3),
+ merged_model_name: Optional[str] = Body(description="Name of destination model"),
+ alpha: Optional[float] = Body(description="Alpha weighting strength to apply to 2d and 3d models", default=0.5),
+ interp: Optional[MergeInterpolationMethod] = Body(description="Interpolation method"),
+ force: Optional[bool] = Body(description="Force merging of models created with different versions of diffusers", default=False),
+) -> MergeModelResponse:
+ """Convert a checkpoint model into a diffusers model"""
+ logger = ApiDependencies.invoker.services.logger
+ try:
+ logger.info(f"Merging models: {model_names}")
+ result = ApiDependencies.invoker.services.model_manager.merge_models(model_names,
+ base_model,
+ merged_model_name or "+".join(model_names),
+ alpha,
+ interp,
+ force)
+ model_raw = ApiDependencies.invoker.services.model_manager.list_model(result.name,
+ base_model = base_model,
+ model_type = ModelType.Main,
+ )
+ response = parse_obj_as(ConvertModelResponse, model_raw)
+ except KeyError:
+ raise HTTPException(status_code=404, detail=f"One or more of the models '{model_names}' not found")
+ except ValueError as e:
+ raise HTTPException(status_code=400, detail=str(e))
+ return response
diff --git a/invokeai/app/api_app.py b/invokeai/app/api_app.py
index e14c58bab7..9bcded1c49 100644
--- a/invokeai/app/api_app.py
+++ b/invokeai/app/api_app.py
@@ -1,5 +1,6 @@
-# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
+# Copyright (c) 2022-2023 Kyle Schouviller (https://github.com/kyle0654) and the InvokeAI Team
import asyncio
+import sys
from inspect import signature
import uvicorn
@@ -20,13 +21,31 @@ from ..backend.util.logging import InvokeAILogger
app_config = InvokeAIAppConfig.get_config()
app_config.parse_args()
logger = InvokeAILogger.getLogger(config=app_config)
+from invokeai.version.invokeai_version import __version__
+
+# we call this early so that the message appears before
+# other invokeai initialization messages
+if app_config.version:
+ print(f'InvokeAI version {__version__}')
+ sys.exit(0)
import invokeai.frontend.web as web_dir
+import mimetypes
from .api.dependencies import ApiDependencies
-from .api.routers import sessions, models, images, boards, board_images
+from .api.routers import sessions, models, images, boards, board_images, app_info
from .api.sockets import SocketIO
from .invocations.baseinvocation import BaseInvocation
+
+
+import torch
+if torch.backends.mps.is_available():
+ import invokeai.backend.util.mps_fixes
+
+# fix for windows mimetypes registry entries being borked
+# see https://github.com/invoke-ai/InvokeAI/discussions/3684#discussioncomment-6391352
+mimetypes.add_type('application/javascript', '.js')
+mimetypes.add_type('text/css', '.css')
# Create the app
# TODO: create this all in a method so configuration/etc. can be passed in?
@@ -82,6 +101,8 @@ app.include_router(boards.boards_router, prefix="/api")
app.include_router(board_images.board_images_router, prefix="/api")
+app.include_router(app_info.app_router, prefix='/api')
+
# Build a custom OpenAPI to include all outputs
# TODO: can outputs be included on metadata of invocation schemas somehow?
def custom_openapi():
diff --git a/invokeai/app/cli/commands.py b/invokeai/app/cli/commands.py
index 10d1ead677..bffb2988dc 100644
--- a/invokeai/app/cli/commands.py
+++ b/invokeai/app/cli/commands.py
@@ -47,7 +47,7 @@ def add_parsers(
commands: list[type],
command_field: str = "type",
exclude_fields: list[str] = ["id", "type"],
- add_arguments: Callable[[argparse.ArgumentParser], None]|None = None
+ add_arguments: Union[Callable[[argparse.ArgumentParser], None],None] = None
):
"""Adds parsers for each command to the subparsers"""
@@ -72,7 +72,7 @@ def add_parsers(
def add_graph_parsers(
subparsers,
graphs: list[LibraryGraph],
- add_arguments: Callable[[argparse.ArgumentParser], None]|None = None
+ add_arguments: Union[Callable[[argparse.ArgumentParser], None], None] = None
):
for graph in graphs:
command_parser = subparsers.add_parser(graph.name, help=graph.description)
diff --git a/invokeai/app/cli_app.py b/invokeai/app/cli_app.py
index 26e058166b..888d36c4bf 100644
--- a/invokeai/app/cli_app.py
+++ b/invokeai/app/cli_app.py
@@ -1,12 +1,11 @@
# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
import argparse
-import os
import re
import shlex
import sys
import time
-from typing import Union, get_type_hints
+from typing import Union, get_type_hints, Optional
from pydantic import BaseModel, ValidationError
from pydantic.fields import Field
@@ -17,10 +16,24 @@ from invokeai.backend.util.logging import InvokeAILogger
config = InvokeAIAppConfig.get_config()
config.parse_args()
logger = InvokeAILogger().getLogger(config=config)
+from invokeai.version.invokeai_version import __version__
+# we call this early so that the message appears before other invokeai initialization messages
+if config.version:
+ print(f'InvokeAI version {__version__}')
+ sys.exit(0)
+
+from invokeai.app.services.board_image_record_storage import (
+ SqliteBoardImageRecordStorage,
+)
+from invokeai.app.services.board_images import (
+ BoardImagesService,
+ BoardImagesServiceDependencies,
+)
+from invokeai.app.services.board_record_storage import SqliteBoardRecordStorage
+from invokeai.app.services.boards import BoardService, BoardServiceDependencies
from invokeai.app.services.image_record_storage import SqliteImageRecordStorage
-from invokeai.app.services.images import ImageService
-from invokeai.app.services.metadata import CoreMetadataService
+from invokeai.app.services.images import ImageService, ImageServiceDependencies
from invokeai.app.services.resource_name import SimpleNameService
from invokeai.app.services.urls import LocalUrlService
from .services.default_graphs import (default_text_to_image_graph_id,
@@ -44,6 +57,10 @@ from .services.processor import DefaultInvocationProcessor
from .services.restoration_services import RestorationServices
from .services.sqlite import SqliteItemStorage
+import torch
+if torch.backends.mps.is_available():
+ import invokeai.backend.util.mps_fixes
+
class CliCommand(BaseModel):
command: Union[BaseCommand.get_commands() + BaseInvocation.get_invocations()] = Field(discriminator="type") # type: ignore
@@ -196,6 +213,7 @@ def invoke_all(context: CliContext):
raise SessionError()
def invoke_cli():
+ logger.info(f'InvokeAI version {__version__}')
# get the optional list of invocations to execute on the command line
parser = config.get_parser()
parser.add_argument('commands',nargs='*')
@@ -225,26 +243,52 @@ def invoke_cli():
)
urls = LocalUrlService()
- metadata = CoreMetadataService()
image_record_storage = SqliteImageRecordStorage(db_location)
image_file_storage = DiskImageFileStorage(f"{output_folder}/images")
names = SimpleNameService()
- images = ImageService(
- image_record_storage=image_record_storage,
- image_file_storage=image_file_storage,
- metadata=metadata,
- url=urls,
- logger=logger,
- names=names,
- graph_execution_manager=graph_execution_manager,
+ board_record_storage = SqliteBoardRecordStorage(db_location)
+ board_image_record_storage = SqliteBoardImageRecordStorage(db_location)
+
+ boards = BoardService(
+ services=BoardServiceDependencies(
+ board_image_record_storage=board_image_record_storage,
+ board_record_storage=board_record_storage,
+ image_record_storage=image_record_storage,
+ url=urls,
+ logger=logger,
+ )
)
+ board_images = BoardImagesService(
+ services=BoardImagesServiceDependencies(
+ board_image_record_storage=board_image_record_storage,
+ board_record_storage=board_record_storage,
+ image_record_storage=image_record_storage,
+ url=urls,
+ logger=logger,
+ )
+ )
+
+ images = ImageService(
+ services=ImageServiceDependencies(
+ board_image_record_storage=board_image_record_storage,
+ image_record_storage=image_record_storage,
+ image_file_storage=image_file_storage,
+ url=urls,
+ logger=logger,
+ names=names,
+ graph_execution_manager=graph_execution_manager,
+ )
+ )
+
services = InvocationServices(
model_manager=model_manager,
events=events,
latents = ForwardCacheLatentsStorage(DiskLatentsStorage(f'{output_folder}/latents')),
images=images,
+ boards=boards,
+ board_images=board_images,
queue=MemoryInvocationQueue(),
graph_library=SqliteItemStorage[LibraryGraph](
filename=db_location, table_name="graphs"
@@ -311,7 +355,7 @@ def invoke_cli():
# Parse invocation
command: CliCommand = None # type:ignore
- system_graph: LibraryGraph|None = None
+ system_graph: Optional[LibraryGraph] = None
if args['type'] in system_graph_names:
system_graph = next(filter(lambda g: g.name == args['type'], system_graphs))
invocation = GraphInvocation(graph=system_graph.graph, id=str(current_id))
diff --git a/invokeai/app/invocations/baseinvocation.py b/invokeai/app/invocations/baseinvocation.py
index 4ce3e839b6..4c7314bd2b 100644
--- a/invokeai/app/invocations/baseinvocation.py
+++ b/invokeai/app/invocations/baseinvocation.py
@@ -4,9 +4,10 @@ from __future__ import annotations
from abc import ABC, abstractmethod
from inspect import signature
-from typing import get_args, get_type_hints, Dict, List, Literal, TypedDict, TYPE_CHECKING
+from typing import (TYPE_CHECKING, Dict, List, Literal, TypedDict, get_args,
+ get_type_hints)
-from pydantic import BaseModel, Field
+from pydantic import BaseConfig, BaseModel, Field
if TYPE_CHECKING:
from ..services.invocation_services import InvocationServices
@@ -65,8 +66,13 @@ class BaseInvocation(ABC, BaseModel):
@classmethod
def get_invocations_map(cls):
# Get the type strings out of the literals and into a dictionary
- return dict(map(lambda t: (get_args(get_type_hints(t)['type'])[0], t),BaseInvocation.get_all_subclasses()))
-
+ return dict(
+ map(
+ lambda t: (get_args(get_type_hints(t)["type"])[0], t),
+ BaseInvocation.get_all_subclasses(),
+ )
+ )
+
@classmethod
def get_output_type(cls):
return signature(cls.invoke).return_annotation
@@ -75,11 +81,11 @@ class BaseInvocation(ABC, BaseModel):
def invoke(self, context: InvocationContext) -> BaseInvocationOutput:
"""Invoke with provided context and return outputs."""
pass
-
- #fmt: off
+
+ # fmt: off
id: str = Field(description="The id of this node. Must be unique among all nodes.")
is_intermediate: bool = Field(default=False, description="Whether or not this node is an intermediate node.")
- #fmt: on
+ # fmt: on
# TODO: figure out a better way to provide these hints
@@ -97,16 +103,20 @@ class UIConfig(TypedDict, total=False):
"latents",
"model",
"control",
+ "image_collection",
+ "vae_model",
+ "lora_model",
],
]
tags: List[str]
title: str
+
class CustomisedSchemaExtra(TypedDict):
ui: UIConfig
-class InvocationConfig(BaseModel.Config):
+class InvocationConfig(BaseConfig):
"""Customizes pydantic's BaseModel.Config class for use by Invocations.
Provide `schema_extra` a `ui` dict to add hints for generated UIs.
diff --git a/invokeai/app/invocations/collections.py b/invokeai/app/invocations/collections.py
index 891f217317..33bde42d69 100644
--- a/invokeai/app/invocations/collections.py
+++ b/invokeai/app/invocations/collections.py
@@ -4,13 +4,16 @@ from typing import Literal
import numpy as np
from pydantic import Field, validator
+from invokeai.app.models.image import ImageField
from invokeai.app.util.misc import SEED_MAX, get_random_seed
from .baseinvocation import (
BaseInvocation,
+ InvocationConfig,
InvocationContext,
BaseInvocationOutput,
+ UIConfig,
)
@@ -22,6 +25,7 @@ class IntCollectionOutput(BaseInvocationOutput):
# Outputs
collection: list[int] = Field(default=[], description="The int collection")
+
class FloatCollectionOutput(BaseInvocationOutput):
"""A collection of floats"""
@@ -31,6 +35,18 @@ class FloatCollectionOutput(BaseInvocationOutput):
collection: list[float] = Field(default=[], description="The float collection")
+class ImageCollectionOutput(BaseInvocationOutput):
+ """A collection of images"""
+
+ type: Literal["image_collection"] = "image_collection"
+
+ # Outputs
+ collection: list[ImageField] = Field(default=[], description="The output images")
+
+ class Config:
+ schema_extra = {"required": ["type", "collection"]}
+
+
class RangeInvocation(BaseInvocation):
"""Creates a range of numbers from start to stop with step"""
@@ -92,3 +108,27 @@ class RandomRangeInvocation(BaseInvocation):
return IntCollectionOutput(
collection=list(rng.integers(low=self.low, high=self.high, size=self.size))
)
+
+
+class ImageCollectionInvocation(BaseInvocation):
+ """Load a collection of images and provide it as output."""
+
+ # fmt: off
+ type: Literal["image_collection"] = "image_collection"
+
+ # Inputs
+ images: list[ImageField] = Field(
+ default=[], description="The image collection to load"
+ )
+ # fmt: on
+ def invoke(self, context: InvocationContext) -> ImageCollectionOutput:
+ return ImageCollectionOutput(collection=self.images)
+
+ class Config(InvocationConfig):
+ schema_extra = {
+ "ui": {
+ "type_hints": {
+ "images": "image_collection",
+ }
+ },
+ }
diff --git a/invokeai/app/invocations/compel.py b/invokeai/app/invocations/compel.py
index 4c1a4448de..dd4d5ce30b 100644
--- a/invokeai/app/invocations/compel.py
+++ b/invokeai/app/invocations/compel.py
@@ -1,6 +1,5 @@
-from typing import Literal, Optional, Union
+from typing import Literal, Optional, Union, List
from pydantic import BaseModel, Field
-from contextlib import ExitStack
import re
from .baseinvocation import BaseInvocation, BaseInvocationOutput, InvocationContext, InvocationConfig
@@ -10,17 +9,25 @@ from ...backend.util.devices import torch_dtype
from ...backend.stable_diffusion.diffusion import InvokeAIDiffuserComponent
from ...backend.model_management import BaseModelType, ModelType, SubModelType, ModelPatcher
+import torch
from compel import Compel
-from compel.prompt_parser import (
- Blend,
- CrossAttentionControlSubstitute,
- FlattenedPrompt,
- Fragment, Conjunction,
-)
+from compel.prompt_parser import (Blend, Conjunction,
+ CrossAttentionControlSubstitute,
+ FlattenedPrompt, Fragment)
+from ...backend.util.devices import torch_dtype
+from ...backend.model_management import ModelType
+from ...backend.model_management.models import ModelNotFoundException
+from ...backend.model_management.lora import ModelPatcher
+from ...backend.stable_diffusion.diffusion import InvokeAIDiffuserComponent
+from .baseinvocation import (BaseInvocation, BaseInvocationOutput,
+ InvocationConfig, InvocationContext)
+from .model import ClipField
class ConditioningField(BaseModel):
- conditioning_name: Optional[str] = Field(default=None, description="The name of conditioning data")
+ conditioning_name: Optional[str] = Field(
+ default=None, description="The name of conditioning data")
+
class Config:
schema_extra = {"required": ["conditioning_name"]}
@@ -50,86 +57,111 @@ class CompelInvocation(BaseInvocation):
"title": "Prompt (Compel)",
"tags": ["prompt", "compel"],
"type_hints": {
- "model": "model"
+ "model": "model"
}
},
}
+ @torch.no_grad()
def invoke(self, context: InvocationContext) -> CompelOutput:
-
tokenizer_info = context.services.model_manager.get_model(
**self.clip.tokenizer.dict(),
)
text_encoder_info = context.services.model_manager.get_model(
**self.clip.text_encoder.dict(),
)
- with tokenizer_info as orig_tokenizer,\
- text_encoder_info as text_encoder,\
- ExitStack() as stack:
- loras = [(stack.enter_context(context.services.model_manager.get_model(**lora.dict(exclude={"weight"}))), lora.weight) for lora in self.clip.loras]
+ def _lora_loader():
+ for lora in self.clip.loras:
+ lora_info = context.services.model_manager.get_model(
+ **lora.dict(exclude={"weight"}))
+ yield (lora_info.context.model, lora.weight)
+ del lora_info
+ return
- ti_list = []
- for trigger in re.findall(r"<[a-zA-Z0-9., _-]+>", self.prompt):
- name = trigger[1:-1]
- try:
- ti_list.append(
- stack.enter_context(
- context.services.model_manager.get_model(
- model_name=name,
- base_model=self.clip.text_encoder.base_model,
- model_type=ModelType.TextualInversion,
- )
- )
- )
- except Exception:
- #print(e)
- #import traceback
- #print(traceback.format_exc())
- print(f"Warn: trigger: \"{trigger}\" not found")
+ #loras = [(context.services.model_manager.get_model(**lora.dict(exclude={"weight"})).context.model, lora.weight) for lora in self.clip.loras]
- with ModelPatcher.apply_lora_text_encoder(text_encoder, loras),\
- ModelPatcher.apply_ti(orig_tokenizer, text_encoder, ti_list) as (tokenizer, ti_manager):
-
- compel = Compel(
- tokenizer=tokenizer,
- text_encoder=text_encoder,
- textual_inversion_manager=ti_manager,
- dtype_for_device_getter=torch_dtype,
- truncate_long_prompts=True, # TODO:
+ ti_list = []
+ for trigger in re.findall(r"<[a-zA-Z0-9., _-]+>", self.prompt):
+ name = trigger[1:-1]
+ try:
+ ti_list.append(
+ context.services.model_manager.get_model(
+ model_name=name,
+ base_model=self.clip.text_encoder.base_model,
+ model_type=ModelType.TextualInversion,
+ ).context.model
)
-
- conjunction = Compel.parse_prompt_string(self.prompt)
- prompt: Union[FlattenedPrompt, Blend] = conjunction.prompts[0]
+ except ModelNotFoundException:
+ # print(e)
+ #import traceback
+ #print(traceback.format_exc())
+ print(f"Warn: trigger: \"{trigger}\" not found")
- if context.services.configuration.log_tokenization:
- log_tokenization_for_prompt_object(prompt, tokenizer)
+ with ModelPatcher.apply_lora_text_encoder(text_encoder_info.context.model, _lora_loader()),\
+ ModelPatcher.apply_ti(tokenizer_info.context.model, text_encoder_info.context.model, ti_list) as (tokenizer, ti_manager),\
+ ModelPatcher.apply_clip_skip(text_encoder_info.context.model, self.clip.skipped_layers),\
+ text_encoder_info as text_encoder:
- c, options = compel.build_conditioning_tensor_for_prompt_object(prompt)
-
- # TODO: long prompt support
- #if not self.truncate_long_prompts:
- # [c, uc] = compel.pad_conditioning_tensors_to_same_length([c, uc])
- ec = InvokeAIDiffuserComponent.ExtraConditioningInfo(
- tokens_count_including_eos_bos=get_max_token_count(tokenizer, conjunction),
- cross_attention_control_args=options.get("cross_attention_control", None),
- )
-
- conditioning_name = f"{context.graph_execution_state_id}_{self.id}_conditioning"
-
- # TODO: hacky but works ;D maybe rename latents somehow?
- context.services.latents.save(conditioning_name, (c, ec))
-
- return CompelOutput(
- conditioning=ConditioningField(
- conditioning_name=conditioning_name,
- ),
+ compel = Compel(
+ tokenizer=tokenizer,
+ text_encoder=text_encoder,
+ textual_inversion_manager=ti_manager,
+ dtype_for_device_getter=torch_dtype,
+ truncate_long_prompts=True, # TODO:
)
+ conjunction = Compel.parse_prompt_string(self.prompt)
+ prompt: Union[FlattenedPrompt, Blend] = conjunction.prompts[0]
+
+ if context.services.configuration.log_tokenization:
+ log_tokenization_for_prompt_object(prompt, tokenizer)
+
+ c, options = compel.build_conditioning_tensor_for_prompt_object(
+ prompt)
+
+ # TODO: long prompt support
+ # if not self.truncate_long_prompts:
+ # [c, uc] = compel.pad_conditioning_tensors_to_same_length([c, uc])
+ ec = InvokeAIDiffuserComponent.ExtraConditioningInfo(
+ tokens_count_including_eos_bos=get_max_token_count(
+ tokenizer, conjunction),
+ cross_attention_control_args=options.get(
+ "cross_attention_control", None),)
+
+ conditioning_name = f"{context.graph_execution_state_id}_{self.id}_conditioning"
+
+ # TODO: hacky but works ;D maybe rename latents somehow?
+ context.services.latents.save(conditioning_name, (c, ec))
+
+ return CompelOutput(
+ conditioning=ConditioningField(
+ conditioning_name=conditioning_name,
+ ),
+ )
+
+class ClipSkipInvocationOutput(BaseInvocationOutput):
+ """Clip skip node output"""
+ type: Literal["clip_skip_output"] = "clip_skip_output"
+ clip: ClipField = Field(None, description="Clip with skipped layers")
+
+class ClipSkipInvocation(BaseInvocation):
+ """Skip layers in clip text_encoder model."""
+ type: Literal["clip_skip"] = "clip_skip"
+
+ clip: ClipField = Field(None, description="Clip to use")
+ skipped_layers: int = Field(0, description="Number of layers to skip in text_encoder")
+
+ def invoke(self, context: InvocationContext) -> ClipSkipInvocationOutput:
+ self.clip.skipped_layers += self.skipped_layers
+ return ClipSkipInvocationOutput(
+ clip=self.clip,
+ )
+
def get_max_token_count(
- tokenizer, prompt: Union[FlattenedPrompt, Blend, Conjunction], truncate_if_too_long=False
-) -> int:
+ tokenizer, prompt: Union[FlattenedPrompt, Blend, Conjunction],
+ truncate_if_too_long=False) -> int:
if type(prompt) is Blend:
blend: Blend = prompt
return max(
@@ -148,13 +180,13 @@ def get_max_token_count(
)
else:
return len(
- get_tokens_for_prompt_object(tokenizer, prompt, truncate_if_too_long)
- )
+ get_tokens_for_prompt_object(
+ tokenizer, prompt, truncate_if_too_long))
def get_tokens_for_prompt_object(
tokenizer, parsed_prompt: FlattenedPrompt, truncate_if_too_long=True
-) -> [str]:
+) -> List[str]:
if type(parsed_prompt) is Blend:
raise ValueError(
"Blend is not supported here - you need to get tokens for each of its .children"
@@ -183,7 +215,7 @@ def log_tokenization_for_conjunction(
):
display_label_prefix = display_label_prefix or ""
for i, p in enumerate(c.prompts):
- if len(c.prompts)>1:
+ if len(c.prompts) > 1:
this_display_label_prefix = f"{display_label_prefix}(conjunction part {i + 1}, weight={c.weights[i]})"
else:
this_display_label_prefix = display_label_prefix
@@ -238,7 +270,8 @@ def log_tokenization_for_prompt_object(
)
-def log_tokenization_for_text(text, tokenizer, display_label=None, truncate_if_too_long=False):
+def log_tokenization_for_text(
+ text, tokenizer, display_label=None, truncate_if_too_long=False):
"""shows how the prompt is tokenized
# usually tokens have '' to indicate end-of-word,
# but for readability it has been replaced with ' '
diff --git a/invokeai/app/invocations/controlnet_image_processors.py b/invokeai/app/invocations/controlnet_image_processors.py
index f40954ebb6..c37dcda998 100644
--- a/invokeai/app/invocations/controlnet_image_processors.py
+++ b/invokeai/app/invocations/controlnet_image_processors.py
@@ -1,11 +1,12 @@
-# InvokeAI nodes for ControlNet image preprocessors
+# Invocations for ControlNet image preprocessors
# initial implementation by Gregg Helt, 2023
# heavily leverages controlnet_aux package: https://github.com/patrickvonplaten/controlnet_aux
-from builtins import float
+from builtins import float, bool
+import cv2
import numpy as np
-from typing import Literal, Optional, Union, List
-from PIL import Image, ImageFilter, ImageOps
+from typing import Literal, Optional, Union, List, Dict
+from PIL import Image
from pydantic import BaseModel, Field, validator
from ..models.image import ImageField, ImageCategory, ResourceOrigin
@@ -29,8 +30,13 @@ from controlnet_aux import (
ContentShuffleDetector,
ZoeDetector,
MediapipeFaceDetector,
+ SamDetector,
+ LeresDetector,
)
+from controlnet_aux.util import HWC3, ade_palette
+
+
from .image import ImageOutput, PILInvocationConfig
CONTROLNET_DEFAULT_MODELS = [
@@ -94,6 +100,10 @@ CONTROLNET_DEFAULT_MODELS = [
]
CONTROLNET_NAME_VALUES = Literal[tuple(CONTROLNET_DEFAULT_MODELS)]
+CONTROLNET_MODE_VALUES = Literal[tuple(["balanced", "more_prompt", "more_control", "unbalanced"])]
+# crop and fill options not ready yet
+# CONTROLNET_RESIZE_VALUES = Literal[tuple(["just_resize", "crop_resize", "fill_resize"])]
+
class ControlField(BaseModel):
image: ImageField = Field(default=None, description="The control image")
@@ -104,6 +114,9 @@ class ControlField(BaseModel):
description="When the ControlNet is first applied (% of total steps)")
end_step_percent: float = Field(default=1, ge=0, le=1,
description="When the ControlNet is last applied (% of total steps)")
+ control_mode: CONTROLNET_MODE_VALUES = Field(default="balanced", description="The control mode to use")
+ # resize_mode: CONTROLNET_RESIZE_VALUES = Field(default="just_resize", description="The resize mode to use")
+
@validator("control_weight")
def abs_le_one(cls, v):
"""validate that all abs(values) are <=1"""
@@ -144,11 +157,11 @@ class ControlNetInvocation(BaseInvocation):
control_model: CONTROLNET_NAME_VALUES = Field(default="lllyasviel/sd-controlnet-canny",
description="control model used")
control_weight: Union[float, List[float]] = Field(default=1.0, description="The weight given to the ControlNet")
- # TODO: add support in backend core for begin_step_percent, end_step_percent, guess_mode
begin_step_percent: float = Field(default=0, ge=0, le=1,
description="When the ControlNet is first applied (% of total steps)")
end_step_percent: float = Field(default=1, ge=0, le=1,
description="When the ControlNet is last applied (% of total steps)")
+ control_mode: CONTROLNET_MODE_VALUES = Field(default="balanced", description="The control mode used")
# fmt: on
class Config(InvocationConfig):
@@ -166,7 +179,6 @@ class ControlNetInvocation(BaseInvocation):
}
def invoke(self, context: InvocationContext) -> ControlOutput:
-
return ControlOutput(
control=ControlField(
image=self.image,
@@ -174,10 +186,11 @@ class ControlNetInvocation(BaseInvocation):
control_weight=self.control_weight,
begin_step_percent=self.begin_step_percent,
end_step_percent=self.end_step_percent,
+ control_mode=self.control_mode,
),
)
-# TODO: move image processors to separate file (image_analysis.py
+
class ImageProcessorInvocation(BaseInvocation, PILInvocationConfig):
"""Base class for invocations that preprocess images for ControlNet"""
@@ -409,9 +422,9 @@ class ContentShuffleImageProcessorInvocation(ImageProcessorInvocation, PILInvoca
# Inputs
detect_resolution: int = Field(default=512, ge=0, description="The pixel resolution for detection")
image_resolution: int = Field(default=512, ge=0, description="The pixel resolution for the output image")
- h: Union[int, None] = Field(default=512, ge=0, description="Content shuffle `h` parameter")
- w: Union[int, None] = Field(default=512, ge=0, description="Content shuffle `w` parameter")
- f: Union[int, None] = Field(default=256, ge=0, description="Content shuffle `f` parameter")
+ h: Optional[int] = Field(default=512, ge=0, description="Content shuffle `h` parameter")
+ w: Optional[int] = Field(default=512, ge=0, description="Content shuffle `w` parameter")
+ f: Optional[int] = Field(default=256, ge=0, description="Content shuffle `f` parameter")
# fmt: on
def run_processor(self, image):
@@ -449,6 +462,104 @@ class MediapipeFaceProcessorInvocation(ImageProcessorInvocation, PILInvocationCo
# fmt: on
def run_processor(self, image):
+ # MediaPipeFaceDetector throws an error if image has alpha channel
+ # so convert to RGB if needed
+ if image.mode == 'RGBA':
+ image = image.convert('RGB')
mediapipe_face_processor = MediapipeFaceDetector()
processed_image = mediapipe_face_processor(image, max_faces=self.max_faces, min_confidence=self.min_confidence)
return processed_image
+
+class LeresImageProcessorInvocation(ImageProcessorInvocation, PILInvocationConfig):
+ """Applies leres processing to image"""
+ # fmt: off
+ type: Literal["leres_image_processor"] = "leres_image_processor"
+ # Inputs
+ thr_a: float = Field(default=0, description="Leres parameter `thr_a`")
+ thr_b: float = Field(default=0, description="Leres parameter `thr_b`")
+ boost: bool = Field(default=False, description="Whether to use boost mode")
+ detect_resolution: int = Field(default=512, ge=0, description="The pixel resolution for detection")
+ image_resolution: int = Field(default=512, ge=0, description="The pixel resolution for the output image")
+ # fmt: on
+
+ def run_processor(self, image):
+ leres_processor = LeresDetector.from_pretrained("lllyasviel/Annotators")
+ processed_image = leres_processor(image,
+ thr_a=self.thr_a,
+ thr_b=self.thr_b,
+ boost=self.boost,
+ detect_resolution=self.detect_resolution,
+ image_resolution=self.image_resolution)
+ return processed_image
+
+
+class TileResamplerProcessorInvocation(ImageProcessorInvocation, PILInvocationConfig):
+
+ # fmt: off
+ type: Literal["tile_image_processor"] = "tile_image_processor"
+ # Inputs
+ #res: int = Field(default=512, ge=0, le=1024, description="The pixel resolution for each tile")
+ down_sampling_rate: float = Field(default=1.0, ge=1.0, le=8.0, description="Down sampling rate")
+ # fmt: on
+
+ # tile_resample copied from sd-webui-controlnet/scripts/processor.py
+ def tile_resample(self,
+ np_img: np.ndarray,
+ res=512, # never used?
+ down_sampling_rate=1.0,
+ ):
+ np_img = HWC3(np_img)
+ if down_sampling_rate < 1.1:
+ return np_img
+ H, W, C = np_img.shape
+ H = int(float(H) / float(down_sampling_rate))
+ W = int(float(W) / float(down_sampling_rate))
+ np_img = cv2.resize(np_img, (W, H), interpolation=cv2.INTER_AREA)
+ return np_img
+
+ def run_processor(self, img):
+ np_img = np.array(img, dtype=np.uint8)
+ processed_np_image = self.tile_resample(np_img,
+ #res=self.tile_size,
+ down_sampling_rate=self.down_sampling_rate
+ )
+ processed_image = Image.fromarray(processed_np_image)
+ return processed_image
+
+
+
+
+class SegmentAnythingProcessorInvocation(ImageProcessorInvocation, PILInvocationConfig):
+ """Applies segment anything processing to image"""
+ # fmt: off
+ type: Literal["segment_anything_processor"] = "segment_anything_processor"
+ # fmt: on
+
+ def run_processor(self, image):
+ # segment_anything_processor = SamDetector.from_pretrained("ybelkada/segment-anything", subfolder="checkpoints")
+ segment_anything_processor = SamDetectorReproducibleColors.from_pretrained("ybelkada/segment-anything", subfolder="checkpoints")
+ np_img = np.array(image, dtype=np.uint8)
+ processed_image = segment_anything_processor(np_img)
+ return processed_image
+
+class SamDetectorReproducibleColors(SamDetector):
+
+ # overriding SamDetector.show_anns() method to use reproducible colors for segmentation image
+ # base class show_anns() method randomizes colors,
+ # which seems to also lead to non-reproducible image generation
+ # so using ADE20k color palette instead
+ def show_anns(self, anns: List[Dict]):
+ if len(anns) == 0:
+ return
+ sorted_anns = sorted(anns, key=(lambda x: x['area']), reverse=True)
+ h, w = anns[0]['segmentation'].shape
+ final_img = Image.fromarray(np.zeros((h, w, 3), dtype=np.uint8), mode="RGB")
+ palette = ade_palette()
+ for i, ann in enumerate(sorted_anns):
+ m = ann['segmentation']
+ img = np.empty((m.shape[0], m.shape[1], 3), dtype=np.uint8)
+ # doing modulo just in case number of annotated regions exceeds number of colors in palette
+ ann_color = palette[i % len(palette)]
+ img[:, :] = ann_color
+ final_img.paste(Image.fromarray(img, mode="RGB"), (0, 0), Image.fromarray(np.uint8(m * 255)))
+ return np.array(final_img, dtype=np.uint8)
diff --git a/invokeai/app/invocations/generate.py b/invokeai/app/invocations/generate.py
index 7135e66a02..6cdb83effc 100644
--- a/invokeai/app/invocations/generate.py
+++ b/invokeai/app/invocations/generate.py
@@ -1,11 +1,10 @@
# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
from functools import partial
-from typing import Literal, Optional, Union, get_args
+from typing import Literal, Optional, get_args
import torch
-from diffusers import ControlNetModel
-from pydantic import BaseModel, Field
+from pydantic import Field
from invokeai.app.models.image import (ColorField, ImageCategory, ImageField,
ResourceOrigin)
@@ -18,7 +17,6 @@ from ..util.step_callback import stable_diffusion_step_callback
from .baseinvocation import BaseInvocation, InvocationConfig, InvocationContext
from .image import ImageOutput
-import re
from ...backend.model_management.lora import ModelPatcher
from ...backend.stable_diffusion.diffusers_pipeline import StableDiffusionGeneratorPipeline
from .model import UNetField, VaeField
@@ -76,7 +74,7 @@ class InpaintInvocation(BaseInvocation):
vae: VaeField = Field(default=None, description="Vae model")
# Inputs
- image: Union[ImageField, None] = Field(description="The input image")
+ image: Optional[ImageField] = Field(description="The input image")
strength: float = Field(
default=0.75, gt=0, le=1, description="The strength of the original image"
)
@@ -86,7 +84,7 @@ class InpaintInvocation(BaseInvocation):
)
# Inputs
- mask: Union[ImageField, None] = Field(description="The mask")
+ mask: Optional[ImageField] = Field(description="The mask")
seam_size: int = Field(default=96, ge=1, description="The seam inpaint size (px)")
seam_blur: int = Field(
default=16, ge=0, description="The seam inpaint blur radius (px)"
@@ -156,40 +154,42 @@ class InpaintInvocation(BaseInvocation):
@contextmanager
def load_model_old_way(self, context, scheduler):
+ def _lora_loader():
+ for lora in self.unet.loras:
+ lora_info = context.services.model_manager.get_model(
+ **lora.dict(exclude={"weight"}))
+ yield (lora_info.context.model, lora.weight)
+ del lora_info
+ return
+
unet_info = context.services.model_manager.get_model(**self.unet.unet.dict())
vae_info = context.services.model_manager.get_model(**self.vae.vae.dict())
- #unet = unet_info.context.model
- #vae = vae_info.context.model
+ with vae_info as vae,\
+ ModelPatcher.apply_lora_unet(unet_info.context.model, _lora_loader()),\
+ unet_info as unet:
- with ExitStack() as stack:
- loras = [(stack.enter_context(context.services.model_manager.get_model(**lora.dict(exclude={"weight"}))), lora.weight) for lora in self.unet.loras]
+ device = context.services.model_manager.mgr.cache.execution_device
+ dtype = context.services.model_manager.mgr.cache.precision
- with vae_info as vae,\
- unet_info as unet,\
- ModelPatcher.apply_lora_unet(unet, loras):
+ pipeline = StableDiffusionGeneratorPipeline(
+ vae=vae,
+ text_encoder=None,
+ tokenizer=None,
+ unet=unet,
+ scheduler=scheduler,
+ safety_checker=None,
+ feature_extractor=None,
+ requires_safety_checker=False,
+ precision="float16" if dtype == torch.float16 else "float32",
+ execution_device=device,
+ )
- device = context.services.model_manager.mgr.cache.execution_device
- dtype = context.services.model_manager.mgr.cache.precision
-
- pipeline = StableDiffusionGeneratorPipeline(
- vae=vae,
- text_encoder=None,
- tokenizer=None,
- unet=unet,
- scheduler=scheduler,
- safety_checker=None,
- feature_extractor=None,
- requires_safety_checker=False,
- precision="float16" if dtype == torch.float16 else "float32",
- execution_device=device,
- )
-
- yield OldModelInfo(
- name=self.unet.unet.model_name,
- hash="",
- model=pipeline,
- )
+ yield OldModelInfo(
+ name=self.unet.unet.model_name,
+ hash="",
+ model=pipeline,
+ )
def invoke(self, context: InvocationContext) -> ImageOutput:
image = (
@@ -228,21 +228,21 @@ class InpaintInvocation(BaseInvocation):
), # Shorthand for passing all of the parameters above manually
)
- # Outputs is an infinite iterator that will return a new InvokeAIGeneratorOutput object
- # each time it is called. We only need the first one.
- generator_output = next(outputs)
+ # Outputs is an infinite iterator that will return a new InvokeAIGeneratorOutput object
+ # each time it is called. We only need the first one.
+ generator_output = next(outputs)
- image_dto = context.services.images.create(
- image=generator_output.image,
- image_origin=ResourceOrigin.INTERNAL,
- image_category=ImageCategory.GENERAL,
- session_id=context.graph_execution_state_id,
- node_id=self.id,
- is_intermediate=self.is_intermediate,
- )
+ image_dto = context.services.images.create(
+ image=generator_output.image,
+ image_origin=ResourceOrigin.INTERNAL,
+ image_category=ImageCategory.GENERAL,
+ session_id=context.graph_execution_state_id,
+ node_id=self.id,
+ is_intermediate=self.is_intermediate,
+ )
- return ImageOutput(
- image=ImageField(image_name=image_dto.image_name),
- width=image_dto.width,
- height=image_dto.height,
- )
+ return ImageOutput(
+ image=ImageField(image_name=image_dto.image_name),
+ width=image_dto.width,
+ height=image_dto.height,
+ )
diff --git a/invokeai/app/invocations/image.py b/invokeai/app/invocations/image.py
index f85669eab1..1dbbff1c84 100644
--- a/invokeai/app/invocations/image.py
+++ b/invokeai/app/invocations/image.py
@@ -1,7 +1,6 @@
# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
-import io
-from typing import Literal, Optional, Union
+from typing import Literal, Optional
import numpy
from PIL import Image, ImageFilter, ImageOps, ImageChops
@@ -67,7 +66,7 @@ class LoadImageInvocation(BaseInvocation):
type: Literal["load_image"] = "load_image"
# Inputs
- image: Union[ImageField, None] = Field(
+ image: Optional[ImageField] = Field(
default=None, description="The image to load"
)
# fmt: on
@@ -87,7 +86,7 @@ class ShowImageInvocation(BaseInvocation):
type: Literal["show_image"] = "show_image"
# Inputs
- image: Union[ImageField, None] = Field(
+ image: Optional[ImageField] = Field(
default=None, description="The image to show"
)
@@ -112,7 +111,7 @@ class ImageCropInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_crop"] = "img_crop"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to crop")
+ image: Optional[ImageField] = Field(default=None, description="The image to crop")
x: int = Field(default=0, description="The left x coordinate of the crop rectangle")
y: int = Field(default=0, description="The top y coordinate of the crop rectangle")
width: int = Field(default=512, gt=0, description="The width of the crop rectangle")
@@ -150,8 +149,8 @@ class ImagePasteInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_paste"] = "img_paste"
# Inputs
- base_image: Union[ImageField, None] = Field(default=None, description="The base image")
- image: Union[ImageField, None] = Field(default=None, description="The image to paste")
+ base_image: Optional[ImageField] = Field(default=None, description="The base image")
+ image: Optional[ImageField] = Field(default=None, description="The image to paste")
mask: Optional[ImageField] = Field(default=None, description="The mask to use when pasting")
x: int = Field(default=0, description="The left x coordinate at which to paste the image")
y: int = Field(default=0, description="The top y coordinate at which to paste the image")
@@ -203,7 +202,7 @@ class MaskFromAlphaInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["tomask"] = "tomask"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to create the mask from")
+ image: Optional[ImageField] = Field(default=None, description="The image to create the mask from")
invert: bool = Field(default=False, description="Whether or not to invert the mask")
# fmt: on
@@ -237,8 +236,8 @@ class ImageMultiplyInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_mul"] = "img_mul"
# Inputs
- image1: Union[ImageField, None] = Field(default=None, description="The first image to multiply")
- image2: Union[ImageField, None] = Field(default=None, description="The second image to multiply")
+ image1: Optional[ImageField] = Field(default=None, description="The first image to multiply")
+ image2: Optional[ImageField] = Field(default=None, description="The second image to multiply")
# fmt: on
def invoke(self, context: InvocationContext) -> ImageOutput:
@@ -273,7 +272,7 @@ class ImageChannelInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_chan"] = "img_chan"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to get the channel from")
+ image: Optional[ImageField] = Field(default=None, description="The image to get the channel from")
channel: IMAGE_CHANNELS = Field(default="A", description="The channel to get")
# fmt: on
@@ -308,7 +307,7 @@ class ImageConvertInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_conv"] = "img_conv"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to convert")
+ image: Optional[ImageField] = Field(default=None, description="The image to convert")
mode: IMAGE_MODES = Field(default="L", description="The mode to convert to")
# fmt: on
@@ -340,7 +339,7 @@ class ImageBlurInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_blur"] = "img_blur"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to blur")
+ image: Optional[ImageField] = Field(default=None, description="The image to blur")
radius: float = Field(default=8.0, ge=0, description="The blur radius")
blur_type: Literal["gaussian", "box"] = Field(default="gaussian", description="The type of blur")
# fmt: on
@@ -398,7 +397,7 @@ class ImageResizeInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_resize"] = "img_resize"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to resize")
+ image: Optional[ImageField] = Field(default=None, description="The image to resize")
width: int = Field(ge=64, multiple_of=8, description="The width to resize to (px)")
height: int = Field(ge=64, multiple_of=8, description="The height to resize to (px)")
resample_mode: PIL_RESAMPLING_MODES = Field(default="bicubic", description="The resampling mode")
@@ -437,7 +436,7 @@ class ImageScaleInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_scale"] = "img_scale"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to scale")
+ image: Optional[ImageField] = Field(default=None, description="The image to scale")
scale_factor: float = Field(gt=0, description="The factor by which to scale the image")
resample_mode: PIL_RESAMPLING_MODES = Field(default="bicubic", description="The resampling mode")
# fmt: on
@@ -477,7 +476,7 @@ class ImageLerpInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_lerp"] = "img_lerp"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to lerp")
+ image: Optional[ImageField] = Field(default=None, description="The image to lerp")
min: int = Field(default=0, ge=0, le=255, description="The minimum output value")
max: int = Field(default=255, ge=0, le=255, description="The maximum output value")
# fmt: on
@@ -513,7 +512,7 @@ class ImageInverseLerpInvocation(BaseInvocation, PILInvocationConfig):
type: Literal["img_ilerp"] = "img_ilerp"
# Inputs
- image: Union[ImageField, None] = Field(default=None, description="The image to lerp")
+ image: Optional[ImageField] = Field(default=None, description="The image to lerp")
min: int = Field(default=0, ge=0, le=255, description="The minimum input value")
max: int = Field(default=255, ge=0, le=255, description="The maximum input value")
# fmt: on
diff --git a/invokeai/app/invocations/infill.py b/invokeai/app/invocations/infill.py
index ad67594c29..2640b2531d 100644
--- a/invokeai/app/invocations/infill.py
+++ b/invokeai/app/invocations/infill.py
@@ -1,6 +1,6 @@
# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654) and the InvokeAI Team
-from typing import Literal, Union, get_args
+from typing import Literal, Optional, get_args
import numpy as np
import math
@@ -68,7 +68,7 @@ def get_tile_images(image: np.ndarray, width=8, height=8):
def tile_fill_missing(
- im: Image.Image, tile_size: int = 16, seed: Union[int, None] = None
+ im: Image.Image, tile_size: int = 16, seed: Optional[int] = None
) -> Image.Image:
# Only fill if there's an alpha layer
if im.mode != "RGBA":
@@ -125,7 +125,7 @@ class InfillColorInvocation(BaseInvocation):
"""Infills transparent areas of an image with a solid color"""
type: Literal["infill_rgba"] = "infill_rgba"
- image: Union[ImageField, None] = Field(
+ image: Optional[ImageField] = Field(
default=None, description="The image to infill"
)
color: ColorField = Field(
@@ -162,7 +162,7 @@ class InfillTileInvocation(BaseInvocation):
type: Literal["infill_tile"] = "infill_tile"
- image: Union[ImageField, None] = Field(
+ image: Optional[ImageField] = Field(
default=None, description="The image to infill"
)
tile_size: int = Field(default=32, ge=1, description="The tile size (px)")
@@ -202,7 +202,7 @@ class InfillPatchMatchInvocation(BaseInvocation):
type: Literal["infill_patchmatch"] = "infill_patchmatch"
- image: Union[ImageField, None] = Field(
+ image: Optional[ImageField] = Field(
default=None, description="The image to infill"
)
diff --git a/invokeai/app/invocations/latent.py b/invokeai/app/invocations/latent.py
index 942dae77f0..118f2e5e9d 100644
--- a/invokeai/app/invocations/latent.py
+++ b/invokeai/app/invocations/latent.py
@@ -1,21 +1,18 @@
# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654)
-from contextlib import ExitStack
from typing import List, Literal, Optional, Union
import einops
-
-from pydantic import BaseModel, Field, validator
import torch
-from diffusers import ControlNetModel, DPMSolverMultistepScheduler
+from diffusers import ControlNetModel
from diffusers.image_processor import VaeImageProcessor
from diffusers.schedulers import SchedulerMixin as Scheduler
+from pydantic import BaseModel, Field, validator
-from invokeai.app.util.misc import SEED_MAX, get_random_seed
+from invokeai.app.invocations.metadata import CoreMetadata
from invokeai.app.util.step_callback import stable_diffusion_step_callback
-from ..models.image import ImageCategory, ImageField, ResourceOrigin
-from ...backend.image_util.seamless import configure_model_padding
+from ...backend.model_management.lora import ModelPatcher
from ...backend.stable_diffusion import PipelineIntermediateState
from ...backend.stable_diffusion.diffusers_pipeline import (
ConditioningData, ControlNetData, StableDiffusionGeneratorPipeline,
@@ -25,6 +22,7 @@ from ...backend.stable_diffusion.diffusion.shared_invokeai_diffusion import \
from ...backend.stable_diffusion.schedulers import SCHEDULER_MAP
from ...backend.util.devices import choose_torch_device, torch_dtype
from ...backend.model_management import ModelPatcher
+from ..models.image import ImageCategory, ImageField, ResourceOrigin
from .baseinvocation import (BaseInvocation, BaseInvocationOutput,
InvocationConfig, InvocationContext)
from .compel import ConditioningField
@@ -32,14 +30,17 @@ from .controlnet_image_processors import ControlField
from .image import ImageOutput
from .model import ModelInfo, UNetField, VaeField
+
class LatentsField(BaseModel):
"""A latents field used for passing latents between invocations"""
- latents_name: Optional[str] = Field(default=None, description="The name of the latents")
+ latents_name: Optional[str] = Field(
+ default=None, description="The name of the latents")
class Config:
schema_extra = {"required": ["latents_name"]}
+
class LatentsOutput(BaseInvocationOutput):
"""Base class for invocations that output latents"""
#fmt: off
@@ -53,29 +54,11 @@ class LatentsOutput(BaseInvocationOutput):
def build_latents_output(latents_name: str, latents: torch.Tensor):
- return LatentsOutput(
- latents=LatentsField(latents_name=latents_name),
- width=latents.shape[3] * 8,
- height=latents.shape[2] * 8,
- )
-
-class NoiseOutput(BaseInvocationOutput):
- """Invocation noise output"""
- #fmt: off
- type: Literal["noise_output"] = "noise_output"
-
- # Inputs
- noise: LatentsField = Field(default=None, description="The output noise")
- width: int = Field(description="The width of the noise in pixels")
- height: int = Field(description="The height of the noise in pixels")
- #fmt: on
-
-def build_noise_output(latents_name: str, latents: torch.Tensor):
- return NoiseOutput(
- noise=LatentsField(latents_name=latents_name),
- width=latents.size()[3] * 8,
- height=latents.size()[2] * 8,
- )
+ return LatentsOutput(
+ latents=LatentsField(latents_name=latents_name),
+ width=latents.size()[3] * 8,
+ height=latents.size()[2] * 8,
+ )
SAMPLER_NAME_VALUES = Literal[
@@ -83,84 +66,30 @@ SAMPLER_NAME_VALUES = Literal[
]
-
def get_scheduler(
context: InvocationContext,
scheduler_info: ModelInfo,
scheduler_name: str,
) -> Scheduler:
- scheduler_class, scheduler_extra_config = SCHEDULER_MAP.get(scheduler_name, SCHEDULER_MAP['ddim'])
- orig_scheduler_info = context.services.model_manager.get_model(**scheduler_info.dict())
+ scheduler_class, scheduler_extra_config = SCHEDULER_MAP.get(
+ scheduler_name, SCHEDULER_MAP['ddim'])
+ orig_scheduler_info = context.services.model_manager.get_model(
+ **scheduler_info.dict())
with orig_scheduler_info as orig_scheduler:
scheduler_config = orig_scheduler.config
-
+
if "_backup" in scheduler_config:
scheduler_config = scheduler_config["_backup"]
- scheduler_config = {**scheduler_config, **scheduler_extra_config, "_backup": scheduler_config}
+ scheduler_config = {**scheduler_config, **
+ scheduler_extra_config, "_backup": scheduler_config}
scheduler = scheduler_class.from_config(scheduler_config)
-
+
# hack copied over from generate.py
if not hasattr(scheduler, 'uses_inpainting_model'):
scheduler.uses_inpainting_model = lambda: False
return scheduler
-def get_noise(width:int, height:int, device:torch.device, seed:int = 0, latent_channels:int=4, use_mps_noise:bool=False, downsampling_factor:int = 8):
- # limit noise to only the diffusion image channels, not the mask channels
- input_channels = min(latent_channels, 4)
- use_device = "cpu" if (use_mps_noise or device.type == "mps") else device
- generator = torch.Generator(device=use_device).manual_seed(seed)
- x = torch.randn(
- [
- 1,
- input_channels,
- height // downsampling_factor,
- width // downsampling_factor,
- ],
- dtype=torch_dtype(device),
- device=use_device,
- generator=generator,
- ).to(device)
- # if self.perlin > 0.0:
- # perlin_noise = self.get_perlin_noise(
- # width // self.downsampling_factor, height // self.downsampling_factor
- # )
- # x = (1 - self.perlin) * x + self.perlin * perlin_noise
- return x
-
-class NoiseInvocation(BaseInvocation):
- """Generates latent noise."""
-
- type: Literal["noise"] = "noise"
-
- # Inputs
- seed: int = Field(ge=0, le=SEED_MAX, description="The seed to use", default_factory=get_random_seed)
- width: int = Field(default=512, multiple_of=8, gt=0, description="The width of the resulting noise", )
- height: int = Field(default=512, multiple_of=8, gt=0, description="The height of the resulting noise", )
-
-
- # Schema customisation
- class Config(InvocationConfig):
- schema_extra = {
- "ui": {
- "tags": ["latents", "noise"],
- },
- }
-
- @validator("seed", pre=True)
- def modulo_seed(cls, v):
- """Returns the seed modulo SEED_MAX to ensure it is within the valid range."""
- return v % SEED_MAX
-
- def invoke(self, context: InvocationContext) -> NoiseOutput:
- device = torch.device(choose_torch_device())
- noise = get_noise(self.width, self.height, device, self.seed)
-
- name = f'{context.graph_execution_state_id}__{self.id}'
- context.services.latents.save(name, noise)
- return build_noise_output(latents_name=name, latents=noise)
-
-
# Text to image
class TextToLatentsInvocation(BaseInvocation):
"""Generates latents from conditionings."""
@@ -199,18 +128,18 @@ class TextToLatentsInvocation(BaseInvocation):
"ui": {
"tags": ["latents"],
"type_hints": {
- "model": "model",
- "control": "control",
- # "cfg_scale": "float",
- "cfg_scale": "number"
+ "model": "model",
+ "control": "control",
+ # "cfg_scale": "float",
+ "cfg_scale": "number"
}
},
}
# TODO: pass this an emitter method or something? or a session for dispatching?
def dispatch_progress(
- self, context: InvocationContext, source_node_id: str, intermediate_state: PipelineIntermediateState
- ) -> None:
+ self, context: InvocationContext, source_node_id: str,
+ intermediate_state: PipelineIntermediateState) -> None:
stable_diffusion_step_callback(
context=context,
intermediate_state=intermediate_state,
@@ -218,9 +147,12 @@ class TextToLatentsInvocation(BaseInvocation):
source_node_id=source_node_id,
)
- def get_conditioning_data(self, context: InvocationContext, scheduler) -> ConditioningData:
- c, extra_conditioning_info = context.services.latents.get(self.positive_conditioning.conditioning_name)
- uc, _ = context.services.latents.get(self.negative_conditioning.conditioning_name)
+ def get_conditioning_data(
+ self, context: InvocationContext, scheduler) -> ConditioningData:
+ c, extra_conditioning_info = context.services.latents.get(
+ self.positive_conditioning.conditioning_name)
+ uc, _ = context.services.latents.get(
+ self.negative_conditioning.conditioning_name)
conditioning_data = ConditioningData(
unconditioned_embeddings=uc,
@@ -228,10 +160,10 @@ class TextToLatentsInvocation(BaseInvocation):
guidance_scale=self.cfg_scale,
extra=extra_conditioning_info,
postprocessing_settings=PostprocessingSettings(
- threshold=0.0,#threshold,
- warmup=0.2,#warmup,
- h_symmetry_time_pct=None,#h_symmetry_time_pct,
- v_symmetry_time_pct=None#v_symmetry_time_pct,
+ threshold=0.0, # threshold,
+ warmup=0.2, # warmup,
+ h_symmetry_time_pct=None, # h_symmetry_time_pct,
+ v_symmetry_time_pct=None # v_symmetry_time_pct,
),
)
@@ -239,31 +171,32 @@ class TextToLatentsInvocation(BaseInvocation):
scheduler,
# for ddim scheduler
- eta=0.0, #ddim_eta
+ eta=0.0, # ddim_eta
# for ancestral and sde schedulers
generator=torch.Generator(device=uc.device).manual_seed(0),
)
return conditioning_data
- def create_pipeline(self, unet, scheduler) -> StableDiffusionGeneratorPipeline:
+ def create_pipeline(
+ self, unet, scheduler) -> StableDiffusionGeneratorPipeline:
# TODO:
- #configure_model_padding(
+ # configure_model_padding(
# unet,
# self.seamless,
# self.seamless_axes,
- #)
+ # )
class FakeVae:
class FakeVaeConfig:
def __init__(self):
self.block_out_channels = [0]
-
+
def __init__(self):
self.config = FakeVae.FakeVaeConfig()
return StableDiffusionGeneratorPipeline(
- vae=FakeVae(), # TODO: oh...
+ vae=FakeVae(), # TODO: oh...
text_encoder=None,
tokenizer=None,
unet=unet,
@@ -273,11 +206,12 @@ class TextToLatentsInvocation(BaseInvocation):
requires_safety_checker=False,
precision="float16" if unet.dtype == torch.float16 else "float32",
)
-
+
def prep_control_data(
self,
context: InvocationContext,
- model: StableDiffusionGeneratorPipeline, # really only need model for dtype and device
+ # really only need model for dtype and device
+ model: StableDiffusionGeneratorPipeline,
control_input: List[ControlField],
latents_shape: List[int],
do_classifier_free_guidance: bool = True,
@@ -287,19 +221,14 @@ class TextToLatentsInvocation(BaseInvocation):
control_height_resize = latents_shape[2] * 8
control_width_resize = latents_shape[3] * 8
if control_input is None:
- # print("control input is None")
control_list = None
elif isinstance(control_input, list) and len(control_input) == 0:
- # print("control input is empty list")
control_list = None
elif isinstance(control_input, ControlField):
- # print("control input is ControlField")
control_list = [control_input]
elif isinstance(control_input, list) and len(control_input) > 0 and isinstance(control_input[0], ControlField):
- # print("control input is list[ControlField]")
control_list = control_input
else:
- # print("input control is unrecognized:", type(self.control))
control_list = None
if (control_list is None):
control_data = None
@@ -318,15 +247,17 @@ class TextToLatentsInvocation(BaseInvocation):
print("Using HF model subfolders")
print(" control_name: ", control_name)
print(" control_subfolder: ", control_subfolder)
- control_model = ControlNetModel.from_pretrained(control_name,
- subfolder=control_subfolder,
- torch_dtype=model.unet.dtype).to(model.device)
+ control_model = ControlNetModel.from_pretrained(
+ control_name, subfolder=control_subfolder,
+ torch_dtype=model.unet.dtype).to(
+ model.device)
else:
- control_model = ControlNetModel.from_pretrained(control_info.control_model,
- torch_dtype=model.unet.dtype).to(model.device)
+ control_model = ControlNetModel.from_pretrained(
+ control_info.control_model, torch_dtype=model.unet.dtype).to(model.device)
control_models.append(control_model)
control_image_field = control_info.image
- input_image = context.services.images.get_pil_image(control_image_field.image_name)
+ input_image = context.services.images.get_pil_image(
+ control_image_field.image_name)
# self.image.image_type, self.image.image_name
# FIXME: still need to test with different widths, heights, devices, dtypes
# and add in batch_size, num_images_per_prompt?
@@ -341,41 +272,52 @@ class TextToLatentsInvocation(BaseInvocation):
# num_images_per_prompt=num_images_per_prompt,
device=control_model.device,
dtype=control_model.dtype,
+ control_mode=control_info.control_mode,
)
- control_item = ControlNetData(model=control_model,
- image_tensor=control_image,
- weight=control_info.control_weight,
- begin_step_percent=control_info.begin_step_percent,
- end_step_percent=control_info.end_step_percent)
+ control_item = ControlNetData(
+ model=control_model, image_tensor=control_image,
+ weight=control_info.control_weight,
+ begin_step_percent=control_info.begin_step_percent,
+ end_step_percent=control_info.end_step_percent,
+ control_mode=control_info.control_mode,)
control_data.append(control_item)
# MultiControlNetModel has been refactored out, just need list[ControlNetData]
return control_data
+ @torch.no_grad()
def invoke(self, context: InvocationContext) -> LatentsOutput:
noise = context.services.latents.get(self.noise.latents_name)
# Get the source node id (we are invoking the prepared node)
- graph_execution_state = context.services.graph_execution_manager.get(context.graph_execution_state_id)
+ graph_execution_state = context.services.graph_execution_manager.get(
+ context.graph_execution_state_id)
source_node_id = graph_execution_state.prepared_source_mapping[self.id]
def step_callback(state: PipelineIntermediateState):
self.dispatch_progress(context, source_node_id, state)
- unet_info = context.services.model_manager.get_model(**self.unet.unet.dict())
- with unet_info as unet,\
- ExitStack() as stack:
+ def _lora_loader():
+ for lora in self.unet.loras:
+ lora_info = context.services.model_manager.get_model(
+ **lora.dict(exclude={"weight"}))
+ yield (lora_info.context.model, lora.weight)
+ del lora_info
+ return
+
+ unet_info = context.services.model_manager.get_model(
+ **self.unet.unet.dict())
+ with ModelPatcher.apply_lora_unet(unet_info.context.model, _lora_loader()),\
+ unet_info as unet:
scheduler = get_scheduler(
context=context,
scheduler_info=self.unet.scheduler,
scheduler_name=self.scheduler,
)
-
+
pipeline = self.create_pipeline(unet, scheduler)
conditioning_data = self.get_conditioning_data(context, scheduler)
- loras = [(stack.enter_context(context.services.model_manager.get_model(**lora.dict(exclude={"weight"}))), lora.weight) for lora in self.unet.loras]
-
control_data = self.prep_control_data(
model=pipeline, context=context, control_input=self.control,
latents_shape=noise.shape,
@@ -383,16 +325,15 @@ class TextToLatentsInvocation(BaseInvocation):
do_classifier_free_guidance=True,
)
- with ModelPatcher.apply_lora_unet(pipeline.unet, loras):
- # TODO: Verify the noise is the right size
- result_latents, result_attention_map_saver = pipeline.latents_from_embeddings(
- latents=torch.zeros_like(noise, dtype=torch_dtype(unet.device)),
- noise=noise,
- num_inference_steps=self.steps,
- conditioning_data=conditioning_data,
- control_data=control_data, # list[ControlNetData]
- callback=step_callback,
- )
+ # TODO: Verify the noise is the right size
+ result_latents, result_attention_map_saver = pipeline.latents_from_embeddings(
+ latents=torch.zeros_like(noise, dtype=torch_dtype(unet.device)),
+ noise=noise,
+ num_inference_steps=self.steps,
+ conditioning_data=conditioning_data,
+ control_data=control_data, # list[ControlNetData]
+ callback=step_callback,
+ )
# https://discuss.huggingface.co/t/memory-usage-by-later-pipeline-stages/23699
torch.cuda.empty_cache()
@@ -401,14 +342,18 @@ class TextToLatentsInvocation(BaseInvocation):
context.services.latents.save(name, result_latents)
return build_latents_output(latents_name=name, latents=result_latents)
+
class LatentsToLatentsInvocation(TextToLatentsInvocation):
"""Generates latents using latents as base image."""
type: Literal["l2l"] = "l2l"
# Inputs
- latents: Optional[LatentsField] = Field(description="The latents to use as a base image")
- strength: float = Field(default=0.7, ge=0, le=1, description="The strength of the latents to use")
+ latents: Optional[LatentsField] = Field(
+ description="The latents to use as a base image")
+ strength: float = Field(
+ default=0.7, ge=0, le=1,
+ description="The strength of the latents to use")
# Schema customisation
class Config(InvocationConfig):
@@ -423,23 +368,31 @@ class LatentsToLatentsInvocation(TextToLatentsInvocation):
},
}
+ @torch.no_grad()
def invoke(self, context: InvocationContext) -> LatentsOutput:
noise = context.services.latents.get(self.noise.latents_name)
latent = context.services.latents.get(self.latents.latents_name)
# Get the source node id (we are invoking the prepared node)
- graph_execution_state = context.services.graph_execution_manager.get(context.graph_execution_state_id)
+ graph_execution_state = context.services.graph_execution_manager.get(
+ context.graph_execution_state_id)
source_node_id = graph_execution_state.prepared_source_mapping[self.id]
def step_callback(state: PipelineIntermediateState):
self.dispatch_progress(context, source_node_id, state)
- unet_info = context.services.model_manager.get_model(
- **self.unet.unet.dict(),
- )
+ def _lora_loader():
+ for lora in self.unet.loras:
+ lora_info = context.services.model_manager.get_model(
+ **lora.dict(exclude={"weight"}))
+ yield (lora_info.context.model, lora.weight)
+ del lora_info
+ return
- with unet_info as unet,\
- ExitStack() as stack:
+ unet_info = context.services.model_manager.get_model(
+ **self.unet.unet.dict())
+ with ModelPatcher.apply_lora_unet(unet_info.context.model, _lora_loader()),\
+ unet_info as unet:
scheduler = get_scheduler(
context=context,
@@ -449,7 +402,7 @@ class LatentsToLatentsInvocation(TextToLatentsInvocation):
pipeline = self.create_pipeline(unet, scheduler)
conditioning_data = self.get_conditioning_data(context, scheduler)
-
+
control_data = self.prep_control_data(
model=pipeline, context=context, control_input=self.control,
latents_shape=noise.shape,
@@ -459,8 +412,7 @@ class LatentsToLatentsInvocation(TextToLatentsInvocation):
# TODO: Verify the noise is the right size
initial_latents = latent if self.strength < 1.0 else torch.zeros_like(
- latent, device=unet.device, dtype=latent.dtype
- )
+ latent, device=unet.device, dtype=latent.dtype)
timesteps, _ = pipeline.get_img2img_timesteps(
self.steps,
@@ -468,18 +420,15 @@ class LatentsToLatentsInvocation(TextToLatentsInvocation):
device=unet.device,
)
- loras = [(stack.enter_context(context.services.model_manager.get_model(**lora.dict(exclude={"weight"}))), lora.weight) for lora in self.unet.loras]
-
- with ModelPatcher.apply_lora_unet(pipeline.unet, loras):
- result_latents, result_attention_map_saver = pipeline.latents_from_embeddings(
- latents=initial_latents,
- timesteps=timesteps,
- noise=noise,
- num_inference_steps=self.steps,
- conditioning_data=conditioning_data,
- control_data=control_data, # list[ControlNetData]
- callback=step_callback
- )
+ result_latents, result_attention_map_saver = pipeline.latents_from_embeddings(
+ latents=initial_latents,
+ timesteps=timesteps,
+ noise=noise,
+ num_inference_steps=self.steps,
+ conditioning_data=conditioning_data,
+ control_data=control_data, # list[ControlNetData]
+ callback=step_callback
+ )
# https://discuss.huggingface.co/t/memory-usage-by-later-pipeline-stages/23699
torch.cuda.empty_cache()
@@ -496,9 +445,14 @@ class LatentsToImageInvocation(BaseInvocation):
type: Literal["l2i"] = "l2i"
# Inputs
- latents: Optional[LatentsField] = Field(description="The latents to generate an image from")
+ latents: Optional[LatentsField] = Field(
+ description="The latents to generate an image from")
vae: VaeField = Field(default=None, description="Vae submodel")
- tiled: bool = Field(default=False, description="Decode latents by overlaping tiles(less memory consumption)")
+ tiled: bool = Field(
+ default=False,
+ description="Decode latents by overlaping tiles(less memory consumption)")
+ metadata: Optional[CoreMetadata] = Field(default=None, description="Optional core metadata to be written to the image")
+
# Schema customisation
class Config(InvocationConfig):
@@ -529,7 +483,7 @@ class LatentsToImageInvocation(BaseInvocation):
# copied from diffusers pipeline
latents = latents / vae.config.scaling_factor
image = vae.decode(latents, return_dict=False)[0]
- image = (image / 2 + 0.5).clamp(0, 1) # denormalize
+ image = (image / 2 + 0.5).clamp(0, 1) # denormalize
# we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
np_image = image.cpu().permute(0, 2, 3, 1).float().numpy()
@@ -543,6 +497,8 @@ class LatentsToImageInvocation(BaseInvocation):
image_category=ImageCategory.GENERAL,
node_id=self.id,
session_id=context.graph_execution_state_id,
+ is_intermediate=self.is_intermediate,
+ metadata=self.metadata.dict() if self.metadata else None,
)
return ImageOutput(
@@ -551,9 +507,9 @@ class LatentsToImageInvocation(BaseInvocation):
height=image_dto.height,
)
-LATENTS_INTERPOLATION_MODE = Literal[
- "nearest", "linear", "bilinear", "bicubic", "trilinear", "area", "nearest-exact"
-]
+
+LATENTS_INTERPOLATION_MODE = Literal["nearest", "linear",
+ "bilinear", "bicubic", "trilinear", "area", "nearest-exact"]
class ResizeLatentsInvocation(BaseInvocation):
@@ -562,21 +518,25 @@ class ResizeLatentsInvocation(BaseInvocation):
type: Literal["lresize"] = "lresize"
# Inputs
- latents: Optional[LatentsField] = Field(description="The latents to resize")
- width: int = Field(ge=64, multiple_of=8, description="The width to resize to (px)")
- height: int = Field(ge=64, multiple_of=8, description="The height to resize to (px)")
- mode: LATENTS_INTERPOLATION_MODE = Field(default="bilinear", description="The interpolation mode")
- antialias: bool = Field(default=False, description="Whether or not to antialias (applied in bilinear and bicubic modes only)")
+ latents: Optional[LatentsField] = Field(
+ description="The latents to resize")
+ width: int = Field(
+ ge=64, multiple_of=8, description="The width to resize to (px)")
+ height: int = Field(
+ ge=64, multiple_of=8, description="The height to resize to (px)")
+ mode: LATENTS_INTERPOLATION_MODE = Field(
+ default="bilinear", description="The interpolation mode")
+ antialias: bool = Field(
+ default=False,
+ description="Whether or not to antialias (applied in bilinear and bicubic modes only)")
def invoke(self, context: InvocationContext) -> LatentsOutput:
latents = context.services.latents.get(self.latents.latents_name)
resized_latents = torch.nn.functional.interpolate(
- latents,
- size=(self.height // 8, self.width // 8),
- mode=self.mode,
- antialias=self.antialias if self.mode in ["bilinear", "bicubic"] else False,
- )
+ latents, size=(self.height // 8, self.width // 8),
+ mode=self.mode, antialias=self.antialias
+ if self.mode in ["bilinear", "bicubic"] else False,)
# https://discuss.huggingface.co/t/memory-usage-by-later-pipeline-stages/23699
torch.cuda.empty_cache()
@@ -593,21 +553,24 @@ class ScaleLatentsInvocation(BaseInvocation):
type: Literal["lscale"] = "lscale"
# Inputs
- latents: Optional[LatentsField] = Field(description="The latents to scale")
- scale_factor: float = Field(gt=0, description="The factor by which to scale the latents")
- mode: LATENTS_INTERPOLATION_MODE = Field(default="bilinear", description="The interpolation mode")
- antialias: bool = Field(default=False, description="Whether or not to antialias (applied in bilinear and bicubic modes only)")
+ latents: Optional[LatentsField] = Field(
+ description="The latents to scale")
+ scale_factor: float = Field(
+ gt=0, description="The factor by which to scale the latents")
+ mode: LATENTS_INTERPOLATION_MODE = Field(
+ default="bilinear", description="The interpolation mode")
+ antialias: bool = Field(
+ default=False,
+ description="Whether or not to antialias (applied in bilinear and bicubic modes only)")
def invoke(self, context: InvocationContext) -> LatentsOutput:
latents = context.services.latents.get(self.latents.latents_name)
# resizing
resized_latents = torch.nn.functional.interpolate(
- latents,
- scale_factor=self.scale_factor,
- mode=self.mode,
- antialias=self.antialias if self.mode in ["bilinear", "bicubic"] else False,
- )
+ latents, scale_factor=self.scale_factor, mode=self.mode,
+ antialias=self.antialias
+ if self.mode in ["bilinear", "bicubic"] else False,)
# https://discuss.huggingface.co/t/memory-usage-by-later-pipeline-stages/23699
torch.cuda.empty_cache()
@@ -624,9 +587,11 @@ class ImageToLatentsInvocation(BaseInvocation):
type: Literal["i2l"] = "i2l"
# Inputs
- image: Union[ImageField, None] = Field(description="The image to encode")
+ image: Optional[ImageField] = Field(description="The image to encode")
vae: VaeField = Field(default=None, description="Vae submodel")
- tiled: bool = Field(default=False, description="Encode latents by overlaping tiles(less memory consumption)")
+ tiled: bool = Field(
+ default=False,
+ description="Encode latents by overlaping tiles(less memory consumption)")
# Schema customisation
class Config(InvocationConfig):
diff --git a/invokeai/app/invocations/metadata.py b/invokeai/app/invocations/metadata.py
new file mode 100644
index 0000000000..b7639c56c7
--- /dev/null
+++ b/invokeai/app/invocations/metadata.py
@@ -0,0 +1,124 @@
+from typing import Literal, Optional, Union
+
+from pydantic import BaseModel, Field
+
+from invokeai.app.invocations.baseinvocation import (BaseInvocation,
+ BaseInvocationOutput,
+ InvocationContext)
+from invokeai.app.invocations.controlnet_image_processors import ControlField
+from invokeai.app.invocations.model import (LoRAModelField, MainModelField,
+ VAEModelField)
+
+
+class LoRAMetadataField(BaseModel):
+ """LoRA metadata for an image generated in InvokeAI."""
