chore(ui): lint

Updated by "Remove blank strings" hook in Weblate.

Co-authored-by: Hosted Weblate <hosted@weblate.org>
Translate-URL: https://hosted.weblate.org/projects/invokeai/web-ui/
Translation: InvokeAI/Web UI

.github/pull_request_template.md

@@ -19,3 +19,4 @@
 - [ ] _The PR has a short but descriptive title, suitable for a changelog_
 - [ ] _Tests added / updated (if applicable)_
 - [ ] _Documentation added / updated (if applicable)_
+- [ ] _Updated `What's New` copy (if doing a release after this PR)_

README.md

@@ -105,7 +105,7 @@ Invoke features an organized gallery system for easily storing, accessing, and r
 ### Other features
 
 - Support for both ckpt and diffusers models
-- SD1.5, SD2.0, and SDXL support
+- SD1.5, SD2.0, SDXL, and FLUX support
 - Upscaling Tools
 - Embedding Manager & Support
 - Model Manager & Support

docker/Dockerfile

@@ -38,9 +38,9 @@ RUN --mount=type=cache,target=/root/.cache/pip \
     if [ "$TARGETPLATFORM" = "linux/arm64" ] || [ "$GPU_DRIVER" = "cpu" ]; then \
         extra_index_url_arg="--extra-index-url https://download.pytorch.org/whl/cpu"; \
     elif [ "$GPU_DRIVER" = "rocm" ]; then \
-        extra_index_url_arg="--extra-index-url https://download.pytorch.org/whl/rocm5.6"; \
+        extra_index_url_arg="--extra-index-url https://download.pytorch.org/whl/rocm6.1"; \
     else \
-        extra_index_url_arg="--extra-index-url https://download.pytorch.org/whl/cu121"; \
+        extra_index_url_arg="--extra-index-url https://download.pytorch.org/whl/cu124"; \
     fi &&\
 
     # xformers + triton fails to install on arm64

docker/docker-compose.yml

@@ -1,7 +1,7 @@
 # Copyright (c) 2023 Eugene Brodsky https://github.com/ebr
 
 x-invokeai: &invokeai
-    image: "local/invokeai:latest"
+    image: "ghcr.io/invoke-ai/invokeai:latest"
     build:
       context: ..
       dockerfile: docker/Dockerfile

docs/contributing/INVOCATIONS.md

@@ -144,7 +144,7 @@ As you might have noticed, we added two new arguments to the `InputField`
 definition for `width` and `height`, called `gt` and `le`. They stand for
 _greater than or equal to_ and _less than or equal to_.
 
-These impose contraints on those fields, and will raise an exception if the
+These impose constraints on those fields, and will raise an exception if the
 values do not meet the constraints. Field constraints are provided by
 **pydantic**, so anything you see in the **pydantic docs** will work.
 

docs/contributing/LOCAL_DEVELOPMENT.md

@@ -239,7 +239,7 @@ Consult the
 get it set up.
 
 Suggest using VSCode's included settings sync so that your remote dev host has
-all the same app settings and extensions automagically.
+all the same app settings and extensions automatically.
 
 ##### One remote dev gotcha
 

docs/contributing/contribution_guides/development.md

@@ -2,7 +2,7 @@
 
 ## **What do I need to know to help?**
 
-If you are looking to help to with a code contribution, InvokeAI uses several different technologies under the hood: Python (Pydantic, FastAPI, diffusers) and Typescript (React, Redux Toolkit, ChakraUI, Mantine, Konva). Familiarity with StableDiffusion and image generation concepts is helpful, but not essential. 
+If you are looking to help with a code contribution, InvokeAI uses several different technologies under the hood: Python (Pydantic, FastAPI, diffusers) and Typescript (React, Redux Toolkit, ChakraUI, Mantine, Konva). Familiarity with StableDiffusion and image generation concepts is helpful, but not essential. 
 
 
 ## **Get Started**

docs/contributing/contribution_guides/newContributorChecklist.md

@@ -5,7 +5,7 @@ If you're a new contributor to InvokeAI or Open Source Projects, this is the gui
 ## New Contributor Checklist
 
 - [x] Set up your local development environment & fork of InvokAI by following [the steps outlined here](../dev-environment.md)
-- [x] Set up your local tooling with [this guide](InvokeAI/contributing/LOCAL_DEVELOPMENT/#developing-invokeai-in-vscode). Feel free to skip this step if you already have tooling you're comfortable with.
+- [x] Set up your local tooling with [this guide](../LOCAL_DEVELOPMENT.md). Feel free to skip this step if you already have tooling you're comfortable with.
 - [x] Familiarize yourself with [Git](https://www.atlassian.com/git) & our project structure by reading through the [development documentation](development.md)
 - [x] Join the [#dev-chat](https://discord.com/channels/1020123559063990373/1049495067846524939) channel of the Discord
 - [x] Choose an issue to work on! This can be achieved by asking in the #dev-chat channel, tackling a [good first issue](https://github.com/invoke-ai/InvokeAI/contribute) or finding an item on the [roadmap](https://github.com/orgs/invoke-ai/projects/7). If nothing in any of those places catches your eye, feel free to work on something of interest to you!

docs/contributing/contribution_guides/tutorials.md

@@ -1,6 +1,6 @@
 # Tutorials
 
-Tutorials help new & existing users expand their abilty to use InvokeAI to the full extent of our features and services.  
+Tutorials help new & existing users expand their ability to use InvokeAI to the full extent of our features and services.  
 