+ lora: LoRAModelField = Field(description="The LoRA model")
+ weight: float = Field(description="The weight of the LoRA model")
+
+
+class CoreMetadata(BaseModel):
+ """Core generation metadata for an image generated in InvokeAI."""
+
+ generation_mode: str = Field(description="The generation mode that output this image",)
+ positive_prompt: str = Field(description="The positive prompt parameter")
+ negative_prompt: str = Field(description="The negative prompt parameter")
+ width: int = Field(description="The width parameter")
+ height: int = Field(description="The height parameter")
+ seed: int = Field(description="The seed used for noise generation")
+ rand_device: str = Field(description="The device used for random number generation")
+ cfg_scale: float = Field(description="The classifier-free guidance scale parameter")
+ steps: int = Field(description="The number of steps used for inference")
+ scheduler: str = Field(description="The scheduler used for inference")
+ clip_skip: int = Field(description="The number of skipped CLIP layers",)
+ model: MainModelField = Field(description="The main model used for inference")
+ controlnets: list[ControlField]= Field(description="The ControlNets used for inference")
+ loras: list[LoRAMetadataField] = Field(description="The LoRAs used for inference")
+ strength: Union[float, None] = Field(
+ default=None,
+ description="The strength used for latents-to-latents",
+ )
+ init_image: Union[str, None] = Field(
+ default=None, description="The name of the initial image"
+ )
+ vae: Union[VAEModelField, None] = Field(
+ default=None,
+ description="The VAE used for decoding, if the main model's default was not used",
+ )
+
+
+class ImageMetadata(BaseModel):
+ """An image's generation metadata"""
+
+ metadata: Optional[dict] = Field(
+ default=None,
+ description="The image's core metadata, if it was created in the Linear or Canvas UI",
+ )
+ graph: Optional[dict] = Field(
+ default=None, description="The graph that created the image"
+ )
+
+
+class MetadataAccumulatorOutput(BaseInvocationOutput):
+ """The output of the MetadataAccumulator node"""
+
+ type: Literal["metadata_accumulator_output"] = "metadata_accumulator_output"
+
+ metadata: CoreMetadata = Field(description="The core metadata for the image")
+
+
+class MetadataAccumulatorInvocation(BaseInvocation):
+ """Outputs a Core Metadata Object"""
+
+ type: Literal["metadata_accumulator"] = "metadata_accumulator"
+
+ generation_mode: str = Field(description="The generation mode that output this image",)
+ positive_prompt: str = Field(description="The positive prompt parameter")
+ negative_prompt: str = Field(description="The negative prompt parameter")
+ width: int = Field(description="The width parameter")
+ height: int = Field(description="The height parameter")
+ seed: int = Field(description="The seed used for noise generation")
+ rand_device: str = Field(description="The device used for random number generation")
+ cfg_scale: float = Field(description="The classifier-free guidance scale parameter")
+ steps: int = Field(description="The number of steps used for inference")
+ scheduler: str = Field(description="The scheduler used for inference")
+ clip_skip: int = Field(description="The number of skipped CLIP layers",)
+ model: MainModelField = Field(description="The main model used for inference")
+ controlnets: list[ControlField]= Field(description="The ControlNets used for inference")
+ loras: list[LoRAMetadataField] = Field(description="The LoRAs used for inference")
+ strength: Union[float, None] = Field(
+ default=None,
+ description="The strength used for latents-to-latents",
+ )
+ init_image: Union[str, None] = Field(
+ default=None, description="The name of the initial image"
+ )
+ vae: Union[VAEModelField, None] = Field(
+ default=None,
+ description="The VAE used for decoding, if the main model's default was not used",
+ )
+
+
+ def invoke(self, context: InvocationContext) -> MetadataAccumulatorOutput:
+ """Collects and outputs a CoreMetadata object"""
+
+ return MetadataAccumulatorOutput(
+ metadata=CoreMetadata(
+ generation_mode=self.generation_mode,
+ positive_prompt=self.positive_prompt,
+ negative_prompt=self.negative_prompt,
+ width=self.width,
+ height=self.height,
+ seed=self.seed,
+ rand_device=self.rand_device,
+ cfg_scale=self.cfg_scale,
+ steps=self.steps,
+ scheduler=self.scheduler,
+ model=self.model,
+ strength=self.strength,
+ init_image=self.init_image,
+ vae=self.vae,
+ controlnets=self.controlnets,
+ loras=self.loras,
+ clip_skip=self.clip_skip,
+ )
+ )
diff --git a/invokeai/app/invocations/model.py b/invokeai/app/invocations/model.py
index 6bc4eb93a4..2ea5bf9687 100644
--- a/invokeai/app/invocations/model.py
+++ b/invokeai/app/invocations/model.py
@@ -1,31 +1,39 @@
-from typing import Literal, Optional, Union, List
-from pydantic import BaseModel, Field
import copy
+from typing import List, Literal, Optional, Union
-from .baseinvocation import BaseInvocation, BaseInvocationOutput, InvocationContext, InvocationConfig
+from pydantic import BaseModel, Field
-from ...backend.util.devices import choose_torch_device, torch_dtype
from ...backend.model_management import BaseModelType, ModelType, SubModelType
+from .baseinvocation import (BaseInvocation, BaseInvocationOutput,
+ InvocationConfig, InvocationContext)
+
class ModelInfo(BaseModel):
model_name: str = Field(description="Info to load submodel")
base_model: BaseModelType = Field(description="Base model")
model_type: ModelType = Field(description="Info to load submodel")
- submodel: Optional[SubModelType] = Field(description="Info to load submodel")
+ submodel: Optional[SubModelType] = Field(
+ default=None, description="Info to load submodel"
+ )
+
class LoraInfo(ModelInfo):
weight: float = Field(description="Lora's weight which to use when apply to model")
+
class UNetField(BaseModel):
unet: ModelInfo = Field(description="Info to load unet submodel")
scheduler: ModelInfo = Field(description="Info to load scheduler submodel")
loras: List[LoraInfo] = Field(description="Loras to apply on model loading")
+
class ClipField(BaseModel):
tokenizer: ModelInfo = Field(description="Info to load tokenizer submodel")
text_encoder: ModelInfo = Field(description="Info to load text_encoder submodel")
+ skipped_layers: int = Field(description="Number of skipped layers in text_encoder")
loras: List[LoraInfo] = Field(description="Loras to apply on model loading")
+
class VaeField(BaseModel):
# TODO: better naming?
vae: ModelInfo = Field(description="Info to load vae submodel")
@@ -34,46 +42,51 @@ class VaeField(BaseModel):
class ModelLoaderOutput(BaseInvocationOutput):
"""Model loader output"""
- #fmt: off
+ # fmt: off
type: Literal["model_loader_output"] = "model_loader_output"
unet: UNetField = Field(default=None, description="UNet submodel")
clip: ClipField = Field(default=None, description="Tokenizer and text_encoder submodels")
vae: VaeField = Field(default=None, description="Vae submodel")
- #fmt: on
+ # fmt: on
-class PipelineModelField(BaseModel):
- """Pipeline model field"""
+class MainModelField(BaseModel):
+ """Main model field"""
model_name: str = Field(description="Name of the model")
base_model: BaseModelType = Field(description="Base model")
-class PipelineModelLoaderInvocation(BaseInvocation):
- """Loads a pipeline model, outputting its submodels."""
+class LoRAModelField(BaseModel):
+ """LoRA model field"""
- type: Literal["pipeline_model_loader"] = "pipeline_model_loader"
+ model_name: str = Field(description="Name of the LoRA model")
+ base_model: BaseModelType = Field(description="Base model")
- model: PipelineModelField = Field(description="The model to load")
+
+class MainModelLoaderInvocation(BaseInvocation):
+ """Loads a main model, outputting its submodels."""
+
+ type: Literal["main_model_loader"] = "main_model_loader"
+
+ model: MainModelField = Field(description="The model to load")
# TODO: precision?
# Schema customisation
class Config(InvocationConfig):
schema_extra = {
"ui": {
+ "title": "Model Loader",
"tags": ["model", "loader"],
- "type_hints": {
- "model": "model"
- }
+ "type_hints": {"model": "model"},
},
}
def invoke(self, context: InvocationContext) -> ModelLoaderOutput:
-
base_model = self.model.base_model
model_name = self.model.model_name
- model_type = ModelType.Pipeline
+ model_type = ModelType.Main
# TODO: not found exceptions
if not context.services.model_manager.model_exists(
@@ -112,7 +125,6 @@ class PipelineModelLoaderInvocation(BaseInvocation):
)
"""
-
return ModelLoaderOutput(
unet=UNetField(
unet=ModelInfo(
@@ -143,6 +155,7 @@ class PipelineModelLoaderInvocation(BaseInvocation):
submodel=SubModelType.TextEncoder,
),
loras=[],
+ skipped_layers=0,
),
vae=VaeField(
vae=ModelInfo(
@@ -151,47 +164,69 @@ class PipelineModelLoaderInvocation(BaseInvocation):
model_type=model_type,
submodel=SubModelType.Vae,
),
- )
+ ),
)
+
class LoraLoaderOutput(BaseInvocationOutput):
"""Model loader output"""
- #fmt: off
+ # fmt: off
type: Literal["lora_loader_output"] = "lora_loader_output"
unet: Optional[UNetField] = Field(default=None, description="UNet submodel")
clip: Optional[ClipField] = Field(default=None, description="Tokenizer and text_encoder submodels")
- #fmt: on
+ # fmt: on
+
class LoraLoaderInvocation(BaseInvocation):
"""Apply selected lora to unet and text_encoder."""
type: Literal["lora_loader"] = "lora_loader"
- lora_name: str = Field(description="Lora model name")
+ lora: Union[LoRAModelField, None] = Field(
+ default=None, description="Lora model name"
+ )
weight: float = Field(default=0.75, description="With what weight to apply lora")
unet: Optional[UNetField] = Field(description="UNet model for applying lora")
clip: Optional[ClipField] = Field(description="Clip model for applying lora")
+ class Config(InvocationConfig):
+ schema_extra = {
+ "ui": {
+ "title": "Lora Loader",
+ "tags": ["lora", "loader"],
+ "type_hints": {"lora": "lora_model"},
+ },
+ }
+
def invoke(self, context: InvocationContext) -> LoraLoaderOutput:
+ if self.lora is None:
+ raise Exception("No LoRA provided")
+
+ base_model = self.lora.base_model
+ lora_name = self.lora.model_name
# TODO: ui rewrite
base_model = BaseModelType.StableDiffusion1
if not context.services.model_manager.model_exists(
base_model=base_model,
- model_name=self.lora_name,
+ model_name=lora_name,
model_type=ModelType.Lora,
):
- raise Exception(f"Unkown lora name: {self.lora_name}!")
+ raise Exception(f"Unkown lora name: {lora_name}!")
- if self.unet is not None and any(lora.model_name == self.lora_name for lora in self.unet.loras):
- raise Exception(f"Lora \"{self.lora_name}\" already applied to unet")
+ if self.unet is not None and any(
+ lora.model_name == lora_name for lora in self.unet.loras
+ ):
+ raise Exception(f'Lora "{lora_name}" already applied to unet')
- if self.clip is not None and any(lora.model_name == self.lora_name for lora in self.clip.loras):
- raise Exception(f"Lora \"{self.lora_name}\" already applied to clip")
+ if self.clip is not None and any(
+ lora.model_name == lora_name for lora in self.clip.loras
+ ):
+ raise Exception(f'Lora "{lora_name}" already applied to clip')
output = LoraLoaderOutput()
@@ -200,7 +235,7 @@ class LoraLoaderInvocation(BaseInvocation):
output.unet.loras.append(
LoraInfo(
base_model=base_model,
- model_name=self.lora_name,
+ model_name=lora_name,
model_type=ModelType.Lora,
submodel=None,
weight=self.weight,
@@ -212,7 +247,7 @@ class LoraLoaderInvocation(BaseInvocation):
output.clip.loras.append(
LoraInfo(
base_model=base_model,
- model_name=self.lora_name,
+ model_name=lora_name,
model_type=ModelType.Lora,
submodel=None,
weight=self.weight,
@@ -221,3 +256,58 @@ class LoraLoaderInvocation(BaseInvocation):
return output
+
+class VAEModelField(BaseModel):
+ """Vae model field"""
+
+ model_name: str = Field(description="Name of the model")
+ base_model: BaseModelType = Field(description="Base model")
+
+
+class VaeLoaderOutput(BaseInvocationOutput):
+ """Model loader output"""
+
+ # fmt: off
+ type: Literal["vae_loader_output"] = "vae_loader_output"
+
+ vae: VaeField = Field(default=None, description="Vae model")
+ # fmt: on
+
+
+class VaeLoaderInvocation(BaseInvocation):
+ """Loads a VAE model, outputting a VaeLoaderOutput"""
+
+ type: Literal["vae_loader"] = "vae_loader"
+
+ vae_model: VAEModelField = Field(description="The VAE to load")
+
+ # Schema customisation
+ class Config(InvocationConfig):
+ schema_extra = {
+ "ui": {
+ "title": "VAE Loader",
+ "tags": ["vae", "loader"],
+ "type_hints": {"vae_model": "vae_model"},
+ },
+ }
+
+ def invoke(self, context: InvocationContext) -> VaeLoaderOutput:
+ base_model = self.vae_model.base_model
+ model_name = self.vae_model.model_name
+ model_type = ModelType.Vae
+
+ if not context.services.model_manager.model_exists(
+ base_model=base_model,
+ model_name=model_name,
+ model_type=model_type,
+ ):
+ raise Exception(f"Unkown vae name: {model_name}!")
+ return VaeLoaderOutput(
+ vae=VaeField(
+ vae=ModelInfo(
+ model_name=model_name,
+ base_model=base_model,
+ model_type=model_type,
+ )
+ )
+ )
diff --git a/invokeai/app/invocations/noise.py b/invokeai/app/invocations/noise.py
new file mode 100644
index 0000000000..0d62ada34e
--- /dev/null
+++ b/invokeai/app/invocations/noise.py
@@ -0,0 +1,134 @@
+# Copyright (c) 2023 Kyle Schouviller (https://github.com/kyle0654) & the InvokeAI Team
+
+import math
+from typing import Literal
+
+from pydantic import Field, validator
+import torch
+from invokeai.app.invocations.latent import LatentsField
+
+from invokeai.app.util.misc import SEED_MAX, get_random_seed
+from ...backend.util.devices import choose_torch_device, torch_dtype
+from .baseinvocation import (
+ BaseInvocation,
+ BaseInvocationOutput,
+ InvocationConfig,
+ InvocationContext,
+)
+
+"""
+Utilities
+"""
+
+
+def get_noise(
+ width: int,
+ height: int,
+ device: torch.device,
+ seed: int = 0,
+ latent_channels: int = 4,
+ downsampling_factor: int = 8,
+ use_cpu: bool = True,
+ perlin: float = 0.0,
+):
+ """Generate noise for a given image size."""
+ noise_device_type = "cpu" if use_cpu else device.type
+
+ # limit noise to only the diffusion image channels, not the mask channels
+ input_channels = min(latent_channels, 4)
+ generator = torch.Generator(device=noise_device_type).manual_seed(seed)
+
+ noise_tensor = torch.randn(
+ [
+ 1,
+ input_channels,
+ height // downsampling_factor,
+ width // downsampling_factor,
+ ],
+ dtype=torch_dtype(device),
+ device=noise_device_type,
+ generator=generator,
+ ).to(device)
+
+ return noise_tensor
+
+
+"""
+Nodes
+"""
+
+
+class NoiseOutput(BaseInvocationOutput):
+ """Invocation noise output"""
+
+ # fmt: off
+ type: Literal["noise_output"] = "noise_output"
+
+ # Inputs
+ noise: LatentsField = Field(default=None, description="The output noise")
+ width: int = Field(description="The width of the noise in pixels")
+ height: int = Field(description="The height of the noise in pixels")
+ # fmt: on
+
+
+def build_noise_output(latents_name: str, latents: torch.Tensor):
+ return NoiseOutput(
+ noise=LatentsField(latents_name=latents_name),
+ width=latents.size()[3] * 8,
+ height=latents.size()[2] * 8,
+ )
+
+
+class NoiseInvocation(BaseInvocation):
+ """Generates latent noise."""
+
+ type: Literal["noise"] = "noise"
+
+ # Inputs
+ seed: int = Field(
+ ge=0,
+ le=SEED_MAX,
+ description="The seed to use",
+ default_factory=get_random_seed,
+ )
+ width: int = Field(
+ default=512,
+ multiple_of=8,
+ gt=0,
+ description="The width of the resulting noise",
+ )
+ height: int = Field(
+ default=512,
+ multiple_of=8,
+ gt=0,
+ description="The height of the resulting noise",
+ )
+ use_cpu: bool = Field(
+ default=True,
+ description="Use CPU for noise generation (for reproducible results across platforms)",
+ )
+
+ # Schema customisation
+ class Config(InvocationConfig):
+ schema_extra = {
+ "ui": {
+ "tags": ["latents", "noise"],
+ },
+ }
+
+ @validator("seed", pre=True)
+ def modulo_seed(cls, v):
+ """Returns the seed modulo SEED_MAX to ensure it is within the valid range."""
+ return v % SEED_MAX
+
+ def invoke(self, context: InvocationContext) -> NoiseOutput:
+ noise = get_noise(
+ width=self.width,
+ height=self.height,
+ device=choose_torch_device(),
+ seed=self.seed,
+ use_cpu=self.use_cpu,
+ )
+ name = f"{context.graph_execution_state_id}__{self.id}"
+ context.services.latents.save(name, noise)
+ return build_noise_output(latents_name=name, latents=noise)
diff --git a/invokeai/app/invocations/param_easing.py b/invokeai/app/invocations/param_easing.py
index 1ff6261b88..e79763a35e 100644
--- a/invokeai/app/invocations/param_easing.py
+++ b/invokeai/app/invocations/param_easing.py
@@ -133,20 +133,19 @@ class StepParamEasingInvocation(BaseInvocation):
postlist = list(num_poststeps * [self.post_end_value])
if log_diagnostics:
- logger = InvokeAILogger.getLogger(name="StepParamEasing")
- logger.debug("start_step: " + str(start_step))
- logger.debug("end_step: " + str(end_step))
- logger.debug("num_easing_steps: " + str(num_easing_steps))
- logger.debug("num_presteps: " + str(num_presteps))
- logger.debug("num_poststeps: " + str(num_poststeps))
- logger.debug("prelist size: " + str(len(prelist)))
- logger.debug("postlist size: " + str(len(postlist)))
- logger.debug("prelist: " + str(prelist))
- logger.debug("postlist: " + str(postlist))
+ context.services.logger.debug("start_step: " + str(start_step))
+ context.services.logger.debug("end_step: " + str(end_step))
+ context.services.logger.debug("num_easing_steps: " + str(num_easing_steps))
+ context.services.logger.debug("num_presteps: " + str(num_presteps))
+ context.services.logger.debug("num_poststeps: " + str(num_poststeps))
+ context.services.logger.debug("prelist size: " + str(len(prelist)))
+ context.services.logger.debug("postlist size: " + str(len(postlist)))
+ context.services.logger.debug("prelist: " + str(prelist))
+ context.services.logger.debug("postlist: " + str(postlist))
easing_class = EASING_FUNCTIONS_MAP[self.easing]
if log_diagnostics:
- logger.debug("easing class: " + str(easing_class))
+ context.services.logger.debug("easing class: " + str(easing_class))
easing_list = list()
if self.mirror: # "expected" mirroring
# if number of steps is even, squeeze duration down to (number_of_steps)/2
@@ -156,7 +155,7 @@ class StepParamEasingInvocation(BaseInvocation):
# but if even then number_of_steps/2 === ceil(number_of_steps/2), so can just use ceil always
base_easing_duration = int(np.ceil(num_easing_steps/2.0))
- if log_diagnostics: logger.debug("base easing duration: " + str(base_easing_duration))
+ if log_diagnostics: context.services.logger.debug("base easing duration: " + str(base_easing_duration))
even_num_steps = (num_easing_steps % 2 == 0) # even number of steps
easing_function = easing_class(start=self.start_value,
end=self.end_value,
@@ -166,14 +165,14 @@ class StepParamEasingInvocation(BaseInvocation):
easing_val = easing_function.ease(step_index)
base_easing_vals.append(easing_val)
if log_diagnostics:
- logger.debug("step_index: " + str(step_index) + ", easing_val: " + str(easing_val))
+ context.services.logger.debug("step_index: " + str(step_index) + ", easing_val: " + str(easing_val))
if even_num_steps:
mirror_easing_vals = list(reversed(base_easing_vals))
else:
mirror_easing_vals = list(reversed(base_easing_vals[0:-1]))
if log_diagnostics:
- logger.debug("base easing vals: " + str(base_easing_vals))
- logger.debug("mirror easing vals: " + str(mirror_easing_vals))
+ context.services.logger.debug("base easing vals: " + str(base_easing_vals))
+ context.services.logger.debug("mirror easing vals: " + str(mirror_easing_vals))
easing_list = base_easing_vals + mirror_easing_vals
# FIXME: add alt_mirror option (alternative to default or mirror), or remove entirely
@@ -206,12 +205,12 @@ class StepParamEasingInvocation(BaseInvocation):
step_val = easing_function.ease(step_index)
easing_list.append(step_val)
if log_diagnostics:
- logger.debug("step_index: " + str(step_index) + ", easing_val: " + str(step_val))
+ context.services.logger.debug("step_index: " + str(step_index) + ", easing_val: " + str(step_val))
if log_diagnostics:
- logger.debug("prelist size: " + str(len(prelist)))
- logger.debug("easing_list size: " + str(len(easing_list)))
- logger.debug("postlist size: " + str(len(postlist)))
+ context.services.logger.debug("prelist size: " + str(len(prelist)))
+ context.services.logger.debug("easing_list size: " + str(len(easing_list)))
+ context.services.logger.debug("postlist size: " + str(len(postlist)))
param_list = prelist + easing_list + postlist
diff --git a/invokeai/app/invocations/reconstruct.py b/invokeai/app/invocations/reconstruct.py
index 4185de3fd3..84ace2eefe 100644
--- a/invokeai/app/invocations/reconstruct.py
+++ b/invokeai/app/invocations/reconstruct.py
@@ -1,4 +1,4 @@
-from typing import Literal, Union
+from typing import Literal, Optional
from pydantic import Field
@@ -15,7 +15,7 @@ class RestoreFaceInvocation(BaseInvocation):
type: Literal["restore_face"] = "restore_face"
# Inputs
- image: Union[ImageField, None] = Field(description="The input image")
+ image: Optional[ImageField] = Field(description="The input image")
strength: float = Field(default=0.75, gt=0, le=1, description="The strength of the restoration" )
# fmt: on
diff --git a/invokeai/app/invocations/upscale.py b/invokeai/app/invocations/upscale.py
index 42f85fd18d..15497343ae 100644
--- a/invokeai/app/invocations/upscale.py
+++ b/invokeai/app/invocations/upscale.py
@@ -1,6 +1,6 @@
# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
-from typing import Literal, Union
+from typing import Literal, Optional
from pydantic import Field
@@ -16,7 +16,7 @@ class UpscaleInvocation(BaseInvocation):
type: Literal["upscale"] = "upscale"
# Inputs
- image: Union[ImageField, None] = Field(description="The input image", default=None)
+ image: Optional[ImageField] = Field(description="The input image", default=None)
strength: float = Field(default=0.75, gt=0, le=1, description="The strength")
level: Literal[2, 4] = Field(default=2, description="The upscale level")
# fmt: on
diff --git a/invokeai/app/models/metadata.py b/invokeai/app/models/metadata.py
deleted file mode 100644
index 8d90ca0bc8..0000000000
--- a/invokeai/app/models/metadata.py
+++ /dev/null
@@ -1,93 +0,0 @@
-from typing import Optional, Union, List
-from pydantic import BaseModel, Extra, Field, StrictFloat, StrictInt, StrictStr
-
-
-class ImageMetadata(BaseModel):
- """
- Core generation metadata for an image/tensor generated in InvokeAI.
-
- Also includes any metadata from the image's PNG tEXt chunks.
-
- Generated by traversing the execution graph, collecting the parameters of the nearest ancestors
- of a given node.
-
- Full metadata may be accessed by querying for the session in the `graph_executions` table.
- """
-
- class Config:
- extra = Extra.allow
- """
- This lets the ImageMetadata class accept arbitrary additional fields. The CoreMetadataService
- won't add any fields that are not already defined, but other a different metadata service
- implementation might.
- """
-
- type: Optional[StrictStr] = Field(
- default=None,
- description="The type of the ancestor node of the image output node.",
- )
- """The type of the ancestor node of the image output node."""
- positive_conditioning: Optional[StrictStr] = Field(
- default=None, description="The positive conditioning."
- )
- """The positive conditioning"""
- negative_conditioning: Optional[StrictStr] = Field(
- default=None, description="The negative conditioning."
- )
- """The negative conditioning"""
- width: Optional[StrictInt] = Field(
- default=None, description="Width of the image/latents in pixels."
- )
- """Width of the image/latents in pixels"""
- height: Optional[StrictInt] = Field(
- default=None, description="Height of the image/latents in pixels."
- )
- """Height of the image/latents in pixels"""
- seed: Optional[StrictInt] = Field(
- default=None, description="The seed used for noise generation."
- )
- """The seed used for noise generation"""
- # cfg_scale: Optional[StrictFloat] = Field(
- # cfg_scale: Union[float, list[float]] = Field(
- cfg_scale: Union[StrictFloat, List[StrictFloat]] = Field(
- default=None, description="The classifier-free guidance scale."
- )
- """The classifier-free guidance scale"""
- steps: Optional[StrictInt] = Field(
- default=None, description="The number of steps used for inference."
- )
- """The number of steps used for inference"""
- scheduler: Optional[StrictStr] = Field(
- default=None, description="The scheduler used for inference."
- )
- """The scheduler used for inference"""
- model: Optional[StrictStr] = Field(
- default=None, description="The model used for inference."
- )
- """The model used for inference"""
- strength: Optional[StrictFloat] = Field(
- default=None,
- description="The strength used for image-to-image/latents-to-latents.",
- )
- """The strength used for image-to-image/latents-to-latents."""
- latents: Optional[StrictStr] = Field(
- default=None, description="The ID of the initial latents."
- )
- """The ID of the initial latents"""
- vae: Optional[StrictStr] = Field(
- default=None, description="The VAE used for decoding."
- )
- """The VAE used for decoding"""
- unet: Optional[StrictStr] = Field(
- default=None, description="The UNet used dor inference."
- )
- """The UNet used dor inference"""
- clip: Optional[StrictStr] = Field(
- default=None, description="The CLIP Encoder used for conditioning."
- )
- """The CLIP Encoder used for conditioning"""
- extra: Optional[StrictStr] = Field(
- default=None,
- description="Uploaded image metadata, extracted from the PNG tEXt chunk.",
- )
- """Uploaded image metadata, extracted from the PNG tEXt chunk."""
diff --git a/invokeai/app/services/board_image_record_storage.py b/invokeai/app/services/board_image_record_storage.py
index 7aff41860c..197a639157 100644
--- a/invokeai/app/services/board_image_record_storage.py
+++ b/invokeai/app/services/board_image_record_storage.py
@@ -1,8 +1,7 @@
from abc import ABC, abstractmethod
import sqlite3
import threading
-from typing import Union, cast
-from invokeai.app.services.board_record_storage import BoardRecord
+from typing import Optional, cast
from invokeai.app.services.image_record_storage import OffsetPaginatedResults
from invokeai.app.services.models.image_record import (
@@ -44,7 +43,7 @@ class BoardImageRecordStorageBase(ABC):
def get_board_for_image(
self,
image_name: str,
- ) -> Union[str, None]:
+ ) -> Optional[str]:
"""Gets an image's board id, if it has one."""
pass
@@ -215,7 +214,7 @@ class SqliteBoardImageRecordStorage(BoardImageRecordStorageBase):
def get_board_for_image(
self,
image_name: str,
- ) -> Union[str, None]:
+ ) -> Optional[str]:
try:
self._lock.acquire()
self._cursor.execute(
diff --git a/invokeai/app/services/board_images.py b/invokeai/app/services/board_images.py
index 072effbfae..1ba225338b 100644
--- a/invokeai/app/services/board_images.py
+++ b/invokeai/app/services/board_images.py
@@ -1,6 +1,6 @@
from abc import ABC, abstractmethod
from logging import Logger
-from typing import List, Union
+from typing import List, Union, Optional
from invokeai.app.services.board_image_record_storage import BoardImageRecordStorageBase
from invokeai.app.services.board_record_storage import (
BoardRecord,
@@ -49,7 +49,7 @@ class BoardImagesServiceABC(ABC):
def get_board_for_image(
self,
image_name: str,
- ) -> Union[str, None]:
+ ) -> Optional[str]:
"""Gets an image's board id, if it has one."""
pass
@@ -126,13 +126,13 @@ class BoardImagesService(BoardImagesServiceABC):
def get_board_for_image(
self,
image_name: str,
- ) -> Union[str, None]:
+ ) -> Optional[str]:
board_id = self._services.board_image_records.get_board_for_image(image_name)
return board_id
def board_record_to_dto(
- board_record: BoardRecord, cover_image_name: str | None, image_count: int
+ board_record: BoardRecord, cover_image_name: Optional[str], image_count: int
) -> BoardDTO:
"""Converts a board record to a board DTO."""
return BoardDTO(
diff --git a/invokeai/app/services/config.py b/invokeai/app/services/config.py
index 014119289f..6d27c013f1 100644
--- a/invokeai/app/services/config.py
+++ b/invokeai/app/services/config.py
@@ -15,7 +15,7 @@ InvokeAI:
conf_path: configs/models.yaml
legacy_conf_dir: configs/stable-diffusion
outdir: outputs
- autoconvert_dir: null
+ autoimport_dir: null
Models:
model: stable-diffusion-1.5
embeddings: true
@@ -23,7 +23,8 @@ InvokeAI:
xformers_enabled: false
sequential_guidance: false
precision: float16
- max_loaded_models: 4
+ max_cache_size: 6
+ max_vram_cache_size: 2.7
always_use_cpu: false
free_gpu_mem: false
Features:
@@ -168,9 +169,10 @@ from argparse import ArgumentParser
from omegaconf import OmegaConf, DictConfig
from pathlib import Path
from pydantic import BaseSettings, Field, parse_obj_as
-from typing import ClassVar, Dict, List, Literal, Union, get_origin, get_type_hints, get_args
+from typing import ClassVar, Dict, List, Set, Literal, Union, get_origin, get_type_hints, get_args
INIT_FILE = Path('invokeai.yaml')
+MODEL_CORE = Path('models/core')
DB_FILE = Path('invokeai.db')
LEGACY_INIT_FILE = Path('invokeai.init')
@@ -228,10 +230,10 @@ class InvokeAISettings(BaseSettings):
upcase_environ = dict()
for key,value in os.environ.items():
upcase_environ[key.upper()] = value
-
+
fields = cls.__fields__
cls.argparse_groups = {}
-
+
for name, field in fields.items():
if name not in cls._excluded():
current_default = field.default
@@ -269,7 +271,8 @@ class InvokeAISettings(BaseSettings):
@classmethod
def _excluded(self)->List[str]:
- return ['type','initconf']
+ # combination of deprecated parameters and internal ones
+ return ['type','initconf', 'gpu_mem_reserved', 'max_loaded_models', 'version']
class Config:
env_file_encoding = 'utf-8'
@@ -324,16 +327,11 @@ class InvokeAISettings(BaseSettings):
help=field.field_info.description,
)
def _find_root()->Path:
+ venv = Path(os.environ.get("VIRTUAL_ENV") or ".")
if os.environ.get("INVOKEAI_ROOT"):
root = Path(os.environ.get("INVOKEAI_ROOT")).resolve()
- elif (
- os.environ.get("VIRTUAL_ENV")
- and (Path(os.environ.get("VIRTUAL_ENV"), "..", INIT_FILE).exists()
- or
- Path(os.environ.get("VIRTUAL_ENV"), "..", LEGACY_INIT_FILE).exists()
- )
- ):
- root = Path(os.environ.get("VIRTUAL_ENV"), "..").resolve()
+ elif any([(venv.parent/x).exists() for x in [INIT_FILE, LEGACY_INIT_FILE, MODEL_CORE]]):
+ root = (venv.parent).resolve()
else:
root = Path("~/invokeai").expanduser().resolve()
return root
@@ -348,7 +346,7 @@ setting environment variables INVOKEAI_.
'''
singleton_config: ClassVar[InvokeAIAppConfig] = None
singleton_init: ClassVar[Dict] = None
-
+
#fmt: off
type: Literal["InvokeAI"] = "InvokeAI"
host : str = Field(default="127.0.0.1", description="IP address to bind to", category='Web Server')
@@ -367,33 +365,41 @@ setting environment variables INVOKEAI_.
always_use_cpu : bool = Field(default=False, description="If true, use the CPU for rendering even if a GPU is available.", category='Memory/Performance')
free_gpu_mem : bool = Field(default=False, description="If true, purge model from GPU after each generation.", category='Memory/Performance')
- max_loaded_models : int = Field(default=2, gt=0, description="Maximum number of models to keep in memory for rapid switching", category='Memory/Performance')
+ max_loaded_models : int = Field(default=3, gt=0, description="(DEPRECATED: use max_cache_size) Maximum number of models to keep in memory for rapid switching", category='DEPRECATED')
+ max_cache_size : float = Field(default=6.0, gt=0, description="Maximum memory amount used by model cache for rapid switching", category='Memory/Performance')
+ max_vram_cache_size : float = Field(default=2.75, ge=0, description="Amount of VRAM reserved for model storage", category='Memory/Performance')
+ gpu_mem_reserved : float = Field(default=2.75, ge=0, description="DEPRECATED: use max_vram_cache_size. Amount of VRAM reserved for model storage", category='DEPRECATED')
precision : Literal[tuple(['auto','float16','float32','autocast'])] = Field(default='float16',description='Floating point precision', category='Memory/Performance')
sequential_guidance : bool = Field(default=False, description="Whether to calculate guidance in serial instead of in parallel, lowering memory requirements", category='Memory/Performance')
xformers_enabled : bool = Field(default=True, description="Enable/disable memory-efficient attention", category='Memory/Performance')
tiled_decode : bool = Field(default=False, description="Whether to enable tiled VAE decode (reduces memory consumption with some performance penalty)", category='Memory/Performance')
root : Path = Field(default=_find_root(), description='InvokeAI runtime root directory', category='Paths')
- autoconvert_dir : Path = Field(default=None, description='Path to a directory of ckpt files to be converted into diffusers and imported on startup.', category='Paths')
+ autoimport_dir : Path = Field(default='autoimport/main', description='Path to a directory of models files to be imported on startup.', category='Paths')
+ lora_dir : Path = Field(default='autoimport/lora', description='Path to a directory of LoRA/LyCORIS models to be imported on startup.', category='Paths')
+ embedding_dir : Path = Field(default='autoimport/embedding', description='Path to a directory of Textual Inversion embeddings to be imported on startup.', category='Paths')
+ controlnet_dir : Path = Field(default='autoimport/controlnet', description='Path to a directory of ControlNet embeddings to be imported on startup.', category='Paths')
conf_path : Path = Field(default='configs/models.yaml', description='Path to models definition file', category='Paths')
- models_dir : Path = Field(default='./models', description='Path to the models directory', category='Paths')
+ models_dir : Path = Field(default='models', description='Path to the models directory', category='Paths')
legacy_conf_dir : Path = Field(default='configs/stable-diffusion', description='Path to directory of legacy checkpoint config files', category='Paths')
db_dir : Path = Field(default='databases', description='Path to InvokeAI databases directory', category='Paths')
outdir : Path = Field(default='outputs', description='Default folder for output images', category='Paths')
from_file : Path = Field(default=None, description='Take command input from the indicated file (command-line client only)', category='Paths')
use_memory_db : bool = Field(default=False, description='Use in-memory database for storing image metadata', category='Paths')
-
+
model : str = Field(default='stable-diffusion-1.5', description='Initial model name', category='Models')
-
+
log_handlers : List[str] = Field(default=["console"], description='Log handler. Valid options are "console", "file=", "syslog=path|address:host:port", "http="', category="Logging")
# note - would be better to read the log_format values from logging.py, but this creates circular dependencies issues
log_format : Literal[tuple(['plain','color','syslog','legacy'])] = Field(default="color", description='Log format. Use "plain" for text-only, "color" for colorized output, "legacy" for 2.3-style logging and "syslog" for syslog-style', category="Logging")
log_level : Literal[tuple(["debug","info","warning","error","critical"])] = Field(default="debug", description="Emit logging messages at this level or higher", category="Logging")
+
+ version : bool = Field(default=False, description="Show InvokeAI version and exit", category="Other")
#fmt: on
def parse_args(self, argv: List[str]=None, conf: DictConfig = None, clobber=False):
'''
- Update settings with contents of init file, environment, and
+ Update settings with contents of init file, environment, and
command-line settings.
:param conf: alternate Omegaconf dictionary object
:param argv: aternate sys.argv list
@@ -408,7 +414,7 @@ setting environment variables INVOKEAI_.
except:
pass
InvokeAISettings.initconf = conf
-
+
# parse args again in order to pick up settings in configuration file
super().parse_args(argv)
@@ -428,7 +434,7 @@ setting environment variables INVOKEAI_.
cls.singleton_config = cls(**kwargs)
cls.singleton_init = kwargs
return cls.singleton_config
-
+
@property
def root_path(self)->Path:
'''
diff --git a/invokeai/app/services/default_graphs.py b/invokeai/app/services/default_graphs.py
index 5eda5e957d..92263751b7 100644
--- a/invokeai/app/services/default_graphs.py
+++ b/invokeai/app/services/default_graphs.py
@@ -1,4 +1,5 @@
-from ..invocations.latent import LatentsToImageInvocation, NoiseInvocation, TextToLatentsInvocation
+from ..invocations.latent import LatentsToImageInvocation, TextToLatentsInvocation
+from ..invocations.noise import NoiseInvocation
from ..invocations.compel import CompelInvocation
from ..invocations.params import ParamIntInvocation
from .graph import Edge, EdgeConnection, ExposedNodeInput, ExposedNodeOutput, Graph, LibraryGraph
diff --git a/invokeai/app/services/events.py b/invokeai/app/services/events.py
index e578a24006..6c516c9b74 100644
--- a/invokeai/app/services/events.py
+++ b/invokeai/app/services/events.py
@@ -1,10 +1,9 @@
# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
-from typing import Any
+from typing import Any, Optional
from invokeai.app.models.image import ProgressImage
from invokeai.app.util.misc import get_timestamp
from invokeai.app.services.model_manager_service import BaseModelType, ModelType, SubModelType, ModelInfo
-from invokeai.app.models.exceptions import CanceledException
class EventServiceBase:
session_event: str = "session_event"
@@ -28,7 +27,7 @@ class EventServiceBase:
graph_execution_state_id: str,
node: dict,
source_node_id: str,
- progress_image: ProgressImage | None,
+ progress_image: Optional[ProgressImage],
step: int,
total_steps: int,
) -> None:
diff --git a/invokeai/app/services/graph.py b/invokeai/app/services/graph.py
index e3cd3d47ce..24096da29b 100644
--- a/invokeai/app/services/graph.py
+++ b/invokeai/app/services/graph.py
@@ -3,7 +3,6 @@
import copy
import itertools
import uuid
-from types import NoneType
from typing import (
Annotated,
Any,
@@ -26,6 +25,8 @@ from ..invocations.baseinvocation import (
InvocationContext,
)
+# in 3.10 this would be "from types import NoneType"
+NoneType = type(None)
class EdgeConnection(BaseModel):
node_id: str = Field(description="The id of the node for this edge connection")
@@ -60,8 +61,6 @@ def get_input_field(node: BaseInvocation, field: str) -> Any:
node_input_field = node_inputs.get(field) or None
return node_input_field
-from typing import Optional, Union, List, get_args
-
def is_union_subtype(t1, t2):
t1_args = get_args(t1)
t2_args = get_args(t2)
@@ -846,7 +845,7 @@ class GraphExecutionState(BaseModel):
]
}
- def next(self) -> BaseInvocation | None:
+ def next(self) -> Optional[BaseInvocation]:
"""Gets the next node ready to execute."""
# TODO: enable multiple nodes to execute simultaneously by tracking currently executing nodes
diff --git a/invokeai/app/services/image_file_storage.py b/invokeai/app/services/image_file_storage.py
index b90b9b2f8b..60ae613748 100644
--- a/invokeai/app/services/image_file_storage.py
+++ b/invokeai/app/services/image_file_storage.py
@@ -1,15 +1,14 @@
# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654) and the InvokeAI Team
+import json
from abc import ABC, abstractmethod
from pathlib import Path
from queue import Queue
-from typing import Dict, Optional
+from typing import Dict, Optional, Union
-from PIL.Image import Image as PILImageType
from PIL import Image, PngImagePlugin
+from PIL.Image import Image as PILImageType
from send2trash import send2trash
-from invokeai.app.models.image import ResourceOrigin
-from invokeai.app.models.metadata import ImageMetadata
from invokeai.app.util.thumbnails import get_thumbnail_name, make_thumbnail
@@ -60,7 +59,8 @@ class ImageFileStorageBase(ABC):
self,
image: PILImageType,
image_name: str,
- metadata: Optional[ImageMetadata] = None,
+ metadata: Optional[dict] = None,
+ graph: Optional[dict] = None,
thumbnail_size: int = 256,
) -> None:
"""Saves an image and a 256x256 WEBP thumbnail. Returns a tuple of the image name, thumbnail name, and created timestamp."""
@@ -80,13 +80,15 @@ class DiskImageFileStorage(ImageFileStorageBase):
__cache: Dict[Path, PILImageType]
__max_cache_size: int
- def __init__(self, output_folder: str | Path):
+ def __init__(self, output_folder: Union[str, Path]):
self.__cache = dict()
self.__cache_ids = Queue()
self.__max_cache_size = 10 # TODO: get this from config
- self.__output_folder: Path = output_folder if isinstance(output_folder, Path) else Path(output_folder)
- self.__thumbnails_folder = self.__output_folder / 'thumbnails'
+ self.__output_folder: Path = (
+ output_folder if isinstance(output_folder, Path) else Path(output_folder)
+ )
+ self.__thumbnails_folder = self.__output_folder / "thumbnails"
# Validate required output folders at launch
self.__validate_storage_folders()
@@ -94,7 +96,7 @@ class DiskImageFileStorage(ImageFileStorageBase):
def get(self, image_name: str) -> PILImageType:
try:
image_path = self.get_path(image_name)
-
+
cache_item = self.__get_cache(image_path)
if cache_item:
return cache_item
@@ -109,20 +111,22 @@ class DiskImageFileStorage(ImageFileStorageBase):
self,
image: PILImageType,
image_name: str,
- metadata: Optional[ImageMetadata] = None,
+ metadata: Optional[dict] = None,
+ graph: Optional[dict] = None,
thumbnail_size: int = 256,
) -> None:
try:
self.__validate_storage_folders()
image_path = self.get_path(image_name)
+ pnginfo = PngImagePlugin.PngInfo()
+
if metadata is not None:
- pnginfo = PngImagePlugin.PngInfo()
- pnginfo.add_text("invokeai", metadata.json())
- image.save(image_path, "PNG", pnginfo=pnginfo)
- else:
- image.save(image_path, "PNG")
+ pnginfo.add_text("invokeai_metadata", json.dumps(metadata))
+ if graph is not None:
+ pnginfo.add_text("invokeai_graph", json.dumps(graph))
+ image.save(image_path, "PNG", pnginfo=pnginfo)
thumbnail_name = get_thumbnail_name(image_name)
thumbnail_path = self.get_path(thumbnail_name, thumbnail=True)
thumbnail_image = make_thumbnail(image, thumbnail_size)
@@ -155,31 +159,33 @@ class DiskImageFileStorage(ImageFileStorageBase):
# TODO: make this a bit more flexible for e.g. cloud storage
def get_path(self, image_name: str, thumbnail: bool = False) -> Path:
path = self.__output_folder / image_name
-
+
if thumbnail:
thumbnail_name = get_thumbnail_name(image_name)
path = self.__thumbnails_folder / thumbnail_name
return path
- def validate_path(self, path: str | Path) -> bool:
+ def validate_path(self, path: Union[str, Path]) -> bool:
"""Validates the path given for an image or thumbnail."""
path = path if isinstance(path, Path) else Path(path)
return path.exists()
-
+
def __validate_storage_folders(self) -> None:
"""Checks if the required output folders exist and create them if they don't"""
folders: list[Path] = [self.__output_folder, self.__thumbnails_folder]
for folder in folders:
folder.mkdir(parents=True, exist_ok=True)
- def __get_cache(self, image_name: Path) -> PILImageType | None:
+ def __get_cache(self, image_name: Path) -> Optional[PILImageType]:
return None if image_name not in self.__cache else self.__cache[image_name]
def __set_cache(self, image_name: Path, image: PILImageType):
if not image_name in self.__cache:
self.__cache[image_name] = image
- self.__cache_ids.put(image_name) # TODO: this should refresh position for LRU cache
+ self.__cache_ids.put(
+ image_name
+ ) # TODO: this should refresh position for LRU cache
if len(self.__cache) > self.__max_cache_size:
cache_id = self.__cache_ids.get()
if cache_id in self.__cache:
diff --git a/invokeai/app/services/image_record_storage.py b/invokeai/app/services/image_record_storage.py
index c34d2ca5c8..7b37307ce8 100644
--- a/invokeai/app/services/image_record_storage.py
+++ b/invokeai/app/services/image_record_storage.py
@@ -1,23 +1,16 @@
+import json
+import sqlite3
+import threading
from abc import ABC, abstractmethod
from datetime import datetime
from typing import Generic, Optional, TypeVar, cast
-import sqlite3
-import threading
-from typing import Optional, Union
from pydantic import BaseModel, Field
from pydantic.generics import GenericModel
-from invokeai.app.models.metadata import ImageMetadata
-from invokeai.app.models.image import (
- ImageCategory,
- ResourceOrigin,
-)
+from invokeai.app.models.image import ImageCategory, ResourceOrigin
from invokeai.app.services.models.image_record import (
- ImageRecord,
- ImageRecordChanges,
- deserialize_image_record,
-)
+ ImageRecord, ImageRecordChanges, deserialize_image_record)
T = TypeVar("T", bound=BaseModel)
@@ -55,6 +48,28 @@ class ImageRecordDeleteException(Exception):
super().__init__(message)
+IMAGE_DTO_COLS = ", ".join(
+ list(
+ map(
+ lambda c: "images." + c,
+ [
+ "image_name",
+ "image_origin",
+ "image_category",
+ "width",
+ "height",
+ "session_id",
+ "node_id",
+ "is_intermediate",
+ "created_at",
+ "updated_at",
+ "deleted_at",
+ ],
+ )
+ )
+)
+
+
class ImageRecordStorageBase(ABC):
"""Low-level service responsible for interfacing with the image record store."""
@@ -65,6 +80,11 @@ class ImageRecordStorageBase(ABC):
"""Gets an image record."""
pass
+ @abstractmethod
+ def get_metadata(self, image_name: str) -> Optional[dict]:
+ """Gets an image's metadata'."""
+ pass
+
@abstractmethod
def update(
self,
@@ -94,6 +114,11 @@ class ImageRecordStorageBase(ABC):
"""Deletes an image record."""
pass
+ @abstractmethod
+ def delete_many(self, image_names: list[str]) -> None:
+ """Deletes many image records."""
+ pass
+
@abstractmethod
def save(
self,
@@ -104,14 +129,14 @@ class ImageRecordStorageBase(ABC):
height: int,
session_id: Optional[str],
node_id: Optional[str],
- metadata: Optional[ImageMetadata],
+ metadata: Optional[dict],
is_intermediate: bool = False,
) -> datetime:
"""Saves an image record."""
pass
@abstractmethod
- def get_most_recent_image_for_board(self, board_id: str) -> ImageRecord | None:
+ def get_most_recent_image_for_board(self, board_id: str) -> Optional[ImageRecord]:
"""Gets the most recent image for a board."""
pass
@@ -158,7 +183,6 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
node_id TEXT,
metadata TEXT,
is_intermediate BOOLEAN DEFAULT FALSE,
- board_id TEXT,
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')),
@@ -203,19 +227,19 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
"""
)
- def get(self, image_name: str) -> Union[ImageRecord, None]:
+ def get(self, image_name: str) -> Optional[ImageRecord]:
try:
self._lock.acquire()
self._cursor.execute(
f"""--sql
- SELECT * FROM images
+ SELECT {IMAGE_DTO_COLS} FROM images
WHERE image_name = ?;
""",
(image_name,),
)
- result = cast(Union[sqlite3.Row, None], self._cursor.fetchone())
+ result = cast(Optional[sqlite3.Row], self._cursor.fetchone())
except sqlite3.Error as e:
self._conn.rollback()
raise ImageRecordNotFoundException from e
@@ -227,6 +251,28 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
return deserialize_image_record(dict(result))
+ def get_metadata(self, image_name: str) -> Optional[dict]:
+ try:
+ self._lock.acquire()
+
+ self._cursor.execute(
+ f"""--sql
+ SELECT images.metadata FROM images
+ WHERE image_name = ?;
+ """,
+ (image_name,),
+ )
+
+ result = cast(Optional[sqlite3.Row], self._cursor.fetchone())
+ if not result or not result[0]:
+ return None
+ return json.loads(result[0])
+ except sqlite3.Error as e:
+ self._conn.rollback()
+ raise ImageRecordNotFoundException from e
+ finally:
+ self._lock.release()
+
def update(
self,
image_name: str,
@@ -294,8 +340,8 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
WHERE 1=1
"""
- images_query = """--sql
- SELECT images.*
+ images_query = f"""--sql
+ SELECT {IMAGE_DTO_COLS}
FROM images
LEFT JOIN board_images ON board_images.image_name = images.image_name
WHERE 1=1
@@ -385,6 +431,25 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
finally:
self._lock.release()
+ def delete_many(self, image_names: list[str]) -> None:
+ try:
+ placeholders = ",".join("?" for _ in image_names)
+
+ self._lock.acquire()
+
+ # Construct the SQLite query with the placeholders
+ query = f"DELETE FROM images WHERE image_name IN ({placeholders})"
+
+ # Execute the query with the list of IDs as parameters
+ self._cursor.execute(query, image_names)
+
+ self._conn.commit()
+ except sqlite3.Error as e:
+ self._conn.rollback()
+ raise ImageRecordDeleteException from e
+ finally:
+ self._lock.release()
+
def save(
self,
image_name: str,
@@ -394,12 +459,12 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
width: int,
height: int,
node_id: Optional[str],
- metadata: Optional[ImageMetadata],
+ metadata: Optional[dict],
is_intermediate: bool = False,
) -> datetime:
try:
metadata_json = (
- None if metadata is None else metadata.json(exclude_none=True)
+ None if metadata is None else json.dumps(metadata)
)
self._lock.acquire()
self._cursor.execute(
@@ -449,9 +514,7 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
finally:
self._lock.release()
- def get_most_recent_image_for_board(
- self, board_id: str
- ) -> Union[ImageRecord, None]:
+ def get_most_recent_image_for_board(self, board_id: str) -> Optional[ImageRecord]:
try:
self._lock.acquire()
self._cursor.execute(
@@ -466,7 +529,7 @@ class SqliteImageRecordStorage(ImageRecordStorageBase):
(board_id,),
)
- result = cast(Union[sqlite3.Row, None], self._cursor.fetchone())
+ result = cast(Optional[sqlite3.Row], self._cursor.fetchone())
finally:
self._lock.release()
if result is None:
diff --git a/invokeai/app/services/images.py b/invokeai/app/services/images.py
index 542f874f1d..a7d0b6ddee 100644
--- a/invokeai/app/services/images.py
+++ b/invokeai/app/services/images.py
@@ -1,39 +1,30 @@
+import json
from abc import ABC, abstractmethod
from logging import Logger
-from typing import Optional, TYPE_CHECKING, Union
+from typing import TYPE_CHECKING, Optional
+
from PIL.Image import Image as PILImageType
-from invokeai.app.models.image import (
- ImageCategory,
- ResourceOrigin,
- InvalidImageCategoryException,
- InvalidOriginException,
-)
-from invokeai.app.models.metadata import ImageMetadata
-from invokeai.app.services.board_image_record_storage import BoardImageRecordStorageBase
-from invokeai.app.services.image_record_storage import (
- ImageRecordDeleteException,
- ImageRecordNotFoundException,
- ImageRecordSaveException,
- ImageRecordStorageBase,
- OffsetPaginatedResults,
-)
-from invokeai.app.services.models.image_record import (
- ImageRecord,
- ImageDTO,
- ImageRecordChanges,
- image_record_to_dto,
-)
+from invokeai.app.invocations.metadata import ImageMetadata
+from invokeai.app.models.image import (ImageCategory,
+ InvalidImageCategoryException,
+ InvalidOriginException, ResourceOrigin)
+from invokeai.app.services.board_image_record_storage import \
+ BoardImageRecordStorageBase
+from invokeai.app.services.graph import Graph
from invokeai.app.services.image_file_storage import (
- ImageFileDeleteException,
- ImageFileNotFoundException,
- ImageFileSaveException,
- ImageFileStorageBase,
-)
-from invokeai.app.services.item_storage import ItemStorageABC, PaginatedResults
-from invokeai.app.services.metadata import MetadataServiceBase
+ ImageFileDeleteException, ImageFileNotFoundException,
+ ImageFileSaveException, ImageFileStorageBase)
+from invokeai.app.services.image_record_storage import (
+ ImageRecordDeleteException, ImageRecordNotFoundException,
+ ImageRecordSaveException, ImageRecordStorageBase, OffsetPaginatedResults)
+from invokeai.app.services.item_storage import ItemStorageABC
+from invokeai.app.services.models.image_record import (ImageDTO, ImageRecord,
+ ImageRecordChanges,
+ image_record_to_dto)
from invokeai.app.services.resource_name import NameServiceBase
from invokeai.app.services.urls import UrlServiceBase
+from invokeai.app.util.metadata import get_metadata_graph_from_raw_session
if TYPE_CHECKING:
from invokeai.app.services.graph import GraphExecutionState
@@ -51,6 +42,7 @@ class ImageServiceABC(ABC):
node_id: Optional[str] = None,
session_id: Optional[str] = None,
is_intermediate: bool = False,
+ metadata: Optional[dict] = None,
) -> ImageDTO:
"""Creates an image, storing the file and its metadata."""
pass
@@ -79,6 +71,11 @@ class ImageServiceABC(ABC):
"""Gets an image DTO."""
pass
+ @abstractmethod
+ def get_metadata(self, image_name: str) -> ImageMetadata:
+ """Gets an image's metadata."""
+ pass
+
@abstractmethod
def get_path(self, image_name: str, thumbnail: bool = False) -> str:
"""Gets an image's path."""
@@ -112,6 +109,11 @@ class ImageServiceABC(ABC):
"""Deletes an image."""
pass
+ @abstractmethod
+ def delete_images_on_board(self, board_id: str):
+ """Deletes all images on a board."""
+ pass
+
class ImageServiceDependencies:
"""Service dependencies for the ImageService."""
@@ -119,7 +121,6 @@ class ImageServiceDependencies:
image_records: ImageRecordStorageBase
image_files: ImageFileStorageBase
board_image_records: BoardImageRecordStorageBase
- metadata: MetadataServiceBase
urls: UrlServiceBase
logger: Logger
names: NameServiceBase
@@ -130,7 +131,6 @@ class ImageServiceDependencies:
image_record_storage: ImageRecordStorageBase,
image_file_storage: ImageFileStorageBase,
board_image_record_storage: BoardImageRecordStorageBase,
- metadata: MetadataServiceBase,
url: UrlServiceBase,
logger: Logger,
names: NameServiceBase,
@@ -139,7 +139,6 @@ class ImageServiceDependencies:
self.image_records = image_record_storage
self.image_files = image_file_storage
self.board_image_records = board_image_record_storage
- self.metadata = metadata
self.urls = url
self.logger = logger
self.names = names
@@ -160,6 +159,7 @@ class ImageService(ImageServiceABC):
node_id: Optional[str] = None,
session_id: Optional[str] = None,
is_intermediate: bool = False,
+ metadata: Optional[dict] = None,
) -> ImageDTO:
if image_origin not in ResourceOrigin:
raise InvalidOriginException
@@ -169,7 +169,16 @@ class ImageService(ImageServiceABC):
image_name = self._services.names.create_image_name()
- metadata = self._get_metadata(session_id, node_id)
+ graph = None
+
+ if session_id is not None:
+ session_raw = self._services.graph_execution_manager.get_raw(session_id)
+ if session_raw is not None:
+ try:
+ graph = get_metadata_graph_from_raw_session(session_raw)
+ except Exception as e:
+ self._services.logger.warn(f"Failed to parse session graph: {e}")
+ graph = None
(width, height) = image.size
@@ -186,14 +195,12 @@ class ImageService(ImageServiceABC):
is_intermediate=is_intermediate,
# Nullable fields
node_id=node_id,
- session_id=session_id,
metadata=metadata,
+ session_id=session_id,
)
self._services.image_files.save(
- image_name=image_name,
- image=image,
- metadata=metadata,
+ image_name=image_name, image=image, metadata=metadata, graph=graph
)
image_dto = self.get_dto(image_name)
@@ -263,6 +270,34 @@ class ImageService(ImageServiceABC):
self._services.logger.error("Problem getting image DTO")
raise e
+ def get_metadata(self, image_name: str) -> Optional[ImageMetadata]:
+ try:
+ image_record = self._services.image_records.get(image_name)
+
+ if not image_record.session_id:
+ return ImageMetadata()
+
+ session_raw = self._services.graph_execution_manager.get_raw(
+ image_record.session_id
+ )
+ graph = None
+
+ if session_raw:
+ try:
+ graph = get_metadata_graph_from_raw_session(session_raw)
+ except Exception as e:
+ self._services.logger.warn(f"Failed to parse session graph: {e}")
+ graph = None
+
+ metadata = self._services.image_records.get_metadata(image_name)
+ return ImageMetadata(graph=graph, metadata=metadata)
+ except ImageRecordNotFoundException:
+ self._services.logger.error("Image record not found")
+ raise
+ except Exception as e:
+ self._services.logger.error("Problem getting image DTO")
+ raise e
+
def get_path(self, image_name: str, thumbnail: bool = False) -> str:
try:
return self._services.image_files.get_path(image_name, thumbnail)
@@ -341,14 +376,24 @@ class ImageService(ImageServiceABC):
self._services.logger.error("Problem deleting image record and file")
raise e
- def _get_metadata(
- self, session_id: Optional[str] = None, node_id: Optional[str] = None
- ) -> Union[ImageMetadata, None]:
- """Get the metadata for a node."""
- metadata = None
-
- if node_id is not None and session_id is not None:
- session = self._services.graph_execution_manager.get(session_id)
- metadata = self._services.metadata.create_image_metadata(session, node_id)
-
- return metadata
+ def delete_images_on_board(self, board_id: str):
+ try:
+ images = self._services.board_image_records.get_images_for_board(board_id)
+ image_name_list = list(
+ map(
+ lambda r: r.image_name,
+ images.items,
+ )
+ )
+ for image_name in image_name_list:
+ self._services.image_files.delete(image_name)
+ self._services.image_records.delete_many(image_name_list)
+ except ImageRecordDeleteException:
+ self._services.logger.error(f"Failed to delete image records")
+ raise
+ except ImageFileDeleteException:
+ self._services.logger.error(f"Failed to delete image files")
+ raise
+ except Exception as e:
+ self._services.logger.error("Problem deleting image records and files")
+ raise e
diff --git a/invokeai/app/services/invocation_queue.py b/invokeai/app/services/invocation_queue.py
index acfda6b90b..eb78a542a6 100644
--- a/invokeai/app/services/invocation_queue.py
+++ b/invokeai/app/services/invocation_queue.py
@@ -5,7 +5,7 @@ from abc import ABC, abstractmethod
from queue import Queue
from pydantic import BaseModel, Field
-
+from typing import Optional
class InvocationQueueItem(BaseModel):
graph_execution_state_id: str = Field(description="The ID of the graph execution state")
@@ -22,7 +22,7 @@ class InvocationQueueABC(ABC):
pass
@abstractmethod
- def put(self, item: InvocationQueueItem | None) -> None:
+ def put(self, item: Optional[InvocationQueueItem]) -> None:
pass
@abstractmethod
@@ -57,7 +57,7 @@ class MemoryInvocationQueue(InvocationQueueABC):
return item
- def put(self, item: InvocationQueueItem | None) -> None:
+ def put(self, item: Optional[InvocationQueueItem]) -> None:
self.__queue.put(item)
def cancel(self, graph_execution_state_id: str) -> None:
diff --git a/invokeai/app/services/invocation_services.py b/invokeai/app/services/invocation_services.py
index 10d1d91920..4e1da3b040 100644
--- a/invokeai/app/services/invocation_services.py
+++ b/invokeai/app/services/invocation_services.py
@@ -7,7 +7,7 @@ if TYPE_CHECKING:
from invokeai.app.services.board_images import BoardImagesServiceABC
from invokeai.app.services.boards import BoardServiceABC
from invokeai.app.services.images import ImageServiceABC
- from invokeai.backend import ModelManager
+ from invokeai.app.services.model_manager_service import ModelManagerServiceBase
from invokeai.app.services.events import EventServiceBase
from invokeai.app.services.latent_storage import LatentsStorageBase
from invokeai.app.services.restoration_services import RestorationServices
@@ -22,46 +22,47 @@ class InvocationServices:
"""Services that can be used by invocations"""
# TODO: Just forward-declared everything due to circular dependencies. Fix structure.
- events: "EventServiceBase"
- latents: "LatentsStorageBase"
- queue: "InvocationQueueABC"
- model_manager: "ModelManager"
- restoration: "RestorationServices"
- configuration: "InvokeAISettings"
- images: "ImageServiceABC"
- boards: "BoardServiceABC"
board_images: "BoardImagesServiceABC"
- graph_library: "ItemStorageABC"["LibraryGraph"]
+ boards: "BoardServiceABC"
+ configuration: "InvokeAISettings"
+ events: "EventServiceBase"
graph_execution_manager: "ItemStorageABC"["GraphExecutionState"]
+ graph_library: "ItemStorageABC"["LibraryGraph"]
+ images: "ImageServiceABC"
+ latents: "LatentsStorageBase"
+ logger: "Logger"
+ model_manager: "ModelManagerServiceBase"
processor: "InvocationProcessorABC"
+ queue: "InvocationQueueABC"
+ restoration: "RestorationServices"
def __init__(
self,
- model_manager: "ModelManager",
- events: "EventServiceBase",
- logger: "Logger",
- latents: "LatentsStorageBase",
- images: "ImageServiceABC",
- boards: "BoardServiceABC",
board_images: "BoardImagesServiceABC",
- queue: "InvocationQueueABC",
- graph_library: "ItemStorageABC"["LibraryGraph"],
- graph_execution_manager: "ItemStorageABC"["GraphExecutionState"],
- processor: "InvocationProcessorABC",
- restoration: "RestorationServices",
+ boards: "BoardServiceABC",
configuration: "InvokeAISettings",
+ events: "EventServiceBase",
+ graph_execution_manager: "ItemStorageABC"["GraphExecutionState"],
+ graph_library: "ItemStorageABC"["LibraryGraph"],
+ images: "ImageServiceABC",
+ latents: "LatentsStorageBase",
+ logger: "Logger",
+ model_manager: "ModelManagerServiceBase",
+ processor: "InvocationProcessorABC",
+ queue: "InvocationQueueABC",
+ restoration: "RestorationServices",
):
- self.model_manager = model_manager
- self.events = events
- self.logger = logger
- self.latents = latents
- self.images = images
- self.boards = boards
self.board_images = board_images
- self.queue = queue
- self.graph_library = graph_library
- self.graph_execution_manager = graph_execution_manager
- self.processor = processor
- self.restoration = restoration
- self.configuration = configuration
self.boards = boards
+ self.boards = boards
+ self.configuration = configuration
+ self.events = events
+ self.graph_execution_manager = graph_execution_manager
+ self.graph_library = graph_library
+ self.images = images
+ self.latents = latents
+ self.logger = logger
+ self.model_manager = model_manager
+ self.processor = processor
+ self.queue = queue
+ self.restoration = restoration
diff --git a/invokeai/app/services/invoker.py b/invokeai/app/services/invoker.py
index f12ba79c15..951d3b17c4 100644
--- a/invokeai/app/services/invoker.py
+++ b/invokeai/app/services/invoker.py
@@ -1,14 +1,11 @@
# Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
from abc import ABC
-from threading import Event, Thread
+from typing import Optional
-from ..invocations.baseinvocation import InvocationContext
from .graph import Graph, GraphExecutionState
-from .invocation_queue import InvocationQueueABC, InvocationQueueItem
+from .invocation_queue import InvocationQueueItem
from .invocation_services import InvocationServices
-from .item_storage import ItemStorageABC
-
class Invoker:
"""The invoker, used to execute invocations"""
@@ -21,7 +18,7 @@ class Invoker:
def invoke(
self, graph_execution_state: GraphExecutionState, invoke_all: bool = False
- ) -> str | None:
+ ) -> Optional[str]:
"""Determines the next node to invoke and enqueues it, preparing if needed.
Returns the id of the queued node, or `None` if there are no nodes left to enqueue."""
@@ -45,7 +42,7 @@ class Invoker:
return invocation.id
- def create_execution_state(self, graph: Graph | None = None) -> GraphExecutionState:
+ def create_execution_state(self, graph: Optional[Graph] = None) -> GraphExecutionState:
"""Creates a new execution state for the given graph"""
new_state = GraphExecutionState(graph=Graph() if graph is None else graph)
self.services.graph_execution_manager.set(new_state)
diff --git a/invokeai/app/services/item_storage.py b/invokeai/app/services/item_storage.py
index 394f67797d..709d88bf97 100644
--- a/invokeai/app/services/item_storage.py
+++ b/invokeai/app/services/item_storage.py
@@ -1,5 +1,5 @@
from abc import ABC, abstractmethod
-from typing import Callable, Generic, TypeVar
+from typing import Callable, Generic, Optional, TypeVar
from pydantic import BaseModel, Field
from pydantic.generics import GenericModel
@@ -29,14 +29,22 @@ class ItemStorageABC(ABC, Generic[T]):
@abstractmethod
def get(self, item_id: str) -> T:
+ """Gets the item, parsing it into a Pydantic model"""
+ pass
+
+ @abstractmethod
+ def get_raw(self, item_id: str) -> Optional[str]:
+ """Gets the raw item as a string, skipping Pydantic parsing"""
pass
@abstractmethod
def set(self, item: T) -> None:
+ """Sets the item"""
pass
@abstractmethod
def list(self, page: int = 0, per_page: int = 10) -> PaginatedResults[T]:
+ """Gets a paginated list of items"""
pass
@abstractmethod
diff --git a/invokeai/app/services/latent_storage.py b/invokeai/app/services/latent_storage.py
index 17d35d7c33..0e23d6d018 100644
--- a/invokeai/app/services/latent_storage.py
+++ b/invokeai/app/services/latent_storage.py
@@ -3,7 +3,7 @@
from abc import ABC, abstractmethod
from pathlib import Path
from queue import Queue
-from typing import Dict
+from typing import Dict, Union, Optional
import torch
@@ -55,7 +55,7 @@ class ForwardCacheLatentsStorage(LatentsStorageBase):
if name in self.__cache:
del self.__cache[name]
- def __get_cache(self, name: str) -> torch.Tensor|None:
+ def __get_cache(self, name: str) -> Optional[torch.Tensor]:
return None if name not in self.__cache else self.__cache[name]
def __set_cache(self, name: str, data: torch.Tensor):
@@ -69,9 +69,9 @@ class ForwardCacheLatentsStorage(LatentsStorageBase):
class DiskLatentsStorage(LatentsStorageBase):
"""Stores latents in a folder on disk without caching"""
- __output_folder: str | Path
+ __output_folder: Union[str, Path]
- def __init__(self, output_folder: str | Path):
+ def __init__(self, output_folder: Union[str, Path]):
self.__output_folder = output_folder if isinstance(output_folder, Path) else Path(output_folder)
self.__output_folder.mkdir(parents=True, exist_ok=True)
@@ -91,4 +91,4 @@ class DiskLatentsStorage(LatentsStorageBase):
def get_path(self, name: str) -> Path:
return self.__output_folder / name
-
\ No newline at end of file
+
diff --git a/invokeai/app/services/metadata.py b/invokeai/app/services/metadata.py
deleted file mode 100644
index aa7bc2291a..0000000000
--- a/invokeai/app/services/metadata.py
+++ /dev/null
@@ -1,142 +0,0 @@
-from abc import ABC, abstractmethod
-from typing import Any, Union
-import networkx as nx
-
-from invokeai.app.models.metadata import ImageMetadata
-from invokeai.app.services.graph import Graph, GraphExecutionState
-
-
-class MetadataServiceBase(ABC):
- """Handles building metadata for nodes, images, and outputs."""
-
- @abstractmethod
- def create_image_metadata(
- self, session: GraphExecutionState, node_id: str
- ) -> ImageMetadata:
- """Builds an ImageMetadata object for a node."""
- pass
-
-
-class CoreMetadataService(MetadataServiceBase):
- _ANCESTOR_TYPES = ["t2l", "l2l"]
- """The ancestor types that contain the core metadata"""
-
- _ANCESTOR_PARAMS = ["type", "steps", "model", "cfg_scale", "scheduler", "strength"]
- """The core metadata parameters in the ancestor types"""
-
- _NOISE_FIELDS = ["seed", "width", "height"]
- """The core metadata parameters in the noise node"""
-
- def create_image_metadata(
- self, session: GraphExecutionState, node_id: str
- ) -> ImageMetadata:
- metadata = self._build_metadata_from_graph(session, node_id)
-
- return metadata
-
- def _find_nearest_ancestor(self, G: nx.DiGraph, node_id: str) -> Union[str, None]:
- """
- Finds the id of the nearest ancestor (of a valid type) of a given node.
-
- Parameters:
- G (nx.DiGraph): The execution graph, converted in to a networkx DiGraph. Its nodes must
- have the same data as the execution graph.
- node_id (str): The ID of the node.
-
- Returns:
- str | None: The ID of the nearest ancestor, or None if there are no valid ancestors.
- """
-
- # Retrieve the node from the graph
- node = G.nodes[node_id]
-
- # If the node type is one of the core metadata node types, return its id
- if node.get("type") in self._ANCESTOR_TYPES:
- return node.get("id")
-
- # Else, look for the ancestor in the predecessor nodes
- for predecessor in G.predecessors(node_id):
- result = self._find_nearest_ancestor(G, predecessor)
- if result:
- return result
-
- # If there are no valid ancestors, return None
- return None
-
- def _get_additional_metadata(
- self, graph: Graph, node_id: str
- ) -> Union[dict[str, Any], None]:
- """
- Returns additional metadata for a given node.
-
- Parameters:
- graph (Graph): The execution graph.
- node_id (str): The ID of the node.
-
- Returns:
- dict[str, Any] | None: A dictionary of additional metadata.
- """
-
- metadata = {}
-
- # Iterate over all edges in the graph
- for edge in graph.edges:
- dest_node_id = edge.destination.node_id
- dest_field = edge.destination.field
- source_node_dict = graph.nodes[edge.source.node_id].dict()
-
- # If the destination node ID matches the given node ID, gather necessary metadata
- if dest_node_id == node_id:
- # Prompt
- if dest_field == "positive_conditioning":
- metadata["positive_conditioning"] = source_node_dict.get("prompt")
- # Negative prompt
- if dest_field == "negative_conditioning":
- metadata["negative_conditioning"] = source_node_dict.get("prompt")
- # Seed, width and height
- if dest_field == "noise":
- for field in self._NOISE_FIELDS:
- metadata[field] = source_node_dict.get(field)
- return metadata
-
- def _build_metadata_from_graph(
- self, session: GraphExecutionState, node_id: str
- ) -> ImageMetadata:
- """
- Builds an ImageMetadata object for a node.
-
- Parameters:
- session (GraphExecutionState): The session.
- node_id (str): The ID of the node.
-
- Returns:
- ImageMetadata: The metadata for the node.
- """
-
- # We need to do all the traversal on the execution graph
- graph = session.execution_graph
-
- # Find the nearest `t2l`/`l2l` ancestor of the given node
- ancestor_id = self._find_nearest_ancestor(graph.nx_graph_with_data(), node_id)
-
- # If no ancestor was found, return an empty ImageMetadata object
- if ancestor_id is None:
- return ImageMetadata()
-
- ancestor_node = graph.get_node(ancestor_id)
-
- # Grab all the core metadata from the ancestor node
- ancestor_metadata = {
- param: val
- for param, val in ancestor_node.dict().items()
- if param in self._ANCESTOR_PARAMS
- }
-
- # Get this image's prompts and noise parameters
- addl_metadata = self._get_additional_metadata(graph, ancestor_id)
-
- # If additional metadata was found, add it to the main metadata
- if addl_metadata is not None:
- ancestor_metadata.update(addl_metadata)
-
- return ImageMetadata(**ancestor_metadata)
diff --git a/invokeai/app/services/model_manager_service.py b/invokeai/app/services/model_manager_service.py
index 8b46b17ad0..1b1c43dc11 100644
--- a/invokeai/app/services/model_manager_service.py
+++ b/invokeai/app/services/model_manager_service.py
@@ -2,22 +2,29 @@
from __future__ import annotations
-import torch
from abc import ABC, abstractmethod
from pathlib import Path
-from typing import Optional, Union, Callable, List, Tuple, types, TYPE_CHECKING
-from dataclasses import dataclass
+from pydantic import Field
+from typing import Optional, Union, Callable, List, Tuple, TYPE_CHECKING
+from types import ModuleType
-from invokeai.backend.model_management.model_manager import (
+from invokeai.backend.model_management import (
ModelManager,
BaseModelType,
ModelType,
SubModelType,
ModelInfo,
+ AddModelResult,
+ SchedulerPredictionType,
+ ModelMerger,
+ MergeInterpolationMethod,
)
+
+
+import torch
from invokeai.app.models.exceptions import CanceledException
-from .config import InvokeAIAppConfig
from ...backend.util import choose_precision, choose_torch_device
+from .config import InvokeAIAppConfig
if TYPE_CHECKING:
from ..invocations.baseinvocation import BaseInvocation, InvocationContext
@@ -30,16 +37,16 @@ class ModelManagerServiceBase(ABC):
def __init__(
self,
config: InvokeAIAppConfig,
- logger: types.ModuleType,
+ logger: ModuleType,
):
"""
- Initialize with the path to the models.yaml config file.
+ Initialize with the path to the models.yaml config file.
Optional parameters are the torch device type, precision, max_models,
and sequential_offload boolean. Note that the default device
type and precision are set up for a CUDA system running at half precision.
"""
pass
-
+
@abstractmethod
def get_model(
self,
@@ -50,8 +57,8 @@ class ModelManagerServiceBase(ABC):
node: Optional[BaseInvocation] = None,
context: Optional[InvocationContext] = None,
) -> ModelInfo:
- """Retrieve the indicated model with name and type.
- submodel can be used to get a part (such as the vae)
+ """Retrieve the indicated model with name and type.
+ submodel can be used to get a part (such as the vae)
of a diffusers pipeline."""
pass
@@ -73,13 +80,7 @@ class ModelManagerServiceBase(ABC):
def model_info(self, model_name: str, base_model: BaseModelType, model_type: ModelType) -> dict:
"""
Given a model name returns a dict-like (OmegaConf) object describing it.
- """
- pass
-
- @abstractmethod
- def model_names(self) -> List[Tuple[str, BaseModelType, ModelType]]:
- """
- Returns a list of all the model names known.
+ Uses the exact format as the omegaconf stanza.
"""
pass
@@ -101,7 +102,20 @@ class ModelManagerServiceBase(ABC):
}
"""
pass
+
+ @abstractmethod
+ def list_model(self, model_name: str, base_model: BaseModelType, model_type: ModelType) -> dict:
+ """
+ Return information about the model using the same format as list_models()
+ """
+ pass
+ @abstractmethod
+ def model_names(self) -> List[Tuple[str, BaseModelType, ModelType]]:
+ """
+ Returns a list of all the model names known.
+ """
+ pass
@abstractmethod
def add_model(
@@ -111,16 +125,34 @@ class ModelManagerServiceBase(ABC):
model_type: ModelType,
model_attributes: dict,
clobber: bool = False
- ) -> None:
+ ) -> AddModelResult:
"""
Update the named model with a dictionary of attributes. Will fail with an
assertion error if the name already exists. Pass clobber=True to overwrite.
+ On a successful update, the config will be changed in memory. Will fail
+ with an assertion error if provided attributes are incorrect or
+ the model name is missing. Call commit() to write changes to disk.
+ """
+ pass
+
+ @abstractmethod
+ def update_model(
+ self,
+ model_name: str,
+ base_model: BaseModelType,
+ model_type: ModelType,
+ model_attributes: dict,
+ ) -> AddModelResult:
+ """
+ Update the named model with a dictionary of attributes. Will fail with a
+ KeyErrorException if the name does not already exist.
+
On a successful update, the config will be changed in memory. Will fail
with an assertion error if provided attributes are incorrect or
the model name is missing. Call commit() to write changes to disk.
"""
pass
-
+
@abstractmethod
def del_model(
self,
@@ -129,14 +161,78 @@ class ModelManagerServiceBase(ABC):
model_type: ModelType,
):
"""
- Delete the named model from configuration. If delete_files is true,
- then the underlying weight file or diffusers directory will be deleted
+ Delete the named model from configuration. If delete_files is true,
+ then the underlying weight file or diffusers directory will be deleted
as well. Call commit() to write to disk.
"""
pass
@abstractmethod
- def commit(self, conf_file: Path = None) -> None:
+ def convert_model(
+ self,
+ model_name: str,
+ base_model: BaseModelType,
+ model_type: Union[ModelType.Main,ModelType.Vae],
+ ) -> AddModelResult:
+ """
+ Convert a checkpoint file into a diffusers folder, deleting the cached
+ version and deleting the original checkpoint file if it is in the models
+ directory.
+ :param model_name: Name of the model to convert
+ :param base_model: Base model type
+ :param model_type: Type of model ['vae' or 'main']
+
+ This will raise a ValueError unless the model is not a checkpoint. It will
+ also raise a ValueError in the event that there is a similarly-named diffusers
+ directory already in place.
+ """
+ pass
+
+ @abstractmethod
+ def heuristic_import(self,
+ items_to_import: set[str],
+ prediction_type_helper: Optional[Callable[[Path],SchedulerPredictionType]]=None,
+ )->dict[str, AddModelResult]:
+ '''Import a list of paths, repo_ids or URLs. Returns the set of
+ successfully imported items.
+ :param items_to_import: Set of strings corresponding to models to be imported.
+ :param prediction_type_helper: A callback that receives the Path of a Stable Diffusion 2 checkpoint model and returns a SchedulerPredictionType.
+
+ The prediction type helper is necessary to distinguish between
+ models based on Stable Diffusion 2 Base (requiring
+ SchedulerPredictionType.Epsilson) and Stable Diffusion 768
+ (requiring SchedulerPredictionType.VPrediction). It is
+ generally impossible to do this programmatically, so the
+ prediction_type_helper usually asks the user to choose.
+
+ The result is a set of successfully installed models. Each element
+ of the set is a dict corresponding to the newly-created OmegaConf stanza for
+ that model.
+ '''
+ pass
+
+ @abstractmethod
+ def merge_models(
+ self,
+ model_names: List[str] = Field(default=None, min_items=2, max_items=3, description="List of model names to merge"),
+ base_model: Union[BaseModelType,str] = Field(default=None, description="Base model shared by all models to be merged"),
+ merged_model_name: str = Field(default=None, description="Name of destination model after merging"),
+ alpha: Optional[float] = 0.5,
+ interp: Optional[MergeInterpolationMethod] = None,
+ force: Optional[bool] = False,
+ ) -> AddModelResult:
+ """
+ Merge two to three diffusrs pipeline models and save as a new model.
+ :param model_names: List of 2-3 models to merge
+ :param base_model: Base model to use for all models
+ :param merged_model_name: Name of destination merged model
+ :param alpha: Alpha strength to apply to 2d and 3d model
+ :param interp: Interpolation method. None (default)
+ """
+ pass
+
+ @abstractmethod
+ def commit(self, conf_file: Optional[Path] = None) -> None:
"""
Write current configuration out to the indicated file.
If no conf_file is provided, then replaces the
@@ -150,10 +246,10 @@ class ModelManagerService(ModelManagerServiceBase):
def __init__(
self,
config: InvokeAIAppConfig,
- logger: types.ModuleType,
+ logger: ModuleType,
):
"""
- Initialize with the path to the models.yaml config file.
+ Initialize with the path to the models.yaml config file.
Optional parameters are the torch device type, precision, max_models,
and sequential_offload boolean. Note that the default device
type and precision are set up for a CUDA system running at half precision.
@@ -162,12 +258,12 @@ class ModelManagerService(ModelManagerServiceBase):
config_file = config.model_conf_path
else:
config_file = config.root_dir / "configs/models.yaml"
- if not config_file.exists():
- raise IOError(f"The file {config_file} could not be found.")
-
+
logger.debug(f'config file={config_file}')
device = torch.device(choose_torch_device())
+ logger.debug(f'GPU device = {device}')
+
precision = config.precision
if precision == "auto":
precision = choose_precision(device)
@@ -183,6 +279,8 @@ class ModelManagerService(ModelManagerServiceBase):
if hasattr(config,'max_cache_size') \
else config.max_loaded_models * 2.5
+ logger.debug(f"Maximum RAM cache size: {max_cache_size} GiB")
+
sequential_offload = config.sequential_guidance
self.mgr = ModelManager(
@@ -238,7 +336,7 @@ class ModelManagerService(ModelManagerServiceBase):
submodel=submodel,
model_info=model_info
)
-
+
return model_info
def model_exists(
@@ -274,12 +372,19 @@ class ModelManagerService(ModelManagerServiceBase):
base_model: Optional[BaseModelType] = None,
model_type: Optional[ModelType] = None
) -> list[dict]:
- # ) -> dict:
"""
Return a list of models.
"""
return self.mgr.list_models(base_model, model_type)
+ def list_model(self, model_name: str, base_model: BaseModelType, model_type: ModelType) -> dict:
+ """
+ Return information about the model using the same format as list_models()
+ """
+ return self.mgr.list_model(model_name=model_name,
+ base_model=base_model,
+ model_type=model_type)
+
def add_model(
self,
model_name: str,
@@ -291,13 +396,32 @@ class ModelManagerService(ModelManagerServiceBase):
"""
Update the named model with a dictionary of attributes. Will fail with an
assertion error if the name already exists. Pass clobber=True to overwrite.
+ On a successful update, the config will be changed in memory. Will fail
+ with an assertion error if provided attributes are incorrect or
+ the model name is missing. Call commit() to write changes to disk.
+ """
+ self.logger.debug(f'add/update model {model_name}')
+ return self.mgr.add_model(model_name, base_model, model_type, model_attributes, clobber)
+
+ def update_model(
+ self,
+ model_name: str,
+ base_model: BaseModelType,
+ model_type: ModelType,
+ model_attributes: dict,
+ ) -> AddModelResult:
+ """
+ Update the named model with a dictionary of attributes. Will fail with a
+ KeyError exception if the name does not already exist.
On a successful update, the config will be changed in memory. Will fail
with an assertion error if provided attributes are incorrect or
the model name is missing. Call commit() to write changes to disk.
"""
- return self.mgr.add_model(model_name, base_model, model_type, model_attributes, clobber)
-
-
+ self.logger.debug(f'update model {model_name}')
+ if not self.model_exists(model_name, base_model, model_type):
+ raise KeyError(f"Unknown model {model_name}")
+ return self.add_model(model_name, base_model, model_type, model_attributes, clobber=True)
+
def del_model(
self,
model_name: str,
@@ -305,12 +429,33 @@ class ModelManagerService(ModelManagerServiceBase):
model_type: ModelType,
):
"""
- Delete the named model from configuration. If delete_files is true,
- then the underlying weight file or diffusers directory will be deleted
+ Delete the named model from configuration. If delete_files is true,
+ then the underlying weight file or diffusers directory will be deleted
as well. Call commit() to write to disk.
"""
+ self.logger.debug(f'delete model {model_name}')
self.mgr.del_model(model_name, base_model, model_type)
+ def convert_model(
+ self,
+ model_name: str,
+ base_model: BaseModelType,
+ model_type: Union[ModelType.Main,ModelType.Vae],
+ ) -> AddModelResult:
+ """
+ Convert a checkpoint file into a diffusers folder, deleting the cached
+ version and deleting the original checkpoint file if it is in the models
+ directory.
+ :param model_name: Name of the model to convert
+ :param base_model: Base model type
+ :param model_type: Type of model ['vae' or 'main']
+
+ This will raise a ValueError unless the model is not a checkpoint. It will
+ also raise a ValueError in the event that there is a similarly-named diffusers
+ directory already in place.
+ """
+ self.logger.debug(f'convert model {model_name}')
+ return self.mgr.convert_model(model_name, base_model, model_type)
def commit(self, conf_file: Optional[Path]=None):
"""
@@ -360,4 +505,56 @@ class ModelManagerService(ModelManagerServiceBase):
@property
def logger(self):
return self.mgr.logger
-
+
+ def heuristic_import(self,
+ items_to_import: set[str],
+ prediction_type_helper: Optional[Callable[[Path],SchedulerPredictionType]]=None,
+ )->dict[str, AddModelResult]:
+ '''Import a list of paths, repo_ids or URLs. Returns the set of
+ successfully imported items.
+ :param items_to_import: Set of strings corresponding to models to be imported.
+ :param prediction_type_helper: A callback that receives the Path of a Stable Diffusion 2 checkpoint model and returns a SchedulerPredictionType.
+
+ The prediction type helper is necessary to distinguish between
+ models based on Stable Diffusion 2 Base (requiring
+ SchedulerPredictionType.Epsilson) and Stable Diffusion 768
+ (requiring SchedulerPredictionType.VPrediction). It is
+ generally impossible to do this programmatically, so the
+ prediction_type_helper usually asks the user to choose.
+
+ The result is a set of successfully installed models. Each element
+ of the set is a dict corresponding to the newly-created OmegaConf stanza for
+ that model.
+ '''
+ return self.mgr.heuristic_import(items_to_import, prediction_type_helper)
+
+ def merge_models(
+ self,
+ model_names: List[str] = Field(default=None, min_items=2, max_items=3, description="List of model names to merge"),
+ base_model: Union[BaseModelType,str] = Field(default=None, description="Base model shared by all models to be merged"),
+ merged_model_name: str = Field(default=None, description="Name of destination model after merging"),
+ alpha: Optional[float] = 0.5,
+ interp: Optional[MergeInterpolationMethod] = None,
+ force: Optional[bool] = False,
+ ) -> AddModelResult:
+ """
+ Merge two to three diffusrs pipeline models and save as a new model.
+ :param model_names: List of 2-3 models to merge
+ :param base_model: Base model to use for all models
+ :param merged_model_name: Name of destination merged model
+ :param alpha: Alpha strength to apply to 2d and 3d model
+ :param interp: Interpolation method. None (default)
+ """
+ merger = ModelMerger(self.mgr)
+ try:
+ result = merger.merge_diffusion_models_and_save(
+ model_names = model_names,
+ base_model = base_model,
+ merged_model_name = merged_model_name,
+ alpha = alpha,
+ interp = interp,
+ force = force,
+ )
+ except AssertionError as e:
+ raise ValueError(e)
+ return result
diff --git a/invokeai/app/services/models/image_record.py b/invokeai/app/services/models/image_record.py
index cc02016cf9..cf10f6e8b2 100644
--- a/invokeai/app/services/models/image_record.py
+++ b/invokeai/app/services/models/image_record.py
@@ -1,13 +1,14 @@
import datetime
from typing import Optional, Union
+
from pydantic import BaseModel, Extra, Field, StrictBool, StrictStr
+
from invokeai.app.models.image import ImageCategory, ResourceOrigin
-from invokeai.app.models.metadata import ImageMetadata
from invokeai.app.util.misc import get_iso_timestamp
class ImageRecord(BaseModel):
- """Deserialized image record."""
+ """Deserialized image record without metadata."""
image_name: str = Field(description="The unique name of the image.")
"""The unique name of the image."""
@@ -43,11 +44,6 @@ class ImageRecord(BaseModel):
description="The node ID that generated this image, if it is a generated image.",
)
"""The node ID that generated this image, if it is a generated image."""
- metadata: Optional[ImageMetadata] = Field(
- default=None,
- description="A limited subset of the image's generation metadata. Retrieve the image's session for full metadata.",
- )
- """A limited subset of the image's generation metadata. Retrieve the image's session for full metadata."""
class ImageRecordChanges(BaseModel, extra=Extra.forbid):
@@ -88,7 +84,7 @@ class ImageUrlsDTO(BaseModel):
class ImageDTO(ImageRecord, ImageUrlsDTO):
"""Deserialized image record, enriched for the frontend."""
- board_id: Union[str, None] = Field(
+ board_id: Optional[str] = Field(
description="The id of the board the image belongs to, if one exists."
)
"""The id of the board the image belongs to, if one exists."""
@@ -96,7 +92,7 @@ class ImageDTO(ImageRecord, ImageUrlsDTO):
def image_record_to_dto(
- image_record: ImageRecord, image_url: str, thumbnail_url: str, board_id: Union[str, None]
+ image_record: ImageRecord, image_url: str, thumbnail_url: str, board_id: Optional[str]
) -> ImageDTO:
"""Converts an image record to an image DTO."""
return ImageDTO(
@@ -112,6 +108,7 @@ def deserialize_image_record(image_dict: dict) -> ImageRecord:
# Retrieve all the values, setting "reasonable" defaults if they are not present.
+ # TODO: do we really need to handle default values here? ideally the data is the correct shape...
image_name = image_dict.get("image_name", "unknown")
image_origin = ResourceOrigin(
image_dict.get("image_origin", ResourceOrigin.INTERNAL.value)
@@ -128,13 +125,6 @@ def deserialize_image_record(image_dict: dict) -> ImageRecord:
deleted_at = image_dict.get("deleted_at", get_iso_timestamp())
is_intermediate = image_dict.get("is_intermediate", False)
- raw_metadata = image_dict.get("metadata")
-
- if raw_metadata is not None:
- metadata = ImageMetadata.parse_raw(raw_metadata)
- else:
- metadata = None
-
return ImageRecord(
image_name=image_name,
image_origin=image_origin,
@@ -143,7 +133,6 @@ def deserialize_image_record(image_dict: dict) -> ImageRecord:
height=height,
session_id=session_id,
node_id=node_id,
- metadata=metadata,
created_at=created_at,
updated_at=updated_at,
deleted_at=deleted_at,
diff --git a/invokeai/app/services/processor.py b/invokeai/app/services/processor.py
index 9e3b5a0a30..e11eb84b3d 100644
--- a/invokeai/app/services/processor.py
+++ b/invokeai/app/services/processor.py
@@ -104,6 +104,7 @@ class DefaultInvocationProcessor(InvocationProcessorABC):
except Exception as e:
error = traceback.format_exc()
+ logger.error(error)
# Save error
graph_execution_state.set_node_error(invocation.id, error)
diff --git a/invokeai/app/services/sqlite.py b/invokeai/app/services/sqlite.py
index a62fff88a1..8902415096 100644
--- a/invokeai/app/services/sqlite.py
+++ b/invokeai/app/services/sqlite.py
@@ -1,6 +1,6 @@
import sqlite3
from threading import Lock
-from typing import Generic, TypeVar, Union, get_args
+from typing import Generic, Optional, TypeVar, get_args
from pydantic import BaseModel, parse_raw_as
@@ -63,7 +63,7 @@ class SqliteItemStorage(ItemStorageABC, Generic[T]):
self._lock.release()
self._on_changed(item)
- def get(self, id: str) -> Union[T, None]:
+ def get(self, id: str) -> Optional[T]:
try:
self._lock.acquire()
self._cursor.execute(
@@ -78,6 +78,21 @@ class SqliteItemStorage(ItemStorageABC, Generic[T]):
return self._parse_item(result[0])
+ def get_raw(self, id: str) -> Optional[str]:
+ try:
+ self._lock.acquire()
+ self._cursor.execute(
+ f"""SELECT item FROM {self._table_name} WHERE id = ?;""", (str(id),)
+ )
+ result = self._cursor.fetchone()
+ finally:
+ self._lock.release()
+
+ if not result:
+ return None
+
+ return result[0]
+
def delete(self, id: str):
try:
self._lock.acquire()
diff --git a/invokeai/app/services/urls.py b/invokeai/app/services/urls.py
index 5920e9e6c1..73d8ddadf4 100644
--- a/invokeai/app/services/urls.py
+++ b/invokeai/app/services/urls.py
@@ -22,4 +22,4 @@ class LocalUrlService(UrlServiceBase):
if thumbnail:
return f"{self._base_url}/images/{image_basename}/thumbnail"
- return f"{self._base_url}/images/{image_basename}"
+ return f"{self._base_url}/images/{image_basename}/full"
diff --git a/invokeai/app/util/metadata.py b/invokeai/app/util/metadata.py
new file mode 100644
index 0000000000..5ca5f14e12
--- /dev/null
+++ b/invokeai/app/util/metadata.py
@@ -0,0 +1,55 @@
+import json
+from typing import Optional
+
+from pydantic import ValidationError
+
+from invokeai.app.services.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 (
+ type(graph) is not dict
+ or "nodes" not in graph
+ or type(graph["nodes"]) is not dict
+ or "edges" not in graph
+ or type(graph["edges"]) is not 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
diff --git a/invokeai/backend/generator/base.py b/invokeai/backend/generator/base.py
index 462b1a4f4b..99a63d56a3 100644
--- a/invokeai/backend/generator/base.py
+++ b/invokeai/backend/generator/base.py
@@ -21,7 +21,7 @@ from PIL import Image, ImageChops, ImageFilter
from accelerate.utils import set_seed
from diffusers import DiffusionPipeline
from tqdm import trange
-from typing import Callable, List, Iterator, Optional, Type
+from typing import Callable, List, Iterator, Optional, Type, Union
from dataclasses import dataclass, field
from diffusers.schedulers import SchedulerMixin as Scheduler
@@ -178,7 +178,7 @@ class InvokeAIGenerator(metaclass=ABCMeta):
# ------------------------------------
class Img2Img(InvokeAIGenerator):
def generate(self,
- init_image: Image.Image | torch.FloatTensor,
+ init_image: Union[Image.Image, torch.FloatTensor],
strength: float=0.75,
**keyword_args
)->Iterator[InvokeAIGeneratorOutput]:
@@ -195,7 +195,7 @@ class Img2Img(InvokeAIGenerator):
# Takes all the arguments of Img2Img and adds the mask image and the seam/infill stuff
class Inpaint(Img2Img):
def generate(self,
- mask_image: Image.Image | torch.FloatTensor,
+ mask_image: Union[Image.Image, torch.FloatTensor],
# Seam settings - when 0, doesn't fill seam
seam_size: int = 96,
seam_blur: int = 16,
@@ -570,28 +570,16 @@ class Generator:
device = self.model.device
# limit noise to only the diffusion image channels, not the mask channels
input_channels = min(self.latent_channels, 4)
- if self.use_mps_noise or device.type == "mps":
- x = torch.randn(
- [
- 1,
- input_channels,
- height // self.downsampling_factor,
- width // self.downsampling_factor,
- ],
- dtype=self.torch_dtype(),
- device="cpu",
- ).to(device)
- else:
- x = torch.randn(
- [
- 1,
- input_channels,
- height // self.downsampling_factor,
- width // self.downsampling_factor,
- ],
- dtype=self.torch_dtype(),
- device=device,
- )
+ x = torch.randn(
+ [
+ 1,
+ input_channels,
+ height // self.downsampling_factor,
+ width // self.downsampling_factor,
+ ],
+ dtype=self.torch_dtype(),
+ device=device,
+ )
if self.perlin > 0.0:
perlin_noise = self.get_perlin_noise(
width // self.downsampling_factor, height // self.downsampling_factor
diff --git a/invokeai/backend/generator/img2img.py b/invokeai/backend/generator/img2img.py
index 1cfbeb66c0..b3b0e8f510 100644
--- a/invokeai/backend/generator/img2img.py
+++ b/invokeai/backend/generator/img2img.py
@@ -88,10 +88,7 @@ class Img2Img(Generator):
def get_noise_like(self, like: torch.Tensor):
device = like.device
- if device.type == "mps":
- x = torch.randn_like(like, device="cpu").to(device)
- else:
- x = torch.randn_like(like, device=device)
+ x = torch.randn_like(like, device=device)
if self.perlin > 0.0:
shape = like.shape
x = (1 - self.perlin) * x + self.perlin * self.get_perlin_noise(
diff --git a/invokeai/backend/generator/inpaint.py b/invokeai/backend/generator/inpaint.py
index eaf4047109..c91fe0c6a7 100644
--- a/invokeai/backend/generator/inpaint.py
+++ b/invokeai/backend/generator/inpaint.py
@@ -4,11 +4,10 @@ invokeai.backend.generator.inpaint descends from .generator
from __future__ import annotations
import math
-from typing import Tuple, Union
+from typing import Tuple, Union, Optional
import cv2
import numpy as np
-import PIL
import torch
from PIL import Image, ImageChops, ImageFilter, ImageOps
@@ -76,7 +75,7 @@ class Inpaint(Img2Img):
return im_patched
def tile_fill_missing(
- self, im: Image.Image, tile_size: int = 16, seed: Union[int, None] = None
+ self, im: Image.Image, tile_size: int = 16, seed: Optional[int] = None
) -> Image.Image:
# Only fill if there's an alpha layer
if im.mode != "RGBA":
@@ -203,8 +202,8 @@ class Inpaint(Img2Img):
cfg_scale,
ddim_eta,
conditioning,
- init_image: Image.Image | torch.FloatTensor,
- mask_image: Image.Image | torch.FloatTensor,
+ init_image: Union[Image.Image, torch.FloatTensor],
+ mask_image: Union[Image.Image, torch.FloatTensor],
strength: float,
mask_blur_radius: int = 8,
# Seam settings - when 0, doesn't fill seam
diff --git a/invokeai/backend/install/invokeai_configure.py b/invokeai/backend/install/invokeai_configure.py
index 603760c0c1..adc985c655 100755
--- a/invokeai/backend/install/invokeai_configure.py
+++ b/invokeai/backend/install/invokeai_configure.py
@@ -7,8 +7,6 @@
# Coauthor: Kevin Turner http://github.com/keturn
#
import sys
-print("Loading Python libraries...\n",file=sys.stderr)
-
import argparse
import io
import os
@@ -16,6 +14,7 @@ import shutil
import textwrap
import traceback
import warnings
+import yaml
from argparse import Namespace
from pathlib import Path
from shutil import get_terminal_size
@@ -25,6 +24,7 @@ from urllib import request
import npyscreen
import transformers
from diffusers import AutoencoderKL
+from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
from huggingface_hub import HfFolder
from huggingface_hub import login as hf_hub_login
from omegaconf import OmegaConf
@@ -34,6 +34,8 @@ from transformers import (
CLIPSegForImageSegmentation,
CLIPTextModel,
CLIPTokenizer,
+ AutoFeatureExtractor,
+ BertTokenizerFast,
)
import invokeai.configs as configs
@@ -43,6 +45,7 @@ from invokeai.app.services.config import (
from invokeai.backend.util.logging import InvokeAILogger
from invokeai.frontend.install.model_install import addModelsForm, process_and_execute
from invokeai.frontend.install.widgets import (
+ SingleSelectColumns,
CenteredButtonPress,
IntTitleSlider,
set_min_terminal_size,
@@ -52,12 +55,13 @@ from invokeai.frontend.install.widgets import (
)
from invokeai.backend.install.legacy_arg_parsing import legacy_parser
from invokeai.backend.install.model_install_backend import (
- default_dataset,
- download_from_hf,
- hf_download_with_resume,
- recommended_datasets,
- UserSelections,
+ hf_download_from_pretrained,
+ InstallSelections,
+ ModelInstall,
)
+from invokeai.backend.model_management.model_probe import (
+ ModelType, BaseModelType
+ )
warnings.filterwarnings("ignore")
transformers.logging.set_verbosity_error()
@@ -73,7 +77,7 @@ Weights_dir = "ldm/stable-diffusion-v1/"
Default_config_file = config.model_conf_path
SD_Configs = config.legacy_conf_path
-PRECISION_CHOICES = ['auto','float16','float32','autocast']
+PRECISION_CHOICES = ['auto','float16','float32']
INIT_FILE_PREAMBLE = """# InvokeAI initialization file
# This is the InvokeAI initialization file, which contains command-line default values.
@@ -81,7 +85,7 @@ INIT_FILE_PREAMBLE = """# InvokeAI initialization file
# or renaming it and then running invokeai-configure again.
"""
-logger=None
+logger=InvokeAILogger.getLogger()
# --------------------------------------------
def postscript(errors: None):
@@ -162,75 +166,91 @@ class ProgressBar:
# ---------------------------------------------
def download_with_progress_bar(model_url: str, model_dest: str, label: str = "the"):
try:
- print(f"Installing {label} model file {model_url}...", end="", file=sys.stderr)
+ logger.info(f"Installing {label} model file {model_url}...")
if not os.path.exists(model_dest):
os.makedirs(os.path.dirname(model_dest), exist_ok=True)
request.urlretrieve(
model_url, model_dest, ProgressBar(os.path.basename(model_dest))
)
- print("...downloaded successfully", file=sys.stderr)
+ logger.info("...downloaded successfully")
else:
- print("...exists", file=sys.stderr)
+ logger.info("...exists")
except Exception:
- print("...download failed", file=sys.stderr)
- print(f"Error downloading {label} model", file=sys.stderr)
+ logger.info("...download failed")
+ logger.info(f"Error downloading {label} model")
print(traceback.format_exc(), file=sys.stderr)
-# ---------------------------------------------
-# this will preload the Bert tokenizer fles
-def download_bert():
- print("Installing bert tokenizer...", file=sys.stderr)
- with warnings.catch_warnings():
- warnings.filterwarnings("ignore", category=DeprecationWarning)
- from transformers import BertTokenizerFast
+def download_conversion_models():
+ target_dir = config.root_path / 'models/core/convert'
+ kwargs = dict() # for future use
+ try:
+ logger.info('Downloading core tokenizers and text encoders')
- download_from_hf(BertTokenizerFast, "bert-base-uncased")
+ # bert
+ with warnings.catch_warnings():
+ warnings.filterwarnings("ignore", category=DeprecationWarning)
+ bert = BertTokenizerFast.from_pretrained("bert-base-uncased", **kwargs)
+ bert.save_pretrained(target_dir / 'bert-base-uncased', safe_serialization=True)
+
+ # sd-1
+ repo_id = 'openai/clip-vit-large-patch14'
+ hf_download_from_pretrained(CLIPTokenizer, repo_id, target_dir / 'clip-vit-large-patch14')
+ hf_download_from_pretrained(CLIPTextModel, repo_id, target_dir / 'clip-vit-large-patch14')
+ # sd-2
+ repo_id = "stabilityai/stable-diffusion-2"
+ pipeline = CLIPTokenizer.from_pretrained(repo_id, subfolder="tokenizer", **kwargs)
+ pipeline.save_pretrained(target_dir / 'stable-diffusion-2-clip' / 'tokenizer', safe_serialization=True)
-# ---------------------------------------------
-def download_sd1_clip():
- print("Installing SD1 clip model...", file=sys.stderr)
- version = "openai/clip-vit-large-patch14"
- download_from_hf(CLIPTokenizer, version)
- download_from_hf(CLIPTextModel, version)
+ pipeline = CLIPTextModel.from_pretrained(repo_id, subfolder="text_encoder", **kwargs)
+ pipeline.save_pretrained(target_dir / 'stable-diffusion-2-clip' / 'text_encoder', safe_serialization=True)
+ # VAE
+ logger.info('Downloading stable diffusion VAE')
+ vae = AutoencoderKL.from_pretrained('stabilityai/sd-vae-ft-mse', **kwargs)
+ vae.save_pretrained(target_dir / 'sd-vae-ft-mse', safe_serialization=True)
-# ---------------------------------------------
-def download_sd2_clip():
- version = "stabilityai/stable-diffusion-2"
- print("Installing SD2 clip model...", file=sys.stderr)
- download_from_hf(CLIPTokenizer, version, subfolder="tokenizer")
- download_from_hf(CLIPTextModel, version, subfolder="text_encoder")
+ # safety checking
+ logger.info('Downloading safety checker')
+ repo_id = "CompVis/stable-diffusion-safety-checker"
+ pipeline = AutoFeatureExtractor.from_pretrained(repo_id,**kwargs)
+ pipeline.save_pretrained(target_dir / 'stable-diffusion-safety-checker', safe_serialization=True)
+ pipeline = StableDiffusionSafetyChecker.from_pretrained(repo_id,**kwargs)
+ pipeline.save_pretrained(target_dir / 'stable-diffusion-safety-checker', safe_serialization=True)
+ except KeyboardInterrupt:
+ raise
+ except Exception as e:
+ logger.error(str(e))
# ---------------------------------------------
def download_realesrgan():
- print("Installing models from RealESRGAN...", file=sys.stderr)
+ logger.info("Installing models from RealESRGAN...")
model_url = "https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-x4v3.pth"
wdn_model_url = "https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-wdn-x4v3.pth"
- model_dest = config.root_path / "models/realesrgan/realesr-general-x4v3.pth"
- wdn_model_dest = config.root_path / "models/realesrgan/realesr-general-wdn-x4v3.pth"
+ model_dest = config.root_path / "models/core/upscaling/realesrgan/realesr-general-x4v3.pth"
+ wdn_model_dest = config.root_path / "models/core/upscaling/realesrgan/realesr-general-wdn-x4v3.pth"
download_with_progress_bar(model_url, str(model_dest), "RealESRGAN")
download_with_progress_bar(wdn_model_url, str(wdn_model_dest), "RealESRGANwdn")
def download_gfpgan():
- print("Installing GFPGAN models...", file=sys.stderr)
+ logger.info("Installing GFPGAN models...")
for model in (
[
"https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.4.pth",
- "./models/gfpgan/GFPGANv1.4.pth",
+ "./models/core/face_restoration/gfpgan/GFPGANv1.4.pth",
],
[
"https://github.com/xinntao/facexlib/releases/download/v0.1.0/detection_Resnet50_Final.pth",
- "./models/gfpgan/weights/detection_Resnet50_Final.pth",
+ "./models/core/face_restoration/gfpgan/weights/detection_Resnet50_Final.pth",
],
[
"https://github.com/xinntao/facexlib/releases/download/v0.2.2/parsing_parsenet.pth",
- "./models/gfpgan/weights/parsing_parsenet.pth",
+ "./models/core/face_restoration/gfpgan/weights/parsing_parsenet.pth",
],
):
model_url, model_dest = model[0], config.root_path / model[1]
@@ -239,70 +259,32 @@ def download_gfpgan():
# ---------------------------------------------
def download_codeformer():
- print("Installing CodeFormer model file...", file=sys.stderr)
+ logger.info("Installing CodeFormer model file...")
model_url = (
"https://github.com/sczhou/CodeFormer/releases/download/v0.1.0/codeformer.pth"
)
- model_dest = config.root_path / "models/codeformer/codeformer.pth"
+ model_dest = config.root_path / "models/core/face_restoration/codeformer/codeformer.pth"
download_with_progress_bar(model_url, str(model_dest), "CodeFormer")
# ---------------------------------------------
def download_clipseg():
- print("Installing clipseg model for text-based masking...", file=sys.stderr)
+ logger.info("Installing clipseg model for text-based masking...")
CLIPSEG_MODEL = "CIDAS/clipseg-rd64-refined"
try:
- download_from_hf(AutoProcessor, CLIPSEG_MODEL)
- download_from_hf(CLIPSegForImageSegmentation, CLIPSEG_MODEL)
+ hf_download_from_pretrained(AutoProcessor, CLIPSEG_MODEL, config.root_path / 'models/core/misc/clipseg')
+ hf_download_from_pretrained(CLIPSegForImageSegmentation, CLIPSEG_MODEL, config.root_path / 'models/core/misc/clipseg')
except Exception:
- print("Error installing clipseg model:")
- print(traceback.format_exc())
+ logger.info("Error installing clipseg model:")
+ logger.info(traceback.format_exc())
-# -------------------------------------
-def download_safety_checker():
- print("Installing model for NSFW content detection...", file=sys.stderr)
- try:
- from diffusers.pipelines.stable_diffusion.safety_checker import (
- StableDiffusionSafetyChecker,
- )
- from transformers import AutoFeatureExtractor
- except ModuleNotFoundError:
- print("Error installing NSFW checker model:")
- print(traceback.format_exc())
- return
- safety_model_id = "CompVis/stable-diffusion-safety-checker"
- print("AutoFeatureExtractor...", file=sys.stderr)
- download_from_hf(AutoFeatureExtractor, safety_model_id)
- print("StableDiffusionSafetyChecker...", file=sys.stderr)
- download_from_hf(StableDiffusionSafetyChecker, safety_model_id)
-
-
-# -------------------------------------
-def download_vaes():
- print("Installing stabilityai VAE...", file=sys.stderr)
- try:
- # first the diffusers version
- repo_id = "stabilityai/sd-vae-ft-mse"
- args = dict(
- cache_dir=config.cache_dir,
- )
- if not AutoencoderKL.from_pretrained(repo_id, **args):
- raise Exception(f"download of {repo_id} failed")
-
- repo_id = "stabilityai/sd-vae-ft-mse-original"
- model_name = "vae-ft-mse-840000-ema-pruned.ckpt"
- # next the legacy checkpoint version
- if not hf_download_with_resume(
- repo_id=repo_id,
- model_name=model_name,
- model_dir=str(config.root_path / Model_dir / Weights_dir),
- ):
- raise Exception(f"download of {model_name} failed")
- except Exception as e:
- print(f"Error downloading StabilityAI standard VAE: {str(e)}", file=sys.stderr)
- print(traceback.format_exc(), file=sys.stderr)
-
+def download_support_models():
+ download_realesrgan()
+ download_gfpgan()
+ download_codeformer()
+ download_clipseg()
+ download_conversion_models()
# -------------------------------------
def get_root(root: str = None) -> str:
@@ -378,9 +360,7 @@ Use cursor arrows to make a checkbox selection, and space to toggle.
scroll_exit=True,
)
self.nextrely += 1
- label = """If you have an account at HuggingFace you may optionally paste your access token here
-to allow InvokeAI to download restricted styles & subjects from the "Concept Library". See https://huggingface.co/settings/tokens.
-"""
+ label = """HuggingFace access token (OPTIONAL) for automatic model downloads. See https://huggingface.co/settings/tokens."""
for line in textwrap.wrap(label,width=window_width-6):
self.add_widget_intelligent(
npyscreen.FixedText,
@@ -442,6 +422,7 @@ to allow InvokeAI to download restricted styles & subjects from the "Concept Lib
)
self.precision = self.add_widget_intelligent(
npyscreen.TitleSelectOne,
+ columns = 2,
name="Precision",
values=PRECISION_CHOICES,
value=PRECISION_CHOICES.index(precision),
@@ -449,13 +430,13 @@ to allow InvokeAI to download restricted styles & subjects from the "Concept Lib
max_height=len(PRECISION_CHOICES) + 1,
scroll_exit=True,
)
- self.max_loaded_models = self.add_widget_intelligent(
+ self.max_cache_size = self.add_widget_intelligent(
IntTitleSlider,
- name="Number of models to cache in CPU memory (each will use 2-4 GB!)",
- value=old_opts.max_loaded_models,
- out_of=10,
- lowest=1,
- begin_entry_at=4,
+ name="Size of the RAM cache used for fast model switching (GB)",
+ value=old_opts.max_cache_size,
+ out_of=20,
+ lowest=3,
+ begin_entry_at=6,
scroll_exit=True,
)
self.nextrely += 1
@@ -465,39 +446,19 @@ to allow InvokeAI to download restricted styles & subjects from the "Concept Lib
editable=False,
color="CONTROL",
)
- self.embedding_dir = self.add_widget_intelligent(
- npyscreen.TitleFilename,
- name=" Textual Inversion Embeddings:",
- value=str(default_embedding_dir()),
- select_dir=True,
- must_exist=False,
- use_two_lines=False,
- labelColor="GOOD",
- begin_entry_at=32,
- scroll_exit=True,
- )
- self.lora_dir = self.add_widget_intelligent(
- npyscreen.TitleFilename,
- name=" LoRA and LyCORIS:",
- value=str(default_lora_dir()),
- select_dir=True,
- must_exist=False,
- use_two_lines=False,
- labelColor="GOOD",
- begin_entry_at=32,
- scroll_exit=True,
- )
- self.controlnet_dir = self.add_widget_intelligent(
- npyscreen.TitleFilename,
- name=" ControlNets:",
- value=str(default_controlnet_dir()),
- select_dir=True,
- must_exist=False,
- use_two_lines=False,
- labelColor="GOOD",
- begin_entry_at=32,
- scroll_exit=True,
- )
+ self.autoimport_dirs = {}
+ for description, config_name, path in autoimport_paths(old_opts):
+ self.autoimport_dirs[config_name] = self.add_widget_intelligent(
+ npyscreen.TitleFilename,
+ name=description+':',
+ value=str(path),
+ select_dir=True,
+ must_exist=False,
+ use_two_lines=False,
+ labelColor="GOOD",
+ begin_entry_at=32,
+ scroll_exit=True
+ )
self.nextrely += 1
self.add_widget_intelligent(
npyscreen.TitleFixedText,
@@ -562,10 +523,6 @@ https://huggingface.co/spaces/CompVis/stable-diffusion-license
bad_fields.append(
f"The output directory does not seem to be valid. Please check that {str(Path(opt.outdir).parent)} is an existing directory."
)
- if not Path(opt.embedding_dir).parent.exists():
- bad_fields.append(
- f"The embedding directory does not seem to be valid. Please check that {str(Path(opt.embedding_dir).parent)} is an existing directory."
- )
if len(bad_fields) > 0:
message = "The following problems were detected and must be corrected:\n"
for problem in bad_fields:
@@ -582,22 +539,22 @@ https://huggingface.co/spaces/CompVis/stable-diffusion-license
"outdir",
"nsfw_checker",
"free_gpu_mem",
- "max_loaded_models",
+ "max_cache_size",
"xformers_enabled",
"always_use_cpu",
- "embedding_dir",
- "lora_dir",
- "controlnet_dir",
]:
setattr(new_opts, attr, getattr(self, attr).value)
+ for attr in self.autoimport_dirs:
+ directory = Path(self.autoimport_dirs[attr].value)
+ if directory.is_relative_to(config.root_path):
+ directory = directory.relative_to(config.root_path)
+ setattr(new_opts, attr, directory)
+
new_opts.hf_token = self.hf_token.value
new_opts.license_acceptance = self.license_acceptance.value
new_opts.precision = PRECISION_CHOICES[self.precision.value[0]]
- # widget library workaround to make max_loaded_models an int rather than a float
- new_opts.max_loaded_models = int(new_opts.max_loaded_models)
-
return new_opts
@@ -607,7 +564,8 @@ class EditOptApplication(npyscreen.NPSAppManaged):
self.program_opts = program_opts
self.invokeai_opts = invokeai_opts
self.user_cancelled = False
- self.user_selections = default_user_selections(program_opts)
+ self.autoload_pending = True
+ self.install_selections = default_user_selections(program_opts)
def onStart(self):
npyscreen.setTheme(npyscreen.Themes.DefaultTheme)
@@ -642,41 +600,62 @@ def default_startup_options(init_file: Path) -> Namespace:
opts.nsfw_checker = True
return opts
-def default_user_selections(program_opts: Namespace) -> UserSelections:
- return UserSelections(
- install_models=default_dataset()
+def default_user_selections(program_opts: Namespace) -> InstallSelections:
+ installer = ModelInstall(config)
+ models = installer.all_models()
+ return InstallSelections(
+ install_models=[models[installer.default_model()].path or models[installer.default_model()].repo_id]
if program_opts.default_only
- else recommended_datasets()
+ else [models[x].path or models[x].repo_id for x in installer.recommended_models()]
if program_opts.yes_to_all
- else dict(),
- purge_deleted_models=False,
- scan_directory=None,
- autoscan_on_startup=None,
+ else list(),
+# scan_directory=None,
+# autoscan_on_startup=None,
)
-
+# -------------------------------------
+def autoimport_paths(config: InvokeAIAppConfig):
+ return [
+ ('Checkpoints & diffusers models', 'autoimport_dir', config.root_path / config.autoimport_dir),
+ ('LoRA/LyCORIS models', 'lora_dir', config.root_path / config.lora_dir),
+ ('Controlnet models', 'controlnet_dir', config.root_path / config.controlnet_dir),
+ ('Textual Inversion Embeddings', 'embedding_dir', config.root_path / config.embedding_dir),
+ ]
+
# -------------------------------------
def initialize_rootdir(root: Path, yes_to_all: bool = False):
- print("** INITIALIZING INVOKEAI RUNTIME DIRECTORY **")
-
+ logger.info("** INITIALIZING INVOKEAI RUNTIME DIRECTORY **")
for name in (
"models",
- "configs",
- "embeddings",
"databases",
- "loras",
- "controlnets",
"text-inversion-output",
"text-inversion-training-data",
+ "configs"
):
os.makedirs(os.path.join(root, name), exist_ok=True)
+ for model_type in ModelType:
+ Path(root, 'autoimport', model_type.value).mkdir(parents=True, exist_ok=True)
configs_src = Path(configs.__path__[0])
configs_dest = root / "configs"
if not os.path.samefile(configs_src, configs_dest):
shutil.copytree(configs_src, configs_dest, dirs_exist_ok=True)
+ dest = root / 'models'
+ for model_base in BaseModelType:
+ for model_type in ModelType:
+ path = dest / model_base.value / model_type.value
+ path.mkdir(parents=True, exist_ok=True)
+ path = dest / 'core'
+ path.mkdir(parents=True, exist_ok=True)
+ with open(root / 'configs' / 'models.yaml','w') as yaml_file:
+ yaml_file.write(yaml.dump({'__metadata__':
+ {'version':'3.0.0'}
+ }
+ )
+ )
+
# -------------------------------------
def run_console_ui(
program_opts: Namespace, initfile: Path = None
@@ -699,7 +678,7 @@ def run_console_ui(
if editApp.user_cancelled:
return (None, None)
else:
- return (editApp.new_opts, editApp.user_selections)
+ return (editApp.new_opts, editApp.install_selections)
# -------------------------------------
@@ -722,18 +701,6 @@ def write_opts(opts: Namespace, init_file: Path):
def default_output_dir() -> Path:
return config.root_path / "outputs"
-# -------------------------------------
-def default_embedding_dir() -> Path:
- return config.root_path / "embeddings"
-
-# -------------------------------------
-def default_lora_dir() -> Path:
- return config.root_path / "loras"
-
-# -------------------------------------
-def default_controlnet_dir() -> Path:
- return config.root_path / "controlnets"
-
# -------------------------------------
def write_default_options(program_opts: Namespace, initfile: Path):
opt = default_startup_options(initfile)
@@ -758,14 +725,42 @@ def migrate_init_file(legacy_format:Path):
new.nsfw_checker = old.safety_checker
new.xformers_enabled = old.xformers
new.conf_path = old.conf
- new.embedding_dir = old.embedding_path
+ new.root = legacy_format.parent.resolve()
invokeai_yaml = legacy_format.parent / 'invokeai.yaml'
with open(invokeai_yaml,"w", encoding="utf-8") as outfile:
outfile.write(new.to_yaml())
- legacy_format.replace(legacy_format.parent / 'invokeai.init.old')
+ legacy_format.replace(legacy_format.parent / 'invokeai.init.orig')
+# -------------------------------------
+def migrate_models(root: Path):
+ from invokeai.backend.install.migrate_to_3 import do_migrate
+ do_migrate(root, root)
+
+def migrate_if_needed(opt: Namespace, root: Path)->bool:
+ # We check for to see if the runtime directory is correctly initialized.
+ old_init_file = root / 'invokeai.init'
+ new_init_file = root / 'invokeai.yaml'
+ old_hub = root / 'models/hub'
+ migration_needed = old_init_file.exists() and not new_init_file.exists() or old_hub.exists()
+
+ if migration_needed:
+ if opt.yes_to_all or \
+ yes_or_no(f'{str(config.root_path)} appears to be a 2.3 format root directory. Convert to version 3.0?'):
+
+ logger.info('** Migrating invokeai.init to invokeai.yaml')
+ migrate_init_file(old_init_file)
+ config.parse_args(argv=[],conf=OmegaConf.load(new_init_file))
+
+ if old_hub.exists():
+ migrate_models(config.root_path)
+ else:
+ print('Cannot continue without conversion. Aborting.')
+
+ return migration_needed
+
+
# -------------------------------------
def main():
parser = argparse.ArgumentParser(description="InvokeAI model downloader")
@@ -831,20 +826,16 @@ def main():
errors = set()
try:
- models_to_download = default_user_selections(opt)
-
- # We check for to see if the runtime directory is correctly initialized.
- old_init_file = config.root_path / 'invokeai.init'
- new_init_file = config.root_path / 'invokeai.yaml'
- if old_init_file.exists() and not new_init_file.exists():
- print('** Migrating invokeai.init to invokeai.yaml')
- migrate_init_file(old_init_file)
- # Load new init file into config
- config.parse_args(argv=[],conf=OmegaConf.load(new_init_file))
+ # if we do a root migration/upgrade, then we are keeping previous
+ # configuration and we are done.
+ if migrate_if_needed(opt, config.root_path):
+ sys.exit(0)
if not config.model_conf_path.exists():
initialize_rootdir(config.root_path, opt.yes_to_all)
+ models_to_download = default_user_selections(opt)
+ new_init_file = config.root_path / 'invokeai.yaml'
if opt.yes_to_all:
write_default_options(opt, new_init_file)
init_options = Namespace(
@@ -855,29 +846,21 @@ def main():
if init_options:
write_opts(init_options, new_init_file)
else:
- print(
+ logger.info(
'\n** CANCELLED AT USER\'S REQUEST. USE THE "invoke.sh" LAUNCHER TO RUN LATER **\n'
)
sys.exit(0)
-
+
if opt.skip_support_models:
- print("\n** SKIPPING SUPPORT MODEL DOWNLOADS PER USER REQUEST **")
+ logger.info("SKIPPING SUPPORT MODEL DOWNLOADS PER USER REQUEST")
else:
- print("\n** CHECKING/UPDATING SUPPORT MODELS **")
- download_bert()
- download_sd1_clip()
- download_sd2_clip()
- download_realesrgan()
- download_gfpgan()
- download_codeformer()
- download_clipseg()
- download_safety_checker()
- download_vaes()
+ logger.info("CHECKING/UPDATING SUPPORT MODELS")
+ download_support_models()
if opt.skip_sd_weights:
- print("\n** SKIPPING DIFFUSION WEIGHTS DOWNLOAD PER USER REQUEST **")
+ logger.info("\n** SKIPPING DIFFUSION WEIGHTS DOWNLOAD PER USER REQUEST **")
elif models_to_download:
- print("\n** DOWNLOADING DIFFUSION WEIGHTS **")
+ logger.info("\n** DOWNLOADING DIFFUSION WEIGHTS **")
process_and_execute(opt, models_to_download)
postscript(errors=errors)
diff --git a/invokeai/backend/install/legacy_arg_parsing.py b/invokeai/backend/install/legacy_arg_parsing.py
index 4a58ff8336..684c50c77d 100644
--- a/invokeai/backend/install/legacy_arg_parsing.py
+++ b/invokeai/backend/install/legacy_arg_parsing.py
@@ -4,6 +4,8 @@ import argparse
import shlex
from argparse import ArgumentParser
+# note that this includes both old sampler names and new scheduler names
+# in order to be able to parse both 2.0 and 3.0-pre-nodes versions of invokeai.init
SAMPLER_CHOICES = [
"ddim",
"ddpm",
@@ -27,6 +29,15 @@ SAMPLER_CHOICES = [
"dpmpp_sde",
"dpmpp_sde_k",
"unipc",
+ "k_dpm_2_a",
+ "k_dpm_2",
+ "k_dpmpp_2_a",
+ "k_dpmpp_2",
+ "k_euler_a",
+ "k_euler",
+ "k_heun",
+ "k_lms",
+ "plms",
]
PRECISION_CHOICES = [
diff --git a/invokeai/backend/install/migrate_to_3.py b/invokeai/backend/install/migrate_to_3.py
new file mode 100644
index 0000000000..4500d7b7cd
--- /dev/null
+++ b/invokeai/backend/install/migrate_to_3.py
@@ -0,0 +1,606 @@
+'''
+Migrate the models directory and models.yaml file from an existing
+InvokeAI 2.3 installation to 3.0.0.
+'''
+
+import os
+import argparse
+import shutil
+import yaml
+
+import transformers
+import diffusers
+import warnings
+
+from dataclasses import dataclass
+from pathlib import Path
+from omegaconf import OmegaConf, DictConfig
+from typing import Union
+
+from diffusers import StableDiffusionPipeline, AutoencoderKL
+from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
+from transformers import (
+ CLIPTextModel,
+ CLIPTokenizer,
+ AutoFeatureExtractor,
+ BertTokenizerFast,
+)
+
+import invokeai.backend.util.logging as logger
+from invokeai.app.services.config import InvokeAIAppConfig
+from invokeai.backend.model_management import ModelManager
+from invokeai.backend.model_management.model_probe import (
+ ModelProbe, ModelType, BaseModelType, ModelProbeInfo
+ )
+
+warnings.filterwarnings("ignore")
+transformers.logging.set_verbosity_error()
+diffusers.logging.set_verbosity_error()
+
+# holder for paths that we will migrate
+@dataclass
+class ModelPaths:
+ models: Path
+ embeddings: Path
+ loras: Path
+ controlnets: Path
+
+class MigrateTo3(object):
+ def __init__(self,
+ from_root: Path,
+ to_models: Path,
+ model_manager: ModelManager,
+ src_paths: ModelPaths,
+ ):
+ self.root_directory = from_root
+ self.dest_models = to_models
+ self.mgr = model_manager
+ self.src_paths = src_paths
+
+ @classmethod
+ def initialize_yaml(cls, yaml_file: Path):
+ with open(yaml_file, 'w') as file:
+ file.write(
+ yaml.dump(
+ {
+ '__metadata__': {'version':'3.0.0'}
+ }
+ )
+ )
+
+ def create_directory_structure(self):
+ '''
+ Create the basic directory structure for the models folder.
+ '''
+ for model_base in [BaseModelType.StableDiffusion1,BaseModelType.StableDiffusion2]:
+ for model_type in [ModelType.Main, ModelType.Vae, ModelType.Lora,
+ ModelType.ControlNet,ModelType.TextualInversion]:
+ path = self.dest_models / model_base.value / model_type.value
+ path.mkdir(parents=True, exist_ok=True)
+ path = self.dest_models / 'core'
+ path.mkdir(parents=True, exist_ok=True)
+
+ @staticmethod
+ def copy_file(src:Path,dest:Path):
+ '''
+ copy a single file with logging
+ '''
+ if dest.exists():
+ logger.info(f'Skipping existing {str(dest)}')
+ return
+ logger.info(f'Copying {str(src)} to {str(dest)}')
+ try:
+ shutil.copy(src, dest)
+ except Exception as e:
+ logger.error(f'COPY FAILED: {str(e)}')
+
+ @staticmethod
+ def copy_dir(src:Path,dest:Path):
+ '''
+ Recursively copy a directory with logging
+ '''
+ if dest.exists():
+ logger.info(f'Skipping existing {str(dest)}')
+ return
+
+ logger.info(f'Copying {str(src)} to {str(dest)}')
+ try:
+ shutil.copytree(src, dest)
+ except Exception as e:
+ logger.error(f'COPY FAILED: {str(e)}')
+
+ def migrate_models(self, src_dir: Path):
+ '''
+ Recursively walk through src directory, probe anything
+ that looks like a model, and copy the model into the
+ appropriate location within the destination models directory.
+ '''
+ directories_scanned = set()
+ for root, dirs, files in os.walk(src_dir):
+ for d in dirs:
+ try:
+ model = Path(root,d)
+ info = ModelProbe().heuristic_probe(model)
+ if not info:
+ continue
+ dest = self._model_probe_to_path(info) / model.name
+ self.copy_dir(model, dest)
+ directories_scanned.add(model)
+ except Exception as e:
+ logger.error(str(e))
+ except KeyboardInterrupt:
+ raise
+ except Exception as e:
+ logger.error(str(e))
+ for f in files:
+ # don't copy raw learned_embeds.bin or pytorch_lora_weights.bin
+ # let them be copied as part of a tree copy operation
+ try:
+ if f in {'learned_embeds.bin','pytorch_lora_weights.bin'}:
+ continue
+ model = Path(root,f)
+ if model.parent in directories_scanned:
+ continue
+ info = ModelProbe().heuristic_probe(model)
+ if not info:
+ continue
+ dest = self._model_probe_to_path(info) / f
+ self.copy_file(model, dest)
+ except Exception as e:
+ logger.error(str(e))
+ except KeyboardInterrupt:
+ raise
+ except Exception as e:
+ logger.error(str(e))
+
+ def migrate_support_models(self):
+ '''
+ Copy the clipseg, upscaler, and restoration models to their new
+ locations.
+ '''
+ dest_directory = self.dest_models
+ if (self.root_directory / 'models/clipseg').exists():
+ self.copy_dir(self.root_directory / 'models/clipseg', dest_directory / 'core/misc/clipseg')
+ if (self.root_directory / 'models/realesrgan').exists():
+ self.copy_dir(self.root_directory / 'models/realesrgan', dest_directory / 'core/upscaling/realesrgan')
+ for d in ['codeformer','gfpgan']:
+ path = self.root_directory / 'models' / d
+ if path.exists():
+ self.copy_dir(path,dest_directory / f'core/face_restoration/{d}')
+
+ def migrate_tuning_models(self):
+ '''
+ Migrate the embeddings, loras and controlnets directories to their new homes.
+ '''
+ for src in [self.src_paths.embeddings, self.src_paths.loras, self.src_paths.controlnets]:
+ if not src:
+ continue
+ if src.is_dir():
+ logger.info(f'Scanning {src}')
+ self.migrate_models(src)
+ else:
+ logger.info(f'{src} directory not found; skipping')
+ continue
+
+ def migrate_conversion_models(self):
+ '''
+ Migrate all the models that are needed by the ckpt_to_diffusers conversion
+ script.
+ '''
+
+ dest_directory = self.dest_models
+ kwargs = dict(
+ cache_dir = self.root_directory / 'models/hub',
+ #local_files_only = True
+ )
+ try:
+ logger.info('Migrating core tokenizers and text encoders')
+ target_dir = dest_directory / 'core' / 'convert'
+
+ self._migrate_pretrained(BertTokenizerFast,
+ repo_id='bert-base-uncased',
+ dest = target_dir / 'bert-base-uncased',
+ **kwargs)
+
+ # sd-1
+ repo_id = 'openai/clip-vit-large-patch14'
+ self._migrate_pretrained(CLIPTokenizer,
+ repo_id= repo_id,
+ dest= target_dir / 'clip-vit-large-patch14',
+ **kwargs)
+ self._migrate_pretrained(CLIPTextModel,
+ repo_id = repo_id,
+ dest = target_dir / 'clip-vit-large-patch14',
+ force = True,
+ **kwargs)
+
+ # sd-2
+ repo_id = "stabilityai/stable-diffusion-2"
+ self._migrate_pretrained(CLIPTokenizer,
+ repo_id = repo_id,
+ dest = target_dir / 'stable-diffusion-2-clip' / 'tokenizer',
+ **{'subfolder':'tokenizer',**kwargs}
+ )
+ self._migrate_pretrained(CLIPTextModel,
+ repo_id = repo_id,
+ dest = target_dir / 'stable-diffusion-2-clip' / 'text_encoder',
+ **{'subfolder':'text_encoder',**kwargs}
+ )
+
+ # VAE
+ logger.info('Migrating stable diffusion VAE')
+ self._migrate_pretrained(AutoencoderKL,
+ repo_id = 'stabilityai/sd-vae-ft-mse',
+ dest = target_dir / 'sd-vae-ft-mse',
+ **kwargs)
+
+ # safety checking
+ logger.info('Migrating safety checker')
+ repo_id = "CompVis/stable-diffusion-safety-checker"
+ self._migrate_pretrained(AutoFeatureExtractor,
+ repo_id = repo_id,
+ dest = target_dir / 'stable-diffusion-safety-checker',
+ **kwargs)
+ self._migrate_pretrained(StableDiffusionSafetyChecker,
+ repo_id = repo_id,
+ dest = target_dir / 'stable-diffusion-safety-checker',
+ **kwargs)
+ except KeyboardInterrupt:
+ raise
+ except Exception as e:
+ logger.error(str(e))
+
+ def _model_probe_to_path(self, info: ModelProbeInfo)->Path:
+ return Path(self.dest_models, info.base_type.value, info.model_type.value)
+
+ def _migrate_pretrained(self, model_class, repo_id: str, dest: Path, force:bool=False, **kwargs):
+ if dest.exists() and not force:
+ logger.info(f'Skipping existing {dest}')
+ return
+ model = model_class.from_pretrained(repo_id, **kwargs)
+ self._save_pretrained(model, dest, overwrite=force)
+
+ def _save_pretrained(self, model, dest: Path, overwrite: bool=False):
+ model_name = dest.name
+ if overwrite:
+ model.save_pretrained(dest, safe_serialization=True)
+ else:
+ download_path = dest.with_name(f'{model_name}.downloading')
+ model.save_pretrained(download_path, safe_serialization=True)
+ download_path.replace(dest)
+
+ def _download_vae(self, repo_id: str, subfolder:str=None)->Path:
+ vae = AutoencoderKL.from_pretrained(repo_id, cache_dir=self.root_directory / 'models/hub', subfolder=subfolder)
+ info = ModelProbe().heuristic_probe(vae)
+ _, model_name = repo_id.split('/')
+ dest = self._model_probe_to_path(info) / self.unique_name(model_name, info)
+ vae.save_pretrained(dest, safe_serialization=True)
+ return dest
+
+ def _vae_path(self, vae: Union[str,dict])->Path:
+ '''
+ Convert 2.3 VAE stanza to a straight path.
+ '''
+ vae_path = None
+
+ # First get a path
+ if isinstance(vae,str):
+ vae_path = vae
+
+ elif isinstance(vae,DictConfig):
+ if p := vae.get('path'):
+ vae_path = p
+ elif repo_id := vae.get('repo_id'):
+ if repo_id=='stabilityai/sd-vae-ft-mse': # this guy is already downloaded
+ vae_path = 'models/core/convert/sd-vae-ft-mse'
+ return vae_path
+ else:
+ vae_path = self._download_vae(repo_id, vae.get('subfolder'))
+
+ assert vae_path is not None, "Couldn't find VAE for this model"
+
+ # if the VAE is in the old models directory, then we must move it into the new
+ # one. VAEs outside of this directory can stay where they are.
+ vae_path = Path(vae_path)
+ if vae_path.is_relative_to(self.src_paths.models):
+ info = ModelProbe().heuristic_probe(vae_path)
+ dest = self._model_probe_to_path(info) / vae_path.name
+ if not dest.exists():
+ if vae_path.is_dir():
+ self.copy_dir(vae_path,dest)
+ else:
+ self.copy_file(vae_path,dest)
+ vae_path = dest
+
+ if vae_path.is_relative_to(self.dest_models):
+ rel_path = vae_path.relative_to(self.dest_models)
+ return Path('models',rel_path)
+ else:
+ return vae_path
+
+ def migrate_repo_id(self, repo_id: str, model_name: str=None, **extra_config):
+ '''
+ Migrate a locally-cached diffusers pipeline identified with a repo_id
+ '''
+ dest_dir = self.dest_models
+
+ cache = self.root_directory / 'models/hub'
+ kwargs = dict(
+ cache_dir = cache,
+ safety_checker = None,
+ # local_files_only = True,
+ )
+
+ owner,repo_name = repo_id.split('/')
+ model_name = model_name or repo_name
+ model = cache / '--'.join(['models',owner,repo_name])
+
+ if len(list(model.glob('snapshots/**/model_index.json')))==0:
+ return
+ revisions = [x.name for x in model.glob('refs/*')]
+
+ # if an fp16 is available we use that
+ revision = 'fp16' if len(revisions) > 1 and 'fp16' in revisions else revisions[0]
+ pipeline = StableDiffusionPipeline.from_pretrained(
+ repo_id,
+ revision=revision,
+ **kwargs)
+
+ info = ModelProbe().heuristic_probe(pipeline)
+ if not info:
+ return
+
+ if self.mgr.model_exists(model_name, info.base_type, info.model_type):
+ logger.warning(f'A model named {model_name} already exists at the destination. Skipping migration.')
+ return
+
+ dest = self._model_probe_to_path(info) / model_name
+ self._save_pretrained(pipeline, dest)
+
+ rel_path = Path('models',dest.relative_to(dest_dir))
+ self._add_model(model_name, info, rel_path, **extra_config)
+
+ def migrate_path(self, location: Path, model_name: str=None, **extra_config):
+ '''
+ Migrate a model referred to using 'weights' or 'path'
+ '''
+
+ # handle relative paths
+ dest_dir = self.dest_models
+ location = self.root_directory / location
+ model_name = model_name or location.stem
+
+ info = ModelProbe().heuristic_probe(location)
+ if not info:
+ return
+
+ if self.mgr.model_exists(model_name, info.base_type, info.model_type):
+ logger.warning(f'A model named {model_name} already exists at the destination. Skipping migration.')
+ return
+
+ # uh oh, weights is in the old models directory - move it into the new one
+ if Path(location).is_relative_to(self.src_paths.models):
+ dest = Path(dest_dir, info.base_type.value, info.model_type.value, location.name)
+ if location.is_dir():
+ self.copy_dir(location,dest)
+ else:
+ self.copy_file(location,dest)
+ location = Path('models', info.base_type.value, info.model_type.value, location.name)
+
+ self._add_model(model_name, info, location, **extra_config)
+
+ def _add_model(self,
+ model_name: str,
+ info: ModelProbeInfo,
+ location: Path,
+ **extra_config):
+ if info.model_type != ModelType.Main:
+ return
+
+ self.mgr.add_model(
+ model_name = model_name,
+ base_model = info.base_type,
+ model_type = info.model_type,
+ clobber = True,
+ model_attributes = {
+ 'path': str(location),
+ 'description': f'A {info.base_type.value} {info.model_type.value} model',
+ 'model_format': info.format,
+ 'variant': info.variant_type.value,
+ **extra_config,
+ }
+ )
+
+ def migrate_defined_models(self):
+ '''
+ Migrate models defined in models.yaml
+ '''
+ # find any models referred to in old models.yaml
+ conf = OmegaConf.load(self.root_directory / 'configs/models.yaml')
+
+ for model_name, stanza in conf.items():
+
+ try:
+ passthru_args = {}
+
+ if vae := stanza.get('vae'):
+ try:
+ passthru_args['vae'] = str(self._vae_path(vae))
+ except Exception as e:
+ logger.warning(f'Could not find a VAE matching "{vae}" for model "{model_name}"')
+ logger.warning(str(e))
+
+ if config := stanza.get('config'):
+ passthru_args['config'] = config
+
+ if description:= stanza.get('description'):
+ passthru_args['description'] = description
+
+ if repo_id := stanza.get('repo_id'):
+ logger.info(f'Migrating diffusers model {model_name}')
+ self.migrate_repo_id(repo_id, model_name, **passthru_args)
+
+ elif location := stanza.get('weights'):
+ logger.info(f'Migrating checkpoint model {model_name}')
+ self.migrate_path(Path(location), model_name, **passthru_args)
+
+ elif location := stanza.get('path'):
+ logger.info(f'Migrating diffusers model {model_name}')
+ self.migrate_path(Path(location), model_name, **passthru_args)
+
+ except KeyboardInterrupt:
+ raise
+ except Exception as e:
+ logger.error(str(e))
+
+ def migrate(self):
+ self.create_directory_structure()
+ # the configure script is doing this
+ self.migrate_support_models()
+ self.migrate_conversion_models()
+ self.migrate_tuning_models()
+ self.migrate_defined_models()
+
+def _parse_legacy_initfile(root: Path, initfile: Path)->ModelPaths:
+ '''
+ Returns tuple of (embedding_path, lora_path, controlnet_path)
+ '''
+ parser = argparse.ArgumentParser(fromfile_prefix_chars='@')
+ parser.add_argument(
+ '--embedding_directory',
+ '--embedding_path',
+ type=Path,
+ dest='embedding_path',
+ default=Path('embeddings'),
+ )
+ parser.add_argument(
+ '--lora_directory',
+ dest='lora_path',
+ type=Path,
+ default=Path('loras'),
+ )
+ opt,_ = parser.parse_known_args([f'@{str(initfile)}'])
+ return ModelPaths(
+ models = root / 'models',
+ embeddings = root / str(opt.embedding_path).strip('"'),
+ loras = root / str(opt.lora_path).strip('"'),
+ controlnets = root / 'controlnets',
+ )
+
+def _parse_legacy_yamlfile(root: Path, initfile: Path)->ModelPaths:
+ '''
+ Returns tuple of (embedding_path, lora_path, controlnet_path)
+ '''
+ # Don't use the config object because it is unforgiving of version updates
+ # Just use omegaconf directly
+ opt = OmegaConf.load(initfile)
+ paths = opt.InvokeAI.Paths
+ models = paths.get('models_dir','models')
+ embeddings = paths.get('embedding_dir','embeddings')
+ loras = paths.get('lora_dir','loras')
+ controlnets = paths.get('controlnet_dir','controlnets')
+ return ModelPaths(
+ models = root / models,
+ embeddings = root / embeddings,
+ loras = root /loras,
+ controlnets = root / controlnets,
+ )
+
+def get_legacy_embeddings(root: Path) -> ModelPaths:
+ path = root / 'invokeai.init'
+ if path.exists():
+ return _parse_legacy_initfile(root, path)
+ path = root / 'invokeai.yaml'
+ if path.exists():
+ return _parse_legacy_yamlfile(root, path)
+
+def do_migrate(src_directory: Path, dest_directory: Path):
+ """
+ Migrate models from src to dest InvokeAI root directories
+ """
+ config_file = dest_directory / 'configs' / 'models.yaml.3'
+ dest_models = dest_directory / 'models.3'
+
+ version_3 = (dest_directory / 'models' / 'core').exists()
+
+ # Here we create the destination models.yaml file.
+ # If we are writing into a version 3 directory and the
+ # file already exists, then we write into a copy of it to
+ # avoid deleting its previous customizations. Otherwise we
+ # create a new empty one.
+ if version_3: # write into the dest directory
+ try:
+ shutil.copy(dest_directory / 'configs' / 'models.yaml', config_file)
+ except:
+ MigrateTo3.initialize_yaml(config_file)
+ mgr = ModelManager(config_file) # important to initialize BEFORE moving the models directory
+ (dest_directory / 'models').replace(dest_models)
+ else:
+ MigrateTo3.initialize_yaml(config_file)
+ mgr = ModelManager(config_file)
+
+ paths = get_legacy_embeddings(src_directory)
+ migrator = MigrateTo3(
+ from_root = src_directory,
+ to_models = dest_models,
+ model_manager = mgr,
+ src_paths = paths
+ )
+ migrator.migrate()
+ print("Migration successful.")
+
+ if not version_3:
+ (dest_directory / 'models').replace(src_directory / 'models.orig')
+ print(f'Original models directory moved to {dest_directory}/models.orig')
+
+ (dest_directory / 'configs' / 'models.yaml').replace(src_directory / 'configs' / 'models.yaml.orig')
+ print(f'Original models.yaml file moved to {dest_directory}/configs/models.yaml.orig')
+
+ config_file.replace(config_file.with_suffix(''))
+ dest_models.replace(dest_models.with_suffix(''))
+
+def main():
+ parser = argparse.ArgumentParser(prog="invokeai-migrate3",
+ description="""
+This will copy and convert the models directory and the configs/models.yaml from the InvokeAI 2.3 format
+'--from-directory' root to the InvokeAI 3.0 '--to-directory' root. These may be abbreviated '--from' and '--to'.a
+
+The old models directory and config file will be renamed 'models.orig' and 'models.yaml.orig' respectively.
+It is safe to provide the same directory for both arguments, but it is better to use the invokeai_configure
+script, which will perform a full upgrade in place."""
+ )
+ parser.add_argument('--from-directory',
+ dest='src_root',
+ type=Path,
+ required=True,
+ help='Source InvokeAI 2.3 root directory (containing "invokeai.init" or "invokeai.yaml")'
+ )
+ parser.add_argument('--to-directory',
+ dest='dest_root',
+ type=Path,
+ required=True,
+ help='Destination InvokeAI 3.0 directory (containing "invokeai.yaml")'
+ )
+ args = parser.parse_args()
+ src_root = args.src_root
+ assert src_root.is_dir(), f"{src_root} is not a valid directory"
+ assert (src_root / 'models').is_dir(), f"{src_root} does not contain a 'models' subdirectory"
+ assert (src_root / 'models' / 'hub').exists(), f"{src_root} does not contain a version 2.3 models directory"
+ assert (src_root / 'invokeai.init').exists() or (src_root / 'invokeai.yaml').exists(), f"{src_root} does not contain an InvokeAI init file."
+
+ dest_root = args.dest_root
+ assert dest_root.is_dir(), f"{dest_root} is not a valid directory"
+ config = InvokeAIAppConfig.get_config()
+ config.parse_args(['--root',str(dest_root)])
+
+ # TODO: revisit
+ # assert (dest_root / 'models').is_dir(), f"{dest_root} does not contain a 'models' subdirectory"
+ # assert (dest_root / 'invokeai.yaml').exists(), f"{dest_root} does not contain an InvokeAI init file."
+
+ do_migrate(src_root,dest_root)
+
+if __name__ == '__main__':
+ main()
+
+
+
diff --git a/invokeai/backend/install/model_install_backend.py b/invokeai/backend/install/model_install_backend.py
index 18964c774e..b6f6d62d97 100644
--- a/invokeai/backend/install/model_install_backend.py
+++ b/invokeai/backend/install/model_install_backend.py
@@ -2,46 +2,37 @@
Utility (backend) functions used by model_install.py
"""
import os
-import re
import shutil
-import sys
import warnings
from dataclasses import dataclass,field
from pathlib import Path
-from tempfile import TemporaryFile
-from typing import List, Dict, Callable
+from tempfile import TemporaryDirectory
+from typing import List, Dict, Callable, Union, Set
import requests
-from diffusers import AutoencoderKL
-from huggingface_hub import hf_hub_url, HfFolder
+from diffusers import StableDiffusionPipeline
+from diffusers import logging as dlogging
+from huggingface_hub import hf_hub_url, HfFolder, HfApi
from omegaconf import OmegaConf
-from omegaconf.dictconfig import DictConfig
from tqdm import tqdm
import invokeai.configs as configs
-
from invokeai.app.services.config import InvokeAIAppConfig
-from ..stable_diffusion import StableDiffusionGeneratorPipeline
+from invokeai.backend.model_management import ModelManager, ModelType, BaseModelType, ModelVariantType, AddModelResult
+from invokeai.backend.model_management.model_probe import ModelProbe, SchedulerPredictionType, ModelProbeInfo
+from invokeai.backend.util import download_with_resume
from ..util.logging import InvokeAILogger
warnings.filterwarnings("ignore")
# --------------------------globals-----------------------
config = InvokeAIAppConfig.get_config()
-
-Model_dir = "models"
-Weights_dir = "ldm/stable-diffusion-v1/"
+logger = InvokeAILogger.getLogger(name='InvokeAI')
# the initial "configs" dir is now bundled in the `invokeai.configs` package
Dataset_path = Path(configs.__path__[0]) / "INITIAL_MODELS.yaml"
-# initial models omegaconf
-Datasets = None
-
-# logger
-logger = InvokeAILogger.getLogger(name='InvokeAI')
-
Config_preamble = """
# This file describes the alternative machine learning models
# available to InvokeAI script.
@@ -52,6 +43,24 @@ Config_preamble = """
# was trained on.
"""
+LEGACY_CONFIGS = {
+ BaseModelType.StableDiffusion1: {
+ ModelVariantType.Normal: 'v1-inference.yaml',
+ ModelVariantType.Inpaint: 'v1-inpainting-inference.yaml',
+ },
+
+ BaseModelType.StableDiffusion2: {
+ ModelVariantType.Normal: {
+ SchedulerPredictionType.Epsilon: 'v2-inference.yaml',
+ SchedulerPredictionType.VPrediction: 'v2-inference-v.yaml',
+ },
+ ModelVariantType.Inpaint: {
+ SchedulerPredictionType.Epsilon: 'v2-inpainting-inference.yaml',
+ SchedulerPredictionType.VPrediction: 'v2-inpainting-inference-v.yaml',
+ }
+ }
+}
+
@dataclass
class ModelInstallList:
'''Class for listing models to be installed/removed'''
@@ -59,133 +68,332 @@ class ModelInstallList:
remove_models: List[str] = field(default_factory=list)
@dataclass
-class UserSelections():
+class InstallSelections():
install_models: List[str]= field(default_factory=list)
remove_models: List[str]=field(default_factory=list)
- purge_deleted_models: bool=field(default_factory=list)
- install_cn_models: List[str] = field(default_factory=list)
- remove_cn_models: List[str] = field(default_factory=list)
- install_lora_models: List[str] = field(default_factory=list)
- remove_lora_models: List[str] = field(default_factory=list)
- install_ti_models: List[str] = field(default_factory=list)
- remove_ti_models: List[str] = field(default_factory=list)
- scan_directory: Path = None
- autoscan_on_startup: bool=False
- import_model_paths: str=None
+# scan_directory: Path = None
+# autoscan_on_startup: bool=False
+
+@dataclass
+class ModelLoadInfo():
+ name: str
+ model_type: ModelType
+ base_type: BaseModelType
+ path: Path = None
+ repo_id: str = None
+ description: str = ''
+ installed: bool = False
+ recommended: bool = False
+ default: bool = False
+
+class ModelInstall(object):
+ def __init__(self,
+ config:InvokeAIAppConfig,
+ prediction_type_helper: Callable[[Path],SchedulerPredictionType]=None,
+ model_manager: ModelManager = None,
+ access_token:str = None):
+ self.config = config
+ self.mgr = model_manager or ModelManager(config.model_conf_path)
+ self.datasets = OmegaConf.load(Dataset_path)
+ self.prediction_helper = prediction_type_helper
+ self.access_token = access_token or HfFolder.get_token()
+ self.reverse_paths = self._reverse_paths(self.datasets)
+
+ def all_models(self)->Dict[str,ModelLoadInfo]:
+ '''
+ Return dict of model_key=>ModelLoadInfo objects.
+ This method consolidates and simplifies the entries in both
+ models.yaml and INITIAL_MODELS.yaml so that they can
+ be treated uniformly. It also sorts the models alphabetically
+ by their name, to improve the display somewhat.
+ '''
+ model_dict = dict()
-def default_config_file():
- return config.model_conf_path
+ # first populate with the entries in INITIAL_MODELS.yaml
+ for key, value in self.datasets.items():
+ name,base,model_type = ModelManager.parse_key(key)
+ value['name'] = name
+ value['base_type'] = base
+ value['model_type'] = model_type
+ model_dict[key] = ModelLoadInfo(**value)
-def sd_configs():
- return config.legacy_conf_path
-
-def initial_models():
- global Datasets
- if Datasets:
- return Datasets
- return (Datasets := OmegaConf.load(Dataset_path)['diffusers'])
-
-def install_requested_models(
- diffusers: ModelInstallList = None,
- controlnet: ModelInstallList = None,
- lora: ModelInstallList = None,
- ti: ModelInstallList = None,
- cn_model_map: Dict[str,str] = None, # temporary - move to model manager
- scan_directory: Path = None,
- external_models: List[str] = None,
- scan_at_startup: bool = False,
- precision: str = "float16",
- purge_deleted: bool = False,
- config_file_path: Path = None,
- model_config_file_callback: Callable[[Path],Path] = None
-):
- """
- Entry point for installing/deleting starter models, or installing external models.
- """
- access_token = HfFolder.get_token()
- config_file_path = config_file_path or default_config_file()
- if not config_file_path.exists():
- open(config_file_path, "w")
-
- # prevent circular import here
- from ..model_management import ModelManager
- model_manager = ModelManager(OmegaConf.load(config_file_path), precision=precision)
- if controlnet:
- model_manager.install_controlnet_models(controlnet.install_models, access_token=access_token)
- model_manager.delete_controlnet_models(controlnet.remove_models)
-
- if lora:
- model_manager.install_lora_models(lora.install_models, access_token=access_token)
- model_manager.delete_lora_models(lora.remove_models)
-
- if ti:
- model_manager.install_ti_models(ti.install_models, access_token=access_token)
- model_manager.delete_ti_models(ti.remove_models)
-
- if diffusers:
- # TODO: Replace next three paragraphs with calls into new model manager
- if diffusers.remove_models and len(diffusers.remove_models) > 0:
- logger.info("Processing requested deletions")
- for model in diffusers.remove_models:
- logger.info(f"{model}...")
- model_manager.del_model(model, delete_files=purge_deleted)
- model_manager.commit(config_file_path)
-
- if diffusers.install_models and len(diffusers.install_models) > 0:
- logger.info("Installing requested models")
- downloaded_paths = download_weight_datasets(
- models=diffusers.install_models,
- access_token=None,
- precision=precision,
- )
- successful = {x:v for x,v in downloaded_paths.items() if v is not None}
- if len(successful) > 0:
- update_config_file(successful, config_file_path)
- if len(successful) < len(diffusers.install_models):
- unsuccessful = [x for x in downloaded_paths if downloaded_paths[x] is None]
- logger.warning(f"Some of the model downloads were not successful: {unsuccessful}")
-
- # due to above, we have to reload the model manager because conf file
- # was changed behind its back
- model_manager = ModelManager(OmegaConf.load(config_file_path), precision=precision)
-
- external_models = external_models or list()
- if scan_directory:
- external_models.append(str(scan_directory))
-
- if len(external_models) > 0:
- logger.info("INSTALLING EXTERNAL MODELS")
- for path_url_or_repo in external_models:
- try:
- logger.debug(f'In install_requested_models; callback = {model_config_file_callback}')
- model_manager.heuristic_import(
- path_url_or_repo,
- commit_to_conf=config_file_path,
- config_file_callback = model_config_file_callback,
+ # supplement with entries in models.yaml
+ installed_models = self.mgr.list_models()
+ for md in installed_models:
+ base = md['base_model']
+ model_type = md['model_type']
+ name = md['model_name']
+ key = ModelManager.create_key(name, base, model_type)
+ if key in model_dict:
+ model_dict[key].installed = True
+ else:
+ model_dict[key] = ModelLoadInfo(
+ name = name,
+ base_type = base,
+ model_type = model_type,
+ path = value.get('path'),
+ installed = True,
)
- except KeyboardInterrupt:
- sys.exit(-1)
- except Exception:
+ return {x : model_dict[x] for x in sorted(model_dict.keys(),key=lambda y: model_dict[y].name.lower())}
+
+ def starter_models(self)->Set[str]:
+ models = set()
+ for key, value in self.datasets.items():
+ name,base,model_type = ModelManager.parse_key(key)
+ if model_type==ModelType.Main:
+ models.add(key)
+ return models
+
+ def recommended_models(self)->Set[str]:
+ starters = self.starter_models()
+ return set([x for x in starters if self.datasets[x].get('recommended',False)])
+
+ def default_model(self)->str:
+ starters = self.starter_models()
+ defaults = [x for x in starters if self.datasets[x].get('default',False)]
+ return defaults[0]
+
+ def install(self, selections: InstallSelections):
+ verbosity = dlogging.get_verbosity() # quench NSFW nags
+ dlogging.set_verbosity_error()
+
+ job = 1
+ jobs = len(selections.remove_models) + len(selections.install_models)
+
+ # remove requested models
+ for key in selections.remove_models:
+ name,base,mtype = self.mgr.parse_key(key)
+ logger.info(f'Deleting {mtype} model {name} [{job}/{jobs}]')
+ try:
+ self.mgr.del_model(name,base,mtype)
+ except FileNotFoundError as e:
+ logger.warning(e)
+ job += 1
+
+ # add requested models
+ for path in selections.install_models:
+ logger.info(f'Installing {path} [{job}/{jobs}]')
+ try:
+ self.heuristic_import(path)
+ except (ValueError, KeyError) as e:
+ logger.error(str(e))
+ job += 1
+
+ dlogging.set_verbosity(verbosity)
+ self.mgr.commit()
+
+ def heuristic_import(self,
+ model_path_id_or_url: Union[str,Path],
+ models_installed: Set[Path]=None,
+ )->Dict[str, AddModelResult]:
+ '''
+ :param model_path_id_or_url: A Path to a local model to import, or a string representing its repo_id or URL
+ :param models_installed: Set of installed models, used for recursive invocation
+ Returns a set of dict objects corresponding to newly-created stanzas in models.yaml.
+ '''
+
+ if not models_installed:
+ models_installed = dict()
+
+ # A little hack to allow nested routines to retrieve info on the requested ID
+ self.current_id = model_path_id_or_url
+ path = Path(model_path_id_or_url)
+ # checkpoint file, or similar
+ if path.is_file():
+ models_installed.update({str(path):self._install_path(path)})
+
+ # folders style or similar
+ elif path.is_dir() and any([(path/x).exists() for x in \
+ {'config.json','model_index.json','learned_embeds.bin','pytorch_lora_weights.bin'}
+ ]
+ ):
+ models_installed.update(self._install_path(path))
+
+ # recursive scan
+ elif path.is_dir():
+ for child in path.iterdir():
+ self.heuristic_import(child, models_installed=models_installed)
+
+ # huggingface repo
+ elif len(str(model_path_id_or_url).split('/')) == 2:
+ models_installed.update({str(model_path_id_or_url): self._install_repo(str(model_path_id_or_url))})
+
+ # a URL
+ elif str(model_path_id_or_url).startswith(("http:", "https:", "ftp:")):
+ models_installed.update({str(model_path_id_or_url): self._install_url(model_path_id_or_url)})
+
+ else:
+ raise KeyError(f'{str(model_path_id_or_url)} is not recognized as a local path, repo ID or URL. Skipping')
+
+ return models_installed
+
+ # install a model from a local path. The optional info parameter is there to prevent
+ # the model from being probed twice in the event that it has already been probed.
+ def _install_path(self, path: Path, info: ModelProbeInfo=None)->AddModelResult:
+ info = info or ModelProbe().heuristic_probe(path,self.prediction_helper)
+ if not info:
+ logger.warning(f'Unable to parse format of {path}')
+ return None
+ model_name = path.stem if path.is_file() else path.name
+ if self.mgr.model_exists(model_name, info.base_type, info.model_type):
+ raise ValueError(f'A model named "{model_name}" is already installed.')
+ attributes = self._make_attributes(path,info)
+ return self.mgr.add_model(model_name = model_name,
+ base_model = info.base_type,
+ model_type = info.model_type,
+ model_attributes = attributes,
+ )
+
+ def _install_url(self, url: str)->AddModelResult:
+ with TemporaryDirectory(dir=self.config.models_path) as staging:
+ location = download_with_resume(url,Path(staging))
+ if not location:
+ logger.error(f'Unable to download {url}. Skipping.')
+ info = ModelProbe().heuristic_probe(location)
+ dest = self.config.models_path / info.base_type.value / info.model_type.value / location.name
+ models_path = shutil.move(location,dest)
+
+ # staged version will be garbage-collected at this time
+ return self._install_path(Path(models_path), info)
+
+ def _install_repo(self, repo_id: str)->AddModelResult:
+ hinfo = HfApi().model_info(repo_id)
+
+ # we try to figure out how to download this most economically
+ # list all the files in the repo
+ files = [x.rfilename for x in hinfo.siblings]
+ location = None
+
+ with TemporaryDirectory(dir=self.config.models_path) as staging:
+ staging = Path(staging)
+ if 'model_index.json' in files:
+ location = self._download_hf_pipeline(repo_id, staging) # pipeline
+ else:
+ for suffix in ['safetensors','bin']:
+ if f'pytorch_lora_weights.{suffix}' in files:
+ location = self._download_hf_model(repo_id, ['pytorch_lora_weights.bin'], staging) # LoRA
+ break
+ elif self.config.precision=='float16' and f'diffusion_pytorch_model.fp16.{suffix}' in files: # vae, controlnet or some other standalone
+ files = ['config.json', f'diffusion_pytorch_model.fp16.{suffix}']
+ location = self._download_hf_model(repo_id, files, staging)
+ break
+ elif f'diffusion_pytorch_model.{suffix}' in files:
+ files = ['config.json', f'diffusion_pytorch_model.{suffix}']
+ location = self._download_hf_model(repo_id, files, staging)
+ break
+ elif f'learned_embeds.{suffix}' in files:
+ location = self._download_hf_model(repo_id, [f'learned_embeds.{suffix}'], staging)
+ break
+ if not location:
+ logger.warning(f'Could not determine type of repo {repo_id}. Skipping install.')
+ return {}
+
+ info = ModelProbe().heuristic_probe(location, self.prediction_helper)
+ if not info:
+ logger.warning(f'Could not probe {location}. Skipping install.')
+ return {}
+ dest = self.config.models_path / info.base_type.value / info.model_type.value / self._get_model_name(repo_id,location)
+ if dest.exists():
+ shutil.rmtree(dest)
+ shutil.copytree(location,dest)
+ return self._install_path(dest, info)
+
+ def _get_model_name(self,path_name: str, location: Path)->str:
+ '''
+ Calculate a name for the model - primitive implementation.
+ '''
+ if key := self.reverse_paths.get(path_name):
+ (name, base, mtype) = ModelManager.parse_key(key)
+ return name
+ else:
+ return location.stem
+
+ def _make_attributes(self, path: Path, info: ModelProbeInfo)->dict:
+ model_name = path.name if path.is_dir() else path.stem
+ description = f'{info.base_type.value} {info.model_type.value} model {model_name}'
+ if key := self.reverse_paths.get(self.current_id):
+ if key in self.datasets:
+ description = self.datasets[key].get('description') or description
+
+ rel_path = self.relative_to_root(path)
+
+ attributes = dict(
+ path = str(rel_path),
+ description = str(description),
+ model_format = info.format,
+ )
+ if info.model_type == ModelType.Main:
+ attributes.update(dict(variant = info.variant_type,))
+ if info.format=="checkpoint":
+ try:
+ possible_conf = path.with_suffix('.yaml')
+ if possible_conf.exists():
+ legacy_conf = str(self.relative_to_root(possible_conf))
+ elif info.base_type == BaseModelType.StableDiffusion2:
+ legacy_conf = Path(self.config.legacy_conf_dir, LEGACY_CONFIGS[info.base_type][info.variant_type][info.prediction_type])
+ else:
+ legacy_conf = Path(self.config.legacy_conf_dir, LEGACY_CONFIGS[info.base_type][info.variant_type])
+ except KeyError:
+ legacy_conf = Path(self.config.legacy_conf_dir, 'v1-inference.yaml') # best guess
+
+ attributes.update(
+ dict(
+ config = str(legacy_conf)
+ )
+ )
+ return attributes
+
+ def relative_to_root(self, path: Path)->Path:
+ root = self.config.root_path
+ if path.is_relative_to(root):
+ return path.relative_to(root)
+ else:
+ return path
+
+ def _download_hf_pipeline(self, repo_id: str, staging: Path)->Path:
+ '''
+ This retrieves a StableDiffusion model from cache or remote and then
+ does a save_pretrained() to the indicated staging area.
+ '''
+ _,name = repo_id.split("/")
+ revisions = ['fp16','main'] if self.config.precision=='float16' else ['main']
+ model = None
+ for revision in revisions:
+ try:
+ model = StableDiffusionPipeline.from_pretrained(repo_id,revision=revision,safety_checker=None)
+ except: # most errors are due to fp16 not being present. Fix this to catch other errors
pass
+ if model:
+ break
+ if not model:
+ logger.error(f'Diffusers model {repo_id} could not be downloaded. Skipping.')
+ return None
+ model.save_pretrained(staging / name, safe_serialization=True)
+ return staging / name
- if scan_at_startup and scan_directory.is_dir():
- update_autoconvert_dir(scan_directory)
- else:
- update_autoconvert_dir(None)
-
-def update_autoconvert_dir(autodir: Path):
- '''
- Update the "autoconvert_dir" option in invokeai.yaml
- '''
- invokeai_config_path = config.init_file_path
- conf = OmegaConf.load(invokeai_config_path)
- conf.InvokeAI.Paths.autoconvert_dir = str(autodir) if autodir else None
- yaml = OmegaConf.to_yaml(conf)
- tmpfile = invokeai_config_path.parent / "new_config.tmp"
- with open(tmpfile, "w", encoding="utf-8") as outfile:
- outfile.write(yaml)
- tmpfile.replace(invokeai_config_path)
+ def _download_hf_model(self, repo_id: str, files: List[str], staging: Path)->Path:
+ _,name = repo_id.split("/")
+ location = staging / name
+ paths = list()
+ for filename in files:
+ p = hf_download_with_resume(repo_id,
+ model_dir=location,
+ model_name=filename,
+ access_token = self.access_token
+ )
+ if p:
+ paths.append(p)
+ else:
+ logger.warning(f'Could not download {filename} from {repo_id}.')
+
+ return location if len(paths)>0 else None
+ @classmethod
+ def _reverse_paths(cls,datasets)->dict:
+ '''
+ Reverse mapping from repo_id/path to destination name.
+ '''
+ return {v.get('path') or v.get('repo_id') : k for k, v in datasets.items()}
# -------------------------------------
def yes_or_no(prompt: str, default_yes=True):
@@ -197,133 +405,19 @@ def yes_or_no(prompt: str, default_yes=True):
return response[0] in ("y", "Y")
# ---------------------------------------------
-def recommended_datasets() -> List['str']:
- datasets = set()
- for ds in initial_models().keys():
- if initial_models()[ds].get("recommended", False):
- datasets.add(ds)
- return list(datasets)
-
-# ---------------------------------------------
-def default_dataset() -> dict:
- datasets = set()
- for ds in initial_models().keys():
- if initial_models()[ds].get("default", False):
- datasets.add(ds)
- return list(datasets)
-
-
-# ---------------------------------------------
-def all_datasets() -> dict:
- datasets = dict()
- for ds in initial_models().keys():
- datasets[ds] = True
- return datasets
-
-
-# ---------------------------------------------
-# look for legacy model.ckpt in models directory and offer to
-# normalize its name
-def migrate_models_ckpt():
- model_path = os.path.join(config.root_dir, Model_dir, Weights_dir)
- if not os.path.exists(os.path.join(model_path, "model.ckpt")):
- return
- new_name = initial_models()["stable-diffusion-1.4"]["file"]
- logger.warning(
- 'The Stable Diffusion v4.1 "model.ckpt" is already installed. The name will be changed to {new_name} to avoid confusion.'
- )
- logger.warning(f"model.ckpt => {new_name}")
- os.replace(
- os.path.join(model_path, "model.ckpt"), os.path.join(model_path, new_name)
- )
-
-
-# ---------------------------------------------
-def download_weight_datasets(
- models: List[str], access_token: str, precision: str = "float32"
-):
- migrate_models_ckpt()
- successful = dict()
- for mod in models:
- logger.info(f"Downloading {mod}:")
- successful[mod] = _download_repo_or_file(
- initial_models()[mod], access_token, precision=precision
- )
- return successful
-
-
-def _download_repo_or_file(
- mconfig: DictConfig, access_token: str, precision: str = "float32"
-) -> Path:
- path = None
- if mconfig["format"] == "ckpt":
- path = _download_ckpt_weights(mconfig, access_token)
- else:
- path = _download_diffusion_weights(mconfig, access_token, precision=precision)
- if "vae" in mconfig and "repo_id" in mconfig["vae"]:
- _download_diffusion_weights(
- mconfig["vae"], access_token, precision=precision
- )
- return path
-
-def _download_ckpt_weights(mconfig: DictConfig, access_token: str) -> Path:
- repo_id = mconfig["repo_id"]
- filename = mconfig["file"]
- cache_dir = os.path.join(config.root_dir, Model_dir, Weights_dir)
- return hf_download_with_resume(
- repo_id=repo_id,
- model_dir=cache_dir,
- model_name=filename,
- access_token=access_token,
- )
-
-
-# ---------------------------------------------
-def download_from_hf(
- model_class: object, model_name: str, **kwargs
+def hf_download_from_pretrained(
+ model_class: object, model_name: str, destination: Path, **kwargs
):
logger = InvokeAILogger.getLogger('InvokeAI')
logger.addFilter(lambda x: 'fp16 is not a valid' not in x.getMessage())
- path = config.cache_dir
model = model_class.from_pretrained(
model_name,
- cache_dir=path,
resume_download=True,
**kwargs,
)
- model_name = "--".join(("models", *model_name.split("/")))
- return path / model_name if model else None
-
-
-def _download_diffusion_weights(
- mconfig: DictConfig, access_token: str, precision: str = "float32"
-):
- repo_id = mconfig["repo_id"]
- model_class = (
- StableDiffusionGeneratorPipeline
- if mconfig.get("format", None) == "diffusers"
- else AutoencoderKL
- )
- extra_arg_list = [{"revision": "fp16"}, {}] if precision == "float16" else [{}]
- path = None
- for extra_args in extra_arg_list:
- try:
- path = download_from_hf(
- model_class,
- repo_id,
- safety_checker=None,
- **extra_args,
- )
- except OSError as e:
- if 'Revision Not Found' in str(e):
- pass
- else:
- logger.error(str(e))
- if path:
- break
- return path
-
+ model.save_pretrained(destination, safe_serialization=True)
+ return destination
# ---------------------------------------------
def hf_download_with_resume(
@@ -383,128 +477,3 @@ def hf_download_with_resume(
return model_dest
-# ---------------------------------------------
-def update_config_file(successfully_downloaded: dict, config_file: Path):
- config_file = (
- Path(config_file) if config_file is not None else default_config_file()
- )
-
- # In some cases (incomplete setup, etc), the default configs directory might be missing.
- # Create it if it doesn't exist.
- # this check is ignored if opt.config_file is specified - user is assumed to know what they
- # are doing if they are passing a custom config file from elsewhere.
- if config_file is default_config_file() and not config_file.parent.exists():
- configs_src = Dataset_path.parent
- configs_dest = default_config_file().parent
- shutil.copytree(configs_src, configs_dest, dirs_exist_ok=True)
-
- yaml = new_config_file_contents(successfully_downloaded, config_file)
-
- try:
- backup = None
- if os.path.exists(config_file):
- logger.warning(
- f"{config_file.name} exists. Renaming to {config_file.stem}.yaml.orig"
- )
- backup = config_file.with_suffix(".yaml.orig")
- ## Ugh. Windows is unable to overwrite an existing backup file, raises a WinError 183
- if sys.platform == "win32" and backup.is_file():
- backup.unlink()
- config_file.rename(backup)
-
- with TemporaryFile() as tmp:
- tmp.write(Config_preamble.encode())
- tmp.write(yaml.encode())
-
- with open(str(config_file.expanduser().resolve()), "wb") as new_config:
- tmp.seek(0)
- new_config.write(tmp.read())
-
- except Exception as e:
- logger.error(f"Error creating config file {config_file}: {str(e)}")
- if backup is not None:
- logger.info("restoring previous config file")
- ## workaround, for WinError 183, see above
- if sys.platform == "win32" and config_file.is_file():
- config_file.unlink()
- backup.rename(config_file)
- return
-
- logger.info(f"Successfully created new configuration file {config_file}")
-
-
-# ---------------------------------------------
-def new_config_file_contents(
- successfully_downloaded: dict,
- config_file: Path,
-) -> str:
- if config_file.exists():
- conf = OmegaConf.load(str(config_file.expanduser().resolve()))
- else:
- conf = OmegaConf.create()
-
- default_selected = None
- for model in successfully_downloaded:
- # a bit hacky - what we are doing here is seeing whether a checkpoint
- # version of the model was previously defined, and whether the current
- # model is a diffusers (indicated with a path)
- if conf.get(model) and Path(successfully_downloaded[model]).is_dir():
- delete_weights(model, conf[model])
-
- stanza = {}
- mod = initial_models()[model]
- stanza["description"] = mod["description"]
- stanza["repo_id"] = mod["repo_id"]
- stanza["format"] = mod["format"]
- # diffusers don't need width and height (probably .ckpt doesn't either)
- # so we no longer require these in INITIAL_MODELS.yaml
- if "width" in mod:
- stanza["width"] = mod["width"]
- if "height" in mod:
- stanza["height"] = mod["height"]
- if "file" in mod:
- stanza["weights"] = os.path.relpath(
- successfully_downloaded[model], start=config.root_dir
- )
- stanza["config"] = os.path.normpath(
- os.path.join(sd_configs(), mod["config"])
- )
- if "vae" in mod:
- if "file" in mod["vae"]:
- stanza["vae"] = os.path.normpath(
- os.path.join(Model_dir, Weights_dir, mod["vae"]["file"])
- )
- else:
- stanza["vae"] = mod["vae"]
- if mod.get("default", False):
- stanza["default"] = True
- default_selected = True
-
- conf[model] = stanza
-
- # if no default model was chosen, then we select the first
- # one in the list
- if not default_selected:
- conf[list(successfully_downloaded.keys())[0]]["default"] = True
-
- return OmegaConf.to_yaml(conf)
-
-
-# ---------------------------------------------
-def delete_weights(model_name: str, conf_stanza: dict):
- if not (weights := conf_stanza.get("weights")):
- return
- if re.match("/VAE/", conf_stanza.get("config")):
- return
-
- logger.warning(
- f"\nThe checkpoint version of {model_name} is superseded by the diffusers version. Deleting the original file {weights}?"
- )
-
- weights = Path(weights)
- if not weights.is_absolute():
- weights = config.root_dir / weights
- try:
- weights.unlink()
- except OSError as e:
- logger.error(str(e))
diff --git a/invokeai/backend/model_management/__init__.py b/invokeai/backend/model_management/__init__.py
index e96b064658..e2eef922f4 100644
--- a/invokeai/backend/model_management/__init__.py
+++ b/invokeai/backend/model_management/__init__.py
@@ -1,7 +1,9 @@
"""
Initialization file for invokeai.backend.model_management
"""
-from .model_manager import ModelManager, ModelInfo
+from .model_manager import ModelManager, ModelInfo, AddModelResult, SchedulerPredictionType
from .model_cache import ModelCache
from .models import BaseModelType, ModelType, SubModelType, ModelVariantType
from .lora import ModelPatcher, ONNXModelPatcher
+from .model_merge import ModelMerger, MergeInterpolationMethod
+
diff --git a/invokeai/backend/model_management/convert_ckpt_to_diffusers.py b/invokeai/backend/model_management/convert_ckpt_to_diffusers.py
index db099acbb8..e3e64940de 100644
--- a/invokeai/backend/model_management/convert_ckpt_to_diffusers.py
+++ b/invokeai/backend/model_management/convert_ckpt_to_diffusers.py
@@ -29,8 +29,8 @@ import invokeai.backend.util.logging as logger
from invokeai.app.services.config import InvokeAIAppConfig
from .model_manager import ModelManager
-from .model_cache import ModelCache
-from .models import SchedulerPredictionType, BaseModelType, ModelVariantType
+from picklescan.scanner import scan_file_path
+from .models import BaseModelType, ModelVariantType
try:
from omegaconf import OmegaConf
@@ -73,7 +73,9 @@ from transformers import (
from ..stable_diffusion import StableDiffusionGeneratorPipeline
-MODEL_ROOT = None
+# TODO: redo in future
+#CONVERT_MODEL_ROOT = InvokeAIAppConfig.get_config().models_path / "core" / "convert"
+CONVERT_MODEL_ROOT = InvokeAIAppConfig.get_config().root_path / "models" / "core" / "convert"
def shave_segments(path, n_shave_prefix_segments=1):
"""
@@ -159,17 +161,17 @@ def renew_vae_attention_paths(old_list, n_shave_prefix_segments=0):
new_item = new_item.replace("norm.weight", "group_norm.weight")
new_item = new_item.replace("norm.bias", "group_norm.bias")
- new_item = new_item.replace("q.weight", "query.weight")
- new_item = new_item.replace("q.bias", "query.bias")
+ new_item = new_item.replace("q.weight", "to_q.weight")
+ new_item = new_item.replace("q.bias", "to_q.bias")
- new_item = new_item.replace("k.weight", "key.weight")
- new_item = new_item.replace("k.bias", "key.bias")
+ new_item = new_item.replace("k.weight", "to_k.weight")
+ new_item = new_item.replace("k.bias", "to_k.bias")
- new_item = new_item.replace("v.weight", "value.weight")
- new_item = new_item.replace("v.bias", "value.bias")
+ new_item = new_item.replace("v.weight", "to_v.weight")
+ new_item = new_item.replace("v.bias", "to_v.bias")
- new_item = new_item.replace("proj_out.weight", "proj_attn.weight")
- new_item = new_item.replace("proj_out.bias", "proj_attn.bias")
+ new_item = new_item.replace("proj_out.weight", "to_out.0.weight")
+ new_item = new_item.replace("proj_out.bias", "to_out.0.bias")
new_item = shave_segments(
new_item, n_shave_prefix_segments=n_shave_prefix_segments
@@ -184,7 +186,6 @@ def assign_to_checkpoint(
paths,
checkpoint,
old_checkpoint,
- attention_paths_to_split=None,
additional_replacements=None,
config=None,
):
@@ -199,35 +200,9 @@ def assign_to_checkpoint(
paths, list
), "Paths should be a list of dicts containing 'old' and 'new' keys."
- # Splits the attention layers into three variables.
- if attention_paths_to_split is not None:
- for path, path_map in attention_paths_to_split.items():
- old_tensor = old_checkpoint[path]
- channels = old_tensor.shape[0] // 3
-
- target_shape = (-1, channels) if len(old_tensor.shape) == 3 else (-1)
-
- num_heads = old_tensor.shape[0] // config["num_head_channels"] // 3
-
- old_tensor = old_tensor.reshape(
- (num_heads, 3 * channels // num_heads) + old_tensor.shape[1:]
- )
- query, key, value = old_tensor.split(channels // num_heads, dim=1)
-
- checkpoint[path_map["query"]] = query.reshape(target_shape)
- checkpoint[path_map["key"]] = key.reshape(target_shape)
- checkpoint[path_map["value"]] = value.reshape(target_shape)
-
for path in paths:
new_path = path["new"]
- # These have already been assigned
- if (
- attention_paths_to_split is not None
- and new_path in attention_paths_to_split
- ):
- continue
-
# Global renaming happens here
new_path = new_path.replace("middle_block.0", "mid_block.resnets.0")
new_path = new_path.replace("middle_block.1", "mid_block.attentions.0")
@@ -246,14 +221,14 @@ def assign_to_checkpoint(
def conv_attn_to_linear(checkpoint):
keys = list(checkpoint.keys())
- attn_keys = ["query.weight", "key.weight", "value.weight"]
+ attn_keys = ["to_q.weight", "to_k.weight", "to_v.weight"]
for key in keys:
if ".".join(key.split(".")[-2:]) in attn_keys:
if checkpoint[key].ndim > 2:
checkpoint[key] = checkpoint[key][:, :, 0, 0]
- elif "proj_attn.weight" in key:
+ elif "to_out.0.weight" in key:
if checkpoint[key].ndim > 2:
- checkpoint[key] = checkpoint[key][:, :, 0]
+ checkpoint[key] = checkpoint[key][:, :, 0, 0]
def create_unet_diffusers_config(original_config, image_size: int):
@@ -632,7 +607,7 @@ def convert_ldm_vae_checkpoint(checkpoint, config):
else:
vae_state_dict = checkpoint
- new_checkpoint = convert_ldm_vae_state_dict(vae_state_dict,config)
+ new_checkpoint = convert_ldm_vae_state_dict(vae_state_dict, config)
return new_checkpoint
def convert_ldm_vae_state_dict(vae_state_dict, config):
@@ -855,7 +830,7 @@ def convert_ldm_bert_checkpoint(checkpoint, config):
def convert_ldm_clip_checkpoint(checkpoint):
- text_model = CLIPTextModel.from_pretrained(MODEL_ROOT / 'clip-vit-large-patch14')
+ text_model = CLIPTextModel.from_pretrained(CONVERT_MODEL_ROOT / 'clip-vit-large-patch14')
keys = list(checkpoint.keys())
text_model_dict = {}
@@ -909,7 +884,7 @@ textenc_pattern = re.compile("|".join(protected.keys()))
def convert_open_clip_checkpoint(checkpoint):
text_model = CLIPTextModel.from_pretrained(
- MODEL_ROOT / 'stable-diffusion-2-clip',
+ CONVERT_MODEL_ROOT / 'stable-diffusion-2-clip',
subfolder='text_encoder',
)
@@ -976,7 +951,7 @@ def convert_open_clip_checkpoint(checkpoint):
return text_model
-def replace_checkpoint_vae(checkpoint, vae_path:str):
+def replace_checkpoint_vae(checkpoint, vae_path: str):
if vae_path.endswith(".safetensors"):
vae_ckpt = load_file(vae_path)
else:
@@ -986,7 +961,7 @@ def replace_checkpoint_vae(checkpoint, vae_path:str):
new_key = f'first_stage_model.{vae_key}'
checkpoint[new_key] = state_dict[vae_key]
-def convert_ldm_vae_to_diffusers(checkpoint, vae_config: DictConfig, image_size: int)->AutoencoderKL:
+def convert_ldm_vae_to_diffusers(checkpoint, vae_config: DictConfig, image_size: int) -> AutoencoderKL:
vae_config = create_vae_diffusers_config(
vae_config, image_size=image_size
)
@@ -1006,8 +981,6 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
original_config_file: str,
extract_ema: bool = True,
precision: torch.dtype = torch.float32,
- upcast_attention: bool = False,
- prediction_type: SchedulerPredictionType = SchedulerPredictionType.Epsilon,
scan_needed: bool = True,
) -> StableDiffusionPipeline:
"""
@@ -1021,8 +994,6 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
:param checkpoint_path: Path to `.ckpt` file.
:param original_config_file: Path to `.yaml` config file corresponding to the original architecture.
If `None`, will be automatically inferred by looking for a key that only exists in SD2.0 models.
- :param prediction_type: The prediction type that the model was trained on. Use `'epsilon'` for Stable Diffusion
- v1.X and Stable Diffusion v2 Base. Use `'v-prediction'` for Stable Diffusion v2.
:param scheduler_type: Type of scheduler to use. Should be one of `["pndm", "lms", "heun", "euler",
"euler-ancestral", "dpm", "ddim"]`. :param model_type: The pipeline type. `None` to automatically infer, or one of
`["FrozenOpenCLIPEmbedder", "FrozenCLIPEmbedder"]`. :param extract_ema: Only relevant for
@@ -1030,21 +1001,23 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
or not. Defaults to `False`. Pass `True` to extract the EMA weights. EMA weights usually yield higher
quality images for inference. Non-EMA weights are usually better to continue fine-tuning.
:param precision: precision to use - torch.float16, torch.float32 or torch.autocast
- :param upcast_attention: Whether the attention computation should always be upcasted. This is necessary when
- running stable diffusion 2.1.
"""
- config = InvokeAIAppConfig.get_config()
+ if not isinstance(checkpoint_path, Path):
+ checkpoint_path = Path(checkpoint_path)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
verbosity = dlogging.get_verbosity()
dlogging.set_verbosity_error()
- if str(checkpoint_path).endswith(".safetensors"):
+ if checkpoint_path.suffix == ".safetensors":
checkpoint = load_file(checkpoint_path)
else:
if scan_needed:
- ModelCache.scan_model(checkpoint_path, checkpoint_path)
+ # scan model
+ scan_result = scan_file_path(checkpoint_path)
+ if scan_result.infected_files != 0:
+ raise "The model {checkpoint_path} is potentially infected by malware. Aborting import."
checkpoint = torch.load(checkpoint_path)
# sometimes there is a state_dict key and sometimes not
@@ -1053,9 +1026,13 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
original_config = OmegaConf.load(original_config_file)
- if model_version == BaseModelType.StableDiffusion2 and prediction_type == SchedulerPredictionType.VPrediction:
+ if model_version == BaseModelType.StableDiffusion2 and original_config["model"]["params"]["parameterization"] == "v":
+ prediction_type = "v_prediction"
+ upcast_attention = True
image_size = 768
else:
+ prediction_type = "epsilon"
+ upcast_attention = False
image_size = 512
#
@@ -1110,7 +1087,7 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
if model_type == "FrozenOpenCLIPEmbedder":
text_model = convert_open_clip_checkpoint(checkpoint)
tokenizer = CLIPTokenizer.from_pretrained(
- MODEL_ROOT / 'stable-diffusion-2-clip',
+ CONVERT_MODEL_ROOT / 'stable-diffusion-2-clip',
subfolder='tokenizer',
)
pipe = StableDiffusionPipeline(
@@ -1126,9 +1103,9 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
elif model_type in ["FrozenCLIPEmbedder", "WeightedFrozenCLIPEmbedder"]:
text_model = convert_ldm_clip_checkpoint(checkpoint)
- tokenizer = CLIPTokenizer.from_pretrained(MODEL_ROOT / 'clip-vit-large-patch14')
- safety_checker = StableDiffusionSafetyChecker.from_pretrained(MODEL_ROOT / 'stable-diffusion-safety-checker')
- feature_extractor = AutoFeatureExtractor.from_pretrained(MODEL_ROOT / 'stable-diffusion-safety-checker')
+ tokenizer = CLIPTokenizer.from_pretrained(CONVERT_MODEL_ROOT / 'clip-vit-large-patch14')
+ safety_checker = StableDiffusionSafetyChecker.from_pretrained(CONVERT_MODEL_ROOT / 'stable-diffusion-safety-checker')
+ feature_extractor = AutoFeatureExtractor.from_pretrained(CONVERT_MODEL_ROOT / 'stable-diffusion-safety-checker')
pipe = StableDiffusionPipeline(
vae=vae.to(precision),
text_encoder=text_model.to(precision),
@@ -1142,7 +1119,7 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
else:
text_config = create_ldm_bert_config(original_config)
text_model = convert_ldm_bert_checkpoint(checkpoint, text_config)
- tokenizer = BertTokenizerFast.from_pretrained(MODEL_ROOT / "bert-base-uncased")
+ tokenizer = BertTokenizerFast.from_pretrained(CONVERT_MODEL_ROOT / "bert-base-uncased")
pipe = LDMTextToImagePipeline(
vqvae=vae,
bert=text_model,
@@ -1158,7 +1135,6 @@ def load_pipeline_from_original_stable_diffusion_ckpt(
def convert_ckpt_to_diffusers(
checkpoint_path: Union[str, Path],
dump_path: Union[str, Path],
- model_root: Union[str, Path],
**kwargs,
):
"""
@@ -1166,9 +1142,6 @@ def convert_ckpt_to_diffusers(
and in addition a path-like object indicating the location of the desired diffusers
model to be written.
"""
- # setting global here to avoid massive changes late at night
- global MODEL_ROOT
- MODEL_ROOT = Path(model_root) / 'core/convert'
pipe = load_pipeline_from_original_stable_diffusion_ckpt(checkpoint_path, **kwargs)
pipe.save_pretrained(
diff --git a/invokeai/backend/model_management/lora.py b/invokeai/backend/model_management/lora.py
index b6187469c5..37e5a6fa40 100644
--- a/invokeai/backend/model_management/lora.py
+++ b/invokeai/backend/model_management/lora.py
@@ -1,9 +1,9 @@
from __future__ import annotations
import copy
-from pathlib import Path
from contextlib import contextmanager
-from typing import Optional, Dict, Tuple, Any
+from typing import Optional, Dict, Tuple, Any, Union, List
+from pathlib import Path
import torch
from safetensors.torch import load_file
@@ -16,6 +16,9 @@ from onnxruntime import OrtValue
import numpy as np
from compel.embeddings_provider import BaseTextualInversionManager
+from diffusers.models import UNet2DConditionModel
+from safetensors.torch import load_file
+from transformers import CLIPTextModel, CLIPTokenizer
# TODO: rename and split this file
@@ -129,8 +132,8 @@ class LoRALayer(LoRALayerBase):
def get_weight(self):
if self.mid is not None:
- up = self.up.reshape(up.shape[0], up.shape[1])
- down = self.down.reshape(up.shape[0], up.shape[1])
+ up = self.up.reshape(self.up.shape[0], self.up.shape[1])
+ down = self.down.reshape(self.down.shape[0], self.down.shape[1])
weight = torch.einsum("m n w h, i m, n j -> i j w h", self.mid, up, down)
else:
weight = self.up.reshape(self.up.shape[0], -1) @ self.down.reshape(self.down.shape[0], -1)
@@ -171,7 +174,7 @@ class LoHALayer(LoRALayerBase):
layer_key: str,
values: dict,
):
- super().__init__(module_key, rank, alpha, bias)
+ super().__init__(layer_key, values)
self.w1_a = values["hada_w1_a"]
self.w1_b = values["hada_w1_b"]
@@ -244,7 +247,7 @@ class LoKRLayer(LoRALayerBase):
layer_key: str,
values: dict,
):
- super().__init__(module_key, rank, alpha, bias)
+ super().__init__(layer_key, values)
if "lokr_w1" in values:
self.w1 = values["lokr_w1"]
@@ -291,7 +294,7 @@ class LoKRLayer(LoRALayerBase):
if len(w2.shape) == 4:
w1 = w1.unsqueeze(2).unsqueeze(2)
w2 = w2.contiguous()
- weight = torch.kron(w1, w2)#.reshape(module.weight.shape) # TODO: can we remove reshape?
+ weight = torch.kron(w1, w2)
return weight
@@ -416,7 +419,7 @@ class LoRAModel: #(torch.nn.Module):
else:
# TODO: diff/ia3/... format
print(
- f">> Encountered unknown lora layer module in {self.name}: {layer_key}"
+ f">> Encountered unknown lora layer module in {model.name}: {layer_key}"
)
return
@@ -476,7 +479,7 @@ class ModelPatcher:
submodule_name += "_" + key_parts.pop(0)
module = module.get_submodule(submodule_name)
- module_key = module_key.rstrip(".")
+ module_key = (module_key + "." + submodule_name).lstrip(".")
return (module_key, module)
@@ -530,23 +533,37 @@ class ModelPatcher:
loras: List[Tuple[LoraModel, float]],
prefix: str,
):
- hooks = dict()
+ original_weights = dict()
try:
- for lora, lora_weight in loras:
- for layer_key, layer in lora.layers.items():
- if not layer_key.startswith(prefix):
- continue
+ with torch.no_grad():
+ for lora, lora_weight in loras:
+ #assert lora.device.type == "cpu"
+ for layer_key, layer in lora.layers.items():
+ if not layer_key.startswith(prefix):
+ continue
- module_key, module = cls._resolve_lora_key(model, layer_key, prefix)
- if module_key not in hooks:
- hooks[module_key] = module.register_forward_hook(cls._lora_forward_hook(loras, layer_key))
+ module_key, module = cls._resolve_lora_key(model, layer_key, prefix)
+ if module_key not in original_weights:
+ original_weights[module_key] = module.weight.detach().to(device="cpu", copy=True)
+
+ # enable autocast to calc fp16 loras on cpu
+ #with torch.autocast(device_type="cpu"):
+ layer.to(dtype=torch.float32)
+ layer_scale = layer.alpha / layer.rank if (layer.alpha and layer.rank) else 1.0
+ layer_weight = layer.get_weight() * lora_weight * layer_scale
+
+ if module.weight.shape != layer_weight.shape:
+ # TODO: debug on lycoris
+ layer_weight = layer_weight.reshape(module.weight.shape)
+
+ module.weight += layer_weight.to(device=module.weight.device, dtype=module.weight.dtype)
yield # wait for context manager exit
finally:
- for module_key, hook in hooks.items():
- hook.remove()
- hooks.clear()
+ with torch.no_grad():
+ for module_key, weight in original_weights.items():
+ model.get_submodule(module_key).weight.copy_(weight)
@classmethod
@@ -596,7 +613,7 @@ class ModelPatcher:
f"Cannot load embedding for {trigger}. It was trained on a model with token dimension {embedding.shape[0]}, but the current model has token dimension {model_embeddings.weight.data[token_id].shape[0]}."
)
- model_embeddings.weight.data[token_id] = embedding
+ model_embeddings.weight.data[token_id] = embedding.to(device=text_encoder.device, dtype=text_encoder.dtype)
ti_tokens.append(token_id)
if len(ti_tokens) > 1:
@@ -609,6 +626,24 @@ class ModelPatcher:
text_encoder.resize_token_embeddings(init_tokens_count)
+ @classmethod
+ @contextmanager
+ def apply_clip_skip(
+ cls,
+ text_encoder: CLIPTextModel,
+ clip_skip: int,
+ ):
+ skipped_layers = []
+ try:
+ for i in range(clip_skip):
+ skipped_layers.append(text_encoder.text_model.encoder.layers.pop(-1))
+
+ yield
+
+ finally:
+ while len(skipped_layers) > 0:
+ text_encoder.text_model.encoder.layers.append(skipped_layers.pop())
+
class TextualInversionModel:
name: str
embedding: torch.Tensor # [n, 768]|[n, 1280]
@@ -647,6 +682,9 @@ class TextualInversionModel:
else:
result.embedding = next(iter(state_dict.values()))
+ if len(result.embedding.shape) == 1:
+ result.embedding = result.embedding.unsqueeze(0)
+
if not isinstance(result.embedding, torch.Tensor):
raise ValueError(f"Invalid embeddings file: {file_path.name}")
diff --git a/invokeai/backend/model_management/model_cache.py b/invokeai/backend/model_management/model_cache.py
index 77b6ac5115..e4cba3517e 100644
--- a/invokeai/backend/model_management/model_cache.py
+++ b/invokeai/backend/model_management/model_cache.py
@@ -8,7 +8,7 @@ The cache returns context manager generators designed to load the
model into the GPU within the context, and unload outside the
context. Use like this:
- cache = ModelCache(max_models_cached=6)
+ cache = ModelCache(max_cache_size=7.5)
with cache.get_model('runwayml/stable-diffusion-1-5') as SD1,
cache.get_model('stabilityai/stable-diffusion-2') as SD2:
do_something_in_GPU(SD1,SD2)
@@ -36,6 +36,9 @@ from .models import BaseModelType, ModelType, SubModelType, ModelBase
# Default is roughly enough to hold three fp16 diffusers models in RAM simultaneously
DEFAULT_MAX_CACHE_SIZE = 6.0
+# amount of GPU memory to hold in reserve for use by generations (GB)
+DEFAULT_MAX_VRAM_CACHE_SIZE= 2.75
+
# actual size of a gig
GIG = 1073741824
@@ -82,6 +85,7 @@ class ModelCache(object):
def __init__(
self,
max_cache_size: float=DEFAULT_MAX_CACHE_SIZE,
+ max_vram_cache_size: float=DEFAULT_MAX_VRAM_CACHE_SIZE,
execution_device: torch.device=torch.device('cuda'),
storage_device: torch.device=torch.device('cpu'),
precision: torch.dtype=torch.float16,
@@ -91,7 +95,7 @@ class ModelCache(object):
logger: types.ModuleType = logger
):
'''
- :param max_models: Maximum number of models to cache in CPU RAM [4]
+ :param max_cache_size: Maximum size of the RAM cache [6.0 GB]
:param execution_device: Torch device to load active model into [torch.device('cuda')]
:param storage_device: Torch device to save inactive model in [torch.device('cpu')]
:param precision: Precision for loaded models [torch.float16]
@@ -99,14 +103,11 @@ class ModelCache(object):
:param sequential_offload: Conserve VRAM by loading and unloading each stage of the pipeline sequentially
:param sha_chunksize: Chunksize to use when calculating sha256 model hash
'''
- #max_cache_size = 9999
- execution_device = torch.device('cuda')
-
self.model_infos: Dict[str, ModelBase] = dict()
self.lazy_offloading = lazy_offloading
- #self.sequential_offload: bool=sequential_offload
self.precision: torch.dtype=precision
- self.max_cache_size: int=max_cache_size
+ self.max_cache_size: float=max_cache_size
+ self.max_vram_cache_size: float=max_vram_cache_size
self.execution_device: torch.device=execution_device
self.storage_device: torch.device=storage_device
self.sha_chunksize=sha_chunksize
@@ -128,16 +129,6 @@ class ModelCache(object):
key += f":{submodel_type}"
return key
- #def get_model(
- # self,
- # repo_id_or_path: Union[str, Path],
- # model_type: ModelType = ModelType.Diffusers,
- # subfolder: Path = None,
- # submodel: ModelType = None,
- # revision: str = None,
- # attach_model_part: Tuple[ModelType, str] = (None, None),
- # gpu_load: bool = True,
- #) -> ModelLocker: # ?? what does it return
def _get_model_info(
self,
model_path: str,
@@ -213,14 +204,22 @@ class ModelCache(object):
self._cache_stack.remove(key)
self._cache_stack.append(key)
- return self.ModelLocker(self, key, cache_entry.model, gpu_load)
+ return self.ModelLocker(self, key, cache_entry.model, gpu_load, cache_entry.size)
class ModelLocker(object):
- def __init__(self, cache, key, model, gpu_load):
+ def __init__(self, cache, key, model, gpu_load, size_needed):
+ '''
+ :param cache: The model_cache object
+ :param key: The key of the model to lock in GPU
+ :param model: The model to lock
+ :param gpu_load: True if load into gpu
+ :param size_needed: Size of the model to load
+ '''
self.gpu_load = gpu_load
self.cache = cache
self.key = key
self.model = model
+ self.size_needed = size_needed
self.cache_entry = self.cache._cached_models[self.key]
def __enter__(self) -> Any:
@@ -234,7 +233,7 @@ class ModelCache(object):
try:
if self.cache.lazy_offloading:
- self.cache._offload_unlocked_models()
+ self.cache._offload_unlocked_models(self.size_needed)
if self.model.device != self.cache.execution_device:
self.cache.logger.debug(f'Moving {self.key} into {self.cache.execution_device}')
@@ -349,12 +348,20 @@ class ModelCache(object):
self.logger.debug(f"After unloading: cached_models={len(self._cached_models)}")
-
- def _offload_unlocked_models(self):
- for model_key, cache_entry in self._cached_models.items():
+ def _offload_unlocked_models(self, size_needed: int=0):
+ reserved = self.max_vram_cache_size * GIG
+ vram_in_use = torch.cuda.memory_allocated()
+ self.logger.debug(f'{(vram_in_use/GIG):.2f}GB VRAM used for models; max allowed={(reserved/GIG):.2f}GB')
+ for model_key, cache_entry in sorted(self._cached_models.items(), key=lambda x:x[1].size):
+ if vram_in_use <= reserved:
+ break
if not cache_entry.locked and cache_entry.loaded:
self.logger.debug(f'Offloading {model_key} from {self.execution_device} into {self.storage_device}')
- cache_entry.model.to(self.storage_device)
+ with VRAMUsage() as mem:
+ cache_entry.model.to(self.storage_device)
+ self.logger.debug(f'GPU VRAM freed: {(mem.vram_used/GIG):.2f} GB')
+ vram_in_use += mem.vram_used # note vram_used is negative
+ self.logger.debug(f'{(vram_in_use/GIG):.2f}GB VRAM used for models; max allowed={(reserved/GIG):.2f}GB')
def _local_model_hash(self, model_path: Union[str, Path]) -> str:
sha = hashlib.sha256()
diff --git a/invokeai/backend/model_management/model_install.py b/invokeai/backend/model_management/model_install.py
deleted file mode 100644
index 64c52185f3..0000000000
--- a/invokeai/backend/model_management/model_install.py
+++ /dev/null
@@ -1,118 +0,0 @@
-"""
-Routines for downloading and installing models.
-"""
-import json
-import safetensors
-import safetensors.torch
-import shutil
-import tempfile
-import torch
-import traceback
-from dataclasses import dataclass
-from diffusers import ModelMixin
-from enum import Enum
-from typing import Callable
-from pathlib import Path
-
-import invokeai.backend.util.logging as logger
-from invokeai.app.services.config import InvokeAIAppConfig
-from . import ModelManager
-from .models import BaseModelType, ModelType, VariantType
-from .model_probe import ModelProbe, ModelVariantInfo
-from .model_cache import SilenceWarnings
-
-class ModelInstall(object):
- '''
- This class is able to download and install several different kinds of
- InvokeAI models. The helper function, if provided, is called on to distinguish
- between v2-base and v2-768 stable diffusion pipelines. This usually involves
- asking the user to select the proper type, as there is no way of distinguishing
- the two type of v2 file programmatically (as far as I know).
- '''
- def __init__(self,
- config: InvokeAIAppConfig,
- model_base_helper: Callable[[Path],BaseModelType]=None,
- clobber:bool = False
- ):
- '''
- :param config: InvokeAI configuration object
- :param model_base_helper: A function call that accepts the Path to a checkpoint model and returns a ModelType enum
- :param clobber: If true, models with colliding names will be overwritten
- '''
- self.config = config
- self.clogger = clobber
- self.helper = model_base_helper
- self.prober = ModelProbe()
-
- def install_checkpoint_file(self, checkpoint: Path)->dict:
- '''
- Install the checkpoint file at path and return a
- configuration entry that can be added to `models.yaml`.
- Model checkpoints and VAEs will be converted into
- diffusers before installation. Note that the model manager
- does not hold entries for anything but diffusers pipelines,
- and the configuration file stanzas returned from such models
- can be safely ignored.
- '''
- model_info = self.prober.probe(checkpoint, self.helper)
- if not model_info:
- raise ValueError(f"Unable to determine type of checkpoint file {checkpoint}")
-
- key = ModelManager.create_key(
- model_name = checkpoint.stem,
- base_model = model_info.base_type,
- model_type = model_info.model_type,
- )
- destination_path = self._dest_path(model_info) / checkpoint
- destination_path.parent.mkdir(parents=True, exist_ok=True)
- self._check_for_collision(destination_path)
- stanza = {
- key: dict(
- name = checkpoint.stem,
- description = f'{model_info.model_type} model {checkpoint.stem}',
- base = model_info.base_model.value,
- type = model_info.model_type.value,
- variant = model_info.variant_type.value,
- path = str(destination_path),
- )
- }
-
- # non-pipeline; no conversion needed, just copy into right place
- if model_info.model_type != ModelType.Pipeline:
- shutil.copyfile(checkpoint, destination_path)
- stanza[key].update({'format': 'checkpoint'})
-
- # pipeline - conversion needed here
- else:
- destination_path = self._dest_path(model_info) / checkpoint.stem
- config_file = self._pipeline_type_to_config_file(model_info.model_type)
-
- from .convert_ckpt_to_diffusers import convert_ckpt_to_diffusers
- with SilenceWarnings:
- convert_ckpt_to_diffusers(
- checkpoint,
- destination_path,
- extract_ema=True,
- original_config_file=config_file,
- scan_needed=False,
- )
- stanza[key].update({'format': 'folder',
- 'path': destination_path, # no suffix on this
- })
-
- return stanza
-
-
- def _check_for_collision(self, path: Path):
- if not path.exists():
- return
- if self.clobber:
- shutil.rmtree(path)
- else:
- raise ValueError(f"Destination {path} already exists. Won't overwrite unless clobber=True.")
-
- def _staging_directory(self)->tempfile.TemporaryDirectory:
- return tempfile.TemporaryDirectory(dir=self.config.root_path)
-
-
-
diff --git a/invokeai/backend/model_management/model_manager.py b/invokeai/backend/model_management/model_manager.py
index f9a66a87dd..0476425c8b 100644
--- a/invokeai/backend/model_management/model_manager.py
+++ b/invokeai/backend/model_management/model_manager.py
@@ -1,53 +1,209 @@
"""This module manages the InvokeAI `models.yaml` file, mapping
-symbolic diffusers model names to the paths and repo_ids used
-by the underlying `from_pretrained()` call.
+symbolic diffusers model names to the paths and repo_ids used by the
+underlying `from_pretrained()` call.
-For fetching models, use manager.get_model('symbolic name'). This will
-return a ModelInfo object that contains the following attributes:
-
- * context -- a context manager Generator that loads and locks the
- model into GPU VRAM and returns the model for use.
- See below for usage.
- * name -- symbolic name of the model
- * type -- SubModelType of the model
- * hash -- unique hash for the model
- * location -- path or repo_id of the model
- * revision -- revision of the model if coming from a repo id,
- e.g. 'fp16'
- * precision -- torch precision of the model
+SYNOPSIS:
-Typical usage:
+ mgr = ModelManager('/home/phi/invokeai/configs/models.yaml')
+ sd1_5 = mgr.get_model('stable-diffusion-v1-5',
+ model_type=ModelType.Main,
+ base_model=BaseModelType.StableDiffusion1,
+ submodel_type=SubModelType.Unet)
+ with sd1_5 as unet:
+ run_some_inference(unet)
- from invokeai.backend import ModelManager
+FETCHING MODELS:
- manager = ModelManager(
- config='./configs/models.yaml',
- max_cache_size=8
- ) # gigabytes
+Models are described using four attributes:
- model_info = manager.get_model('stable-diffusion-1.5', SubModelType.Diffusers)
- with model_info.context as my_model:
- my_model.latents_from_embeddings(...)
+ 1) model_name -- the symbolic name for the model
-The manager uses the underlying ModelCache class to keep
-frequently-used models in RAM and move them into GPU as needed for
-generation operations. The optional `max_cache_size` argument
-indicates the maximum size the cache can grow to, in gigabytes. The
-underlying ModelCache object can be accessed using the manager's "cache"
-attribute.
+ 2) ModelType -- an enum describing the type of the model. Currently
+ defined types are:
+ ModelType.Main -- a full model capable of generating images
+ ModelType.Vae -- a VAE model
+ ModelType.Lora -- a LoRA or LyCORIS fine-tune
+ ModelType.TextualInversion -- a textual inversion embedding
+ ModelType.ControlNet -- a ControlNet model
-Because the model manager can return multiple different types of
-models, you may wish to add additional type checking on the class
-of model returned. To do this, provide the option `model_type`
-parameter:
+ 3) BaseModelType -- an enum indicating the stable diffusion base model, one of:
+ BaseModelType.StableDiffusion1
+ BaseModelType.StableDiffusion2
- model_info = manager.get_model(
- 'clip-tokenizer',
- model_type=SubModelType.Tokenizer
- )
+ 4) SubModelType (optional) -- an enum that refers to one of the submodels contained
+ within the main model. Values are:
-This will raise an InvalidModelError if the format defined in the
-config file doesn't match the requested model type.
+ SubModelType.UNet
+ SubModelType.TextEncoder
+ SubModelType.Tokenizer
+ SubModelType.Scheduler
+ SubModelType.SafetyChecker
+
+To fetch a model, use `manager.get_model()`. This takes the symbolic
+name of the model, the ModelType, the BaseModelType and the
+SubModelType. The latter is required for ModelType.Main.
+
+get_model() will return a ModelInfo object that can then be used in
+context to retrieve the model and move it into GPU VRAM (on GPU
+systems).
+
+A typical example is:
+
+ sd1_5 = mgr.get_model('stable-diffusion-v1-5',
+ model_type=ModelType.Main,
+ base_model=BaseModelType.StableDiffusion1,
+ submodel_type=SubModelType.UNet)
+ with sd1_5 as unet:
+ run_some_inference(unet)
+
+The ModelInfo object provides a number of useful fields describing the
+model, including:
+
+ name -- symbolic name of the model
+ base_model -- base model (BaseModelType)
+ type -- model type (ModelType)
+ location -- path to the model file
+ precision -- torch precision of the model
+ hash -- unique sha256 checksum for this model
+
+SUBMODELS:
+
+When fetching a main model, you must specify the submodel. Retrieval
+of full pipelines is not supported.
+
+ vae_info = mgr.get_model('stable-diffusion-1.5',
+ model_type = ModelType.Main,
+ base_model = BaseModelType.StableDiffusion1,
+ submodel_type = SubModelType.Vae
+ )
+ with vae_info as vae:
+ do_something(vae)
+
+This rule does not apply to controlnets, embeddings, loras and standalone
+VAEs, which do not have submodels.
+
+LISTING MODELS
+
+The model_names() method will return a list of Tuples describing each
+model it knows about:
+
+ >> mgr.model_names()
+ [
+ ('stable-diffusion-1.5', , ),
+ ('stable-diffusion-2.1', , ),
+ ('inpaint', , )
+ ('Ink scenery', , )
+ ...
+ ]
+
+The tuple is in the correct order to pass to get_model():
+
+ for m in mgr.model_names():
+ info = get_model(*m)
+
+In contrast, the list_models() method returns a list of dicts, each
+providing information about a model defined in models.yaml. For example:
+
+ >>> models = mgr.list_models()
+ >>> json.dumps(models[0])
+ {"path": "/home/lstein/invokeai-main/models/sd-1/controlnet/canny",
+ "model_format": "diffusers",
+ "name": "canny",
+ "base_model": "sd-1",
+ "type": "controlnet"
+ }
+
+You can filter by model type and base model as shown here:
+
+
+ controlnets = mgr.list_models(model_type=ModelType.ControlNet,
+ base_model=BaseModelType.StableDiffusion1)
+ for c in controlnets:
+ name = c['name']
+ format = c['model_format']
+ path = c['path']
+ type = c['type']
+ # etc
+
+ADDING AND REMOVING MODELS
+
+At startup time, the `models` directory will be scanned for
+checkpoints, diffusers pipelines, controlnets, LoRAs and TI
+embeddings. New entries will be added to the model manager and defunct
+ones removed. Anything that is a main model (ModelType.Main) will be
+added to models.yaml. For scanning to succeed, files need to be in
+their proper places. For example, a controlnet folder built on the
+stable diffusion 2 base, will need to be placed in
+`models/sd-2/controlnet`.
+
+Layout of the `models` directory:
+
+ models
+ ├── sd-1
+ │ ├── controlnet
+ │ ├── lora
+ │ ├── main
+ │ └── embedding
+ ├── sd-2
+ │ ├── controlnet
+ │ ├── lora
+ │ ├── main
+ │ └── embedding
+ └── core
+ ├── face_reconstruction
+ │ ├── codeformer
+ │ └── gfpgan
+ ├── sd-conversion
+ │ ├── clip-vit-large-patch14 - tokenizer, text_encoder subdirs
+ │ ├── stable-diffusion-2 - tokenizer, text_encoder subdirs
+ │ └── stable-diffusion-safety-checker
+ └── upscaling
+ └─── esrgan
+
+
+
+class ConfigMeta(BaseModel):Loras, textual_inversion and controlnet models are not listed
+explicitly in models.yaml, but are added to the in-memory data
+structure at initialization time by scanning the models directory. The
+in-memory data structure can be resynchronized by calling
+`manager.scan_models_directory()`.
+
+Files and folders placed inside the `autoimport` paths (paths
+defined in `invokeai.yaml`) will also be scanned for new models at
+initialization time and added to `models.yaml`. Files will not be
+moved from this location but preserved in-place. These directories
+are:
+
+ configuration default description
+ ------------- ------- -----------
+ autoimport_dir autoimport/main main models
+ lora_dir autoimport/lora LoRA/LyCORIS models
+ embedding_dir autoimport/embedding TI embeddings
+ controlnet_dir autoimport/controlnet ControlNet models
+
+In actuality, models located in any of these directories are scanned
+to determine their type, so it isn't strictly necessary to organize
+the different types in this way. This entry in `invokeai.yaml` will
+recursively scan all subdirectories within `autoimport`, scan models
+files it finds, and import them if recognized.
+
+ Paths:
+ autoimport_dir: autoimport
+
+A model can be manually added using `add_model()` using the model's
+name, base model, type and a dict of model attributes. See
+`invokeai/backend/model_management/models` for the attributes required
+by each model type.
+
+A model can be deleted using `del_model()`, providing the same
+identifying information as `get_model()`
+
+The `heuristic_import()` method will take a set of strings
+corresponding to local paths, remote URLs, and repo_ids, probe the
+object to determine what type of model it is (if any), and import new
+models into the manager. If passed a directory, it will recursively
+scan it for models to import. The return value is a set of the models
+successfully added.
MODELS.YAML
@@ -56,93 +212,18 @@ The general format of a models.yaml section is:
type-of-model/name-of-model:
path: /path/to/local/file/or/directory
description: a description
- format: folder|ckpt|safetensors|pt
- base: SD-1|SD-2
- subfolder: subfolder-name
+ format: diffusers|checkpoint
+ variant: normal|inpaint|depth
The type of model is given in the stanza key, and is one of
-{diffusers, ckpt, vae, text_encoder, tokenizer, unet, scheduler,
-safety_checker, feature_extractor, lora, textual_inversion,
-controlnet}, and correspond to items in the SubModelType enum defined
-in model_cache.py
+{main, vae, lora, controlnet, textual}
-The format indicates whether the model is organized as a folder with
-model subdirectories, or is contained in a single checkpoint or
-safetensors file.
+The format indicates whether the model is organized as a diffusers
+folder with model subdirectories, or is contained in a single
+checkpoint or safetensors file.
-One, but not both, of repo_id and path are provided. repo_id is the
-HuggingFace repository ID of the model, and path points to the file or
-directory on disk.
-
-If subfolder is provided, then the model exists in a subdirectory of
-the main model. These are usually named after the model type, such as
-"unet".
-
-This example summarizes the two ways of getting a non-diffuser model:
-
- text_encoder/clip-test-1:
- format: folder
- path: /path/to/folder
- description: Returns standalone CLIPTextModel
-
- text_encoder/clip-test-2:
- format: folder
- repo_id: /path/to/folder
- subfolder: text_encoder
- description: Returns the text_encoder in the subfolder of the diffusers model (just the encoder in RAM)
-
-SUBMODELS:
-
-It is also possible to fetch an isolated submodel from a diffusers
-model. Use the `submodel` parameter to select which part:
-
- vae = manager.get_model('stable-diffusion-1.5',submodel=SubModelType.Vae)
- with vae.context as my_vae:
- print(type(my_vae))
- # "AutoencoderKL"
-
-DIRECTORY_SCANNING:
-
-Loras, textual_inversion and controlnet models are usually not listed
-explicitly in models.yaml, but are added to the in-memory data
-structure at initialization time by scanning the models directory. The
-in-memory data structure can be resynchronized by calling
-`manager.scan_models_directory`.
-
-DISAMBIGUATION:
-
-You may wish to use the same name for a related family of models. To
-do this, disambiguate the stanza key with the model and and format
-separated by "/". Example:
-
- tokenizer/clip-large:
- format: tokenizer
- path: /path/to/folder
- description: Returns standalone tokenizer
-
- text_encoder/clip-large:
- format: text_encoder
- path: /path/to/folder
- description: Returns standalone text encoder
-
-You can now use the `model_type` argument to indicate which model you
-want:
-
- tokenizer = mgr.get('clip-large',model_type=SubModelType.Tokenizer)
- encoder = mgr.get('clip-large',model_type=SubModelType.TextEncoder)
-
-OTHER FUNCTIONS:
-
-Other methods provided by ModelManager support importing, editing,
-converting and deleting models.
-
-IMPORTANT CHANGES AND LIMITATIONS SINCE 2.3:
-
-1. Only local paths are supported. Repo_ids are no longer accepted. This
-simplifies the logic.
-
-2. VAEs can't be swapped in and out at load time. They must be baked
-into the model when downloaded or converted.
+The path points to a file or directory on disk. If a relative path,
+the root is the InvokeAI ROOTDIR.
"""
from __future__ import annotations
@@ -150,24 +231,27 @@ from __future__ import annotations
import os
import hashlib
import textwrap
+import yaml
from dataclasses import dataclass
-from packaging import version
from pathlib import Path
-from typing import Dict, Optional, List, Tuple, Union, types
-from shutil import rmtree
+from typing import Optional, List, Tuple, Union, Dict, Set, Callable, types
+from shutil import rmtree, move
import torch
-from huggingface_hub import scan_cache_dir
from omegaconf import OmegaConf
from omegaconf.dictconfig import DictConfig
-from pydantic import BaseModel
+from pydantic import BaseModel, Field
import invokeai.backend.util.logging as logger
from invokeai.app.services.config import InvokeAIAppConfig
-from invokeai.backend.util import CUDA_DEVICE, download_with_resume
+from invokeai.backend.util import CUDA_DEVICE, Chdir
from .model_cache import ModelCache, ModelLocker
-from .models import BaseModelType, ModelType, SubModelType, ModelError, MODEL_CLASSES
+from .models import (
+ BaseModelType, ModelType, SubModelType,
+ ModelError, SchedulerPredictionType, MODEL_CLASSES,
+ ModelConfigBase, ModelNotFoundException, InvalidModelException,
+)
# We are only starting to number the config file with release 3.
# The config file version doesn't have to start at release version, but it will help
@@ -183,7 +267,6 @@ class ModelInfo():
hash: str
location: Union[Path, str]
precision: torch.dtype
- revision: str = None
_cache: ModelCache = None
def __enter__(self):
@@ -192,38 +275,14 @@ class ModelInfo():
def __exit__(self,*args, **kwargs):
self.context.__exit__(*args, **kwargs)
-class InvalidModelError(Exception):
- "Raised when an invalid model is requested"
- pass
+class AddModelResult(BaseModel):
+ name: str = Field(description="The name of the model after installation")
+ model_type: ModelType = Field(description="The type of model")
+ base_model: BaseModelType = Field(description="The base model")
+ config: ModelConfigBase = Field(description="The configuration of the model")
MAX_CACHE_SIZE = 6.0 # GB
-
-# layout of the models directory:
-# models
-# ├── sd-1
-# │ ├── controlnet
-# │ ├── lora
-# │ ├── pipeline
-# │ └── textual_inversion
-# ├── sd-2
-# │ ├── controlnet
-# │ ├── lora
-# │ ├── pipeline
-# │ └── textual_inversion
-# └── core
-# ├── face_reconstruction
-# │ ├── codeformer
-# │ └── gfpgan
-# ├── sd-conversion
-# │ ├── clip-vit-large-patch14 - tokenizer, text_encoder subdirs
-# │ ├── stable-diffusion-2 - tokenizer, text_encoder subdirs
-# │ └── stable-diffusion-safety-checker
-# └── upscaling
-# └─── esrgan
-
-
-
class ConfigMeta(BaseModel):
version: str
@@ -249,10 +308,12 @@ class ModelManager(object):
and sequential_offload boolean. Note that the default device
type and precision are set up for a CUDA system running at half precision.
"""
-
self.config_path = None
if isinstance(config, (str, Path)):
self.config_path = Path(config)
+ if not self.config_path.exists():
+ logger.warning(f'The file {self.config_path} was not found. Initializing a new file')
+ self.initialize_model_config(self.config_path)
config = OmegaConf.load(self.config_path)
elif not isinstance(config, DictConfig):
@@ -271,10 +332,11 @@ class ModelManager(object):
self.models[model_key] = model_class.create_config(**model_config)
# check config version number and update on disk/RAM if necessary
- self.globals = InvokeAIAppConfig.get_config()
+ self.app_config = InvokeAIAppConfig.get_config()
self.logger = logger
self.cache = ModelCache(
max_cache_size=max_cache_size,
+ max_vram_cache_size = self.app_config.max_vram_cache_size,
execution_device = device_type,
precision = precision,
sequential_offload = sequential_offload,
@@ -307,7 +369,8 @@ class ModelManager(object):
) -> str:
return f"{base_model}/{model_type}/{model_name}"
- def parse_key(self, model_key: str) -> Tuple[str, BaseModelType, ModelType]:
+ @classmethod
+ def parse_key(cls, model_key: str) -> Tuple[str, BaseModelType, ModelType]:
base_model_str, model_type_str, model_name = model_key.split('/', 2)
try:
model_type = ModelType(model_type_str)
@@ -321,103 +384,72 @@ class ModelManager(object):
return (model_name, base_model, model_type)
+ def _get_model_cache_path(self, model_path):
+ return self.app_config.models_path / ".cache" / hashlib.md5(str(model_path).encode()).hexdigest()
+
+ @classmethod
+ def initialize_model_config(cls, config_path: Path):
+ """Create empty config file"""
+ with open(config_path,'w') as yaml_file:
+ yaml_file.write(yaml.dump({'__metadata__':
+ {'version':'3.0.0'}
+ }
+ )
+ )
+
def get_model(
self,
model_name: str,
base_model: BaseModelType,
model_type: ModelType,
submodel_type: Optional[SubModelType] = None
- ):
+ )->ModelInfo:
"""Given a model named identified in models.yaml, return
an ModelInfo object describing it.
:param model_name: symbolic name of the model in models.yaml
:param model_type: ModelType enum indicating the type of model to return
+ :param base_model: BaseModelType enum indicating the base model used by this model
:param submode_typel: an ModelType enum indicating the portion of
the model to retrieve (e.g. ModelType.Vae)
-
- If not provided, the model_type will be read from the `format` field
- of the corresponding stanza. If provided, the model_type will be used
- to disambiguate stanzas in the configuration file. The default is to
- assume a diffusers pipeline. The behavior is illustrated here:
-
- [models.yaml]
- diffusers/test1:
- repo_id: foo/bar
- description: Typical diffusers pipeline
-
- lora/test1:
- repo_id: /tmp/loras/test1.safetensors
- description: Typical lora file
-
- test1_pipeline = mgr.get_model('test1')
- # returns a StableDiffusionGeneratorPipeline
-
- test1_vae1 = mgr.get_model('test1', submodel=ModelType.Vae)
- # returns the VAE part of a diffusers model as an AutoencoderKL
-
- test1_vae2 = mgr.get_model('test1', model_type=ModelType.Diffusers, submodel=ModelType.Vae)
- # does the same thing as the previous statement. Note that model_type
- # is for the parent model, and submodel is for the part
-
- test1_lora = mgr.get_model('test1', model_type=ModelType.Lora)
- # returns a LoRA embed (as a 'dict' of tensors)
-
- test1_encoder = mgr.get_modelI('test1', model_type=ModelType.TextEncoder)
- # raises an InvalidModelError
-
"""
model_class = MODEL_CLASSES[base_model][model_type]
model_key = self.create_key(model_name, base_model, model_type)
# if model not found try to find it (maybe file just pasted)
if model_key not in self.models:
- # TODO: find by mask or try rescan?
- path_mask = f"/models/{base_model}/{model_type}/{model_name}*"
- if False: # model_path = next(find_by_mask(path_mask)):
- model_path = None # TODO:
- model_config = model_class.probe_config(model_path)
- self.models[model_key] = model_config
- else:
- raise Exception(f"Model not found - {model_key}")
-
- # if it known model check that target path exists (if manualy deleted)
- else:
- # logic repeated twice(in rescan too) any way to optimize?
- if not os.path.exists(self.models[model_key].path):
- if model_class.save_to_config:
- self.models[model_key].error = ModelError.NotFound
- raise Exception(f"Files for model \"{model_key}\" not found")
-
- else:
- self.models.pop(model_key, None)
- raise Exception(f"Model not found - {model_key}")
-
- # reset model errors?
-
-
+ self.scan_models_directory(base_model=base_model, model_type=model_type)
+ if model_key not in self.models:
+ raise ModelNotFoundException(f"Model not found - {model_key}")
model_config = self.models[model_key]
+ model_path = self.app_config.root_path / model_config.path
- # /models/{base_model}/{model_type}/{name}.ckpt or .safentesors
- # /models/{base_model}/{model_type}/{name}/
- model_path = model_config.path
+ if not model_path.exists():
+ if model_class.save_to_config:
+ self.models[model_key].error = ModelError.NotFound
+ raise Exception(f"Files for model \"{model_key}\" not found")
+
+ else:
+ self.models.pop(model_key, None)
+ raise ModelNotFoundException(f"Model not found - {model_key}")
# vae/movq override
# TODO:
if submodel_type is not None and hasattr(model_config, submodel_type):
override_path = getattr(model_config, submodel_type)
if override_path:
- model_path = override_path
+ model_path = self.app_config.root_path / override_path
model_type = submodel_type
submodel_type = None
model_class = MODEL_CLASSES[base_model][model_type]
# TODO: path
# TODO: is it accurate to use path as id
- dst_convert_path = self.globals.models_dir / ".cache" / hashlib.md5(model_path.encode()).hexdigest()
+ dst_convert_path = self._get_model_cache_path(model_path)
+
model_path = model_class.convert_if_required(
base_model=base_model,
- model_path=model_path,
+ model_path=str(model_path), # TODO: refactor str/Path types logic
output_path=dst_convert_path,
config=model_config,
)
@@ -469,22 +501,32 @@ class ModelManager(object):
"""
return [(self.parse_key(x)) for x in self.models.keys()]
+ def list_model(
+ self,
+ model_name: str,
+ base_model: BaseModelType,
+ model_type: ModelType,
+ ) -> dict:
+ """
+ Returns a dict describing one installed model, using
+ the combined format of the list_models() method.
+ """
+ models = self.list_models(base_model,model_type,model_name)
+ return models[0] if models else None
+
def list_models(
self,
base_model: Optional[BaseModelType] = None,
model_type: Optional[ModelType] = None,
+ model_name: Optional[str] = None,
) -> list[dict]:
"""
Return a list of models.
-
- Please use model_manager.models() to get all the model names,
- model_manager.model_info('model-name') to get the stanza for the model
- named 'model-name', and model_manager.config to get the full OmegaConf
- object derived from models.yaml
"""
+ model_keys = [self.create_key(model_name, base_model, model_type)] if model_name else sorted(self.models, key=str.casefold)
models = []
- for model_key in sorted(self.models, key=str.casefold):
+ for model_key in model_keys:
model_config = self.models[model_key]
cur_model_name, cur_base_model, cur_model_type = self.parse_key(model_key)
@@ -496,9 +538,9 @@ class ModelManager(object):
model_dict = dict(
**model_config.dict(exclude_defaults=True),
# OpenAPIModelInfoBase
- name=cur_model_name,
+ model_name=cur_model_name,
base_model=cur_base_model,
- type=cur_model_type,
+ model_type=cur_model_type,
)
models.append(model_dict)
@@ -507,7 +549,7 @@ class ModelManager(object):
def print_models(self) -> None:
"""
- Print a table of models, their descriptions
+ Print a table of models and their descriptions. This needs to be redone
"""
# TODO: redo
for model_type, model_dict in self.list_models().items():
@@ -515,7 +557,7 @@ class ModelManager(object):
line = f'{model_info["name"]:25s} {model_info["type"]:10s} {model_info["description"]}'
print(line)
- # TODO: test when ui implemented
+ # Tested - LS
def del_model(
self,
model_name: str,
@@ -525,15 +567,11 @@ class ModelManager(object):
"""
Delete the named model.
"""
- raise Exception("TODO: del_model") # TODO: redo
model_key = self.create_key(model_name, base_model, model_type)
model_cfg = self.models.pop(model_key, None)
if model_cfg is None:
- self.logger.error(
- f"Unknown model {model_key}"
- )
- return
+ raise KeyError(f"Unknown model {model_key}")
# note: it not garantie to release memory(model can has other references)
cache_ids = self.cache_keys.pop(model_key, [])
@@ -541,14 +579,18 @@ class ModelManager(object):
self.cache.uncache_model(cache_id)
# if model inside invoke models folder - delete files
- if model_cfg.path.startswith("models/") or model_cfg.path.startswith("models\\"):
- model_path = self.globals.root_dir / model_cfg.path
- if model_path.isdir():
- shutil.rmtree(str(model_path))
+ model_path = self.app_config.root_path / model_cfg.path
+ cache_path = self._get_model_cache_path(model_path)
+ if cache_path.exists():
+ rmtree(str(cache_path))
+
+ if model_path.is_relative_to(self.app_config.models_path):
+ if model_path.is_dir():
+ rmtree(str(model_path))
else:
model_path.unlink()
- # TODO: test when ui implemented
+ # LS: tested
def add_model(
self,
model_name: str,
@@ -556,31 +598,107 @@ class ModelManager(object):
model_type: ModelType,
model_attributes: dict,
clobber: bool = False,
- ) -> None:
+ ) -> AddModelResult:
"""
Update the named model with a dictionary of attributes. Will fail with an
assertion error if the name already exists. Pass clobber=True to overwrite.
On a successful update, the config will be changed in memory and the
method will return True. Will fail with an assertion error if provided
attributes are incorrect or the model name is missing.
+
+ The returned dict has the same format as the dict returned by
+ model_info().
"""
model_class = MODEL_CLASSES[base_model][model_type]
model_config = model_class.create_config(**model_attributes)
model_key = self.create_key(model_name, base_model, model_type)
- assert (
- clobber or model_key not in self.models
- ), f'attempt to overwrite existing model definition "{model_key}"'
+ if model_key in self.models and not clobber:
+ raise Exception(f'Attempt to overwrite existing model definition "{model_key}"')
- self.models[model_key] = model_config
-
- if clobber and model_key in self.cache_keys:
- # note: it not garantie to release memory(model can has other references)
+ old_model = self.models.pop(model_key, None)
+ if old_model is not None:
+ # TODO: if path changed and old_model.path inside models folder should we delete this too?
+
+ # remove conversion cache as config changed
+ old_model_path = self.app_config.root_path / old_model.path
+ old_model_cache = self._get_model_cache_path(old_model_path)
+ if old_model_cache.exists():
+ if old_model_cache.is_dir():
+ rmtree(str(old_model_cache))
+ else:
+ old_model_cache.unlink()
+
+ # remove in-memory cache
+ # note: it not guaranteed to release memory(model can has other references)
cache_ids = self.cache_keys.pop(model_key, [])
for cache_id in cache_ids:
self.cache.uncache_model(cache_id)
+ self.models[model_key] = model_config
+ self.commit()
+ return AddModelResult(
+ name = model_name,
+ model_type = model_type,
+ base_model = base_model,
+ config = model_config,
+ )
+
+ def convert_model (
+ self,
+ model_name: str,
+ base_model: BaseModelType,
+ model_type: Union[ModelType.Main,ModelType.Vae],
+ ) -> AddModelResult:
+ '''
+ Convert a checkpoint file into a diffusers folder, deleting the cached
+ version and deleting the original checkpoint file if it is in the models
+ directory.
+ :param model_name: Name of the model to convert
+ :param base_model: Base model type
+ :param model_type: Type of model ['vae' or 'main']
+
+ This will raise a ValueError unless the model is a checkpoint.
+ '''
+ info = self.model_info(model_name, base_model, model_type)
+ if info["model_format"] != "checkpoint":
+ raise ValueError(f"not a checkpoint format model: {model_name}")
+
+ # We are taking advantage of a side effect of get_model() that converts check points
+ # into cached diffusers directories stored at `location`. It doesn't matter
+ # what submodeltype we request here, so we get the smallest.
+ submodel = {"submodel_type": SubModelType.Tokenizer} if model_type==ModelType.Main else {}
+ model = self.get_model(model_name,
+ base_model,
+ model_type,
+ **submodel,
+ )
+ checkpoint_path = self.app_config.root_path / info["path"]
+ old_diffusers_path = self.app_config.models_path / model.location
+ new_diffusers_path = self.app_config.models_path / base_model.value / model_type.value / model_name
+ if new_diffusers_path.exists():
+ raise ValueError(f"A diffusers model already exists at {new_diffusers_path}")
+
+ try:
+ move(old_diffusers_path,new_diffusers_path)
+ info["model_format"] = "diffusers"
+ info["path"] = str(new_diffusers_path.relative_to(self.app_config.root_path))
+ info.pop('config')
+
+ result = self.add_model(model_name, base_model, model_type,
+ model_attributes = info,
+ clobber=True)
+ except:
+ # something went wrong, so don't leave dangling diffusers model in directory or it will cause a duplicate model error!
+ rmtree(new_diffusers_path)
+ raise
+
+ if checkpoint_path.exists() and checkpoint_path.is_relative_to(self.app_config.models_path):
+ checkpoint_path.unlink()
+
+ return result
+
def search_models(self, search_folder):
self.logger.info(f"Finding Models In: {search_folder}")
models_folder_ckpt = Path(search_folder).glob("**/*.ckpt")
@@ -621,7 +739,7 @@ class ModelManager(object):
yaml_str = OmegaConf.to_yaml(data_to_save)
config_file_path = conf_file or self.config_path
assert config_file_path is not None,'no config file path to write to'
- config_file_path = self.globals.root_dir / config_file_path
+ config_file_path = self.app_config.root_path / config_file_path
tmpfile = os.path.join(os.path.dirname(config_file_path), "new_config.tmp")
with open(tmpfile, "w", encoding="utf-8") as outfile:
outfile.write(self.preamble())
@@ -644,42 +762,165 @@ class ModelManager(object):
"""
)
- def scan_models_directory(self):
+ def scan_models_directory(
+ self,
+ base_model: Optional[BaseModelType] = None,
+ model_type: Optional[ModelType] = None,
+ ):
+
loaded_files = set()
new_models_found = False
- for model_key, model_config in list(self.models.items()):
- model_name, base_model, model_type = self.parse_key(model_key)
- model_path = str(self.globals.root / model_config.path)
- if not os.path.exists(model_path):
- model_class = MODEL_CLASSES[base_model][model_type]
- if model_class.save_to_config:
- model_config.error = ModelError.NotFound
+ self.logger.info(f'scanning {self.app_config.models_path} for new models')
+ with Chdir(self.app_config.root_path):
+ for model_key, model_config in list(self.models.items()):
+ model_name, cur_base_model, cur_model_type = self.parse_key(model_key)
+ model_path = self.app_config.root_path.absolute() / model_config.path
+ if not model_path.exists():
+ model_class = MODEL_CLASSES[cur_base_model][cur_model_type]
+ if model_class.save_to_config:
+ model_config.error = ModelError.NotFound
+ self.models.pop(model_key, None)
+ else:
+ self.models.pop(model_key, None)
else:
- self.models.pop(model_key, None)
- else:
- loaded_files.add(model_path)
+ loaded_files.add(model_path)
- for base_model in BaseModelType:
- for model_type in ModelType:
- model_class = MODEL_CLASSES[base_model][model_type]
- models_dir = os.path.join(self.globals.models_path, base_model, model_type)
+ for cur_base_model in BaseModelType:
+ if base_model is not None and cur_base_model != base_model:
+ continue
- if not os.path.exists(models_dir):
- continue # TODO: or create all folders?
-
- for entry_name in os.listdir(models_dir):
- model_path = os.path.join(models_dir, entry_name)
- if model_path not in loaded_files: # TODO: check
- model_name = Path(model_path).stem
- model_key = self.create_key(model_name, base_model, model_type)
+ for cur_model_type in ModelType:
+ if model_type is not None and cur_model_type != model_type:
+ continue
+ model_class = MODEL_CLASSES[cur_base_model][cur_model_type]
+ models_dir = self.app_config.models_path / cur_base_model.value / cur_model_type.value
- if model_key in self.models:
- raise Exception(f"Model with key {model_key} added twice")
+ if not models_dir.exists():
+ continue # TODO: or create all folders?
- model_config: ModelConfigBase = model_class.probe_config(model_path)
- self.models[model_key] = model_config
- new_models_found = True
+ for model_path in models_dir.iterdir():
+ if model_path not in loaded_files: # TODO: check
+ model_name = model_path.name if model_path.is_dir() else model_path.stem
+ model_key = self.create_key(model_name, cur_base_model, cur_model_type)
- if new_models_found:
+ if model_key in self.models:
+ raise Exception(f"Model with key {model_key} added twice")
+
+ if model_path.is_relative_to(self.app_config.root_path):
+ model_path = model_path.relative_to(self.app_config.root_path)
+ try:
+ model_config: ModelConfigBase = model_class.probe_config(str(model_path))
+ self.models[model_key] = model_config
+ new_models_found = True
+ except InvalidModelException:
+ self.logger.warning(f"Not a valid model: {model_path}")
+ except NotImplementedError as e:
+ self.logger.warning(e)
+
+ imported_models = self.autoimport()
+
+ if (new_models_found or imported_models) and self.config_path:
self.commit()
+
+ def autoimport(self)->Dict[str, AddModelResult]:
+ '''
+ Scan the autoimport directory (if defined) and import new models, delete defunct models.
+ '''
+ # avoid circular import
+ from invokeai.backend.install.model_install_backend import ModelInstall
+ from invokeai.frontend.install.model_install import ask_user_for_prediction_type
+
+ installer = ModelInstall(config = self.app_config,
+ model_manager = self,
+ prediction_type_helper = ask_user_for_prediction_type,
+ )
+
+ scanned_dirs = set()
+
+ config = self.app_config
+ known_paths = {(self.app_config.root_path / x['path']) for x in self.list_models()}
+
+ for autodir in [config.autoimport_dir,
+ config.lora_dir,
+ config.embedding_dir,
+ config.controlnet_dir]:
+ if autodir is None:
+ continue
+
+ self.logger.info(f'Scanning {autodir} for models to import')
+ installed = dict()
+
+ autodir = self.app_config.root_path / autodir
+ if not autodir.exists():
+ continue
+
+ items_scanned = 0
+ new_models_found = dict()
+
+ for root, dirs, files in os.walk(autodir):
+ items_scanned += len(dirs) + len(files)
+ for d in dirs:
+ path = Path(root) / d
+ if path in known_paths or path.parent in scanned_dirs:
+ 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'}]):
+ try:
+ new_models_found.update(installer.heuristic_import(path))
+ scanned_dirs.add(path)
+ except ValueError as e:
+ self.logger.warning(str(e))
+
+ for f in files:
+ path = Path(root) / f
+ if path in known_paths or path.parent in scanned_dirs:
+ continue
+ if path.suffix in {'.ckpt','.bin','.pth','.safetensors','.pt'}:
+ try:
+ import_result = installer.heuristic_import(path)
+ new_models_found.update(import_result)
+ except ValueError as e:
+ self.logger.warning(str(e))
+
+ self.logger.info(f'Scanned {items_scanned} files and directories, imported {len(new_models_found)} models')
+ installed.update(new_models_found)
+
+ return installed
+
+ def heuristic_import(self,
+ items_to_import: Set[str],
+ prediction_type_helper: Callable[[Path],SchedulerPredictionType]=None,
+ )->Dict[str, AddModelResult]:
+ '''Import a list of paths, repo_ids or URLs. Returns the set of
+ successfully imported items.
+ :param items_to_import: Set of strings corresponding to models to be imported.
+ :param prediction_type_helper: A callback that receives the Path of a Stable Diffusion 2 checkpoint model and returns a SchedulerPredictionType.
+
+ The prediction type helper is necessary to distinguish between
+ models based on Stable Diffusion 2 Base (requiring
+ SchedulerPredictionType.Epsilson) and Stable Diffusion 768
+ (requiring SchedulerPredictionType.VPrediction). It is
+ generally impossible to do this programmatically, so the
+ prediction_type_helper usually asks the user to choose.
+
+ The result is a set of successfully installed models. Each element
+ of the set is a dict corresponding to the newly-created OmegaConf stanza for
+ that model.
+
+ May return the following exceptions:
+ - KeyError - one or more of the items to import is not a valid path, repo_id or URL
+ - ValueError - a corresponding model already exists
+ '''
+ # avoid circular import here
+ from invokeai.backend.install.model_install_backend import ModelInstall
+ successfully_installed = dict()
+
+ installer = ModelInstall(config = self.app_config,
+ prediction_type_helper = prediction_type_helper,
+ model_manager = self)
+ for thing in items_to_import:
+ installed = installer.heuristic_import(thing)
+ successfully_installed.update(installed)
+ self.commit()
+ return successfully_installed
diff --git a/invokeai/backend/model_management/model_merge.py b/invokeai/backend/model_management/model_merge.py
new file mode 100644
index 0000000000..39f951d2b4
--- /dev/null
+++ b/invokeai/backend/model_management/model_merge.py
@@ -0,0 +1,131 @@
+"""
+invokeai.backend.model_management.model_merge exports:
+merge_diffusion_models() -- combine multiple models by location and return a pipeline object
+merge_diffusion_models_and_commit() -- combine multiple models by ModelManager ID and write to models.yaml
+
+Copyright (c) 2023 Lincoln Stein and the InvokeAI Development Team
+"""
+
+import warnings
+from enum import Enum
+from pathlib import Path
+from diffusers import DiffusionPipeline
+from diffusers import logging as dlogging
+from typing import List, Union
+
+import invokeai.backend.util.logging as logger
+
+from ...backend.model_management import ModelManager, ModelType, BaseModelType, ModelVariantType, AddModelResult
+
+class MergeInterpolationMethod(str, Enum):
+ WeightedSum = "weighted_sum"
+ Sigmoid = "sigmoid"
+ InvSigmoid = "inv_sigmoid"
+ AddDifference = "add_difference"
+
+class ModelMerger(object):
+ def __init__(self, manager: ModelManager):
+ self.manager = manager
+
+ def merge_diffusion_models(
+ self,
+ model_paths: List[Path],
+ alpha: float = 0.5,
+ interp: MergeInterpolationMethod = None,
+ force: bool = False,
+ **kwargs,
+ ) -> DiffusionPipeline:
+ """
+ :param model_paths: up to three models, designated by their local paths or HuggingFace repo_ids
+ :param alpha: The interpolation parameter. Ranges from 0 to 1. It affects the ratio in which the checkpoints are merged. A 0.8 alpha
+ would mean that the first model checkpoints would affect the final result far less than an alpha of 0.2
+ :param interp: The interpolation method to use for the merging. Supports "sigmoid", "inv_sigmoid", "add_difference" and None.
+ Passing None uses the default interpolation which is weighted sum interpolation. For merging three checkpoints, only "add_difference" is supported.
+ :param force: Whether to ignore mismatch in model_config.json for the current models. Defaults to False.
+
+ **kwargs - the default DiffusionPipeline.get_config_dict kwargs:
+ cache_dir, resume_download, force_download, proxies, local_files_only, use_auth_token, revision, torch_dtype, device_map
+ """
+ with warnings.catch_warnings():
+ warnings.simplefilter("ignore")
+ verbosity = dlogging.get_verbosity()
+ dlogging.set_verbosity_error()
+
+ pipe = DiffusionPipeline.from_pretrained(
+ model_paths[0],
+ custom_pipeline="checkpoint_merger",
+ )
+ merged_pipe = pipe.merge(
+ pretrained_model_name_or_path_list=model_paths,
+ alpha=alpha,
+ interp=interp.value if interp else None, #diffusers API treats None as "weighted sum"
+ force=force,
+ **kwargs,
+ )
+ dlogging.set_verbosity(verbosity)
+ return merged_pipe
+
+
+ def merge_diffusion_models_and_save (
+ self,
+ model_names: List[str],
+ base_model: Union[BaseModelType,str],
+ merged_model_name: str,
+ alpha: float = 0.5,
+ interp: MergeInterpolationMethod = None,
+ force: bool = False,
+ **kwargs,
+ ) -> AddModelResult:
+ """
+ :param models: up to three models, designated by their InvokeAI models.yaml model name
+ :param base_model: base model (must be the same for all merged models!)
+ :param merged_model_name: name for new model
+ :param alpha: The interpolation parameter. Ranges from 0 to 1. It affects the ratio in which the checkpoints are merged. A 0.8 alpha
+ would mean that the first model checkpoints would affect the final result far less than an alpha of 0.2
+ :param interp: The interpolation method to use for the merging. Supports "weighted_average", "sigmoid", "inv_sigmoid", "add_difference" and None.
+ Passing None uses the default interpolation which is weighted sum interpolation. For merging three checkpoints, only "add_difference" is supported. Add_difference is A+(B-C).
+ :param force: Whether to ignore mismatch in model_config.json for the current models. Defaults to False.
+
+ **kwargs - the default DiffusionPipeline.get_config_dict kwargs:
+ cache_dir, resume_download, force_download, proxies, local_files_only, use_auth_token, revision, torch_dtype, device_map
+ """
+ model_paths = list()
+ config = self.manager.app_config
+ base_model = BaseModelType(base_model)
+ vae = None
+
+ for mod in model_names:
+ info = self.manager.list_model(mod, base_model=base_model, model_type=ModelType.Main)
+ assert info, f"model {mod}, base_model {base_model}, is unknown"
+ assert info["model_format"] == "diffusers", f"{mod} is not a diffusers model. It must be optimized before merging"
+ assert info["variant"] == "normal", f"{mod} is a {info['variant']} model, which cannot currently be merged"
+ assert len(model_names) <= 2 or \
+ interp==MergeInterpolationMethod.AddDifference, "When merging three models, only the 'add_difference' merge method is supported"
+ # pick up the first model's vae
+ if mod == model_names[0]:
+ vae = info.get("vae")
+ model_paths.extend([config.root_path / info["path"]])
+
+ merge_method = None if interp == 'weighted_sum' else MergeInterpolationMethod(interp)
+ logger.debug(f'interp = {interp}, merge_method={merge_method}')
+ merged_pipe = self.merge_diffusion_models(
+ model_paths, alpha, merge_method, force, **kwargs
+ )
+ dump_path = config.models_path / base_model.value / ModelType.Main.value
+ dump_path.mkdir(parents=True, exist_ok=True)
+ dump_path = dump_path / merged_model_name
+
+ merged_pipe.save_pretrained(dump_path, safe_serialization=1)
+ attributes = dict(
+ path = str(dump_path),
+ description = f"Merge of models {', '.join(model_names)}",
+ model_format = "diffusers",
+ variant = ModelVariantType.Normal.value,
+ vae = vae,
+ )
+ return self.manager.add_model(merged_model_name,
+ base_model = base_model,
+ model_type = ModelType.Main,
+ model_attributes = attributes,
+ clobber = True
+ )
diff --git a/invokeai/backend/model_management/model_probe.py b/invokeai/backend/model_management/model_probe.py
index 54fac5cde1..1da0b59101 100644
--- a/invokeai/backend/model_management/model_probe.py
+++ b/invokeai/backend/model_management/model_probe.py
@@ -1,27 +1,28 @@
import json
-import traceback
import torch
import safetensors.torch
from dataclasses import dataclass
-from enum import Enum
-from diffusers import ModelMixin, ConfigMixin, StableDiffusionPipeline, AutoencoderKL, ControlNetModel
+from diffusers import ModelMixin, ConfigMixin
from pathlib import Path
-from typing import Callable, Literal, Union, Dict
+from typing import Callable, Literal, Union, Dict, Optional
from picklescan.scanner import scan_file_path
-import invokeai.backend.util.logging as logger
-from .models import BaseModelType, ModelType, ModelVariantType, SchedulerPredictionType, SilenceWarnings
+from .models import (
+ BaseModelType, ModelType, ModelVariantType,
+ SchedulerPredictionType, SilenceWarnings,
+)
+from .models.base import read_checkpoint_meta
@dataclass
-class ModelVariantInfo(object):
+class ModelProbeInfo(object):
model_type: ModelType
base_type: BaseModelType
variant_type: ModelVariantType
prediction_type: SchedulerPredictionType
upcast_attention: bool
- format: Literal['folder','checkpoint']
+ format: Literal['diffusers','checkpoint', 'lycoris']
image_size: int
class ProbeBase(object):
@@ -31,19 +32,19 @@ class ProbeBase(object):
class ModelProbe(object):
PROBES = {
- 'folder': { },
+ 'diffusers': { },
'checkpoint': { },
}
CLASS2TYPE = {
- 'StableDiffusionPipeline' : ModelType.Pipeline,
+ 'StableDiffusionPipeline' : ModelType.Main,
'AutoencoderKL' : ModelType.Vae,
'ControlNetModel' : ModelType.ControlNet,
}
@classmethod
def register_probe(cls,
- format: Literal['folder','file'],
+ format: Literal['diffusers','checkpoint'],
model_type: ModelType,
probe_class: ProbeBase):
cls.PROBES[format][model_type] = probe_class
@@ -51,20 +52,20 @@ class ModelProbe(object):
@classmethod
def heuristic_probe(cls,
model: Union[Dict, ModelMixin, Path],
- prediction_type_helper: Callable[[Path],BaseModelType]=None,
- )->ModelVariantInfo:
+ prediction_type_helper: Callable[[Path],SchedulerPredictionType]=None,
+ )->ModelProbeInfo:
if isinstance(model,Path):
return cls.probe(model_path=model,prediction_type_helper=prediction_type_helper)
elif isinstance(model,(dict,ModelMixin,ConfigMixin)):
return cls.probe(model_path=None, model=model, prediction_type_helper=prediction_type_helper)
else:
- raise Exception("model parameter {model} is neither a Path, nor a model")
+ raise ValueError("model parameter {model} is neither a Path, nor a model")
@classmethod
def probe(cls,
model_path: Path,
- model: Union[Dict, ModelMixin] = None,
- prediction_type_helper: Callable[[Path],BaseModelType] = None)->ModelVariantInfo:
+ model: Optional[Union[Dict, ModelMixin]] = None,
+ prediction_type_helper: Optional[Callable[[Path],SchedulerPredictionType]] = None)->ModelProbeInfo:
'''
Probe the model at model_path and return sufficient information about it
to place it somewhere in the models directory hierarchy. If the model is
@@ -74,23 +75,23 @@ class ModelProbe(object):
between V2-Base and V2-768 SD models.
'''
if model_path:
- format = 'folder' if model_path.is_dir() else 'checkpoint'
+ format_type = 'diffusers' if model_path.is_dir() else 'checkpoint'
else:
- format = 'folder' if isinstance(model,(ConfigMixin,ModelMixin)) else 'checkpoint'
-
+ format_type = 'diffusers' if isinstance(model,(ConfigMixin,ModelMixin)) else 'checkpoint'
model_info = None
try:
model_type = cls.get_model_type_from_folder(model_path, model) \
- if format == 'folder' \
+ if format_type == 'diffusers' \
else cls.get_model_type_from_checkpoint(model_path, model)
- probe_class = cls.PROBES[format].get(model_type)
+ probe_class = cls.PROBES[format_type].get(model_type)
if not probe_class:
return None
probe = probe_class(model_path, model, prediction_type_helper)
base_type = probe.get_base_type()
variant_type = probe.get_variant_type()
prediction_type = probe.get_scheduler_prediction_type()
- model_info = ModelVariantInfo(
+ format = probe.get_format()
+ model_info = ModelProbeInfo(
model_type = model_type,
base_type = base_type,
variant_type = variant_type,
@@ -102,32 +103,42 @@ class ModelProbe(object):
and prediction_type==SchedulerPredictionType.VPrediction \
) else 512,
)
- except Exception as e:
- return None
+ except Exception:
+ raise
return model_info
@classmethod
- def get_model_type_from_checkpoint(cls, model_path: Path, checkpoint: dict)->ModelType:
- if model_path.suffix not in ('.bin','.pt','.ckpt','.safetensors'):
+ def get_model_type_from_checkpoint(cls, model_path: Path, checkpoint: dict) -> ModelType:
+ if model_path.suffix not in ('.bin','.pt','.ckpt','.safetensors','.pth'):
return None
- if model_path.name=='learned_embeds.bin':
+
+ if model_path.name == "learned_embeds.bin":
return ModelType.TextualInversion
- checkpoint = checkpoint or cls._scan_and_load_checkpoint(model_path)
- state_dict = checkpoint.get("state_dict") or checkpoint
- if any([x.startswith("model.diffusion_model") for x in state_dict.keys()]):
- return ModelType.Pipeline
- if any([x.startswith("encoder.conv_in") for x in state_dict.keys()]):
- return ModelType.Vae
- if "string_to_token" in state_dict or "emb_params" in state_dict:
- return ModelType.TextualInversion
- if any([x.startswith("lora") for x in state_dict.keys()]):
- return ModelType.Lora
- if any([x.startswith("control_model") for x in state_dict.keys()]):
- return ModelType.ControlNet
- if any([x.startswith("input_blocks") for x in state_dict.keys()]):
- return ModelType.ControlNet
- return None # give up
+
+ ckpt = checkpoint if checkpoint else read_checkpoint_meta(model_path, scan=True)
+ ckpt = ckpt.get("state_dict", ckpt)
+
+ for key in ckpt.keys():
+ 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
+ elif any(key.startswith(v) for v in {"lora_te_", "lora_unet_"}):
+ return ModelType.Lora
+ elif any(key.endswith(v) for v in {"to_k_lora.up.weight", "to_q_lora.down.weight"}):
+ 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"}:
+ return ModelType.TextualInversion
+
+ else:
+ # diffusers-ti
+ if len(ckpt) < 10 and all(isinstance(v, torch.Tensor) for v in ckpt.values()):
+ return ModelType.TextualInversion
+
+ raise ValueError(f"Unable to determine model type for {model_path}")
@classmethod
def get_model_type_from_folder(cls, folder_path: Path, model: ModelMixin)->ModelType:
@@ -157,7 +168,7 @@ class ModelProbe(object):
return type
# give up
- raise ValueError("Unable to determine model type")
+ raise ValueError(f"Unable to determine model type for {folder_path}")
@classmethod
def _scan_and_load_checkpoint(cls,model_path: Path)->dict:
@@ -192,11 +203,14 @@ class ProbeBase(object):
def get_scheduler_prediction_type(self)->SchedulerPredictionType:
pass
+ def get_format(self)->str:
+ pass
+
class CheckpointProbeBase(ProbeBase):
def __init__(self,
checkpoint_path: Path,
checkpoint: dict,
- helper: Callable[[Path],BaseModelType] = None
+ helper: Callable[[Path],SchedulerPredictionType] = None
)->BaseModelType:
self.checkpoint = checkpoint or ModelProbe._scan_and_load_checkpoint(checkpoint_path)
self.checkpoint_path = checkpoint_path
@@ -205,9 +219,12 @@ class CheckpointProbeBase(ProbeBase):
def get_base_type(self)->BaseModelType:
pass
+ def get_format(self)->str:
+ return 'checkpoint'
+
def get_variant_type(self)-> ModelVariantType:
model_type = ModelProbe.get_model_type_from_checkpoint(self.checkpoint_path,self.checkpoint)
- if model_type != ModelType.Pipeline:
+ if model_type != ModelType.Main:
return ModelVariantType.Normal
state_dict = self.checkpoint.get('state_dict') or self.checkpoint
in_channels = state_dict[
@@ -220,7 +237,7 @@ class CheckpointProbeBase(ProbeBase):
elif in_channels == 4:
return ModelVariantType.Normal
else:
- raise Exception("Cannot determine variant type")
+ raise ValueError(f"Cannot determine variant type (in_channels={in_channels}) at {self.checkpoint_path}")
class PipelineCheckpointProbe(CheckpointProbeBase):
def get_base_type(self)->BaseModelType:
@@ -231,7 +248,7 @@ class PipelineCheckpointProbe(CheckpointProbeBase):
return BaseModelType.StableDiffusion1
if key_name in state_dict and state_dict[key_name].shape[-1] == 1024:
return BaseModelType.StableDiffusion2
- raise Exception("Cannot determine base type")
+ raise ValueError("Cannot determine base type")
def get_scheduler_prediction_type(self)->SchedulerPredictionType:
type = self.get_base_type()
@@ -246,7 +263,8 @@ class PipelineCheckpointProbe(CheckpointProbeBase):
return SchedulerPredictionType.Epsilon
elif checkpoint["global_step"] == 110000:
return SchedulerPredictionType.VPrediction
- if self.checkpoint_path and self.helper:
+ if self.checkpoint_path and self.helper \
+ and not self.checkpoint_path.with_suffix('.yaml').exists(): # if a .yaml config file exists, then this step not needed
return self.helper(self.checkpoint_path)
else:
return None
@@ -257,6 +275,9 @@ class VaeCheckpointProbe(CheckpointProbeBase):
return BaseModelType.StableDiffusion1
class LoRACheckpointProbe(CheckpointProbeBase):
+ def get_format(self)->str:
+ return 'lycoris'
+
def get_base_type(self)->BaseModelType:
checkpoint = self.checkpoint
key1 = "lora_te_text_model_encoder_layers_0_mlp_fc1.lora_down.weight"
@@ -276,6 +297,9 @@ class LoRACheckpointProbe(CheckpointProbeBase):
return None
class TextualInversionCheckpointProbe(CheckpointProbeBase):
+ def get_format(self)->str:
+ return None
+
def get_base_type(self)->BaseModelType:
checkpoint = self.checkpoint
if 'string_to_token' in checkpoint:
@@ -305,7 +329,7 @@ class ControlNetCheckpointProbe(CheckpointProbeBase):
return BaseModelType.StableDiffusion2
elif self.checkpoint_path and self.helper:
return self.helper(self.checkpoint_path)
- raise Exception("Unable to determine base type for {self.checkpoint_path}")
+ raise ValueError("Unable to determine base type for {self.checkpoint_path}")
########################################################
# classes for probing folders
@@ -322,17 +346,16 @@ class FolderProbeBase(ProbeBase):
def get_variant_type(self)->ModelVariantType:
return ModelVariantType.Normal
+ def get_format(self)->str:
+ return 'diffusers'
+
class PipelineFolderProbe(FolderProbeBase):
def get_base_type(self)->BaseModelType:
if self.model:
unet_conf = self.model.unet.config
- scheduler_conf = self.model.scheduler.config
else:
with open(self.folder_path / 'unet' / 'config.json','r') as file:
unet_conf = json.load(file)
- with open(self.folder_path / 'scheduler' / 'scheduler_config.json','r') as file:
- scheduler_conf = json.load(file)
-
if unet_conf['cross_attention_dim'] == 768:
return BaseModelType.StableDiffusion1
elif unet_conf['cross_attention_dim'] == 1024:
@@ -381,6 +404,9 @@ class VaeFolderProbe(FolderProbeBase):
return BaseModelType.StableDiffusion1
class TextualInversionFolderProbe(FolderProbeBase):
+ def get_format(self)->str:
+ return None
+
def get_base_type(self)->BaseModelType:
path = self.folder_path / 'learned_embeds.bin'
if not path.exists():
@@ -392,7 +418,7 @@ class ControlNetFolderProbe(FolderProbeBase):
def get_base_type(self)->BaseModelType:
config_file = self.folder_path / 'config.json'
if not config_file.exists():
- raise Exception(f"Cannot determine base type for {self.folder_path}")
+ raise ValueError(f"Cannot determine base type for {self.folder_path}")
with open(config_file,'r') as file:
config = json.load(file)
# no obvious way to distinguish between sd2-base and sd2-768
@@ -401,16 +427,24 @@ class ControlNetFolderProbe(FolderProbeBase):
else BaseModelType.StableDiffusion2
class LoRAFolderProbe(FolderProbeBase):
- # I've never seen one of these in the wild, so this is a noop
- pass
+ def get_base_type(self)->BaseModelType:
+ model_file = None
+ for suffix in ['safetensors','bin']:
+ base_file = self.folder_path / f'pytorch_lora_weights.{suffix}'
+ if base_file.exists():
+ model_file = base_file
+ break
+ if not model_file:
+ raise ValueError('Unknown LoRA format encountered')
+ return LoRACheckpointProbe(model_file,None).get_base_type()
############## register probe classes ######
-ModelProbe.register_probe('folder', ModelType.Pipeline, PipelineFolderProbe)
-ModelProbe.register_probe('folder', ModelType.Vae, VaeFolderProbe)
-ModelProbe.register_probe('folder', ModelType.Lora, LoRAFolderProbe)
-ModelProbe.register_probe('folder', ModelType.TextualInversion, TextualInversionFolderProbe)
-ModelProbe.register_probe('folder', ModelType.ControlNet, ControlNetFolderProbe)
-ModelProbe.register_probe('checkpoint', ModelType.Pipeline, PipelineCheckpointProbe)
+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.TextualInversion, TextualInversionFolderProbe)
+ModelProbe.register_probe('diffusers', ModelType.ControlNet, ControlNetFolderProbe)
+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.TextualInversion, TextualInversionCheckpointProbe)
diff --git a/invokeai/backend/model_management/models/__init__.py b/invokeai/backend/model_management/models/__init__.py
index 0b8cdbfa0d..3599a5a60b 100644
--- a/invokeai/backend/model_management/models/__init__.py
+++ b/invokeai/backend/model_management/models/__init__.py
@@ -2,7 +2,7 @@ import inspect
from enum import Enum
from pydantic import BaseModel
from typing import Literal, get_origin
-from .base import BaseModelType, ModelType, SubModelType, ModelBase, ModelConfigBase, ModelVariantType, SchedulerPredictionType, ModelError, SilenceWarnings
+from .base import BaseModelType, ModelType, SubModelType, ModelBase, ModelConfigBase, ModelVariantType, SchedulerPredictionType, ModelError, SilenceWarnings, ModelNotFoundException, InvalidModelException
from .stable_diffusion import StableDiffusion1Model, StableDiffusion2Model
from .vae import VaeModel
from .lora import LoRAModel
@@ -13,23 +13,23 @@ from .stable_diffusion_onnx import ONNXStableDiffusion1Model, ONNXStableDiffusio
MODEL_CLASSES = {
BaseModelType.StableDiffusion1: {
- ModelType.Pipeline: StableDiffusion1Model,
ModelType.ONNX: ONNXStableDiffusion1Model,
+ ModelType.Main: StableDiffusion1Model,
ModelType.Vae: VaeModel,
ModelType.Lora: LoRAModel,
ModelType.ControlNet: ControlNetModel,
ModelType.TextualInversion: TextualInversionModel,
},
BaseModelType.StableDiffusion2: {
- ModelType.Pipeline: StableDiffusion2Model,
ModelType.ONNX: ONNXStableDiffusion2Model,
+ ModelType.Main: StableDiffusion2Model,
ModelType.Vae: VaeModel,
ModelType.Lora: LoRAModel,
ModelType.ControlNet: ControlNetModel,
ModelType.TextualInversion: TextualInversionModel,
},
#BaseModelType.Kandinsky2_1: {
- # ModelType.Pipeline: Kandinsky2_1Model,
+ # ModelType.Main: Kandinsky2_1Model,
# ModelType.MoVQ: MoVQModel,
# ModelType.Lora: LoRAModel,
# ModelType.ControlNet: ControlNetModel,
@@ -41,9 +41,9 @@ MODEL_CONFIGS = list()
OPENAPI_MODEL_CONFIGS = list()
class OpenAPIModelInfoBase(BaseModel):
- name: str
+ model_name: str
base_model: BaseModelType
- type: ModelType
+ model_type: ModelType
for base_model, models in MODEL_CLASSES.items():
@@ -60,7 +60,7 @@ for base_model, models in MODEL_CLASSES.items():
api_wrapper = type(openapi_cfg_name, (cfg, OpenAPIModelInfoBase), dict(
__annotations__ = dict(
- type=Literal[model_type.value],
+ model_type=Literal[model_type.value],
),
))
@@ -72,7 +72,11 @@ def get_model_config_enums():
enums = list()
for model_config in MODEL_CONFIGS:
- fields = inspect.get_annotations(model_config)
+
+ if hasattr(inspect,'get_annotations'):
+ fields = inspect.get_annotations(model_config)
+ else:
+ fields = model_config.__annotations__
try:
field = fields["model_format"]
except:
diff --git a/invokeai/backend/model_management/models/base.py b/invokeai/backend/model_management/models/base.py
index a3c9b4bc87..78685fb6f8 100644
--- a/invokeai/backend/model_management/models/base.py
+++ b/invokeai/backend/model_management/models/base.py
@@ -1,9 +1,12 @@
+import json
import os
import sys
import typing
import inspect
from enum import Enum
from abc import ABCMeta, abstractmethod
+from pathlib import Path
+from picklescan.scanner import scan_file_path
import torch
import numpy as np
import safetensors.torch
@@ -18,6 +21,11 @@ import onnx
from onnx import numpy_helper
from onnx.external_data_helper import set_external_data
from onnxruntime import InferenceSession, OrtValue, SessionOptions
+class InvalidModelException(Exception):
+ pass
+
+class ModelNotFoundException(Exception):
+ pass
class BaseModelType(str, Enum):
StableDiffusion1 = "sd-1"
@@ -26,7 +34,7 @@ class BaseModelType(str, Enum):
class ModelType(str, Enum):
ONNX = "onnx"
- Pipeline = "pipeline"
+ Main = "main"
Vae = "vae"
Lora = "lora"
ControlNet = "controlnet" # used by model_probe
@@ -66,7 +74,6 @@ class ModelConfigBase(BaseModel):
class Config:
use_enum_values = True
-
class EmptyConfigLoader(ConfigMixin):
@classmethod
def load_config(cls, *args, **kwargs):
@@ -134,7 +141,10 @@ class ModelBase(metaclass=ABCMeta):
if not isinstance(value, type) or not issubclass(value, ModelConfigBase):
continue
- fields = inspect.get_annotations(value)
+ if hasattr(inspect,'get_annotations'):
+ fields = inspect.get_annotations(value)
+ else:
+ fields = value.__annotations__
try:
field = fields["model_format"]
except:
@@ -395,15 +405,18 @@ def _fast_safetensors_reader(path: str):
return checkpoint
-
-def read_checkpoint_meta(path: str):
- if path.endswith(".safetensors"):
+def read_checkpoint_meta(path: Union[str, Path], scan: bool = False):
+ if str(path).endswith(".safetensors"):
try:
checkpoint = _fast_safetensors_reader(path)
except:
# TODO: create issue for support "meta"?
checkpoint = safetensors.torch.load_file(path, device="cpu")
else:
+ if scan:
+ scan_result = scan_file_path(path)
+ if scan_result.infected_files != 0:
+ raise Exception(f"The model file \"{path}\" is potentially infected by malware. Aborting import.")
checkpoint = torch.load(path, map_location=torch.device("meta"))
return checkpoint
diff --git a/invokeai/backend/model_management/models/controlnet.py b/invokeai/backend/model_management/models/controlnet.py
index 9563f87afd..3b73097799 100644
--- a/invokeai/backend/model_management/models/controlnet.py
+++ b/invokeai/backend/model_management/models/controlnet.py
@@ -13,6 +13,7 @@ from .base import (
calc_model_size_by_fs,
calc_model_size_by_data,
classproperty,
+ InvalidModelException,
)
class ControlNetModelFormat(str, Enum):
@@ -73,10 +74,18 @@ class ControlNetModel(ModelBase):
@classmethod
def detect_format(cls, path: str):
+ if not os.path.exists(path):
+ raise ModelNotFoundException()
+
if os.path.isdir(path):
- return ControlNetModelFormat.Diffusers
- else:
- return ControlNetModelFormat.Checkpoint
+ if os.path.exists(os.path.join(path, "config.json")):
+ return ControlNetModelFormat.Diffusers
+
+ if os.path.isfile(path):
+ if any([path.endswith(f".{ext}") for ext in ["safetensors", "ckpt", "pt", "pth"]]):
+ return ControlNetModelFormat.Checkpoint
+
+ raise InvalidModelException(f"Not a valid model: {path}")
@classmethod
def convert_if_required(
diff --git a/invokeai/backend/model_management/models/lora.py b/invokeai/backend/model_management/models/lora.py
index 59feacde06..5387ade0e5 100644
--- a/invokeai/backend/model_management/models/lora.py
+++ b/invokeai/backend/model_management/models/lora.py
@@ -9,6 +9,7 @@ from .base import (
ModelType,
SubModelType,
classproperty,
+ InvalidModelException,
)
# TODO: naming
from ..lora import LoRAModel as LoRAModelRaw
@@ -56,10 +57,18 @@ class LoRAModel(ModelBase):
@classmethod
def detect_format(cls, path: str):
+ if not os.path.exists(path):
+ raise ModelNotFoundException()
+
if os.path.isdir(path):
- return LoRAModelFormat.Diffusers
- else:
- return LoRAModelFormat.LyCORIS
+ if os.path.exists(os.path.join(path, "pytorch_lora_weights.bin")):
+ return LoRAModelFormat.Diffusers
+
+ if os.path.isfile(path):
+ if any([path.endswith(f".{ext}") for ext in ["safetensors", "ckpt", "pt"]]):
+ return LoRAModelFormat.LyCORIS
+
+ raise InvalidModelException(f"Not a valid model: {path}")
@classmethod
def convert_if_required(
diff --git a/invokeai/backend/model_management/models/stable_diffusion.py b/invokeai/backend/model_management/models/stable_diffusion.py
index f169326571..74751a40dd 100644
--- a/invokeai/backend/model_management/models/stable_diffusion.py
+++ b/invokeai/backend/model_management/models/stable_diffusion.py
@@ -16,6 +16,7 @@ from .base import (
SilenceWarnings,
read_checkpoint_meta,
classproperty,
+ InvalidModelException,
)
from invokeai.app.services.config import InvokeAIAppConfig
from omegaconf import OmegaConf
@@ -34,17 +35,17 @@ class StableDiffusion1Model(DiffusersModel):
class CheckpointConfig(ModelConfigBase):
model_format: Literal[StableDiffusion1ModelFormat.Checkpoint]
vae: Optional[str] = Field(None)
- config: Optional[str] = Field(None)
+ config: str
variant: ModelVariantType
def __init__(self, model_path: str, base_model: BaseModelType, model_type: ModelType):
assert base_model == BaseModelType.StableDiffusion1
- assert model_type == ModelType.Pipeline
+ assert model_type == ModelType.Main
super().__init__(
model_path=model_path,
base_model=BaseModelType.StableDiffusion1,
- model_type=ModelType.Pipeline,
+ model_type=ModelType.Main,
)
@classmethod
@@ -69,7 +70,7 @@ class StableDiffusion1Model(DiffusersModel):
in_channels = unet_config['in_channels']
else:
- raise Exception("Not supported stable diffusion diffusers format(possibly onnx?)")
+ raise NotImplementedError(f"{path} is not a supported stable diffusion diffusers format")
else:
raise NotImplementedError(f"Unknown stable diffusion 1.* format: {model_format}")
@@ -81,6 +82,8 @@ class StableDiffusion1Model(DiffusersModel):
else:
raise Exception("Unkown stable diffusion 1.* model format")
+ if ckpt_config_path is None:
+ ckpt_config_path = _select_ckpt_config(BaseModelType.StableDiffusion1, variant)
return cls.create_config(
path=path,
@@ -96,10 +99,18 @@ class StableDiffusion1Model(DiffusersModel):
@classmethod
def detect_format(cls, model_path: str):
+ if not os.path.exists(model_path):
+ raise ModelNotFoundException()
+
if os.path.isdir(model_path):
- return StableDiffusion1ModelFormat.Diffusers
- else:
- return StableDiffusion1ModelFormat.Checkpoint
+ if os.path.exists(os.path.join(model_path, "model_index.json")):
+ return StableDiffusion1ModelFormat.Diffusers
+
+ if os.path.isfile(model_path):
+ if any([model_path.endswith(f".{ext}") for ext in ["safetensors", "ckpt", "pt"]]):
+ return StableDiffusion1ModelFormat.Checkpoint
+
+ raise InvalidModelException(f"Not a valid model: {model_path}")
@classmethod
def convert_if_required(
@@ -109,14 +120,12 @@ class StableDiffusion1Model(DiffusersModel):
config: ModelConfigBase,
base_model: BaseModelType,
) -> str:
- assert model_path == config.path
-
if isinstance(config, cls.CheckpointConfig):
return _convert_ckpt_and_cache(
version=BaseModelType.StableDiffusion1,
model_config=config,
output_path=output_path,
- ) # TODO: args
+ )
else:
return model_path
@@ -131,25 +140,20 @@ class StableDiffusion2Model(DiffusersModel):
model_format: Literal[StableDiffusion2ModelFormat.Diffusers]
vae: Optional[str] = Field(None)
variant: ModelVariantType
- prediction_type: SchedulerPredictionType
- upcast_attention: bool
class CheckpointConfig(ModelConfigBase):
model_format: Literal[StableDiffusion2ModelFormat.Checkpoint]
vae: Optional[str] = Field(None)
- config: Optional[str] = Field(None)
+ config: str
variant: ModelVariantType
- prediction_type: SchedulerPredictionType
- upcast_attention: bool
-
def __init__(self, model_path: str, base_model: BaseModelType, model_type: ModelType):
assert base_model == BaseModelType.StableDiffusion2
- assert model_type == ModelType.Pipeline
+ assert model_type == ModelType.Main
super().__init__(
model_path=model_path,
base_model=BaseModelType.StableDiffusion2,
- model_type=ModelType.Pipeline,
+ model_type=ModelType.Main,
)
@classmethod
@@ -188,13 +192,8 @@ class StableDiffusion2Model(DiffusersModel):
else:
raise Exception("Unkown stable diffusion 2.* model format")
- if variant == ModelVariantType.Normal:
- prediction_type = SchedulerPredictionType.VPrediction
- upcast_attention = True
-
- else:
- prediction_type = SchedulerPredictionType.Epsilon
- upcast_attention = False
+ if ckpt_config_path is None:
+ ckpt_config_path = _select_ckpt_config(BaseModelType.StableDiffusion2, variant)
return cls.create_config(
path=path,
@@ -202,8 +201,6 @@ class StableDiffusion2Model(DiffusersModel):
config=ckpt_config_path,
variant=variant,
- prediction_type=prediction_type,
- upcast_attention=upcast_attention,
)
@classproperty
@@ -212,10 +209,18 @@ class StableDiffusion2Model(DiffusersModel):
@classmethod
def detect_format(cls, model_path: str):
+ if not os.path.exists(model_path):
+ raise ModelNotFoundException()
+
if os.path.isdir(model_path):
- return StableDiffusion2ModelFormat.Diffusers
- else:
- return StableDiffusion2ModelFormat.Checkpoint
+ if os.path.exists(os.path.join(model_path, "model_index.json")):
+ return StableDiffusion2ModelFormat.Diffusers
+
+ if os.path.isfile(model_path):
+ if any([model_path.endswith(f".{ext}") for ext in ["safetensors", "ckpt", "pt"]]):
+ return StableDiffusion2ModelFormat.Checkpoint
+
+ raise InvalidModelException(f"Not a valid model: {model_path}")
@classmethod
def convert_if_required(
@@ -225,14 +230,12 @@ class StableDiffusion2Model(DiffusersModel):
config: ModelConfigBase,
base_model: BaseModelType,
) -> str:
- assert model_path == config.path
-
if isinstance(config, cls.CheckpointConfig):
return _convert_ckpt_and_cache(
version=BaseModelType.StableDiffusion2,
model_config=config,
output_path=output_path,
- ) # TODO: args
+ )
else:
return model_path
@@ -243,18 +246,18 @@ def _select_ckpt_config(version: BaseModelType, variant: ModelVariantType):
ModelVariantType.Inpaint: "v1-inpainting-inference.yaml",
},
BaseModelType.StableDiffusion2: {
- # code further will manually set upcast_attention and v_prediction
- ModelVariantType.Normal: "v2-inference.yaml",
+ ModelVariantType.Normal: "v2-inference-v.yaml", # best guess, as we can't differentiate with base(512)
ModelVariantType.Inpaint: "v2-inpainting-inference.yaml",
ModelVariantType.Depth: "v2-midas-inference.yaml",
}
}
+ app_config = InvokeAIAppConfig.get_config()
try:
- # TODO: path
- #model_config.config = app_config.config_dir / "stable-diffusion" / ckpt_configs[version][model_config.variant]
- #return InvokeAIAppConfig.get_config().legacy_conf_dir / ckpt_configs[version][variant]
- return InvokeAIAppConfig.get_config().root_dir / "configs" / "stable-diffusion" / ckpt_configs[version][variant]
+ config_path = app_config.legacy_conf_path / ckpt_configs[version][variant]
+ if config_path.is_relative_to(app_config.root_path):
+ config_path = config_path.relative_to(app_config.root_path)
+ return str(config_path)
except:
return None
@@ -273,36 +276,14 @@ def _convert_ckpt_and_cache(
"""
app_config = InvokeAIAppConfig.get_config()
- if model_config.config is None:
- model_config.config = _select_ckpt_config(version, model_config.variant)
- if model_config.config is None:
- raise Exception(f"Model variant {model_config.variant} not supported for {version}")
-
-
- weights = app_config.root_dir / model_config.path
- config_file = app_config.root_dir / model_config.config
+ weights = app_config.root_path / model_config.path
+ config_file = app_config.root_path / model_config.config
output_path = Path(output_path)
- if version == BaseModelType.StableDiffusion1:
- upcast_attention = False
- prediction_type = SchedulerPredictionType.Epsilon
-
- elif version == BaseModelType.StableDiffusion2:
- upcast_attention = model_config.upcast_attention
- prediction_type = model_config.prediction_type
-
- else:
- raise Exception(f"Unknown model provided: {version}")
-
-
# return cached version if it exists
if output_path.exists():
return output_path
- # TODO: I think that it more correctly to convert with embedded vae
- # as if user will delete custom vae he will got not embedded but also custom vae
- #vae_ckpt_path, vae_model = self._get_vae_for_conversion(weights, mconfig)
-
# to avoid circular import errors
from ..convert_ckpt_to_diffusers import convert_ckpt_to_diffusers
with SilenceWarnings():
@@ -313,9 +294,6 @@ def _convert_ckpt_and_cache(
model_variant=model_config.variant,
original_config_file=config_file,
extract_ema=True,
- upcast_attention=upcast_attention,
- prediction_type=prediction_type,
scan_needed=True,
- model_root=app_config.models_path,
)
return output_path
diff --git a/invokeai/backend/model_management/models/textual_inversion.py b/invokeai/backend/model_management/models/textual_inversion.py
index 9a032218f0..9cd62bb417 100644
--- a/invokeai/backend/model_management/models/textual_inversion.py
+++ b/invokeai/backend/model_management/models/textual_inversion.py
@@ -8,6 +8,8 @@ from .base import (
ModelType,
SubModelType,
classproperty,
+ ModelNotFoundException,
+ InvalidModelException,
)
# TODO: naming
from ..lora import TextualInversionModel as TextualInversionModelRaw
@@ -37,8 +39,15 @@ class TextualInversionModel(ModelBase):
if child_type is not None:
raise Exception("There is no child models in textual inversion")
+ checkpoint_path = self.model_path
+ if os.path.isdir(checkpoint_path):
+ checkpoint_path = os.path.join(checkpoint_path, "learned_embeds.bin")
+
+ if not os.path.exists(checkpoint_path):
+ raise ModelNotFoundException()
+
model = TextualInversionModelRaw.from_checkpoint(
- file_path=self.model_path,
+ file_path=checkpoint_path,
dtype=torch_dtype,
)
@@ -51,7 +60,18 @@ class TextualInversionModel(ModelBase):
@classmethod
def detect_format(cls, path: str):
- return None
+ if not os.path.exists(path):
+ raise ModelNotFoundException()
+
+ if os.path.isdir(path):
+ if os.path.exists(os.path.join(path, "learned_embeds.bin")):
+ return None # diffusers-ti
+
+ if os.path.isfile(path):
+ if any([path.endswith(f".{ext}") for ext in ["safetensors", "ckpt", "pt"]]):
+ return None
+
+ raise InvalidModelException(f"Not a valid model: {path}")
@classmethod
def convert_if_required(
diff --git a/invokeai/backend/model_management/models/vae.py b/invokeai/backend/model_management/models/vae.py
index 76133b074d..2a5b7cff24 100644
--- a/invokeai/backend/model_management/models/vae.py
+++ b/invokeai/backend/model_management/models/vae.py
@@ -15,6 +15,7 @@ from .base import (
calc_model_size_by_fs,
calc_model_size_by_data,
classproperty,
+ InvalidModelException,
)
from invokeai.app.services.config import InvokeAIAppConfig
from diffusers.utils import is_safetensors_available
@@ -75,10 +76,18 @@ class VaeModel(ModelBase):
@classmethod
def detect_format(cls, path: str):
+ if not os.path.exists(path):
+ raise ModelNotFoundException()
+
if os.path.isdir(path):
- return VaeModelFormat.Diffusers
- else:
- return VaeModelFormat.Checkpoint
+ if os.path.exists(os.path.join(path, "config.json")):
+ return VaeModelFormat.Diffusers
+
+ if os.path.isfile(path):
+ if any([path.endswith(f".{ext}") for ext in ["safetensors", "ckpt", "pt"]]):
+ return VaeModelFormat.Checkpoint
+
+ raise InvalidModelException(f"Not a valid model: {path}")
@classmethod
def convert_if_required(
@@ -137,7 +146,6 @@ def _convert_vae_ckpt_and_cache(
from .stable_diffusion import _select_ckpt_config
# all sd models use same vae settings
config_file = _select_ckpt_config(base_model, ModelVariantType.Normal)
-
else:
raise Exception(f"Vae conversion not supported for model type: {base_model}")
@@ -152,13 +160,12 @@ def _convert_vae_ckpt_and_cache(
if "state_dict" in checkpoint:
checkpoint = checkpoint["state_dict"]
- config = OmegaConf.load(config_file)
+ config = OmegaConf.load(app_config.root_path/config_file)
vae_model = convert_ldm_vae_to_diffusers(
checkpoint = checkpoint,
vae_config = config,
image_size = image_size,
- model_root = app_config.models_path,
)
vae_model.save_pretrained(
output_path,
diff --git a/invokeai/backend/stable_diffusion/diffusers_pipeline.py b/invokeai/backend/stable_diffusion/diffusers_pipeline.py
index 0010f33a0d..1d701b21fd 100644
--- a/invokeai/backend/stable_diffusion/diffusers_pipeline.py
+++ b/invokeai/backend/stable_diffusion/diffusers_pipeline.py
@@ -7,7 +7,7 @@ import secrets
from collections.abc import Sequence
from dataclasses import dataclass, field
from typing import Any, Callable, Generic, List, Optional, Type, TypeVar, Union
-from pydantic import BaseModel, Field
+from pydantic import Field
import einops
import PIL.Image
@@ -17,12 +17,11 @@ import psutil
import torch
import torchvision.transforms as T
from diffusers.models import AutoencoderKL, UNet2DConditionModel
-from diffusers.models.controlnet import ControlNetModel, ControlNetOutput
+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.controlnet import MultiControlNetModel
from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img import (
StableDiffusionImg2ImgPipeline,
@@ -46,7 +45,7 @@ from .diffusion import (
InvokeAIDiffuserComponent,
PostprocessingSettings,
)
-from .offloading import FullyLoadedModelGroup, LazilyLoadedModelGroup, ModelGroup
+from .offloading import FullyLoadedModelGroup, ModelGroup
@dataclass
class PipelineIntermediateState:
@@ -105,7 +104,7 @@ class AddsMaskGuidance:
_debug: Optional[Callable] = None
def __call__(
- self, step_output: BaseOutput | SchedulerOutput, t: torch.Tensor, conditioning
+ self, step_output: Union[BaseOutput, SchedulerOutput], t: torch.Tensor, conditioning
) -> BaseOutput:
output_class = step_output.__class__ # We'll create a new one with masked data.
@@ -128,7 +127,7 @@ class AddsMaskGuidance:
def _t_for_field(self, field_name: str, t):
if field_name == "pred_original_sample":
- return torch.zeros_like(t, dtype=t.dtype) # it represents t=0
+ return self.scheduler.timesteps[-1]
return t
def apply_mask(self, latents: torch.Tensor, t) -> torch.Tensor:
@@ -215,10 +214,12 @@ class GeneratorToCallbackinator(Generic[ParamType, ReturnType, CallbackType]):
@dataclass
class ControlNetData:
model: ControlNetModel = Field(default=None)
- image_tensor: torch.Tensor= Field(default=None)
- weight: Union[float, List[float]]= Field(default=1.0)
+ image_tensor: torch.Tensor = Field(default=None)
+ weight: Union[float, List[float]] = Field(default=1.0)
begin_step_percent: float = Field(default=0.0)
end_step_percent: float = Field(default=1.0)
+ control_mode: str = Field(default="balanced")
+
@dataclass(frozen=True)
class ConditioningData:
@@ -359,37 +360,34 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
):
self.enable_xformers_memory_efficient_attention()
else:
- if torch.backends.mps.is_available():
- # until pytorch #91617 is fixed, slicing is borked on MPS
- # https://github.com/pytorch/pytorch/issues/91617
- # fix is in https://github.com/kulinseth/pytorch/pull/222 but no idea when it will get merged to pytorch mainline.
- pass
+ if self.device.type == "cpu" or self.device.type == "mps":
+ mem_free = psutil.virtual_memory().free
+ elif self.device.type == "cuda":
+ mem_free, _ = torch.cuda.mem_get_info(normalize_device(self.device))
else:
- if self.device.type == "cpu" or self.device.type == "mps":
- mem_free = psutil.virtual_memory().free
- elif self.device.type == "cuda":
- mem_free, _ = torch.cuda.mem_get_info(normalize_device(self.device))
- else:
- raise ValueError(f"unrecognized device {self.device}")
- # input tensor of [1, 4, h/8, w/8]
- # output tensor of [16, (h/8 * w/8), (h/8 * w/8)]
- bytes_per_element_needed_for_baddbmm_duplication = (
- latents.element_size() + 4
- )
- max_size_required_for_baddbmm = (
- 16
- * latents.size(dim=2)
- * latents.size(dim=3)
- * latents.size(dim=2)
- * latents.size(dim=3)
- * bytes_per_element_needed_for_baddbmm_duplication
- )
- if max_size_required_for_baddbmm > (
- mem_free * 3.0 / 4.0
- ): # 3.3 / 4.0 is from old Invoke code
- self.enable_attention_slicing(slice_size="max")
- else:
- self.disable_attention_slicing()
+ raise ValueError(f"unrecognized device {self.device}")
+ # input tensor of [1, 4, h/8, w/8]
+ # output tensor of [16, (h/8 * w/8), (h/8 * w/8)]
+ bytes_per_element_needed_for_baddbmm_duplication = (
+ latents.element_size() + 4
+ )
+ max_size_required_for_baddbmm = (
+ 16
+ * latents.size(dim=2)
+ * latents.size(dim=3)
+ * latents.size(dim=2)
+ * latents.size(dim=3)
+ * bytes_per_element_needed_for_baddbmm_duplication
+ )
+ if max_size_required_for_baddbmm > (
+ mem_free * 3.0 / 4.0
+ ): # 3.3 / 4.0 is from old Invoke code
+ self.enable_attention_slicing(slice_size="max")
+ elif torch.backends.mps.is_available():
+ # diffusers recommends always enabling for mps
+ self.enable_attention_slicing(slice_size="max")
+ else:
+ self.disable_attention_slicing()
def to(self, torch_device: Optional[Union[str, torch.device]] = None, silence_dtype_warnings=False):
# overridden method; types match the superclass.
@@ -599,48 +597,68 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
# TODO: should this scaling happen here or inside self._unet_forward?
# i.e. before or after passing it to InvokeAIDiffuserComponent
- latent_model_input = self.scheduler.scale_model_input(latents, timestep)
+ unet_latent_input = self.scheduler.scale_model_input(latents, timestep)
# default is no controlnet, so set controlnet processing output to None
down_block_res_samples, mid_block_res_sample = None, None
if control_data is not None:
- # FIXME: make sure guidance_scale < 1.0 is handled correctly if doing per-step guidance setting
- # if conditioning_data.guidance_scale > 1.0:
- if conditioning_data.guidance_scale is not None:
- # expand the latents input to control model if doing classifier free guidance
- # (which I think for now is always true, there is conditional elsewhere that stops execution if
- # classifier_free_guidance is <= 1.0 ?)
- latent_control_input = torch.cat([latent_model_input] * 2)
- else:
- latent_control_input = latent_model_input
# control_data should be type List[ControlNetData]
# this loop covers both ControlNet (one ControlNetData in list)
# and MultiControlNet (multiple ControlNetData in list)
for i, control_datum in enumerate(control_data):
- # print("controlnet", i, "==>", type(control_datum))
+ control_mode = control_datum.control_mode
+ # soft_injection and cfg_injection are the two ControlNet control_mode booleans
+ # that are combined at higher level to make control_mode enum
+ # soft_injection determines whether to do per-layer re-weighting adjustment (if True)
+ # or default weighting (if False)
+ soft_injection = (control_mode == "more_prompt" or control_mode == "more_control")
+ # cfg_injection = determines whether to apply ControlNet to only the conditional (if True)
+ # or the default both conditional and unconditional (if False)
+ cfg_injection = (control_mode == "more_control" or control_mode == "unbalanced")
+
first_control_step = math.floor(control_datum.begin_step_percent * total_step_count)
last_control_step = math.ceil(control_datum.end_step_percent * total_step_count)
# only apply controlnet if current step is within the controlnet's begin/end step range
if step_index >= first_control_step and step_index <= last_control_step:
- # print("running controlnet", i, "for step", step_index)
+
+ if cfg_injection:
+ control_latent_input = unet_latent_input
+ else:
+ # expand the latents input to control model if doing classifier free guidance
+ # (which I think for now is always true, there is conditional elsewhere that stops execution if
+ # classifier_free_guidance is <= 1.0 ?)
+ control_latent_input = torch.cat([unet_latent_input] * 2)
+
+ if cfg_injection: # only applying ControlNet to conditional instead of in unconditioned
+ encoder_hidden_states = conditioning_data.text_embeddings
+ else:
+ encoder_hidden_states = torch.cat([conditioning_data.unconditioned_embeddings,
+ conditioning_data.text_embeddings])
if isinstance(control_datum.weight, list):
# if controlnet has multiple weights, use the weight for the current step
controlnet_weight = control_datum.weight[step_index]
else:
# if controlnet has a single weight, use it for all steps
controlnet_weight = control_datum.weight
+
+ # controlnet(s) inference
down_samples, mid_sample = control_datum.model(
- sample=latent_control_input,
+ sample=control_latent_input,
timestep=timestep,
- encoder_hidden_states=torch.cat([conditioning_data.unconditioned_embeddings,
- conditioning_data.text_embeddings]),
+ encoder_hidden_states=encoder_hidden_states,
controlnet_cond=control_datum.image_tensor,
- conditioning_scale=controlnet_weight,
- # cross_attention_kwargs,
- guess_mode=False,
+ conditioning_scale=controlnet_weight, # controlnet specific, NOT the guidance scale
+ guess_mode=soft_injection, # this is still called guess_mode in diffusers ControlNetModel
return_dict=False,
)
+ if cfg_injection:
+ # Inferred ControlNet only for the conditional batch.
+ # To apply the output of ControlNet to both the unconditional and conditional batches,
+ # add 0 to the unconditional batch to keep it unchanged.
+ down_samples = [torch.cat([torch.zeros_like(d), d]) for d in down_samples]
+ mid_sample = torch.cat([torch.zeros_like(mid_sample), mid_sample])
+
if down_block_res_samples is None and mid_block_res_sample is None:
down_block_res_samples, mid_block_res_sample = down_samples, mid_sample
else:
@@ -653,11 +671,11 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
# predict the noise residual
noise_pred = self.invokeai_diffuser.do_diffusion_step(
- latent_model_input,
- t,
- conditioning_data.unconditioned_embeddings,
- conditioning_data.text_embeddings,
- conditioning_data.guidance_scale,
+ x=unet_latent_input,
+ sigma=t,
+ unconditioning=conditioning_data.unconditioned_embeddings,
+ conditioning=conditioning_data.text_embeddings,
+ unconditional_guidance_scale=conditioning_data.guidance_scale,
step_index=step_index,
total_step_count=total_step_count,
down_block_additional_residuals=down_block_res_samples, # from controlnet(s)
@@ -895,20 +913,11 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
def non_noised_latents_from_image(self, init_image, *, device: torch.device, dtype):
init_image = init_image.to(device=device, dtype=dtype)
with torch.inference_mode():
- if device.type == "mps":
- # workaround for torch MPS bug that has been fixed in https://github.com/kulinseth/pytorch/pull/222
- # TODO remove this workaround once kulinseth#222 is merged to pytorch mainline
- self.vae.to(CPU_DEVICE)
- init_image = init_image.to(CPU_DEVICE)
- else:
- self._model_group.load(self.vae)
+ self._model_group.load(self.vae)
init_latent_dist = self.vae.encode(init_image).latent_dist
init_latents = init_latent_dist.sample().to(
dtype=dtype
) # FIXME: uses torch.randn. make reproducible!
- if device.type == "mps":
- self.vae.to(device)
- init_latents = init_latents.to(device)
init_latents = 0.18215 * init_latents
return init_latents
@@ -962,6 +971,7 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
device="cuda",
dtype=torch.float16,
do_classifier_free_guidance=True,
+ control_mode="balanced"
):
if not isinstance(image, torch.Tensor):
@@ -992,6 +1002,7 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
repeat_by = num_images_per_prompt
image = image.repeat_interleave(repeat_by, dim=0)
image = image.to(device=device, dtype=dtype)
- if do_classifier_free_guidance:
+ cfg_injection = (control_mode == "more_control" or control_mode == "unbalanced")
+ if do_classifier_free_guidance and not cfg_injection:
image = torch.cat([image] * 2)
return image
diff --git a/invokeai/backend/stable_diffusion/diffusion/shared_invokeai_diffusion.py b/invokeai/backend/stable_diffusion/diffusion/shared_invokeai_diffusion.py
index f3b09f6a9f..1175475bba 100644
--- a/invokeai/backend/stable_diffusion/diffusion/shared_invokeai_diffusion.py
+++ b/invokeai/backend/stable_diffusion/diffusion/shared_invokeai_diffusion.py
@@ -248,9 +248,6 @@ class InvokeAIDiffuserComponent:
x_twice, sigma_twice, both_conditionings, **kwargs,
)
unconditioned_next_x, conditioned_next_x = both_results.chunk(2)
- if conditioned_next_x.device.type == "mps":
- # prevent a result filled with zeros. seems to be a torch bug.
- conditioned_next_x = conditioned_next_x.clone()
return unconditioned_next_x, conditioned_next_x
def _apply_standard_conditioning_sequentially(
@@ -264,9 +261,6 @@ class InvokeAIDiffuserComponent:
# low-memory sequential path
unconditioned_next_x = self.model_forward_callback(x, sigma, unconditioning, **kwargs)
conditioned_next_x = self.model_forward_callback(x, sigma, conditioning, **kwargs)
- if conditioned_next_x.device.type == "mps":
- # prevent a result filled with zeros. seems to be a torch bug.
- conditioned_next_x = conditioned_next_x.clone()
return unconditioned_next_x, conditioned_next_x
# TODO: looks unused
diff --git a/invokeai/backend/stable_diffusion/offloading.py b/invokeai/backend/stable_diffusion/offloading.py
index 5fc3f765ae..d36b65872a 100644
--- a/invokeai/backend/stable_diffusion/offloading.py
+++ b/invokeai/backend/stable_diffusion/offloading.py
@@ -4,7 +4,7 @@ import warnings
import weakref
from abc import ABCMeta, abstractmethod
from collections.abc import MutableMapping
-from typing import Callable
+from typing import Callable, Union
import torch
from accelerate.utils import send_to_device
@@ -117,7 +117,7 @@ class LazilyLoadedModelGroup(ModelGroup):
"""
_hooks: MutableMapping[torch.nn.Module, RemovableHandle]
- _current_model_ref: Callable[[], torch.nn.Module | _NoModel]
+ _current_model_ref: Callable[[], Union[torch.nn.Module, _NoModel]]
def __init__(self, execution_device: torch.device):
super().__init__(execution_device)
diff --git a/invokeai/backend/util/__init__.py b/invokeai/backend/util/__init__.py
index 84720b1854..fadeff4d75 100644
--- a/invokeai/backend/util/__init__.py
+++ b/invokeai/backend/util/__init__.py
@@ -16,6 +16,7 @@ from .util import (
download_with_resume,
instantiate_from_config,
url_attachment_name,
+ Chdir
)
diff --git a/invokeai/backend/util/devices.py b/invokeai/backend/util/devices.py
index 615209d98d..3fbdaba41a 100644
--- a/invokeai/backend/util/devices.py
+++ b/invokeai/backend/util/devices.py
@@ -4,6 +4,7 @@ from contextlib import nullcontext
import torch
from torch import autocast
+from typing import Union
from invokeai.app.services.config import InvokeAIAppConfig
CPU_DEVICE = torch.device("cpu")
@@ -28,6 +29,8 @@ def choose_precision(device: torch.device) -> str:
device_name = torch.cuda.get_device_name(device)
if not ("GeForce GTX 1660" in device_name or "GeForce GTX 1650" in device_name):
return "float16"
+ elif device.type == "mps":
+ return "float16"
return "float32"
@@ -49,7 +52,7 @@ def choose_autocast(precision):
return nullcontext
-def normalize_device(device: str | torch.device) -> torch.device:
+def normalize_device(device: Union[str, torch.device]) -> torch.device:
"""Ensure device has a device index defined, if appropriate."""
device = torch.device(device)
if device.index is None:
diff --git a/invokeai/backend/util/mps_fixes.py b/invokeai/backend/util/mps_fixes.py
new file mode 100644
index 0000000000..1fc58f9c98
--- /dev/null
+++ b/invokeai/backend/util/mps_fixes.py
@@ -0,0 +1,63 @@
+import torch
+
+
+if torch.backends.mps.is_available():
+ torch.empty = torch.zeros
+
+
+_torch_layer_norm = torch.nn.functional.layer_norm
+def new_layer_norm(input, normalized_shape, weight=None, bias=None, eps=1e-05):
+ if input.device.type == "mps" and input.dtype == torch.float16:
+ input = input.float()
+ if weight is not None:
+ weight = weight.float()
+ if bias is not None:
+ bias = bias.float()
+ return _torch_layer_norm(input, normalized_shape, weight, bias, eps).half()
+ else:
+ return _torch_layer_norm(input, normalized_shape, weight, bias, eps)
+
+torch.nn.functional.layer_norm = new_layer_norm
+
+
+_torch_tensor_permute = torch.Tensor.permute
+def new_torch_tensor_permute(input, *dims):
+ result = _torch_tensor_permute(input, *dims)
+ if input.device == "mps" and input.dtype == torch.float16:
+ result = result.contiguous()
+ return result
+
+torch.Tensor.permute = new_torch_tensor_permute
+
+
+_torch_lerp = torch.lerp
+def new_torch_lerp(input, end, weight, *, out=None):
+ if input.device.type == "mps" and input.dtype == torch.float16:
+ input = input.float()
+ end = end.float()
+ if isinstance(weight, torch.Tensor):
+ weight = weight.float()
+ if out is not None:
+ out_fp32 = torch.zeros_like(out, dtype=torch.float32)
+ else:
+ out_fp32 = None
+ result = _torch_lerp(input, end, weight, out=out_fp32)
+ if out is not None:
+ out.copy_(out_fp32.half())
+ del out_fp32
+ return result.half()
+
+ else:
+ return _torch_lerp(input, end, weight, out=out)
+
+torch.lerp = new_torch_lerp
+
+
+_torch_interpolate = torch.nn.functional.interpolate
+def new_torch_interpolate(input, size=None, scale_factor=None, mode='nearest', align_corners=None, recompute_scale_factor=None, antialias=False):
+ if input.device.type == "mps" and input.dtype == torch.float16:
+ return _torch_interpolate(input.float(), size, scale_factor, mode, align_corners, recompute_scale_factor, antialias).half()
+ else:
+ return _torch_interpolate(input, size, scale_factor, mode, align_corners, recompute_scale_factor, antialias)
+
+torch.nn.functional.interpolate = new_torch_interpolate
diff --git a/invokeai/backend/util/util.py b/invokeai/backend/util/util.py
index 9c6a61a4fa..1cc632e483 100644
--- a/invokeai/backend/util/util.py
+++ b/invokeai/backend/util/util.py
@@ -381,3 +381,18 @@ def image_to_dataURL(image: Image.Image, image_format: str = "PNG") -> str:
buffered.getvalue()
).decode("UTF-8")
return image_base64
+
+class Chdir(object):
+ '''Context manager to chdir to desired directory and change back after context exits:
+ Args:
+ path (Path): The path to the cwd
+ '''
+ def __init__(self, path: Path):
+ self.path = path
+ self.original = Path().absolute()
+
+ def __enter__(self):
+ os.chdir(self.path)
+
+ def __exit__(self,*args):
+ os.chdir(self.original)
diff --git a/invokeai/configs/INITIAL_MODELS.yaml b/invokeai/configs/INITIAL_MODELS.yaml
index 6bf3d4231a..4ba67bc4bc 100644
--- a/invokeai/configs/INITIAL_MODELS.yaml
+++ b/invokeai/configs/INITIAL_MODELS.yaml
@@ -1,107 +1,92 @@
# This file predefines a few models that the user may want to install.
-diffusers:
- stable-diffusion-1.5:
- description: Stable Diffusion version 1.5 diffusers model (4.27 GB)
- repo_id: runwayml/stable-diffusion-v1-5
- format: diffusers
- vae:
- repo_id: stabilityai/sd-vae-ft-mse
- recommended: True
- default: True
- sd-inpainting-1.5:
- description: RunwayML SD 1.5 model optimized for inpainting, diffusers version (4.27 GB)
- repo_id: runwayml/stable-diffusion-inpainting
- format: diffusers
- vae:
- repo_id: stabilityai/sd-vae-ft-mse
- recommended: True
- stable-diffusion-2.1:
- description: Stable Diffusion version 2.1 diffusers model, trained on 768 pixel images (5.21 GB)
- repo_id: stabilityai/stable-diffusion-2-1
- format: diffusers
- recommended: True
- sd-inpainting-2.0:
- description: Stable Diffusion version 2.0 inpainting model (5.21 GB)
- repo_id: stabilityai/stable-diffusion-2-inpainting
- format: diffusers
- recommended: False
- analog-diffusion-1.0:
- description: An SD-1.5 model trained on diverse analog photographs (2.13 GB)
- repo_id: wavymulder/Analog-Diffusion
- format: diffusers
- recommended: false
- deliberate-1.0:
- description: Versatile model that produces detailed images up to 768px (4.27 GB)
- format: diffusers
- repo_id: XpucT/Deliberate
- recommended: False
- d&d-diffusion-1.0:
- description: Dungeons & Dragons characters (2.13 GB)
- format: diffusers
- repo_id: 0xJustin/Dungeons-and-Diffusion
- recommended: False
- dreamlike-photoreal-2.0:
- description: A photorealistic model trained on 768 pixel images based on SD 1.5 (2.13 GB)
- format: diffusers
- repo_id: dreamlike-art/dreamlike-photoreal-2.0
- recommended: False
- inkpunk-1.0:
- description: Stylized illustrations inspired by Gorillaz, FLCL and Shinkawa; prompt with "nvinkpunk" (4.27 GB)
- format: diffusers
- repo_id: Envvi/Inkpunk-Diffusion
- recommended: False
- openjourney-4.0:
- description: An SD 1.5 model fine tuned on Midjourney; prompt with "mdjrny-v4 style" (2.13 GB)
- format: diffusers
- repo_id: prompthero/openjourney
- vae:
- repo_id: stabilityai/sd-vae-ft-mse
- recommended: False
- portrait-plus-1.0:
- description: An SD-1.5 model trained on close range portraits of people; prompt with "portrait+" (2.13 GB)
- format: diffusers
- repo_id: wavymulder/portraitplus
- recommended: False
- seek-art-mega-1.0:
- description: A general use SD-1.5 "anything" model that supports multiple styles (2.1 GB)
- repo_id: coreco/seek.art_MEGA
- format: diffusers
- vae:
- repo_id: stabilityai/sd-vae-ft-mse
- recommended: False
- trinart-2.0:
- description: An SD-1.5 model finetuned with ~40K assorted high resolution manga/anime-style images (2.13 GB)
- repo_id: naclbit/trinart_stable_diffusion_v2
- format: diffusers
- vae:
- repo_id: stabilityai/sd-vae-ft-mse
- recommended: False
- waifu-diffusion-1.4:
- description: An SD-1.5 model trained on 680k anime/manga-style images (2.13 GB)
- repo_id: hakurei/waifu-diffusion
- format: diffusers
- vae:
- repo_id: stabilityai/sd-vae-ft-mse
- recommended: False
-controlnet:
- canny: lllyasviel/control_v11p_sd15_canny
- inpaint: lllyasviel/control_v11p_sd15_inpaint
- mlsd: lllyasviel/control_v11p_sd15_mlsd
- depth: lllyasviel/control_v11f1p_sd15_depth
- normal_bae: lllyasviel/control_v11p_sd15_normalbae
- seg: lllyasviel/control_v11p_sd15_seg
- lineart: lllyasviel/control_v11p_sd15_lineart
- lineart_anime: lllyasviel/control_v11p_sd15s2_lineart_anime
- scribble: lllyasviel/control_v11p_sd15_scribble
- softedge: lllyasviel/control_v11p_sd15_softedge
- shuffle: lllyasviel/control_v11e_sd15_shuffle
- tile: lllyasviel/control_v11f1e_sd15_tile
- ip2p: lllyasviel/control_v11e_sd15_ip2p
-textual_inversion:
- 'EasyNegative': https://huggingface.co/embed/EasyNegative/resolve/main/EasyNegative.safetensors
- 'ahx-beta-453407d': sd-concepts-library/ahx-beta-453407d
-lora:
- 'LowRA': https://civitai.com/api/download/models/63006
- 'Ink scenery': https://civitai.com/api/download/models/83390
- 'sd-model-finetuned-lora-t4': sayakpaul/sd-model-finetuned-lora-t4
-
+sd-1/main/stable-diffusion-v1-5:
+ description: Stable Diffusion version 1.5 diffusers model (4.27 GB)
+ repo_id: runwayml/stable-diffusion-v1-5
+ recommended: True
+ default: True
+sd-1/main/stable-diffusion-inpainting:
+ description: RunwayML SD 1.5 model optimized for inpainting, diffusers version (4.27 GB)
+ repo_id: runwayml/stable-diffusion-inpainting
+ recommended: True
+sd-2/main/stable-diffusion-2-1:
+ description: Stable Diffusion version 2.1 diffusers model, trained on 768 pixel images (5.21 GB)
+ repo_id: stabilityai/stable-diffusion-2-1
+ recommended: True
+sd-2/main/stable-diffusion-2-inpainting:
+ description: Stable Diffusion version 2.0 inpainting model (5.21 GB)
+ repo_id: stabilityai/stable-diffusion-2-inpainting
+ recommended: False
+sd-1/main/Analog-Diffusion:
+ description: An SD-1.5 model trained on diverse analog photographs (2.13 GB)
+ repo_id: wavymulder/Analog-Diffusion
+ recommended: false
+sd-1/main/Deliberate:
+ description: Versatile model that produces detailed images up to 768px (4.27 GB)
+ repo_id: XpucT/Deliberate
+ recommended: False
+sd-1/main/Dungeons-and-Diffusion:
+ description: Dungeons & Dragons characters (2.13 GB)
+ repo_id: 0xJustin/Dungeons-and-Diffusion
+ recommended: False
+sd-1/main/dreamlike-photoreal-2:
+ description: A photorealistic model trained on 768 pixel images based on SD 1.5 (2.13 GB)
+ repo_id: dreamlike-art/dreamlike-photoreal-2.0
+ recommended: False
+sd-1/main/Inkpunk-Diffusion:
+ description: Stylized illustrations inspired by Gorillaz, FLCL and Shinkawa; prompt with "nvinkpunk" (4.27 GB)
+ repo_id: Envvi/Inkpunk-Diffusion
+ recommended: False
+sd-1/main/openjourney:
+ description: An SD 1.5 model fine tuned on Midjourney; prompt with "mdjrny-v4 style" (2.13 GB)
+ repo_id: prompthero/openjourney
+ recommended: False
+sd-1/main/portraitplus:
+ description: An SD-1.5 model trained on close range portraits of people; prompt with "portrait+" (2.13 GB)
+ repo_id: wavymulder/portraitplus
+ recommended: False
+sd-1/main/seek.art_MEGA:
+ repo_id: coreco/seek.art_MEGA
+ description: A general use SD-1.5 "anything" model that supports multiple styles (2.1 GB)
+ recommended: False
+sd-1/main/trinart_stable_diffusion_v2:
+ description: An SD-1.5 model finetuned with ~40K assorted high resolution manga/anime-style images (2.13 GB)
+ repo_id: naclbit/trinart_stable_diffusion_v2
+ recommended: False
+sd-1/main/waifu-diffusion:
+ description: An SD-1.5 model trained on 680k anime/manga-style images (2.13 GB)
+ repo_id: hakurei/waifu-diffusion
+ recommended: False
+sd-1/controlnet/canny:
+ repo_id: lllyasviel/control_v11p_sd15_canny
+sd-1/controlnet/inpaint:
+ repo_id: lllyasviel/control_v11p_sd15_inpaint
+sd-1/controlnet/mlsd:
+ repo_id: lllyasviel/control_v11p_sd15_mlsd
+sd-1/controlnet/depth:
+ repo_id: lllyasviel/control_v11f1p_sd15_depth
+sd-1/controlnet/normal_bae:
+ repo_id: lllyasviel/control_v11p_sd15_normalbae
+sd-1/controlnet/seg:
+ repo_id: lllyasviel/control_v11p_sd15_seg
+sd-1/controlnet/lineart:
+ repo_id: lllyasviel/control_v11p_sd15_lineart
+sd-1/controlnet/lineart_anime:
+ repo_id: lllyasviel/control_v11p_sd15s2_lineart_anime
+sd-1/controlnet/scribble:
+ repo_id: lllyasviel/control_v11p_sd15_scribble
+sd-1/controlnet/softedge:
+ repo_id: lllyasviel/control_v11p_sd15_softedge
+sd-1/controlnet/shuffle:
+ repo_id: lllyasviel/control_v11e_sd15_shuffle
+sd-1/controlnet/tile:
+ repo_id: lllyasviel/control_v11f1e_sd15_tile
+sd-1/controlnet/ip2p:
+ repo_id: lllyasviel/control_v11e_sd15_ip2p
+sd-1/embedding/EasyNegative:
+ path: https://huggingface.co/embed/EasyNegative/resolve/main/EasyNegative.safetensors
+sd-1/embedding/ahx-beta-453407d:
+ repo_id: sd-concepts-library/ahx-beta-453407d
+sd-1/lora/LowRA:
+ path: https://civitai.com/api/download/models/63006
+sd-1/lora/Ink scenery:
+ path: https://civitai.com/api/download/models/83390
diff --git a/invokeai/configs/stable-diffusion/v2-inpainting-inference-v.yaml b/invokeai/configs/stable-diffusion/v2-inpainting-inference-v.yaml
new file mode 100644
index 0000000000..37cda460aa
--- /dev/null
+++ b/invokeai/configs/stable-diffusion/v2-inpainting-inference-v.yaml
@@ -0,0 +1,159 @@
+model:
+ base_learning_rate: 5.0e-05
+ target: ldm.models.diffusion.ddpm.LatentInpaintDiffusion
+ params:
+ linear_start: 0.00085
+ linear_end: 0.0120
+ parameterization: "v"
+ num_timesteps_cond: 1
+ log_every_t: 200
+ timesteps: 1000
+ first_stage_key: "jpg"
+ cond_stage_key: "txt"
+ image_size: 64
+ channels: 4
+ cond_stage_trainable: false
+ conditioning_key: hybrid
+ scale_factor: 0.18215
+ monitor: val/loss_simple_ema
+ finetune_keys: null
+ use_ema: False
+
+ unet_config:
+ target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+ params:
+ use_checkpoint: True
+ image_size: 32 # unused
+ in_channels: 9
+ out_channels: 4
+ model_channels: 320
+ attention_resolutions: [ 4, 2, 1 ]
+ num_res_blocks: 2
+ channel_mult: [ 1, 2, 4, 4 ]
+ num_head_channels: 64 # need to fix for flash-attn
+ use_spatial_transformer: True
+ use_linear_in_transformer: True
+ transformer_depth: 1
+ context_dim: 1024
+ legacy: False
+
+ first_stage_config:
+ target: ldm.models.autoencoder.AutoencoderKL
+ params:
+ embed_dim: 4
+ monitor: val/rec_loss
+ ddconfig:
+ #attn_type: "vanilla-xformers"
+ double_z: true
+ z_channels: 4
+ resolution: 256
+ in_channels: 3
+ out_ch: 3
+ ch: 128
+ ch_mult:
+ - 1
+ - 2
+ - 4
+ - 4
+ num_res_blocks: 2
+ attn_resolutions: [ ]
+ dropout: 0.0
+ lossconfig:
+ target: torch.nn.Identity
+
+ cond_stage_config:
+ target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder
+ params:
+ freeze: True
+ layer: "penultimate"
+
+
+data:
+ target: ldm.data.laion.WebDataModuleFromConfig
+ params:
+ tar_base: null # for concat as in LAION-A
+ p_unsafe_threshold: 0.1
+ filter_word_list: "data/filters.yaml"
+ max_pwatermark: 0.45
+ batch_size: 8
+ num_workers: 6
+ multinode: True
+ min_size: 512
+ train:
+ shards:
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-0/{00000..18699}.tar -"
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-1/{00000..18699}.tar -"
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-2/{00000..18699}.tar -"
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-3/{00000..18699}.tar -"
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-4/{00000..18699}.tar -" #{00000-94333}.tar"
+ shuffle: 10000
+ image_key: jpg
+ image_transforms:
+ - target: torchvision.transforms.Resize
+ params:
+ size: 512
+ interpolation: 3
+ - target: torchvision.transforms.RandomCrop
+ params:
+ size: 512
+ postprocess:
+ target: ldm.data.laion.AddMask
+ params:
+ mode: "512train-large"
+ p_drop: 0.25
+ # NOTE use enough shards to avoid empty validation loops in workers
+ validation:
+ shards:
+ - "pipe:aws s3 cp s3://deep-floyd-s3/datasets/laion_cleaned-part5/{93001..94333}.tar - "
+ shuffle: 0
+ image_key: jpg
+ image_transforms:
+ - target: torchvision.transforms.Resize
+ params:
+ size: 512
+ interpolation: 3
+ - target: torchvision.transforms.CenterCrop
+ params:
+ size: 512
+ postprocess:
+ target: ldm.data.laion.AddMask
+ params:
+ mode: "512train-large"
+ p_drop: 0.25
+
+lightning:
+ find_unused_parameters: True
+ modelcheckpoint:
+ params:
+ every_n_train_steps: 5000
+
+ callbacks:
+ metrics_over_trainsteps_checkpoint:
+ params:
+ every_n_train_steps: 10000
+
+ image_logger:
+ target: main.ImageLogger
+ params:
+ enable_autocast: False
+ disabled: False
+ batch_frequency: 1000
+ max_images: 4
+ increase_log_steps: False
+ log_first_step: False
+ log_images_kwargs:
+ use_ema_scope: False
+ inpaint: False
+ plot_progressive_rows: False
+ plot_diffusion_rows: False
+ N: 4
+ unconditional_guidance_scale: 5.0
+ unconditional_guidance_label: [""]
+ ddim_steps: 50 # todo check these out for depth2img,
+ ddim_eta: 0.0 # todo check these out for depth2img,
+
+ trainer:
+ benchmark: True
+ val_check_interval: 5000000
+ num_sanity_val_steps: 0
+ accumulate_grad_batches: 1
\ No newline at end of file
diff --git a/invokeai/configs/stable-diffusion/v2-inpainting-inference.yaml b/invokeai/configs/stable-diffusion/v2-inpainting-inference.yaml
new file mode 100644
index 0000000000..5aaf13162d
--- /dev/null
+++ b/invokeai/configs/stable-diffusion/v2-inpainting-inference.yaml
@@ -0,0 +1,158 @@
+model:
+ base_learning_rate: 5.0e-05
+ target: ldm.models.diffusion.ddpm.LatentInpaintDiffusion
+ params:
+ linear_start: 0.00085
+ linear_end: 0.0120
+ num_timesteps_cond: 1
+ log_every_t: 200
+ timesteps: 1000
+ first_stage_key: "jpg"
+ cond_stage_key: "txt"
+ image_size: 64
+ channels: 4
+ cond_stage_trainable: false
+ conditioning_key: hybrid
+ scale_factor: 0.18215
+ monitor: val/loss_simple_ema
+ finetune_keys: null
+ use_ema: False
+
+ unet_config:
+ target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+ params:
+ use_checkpoint: True
+ image_size: 32 # unused
+ in_channels: 9
+ out_channels: 4
+ model_channels: 320
+ attention_resolutions: [ 4, 2, 1 ]
+ num_res_blocks: 2
+ channel_mult: [ 1, 2, 4, 4 ]
+ num_head_channels: 64 # need to fix for flash-attn
+ use_spatial_transformer: True
+ use_linear_in_transformer: True
+ transformer_depth: 1
+ context_dim: 1024
+ legacy: False
+
+ first_stage_config:
+ target: ldm.models.autoencoder.AutoencoderKL
+ params:
+ embed_dim: 4
+ monitor: val/rec_loss
+ ddconfig:
+ #attn_type: "vanilla-xformers"
+ double_z: true
+ z_channels: 4
+ resolution: 256
+ in_channels: 3
+ out_ch: 3
+ ch: 128
+ ch_mult:
+ - 1
+ - 2
+ - 4
+ - 4
+ num_res_blocks: 2
+ attn_resolutions: [ ]
+ dropout: 0.0
+ lossconfig:
+ target: torch.nn.Identity
+
+ cond_stage_config:
+ target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder
+ params:
+ freeze: True
+ layer: "penultimate"
+
+
+data:
+ target: ldm.data.laion.WebDataModuleFromConfig
+ params:
+ tar_base: null # for concat as in LAION-A
+ p_unsafe_threshold: 0.1
+ filter_word_list: "data/filters.yaml"
+ max_pwatermark: 0.45
+ batch_size: 8
+ num_workers: 6
+ multinode: True
+ min_size: 512
+ train:
+ shards:
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-0/{00000..18699}.tar -"
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-1/{00000..18699}.tar -"
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-2/{00000..18699}.tar -"
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-3/{00000..18699}.tar -"
+ - "pipe:aws s3 cp s3://stability-aws/laion-a-native/part-4/{00000..18699}.tar -" #{00000-94333}.tar"
+ shuffle: 10000
+ image_key: jpg
+ image_transforms:
+ - target: torchvision.transforms.Resize
+ params:
+ size: 512
+ interpolation: 3
+ - target: torchvision.transforms.RandomCrop
+ params:
+ size: 512
+ postprocess:
+ target: ldm.data.laion.AddMask
+ params:
+ mode: "512train-large"
+ p_drop: 0.25
+ # NOTE use enough shards to avoid empty validation loops in workers
+ validation:
+ shards:
+ - "pipe:aws s3 cp s3://deep-floyd-s3/datasets/laion_cleaned-part5/{93001..94333}.tar - "
+ shuffle: 0
+ image_key: jpg
+ image_transforms:
+ - target: torchvision.transforms.Resize
+ params:
+ size: 512
+ interpolation: 3
+ - target: torchvision.transforms.CenterCrop
+ params:
+ size: 512
+ postprocess:
+ target: ldm.data.laion.AddMask
+ params:
+ mode: "512train-large"
+ p_drop: 0.25
+
+lightning:
+ find_unused_parameters: True
+ modelcheckpoint:
+ params:
+ every_n_train_steps: 5000
+
+ callbacks:
+ metrics_over_trainsteps_checkpoint:
+ params:
+ every_n_train_steps: 10000
+
+ image_logger:
+ target: main.ImageLogger
+ params:
+ enable_autocast: False
+ disabled: False
+ batch_frequency: 1000
+ max_images: 4
+ increase_log_steps: False
+ log_first_step: False
+ log_images_kwargs:
+ use_ema_scope: False
+ inpaint: False
+ plot_progressive_rows: False
+ plot_diffusion_rows: False
+ N: 4
+ unconditional_guidance_scale: 5.0
+ unconditional_guidance_label: [""]
+ ddim_steps: 50 # todo check these out for depth2img,
+ ddim_eta: 0.0 # todo check these out for depth2img,
+
+ trainer:
+ benchmark: True
+ val_check_interval: 5000000
+ num_sanity_val_steps: 0
+ accumulate_grad_batches: 1
\ No newline at end of file
diff --git a/invokeai/frontend/install/invokeai_update.py b/invokeai/frontend/install/invokeai_update.py
index 18ad71957e..f73f670496 100644
--- a/invokeai/frontend/install/invokeai_update.py
+++ b/invokeai/frontend/install/invokeai_update.py
@@ -108,11 +108,11 @@ def main():
print(f':crossed_fingers: Upgrading to [yellow]{tag if tag else release}[/yellow]')
if release:
- cmd = f"pip install 'invokeai{extras} @ {INVOKE_AI_SRC}/{release}.zip' --use-pep517 --upgrade"
+ cmd = f'pip install "invokeai{extras} @ {INVOKE_AI_SRC}/{release}.zip" --use-pep517 --upgrade'
elif tag:
- cmd = f"pip install 'invokeai{extras} @ {INVOKE_AI_TAG}/{tag}.zip' --use-pep517 --upgrade"
+ cmd = f'pip install "invokeai{extras} @ {INVOKE_AI_TAG}/{tag}.zip" --use-pep517 --upgrade'
else:
- cmd = f"pip install 'invokeai{extras} @ {INVOKE_AI_BRANCH}/{branch}.zip' --use-pep517 --upgrade"
+ cmd = f'pip install "invokeai{extras} @ {INVOKE_AI_BRANCH}/{branch}.zip" --use-pep517 --upgrade'
print('')
print('')
if os.system(cmd)==0:
diff --git a/invokeai/frontend/install/model_install.py b/invokeai/frontend/install/model_install.py
index 265c456e3a..736d27a321 100644
--- a/invokeai/frontend/install/model_install.py
+++ b/invokeai/frontend/install/model_install.py
@@ -11,7 +11,6 @@ The work is actually done in backend code in model_install_backend.py.
import argparse
import curses
-import os
import sys
import textwrap
import traceback
@@ -20,28 +19,22 @@ from multiprocessing import Process
from multiprocessing.connection import Connection, Pipe
from pathlib import Path
from shutil import get_terminal_size
-from typing import List
import logging
import npyscreen
import torch
from npyscreen import widget
-from omegaconf import OmegaConf
from invokeai.backend.util.logging import InvokeAILogger
from invokeai.backend.install.model_install_backend import (
- Dataset_path,
- default_config_file,
- default_dataset,
- install_requested_models,
- recommended_datasets,
ModelInstallList,
- UserSelections,
+ InstallSelections,
+ ModelInstall,
+ SchedulerPredictionType,
)
-from invokeai.backend import ModelManager
+from invokeai.backend.model_management import ModelManager, ModelType
from invokeai.backend.util import choose_precision, choose_torch_device
-from invokeai.backend.util.logging import InvokeAILogger
from invokeai.frontend.install.widgets import (
CenteredTitleText,
MultiSelectColumns,
@@ -58,6 +51,7 @@ from invokeai.frontend.install.widgets import (
from invokeai.app.services.config import InvokeAIAppConfig
config = InvokeAIAppConfig.get_config()
+logger = InvokeAILogger.getLogger()
# build a table mapping all non-printable characters to None
# for stripping control characters
@@ -71,8 +65,8 @@ def make_printable(s:str)->str:
return s.translate(NOPRINT_TRANS_TABLE)
class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
- # for responsive resizing - disabled
- # FIX_MINIMUM_SIZE_WHEN_CREATED = False
+ # for responsive resizing set to False, but this seems to cause a crash!
+ FIX_MINIMUM_SIZE_WHEN_CREATED = True
# for persistence
current_tab = 0
@@ -90,25 +84,10 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
if not config.model_conf_path.exists():
with open(config.model_conf_path,'w') as file:
print('# InvokeAI model configuration file',file=file)
- model_manager = ModelManager(config.model_conf_path)
-
- self.starter_models = OmegaConf.load(Dataset_path)['diffusers']
- self.installed_diffusers_models = self.list_additional_diffusers_models(
- model_manager,
- self.starter_models,
- )
- self.installed_cn_models = model_manager.list_controlnet_models()
- self.installed_lora_models = model_manager.list_lora_models()
- self.installed_ti_models = model_manager.list_ti_models()
-
- try:
- self.existing_models = OmegaConf.load(default_config_file())
- except:
- self.existing_models = dict()
-
- self.starter_model_list = list(self.starter_models.keys())
- self.installed_models = dict()
-
+ self.installer = ModelInstall(config)
+ self.all_models = self.installer.all_models()
+ self.starter_models = self.installer.starter_models()
+ self.model_labels = self._get_model_labels()
window_width, window_height = get_terminal_size()
self.nextrely -= 1
@@ -141,39 +120,37 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
scroll_exit = True,
)
self.tabs.on_changed = self._toggle_tables
-
+
top_of_table = self.nextrely
- self.starter_diffusers_models = self.add_starter_diffusers()
+ self.starter_pipelines = self.add_starter_pipelines()
bottom_of_table = self.nextrely
self.nextrely = top_of_table
- self.diffusers_models = self.add_diffusers_widgets(
- predefined_models=self.installed_diffusers_models,
- model_type='Diffusers',
+ self.pipeline_models = self.add_pipeline_widgets(
+ model_type=ModelType.Main,
window_width=window_width,
+ exclude = self.starter_models
)
+ # self.pipeline_models['autoload_pending'] = True
bottom_of_table = max(bottom_of_table,self.nextrely)
self.nextrely = top_of_table
self.controlnet_models = self.add_model_widgets(
- predefined_models=self.installed_cn_models,
- model_type='ControlNet',
+ model_type=ModelType.ControlNet,
window_width=window_width,
)
bottom_of_table = max(bottom_of_table,self.nextrely)
self.nextrely = top_of_table
self.lora_models = self.add_model_widgets(
- predefined_models=self.installed_lora_models,
- model_type="LoRA/LyCORIS",
+ model_type=ModelType.Lora,
window_width=window_width,
)
bottom_of_table = max(bottom_of_table,self.nextrely)
self.nextrely = top_of_table
self.ti_models = self.add_model_widgets(
- predefined_models=self.installed_ti_models,
- model_type="Textual Inversion Embeddings",
+ model_type=ModelType.TextualInversion,
window_width=window_width,
)
bottom_of_table = max(bottom_of_table,self.nextrely)
@@ -184,7 +161,7 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
BufferBox,
name='Log Messages',
editable=False,
- max_height = 16,
+ max_height = 10,
)
self.nextrely += 1
@@ -197,13 +174,14 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
rely=-3,
when_pressed_function=self.on_back,
)
- self.ok_button = self.add_widget_intelligent(
- npyscreen.ButtonPress,
- name=done_label,
- relx=(window_width - len(done_label)) // 2,
- rely=-3,
- when_pressed_function=self.on_execute
- )
+ else:
+ self.ok_button = self.add_widget_intelligent(
+ npyscreen.ButtonPress,
+ name=done_label,
+ relx=(window_width - len(done_label)) // 2,
+ rely=-3,
+ when_pressed_function=self.on_execute
+ )
label = "APPLY CHANGES & EXIT"
self.done = self.add_widget_intelligent(
@@ -220,18 +198,15 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
self._toggle_tables([self.current_tab])
############# diffusers tab ##########
- def add_starter_diffusers(self)->dict[str, npyscreen.widget]:
+ def add_starter_pipelines(self)->dict[str, npyscreen.widget]:
'''Add widgets responsible for selecting diffusers models'''
widgets = dict()
-
- starter_model_labels = self._get_starter_model_labels()
- recommended_models = [
- x
- for x in self.starter_model_list
- if self.starter_models[x].get("recommended", False)
- ]
+ models = self.all_models
+ starters = self.starter_models
+ starter_model_labels = self.model_labels
+
self.installed_models = sorted(
- [x for x in list(self.starter_models.keys()) if x in self.existing_models]
+ [x for x in starters if models[x].installed]
)
widgets.update(
@@ -246,55 +221,46 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
self.nextrely -= 1
# if user has already installed some initial models, then don't patronize them
# by showing more recommendations
- show_recommended = not self.existing_models
+ show_recommended = len(self.installed_models)==0
+ keys = [x for x in models.keys() if x in starters]
widgets.update(
models_selected = self.add_widget_intelligent(
MultiSelectColumns,
columns=1,
name="Install Starter Models",
- values=starter_model_labels,
+ values=[starter_model_labels[x] for x in keys],
value=[
- self.starter_model_list.index(x)
- for x in self.starter_model_list
- if (show_recommended and x in recommended_models)\
- or (x in self.existing_models)
+ keys.index(x)
+ for x in keys
+ if (show_recommended and models[x].recommended) \
+ or (x in self.installed_models)
],
- max_height=len(starter_model_labels) + 1,
+ max_height=len(starters) + 1,
relx=4,
scroll_exit=True,
- )
+ ),
+ models = keys,
)
- widgets.update(
- purge_deleted = self.add_widget_intelligent(
- npyscreen.Checkbox,
- name="Purge unchecked diffusers models from disk",
- value=False,
- scroll_exit=True,
- relx=4,
- )
- )
- widgets['purge_deleted'].when_value_edited = lambda: self.sync_purge_buttons(widgets['purge_deleted'])
-
self.nextrely += 1
return widgets
############# Add a set of model install widgets ########
def add_model_widgets(self,
- predefined_models: dict[str,bool],
- model_type: str,
+ model_type: ModelType,
window_width: int=120,
install_prompt: str=None,
- add_purge_deleted: bool=False,
+ exclude: set=set(),
)->dict[str,npyscreen.widget]:
'''Generic code to create model selection widgets'''
widgets = dict()
- model_list = sorted(predefined_models.keys())
+ model_list = [x for x in self.all_models if self.all_models[x].model_type==model_type and not x in exclude]
+ model_labels = [self.model_labels[x] for x in model_list]
if len(model_list) > 0:
- max_width = max([len(x) for x in model_list])
+ max_width = max([len(x) for x in model_labels])
columns = window_width // (max_width+8) # 8 characters for "[x] " and padding
columns = min(len(model_list),columns) or 1
- prompt = install_prompt or f"Select the desired {model_type} models to install. Unchecked models will be purged from disk."
+ prompt = install_prompt or f"Select the desired {model_type.value.title()} models to install. Unchecked models will be purged from disk."
widgets.update(
label1 = self.add_widget_intelligent(
@@ -310,31 +276,19 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
MultiSelectColumns,
columns=columns,
name=f"Install {model_type} Models",
- values=model_list,
+ values=model_labels,
value=[
model_list.index(x)
for x in model_list
- if predefined_models[x]
+ if self.all_models[x].installed
],
max_height=len(model_list)//columns + 1,
relx=4,
scroll_exit=True,
- )
+ ),
+ models = model_list,
)
- if add_purge_deleted:
- self.nextrely += 1
- widgets.update(
- purge_deleted = self.add_widget_intelligent(
- npyscreen.Checkbox,
- name="Purge unchecked diffusers models from disk",
- value=False,
- scroll_exit=True,
- relx=4,
- )
- )
- widgets['purge_deleted'].when_value_edited = lambda: self.sync_purge_buttons(widgets['purge_deleted'])
-
self.nextrely += 1
widgets.update(
download_ids = self.add_widget_intelligent(
@@ -348,63 +302,33 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
return widgets
### Tab for arbitrary diffusers widgets ###
- def add_diffusers_widgets(self,
- predefined_models: dict[str,bool],
- model_type: str='Diffusers',
- window_width: int=120,
- )->dict[str,npyscreen.widget]:
+ def add_pipeline_widgets(self,
+ model_type: ModelType=ModelType.Main,
+ window_width: int=120,
+ **kwargs,
+ )->dict[str,npyscreen.widget]:
'''Similar to add_model_widgets() but adds some additional widgets at the bottom
to support the autoload directory'''
widgets = self.add_model_widgets(
- predefined_models,
- 'Diffusers',
- window_width,
- install_prompt="Additional diffusers models already installed.",
- add_purge_deleted=True
+ model_type = model_type,
+ window_width = window_width,
+ install_prompt=f"Additional {model_type.value.title()} models already installed.",
+ **kwargs,
)
- label = "Directory to scan for models to automatically import ( autocompletes):"
- self.nextrely += 1
- widgets.update(
- autoload_directory = self.add_widget_intelligent(
- FileBox,
- max_height=3,
- name=label,
- value=str(config.autoconvert_dir) if config.autoconvert_dir else None,
- select_dir=True,
- must_exist=True,
- use_two_lines=False,
- labelColor="DANGER",
- begin_entry_at=len(label)+1,
- scroll_exit=True,
- )
- )
- widgets.update(
- autoscan_on_startup = self.add_widget_intelligent(
- npyscreen.Checkbox,
- name="Scan and import from this directory each time InvokeAI starts",
- value=config.autoconvert_dir is not None,
- relx=4,
- scroll_exit=True,
- )
- )
return widgets
- def sync_purge_buttons(self,checkbox):
- value = checkbox.value
- self.starter_diffusers_models['purge_deleted'].value = value
- self.diffusers_models['purge_deleted'].value = value
-
def resize(self):
super().resize()
- if (s := self.starter_diffusers_models.get("models_selected")):
- s.values = self._get_starter_model_labels()
+ if (s := self.starter_pipelines.get("models_selected")):
+ keys = [x for x in self.all_models.keys() if x in self.starter_models]
+ s.values = [self.model_labels[x] for x in keys]
def _toggle_tables(self, value=None):
selected_tab = value[0]
widgets = [
- self.starter_diffusers_models,
- self.diffusers_models,
+ self.starter_pipelines,
+ self.pipeline_models,
self.controlnet_models,
self.lora_models,
self.ti_models,
@@ -412,34 +336,38 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
for group in widgets:
for k,v in group.items():
- v.hidden = True
- v.editable = False
+ try:
+ v.hidden = True
+ v.editable = False
+ except:
+ pass
for k,v in widgets[selected_tab].items():
- v.hidden = False
- if not isinstance(v,(npyscreen.FixedText, npyscreen.TitleFixedText, CenteredTitleText)):
- v.editable = True
+ try:
+ v.hidden = False
+ if not isinstance(v,(npyscreen.FixedText, npyscreen.TitleFixedText, CenteredTitleText)):
+ v.editable = True
+ except:
+ pass
self.__class__.current_tab = selected_tab # for persistence
self.display()
- def _get_starter_model_labels(self) -> List[str]:
+ def _get_model_labels(self) -> dict[str,str]:
window_width, window_height = get_terminal_size()
- label_width = 25
checkbox_width = 4
spacing_width = 2
+
+ models = self.all_models
+ label_width = max([len(models[x].name) for x in models])
description_width = window_width - label_width - checkbox_width - spacing_width
- im = self.starter_models
- names = self.starter_model_list
- descriptions = [
- im[x].description[0 : description_width - 3] + "..."
- if len(im[x].description) > description_width
- else im[x].description
- for x in names
- ]
- return [
- f"%-{label_width}s %s" % (names[x], descriptions[x])
- for x in range(0, len(names))
- ]
+ result = dict()
+ for x in models.keys():
+ description = models[x].description
+ description = description[0 : description_width - 3] + "..." \
+ if description and len(description) > description_width \
+ else description if description else ''
+ result[x] = f"%-{label_width}s %s" % (models[x].name, description)
+ return result
def _get_columns(self) -> int:
window_width, window_height = get_terminal_size()
@@ -454,10 +382,21 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
)
return min(cols, len(self.installed_models))
+ def confirm_deletions(self, selections: InstallSelections)->bool:
+ remove_models = selections.remove_models
+ if len(remove_models) > 0:
+ mods = "\n".join([ModelManager.parse_key(x)[0] for x in remove_models])
+ return npyscreen.notify_ok_cancel(f"These unchecked models will be deleted from disk. Continue?\n---------\n{mods}")
+ else:
+ return True
+
def on_execute(self):
- self.monitor.entry_widget.buffer(['Processing...'],scroll_end=True)
self.marshall_arguments()
app = self.parentApp
+ if not self.confirm_deletions(app.install_selections):
+ return
+
+ self.monitor.entry_widget.buffer(['Processing...'],scroll_end=True)
self.ok_button.hidden = True
self.display()
@@ -467,7 +406,7 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
target = process_and_execute,
kwargs=dict(
opt = app.program_opts,
- selections = app.user_selections,
+ selections = app.install_selections,
conn_out = child_conn,
)
)
@@ -475,8 +414,8 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
child_conn.close()
self.subprocess_connection = parent_conn
self.subprocess = p
- app.user_selections = UserSelections()
- # process_and_execute(app.opt, app.user_selections)
+ app.install_selections = InstallSelections()
+ # process_and_execute(app.opt, app.install_selections)
def on_back(self):
self.parentApp.switchFormPrevious()
@@ -489,10 +428,12 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
def on_done(self):
self.marshall_arguments()
+ if not self.confirm_deletions(self.parentApp.install_selections):
+ return
self.parentApp.setNextForm(None)
self.parentApp.user_cancelled = False
self.editing = False
-
+
########## This routine monitors the child process that is performing model installation and removal #####
def while_waiting(self):
'''Called during idle periods. Main task is to update the Log Messages box with messages
@@ -548,8 +489,8 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
# rebuild the form, saving and restoring some of the fields that need to be preserved.
saved_messages = self.monitor.entry_widget.values
- autoload_dir = self.diffusers_models['autoload_directory'].value
- autoscan = self.diffusers_models['autoscan_on_startup'].value
+ # autoload_dir = str(config.root_path / self.pipeline_models['autoload_directory'].value)
+ # autoscan = self.pipeline_models['autoscan_on_startup'].value
app.main_form = app.addForm(
"MAIN", addModelsForm, name="Install Stable Diffusion Models", multipage=self.multipage,
@@ -558,23 +499,8 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
app.main_form.monitor.entry_widget.values = saved_messages
app.main_form.monitor.entry_widget.buffer([''],scroll_end=True)
- app.main_form.diffusers_models['autoload_directory'].value = autoload_dir
- app.main_form.diffusers_models['autoscan_on_startup'].value = autoscan
-
- ###############################################################
-
- def list_additional_diffusers_models(self,
- manager: ModelManager,
- starters:dict
- )->dict[str,bool]:
- '''Return a dict of all the currently installed models that are not on the starter list'''
- model_info = manager.list_models()
- additional_models = {
- x:True for x in model_info \
- if model_info[x]['format']=='diffusers' \
- and x not in starters
- }
- return additional_models
+ # app.main_form.pipeline_models['autoload_directory'].value = autoload_dir
+ # app.main_form.pipeline_models['autoscan_on_startup'].value = autoscan
def marshall_arguments(self):
"""
@@ -586,89 +512,40 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
.autoscan_on_startup: True if invokeai should scan and import at startup time
.import_model_paths: list of URLs, repo_ids and file paths to import
"""
- # we're using a global here rather than storing the result in the parentapp
- # due to some bug in npyscreen that is causing attributes to be lost
- selections = self.parentApp.user_selections
+ selections = self.parentApp.install_selections
+ all_models = self.all_models
- # Starter models to install/remove
- starter_models = dict(
- map(
- lambda x: (self.starter_model_list[x], True),
- self.starter_diffusers_models['models_selected'].value,
- )
- )
- selections.purge_deleted_models = self.starter_diffusers_models['purge_deleted'].value or \
- self.diffusers_models['purge_deleted'].value
-
- selections.install_models = [x for x in starter_models if x not in self.existing_models]
- selections.remove_models = [x for x in self.starter_model_list if x in self.existing_models and x not in starter_models]
+ # Defined models (in INITIAL_CONFIG.yaml or models.yaml) to add/remove
+ ui_sections = [self.starter_pipelines, self.pipeline_models,
+ self.controlnet_models, self.lora_models, self.ti_models]
+ for section in ui_sections:
+ if not 'models_selected' in section:
+ continue
+ selected = set([section['models'][x] for x in section['models_selected'].value])
+ models_to_install = [x for x in selected if not self.all_models[x].installed]
+ models_to_remove = [x for x in section['models'] if x not in selected and self.all_models[x].installed]
+ selections.remove_models.extend(models_to_remove)
+ selections.install_models.extend(all_models[x].path or all_models[x].repo_id \
+ for x in models_to_install if all_models[x].path or all_models[x].repo_id)
- # "More" models
- selections.import_model_paths = self.diffusers_models['download_ids'].value.split()
- if diffusers_selected := self.diffusers_models.get('models_selected'):
- selections.remove_models.extend([x
- for x in diffusers_selected.values
- if self.installed_diffusers_models[x]
- and diffusers_selected.values.index(x) not in diffusers_selected.value
- ]
- )
-
- # TODO: REFACTOR THIS REPETITIVE CODE
- if cn_models_selected := self.controlnet_models.get('models_selected'):
- selections.install_cn_models = [cn_models_selected.values[x]
- for x in cn_models_selected.value
- if not self.installed_cn_models[cn_models_selected.values[x]]
- ]
- selections.remove_cn_models = [x
- for x in cn_models_selected.values
- if self.installed_cn_models[x]
- and cn_models_selected.values.index(x) not in cn_models_selected.value
- ]
- if (additional_cns := self.controlnet_models['download_ids'].value.split()):
- valid_cns = [x for x in additional_cns if '/' in x]
- selections.install_cn_models.extend(valid_cns)
+ # models located in the 'download_ids" section
+ for section in ui_sections:
+ if downloads := section.get('download_ids'):
+ selections.install_models.extend(downloads.value.split())
- # same thing, for LoRAs
- if loras_selected := self.lora_models.get('models_selected'):
- selections.install_lora_models = [loras_selected.values[x]
- for x in loras_selected.value
- if not self.installed_lora_models[loras_selected.values[x]]
- ]
- selections.remove_lora_models = [x
- for x in loras_selected.values
- if self.installed_lora_models[x]
- and loras_selected.values.index(x) not in loras_selected.value
- ]
- if (additional_loras := self.lora_models['download_ids'].value.split()):
- selections.install_lora_models.extend(additional_loras)
-
- # same thing, for TIs
- # TODO: refactor
- if tis_selected := self.ti_models.get('models_selected'):
- selections.install_ti_models = [tis_selected.values[x]
- for x in tis_selected.value
- if not self.installed_ti_models[tis_selected.values[x]]
- ]
- selections.remove_ti_models = [x
- for x in tis_selected.values
- if self.installed_ti_models[x]
- and tis_selected.values.index(x) not in tis_selected.value
- ]
-
- if (additional_tis := self.ti_models['download_ids'].value.split()):
- selections.install_ti_models.extend(additional_tis)
-
# load directory and whether to scan on startup
- selections.scan_directory = self.diffusers_models['autoload_directory'].value
- selections.autoscan_on_startup = self.diffusers_models['autoscan_on_startup'].value
-
+ # if self.parentApp.autoload_pending:
+ # selections.scan_directory = str(config.root_path / self.pipeline_models['autoload_directory'].value)
+ # self.parentApp.autoload_pending = False
+ # selections.autoscan_on_startup = self.pipeline_models['autoscan_on_startup'].value
class AddModelApplication(npyscreen.NPSAppManaged):
def __init__(self,opt):
super().__init__()
self.program_opts = opt
self.user_cancelled = False
- self.user_selections = UserSelections()
+ # self.autoload_pending = True
+ self.install_selections = InstallSelections()
def onStart(self):
npyscreen.setTheme(npyscreen.Themes.DefaultTheme)
@@ -687,26 +564,22 @@ class StderrToMessage():
pass
# --------------------------------------------------------
-def ask_user_for_config_file(model_path: Path,
- tui_conn: Connection=None
- )->Path:
+def ask_user_for_prediction_type(model_path: Path,
+ tui_conn: Connection=None
+ )->SchedulerPredictionType:
if tui_conn:
logger.debug('Waiting for user response...')
- return _ask_user_for_cf_tui(model_path, tui_conn)
+ return _ask_user_for_pt_tui(model_path, tui_conn)
else:
- return _ask_user_for_cf_cmdline(model_path)
+ return _ask_user_for_pt_cmdline(model_path)
-def _ask_user_for_cf_cmdline(model_path):
- choices = [
- config.legacy_conf_path / x
- for x in ['v2-inference.yaml','v2-inference-v.yaml']
- ]
- choices.extend([None])
+def _ask_user_for_pt_cmdline(model_path: Path)->SchedulerPredictionType:
+ choices = [SchedulerPredictionType.Epsilon, SchedulerPredictionType.VPrediction, None]
print(
f"""
Please select the type of the V2 checkpoint named {model_path.name}:
-[1] A Stable Diffusion v2.x base model (512 pixels; there should be no 'parameterization:' line in its yaml file)
-[2] A Stable Diffusion v2.x v-predictive model (768 pixels; look for a 'parameterization: "v"' line in its yaml file)
+[1] A model based on Stable Diffusion v2 trained on 512 pixel images (SD-2-base)
+[2] A model based on Stable Diffusion v2 trained on 768 pixel images (SD-2-768)
[3] Skip this model and come back later.
"""
)
@@ -723,7 +596,7 @@ Please select the type of the V2 checkpoint named {model_path.name}:
return
return choice
-def _ask_user_for_cf_tui(model_path: Path, tui_conn: Connection)->Path:
+def _ask_user_for_pt_tui(model_path: Path, tui_conn: Connection)->SchedulerPredictionType:
try:
tui_conn.send_bytes(f'*need v2 config for:{model_path}'.encode('utf-8'))
# note that we don't do any status checking here
@@ -731,20 +604,20 @@ def _ask_user_for_cf_tui(model_path: Path, tui_conn: Connection)->Path:
if response is None:
return None
elif response == 'epsilon':
- return config.legacy_conf_path / 'v2-inference.yaml'
+ return SchedulerPredictionType.epsilon
elif response == 'v':
- return config.legacy_conf_path / 'v2-inference-v.yaml'
+ return SchedulerPredictionType.VPrediction
elif response == 'abort':
logger.info('Conversion aborted')
return None
else:
- return Path(response)
+ return response
except:
return None
# --------------------------------------------------------
def process_and_execute(opt: Namespace,
- selections: UserSelections,
+ selections: InstallSelections,
conn_out: Connection=None,
):
# set up so that stderr is sent to conn_out
@@ -755,34 +628,14 @@ def process_and_execute(opt: Namespace,
logger = InvokeAILogger.getLogger()
logger.handlers.clear()
logger.addHandler(logging.StreamHandler(translator))
-
- models_to_install = selections.install_models
- models_to_remove = selections.remove_models
- directory_to_scan = selections.scan_directory
- scan_at_startup = selections.autoscan_on_startup
- potential_models_to_install = selections.import_model_paths
- install_requested_models(
- diffusers = ModelInstallList(models_to_install, models_to_remove),
- controlnet = ModelInstallList(selections.install_cn_models, selections.remove_cn_models),
- lora = ModelInstallList(selections.install_lora_models, selections.remove_lora_models),
- ti = ModelInstallList(selections.install_ti_models, selections.remove_ti_models),
- scan_directory=Path(directory_to_scan) if directory_to_scan else None,
- external_models=potential_models_to_install,
- scan_at_startup=scan_at_startup,
- precision="float32"
- if opt.full_precision
- else choose_precision(torch.device(choose_torch_device())),
- purge_deleted=selections.purge_deleted_models,
- config_file_path=Path(opt.config_file) if opt.config_file else config.model_conf_path,
- model_config_file_callback = lambda x: ask_user_for_config_file(x,conn_out)
- )
+ installer = ModelInstall(config, prediction_type_helper=lambda x: ask_user_for_prediction_type(x,conn_out))
+ installer.install(selections)
if conn_out:
conn_out.send_bytes('*done*'.encode('utf-8'))
conn_out.close()
-
def do_listings(opt)->bool:
"""List installed models of various sorts, and return
True if any were requested."""
@@ -813,38 +666,32 @@ def select_and_download_models(opt: Namespace):
if opt.full_precision
else choose_precision(torch.device(choose_torch_device()))
)
-
- if do_listings(opt):
- pass
- # this processes command line additions/removals
- elif opt.diffusers or opt.controlnets or opt.textual_inversions or opt.loras:
- action = 'remove_models' if opt.delete else 'install_models'
- diffusers_args = {'diffusers':ModelInstallList(remove_models=opt.diffusers or [])} \
- if opt.delete \
- else {'external_models':opt.diffusers or []}
- install_requested_models(
- **diffusers_args,
- controlnet=ModelInstallList(**{action:opt.controlnets or []}),
- ti=ModelInstallList(**{action:opt.textual_inversions or []}),
- lora=ModelInstallList(**{action:opt.loras or []}),
- precision=precision,
- purge_deleted=True,
- model_config_file_callback=lambda x: ask_user_for_config_file(x),
+ config.precision = precision
+ helper = lambda x: ask_user_for_prediction_type(x)
+ # if do_listings(opt):
+ # pass
+
+ installer = ModelInstall(config, prediction_type_helper=helper)
+ if opt.add or opt.delete:
+ selections = InstallSelections(
+ install_models = opt.add or [],
+ remove_models = opt.delete or []
)
+ installer.install(selections)
elif opt.default_only:
- install_requested_models(
- diffusers=ModelInstallList(install_models=default_dataset()),
- precision=precision,
+ selections = InstallSelections(
+ install_models = installer.default_model()
)
+ installer.install(selections)
elif opt.yes_to_all:
- install_requested_models(
- diffusers=ModelInstallList(install_models=recommended_datasets()),
- precision=precision,
+ selections = InstallSelections(
+ install_models = installer.recommended_models()
)
+ installer.install(selections)
# this is where the TUI is called
else:
- # needed because the torch library is loaded, even though we don't use it
+ # needed to support the probe() method running under a subprocess
torch.multiprocessing.set_start_method("spawn")
# the third argument is needed in the Windows 11 environment in
@@ -861,35 +708,20 @@ def select_and_download_models(opt: Namespace):
installApp.main_form.subprocess.terminate()
installApp.main_form.subprocess = None
raise e
- process_and_execute(opt, installApp.user_selections)
+ process_and_execute(opt, installApp.install_selections)
# -------------------------------------
def main():
parser = argparse.ArgumentParser(description="InvokeAI model downloader")
parser.add_argument(
- "--diffusers",
+ "--add",
nargs="*",
- help="List of URLs or repo_ids of diffusers to install/delete",
- )
- parser.add_argument(
- "--loras",
- nargs="*",
- help="List of URLs or repo_ids of LoRA/LyCORIS models to install/delete",
- )
- parser.add_argument(
- "--controlnets",
- nargs="*",
- help="List of URLs or repo_ids of controlnet models to install/delete",
- )
- parser.add_argument(
- "--textual-inversions",
- nargs="*",
- help="List of URLs or repo_ids of textual inversion embeddings to install/delete",
+ help="List of URLs, local paths or repo_ids of models to install",
)
parser.add_argument(
"--delete",
- action="store_true",
- help="Delete models listed on command line rather than installing them",
+ nargs="*",
+ help="List of names of models to idelete",
)
parser.add_argument(
"--full-precision",
@@ -909,7 +741,7 @@ def main():
parser.add_argument(
"--default_only",
action="store_true",
- help="only install the default model",
+ help="Only install the default model",
)
parser.add_argument(
"--list-models",
@@ -941,7 +773,7 @@ def main():
config.parse_args(invoke_args)
logger = InvokeAILogger().getLogger(config=config)
- if not (config.root_dir / config.conf_path.parent).exists():
+ if not (config.conf_path / 'models.yaml').exists():
logger.info(
"Your InvokeAI root directory is not set up. Calling invokeai-configure."
)
@@ -965,13 +797,15 @@ def main():
logger.error(
"Insufficient vertical space for the interface. Please make your window taller and try again"
)
- elif str(e).startswith("addwstr"):
+ input('Press any key to continue...')
+ except Exception as e:
+ if str(e).startswith("addwstr"):
logger.error(
"Insufficient horizontal space for the interface. Please make your window wider and try again."
)
- except Exception as e:
- print(f'An exception has occurred: {str(e)} Details:')
- print(traceback.format_exc(), file=sys.stderr)
+ else:
+ print(f'An exception has occurred: {str(e)} Details:')
+ print(traceback.format_exc(), file=sys.stderr)
input('Press any key to continue...')
diff --git a/invokeai/frontend/install/widgets.py b/invokeai/frontend/install/widgets.py
index 14167d4ee0..c3bb3542ae 100644
--- a/invokeai/frontend/install/widgets.py
+++ b/invokeai/frontend/install/widgets.py
@@ -17,8 +17,8 @@ from shutil import get_terminal_size
from curses import BUTTON2_CLICKED,BUTTON3_CLICKED
# minimum size for UIs
-MIN_COLS = 120
-MIN_LINES = 50
+MIN_COLS = 130
+MIN_LINES = 45
# -------------------------------------
def set_terminal_size(columns: int, lines: int, launch_command: str=None):
@@ -42,6 +42,18 @@ def set_terminal_size(columns: int, lines: int, launch_command: str=None):
elif OS in ["Darwin", "Linux"]:
_set_terminal_size_unix(width,height)
+ # check whether it worked....
+ ts = get_terminal_size()
+ pause = False
+ if ts.columns < columns:
+ print('\033[1mThis window is too narrow for the user interface. Please make it wider.\033[0m')
+ pause = True
+ if ts.lines < lines:
+ print('\033[1mThis window is too short for the user interface. Please make it taller.\033[0m')
+ pause = True
+ if pause:
+ input('Press any key to continue..')
+
def _set_terminal_size_powershell(width: int, height: int):
script=f'''
$pshost = get-host
@@ -61,6 +73,12 @@ def _set_terminal_size_unix(width: int, height: int):
import fcntl
import termios
+ # These terminals accept the size command and report that the
+ # size changed, but they lie!!!
+ for bad_terminal in ['TERMINATOR_UUID', 'ALACRITTY_WINDOW_ID']:
+ if os.environ.get(bad_terminal):
+ return
+
winsize = struct.pack("HHHH", height, width, 0, 0)
fcntl.ioctl(sys.stdout.fileno(), termios.TIOCSWINSZ, winsize)
sys.stdout.write("\x1b[8;{height};{width}t".format(height=height, width=width))
@@ -75,6 +93,12 @@ def set_min_terminal_size(min_cols: int, min_lines: int, launch_command: str=Non
lines = max(term_lines, min_lines)
set_terminal_size(cols, lines, launch_command)
+ # did it work?
+ term_cols, term_lines = get_terminal_size()
+ if term_cols < cols or term_lines < lines:
+ print(f'This window is too small for optimal display. For best results please enlarge it.')
+ input('After resizing, press any key to continue...')
+
class IntSlider(npyscreen.Slider):
def translate_value(self):
stri = "%2d / %2d" % (self.value, self.out_of)
@@ -378,13 +402,12 @@ def select_stable_diffusion_config_file(
wrap:bool =True,
model_name:str='Unknown',
):
- message = "Please select the correct base model for the V2 checkpoint named {model_name}. Press to skip installation."
+ message = f"Please select the correct base model for the V2 checkpoint named '{model_name}'. Press to skip installation."
title = "CONFIG FILE SELECTION"
options=[
"An SD v2.x base model (512 pixels; no 'parameterization:' line in its yaml file)",
"An SD v2.x v-predictive model (768 pixels; 'parameterization: \"v\"' line in its yaml file)",
"Skip installation for now and come back later",
- "Enter config file path manually",
]
F = ConfirmCancelPopup(
@@ -406,35 +429,17 @@ def select_stable_diffusion_config_file(
mlw.values = message
choice = F.add(
- SingleSelectWithChanged,
+ npyscreen.SelectOne,
values = options,
value = [0],
max_height = len(options)+1,
scroll_exit=True,
)
- file = F.add(
- FileBox,
- name='Path to config file',
- max_height=3,
- hidden=True,
- must_exist=True,
- scroll_exit=True
- )
-
- def toggle_visible(value):
- value = value[0]
- if value==3:
- file.hidden=False
- else:
- file.hidden=True
- F.display()
-
- choice.on_changed = toggle_visible
F.editw = 1
F.edit()
if not F.value:
return None
- assert choice.value[0] in range(0,4),'invalid choice'
- choices = ['epsilon','v','abort',file.value]
+ assert choice.value[0] in range(0,3),'invalid choice'
+ choices = ['epsilon','v','abort']
return choices[choice.value[0]]
diff --git a/invokeai/frontend/legacy_launch_invokeai.py b/invokeai/frontend/legacy_launch_invokeai.py
new file mode 100644
index 0000000000..349fa5b945
--- /dev/null
+++ b/invokeai/frontend/legacy_launch_invokeai.py
@@ -0,0 +1,19 @@
+import os
+import sys
+import argparse
+
+def main():
+ parser = argparse.ArgumentParser()
+ parser.add_argument('--web', action='store_true')
+ opts,_ = parser.parse_known_args()
+
+ if opts.web:
+ sys.argv.pop(sys.argv.index('--web'))
+ from invokeai.app.api_app import invoke_api
+ invoke_api()
+ else:
+ from invokeai.app.cli_app import invoke_cli
+ invoke_cli()
+
+if __name__ == '__main__':
+ main()
diff --git a/invokeai/frontend/merge/__init__.py b/invokeai/frontend/merge/__init__.py
index fb892fd7db..f1fc66c39e 100644
--- a/invokeai/frontend/merge/__init__.py
+++ b/invokeai/frontend/merge/__init__.py
@@ -1,4 +1,5 @@
"""
Initialization file for invokeai.frontend.merge
"""
-from .merge_diffusers import main as invokeai_merge_diffusers, merge_diffusion_models
+from .merge_diffusers import main as invokeai_merge_diffusers
+
diff --git a/invokeai/frontend/merge/merge_diffusers.py b/invokeai/frontend/merge/merge_diffusers.py
index 9da04b97f8..c20d913883 100644
--- a/invokeai/frontend/merge/merge_diffusers.py
+++ b/invokeai/frontend/merge/merge_diffusers.py
@@ -6,9 +6,7 @@ Copyright (c) 2023 Lincoln Stein and the InvokeAI Development Team
"""
import argparse
import curses
-import os
import sys
-import warnings
from argparse import Namespace
from pathlib import Path
from typing import List, Union
@@ -20,99 +18,15 @@ from npyscreen import widget
from omegaconf import OmegaConf
import invokeai.backend.util.logging as logger
-from invokeai.services.config import InvokeAIAppConfig
-from ...backend.model_management import ModelManager
-from ...frontend.install.widgets import FloatTitleSlider
+from invokeai.app.services.config import InvokeAIAppConfig
+from invokeai.backend.model_management import (
+ ModelMerger, MergeInterpolationMethod,
+ ModelManager, ModelType, BaseModelType,
+)
+from invokeai.frontend.install.widgets import FloatTitleSlider, TextBox, SingleSelectColumns
-DEST_MERGED_MODEL_DIR = "merged_models"
config = InvokeAIAppConfig.get_config()
-def merge_diffusion_models(
- model_ids_or_paths: List[Union[str, Path]],
- alpha: float = 0.5,
- interp: str = None,
- force: bool = False,
- **kwargs,
-) -> DiffusionPipeline:
- """
- model_ids_or_paths - up to three models, designated by their local paths or HuggingFace repo_ids
- alpha - The interpolation parameter. Ranges from 0 to 1. It affects the ratio in which the checkpoints are merged. A 0.8 alpha
- would mean that the first model checkpoints would affect the final result far less than an alpha of 0.2
- interp - The interpolation method to use for the merging. Supports "sigmoid", "inv_sigmoid", "add_difference" and None.
- Passing None uses the default interpolation which is weighted sum interpolation. For merging three checkpoints, only "add_difference" is supported.
- force - Whether to ignore mismatch in model_config.json for the current models. Defaults to False.
-
- **kwargs - the default DiffusionPipeline.get_config_dict kwargs:
- cache_dir, resume_download, force_download, proxies, local_files_only, use_auth_token, revision, torch_dtype, device_map
- """
- with warnings.catch_warnings():
- warnings.simplefilter("ignore")
- verbosity = dlogging.get_verbosity()
- dlogging.set_verbosity_error()
-
- pipe = DiffusionPipeline.from_pretrained(
- model_ids_or_paths[0],
- cache_dir=kwargs.get("cache_dir", config.cache_dir),
- custom_pipeline="checkpoint_merger",
- )
- merged_pipe = pipe.merge(
- pretrained_model_name_or_path_list=model_ids_or_paths,
- alpha=alpha,
- interp=interp,
- force=force,
- **kwargs,
- )
- dlogging.set_verbosity(verbosity)
- return merged_pipe
-
-
-def merge_diffusion_models_and_commit(
- models: List["str"],
- merged_model_name: str,
- alpha: float = 0.5,
- interp: str = None,
- force: bool = False,
- **kwargs,
-):
- """
- models - up to three models, designated by their InvokeAI models.yaml model name
- merged_model_name = name for new model
- alpha - The interpolation parameter. Ranges from 0 to 1. It affects the ratio in which the checkpoints are merged. A 0.8 alpha
- would mean that the first model checkpoints would affect the final result far less than an alpha of 0.2
- interp - The interpolation method to use for the merging. Supports "weighted_average", "sigmoid", "inv_sigmoid", "add_difference" and None.
- Passing None uses the default interpolation which is weighted sum interpolation. For merging three checkpoints, only "add_difference" is supported. Add_difference is A+(B-C).
- force - Whether to ignore mismatch in model_config.json for the current models. Defaults to False.
-
- **kwargs - the default DiffusionPipeline.get_config_dict kwargs:
- cache_dir, resume_download, force_download, proxies, local_files_only, use_auth_token, revision, torch_dtype, device_map
- """
- config_file = config.model_conf_path
- model_manager = ModelManager(OmegaConf.load(config_file))
- for mod in models:
- assert mod in model_manager.model_names(), f'** Unknown model "{mod}"'
- assert (
- model_manager.model_info(mod).get("format", None) == "diffusers"
- ), f"** {mod} is not a diffusers model. It must be optimized before merging."
- model_ids_or_paths = [model_manager.model_name_or_path(x) for x in models]
-
- merged_pipe = merge_diffusion_models(
- model_ids_or_paths, alpha, interp, force, **kwargs
- )
- dump_path = config.models_dir / DEST_MERGED_MODEL_DIR
-
- os.makedirs(dump_path, exist_ok=True)
- dump_path = dump_path / merged_model_name
- merged_pipe.save_pretrained(dump_path, safe_serialization=1)
- import_args = dict(
- model_name=merged_model_name, description=f'Merge of models {", ".join(models)}'
- )
- if vae := model_manager.config[models[0]].get("vae", None):
- logger.info(f"Using configured VAE assigned to {models[0]}")
- import_args.update(vae=vae)
- model_manager.import_diffuser_model(dump_path, **import_args)
- model_manager.commit(config_file)
-
-
def _parse_args() -> Namespace:
parser = argparse.ArgumentParser(description="InvokeAI model merging")
parser.add_argument(
@@ -131,10 +45,17 @@ def _parse_args() -> Namespace:
)
parser.add_argument(
"--models",
+ dest="model_names",
type=str,
nargs="+",
help="Two to three model names to be merged",
)
+ parser.add_argument(
+ "--base_model",
+ type=str,
+ choices=[x.value for x in BaseModelType],
+ help="The base model shared by the models to be merged",
+ )
parser.add_argument(
"--merged_model_name",
"--destination",
@@ -192,6 +113,7 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
window_height, window_width = curses.initscr().getmaxyx()
self.model_names = self.get_model_names()
+ self.current_base = 0
max_width = max([len(x) for x in self.model_names])
max_width += 6
horizontal_layout = max_width * 3 < window_width
@@ -208,12 +130,26 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
value="Use up and down arrows to move, to select an item, and to move from one field to the next.",
editable=False,
)
+ self.nextrely += 1
+ self.base_select = self.add_widget_intelligent(
+ SingleSelectColumns,
+ values=[
+ 'Models Built on SD-1.x',
+ 'Models Built on SD-2.x',
+ ],
+ value=[self.current_base],
+ columns = 4,
+ max_height = 2,
+ relx=8,
+ scroll_exit = True,
+ )
+ self.base_select.on_changed = self._populate_models
self.add_widget_intelligent(
npyscreen.FixedText,
value="MODEL 1",
color="GOOD",
editable=False,
- rely=4 if horizontal_layout else None,
+ rely=6 if horizontal_layout else None,
)
self.model1 = self.add_widget_intelligent(
npyscreen.SelectOne,
@@ -222,7 +158,7 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
max_height=len(self.model_names),
max_width=max_width,
scroll_exit=True,
- rely=5,
+ rely=7,
)
self.add_widget_intelligent(
npyscreen.FixedText,
@@ -230,7 +166,7 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
color="GOOD",
editable=False,
relx=max_width + 3 if horizontal_layout else None,
- rely=4 if horizontal_layout else None,
+ rely=6 if horizontal_layout else None,
)
self.model2 = self.add_widget_intelligent(
npyscreen.SelectOne,
@@ -240,7 +176,7 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
max_height=len(self.model_names),
max_width=max_width,
relx=max_width + 3 if horizontal_layout else None,
- rely=5 if horizontal_layout else None,
+ rely=7 if horizontal_layout else None,
scroll_exit=True,
)
self.add_widget_intelligent(
@@ -249,7 +185,7 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
color="GOOD",
editable=False,
relx=max_width * 2 + 3 if horizontal_layout else None,
- rely=4 if horizontal_layout else None,
+ rely=6 if horizontal_layout else None,
)
models_plus_none = self.model_names.copy()
models_plus_none.insert(0, "None")
@@ -262,24 +198,26 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
max_width=max_width,
scroll_exit=True,
relx=max_width * 2 + 3 if horizontal_layout else None,
- rely=5 if horizontal_layout else None,
+ rely=7 if horizontal_layout else None,
)
for m in [self.model1, self.model2, self.model3]:
m.when_value_edited = self.models_changed
self.merged_model_name = self.add_widget_intelligent(
- npyscreen.TitleText,
+ TextBox,
name="Name for merged model:",
labelColor="CONTROL",
+ max_height=3,
value="",
scroll_exit=True,
)
self.force = self.add_widget_intelligent(
npyscreen.Checkbox,
- name="Force merge of incompatible models",
+ name="Force merge of models created by different diffusers library versions",
labelColor="CONTROL",
- value=False,
+ value=True,
scroll_exit=True,
)
+ self.nextrely += 1
self.merge_method = self.add_widget_intelligent(
npyscreen.TitleSelectOne,
name="Merge Method:",
@@ -341,7 +279,8 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
interp = self.interpolations[self.merge_method.value[0]]
args = dict(
- models=models,
+ model_names=models,
+ base_model=tuple(BaseModelType)[self.base_select.value[0]],
alpha=self.alpha.value,
interp=interp,
force=self.force.value,
@@ -379,21 +318,30 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
else:
return True
- def get_model_names(self) -> List[str]:
+ def get_model_names(self, base_model: BaseModelType=None) -> List[str]:
model_names = [
- name
- for name in self.model_manager.model_names()
- if self.model_manager.model_info(name).get("format") == "diffusers"
+ info["name"]
+ for info in self.model_manager.list_models(model_type=ModelType.Main, base_model=base_model)
+ if info["model_format"] == "diffusers"
]
return sorted(model_names)
+ def _populate_models(self,value=None):
+ base_model = tuple(BaseModelType)[value[0]]
+ self.model_names = self.get_model_names(base_model)
+
+ models_plus_none = self.model_names.copy()
+ models_plus_none.insert(0, "None")
+ self.model1.values = self.model_names
+ self.model2.values = self.model_names
+ self.model3.values = models_plus_none
+
+ self.display()
+
class Mergeapp(npyscreen.NPSAppManaged):
- def __init__(self):
+ def __init__(self, model_manager:ModelManager):
super().__init__()
- conf = OmegaConf.load(config.model_conf_path)
- self.model_manager = ModelManager(
- conf, "cpu", "float16"
- ) # precision doesn't really matter here
+ self.model_manager = model_manager
def onStart(self):
npyscreen.setTheme(npyscreen.Themes.ElegantTheme)
@@ -401,44 +349,41 @@ class Mergeapp(npyscreen.NPSAppManaged):
def run_gui(args: Namespace):
- mergeapp = Mergeapp()
+ model_manager = ModelManager(config.model_conf_path)
+ mergeapp = Mergeapp(model_manager)
mergeapp.run()
args = mergeapp.merge_arguments
- merge_diffusion_models_and_commit(**args)
+ merger = ModelMerger(model_manager)
+ merger.merge_diffusion_models_and_save(**args)
logger.info(f'Models merged into new model: "{args["merged_model_name"]}".')
def run_cli(args: Namespace):
assert args.alpha >= 0 and args.alpha <= 1.0, "alpha must be between 0 and 1"
assert (
- args.models and len(args.models) >= 1 and len(args.models) <= 3
+ args.model_names and len(args.model_names) >= 1 and len(args.model_names) <= 3
), "Please provide the --models argument to list 2 to 3 models to merge. Use --help for full usage."
if not args.merged_model_name:
- args.merged_model_name = "+".join(args.models)
+ args.merged_model_name = "+".join(args.model_names)
logger.info(
f'No --merged_model_name provided. Defaulting to "{args.merged_model_name}"'
)
- model_manager = ModelManager(OmegaConf.load(config.model_conf_path))
- assert (
- args.clobber or args.merged_model_name not in model_manager.model_names()
- ), f'A model named "{args.merged_model_name}" already exists. Use --clobber to overwrite.'
+ model_manager = ModelManager(config.model_conf_path)
+ assert (
+ not model_manager.model_exists(args.merged_model_name, args.base_model, ModelType.Main) or args.clobber
+ ), f'A model named "{args.merged_model_name}" already exists. Use --clobber to overwrite.'
- merge_diffusion_models_and_commit(**vars(args))
- logger.info(f'Models merged into new model: "{args.merged_model_name}".')
+ merger = ModelMerger(model_manager)
+ merger.merge_diffusion_models_and_save(**vars(args))
+ logger.info(f'Models merged into new model: "{args.merged_model_name}".')
def main():
args = _parse_args()
- config.root = args.root_dir
-
- cache_dir = config.cache_dir
- os.environ[
- "HF_HOME"
- ] = cache_dir # because not clear the merge pipeline is honoring cache_dir
- args.cache_dir = cache_dir
+ config.parse_args(['--root',str(args.root_dir)])
try:
if args.front_end:
diff --git a/invokeai/frontend/web/.eslintrc.js b/invokeai/frontend/web/.eslintrc.js
index b1a2b6a7e4..34db9d466b 100644
--- a/invokeai/frontend/web/.eslintrc.js
+++ b/invokeai/frontend/web/.eslintrc.js
@@ -36,6 +36,12 @@ module.exports = {
],
'prettier/prettier': ['error', { endOfLine: 'auto' }],
'@typescript-eslint/ban-ts-comment': 'warn',
+ '@typescript-eslint/no-empty-interface': [
+ 'error',
+ {
+ allowSingleExtends: true,
+ },
+ ],
},
settings: {
react: {
diff --git a/invokeai/frontend/web/config/common.ts b/invokeai/frontend/web/config/common.ts
new file mode 100644
index 0000000000..4470224225
--- /dev/null
+++ b/invokeai/frontend/web/config/common.ts
@@ -0,0 +1,12 @@
+import react from '@vitejs/plugin-react-swc';
+import { visualizer } from 'rollup-plugin-visualizer';
+import { PluginOption, UserConfig } from 'vite';
+import eslint from 'vite-plugin-eslint';
+import tsconfigPaths from 'vite-tsconfig-paths';
+
+export const commonPlugins: UserConfig['plugins'] = [
+ react(),
+ eslint(),
+ tsconfigPaths(),
+ visualizer() as unknown as PluginOption,
+];
diff --git a/invokeai/frontend/web/config/vite.app.config.ts b/invokeai/frontend/web/config/vite.app.config.ts
index e6c4df79fd..958313402a 100644
--- a/invokeai/frontend/web/config/vite.app.config.ts
+++ b/invokeai/frontend/web/config/vite.app.config.ts
@@ -1,17 +1,9 @@
-import react from '@vitejs/plugin-react-swc';
-import { visualizer } from 'rollup-plugin-visualizer';
-import { PluginOption, UserConfig } from 'vite';
-import eslint from 'vite-plugin-eslint';
-import tsconfigPaths from 'vite-tsconfig-paths';
+import { UserConfig } from 'vite';
+import { commonPlugins } from './common';
export const appConfig: UserConfig = {
base: './',
- plugins: [
- react(),
- eslint(),
- tsconfigPaths(),
- visualizer() as unknown as PluginOption,
- ],
+ plugins: [...commonPlugins],
build: {
chunkSizeWarningLimit: 1500,
},
diff --git a/invokeai/frontend/web/config/vite.package.config.ts b/invokeai/frontend/web/config/vite.package.config.ts
index f87cce0bc9..0dcccab086 100644
--- a/invokeai/frontend/web/config/vite.package.config.ts
+++ b/invokeai/frontend/web/config/vite.package.config.ts
@@ -1,19 +1,13 @@
-import react from '@vitejs/plugin-react-swc';
import path from 'path';
-import { visualizer } from 'rollup-plugin-visualizer';
-import { PluginOption, UserConfig } from 'vite';
+import { UserConfig } from 'vite';
import dts from 'vite-plugin-dts';
-import eslint from 'vite-plugin-eslint';
-import tsconfigPaths from 'vite-tsconfig-paths';
import cssInjectedByJsPlugin from 'vite-plugin-css-injected-by-js';
+import { commonPlugins } from './common';
export const packageConfig: UserConfig = {
base: './',
plugins: [
- react(),
- eslint(),
- tsconfigPaths(),
- visualizer() as unknown as PluginOption,
+ ...commonPlugins,
dts({
insertTypesEntry: true,
}),
diff --git a/invokeai/frontend/web/dist/assets/App-4c33c38e.css b/invokeai/frontend/web/dist/assets/App-4c33c38e.css
new file mode 100644
index 0000000000..482eef3217
--- /dev/null
+++ b/invokeai/frontend/web/dist/assets/App-4c33c38e.css
@@ -0,0 +1 @@
+.ltr-image-gallery-css-transition-enter{transform:translate(150%)}.ltr-image-gallery-css-transition-enter-active{transform:translate(0);transition:all .12s ease-out}.ltr-image-gallery-css-transition-exit{transform:translate(0)}.ltr-image-gallery-css-transition-exit-active{transform:translate(150%);transition:all .12s ease-out}.rtl-image-gallery-css-transition-enter{transform:translate(-150%)}.rtl-image-gallery-css-transition-enter-active{transform:translate(0);transition:all .12s ease-out}.rtl-image-gallery-css-transition-exit{transform:translate(0)}.rtl-image-gallery-css-transition-exit-active{transform:translate(-150%);transition:all .12s ease-out}.react-flow__container{position:absolute;width:100%;height:100%;top:0;left:0}.react-flow__pane{z-index:1;cursor:-webkit-grab;cursor:grab}.react-flow__pane.selection{cursor:pointer}.react-flow__pane.dragging{cursor:-webkit-grabbing;cursor:grabbing}.react-flow__viewport{transform-origin:0 0;z-index:2;pointer-events:none}.react-flow__renderer{z-index:4}.react-flow__selection{z-index:6}.react-flow__nodesselection-rect:focus,.react-flow__nodesselection-rect:focus-visible{outline:none}.react-flow .react-flow__edges{pointer-events:none;overflow:visible}.react-flow__edge-path,.react-flow__connection-path{stroke:#b1b1b7;stroke-width:1;fill:none}.react-flow__edge{pointer-events:visibleStroke;cursor:pointer}.react-flow__edge.animated path{stroke-dasharray:5;-webkit-animation:dashdraw .5s linear infinite;animation:dashdraw .5s linear infinite}.react-flow__edge.animated path.react-flow__edge-interaction{stroke-dasharray:none;-webkit-animation:none;animation:none}.react-flow__edge.inactive{pointer-events:none}.react-flow__edge.selected,.react-flow__edge:focus,.react-flow__edge:focus-visible{outline:none}.react-flow__edge.selected .react-flow__edge-path,.react-flow__edge:focus .react-flow__edge-path,.react-flow__edge:focus-visible .react-flow__edge-path{stroke:#555}.react-flow__edge-textwrapper{pointer-events:all}.react-flow__edge-textbg{fill:#fff}.react-flow__edge .react-flow__edge-text{pointer-events:none;-webkit-user-select:none;-moz-user-select:none;user-select:none}.react-flow__connection{pointer-events:none}.react-flow__connection .animated{stroke-dasharray:5;-webkit-animation:dashdraw .5s linear infinite;animation:dashdraw .5s linear infinite}.react-flow__connectionline{z-index:1001}.react-flow__nodes{pointer-events:none;transform-origin:0 0}.react-flow__node{position:absolute;-webkit-user-select:none;-moz-user-select:none;user-select:none;pointer-events:all;transform-origin:0 0;box-sizing:border-box;cursor:-webkit-grab;cursor:grab}.react-flow__node.dragging{cursor:-webkit-grabbing;cursor:grabbing}.react-flow__nodesselection{z-index:3;transform-origin:left top;pointer-events:none}.react-flow__nodesselection-rect{position:absolute;pointer-events:all;cursor:-webkit-grab;cursor:grab}.react-flow__handle{position:absolute;pointer-events:none;min-width:5px;min-height:5px;width:6px;height:6px;background:#1a192b;border:1px solid white;border-radius:100%}.react-flow__handle.connectionindicator{pointer-events:all;cursor:crosshair}.react-flow__handle-bottom{top:auto;left:50%;bottom:-4px;transform:translate(-50%)}.react-flow__handle-top{left:50%;top:-4px;transform:translate(-50%)}.react-flow__handle-left{top:50%;left:-4px;transform:translateY(-50%)}.react-flow__handle-right{right:-4px;top:50%;transform:translateY(-50%)}.react-flow__edgeupdater{cursor:move;pointer-events:all}.react-flow__panel{position:absolute;z-index:5;margin:15px}.react-flow__panel.top{top:0}.react-flow__panel.bottom{bottom:0}.react-flow__panel.left{left:0}.react-flow__panel.right{right:0}.react-flow__panel.center{left:50%;transform:translate(-50%)}.react-flow__attribution{font-size:10px;background:rgba(255,255,255,.5);padding:2px 3px;margin:0}.react-flow__attribution a{text-decoration:none;color:#999}@-webkit-keyframes dashdraw{0%{stroke-dashoffset:10}}@keyframes dashdraw{0%{stroke-dashoffset:10}}.react-flow__edgelabel-renderer{position:absolute;width:100%;height:100%;pointer-events:none;-webkit-user-select:none;-moz-user-select:none;user-select:none}.react-flow__edge.updating .react-flow__edge-path{stroke:#777}.react-flow__edge-text{font-size:10px}.react-flow__node.selectable:focus,.react-flow__node.selectable:focus-visible{outline:none}.react-flow__node-default,.react-flow__node-input,.react-flow__node-output,.react-flow__node-group{padding:10px;border-radius:3px;width:150px;font-size:12px;color:#222;text-align:center;border-width:1px;border-style:solid;border-color:#1a192b;background-color:#fff}.react-flow__node-default.selectable:hover,.react-flow__node-input.selectable:hover,.react-flow__node-output.selectable:hover,.react-flow__node-group.selectable:hover{box-shadow:0 1px 4px 1px #00000014}.react-flow__node-default.selectable.selected,.react-flow__node-default.selectable:focus,.react-flow__node-default.selectable:focus-visible,.react-flow__node-input.selectable.selected,.react-flow__node-input.selectable:focus,.react-flow__node-input.selectable:focus-visible,.react-flow__node-output.selectable.selected,.react-flow__node-output.selectable:focus,.react-flow__node-output.selectable:focus-visible,.react-flow__node-group.selectable.selected,.react-flow__node-group.selectable:focus,.react-flow__node-group.selectable:focus-visible{box-shadow:0 0 0 .5px #1a192b}.react-flow__node-group{background-color:#f0f0f040}.react-flow__nodesselection-rect,.react-flow__selection{background:rgba(0,89,220,.08);border:1px dotted rgba(0,89,220,.8)}.react-flow__nodesselection-rect:focus,.react-flow__nodesselection-rect:focus-visible,.react-flow__selection:focus,.react-flow__selection:focus-visible{outline:none}.react-flow__controls{box-shadow:0 0 2px 1px #00000014}.react-flow__controls-button{border:none;background:#fefefe;border-bottom:1px solid #eee;box-sizing:content-box;display:flex;justify-content:center;align-items:center;width:16px;height:16px;cursor:pointer;-webkit-user-select:none;-moz-user-select:none;user-select:none;padding:5px}.react-flow__controls-button:hover{background:#f4f4f4}.react-flow__controls-button svg{width:100%;max-width:12px;max-height:12px}.react-flow__controls-button:disabled{pointer-events:none}.react-flow__controls-button:disabled svg{fill-opacity:.4}.react-flow__minimap{background-color:#fff}.react-flow__resize-control{position:absolute}.react-flow__resize-control.left,.react-flow__resize-control.right{cursor:ew-resize}.react-flow__resize-control.top,.react-flow__resize-control.bottom{cursor:ns-resize}.react-flow__resize-control.top.left,.react-flow__resize-control.bottom.right{cursor:nwse-resize}.react-flow__resize-control.bottom.left,.react-flow__resize-control.top.right{cursor:nesw-resize}.react-flow__resize-control.handle{width:4px;height:4px;border:1px solid #fff;border-radius:1px;background-color:#3367d9;transform:translate(-50%,-50%)}.react-flow__resize-control.handle.left{left:0;top:50%}.react-flow__resize-control.handle.right{left:100%;top:50%}.react-flow__resize-control.handle.top{left:50%;top:0}.react-flow__resize-control.handle.bottom{left:50%;top:100%}.react-flow__resize-control.handle.top.left,.react-flow__resize-control.handle.bottom.left{left:0}.react-flow__resize-control.handle.top.right,.react-flow__resize-control.handle.bottom.right{left:100%}.react-flow__resize-control.line{border-color:#3367d9;border-width:0;border-style:solid}.react-flow__resize-control.line.left,.react-flow__resize-control.line.right{width:1px;transform:translate(-50%);top:0;height:100%}.react-flow__resize-control.line.left{left:0;border-left-width:1px}.react-flow__resize-control.line.right{left:100%;border-right-width:1px}.react-flow__resize-control.line.top,.react-flow__resize-control.line.bottom{height:1px;transform:translateY(-50%);left:0;width:100%}.react-flow__resize-control.line.top{top:0;border-top-width:1px}.react-flow__resize-control.line.bottom{border-bottom-width:1px;top:100%}
diff --git a/invokeai/frontend/web/dist/assets/App-a44d46fe.js b/invokeai/frontend/web/dist/assets/App-a44d46fe.js
new file mode 100644
index 0000000000..eedd37b459
--- /dev/null
+++ b/invokeai/frontend/web/dist/assets/App-a44d46fe.js
@@ -0,0 +1,199 @@
+import{t as mv,b as VI,r as UI,i as GI,a as qI,c as KI,d as Z4,e as J4,f as e3,g as YI,h as XI,j as QI,k as ZI,l as JI,m as eO,n as tO,o as nO,s as rO,p as oO,q as p,u as nt,v as i,I as mo,w as ur,x as Un,y as Be,z as Ot,A as Ru,B as bt,C as Mu,D as Ao,E as bo,F as Xr,G as xr,H as sO,J as R0,K as aO,L as qa,M as su,N as Zi,O as on,P as xa,Q as oh,R as M0,S as Al,T as T0,U as dc,_ as nr,V as iO,W as z,X as t3,Y as n3,Z as vx,$ as r3,a0 as sh,a1 as D0,a2 as fc,a3 as lO,a4 as cO,a5 as uO,a6 as dO,a7 as o3,a8 as s3,a9 as ah,aa as zf,ab as fO,ac as pO,ad as ag,ae as yx,af as a3,ag as i3,ah as Ji,ai as bx,aj as xx,ak as hO,al as mO,am as gO,an as vO,ao as yO,ap as ig,aq as ml,ar as lg,as as bO,at as xO,au as SO,av as A0,aw as ih,ax as $f,ay as wO,az as l3,aA as c3,aB as CO,aC as Tu,aD as N0,aE as PO,aF as kO,aG as _O,aH as jO,aI as EO,aJ as IO,aK as OO,aL as Sx,aM as qe,aN as H,aO as aa,aP as Pe,aQ as ao,aR as Sr,aS as re,aT as B,aU as Du,aV as me,aW as u3,aX as xo,aY as st,aZ as d3,a_ as ai,a$ as z0,b0 as RO,b1 as MO,b2 as TO,b3 as gl,b4 as Ti,b5 as Vs,b6 as f3,b7 as p3,b8 as DO,b9 as AO,ba as Au,bb as Nu,bc as Gl,bd as Gn,be as h3,bf as NO,bg as zO,bh as mn,bi as Et,bj as lh,bk as Ka,bl as wx,bm as ch,bn as $O,bo as LO,bp as Ef,bq as Ie,br as zd,bs as $d,bt as au,bu as iu,bv as Cx,bw as Lf,bx as Px,by as Ff,bz as kx,bA as gv,bB as _x,bC as vv,bD as jx,bE as Bf,bF as Ex,bG as ql,bH as Ix,bI as Kl,bJ as Ox,bK as Hf,bL as FO,bM as Wf,bN as m3,bO as $0,bP as BO,bQ as g3,bR as HO,bS as WO,bT as v3,bU as y3,bV as uh,bW as Oi,bX as yv,bY as VO,bZ as b3,b_ as UO,b$ as x3,c0 as Ja,c1 as Rx,c2 as GO,c3 as Mx,c4 as qO,c5 as vt,c6 as Vf,c7 as KO,c8 as S3,c9 as pc,ca as Wt,cb as So,cc as rr,cd as bv,ce as Us,cf as ir,cg as mr,ch as w3,ci as YO,cj as C3,ck as Tx,cl as XO,cm as QO,cn as ZO,co as JO,cp as e8,cq as t8,cr as L0,cs as os,ct as n8,cu as r8,cv as Dx,cw as Ax,cx as Nx,cy as zx,cz as o8,cA as s8,cB as P3,cC as mu,cD as k3,cE as $x,cF as a8,cG as i8,cH as l8,cI as c8,cJ as u8,cK as d8,cL as _3,cM as f8,cN as Pr,cO as j3,cP as p8,cQ as E3,cR as I3,cS as h8,cT as Lx,cU as m8,cV as g8,cW as v8,cX as y8,cY as b8,cZ as x8,c_ as S8,c$ as w8,d0 as C8,d1 as P8,d2 as k8,d3 as _8,d4 as j8,d5 as E8,d6 as I8,d7 as O8,d8 as R8,d9 as M8,da as T8,db as D8,dc as A8,dd as N8,de as zu,df as z8,dg as Lc,dh as ii,di as $8,dj as L8,dk as F8,dl as B8,dm as H8,dn as W8,dp as O3,dq as V8,dr as U8,ds as ss,dt as G8,du as Fx,dv as Bx,dw as Ya,dx as q8,dy as K8,dz as Y8,dA as R3,dB as X8,dC as Q8,dD as Z8,dE as F0,dF as B0,dG as J8,dH as Hx,dI as eR,dJ as Wx,dK as No,dL as tR,dM as nR,dN as Vx,dO as Ux,dP as rR,dQ as Gx,dR as oR,dS as sR,dT as aR,dU as iR,dV as lR,dW as cR,dX as M3,dY as T3,dZ as D3,d_ as A3,d$ as N3,e0 as $u,e1 as qx,e2 as cg,e3 as uR,e4 as dR,e5 as Yl,e6 as fR,e7 as pR,e8 as z3,e9 as $3,ea as hR,eb as Kx,ec as mR,ed as gR,ee as vR,ef as Mo,eg as Uf,eh as yR,ei as Yx,ej as L3,ek as bR,el as Xx,em as F3,en as H0,eo as Ps,ep as xR,eq as B3,er as SR,es as wR,et as Qx,eu as CR,ev as PR,ew as kR,ex as _R,ey as jR,ez as ER,eA as IR,eB as OR,eC as RR,eD as MR,eE as Zx,eF as TR,eG as Jx,eH as If,eI as DR,eJ as en,eK as Gf,eL as H3,eM as qf,eN as AR,eO as NR,eP as Ws,eQ as W3,eR as W0,eS as Lu,eT as zR,eU as $R,eV as LR,eW as ha,eX as V3,eY as FR,eZ as BR,e_ as U3,e$ as HR,f0 as WR,f1 as VR,f2 as UR,f3 as GR,f4 as qR,f5 as KR,f6 as YR,f7 as XR,f8 as QR,f9 as ZR,fa as e2,fb as JR,fc as e7,fd as t7,fe as n7,ff as r7,fg as o7,fh as ug,fi as gu,fj as Ld,fk as dg,fl as fg,fm as Fd,fn as t2,fo as xv,fp as s7,fq as a7,fr as i7,fs as l7,ft as c7,fu as Kf,fv as G3,fw as q3,fx as u7,fy as d7,fz as K3,fA as Y3,fB as X3,fC as Q3,fD as Z3,fE as J3,fF as e5,fG as t5,fH as Nl,fI as zl,fJ as n5,fK as r5,fL as f7,fM as o5,fN as s5,fO as a5,fP as i5,fQ as l5,fR as c5,fS as u5,fT as V0,fU as p7,fV as n2,fW as h7,fX as m7,fY as Yf,fZ as r2,f_ as o2,f$ as s2,g0 as a2,g1 as g7,g2 as v7,g3 as y7,g4 as b7,g5 as x7,g6 as S7,g7 as w7,g8 as C7,g9 as P7}from"./index-078526aa.js";import{u as d5,a as Sa,b as k7,r as Le,f as _7,g as i2,c as At,d as jr}from"./MantineProvider-8988d217.js";function j7(e,t){if(e==null)return{};var n={},r=Object.keys(e),o,s;for(s=0;s=0)&&(n[o]=e[o]);return n}var l2=1/0,E7=17976931348623157e292;function pg(e){if(!e)return e===0?e:0;if(e=mv(e),e===l2||e===-l2){var t=e<0?-1:1;return t*E7}return e===e?e:0}var I7=4;function c2(e){return VI(e,I7)}var O7=function(){return UI.Date.now()};const hg=O7;var R7="Expected a function",M7=Math.max,T7=Math.min;function D7(e,t,n){var r,o,s,a,u,d,f=0,h=!1,m=!1,v=!0;if(typeof e!="function")throw new TypeError(R7);t=mv(t)||0,GI(n)&&(h=!!n.leading,m="maxWait"in n,s=m?M7(mv(n.maxWait)||0,t):s,v="trailing"in n?!!n.trailing:v);function b(O){var R=r,M=o;return r=o=void 0,f=O,a=e.apply(M,R),a}function w(O){return f=O,u=setTimeout(P,t),h?b(O):a}function C(O){var R=O-d,M=O-f,A=t-R;return m?T7(A,s-M):A}function S(O){var R=O-d,M=O-f;return d===void 0||R>=t||R<0||m&&M>=s}function P(){var O=hg();if(S(O))return k(O);u=setTimeout(P,C(O))}function k(O){return u=void 0,v&&r?b(O):(r=o=void 0,a)}function j(){u!==void 0&&clearTimeout(u),f=0,r=d=o=u=void 0}function E(){return u===void 0?a:k(hg())}function I(){var O=hg(),R=S(O);if(r=arguments,o=this,d=O,R){if(u===void 0)return w(d);if(m)return clearTimeout(u),u=setTimeout(P,t),b(d)}return u===void 0&&(u=setTimeout(P,t)),a}return I.cancel=j,I.flush=E,I}function A7(e,t){var n=-1,r=qI(e)?Array(e.length):[];return KI(e,function(o,s,a){r[++n]=t(o,s,a)}),r}function Ks(e,t){var n=Z4(e)?J4:A7;return n(e,e3(t))}var N7="[object Number]";function vl(e){return typeof e=="number"||YI(e)&&XI(e)==N7}var z7=Array.prototype,$7=z7.splice;function L7(e,t){for(var n=e?t.length:0,r=n-1;n--;){var o=t[n];if(n==r||o!==s){var s=o;QI(o)?$7.call(e,o,1):ZI(e,o)}}return e}var F7=Math.ceil,B7=Math.max;function H7(e,t,n,r){for(var o=-1,s=B7(F7((t-e)/(n||1)),0),a=Array(s);s--;)a[r?s:++o]=e,e+=n;return a}function W7(e){return function(t,n,r){return r&&typeof r!="number"&&JI(t,n,r)&&(n=r=void 0),t=pg(t),n===void 0?(n=t,t=0):n=pg(n),r=r===void 0?ti.jsx(mo,{ref:d,viewBox:t,...o,...u,children:s.length?s:i.jsx("path",{fill:"currentColor",d:n})}));return a.displayName=r,a}var G7=Object.defineProperty,q7=(e,t,n)=>t in e?G7(e,t,{enumerable:!0,configurable:!0,writable:!0,value:n}):e[t]=n,pr=(e,t,n)=>(q7(e,typeof t!="symbol"?t+"":t,n),n);function d2(e){return e.sort((t,n)=>{const r=t.compareDocumentPosition(n);if(r&Node.DOCUMENT_POSITION_FOLLOWING||r&Node.DOCUMENT_POSITION_CONTAINED_BY)return-1;if(r&Node.DOCUMENT_POSITION_PRECEDING||r&Node.DOCUMENT_POSITION_CONTAINS)return 1;if(r&Node.DOCUMENT_POSITION_DISCONNECTED||r&Node.DOCUMENT_POSITION_IMPLEMENTATION_SPECIFIC)throw Error("Cannot sort the given nodes.");return 0})}var K7=e=>typeof e=="object"&&"nodeType"in e&&e.nodeType===Node.ELEMENT_NODE;function f2(e,t,n){let r=e+1;return n&&r>=t&&(r=0),r}function p2(e,t,n){let r=e-1;return n&&r<0&&(r=t),r}var Sv=typeof window<"u"?p.useLayoutEffect:p.useEffect,Xf=e=>e,Y7=class{constructor(){pr(this,"descendants",new Map),pr(this,"register",e=>{if(e!=null)return K7(e)?this.registerNode(e):t=>{this.registerNode(t,e)}}),pr(this,"unregister",e=>{this.descendants.delete(e);const t=d2(Array.from(this.descendants.keys()));this.assignIndex(t)}),pr(this,"destroy",()=>{this.descendants.clear()}),pr(this,"assignIndex",e=>{this.descendants.forEach(t=>{const n=e.indexOf(t.node);t.index=n,t.node.dataset.index=t.index.toString()})}),pr(this,"count",()=>this.descendants.size),pr(this,"enabledCount",()=>this.enabledValues().length),pr(this,"values",()=>Array.from(this.descendants.values()).sort((t,n)=>t.index-n.index)),pr(this,"enabledValues",()=>this.values().filter(e=>!e.disabled)),pr(this,"item",e=>{if(this.count()!==0)return this.values()[e]}),pr(this,"enabledItem",e=>{if(this.enabledCount()!==0)return this.enabledValues()[e]}),pr(this,"first",()=>this.item(0)),pr(this,"firstEnabled",()=>this.enabledItem(0)),pr(this,"last",()=>this.item(this.descendants.size-1)),pr(this,"lastEnabled",()=>{const e=this.enabledValues().length-1;return this.enabledItem(e)}),pr(this,"indexOf",e=>{var t,n;return e&&(n=(t=this.descendants.get(e))==null?void 0:t.index)!=null?n:-1}),pr(this,"enabledIndexOf",e=>e==null?-1:this.enabledValues().findIndex(t=>t.node.isSameNode(e))),pr(this,"next",(e,t=!0)=>{const n=f2(e,this.count(),t);return this.item(n)}),pr(this,"nextEnabled",(e,t=!0)=>{const n=this.item(e);if(!n)return;const r=this.enabledIndexOf(n.node),o=f2(r,this.enabledCount(),t);return this.enabledItem(o)}),pr(this,"prev",(e,t=!0)=>{const n=p2(e,this.count()-1,t);return this.item(n)}),pr(this,"prevEnabled",(e,t=!0)=>{const n=this.item(e);if(!n)return;const r=this.enabledIndexOf(n.node),o=p2(r,this.enabledCount()-1,t);return this.enabledItem(o)}),pr(this,"registerNode",(e,t)=>{if(!e||this.descendants.has(e))return;const n=Array.from(this.descendants.keys()).concat(e),r=d2(n);t!=null&&t.disabled&&(t.disabled=!!t.disabled);const o={node:e,index:-1,...t};this.descendants.set(e,o),this.assignIndex(r)})}};function X7(){const e=p.useRef(new Y7);return Sv(()=>()=>e.current.destroy()),e.current}var[Q7,p5]=ur({name:"DescendantsProvider",errorMessage:"useDescendantsContext must be used within DescendantsProvider"});function Z7(e){const t=p5(),[n,r]=p.useState(-1),o=p.useRef(null);Sv(()=>()=>{o.current&&t.unregister(o.current)},[]),Sv(()=>{if(!o.current)return;const a=Number(o.current.dataset.index);n!=a&&!Number.isNaN(a)&&r(a)});const s=Xf(e?t.register(e):t.register);return{descendants:t,index:n,enabledIndex:t.enabledIndexOf(o.current),register:Un(s,o)}}function G0(){return[Xf(Q7),()=>Xf(p5()),()=>X7(),o=>Z7(o)]}var[J7,dh]=ur({name:"AccordionStylesContext",hookName:"useAccordionStyles",providerName:""}),[eM,q0]=ur({name:"AccordionItemContext",hookName:"useAccordionItemContext",providerName:""}),[tM,zae,nM,rM]=G0(),Zc=nt(function(t,n){const{getButtonProps:r}=q0(),o=r(t,n),a={display:"flex",alignItems:"center",width:"100%",outline:0,...dh().button};return i.jsx(Be.button,{...o,className:Ot("chakra-accordion__button",t.className),__css:a})});Zc.displayName="AccordionButton";function oM(e){const{onChange:t,defaultIndex:n,index:r,allowMultiple:o,allowToggle:s,...a}=e;iM(e),lM(e);const u=nM(),[d,f]=p.useState(-1);p.useEffect(()=>()=>{f(-1)},[]);const[h,m]=Ru({value:r,defaultValue(){return o?n??[]:n??-1},onChange:t});return{index:h,setIndex:m,htmlProps:a,getAccordionItemProps:b=>{let w=!1;return b!==null&&(w=Array.isArray(h)?h.includes(b):h===b),{isOpen:w,onChange:S=>{if(b!==null)if(o&&Array.isArray(h)){const P=S?h.concat(b):h.filter(k=>k!==b);m(P)}else S?m(b):s&&m(-1)}}},focusedIndex:d,setFocusedIndex:f,descendants:u}}var[sM,K0]=ur({name:"AccordionContext",hookName:"useAccordionContext",providerName:"Accordion"});function aM(e){const{isDisabled:t,isFocusable:n,id:r,...o}=e,{getAccordionItemProps:s,setFocusedIndex:a}=K0(),u=p.useRef(null),d=p.useId(),f=r??d,h=`accordion-button-${f}`,m=`accordion-panel-${f}`;cM(e);const{register:v,index:b,descendants:w}=rM({disabled:t&&!n}),{isOpen:C,onChange:S}=s(b===-1?null:b);uM({isOpen:C,isDisabled:t});const P=()=>{S==null||S(!0)},k=()=>{S==null||S(!1)},j=p.useCallback(()=>{S==null||S(!C),a(b)},[b,a,C,S]),E=p.useCallback(M=>{const L={ArrowDown:()=>{const D=w.nextEnabled(b);D==null||D.node.focus()},ArrowUp:()=>{const D=w.prevEnabled(b);D==null||D.node.focus()},Home:()=>{const D=w.firstEnabled();D==null||D.node.focus()},End:()=>{const D=w.lastEnabled();D==null||D.node.focus()}}[M.key];L&&(M.preventDefault(),L(M))},[w,b]),I=p.useCallback(()=>{a(b)},[a,b]),O=p.useCallback(function(A={},L=null){return{...A,type:"button",ref:Un(v,u,L),id:h,disabled:!!t,"aria-expanded":!!C,"aria-controls":m,onClick:bt(A.onClick,j),onFocus:bt(A.onFocus,I),onKeyDown:bt(A.onKeyDown,E)}},[h,t,C,j,I,E,m,v]),R=p.useCallback(function(A={},L=null){return{...A,ref:L,role:"region",id:m,"aria-labelledby":h,hidden:!C}},[h,C,m]);return{isOpen:C,isDisabled:t,isFocusable:n,onOpen:P,onClose:k,getButtonProps:O,getPanelProps:R,htmlProps:o}}function iM(e){const t=e.index||e.defaultIndex,n=t!=null&&!Array.isArray(t)&&e.allowMultiple;Mu({condition:!!n,message:`If 'allowMultiple' is passed, then 'index' or 'defaultIndex' must be an array. You passed: ${typeof t},`})}function lM(e){Mu({condition:!!(e.allowMultiple&&e.allowToggle),message:"If 'allowMultiple' is passed, 'allowToggle' will be ignored. Either remove 'allowToggle' or 'allowMultiple' depending on whether you want multiple accordions visible or not"})}function cM(e){Mu({condition:!!(e.isFocusable&&!e.isDisabled),message:`Using only 'isFocusable', this prop is reserved for situations where you pass 'isDisabled' but you still want the element to receive focus (A11y). Either remove it or pass 'isDisabled' as well.
+ `})}function uM(e){Mu({condition:e.isOpen&&!!e.isDisabled,message:"Cannot open a disabled accordion item"})}function Jc(e){const{isOpen:t,isDisabled:n}=q0(),{reduceMotion:r}=K0(),o=Ot("chakra-accordion__icon",e.className),s=dh(),a={opacity:n?.4:1,transform:t?"rotate(-180deg)":void 0,transition:r?void 0:"transform 0.2s",transformOrigin:"center",...s.icon};return i.jsx(mo,{viewBox:"0 0 24 24","aria-hidden":!0,className:o,__css:a,...e,children:i.jsx("path",{fill:"currentColor",d:"M16.59 8.59L12 13.17 7.41 8.59 6 10l6 6 6-6z"})})}Jc.displayName="AccordionIcon";var eu=nt(function(t,n){const{children:r,className:o}=t,{htmlProps:s,...a}=aM(t),d={...dh().container,overflowAnchor:"none"},f=p.useMemo(()=>a,[a]);return i.jsx(eM,{value:f,children:i.jsx(Be.div,{ref:n,...s,className:Ot("chakra-accordion__item",o),__css:d,children:typeof r=="function"?r({isExpanded:!!a.isOpen,isDisabled:!!a.isDisabled}):r})})});eu.displayName="AccordionItem";var Ri={ease:[.25,.1,.25,1],easeIn:[.4,0,1,1],easeOut:[0,0,.2,1],easeInOut:[.4,0,.2,1]},Fc={scale:{enter:{scale:1},exit:{scale:.95}},fade:{enter:{opacity:1},exit:{opacity:0}},pushLeft:{enter:{x:"100%"},exit:{x:"-30%"}},pushRight:{enter:{x:"-100%"},exit:{x:"30%"}},pushUp:{enter:{y:"100%"},exit:{y:"-30%"}},pushDown:{enter:{y:"-100%"},exit:{y:"30%"}},slideLeft:{position:{left:0,top:0,bottom:0,width:"100%"},enter:{x:0,y:0},exit:{x:"-100%",y:0}},slideRight:{position:{right:0,top:0,bottom:0,width:"100%"},enter:{x:0,y:0},exit:{x:"100%",y:0}},slideUp:{position:{top:0,left:0,right:0,maxWidth:"100vw"},enter:{x:0,y:0},exit:{x:0,y:"-100%"}},slideDown:{position:{bottom:0,left:0,right:0,maxWidth:"100vw"},enter:{x:0,y:0},exit:{x:0,y:"100%"}}};function wv(e){var t;switch((t=e==null?void 0:e.direction)!=null?t:"right"){case"right":return Fc.slideRight;case"left":return Fc.slideLeft;case"bottom":return Fc.slideDown;case"top":return Fc.slideUp;default:return Fc.slideRight}}var Di={enter:{duration:.2,ease:Ri.easeOut},exit:{duration:.1,ease:Ri.easeIn}},vs={enter:(e,t)=>({...e,delay:typeof t=="number"?t:t==null?void 0:t.enter}),exit:(e,t)=>({...e,delay:typeof t=="number"?t:t==null?void 0:t.exit})},dM=e=>e!=null&&parseInt(e.toString(),10)>0,h2={exit:{height:{duration:.2,ease:Ri.ease},opacity:{duration:.3,ease:Ri.ease}},enter:{height:{duration:.3,ease:Ri.ease},opacity:{duration:.4,ease:Ri.ease}}},fM={exit:({animateOpacity:e,startingHeight:t,transition:n,transitionEnd:r,delay:o})=>{var s;return{...e&&{opacity:dM(t)?1:0},height:t,transitionEnd:r==null?void 0:r.exit,transition:(s=n==null?void 0:n.exit)!=null?s:vs.exit(h2.exit,o)}},enter:({animateOpacity:e,endingHeight:t,transition:n,transitionEnd:r,delay:o})=>{var s;return{...e&&{opacity:1},height:t,transitionEnd:r==null?void 0:r.enter,transition:(s=n==null?void 0:n.enter)!=null?s:vs.enter(h2.enter,o)}}},fh=p.forwardRef((e,t)=>{const{in:n,unmountOnExit:r,animateOpacity:o=!0,startingHeight:s=0,endingHeight:a="auto",style:u,className:d,transition:f,transitionEnd:h,...m}=e,[v,b]=p.useState(!1);p.useEffect(()=>{const k=setTimeout(()=>{b(!0)});return()=>clearTimeout(k)},[]),Mu({condition:Number(s)>0&&!!r,message:"startingHeight and unmountOnExit are mutually exclusive. You can't use them together"});const w=parseFloat(s.toString())>0,C={startingHeight:s,endingHeight:a,animateOpacity:o,transition:v?f:{enter:{duration:0}},transitionEnd:{enter:h==null?void 0:h.enter,exit:r?h==null?void 0:h.exit:{...h==null?void 0:h.exit,display:w?"block":"none"}}},S=r?n:!0,P=n||r?"enter":"exit";return i.jsx(Ao,{initial:!1,custom:C,children:S&&i.jsx(bo.div,{ref:t,...m,className:Ot("chakra-collapse",d),style:{overflow:"hidden",display:"block",...u},custom:C,variants:fM,initial:r?"exit":!1,animate:P,exit:"exit"})})});fh.displayName="Collapse";var pM={enter:({transition:e,transitionEnd:t,delay:n}={})=>{var r;return{opacity:1,transition:(r=e==null?void 0:e.enter)!=null?r:vs.enter(Di.enter,n),transitionEnd:t==null?void 0:t.enter}},exit:({transition:e,transitionEnd:t,delay:n}={})=>{var r;return{opacity:0,transition:(r=e==null?void 0:e.exit)!=null?r:vs.exit(Di.exit,n),transitionEnd:t==null?void 0:t.exit}}},h5={initial:"exit",animate:"enter",exit:"exit",variants:pM},hM=p.forwardRef(function(t,n){const{unmountOnExit:r,in:o,className:s,transition:a,transitionEnd:u,delay:d,...f}=t,h=o||r?"enter":"exit",m=r?o&&r:!0,v={transition:a,transitionEnd:u,delay:d};return i.jsx(Ao,{custom:v,children:m&&i.jsx(bo.div,{ref:n,className:Ot("chakra-fade",s),custom:v,...h5,animate:h,...f})})});hM.displayName="Fade";var mM={exit:({reverse:e,initialScale:t,transition:n,transitionEnd:r,delay:o})=>{var s;return{opacity:0,...e?{scale:t,transitionEnd:r==null?void 0:r.exit}:{transitionEnd:{scale:t,...r==null?void 0:r.exit}},transition:(s=n==null?void 0:n.exit)!=null?s:vs.exit(Di.exit,o)}},enter:({transitionEnd:e,transition:t,delay:n})=>{var r;return{opacity:1,scale:1,transition:(r=t==null?void 0:t.enter)!=null?r:vs.enter(Di.enter,n),transitionEnd:e==null?void 0:e.enter}}},m5={initial:"exit",animate:"enter",exit:"exit",variants:mM},gM=p.forwardRef(function(t,n){const{unmountOnExit:r,in:o,reverse:s=!0,initialScale:a=.95,className:u,transition:d,transitionEnd:f,delay:h,...m}=t,v=r?o&&r:!0,b=o||r?"enter":"exit",w={initialScale:a,reverse:s,transition:d,transitionEnd:f,delay:h};return i.jsx(Ao,{custom:w,children:v&&i.jsx(bo.div,{ref:n,className:Ot("chakra-offset-slide",u),...m5,animate:b,custom:w,...m})})});gM.displayName="ScaleFade";var vM={initial:({offsetX:e,offsetY:t,transition:n,transitionEnd:r,delay:o})=>{var s;return{opacity:0,x:e,y:t,transition:(s=n==null?void 0:n.exit)!=null?s:vs.exit(Di.exit,o),transitionEnd:r==null?void 0:r.exit}},enter:({transition:e,transitionEnd:t,delay:n})=>{var r;return{opacity:1,x:0,y:0,transition:(r=e==null?void 0:e.enter)!=null?r:vs.enter(Di.enter,n),transitionEnd:t==null?void 0:t.enter}},exit:({offsetY:e,offsetX:t,transition:n,transitionEnd:r,reverse:o,delay:s})=>{var a;const u={x:t,y:e};return{opacity:0,transition:(a=n==null?void 0:n.exit)!=null?a:vs.exit(Di.exit,s),...o?{...u,transitionEnd:r==null?void 0:r.exit}:{transitionEnd:{...u,...r==null?void 0:r.exit}}}}},Cv={initial:"initial",animate:"enter",exit:"exit",variants:vM},yM=p.forwardRef(function(t,n){const{unmountOnExit:r,in:o,reverse:s=!0,className:a,offsetX:u=0,offsetY:d=8,transition:f,transitionEnd:h,delay:m,...v}=t,b=r?o&&r:!0,w=o||r?"enter":"exit",C={offsetX:u,offsetY:d,reverse:s,transition:f,transitionEnd:h,delay:m};return i.jsx(Ao,{custom:C,children:b&&i.jsx(bo.div,{ref:n,className:Ot("chakra-offset-slide",a),custom:C,...Cv,animate:w,...v})})});yM.displayName="SlideFade";var m2={exit:{duration:.15,ease:Ri.easeInOut},enter:{type:"spring",damping:25,stiffness:180}},bM={exit:({direction:e,transition:t,transitionEnd:n,delay:r})=>{var o;const{exit:s}=wv({direction:e});return{...s,transition:(o=t==null?void 0:t.exit)!=null?o:vs.exit(m2.exit,r),transitionEnd:n==null?void 0:n.exit}},enter:({direction:e,transitionEnd:t,transition:n,delay:r})=>{var o;const{enter:s}=wv({direction:e});return{...s,transition:(o=n==null?void 0:n.enter)!=null?o:vs.enter(m2.enter,r),transitionEnd:t==null?void 0:t.enter}}},g5=p.forwardRef(function(t,n){const{direction:r="right",style:o,unmountOnExit:s,in:a,className:u,transition:d,transitionEnd:f,delay:h,motionProps:m,...v}=t,b=wv({direction:r}),w=Object.assign({position:"fixed"},b.position,o),C=s?a&&s:!0,S=a||s?"enter":"exit",P={transitionEnd:f,transition:d,direction:r,delay:h};return i.jsx(Ao,{custom:P,children:C&&i.jsx(bo.div,{...v,ref:n,initial:"exit",className:Ot("chakra-slide",u),animate:S,exit:"exit",custom:P,variants:bM,style:w,...m})})});g5.displayName="Slide";var tu=nt(function(t,n){const{className:r,motionProps:o,...s}=t,{reduceMotion:a}=K0(),{getPanelProps:u,isOpen:d}=q0(),f=u(s,n),h=Ot("chakra-accordion__panel",r),m=dh();a||delete f.hidden;const v=i.jsx(Be.div,{...f,__css:m.panel,className:h});return a?v:i.jsx(fh,{in:d,...o,children:v})});tu.displayName="AccordionPanel";var v5=nt(function({children:t,reduceMotion:n,...r},o){const s=Xr("Accordion",r),a=xr(r),{htmlProps:u,descendants:d,...f}=oM(a),h=p.useMemo(()=>({...f,reduceMotion:!!n}),[f,n]);return i.jsx(tM,{value:d,children:i.jsx(sM,{value:h,children:i.jsx(J7,{value:s,children:i.jsx(Be.div,{ref:o,...u,className:Ot("chakra-accordion",r.className),__css:s.root,children:t})})})})});v5.displayName="Accordion";var xM={horizontal:{"> *:first-of-type:not(:last-of-type)":{borderEndRadius:0},"> *:not(:first-of-type):not(:last-of-type)":{borderRadius:0},"> *:not(:first-of-type):last-of-type":{borderStartRadius:0}},vertical:{"> *:first-of-type:not(:last-of-type)":{borderBottomRadius:0},"> *:not(:first-of-type):not(:last-of-type)":{borderRadius:0},"> *:not(:first-of-type):last-of-type":{borderTopRadius:0}}},SM={horizontal:e=>({"& > *:not(style) ~ *:not(style)":{marginStart:e}}),vertical:e=>({"& > *:not(style) ~ *:not(style)":{marginTop:e}})},gr=nt(function(t,n){const{size:r,colorScheme:o,variant:s,className:a,spacing:u="0.5rem",isAttached:d,isDisabled:f,orientation:h="horizontal",...m}=t,v=Ot("chakra-button__group",a),b=p.useMemo(()=>({size:r,colorScheme:o,variant:s,isDisabled:f}),[r,o,s,f]);let w={display:"inline-flex",...d?xM[h]:SM[h](u)};const C=h==="vertical";return i.jsx(sO,{value:b,children:i.jsx(Be.div,{ref:n,role:"group",__css:w,className:v,"data-attached":d?"":void 0,"data-orientation":h,flexDir:C?"column":void 0,...m})})});gr.displayName="ButtonGroup";var[$ae,wM]=ur({name:"CheckboxGroupContext",strict:!1});function CM(e){const[t,n]=p.useState(e),[r,o]=p.useState(!1);return e!==t&&(o(!0),n(e)),r}function PM(e){return i.jsx(Be.svg,{width:"1.2em",viewBox:"0 0 12 10",style:{fill:"none",strokeWidth:2,stroke:"currentColor",strokeDasharray:16},...e,children:i.jsx("polyline",{points:"1.5 6 4.5 9 10.5 1"})})}function kM(e){return i.jsx(Be.svg,{width:"1.2em",viewBox:"0 0 24 24",style:{stroke:"currentColor",strokeWidth:4},...e,children:i.jsx("line",{x1:"21",x2:"3",y1:"12",y2:"12"})})}function _M(e){const{isIndeterminate:t,isChecked:n,...r}=e,o=t?kM:PM;return n||t?i.jsx(Be.div,{style:{display:"flex",alignItems:"center",justifyContent:"center",height:"100%"},children:i.jsx(o,{...r})}):null}var[jM,EM]=ur({name:"FormErrorStylesContext",errorMessage:`useFormErrorStyles returned is 'undefined'. Seems you forgot to wrap the components in "" `}),po=nt((e,t)=>{const n=Xr("FormError",e),r=xr(e),o=R0();return o!=null&&o.isInvalid?i.jsx(jM,{value:n,children:i.jsx(Be.div,{...o==null?void 0:o.getErrorMessageProps(r,t),className:Ot("chakra-form__error-message",e.className),__css:{display:"flex",alignItems:"center",...n.text}})}):null});po.displayName="FormErrorMessage";var IM=nt((e,t)=>{const n=EM(),r=R0();if(!(r!=null&&r.isInvalid))return null;const o=Ot("chakra-form__error-icon",e.className);return i.jsx(mo,{ref:t,"aria-hidden":!0,...e,__css:n.icon,className:o,children:i.jsx("path",{fill:"currentColor",d:"M11.983,0a12.206,12.206,0,0,0-8.51,3.653A11.8,11.8,0,0,0,0,12.207,11.779,11.779,0,0,0,11.8,24h.214A12.111,12.111,0,0,0,24,11.791h0A11.766,11.766,0,0,0,11.983,0ZM10.5,16.542a1.476,1.476,0,0,1,1.449-1.53h.027a1.527,1.527,0,0,1,1.523,1.47,1.475,1.475,0,0,1-1.449,1.53h-.027A1.529,1.529,0,0,1,10.5,16.542ZM11,12.5v-6a1,1,0,0,1,2,0v6a1,1,0,1,1-2,0Z"})})});IM.displayName="FormErrorIcon";var Y0={border:"0",clip:"rect(0, 0, 0, 0)",height:"1px",width:"1px",margin:"-1px",padding:"0",overflow:"hidden",whiteSpace:"nowrap",position:"absolute"},y5=Be("span",{baseStyle:Y0});y5.displayName="VisuallyHidden";var OM=Be("input",{baseStyle:Y0});OM.displayName="VisuallyHiddenInput";var g2=!1,Fu=null,Hi=!1,Pv=!1,kv=new Set;function X0(e,t){kv.forEach(n=>n(e,t))}var RM=typeof window<"u"&&window.navigator!=null?/^Mac/.test(window.navigator.platform):!1;function MM(e){return!(e.metaKey||!RM&&e.altKey||e.ctrlKey||e.key==="Control"||e.key==="Shift"||e.key==="Meta")}function v2(e){Hi=!0,MM(e)&&(Fu="keyboard",X0("keyboard",e))}function yl(e){if(Fu="pointer",e.type==="mousedown"||e.type==="pointerdown"){Hi=!0;const t=e.composedPath?e.composedPath()[0]:e.target;let n=!1;try{n=t.matches(":focus-visible")}catch{}if(n)return;X0("pointer",e)}}function TM(e){return e.mozInputSource===0&&e.isTrusted?!0:e.detail===0&&!e.pointerType}function DM(e){TM(e)&&(Hi=!0,Fu="virtual")}function AM(e){e.target===window||e.target===document||(!Hi&&!Pv&&(Fu="virtual",X0("virtual",e)),Hi=!1,Pv=!1)}function NM(){Hi=!1,Pv=!0}function y2(){return Fu!=="pointer"}function zM(){if(typeof window>"u"||g2)return;const{focus:e}=HTMLElement.prototype;HTMLElement.prototype.focus=function(...n){Hi=!0,e.apply(this,n)},document.addEventListener("keydown",v2,!0),document.addEventListener("keyup",v2,!0),document.addEventListener("click",DM,!0),window.addEventListener("focus",AM,!0),window.addEventListener("blur",NM,!1),typeof PointerEvent<"u"?(document.addEventListener("pointerdown",yl,!0),document.addEventListener("pointermove",yl,!0),document.addEventListener("pointerup",yl,!0)):(document.addEventListener("mousedown",yl,!0),document.addEventListener("mousemove",yl,!0),document.addEventListener("mouseup",yl,!0)),g2=!0}function b5(e){zM(),e(y2());const t=()=>e(y2());return kv.add(t),()=>{kv.delete(t)}}function $M(e,t=[]){const n=Object.assign({},e);for(const r of t)r in n&&delete n[r];return n}function x5(e={}){const t=aO(e),{isDisabled:n,isReadOnly:r,isRequired:o,isInvalid:s,id:a,onBlur:u,onFocus:d,"aria-describedby":f}=t,{defaultChecked:h,isChecked:m,isFocusable:v,onChange:b,isIndeterminate:w,name:C,value:S,tabIndex:P=void 0,"aria-label":k,"aria-labelledby":j,"aria-invalid":E,...I}=e,O=$M(I,["isDisabled","isReadOnly","isRequired","isInvalid","id","onBlur","onFocus","aria-describedby"]),R=qa(b),M=qa(u),A=qa(d),[L,D]=p.useState(!1),[V,$]=p.useState(!1),[F,U]=p.useState(!1),[X,N]=p.useState(!1);p.useEffect(()=>b5(D),[]);const G=p.useRef(null),[W,q]=p.useState(!0),[ne,ve]=p.useState(!!h),le=m!==void 0,oe=le?m:ne,Y=p.useCallback(ye=>{if(r||n){ye.preventDefault();return}le||ve(oe?ye.target.checked:w?!0:ye.target.checked),R==null||R(ye)},[r,n,oe,le,w,R]);su(()=>{G.current&&(G.current.indeterminate=!!w)},[w]),Zi(()=>{n&&$(!1)},[n,$]),su(()=>{const ye=G.current;ye!=null&&ye.form&&(ye.form.onreset=()=>{ve(!!h)})},[]);const ae=n&&!v,de=p.useCallback(ye=>{ye.key===" "&&N(!0)},[N]),xe=p.useCallback(ye=>{ye.key===" "&&N(!1)},[N]);su(()=>{if(!G.current)return;G.current.checked!==oe&&ve(G.current.checked)},[G.current]);const se=p.useCallback((ye={},Oe=null)=>{const at=Fe=>{V&&Fe.preventDefault(),N(!0)};return{...ye,ref:Oe,"data-active":on(X),"data-hover":on(F),"data-checked":on(oe),"data-focus":on(V),"data-focus-visible":on(V&&L),"data-indeterminate":on(w),"data-disabled":on(n),"data-invalid":on(s),"data-readonly":on(r),"aria-hidden":!0,onMouseDown:bt(ye.onMouseDown,at),onMouseUp:bt(ye.onMouseUp,()=>N(!1)),onMouseEnter:bt(ye.onMouseEnter,()=>U(!0)),onMouseLeave:bt(ye.onMouseLeave,()=>U(!1))}},[X,oe,n,V,L,F,w,s,r]),ge=p.useCallback((ye={},Oe=null)=>({...O,...ye,ref:Un(Oe,at=>{at&&q(at.tagName==="LABEL")}),onClick:bt(ye.onClick,()=>{var at;W||((at=G.current)==null||at.click(),requestAnimationFrame(()=>{var Fe;(Fe=G.current)==null||Fe.focus({preventScroll:!0})}))}),"data-disabled":on(n),"data-checked":on(oe),"data-invalid":on(s)}),[O,n,oe,s,W]),fe=p.useCallback((ye={},Oe=null)=>({...ye,ref:Un(G,Oe),type:"checkbox",name:C,value:S,id:a,tabIndex:P,onChange:bt(ye.onChange,Y),onBlur:bt(ye.onBlur,M,()=>$(!1)),onFocus:bt(ye.onFocus,A,()=>$(!0)),onKeyDown:bt(ye.onKeyDown,de),onKeyUp:bt(ye.onKeyUp,xe),required:o,checked:oe,disabled:ae,readOnly:r,"aria-label":k,"aria-labelledby":j,"aria-invalid":E?!!E:s,"aria-describedby":f,"aria-disabled":n,style:Y0}),[C,S,a,Y,M,A,de,xe,o,oe,ae,r,k,j,E,s,f,n,P]),ke=p.useCallback((ye={},Oe=null)=>({...ye,ref:Oe,onMouseDown:bt(ye.onMouseDown,LM),"data-disabled":on(n),"data-checked":on(oe),"data-invalid":on(s)}),[oe,n,s]);return{state:{isInvalid:s,isFocused:V,isChecked:oe,isActive:X,isHovered:F,isIndeterminate:w,isDisabled:n,isReadOnly:r,isRequired:o},getRootProps:ge,getCheckboxProps:se,getInputProps:fe,getLabelProps:ke,htmlProps:O}}function LM(e){e.preventDefault(),e.stopPropagation()}var FM={display:"inline-flex",alignItems:"center",justifyContent:"center",verticalAlign:"top",userSelect:"none",flexShrink:0},BM={cursor:"pointer",display:"inline-flex",alignItems:"center",verticalAlign:"top",position:"relative"},HM=xa({from:{opacity:0,strokeDashoffset:16,transform:"scale(0.95)"},to:{opacity:1,strokeDashoffset:0,transform:"scale(1)"}}),WM=xa({from:{opacity:0},to:{opacity:1}}),VM=xa({from:{transform:"scaleX(0.65)"},to:{transform:"scaleX(1)"}}),S5=nt(function(t,n){const r=wM(),o={...r,...t},s=Xr("Checkbox",o),a=xr(t),{spacing:u="0.5rem",className:d,children:f,iconColor:h,iconSize:m,icon:v=i.jsx(_M,{}),isChecked:b,isDisabled:w=r==null?void 0:r.isDisabled,onChange:C,inputProps:S,...P}=a;let k=b;r!=null&&r.value&&a.value&&(k=r.value.includes(a.value));let j=C;r!=null&&r.onChange&&a.value&&(j=oh(r.onChange,C));const{state:E,getInputProps:I,getCheckboxProps:O,getLabelProps:R,getRootProps:M}=x5({...P,isDisabled:w,isChecked:k,onChange:j}),A=CM(E.isChecked),L=p.useMemo(()=>({animation:A?E.isIndeterminate?`${WM} 20ms linear, ${VM} 200ms linear`:`${HM} 200ms linear`:void 0,fontSize:m,color:h,...s.icon}),[h,m,A,E.isIndeterminate,s.icon]),D=p.cloneElement(v,{__css:L,isIndeterminate:E.isIndeterminate,isChecked:E.isChecked});return i.jsxs(Be.label,{__css:{...BM,...s.container},className:Ot("chakra-checkbox",d),...M(),children:[i.jsx("input",{className:"chakra-checkbox__input",...I(S,n)}),i.jsx(Be.span,{__css:{...FM,...s.control},className:"chakra-checkbox__control",...O(),children:D}),f&&i.jsx(Be.span,{className:"chakra-checkbox__label",...R(),__css:{marginStart:u,...s.label},children:f})]})});S5.displayName="Checkbox";var[UM,w5]=ur({name:"EditableStylesContext",errorMessage:`useEditableStyles returned is 'undefined'. Seems you forgot to wrap the components in "" `}),[GM,C5]=ur({name:"EditableContext",errorMessage:"useEditableContext: context is undefined. Seems you forgot to wrap the editable components in ``"}),P5={fontSize:"inherit",fontWeight:"inherit",textAlign:"inherit",bg:"transparent"},k5=nt(function(t,n){const{getInputProps:r}=C5(),o=w5(),s=r(t,n),a=Ot("chakra-editable__input",t.className);return i.jsx(Be.input,{...s,__css:{outline:0,...P5,...o.input},className:a})});k5.displayName="EditableInput";var _5=nt(function(t,n){const{getPreviewProps:r}=C5(),o=w5(),s=r(t,n),a=Ot("chakra-editable__preview",t.className);return i.jsx(Be.span,{...s,__css:{cursor:"text",display:"inline-block",...P5,...o.preview},className:a})});_5.displayName="EditablePreview";function qM(e){return"current"in e}var j5=()=>typeof window<"u";function KM(){var e;const t=navigator.userAgentData;return(e=t==null?void 0:t.platform)!=null?e:navigator.platform}var YM=e=>j5()&&e.test(navigator.vendor),XM=e=>j5()&&e.test(KM()),QM=()=>XM(/mac|iphone|ipad|ipod/i),ZM=()=>QM()&&YM(/apple/i);function E5(e){const{ref:t,elements:n,enabled:r}=e,o=()=>{var s,a;return(a=(s=t.current)==null?void 0:s.ownerDocument)!=null?a:document};M0(o,"pointerdown",s=>{if(!ZM()||!r)return;const a=s.target,d=(n??[t]).some(f=>{const h=qM(f)?f.current:f;return(h==null?void 0:h.contains(a))||h===a});o().activeElement!==a&&d&&(s.preventDefault(),a.focus())})}function b2(e,t){return e?e===t||e.contains(t):!1}function JM(e={}){const{onChange:t,onCancel:n,onSubmit:r,onBlur:o,value:s,isDisabled:a,defaultValue:u,startWithEditView:d,isPreviewFocusable:f=!0,submitOnBlur:h=!0,selectAllOnFocus:m=!0,placeholder:v,onEdit:b,finalFocusRef:w,...C}=e,S=qa(b),P=!!(d&&!a),[k,j]=p.useState(P),[E,I]=Ru({defaultValue:u||"",value:s,onChange:t}),[O,R]=p.useState(E),M=p.useRef(null),A=p.useRef(null),L=p.useRef(null),D=p.useRef(null),V=p.useRef(null);E5({ref:M,enabled:k,elements:[D,V]});const $=!k&&!a;su(()=>{var se,ge;k&&((se=M.current)==null||se.focus(),m&&((ge=M.current)==null||ge.select()))},[]),Zi(()=>{var se,ge,fe,ke;if(!k){w?(se=w.current)==null||se.focus():(ge=L.current)==null||ge.focus();return}(fe=M.current)==null||fe.focus(),m&&((ke=M.current)==null||ke.select()),S==null||S()},[k,S,m]);const F=p.useCallback(()=>{$&&j(!0)},[$]),U=p.useCallback(()=>{R(E)},[E]),X=p.useCallback(()=>{j(!1),I(O),n==null||n(O),o==null||o(O)},[n,o,I,O]),N=p.useCallback(()=>{j(!1),R(E),r==null||r(E),o==null||o(O)},[E,r,o,O]);p.useEffect(()=>{if(k)return;const se=M.current;(se==null?void 0:se.ownerDocument.activeElement)===se&&(se==null||se.blur())},[k]);const G=p.useCallback(se=>{I(se.currentTarget.value)},[I]),W=p.useCallback(se=>{const ge=se.key,ke={Escape:X,Enter:Ae=>{!Ae.shiftKey&&!Ae.metaKey&&N()}}[ge];ke&&(se.preventDefault(),ke(se))},[X,N]),q=p.useCallback(se=>{const ge=se.key,ke={Escape:X}[ge];ke&&(se.preventDefault(),ke(se))},[X]),ne=E.length===0,ve=p.useCallback(se=>{var ge;if(!k)return;const fe=se.currentTarget.ownerDocument,ke=(ge=se.relatedTarget)!=null?ge:fe.activeElement,Ae=b2(D.current,ke),ye=b2(V.current,ke);!Ae&&!ye&&(h?N():X())},[h,N,X,k]),le=p.useCallback((se={},ge=null)=>{const fe=$&&f?0:void 0;return{...se,ref:Un(ge,A),children:ne?v:E,hidden:k,"aria-disabled":Al(a),tabIndex:fe,onFocus:bt(se.onFocus,F,U)}},[a,k,$,f,ne,F,U,v,E]),oe=p.useCallback((se={},ge=null)=>({...se,hidden:!k,placeholder:v,ref:Un(ge,M),disabled:a,"aria-disabled":Al(a),value:E,onBlur:bt(se.onBlur,ve),onChange:bt(se.onChange,G),onKeyDown:bt(se.onKeyDown,W),onFocus:bt(se.onFocus,U)}),[a,k,ve,G,W,U,v,E]),Y=p.useCallback((se={},ge=null)=>({...se,hidden:!k,placeholder:v,ref:Un(ge,M),disabled:a,"aria-disabled":Al(a),value:E,onBlur:bt(se.onBlur,ve),onChange:bt(se.onChange,G),onKeyDown:bt(se.onKeyDown,q),onFocus:bt(se.onFocus,U)}),[a,k,ve,G,q,U,v,E]),ae=p.useCallback((se={},ge=null)=>({"aria-label":"Edit",...se,type:"button",onClick:bt(se.onClick,F),ref:Un(ge,L),disabled:a}),[F,a]),de=p.useCallback((se={},ge=null)=>({...se,"aria-label":"Submit",ref:Un(V,ge),type:"button",onClick:bt(se.onClick,N),disabled:a}),[N,a]),xe=p.useCallback((se={},ge=null)=>({"aria-label":"Cancel",id:"cancel",...se,ref:Un(D,ge),type:"button",onClick:bt(se.onClick,X),disabled:a}),[X,a]);return{isEditing:k,isDisabled:a,isValueEmpty:ne,value:E,onEdit:F,onCancel:X,onSubmit:N,getPreviewProps:le,getInputProps:oe,getTextareaProps:Y,getEditButtonProps:ae,getSubmitButtonProps:de,getCancelButtonProps:xe,htmlProps:C}}var I5=nt(function(t,n){const r=Xr("Editable",t),o=xr(t),{htmlProps:s,...a}=JM(o),{isEditing:u,onSubmit:d,onCancel:f,onEdit:h}=a,m=Ot("chakra-editable",t.className),v=T0(t.children,{isEditing:u,onSubmit:d,onCancel:f,onEdit:h});return i.jsx(GM,{value:a,children:i.jsx(UM,{value:r,children:i.jsx(Be.div,{ref:n,...s,className:m,children:v})})})});I5.displayName="Editable";var O5={exports:{}},eT="SECRET_DO_NOT_PASS_THIS_OR_YOU_WILL_BE_FIRED",tT=eT,nT=tT;function R5(){}function M5(){}M5.resetWarningCache=R5;var rT=function(){function e(r,o,s,a,u,d){if(d!==nT){var f=new Error("Calling PropTypes validators directly is not supported by the `prop-types` package. Use PropTypes.checkPropTypes() to call them. Read more at http://fb.me/use-check-prop-types");throw f.name="Invariant Violation",f}}e.isRequired=e;function t(){return e}var n={array:e,bigint:e,bool:e,func:e,number:e,object:e,string:e,symbol:e,any:e,arrayOf:t,element:e,elementType:e,instanceOf:t,node:e,objectOf:t,oneOf:t,oneOfType:t,shape:t,exact:t,checkPropTypes:M5,resetWarningCache:R5};return n.PropTypes=n,n};O5.exports=rT();var oT=O5.exports;const An=dc(oT);var _v="data-focus-lock",T5="data-focus-lock-disabled",sT="data-no-focus-lock",aT="data-autofocus-inside",iT="data-no-autofocus";function lT(e,t){return typeof e=="function"?e(t):e&&(e.current=t),e}function cT(e,t){var n=p.useState(function(){return{value:e,callback:t,facade:{get current(){return n.value},set current(r){var o=n.value;o!==r&&(n.value=r,n.callback(r,o))}}}})[0];return n.callback=t,n.facade}function D5(e,t){return cT(t||null,function(n){return e.forEach(function(r){return lT(r,n)})})}var mg={width:"1px",height:"0px",padding:0,overflow:"hidden",position:"fixed",top:"1px",left:"1px"},Bs=function(){return Bs=Object.assign||function(t){for(var n,r=1,o=arguments.length;r0&&s[s.length-1])&&(f[0]===6||f[0]===2)){n=0;continue}if(f[0]===3&&(!s||f[1]>s[0]&&f[1]0)&&!(o=r.next()).done;)s.push(o.value)}catch(u){a={error:u}}finally{try{o&&!o.done&&(n=r.return)&&n.call(r)}finally{if(a)throw a.error}}return s}function jv(e,t,n){if(n||arguments.length===2)for(var r=0,o=t.length,s;r=0}).sort(CT)},PT=["button:enabled","select:enabled","textarea:enabled","input:enabled","a[href]","area[href]","summary","iframe","object","embed","audio[controls]","video[controls]","[tabindex]","[contenteditable]","[autofocus]"],e1=PT.join(","),kT="".concat(e1,", [data-focus-guard]"),Z5=function(e,t){return Ys((e.shadowRoot||e).children).reduce(function(n,r){return n.concat(r.matches(t?kT:e1)?[r]:[],Z5(r))},[])},_T=function(e,t){var n;return e instanceof HTMLIFrameElement&&(!((n=e.contentDocument)===null||n===void 0)&&n.body)?ph([e.contentDocument.body],t):[e]},ph=function(e,t){return e.reduce(function(n,r){var o,s=Z5(r,t),a=(o=[]).concat.apply(o,s.map(function(u){return _T(u,t)}));return n.concat(a,r.parentNode?Ys(r.parentNode.querySelectorAll(e1)).filter(function(u){return u===r}):[])},[])},jT=function(e){var t=e.querySelectorAll("[".concat(aT,"]"));return Ys(t).map(function(n){return ph([n])}).reduce(function(n,r){return n.concat(r)},[])},t1=function(e,t){return Ys(e).filter(function(n){return G5(t,n)}).filter(function(n){return xT(n)})},S2=function(e,t){return t===void 0&&(t=new Map),Ys(e).filter(function(n){return q5(t,n)})},Iv=function(e,t,n){return Q5(t1(ph(e,n),t),!0,n)},w2=function(e,t){return Q5(t1(ph(e),t),!1)},ET=function(e,t){return t1(jT(e),t)},$l=function(e,t){return e.shadowRoot?$l(e.shadowRoot,t):Object.getPrototypeOf(e).contains!==void 0&&Object.getPrototypeOf(e).contains.call(e,t)?!0:Ys(e.children).some(function(n){var r;if(n instanceof HTMLIFrameElement){var o=(r=n.contentDocument)===null||r===void 0?void 0:r.body;return o?$l(o,t):!1}return $l(n,t)})},IT=function(e){for(var t=new Set,n=e.length,r=0;r0&&t.add(o),(s&Node.DOCUMENT_POSITION_CONTAINS)>0&&t.add(r)}return e.filter(function(a,u){return!t.has(u)})},J5=function(e){return e.parentNode?J5(e.parentNode):e},n1=function(e){var t=Qf(e);return t.filter(Boolean).reduce(function(n,r){var o=r.getAttribute(_v);return n.push.apply(n,o?IT(Ys(J5(r).querySelectorAll("[".concat(_v,'="').concat(o,'"]:not([').concat(T5,'="disabled"])')))):[r]),n},[])},OT=function(e){try{return e()}catch{return}},vu=function(e){if(e===void 0&&(e=document),!(!e||!e.activeElement)){var t=e.activeElement;return t.shadowRoot?vu(t.shadowRoot):t instanceof HTMLIFrameElement&&OT(function(){return t.contentWindow.document})?vu(t.contentWindow.document):t}},RT=function(e,t){return e===t},MT=function(e,t){return!!Ys(e.querySelectorAll("iframe")).some(function(n){return RT(n,t)})},e6=function(e,t){return t===void 0&&(t=vu(W5(e).ownerDocument)),!t||t.dataset&&t.dataset.focusGuard?!1:n1(e).some(function(n){return $l(n,t)||MT(n,t)})},TT=function(e){e===void 0&&(e=document);var t=vu(e);return t?Ys(e.querySelectorAll("[".concat(sT,"]"))).some(function(n){return $l(n,t)}):!1},DT=function(e,t){return t.filter(X5).filter(function(n){return n.name===e.name}).filter(function(n){return n.checked})[0]||e},r1=function(e,t){return X5(e)&&e.name?DT(e,t):e},AT=function(e){var t=new Set;return e.forEach(function(n){return t.add(r1(n,e))}),e.filter(function(n){return t.has(n)})},C2=function(e){return e[0]&&e.length>1?r1(e[0],e):e[0]},P2=function(e,t){return e.length>1?e.indexOf(r1(e[t],e)):t},t6="NEW_FOCUS",NT=function(e,t,n,r){var o=e.length,s=e[0],a=e[o-1],u=J0(n);if(!(n&&e.indexOf(n)>=0)){var d=n!==void 0?t.indexOf(n):-1,f=r?t.indexOf(r):d,h=r?e.indexOf(r):-1,m=d-f,v=t.indexOf(s),b=t.indexOf(a),w=AT(t),C=n!==void 0?w.indexOf(n):-1,S=C-(r?w.indexOf(r):d),P=P2(e,0),k=P2(e,o-1);if(d===-1||h===-1)return t6;if(!m&&h>=0)return h;if(d<=v&&u&&Math.abs(m)>1)return k;if(d>=b&&u&&Math.abs(m)>1)return P;if(m&&Math.abs(S)>1)return h;if(d<=v)return k;if(d>b)return P;if(m)return Math.abs(m)>1?h:(o+h+m)%o}},zT=function(e){return function(t){var n,r=(n=K5(t))===null||n===void 0?void 0:n.autofocus;return t.autofocus||r!==void 0&&r!=="false"||e.indexOf(t)>=0}},$T=function(e,t,n){var r=e.map(function(s){var a=s.node;return a}),o=S2(r.filter(zT(n)));return o&&o.length?C2(o):C2(S2(t))},Ov=function(e,t){return t===void 0&&(t=[]),t.push(e),e.parentNode&&Ov(e.parentNode.host||e.parentNode,t),t},gg=function(e,t){for(var n=Ov(e),r=Ov(t),o=0;o=0)return s}return!1},n6=function(e,t,n){var r=Qf(e),o=Qf(t),s=r[0],a=!1;return o.filter(Boolean).forEach(function(u){a=gg(a||u,u)||a,n.filter(Boolean).forEach(function(d){var f=gg(s,d);f&&(!a||$l(f,a)?a=f:a=gg(f,a))})}),a},LT=function(e,t){return e.reduce(function(n,r){return n.concat(ET(r,t))},[])},FT=function(e,t){var n=new Map;return t.forEach(function(r){return n.set(r.node,r)}),e.map(function(r){return n.get(r)}).filter(wT)},BT=function(e,t){var n=vu(Qf(e).length>0?document:W5(e).ownerDocument),r=n1(e).filter(Zf),o=n6(n||e,e,r),s=new Map,a=w2(r,s),u=Iv(r,s).filter(function(b){var w=b.node;return Zf(w)});if(!(!u[0]&&(u=a,!u[0]))){var d=w2([o],s).map(function(b){var w=b.node;return w}),f=FT(d,u),h=f.map(function(b){var w=b.node;return w}),m=NT(h,d,n,t);if(m===t6){var v=$T(a,h,LT(r,s));if(v)return{node:v};console.warn("focus-lock: cannot find any node to move focus into");return}return m===void 0?m:f[m]}},HT=function(e){var t=n1(e).filter(Zf),n=n6(e,e,t),r=new Map,o=Iv([n],r,!0),s=Iv(t,r).filter(function(a){var u=a.node;return Zf(u)}).map(function(a){var u=a.node;return u});return o.map(function(a){var u=a.node,d=a.index;return{node:u,index:d,lockItem:s.indexOf(u)>=0,guard:J0(u)}})},WT=function(e,t){"focus"in e&&e.focus(t),"contentWindow"in e&&e.contentWindow&&e.contentWindow.focus()},vg=0,yg=!1,r6=function(e,t,n){n===void 0&&(n={});var r=BT(e,t);if(!yg&&r){if(vg>2){console.error("FocusLock: focus-fighting detected. Only one focus management system could be active. See https://github.com/theKashey/focus-lock/#focus-fighting"),yg=!0,setTimeout(function(){yg=!1},1);return}vg++,WT(r.node,n.focusOptions),vg--}};function o6(e){var t=window,n=t.setImmediate;typeof n<"u"?n(e):setTimeout(e,1)}var VT=function(){return document&&document.activeElement===document.body},UT=function(){return VT()||TT()},Ll=null,Il=null,Fl=null,yu=!1,GT=function(){return!0},qT=function(t){return(Ll.whiteList||GT)(t)},KT=function(t,n){Fl={observerNode:t,portaledElement:n}},YT=function(t){return Fl&&Fl.portaledElement===t};function k2(e,t,n,r){var o=null,s=e;do{var a=r[s];if(a.guard)a.node.dataset.focusAutoGuard&&(o=a);else if(a.lockItem){if(s!==e)return;o=null}else break}while((s+=n)!==t);o&&(o.node.tabIndex=0)}var XT=function(t){return t&&"current"in t?t.current:t},QT=function(t){return t?!!yu:yu==="meanwhile"},ZT=function e(t,n,r){return n&&(n.host===t&&(!n.activeElement||r.contains(n.activeElement))||n.parentNode&&e(t,n.parentNode,r))},JT=function(t,n){return n.some(function(r){return ZT(t,r,r)})},Jf=function(){var t=!1;if(Ll){var n=Ll,r=n.observed,o=n.persistentFocus,s=n.autoFocus,a=n.shards,u=n.crossFrame,d=n.focusOptions,f=r||Fl&&Fl.portaledElement,h=document&&document.activeElement;if(f){var m=[f].concat(a.map(XT).filter(Boolean));if((!h||qT(h))&&(o||QT(u)||!UT()||!Il&&s)&&(f&&!(e6(m)||h&&JT(h,m)||YT(h))&&(document&&!Il&&h&&!s?(h.blur&&h.blur(),document.body.focus()):(t=r6(m,Il,{focusOptions:d}),Fl={})),yu=!1,Il=document&&document.activeElement),document){var v=document&&document.activeElement,b=HT(m),w=b.map(function(C){var S=C.node;return S}).indexOf(v);w>-1&&(b.filter(function(C){var S=C.guard,P=C.node;return S&&P.dataset.focusAutoGuard}).forEach(function(C){var S=C.node;return S.removeAttribute("tabIndex")}),k2(w,b.length,1,b),k2(w,-1,-1,b))}}}return t},s6=function(t){Jf()&&t&&(t.stopPropagation(),t.preventDefault())},o1=function(){return o6(Jf)},eD=function(t){var n=t.target,r=t.currentTarget;r.contains(n)||KT(r,n)},tD=function(){return null},a6=function(){yu="just",setTimeout(function(){yu="meanwhile"},0)},nD=function(){document.addEventListener("focusin",s6),document.addEventListener("focusout",o1),window.addEventListener("blur",a6)},rD=function(){document.removeEventListener("focusin",s6),document.removeEventListener("focusout",o1),window.removeEventListener("blur",a6)};function oD(e){return e.filter(function(t){var n=t.disabled;return!n})}function sD(e){var t=e.slice(-1)[0];t&&!Ll&&nD();var n=Ll,r=n&&t&&t.id===n.id;Ll=t,n&&!r&&(n.onDeactivation(),e.filter(function(o){var s=o.id;return s===n.id}).length||n.returnFocus(!t)),t?(Il=null,(!r||n.observed!==t.observed)&&t.onActivation(),Jf(),o6(Jf)):(rD(),Il=null)}F5.assignSyncMedium(eD);B5.assignMedium(o1);dT.assignMedium(function(e){return e({moveFocusInside:r6,focusInside:e6})});const aD=mT(oD,sD)(tD);var i6=p.forwardRef(function(t,n){return p.createElement(H5,nr({sideCar:aD,ref:n},t))}),l6=H5.propTypes||{};l6.sideCar;j7(l6,["sideCar"]);i6.propTypes={};const _2=i6;var c6=e=>e.hasAttribute("tabindex"),iD=e=>c6(e)&&e.tabIndex===-1;function lD(e){return!!e.getAttribute("disabled")||!!e.getAttribute("aria-disabled")}function u6(e){return e.parentElement&&u6(e.parentElement)?!0:e.hidden}function cD(e){const t=e.getAttribute("contenteditable");return t!=="false"&&t!=null}function d6(e){if(!t3(e)||u6(e)||lD(e))return!1;const{localName:t}=e;if(["input","select","textarea","button"].indexOf(t)>=0)return!0;const r={a:()=>e.hasAttribute("href"),audio:()=>e.hasAttribute("controls"),video:()=>e.hasAttribute("controls")};return t in r?r[t]():cD(e)?!0:c6(e)}function uD(e){return e?t3(e)&&d6(e)&&!iD(e):!1}var dD=["input:not(:disabled):not([disabled])","select:not(:disabled):not([disabled])","textarea:not(:disabled):not([disabled])","embed","iframe","object","a[href]","area[href]","button:not(:disabled):not([disabled])","[tabindex]","audio[controls]","video[controls]","*[tabindex]:not([aria-disabled])","*[contenteditable]"],fD=dD.join(),pD=e=>e.offsetWidth>0&&e.offsetHeight>0;function f6(e){const t=Array.from(e.querySelectorAll(fD));return t.unshift(e),t.filter(n=>d6(n)&&pD(n))}var j2,hD=(j2=_2.default)!=null?j2:_2,p6=e=>{const{initialFocusRef:t,finalFocusRef:n,contentRef:r,restoreFocus:o,children:s,isDisabled:a,autoFocus:u,persistentFocus:d,lockFocusAcrossFrames:f}=e,h=p.useCallback(()=>{t!=null&&t.current?t.current.focus():r!=null&&r.current&&f6(r.current).length===0&&requestAnimationFrame(()=>{var w;(w=r.current)==null||w.focus()})},[t,r]),m=p.useCallback(()=>{var b;(b=n==null?void 0:n.current)==null||b.focus()},[n]),v=o&&!n;return i.jsx(hD,{crossFrame:f,persistentFocus:d,autoFocus:u,disabled:a,onActivation:h,onDeactivation:m,returnFocus:v,children:s})};p6.displayName="FocusLock";function mD(e,t,n,r){const o=n3(t);return p.useEffect(()=>{var s;const a=(s=vx(n))!=null?s:document;if(t)return a.addEventListener(e,o,r),()=>{a.removeEventListener(e,o,r)}},[e,n,r,o,t]),()=>{var s;((s=vx(n))!=null?s:document).removeEventListener(e,o,r)}}function gD(e){const{ref:t,handler:n,enabled:r=!0}=e,o=n3(n),a=p.useRef({isPointerDown:!1,ignoreEmulatedMouseEvents:!1}).current;p.useEffect(()=>{if(!r)return;const u=m=>{bg(m,t)&&(a.isPointerDown=!0)},d=m=>{if(a.ignoreEmulatedMouseEvents){a.ignoreEmulatedMouseEvents=!1;return}a.isPointerDown&&n&&bg(m,t)&&(a.isPointerDown=!1,o(m))},f=m=>{a.ignoreEmulatedMouseEvents=!0,n&&a.isPointerDown&&bg(m,t)&&(a.isPointerDown=!1,o(m))},h=r3(t.current);return h.addEventListener("mousedown",u,!0),h.addEventListener("mouseup",d,!0),h.addEventListener("touchstart",u,!0),h.addEventListener("touchend",f,!0),()=>{h.removeEventListener("mousedown",u,!0),h.removeEventListener("mouseup",d,!0),h.removeEventListener("touchstart",u,!0),h.removeEventListener("touchend",f,!0)}},[n,t,o,a,r])}function bg(e,t){var n;const r=e.target;return r&&!r3(r).contains(r)?!1:!((n=t.current)!=null&&n.contains(r))}var[vD,yD]=ur({name:"InputGroupStylesContext",errorMessage:`useInputGroupStyles returned is 'undefined'. Seems you forgot to wrap the components in "" `}),h6=nt(function(t,n){const r=Xr("Input",t),{children:o,className:s,...a}=xr(t),u=Ot("chakra-input__group",s),d={},f=sh(o),h=r.field;f.forEach(v=>{var b,w;r&&(h&&v.type.id==="InputLeftElement"&&(d.paddingStart=(b=h.height)!=null?b:h.h),h&&v.type.id==="InputRightElement"&&(d.paddingEnd=(w=h.height)!=null?w:h.h),v.type.id==="InputRightAddon"&&(d.borderEndRadius=0),v.type.id==="InputLeftAddon"&&(d.borderStartRadius=0))});const m=f.map(v=>{var b,w;const C=U0({size:((b=v.props)==null?void 0:b.size)||t.size,variant:((w=v.props)==null?void 0:w.variant)||t.variant});return v.type.id!=="Input"?p.cloneElement(v,C):p.cloneElement(v,Object.assign(C,d,v.props))});return i.jsx(Be.div,{className:u,ref:n,__css:{width:"100%",display:"flex",position:"relative",isolation:"isolate"},"data-group":!0,...a,children:i.jsx(vD,{value:r,children:m})})});h6.displayName="InputGroup";var bD=Be("div",{baseStyle:{display:"flex",alignItems:"center",justifyContent:"center",position:"absolute",top:"0",zIndex:2}}),hh=nt(function(t,n){var r,o;const{placement:s="left",...a}=t,u=yD(),d=u.field,h={[s==="left"?"insetStart":"insetEnd"]:"0",width:(r=d==null?void 0:d.height)!=null?r:d==null?void 0:d.h,height:(o=d==null?void 0:d.height)!=null?o:d==null?void 0:d.h,fontSize:d==null?void 0:d.fontSize,...u.element};return i.jsx(bD,{ref:n,__css:h,...a})});hh.id="InputElement";hh.displayName="InputElement";var m6=nt(function(t,n){const{className:r,...o}=t,s=Ot("chakra-input__left-element",r);return i.jsx(hh,{ref:n,placement:"left",className:s,...o})});m6.id="InputLeftElement";m6.displayName="InputLeftElement";var s1=nt(function(t,n){const{className:r,...o}=t,s=Ot("chakra-input__right-element",r);return i.jsx(hh,{ref:n,placement:"right",className:s,...o})});s1.id="InputRightElement";s1.displayName="InputRightElement";var Bu=nt(function(t,n){const{htmlSize:r,...o}=t,s=Xr("Input",o),a=xr(o),u=D0(a),d=Ot("chakra-input",t.className);return i.jsx(Be.input,{size:r,...u,__css:s.field,ref:n,className:d})});Bu.displayName="Input";Bu.id="Input";var Ai=nt(function(t,n){const r=fc("Link",t),{className:o,isExternal:s,...a}=xr(t);return i.jsx(Be.a,{target:s?"_blank":void 0,rel:s?"noopener":void 0,ref:n,className:Ot("chakra-link",o),...a,__css:r})});Ai.displayName="Link";var[xD,g6]=ur({name:"ListStylesContext",errorMessage:`useListStyles returned is 'undefined'. Seems you forgot to wrap the components in "" `}),a1=nt(function(t,n){const r=Xr("List",t),{children:o,styleType:s="none",stylePosition:a,spacing:u,...d}=xr(t),f=sh(o),m=u?{["& > *:not(style) ~ *:not(style)"]:{mt:u}}:{};return i.jsx(xD,{value:r,children:i.jsx(Be.ul,{ref:n,listStyleType:s,listStylePosition:a,role:"list",__css:{...r.container,...m},...d,children:f})})});a1.displayName="List";var SD=nt((e,t)=>{const{as:n,...r}=e;return i.jsx(a1,{ref:t,as:"ol",styleType:"decimal",marginStart:"1em",...r})});SD.displayName="OrderedList";var mh=nt(function(t,n){const{as:r,...o}=t;return i.jsx(a1,{ref:n,as:"ul",styleType:"initial",marginStart:"1em",...o})});mh.displayName="UnorderedList";var To=nt(function(t,n){const r=g6();return i.jsx(Be.li,{ref:n,...t,__css:r.item})});To.displayName="ListItem";var wD=nt(function(t,n){const r=g6();return i.jsx(mo,{ref:n,role:"presentation",...t,__css:r.icon})});wD.displayName="ListIcon";var ma=nt(function(t,n){const{templateAreas:r,gap:o,rowGap:s,columnGap:a,column:u,row:d,autoFlow:f,autoRows:h,templateRows:m,autoColumns:v,templateColumns:b,...w}=t,C={display:"grid",gridTemplateAreas:r,gridGap:o,gridRowGap:s,gridColumnGap:a,gridAutoColumns:v,gridColumn:u,gridRow:d,gridAutoFlow:f,gridAutoRows:h,gridTemplateRows:m,gridTemplateColumns:b};return i.jsx(Be.div,{ref:n,__css:C,...w})});ma.displayName="Grid";var li=Be("div",{baseStyle:{flex:1,justifySelf:"stretch",alignSelf:"stretch"}});li.displayName="Spacer";var Ue=nt(function(t,n){const r=fc("Text",t),{className:o,align:s,decoration:a,casing:u,...d}=xr(t),f=U0({textAlign:t.align,textDecoration:t.decoration,textTransform:t.casing});return i.jsx(Be.p,{ref:n,className:Ot("chakra-text",t.className),...f,...d,__css:r})});Ue.displayName="Text";var _r=nt((e,t)=>i.jsx(lO,{align:"center",...e,direction:"column",ref:t}));_r.displayName="VStack";function E2(e){return cO(e,t=>t==="auto"?"auto":`span ${t}/span ${t}`)}var Rv=nt(function(t,n){const{area:r,colSpan:o,colStart:s,colEnd:a,rowEnd:u,rowSpan:d,rowStart:f,...h}=t,m=U0({gridArea:r,gridColumn:E2(o),gridRow:E2(d),gridColumnStart:s,gridColumnEnd:a,gridRowStart:f,gridRowEnd:u});return i.jsx(Be.div,{ref:n,__css:m,...h})});Rv.displayName="GridItem";var Xl=nt(function(t,n){const r=fc("Badge",t),{className:o,...s}=xr(t);return i.jsx(Be.span,{ref:n,className:Ot("chakra-badge",t.className),...s,__css:{display:"inline-block",whiteSpace:"nowrap",verticalAlign:"middle",...r}})});Xl.displayName="Badge";var v6=Be("div",{baseStyle:{display:"flex",alignItems:"center",justifyContent:"center"}});v6.displayName="Center";var CD={horizontal:{insetStart:"50%",transform:"translateX(-50%)"},vertical:{top:"50%",transform:"translateY(-50%)"},both:{insetStart:"50%",top:"50%",transform:"translate(-50%, -50%)"}};nt(function(t,n){const{axis:r="both",...o}=t;return i.jsx(Be.div,{ref:n,__css:CD[r],...o,position:"absolute"})});var wa=nt(function(t,n){const{borderLeftWidth:r,borderBottomWidth:o,borderTopWidth:s,borderRightWidth:a,borderWidth:u,borderStyle:d,borderColor:f,...h}=fc("Divider",t),{className:m,orientation:v="horizontal",__css:b,...w}=xr(t),C={vertical:{borderLeftWidth:r||a||u||"1px",height:"100%"},horizontal:{borderBottomWidth:o||s||u||"1px",width:"100%"}};return i.jsx(Be.hr,{ref:n,"aria-orientation":v,...w,__css:{...h,border:"0",borderColor:f,borderStyle:d,...C[v],...b},className:Ot("chakra-divider",m)})});wa.displayName="Divider";function PD(e){const{key:t}=e;return t.length===1||t.length>1&&/[^a-zA-Z0-9]/.test(t)}function kD(e={}){const{timeout:t=300,preventDefault:n=()=>!0}=e,[r,o]=p.useState([]),s=p.useRef(),a=()=>{s.current&&(clearTimeout(s.current),s.current=null)},u=()=>{a(),s.current=setTimeout(()=>{o([]),s.current=null},t)};p.useEffect(()=>a,[]);function d(f){return h=>{if(h.key==="Backspace"){const m=[...r];m.pop(),o(m);return}if(PD(h)){const m=r.concat(h.key);n(h)&&(h.preventDefault(),h.stopPropagation()),o(m),f(m.join("")),u()}}}return d}function _D(e,t,n,r){if(t==null)return r;if(!r)return e.find(a=>n(a).toLowerCase().startsWith(t.toLowerCase()));const o=e.filter(s=>n(s).toLowerCase().startsWith(t.toLowerCase()));if(o.length>0){let s;return o.includes(r)?(s=o.indexOf(r)+1,s===o.length&&(s=0),o[s]):(s=e.indexOf(o[0]),e[s])}return r}function jD(){const e=p.useRef(new Map),t=e.current,n=p.useCallback((o,s,a,u)=>{e.current.set(a,{type:s,el:o,options:u}),o.addEventListener(s,a,u)},[]),r=p.useCallback((o,s,a,u)=>{o.removeEventListener(s,a,u),e.current.delete(a)},[]);return p.useEffect(()=>()=>{t.forEach((o,s)=>{r(o.el,o.type,s,o.options)})},[r,t]),{add:n,remove:r}}function xg(e){const t=e.target,{tagName:n,isContentEditable:r}=t;return n!=="INPUT"&&n!=="TEXTAREA"&&r!==!0}function y6(e={}){const{ref:t,isDisabled:n,isFocusable:r,clickOnEnter:o=!0,clickOnSpace:s=!0,onMouseDown:a,onMouseUp:u,onClick:d,onKeyDown:f,onKeyUp:h,tabIndex:m,onMouseOver:v,onMouseLeave:b,...w}=e,[C,S]=p.useState(!0),[P,k]=p.useState(!1),j=jD(),E=N=>{N&&N.tagName!=="BUTTON"&&S(!1)},I=C?m:m||0,O=n&&!r,R=p.useCallback(N=>{if(n){N.stopPropagation(),N.preventDefault();return}N.currentTarget.focus(),d==null||d(N)},[n,d]),M=p.useCallback(N=>{P&&xg(N)&&(N.preventDefault(),N.stopPropagation(),k(!1),j.remove(document,"keyup",M,!1))},[P,j]),A=p.useCallback(N=>{if(f==null||f(N),n||N.defaultPrevented||N.metaKey||!xg(N.nativeEvent)||C)return;const G=o&&N.key==="Enter";s&&N.key===" "&&(N.preventDefault(),k(!0)),G&&(N.preventDefault(),N.currentTarget.click()),j.add(document,"keyup",M,!1)},[n,C,f,o,s,j,M]),L=p.useCallback(N=>{if(h==null||h(N),n||N.defaultPrevented||N.metaKey||!xg(N.nativeEvent)||C)return;s&&N.key===" "&&(N.preventDefault(),k(!1),N.currentTarget.click())},[s,C,n,h]),D=p.useCallback(N=>{N.button===0&&(k(!1),j.remove(document,"mouseup",D,!1))},[j]),V=p.useCallback(N=>{if(N.button!==0)return;if(n){N.stopPropagation(),N.preventDefault();return}C||k(!0),N.currentTarget.focus({preventScroll:!0}),j.add(document,"mouseup",D,!1),a==null||a(N)},[n,C,a,j,D]),$=p.useCallback(N=>{N.button===0&&(C||k(!1),u==null||u(N))},[u,C]),F=p.useCallback(N=>{if(n){N.preventDefault();return}v==null||v(N)},[n,v]),U=p.useCallback(N=>{P&&(N.preventDefault(),k(!1)),b==null||b(N)},[P,b]),X=Un(t,E);return C?{...w,ref:X,type:"button","aria-disabled":O?void 0:n,disabled:O,onClick:R,onMouseDown:a,onMouseUp:u,onKeyUp:h,onKeyDown:f,onMouseOver:v,onMouseLeave:b}:{...w,ref:X,role:"button","data-active":on(P),"aria-disabled":n?"true":void 0,tabIndex:O?void 0:I,onClick:R,onMouseDown:V,onMouseUp:$,onKeyUp:L,onKeyDown:A,onMouseOver:F,onMouseLeave:U}}function ED(e){const t=e.current;if(!t)return!1;const n=uO(t);return!n||t.contains(n)?!1:!!uD(n)}function b6(e,t){const{shouldFocus:n,visible:r,focusRef:o}=t,s=n&&!r;Zi(()=>{if(!s||ED(e))return;const a=(o==null?void 0:o.current)||e.current;let u;if(a)return u=requestAnimationFrame(()=>{a.focus({preventScroll:!0})}),()=>{cancelAnimationFrame(u)}},[s,e,o])}var ID={preventScroll:!0,shouldFocus:!1};function OD(e,t=ID){const{focusRef:n,preventScroll:r,shouldFocus:o,visible:s}=t,a=RD(e)?e.current:e,u=o&&s,d=p.useRef(u),f=p.useRef(s);su(()=>{!f.current&&s&&(d.current=u),f.current=s},[s,u]);const h=p.useCallback(()=>{if(!(!s||!a||!d.current)&&(d.current=!1,!a.contains(document.activeElement)))if(n!=null&&n.current)requestAnimationFrame(()=>{var m;(m=n.current)==null||m.focus({preventScroll:r})});else{const m=f6(a);m.length>0&&requestAnimationFrame(()=>{m[0].focus({preventScroll:r})})}},[s,r,a,n]);Zi(()=>{h()},[h]),M0(a,"transitionend",h)}function RD(e){return"current"in e}function MD(e){const{ref:t,handler:n,enabled:r=!0}=e,o=qa(n),a=p.useRef({isPointerDown:!1,ignoreEmulatedMouseEvents:!1}).current;p.useEffect(()=>{if(!r)return;const u=m=>{Sg(m,t)&&(a.isPointerDown=!0)},d=m=>{if(a.ignoreEmulatedMouseEvents){a.ignoreEmulatedMouseEvents=!1;return}a.isPointerDown&&n&&Sg(m,t)&&(a.isPointerDown=!1,o(m))},f=m=>{a.ignoreEmulatedMouseEvents=!0,n&&a.isPointerDown&&Sg(m,t)&&(a.isPointerDown=!1,o(m))},h=x6(t.current);return h.addEventListener("mousedown",u,!0),h.addEventListener("mouseup",d,!0),h.addEventListener("touchstart",u,!0),h.addEventListener("touchend",f,!0),()=>{h.removeEventListener("mousedown",u,!0),h.removeEventListener("mouseup",d,!0),h.removeEventListener("touchstart",u,!0),h.removeEventListener("touchend",f,!0)}},[n,t,o,a,r])}function Sg(e,t){var n;const r=e.target;return r&&!x6(r).contains(r)?!1:!((n=t.current)!=null&&n.contains(r))}function x6(e){var t;return(t=e==null?void 0:e.ownerDocument)!=null?t:document}function S6(e){const{isOpen:t,ref:n}=e,[r,o]=p.useState(t),[s,a]=p.useState(!1);return p.useEffect(()=>{s||(o(t),a(!0))},[t,s,r]),M0(()=>n.current,"animationend",()=>{o(t)}),{present:!(t?!1:!r),onComplete(){var d;const f=dO(n.current),h=new f.CustomEvent("animationend",{bubbles:!0});(d=n.current)==null||d.dispatchEvent(h)}}}function i1(e){const{wasSelected:t,enabled:n,isSelected:r,mode:o="unmount"}=e;return!!(!n||r||o==="keepMounted"&&t)}var[TD,DD,AD,ND]=G0(),[zD,Hu]=ur({strict:!1,name:"MenuContext"});function $D(e,...t){const n=p.useId(),r=e||n;return p.useMemo(()=>t.map(o=>`${o}-${r}`),[r,t])}function w6(e){var t;return(t=e==null?void 0:e.ownerDocument)!=null?t:document}function I2(e){return w6(e).activeElement===e}function LD(e={}){const{id:t,closeOnSelect:n=!0,closeOnBlur:r=!0,initialFocusRef:o,autoSelect:s=!0,isLazy:a,isOpen:u,defaultIsOpen:d,onClose:f,onOpen:h,placement:m="bottom-start",lazyBehavior:v="unmount",direction:b,computePositionOnMount:w=!1,...C}=e,S=p.useRef(null),P=p.useRef(null),k=AD(),j=p.useCallback(()=>{requestAnimationFrame(()=>{var le;(le=S.current)==null||le.focus({preventScroll:!1})})},[]),E=p.useCallback(()=>{const le=setTimeout(()=>{var oe;if(o)(oe=o.current)==null||oe.focus();else{const Y=k.firstEnabled();Y&&$(Y.index)}});G.current.add(le)},[k,o]),I=p.useCallback(()=>{const le=setTimeout(()=>{const oe=k.lastEnabled();oe&&$(oe.index)});G.current.add(le)},[k]),O=p.useCallback(()=>{h==null||h(),s?E():j()},[s,E,j,h]),{isOpen:R,onOpen:M,onClose:A,onToggle:L}=o3({isOpen:u,defaultIsOpen:d,onClose:f,onOpen:O});MD({enabled:R&&r,ref:S,handler:le=>{var oe;(oe=P.current)!=null&&oe.contains(le.target)||A()}});const D=s3({...C,enabled:R||w,placement:m,direction:b}),[V,$]=p.useState(-1);Zi(()=>{R||$(-1)},[R]),b6(S,{focusRef:P,visible:R,shouldFocus:!0});const F=S6({isOpen:R,ref:S}),[U,X]=$D(t,"menu-button","menu-list"),N=p.useCallback(()=>{M(),j()},[M,j]),G=p.useRef(new Set([]));qD(()=>{G.current.forEach(le=>clearTimeout(le)),G.current.clear()});const W=p.useCallback(()=>{M(),E()},[E,M]),q=p.useCallback(()=>{M(),I()},[M,I]),ne=p.useCallback(()=>{var le,oe;const Y=w6(S.current),ae=(le=S.current)==null?void 0:le.contains(Y.activeElement);if(!(R&&!ae))return;const xe=(oe=k.item(V))==null?void 0:oe.node;xe==null||xe.focus()},[R,V,k]),ve=p.useRef(null);return{openAndFocusMenu:N,openAndFocusFirstItem:W,openAndFocusLastItem:q,onTransitionEnd:ne,unstable__animationState:F,descendants:k,popper:D,buttonId:U,menuId:X,forceUpdate:D.forceUpdate,orientation:"vertical",isOpen:R,onToggle:L,onOpen:M,onClose:A,menuRef:S,buttonRef:P,focusedIndex:V,closeOnSelect:n,closeOnBlur:r,autoSelect:s,setFocusedIndex:$,isLazy:a,lazyBehavior:v,initialFocusRef:o,rafId:ve}}function FD(e={},t=null){const n=Hu(),{onToggle:r,popper:o,openAndFocusFirstItem:s,openAndFocusLastItem:a}=n,u=p.useCallback(d=>{const f=d.key,m={Enter:s,ArrowDown:s,ArrowUp:a}[f];m&&(d.preventDefault(),d.stopPropagation(),m(d))},[s,a]);return{...e,ref:Un(n.buttonRef,t,o.referenceRef),id:n.buttonId,"data-active":on(n.isOpen),"aria-expanded":n.isOpen,"aria-haspopup":"menu","aria-controls":n.menuId,onClick:bt(e.onClick,r),onKeyDown:bt(e.onKeyDown,u)}}function Mv(e){var t;return UD(e)&&!!((t=e==null?void 0:e.getAttribute("role"))!=null&&t.startsWith("menuitem"))}function BD(e={},t=null){const n=Hu();if(!n)throw new Error("useMenuContext: context is undefined. Seems you forgot to wrap component within