 Currently, we have a set of tutorials available on our [YouTube channel](https://www.youtube.com/@invokeai), but as InvokeAI continues to evolve with new updates, we want to ensure that we are giving our users the resources they need to succeed. 
 
@@ -8,4 +8,4 @@ Tutorials can be in the form of videos or article walkthroughs on a subject of y
 
 ## Contributing
 
-Please reach out to @imic or @hipsterusername on [Discord](https://discord.gg/ZmtBAhwWhy) to help create tutorials for InvokeAI.
+Please reach out to @imic or @hipsterusername on [Discord](https://discord.gg/ZmtBAhwWhy) to help create tutorials for InvokeAI.

docs/contributing/dev-environment.md

@@ -17,46 +17,49 @@ If you just want to use Invoke, you should use the [installer][installer link].
 ## Setup
 
 1. Run through the [requirements][requirements link].
-1. [Fork and clone][forking link] the [InvokeAI repo][repo link].
-1. Create an directory for user data (images, models, db, etc). This is typically at `~/invokeai`, but if you already have a non-dev install, you may want to create a separate directory for the dev install.
-1. Create a python virtual environment inside the directory you just created:
+2. [Fork and clone][forking link] the [InvokeAI repo][repo link].
+3. Create an directory for user data (images, models, db, etc). This is typically at `~/invokeai`, but if you already have a non-dev install, you may want to create a separate directory for the dev install.
+4. Create a python virtual environment inside the directory you just created:
 
-   ```sh
-   python3 -m venv .venv --prompt InvokeAI-Dev
-   ```
+      ```sh
+      python3 -m venv .venv --prompt InvokeAI-Dev
+      ```
 
-1. Activate the venv (you'll need to do this every time you want to run the app):
+5. Activate the venv (you'll need to do this every time you want to run the app):
 
-   ```sh
-   source .venv/bin/activate
-   ```
+        ```sh
+        source .venv/bin/activate
+        ```
 
-1. Install the repo as an [editable install][editable install link]:
+6. Install the repo as an [editable install][editable install link]:
 
-   ```sh
-   pip install -e ".[dev,test,xformers]" --use-pep517 --extra-index-url https://download.pytorch.org/whl/cu121
-   ```
+      ```sh
+      pip install -e ".[dev,test,xformers]" --use-pep517 --extra-index-url https://download.pytorch.org/whl/cu121
+      ```
 
-   Refer to the [manual installation][manual install link]] instructions for more determining the correct install options. `xformers` is optional, but `dev` and `test` are not.
+      Refer to the [manual installation][manual install link]] instructions for more determining the correct install options. `xformers` is optional, but `dev` and `test` are not.
 
-1. Install the frontend dev toolchain:
+7. Install the frontend dev toolchain:
 
    - [`nodejs`](https://nodejs.org/) (recommend v20 LTS)
-   - [`pnpm`](https://pnpm.io/installation#installing-a-specific-version) (must be v8 - not v9!)
+   - [`pnpm`](https://pnpm.io/8.x/installation) (must be v8 - not v9!)
 
-1. Do a production build of the frontend:
+8. Do a production build of the frontend:
 
-   ```sh
-   pnpm build
-   ```
+      ```sh
+      cd PATH_TO_INVOKEAI_REPO/invokeai/frontend/web
+      pnpm i
+      pnpm build
+      ```
 
-1. Start the application:
+9. Start the application:
 
-   ```sh
-   python scripts/invokeai-web.py
-   ```
+      ```sh
+      cd PATH_TO_INVOKEAI_REPO
+      python scripts/invokeai-web.py
+      ```
 
-1. Access the UI at `localhost:9090`.
+10. Access the UI at `localhost:9090`.
 
 ## Updating the UI
 

docs/faq.md

@@ -209,7 +209,7 @@ checkpoint models.
 
 To solve this, go to the Model Manager tab (the cube), select the
 checkpoint model that's giving you trouble, and press the "Convert"
-button in the upper right of your browser window. This will conver the
+button in the upper right of your browser window. This will convert the
 checkpoint into a diffusers model, after which loading should be
 faster and less memory-intensive.
 

docs/installation/patchmatch.md

@@ -97,16 +97,16 @@ Prior to installing PyPatchMatch, you need to take the following steps:
    sudo pacman -S --needed base-devel
    ```
 
-2. Install `opencv` and `blas`:
+2. Install `opencv`, `blas`, and required dependencies:
 
    ```sh
-   sudo pacman -S opencv blas
+   sudo pacman -S opencv blas fmt glew vtk hdf5
    ```
 
    or for CUDA support
 
    ```sh
-   sudo pacman -S opencv-cuda blas
+   sudo pacman -S opencv-cuda blas fmt glew vtk hdf5
    ```
 
 3. Fix the naming of the `opencv` package configuration file:

docs/nodes/communityNodes.md

@@ -21,6 +21,7 @@ To use a community workflow, download the `.json` node graph file and load it in
     + [Clothing Mask](#clothing-mask)
     + [Contrast Limited Adaptive Histogram Equalization](#contrast-limited-adaptive-histogram-equalization)
     + [Depth Map from Wavefront OBJ](#depth-map-from-wavefront-obj)
+    + [Enhance Detail](#enhance-detail)
     + [Film Grain](#film-grain)
     + [Generative Grammar-Based Prompt Nodes](#generative-grammar-based-prompt-nodes)
     + [GPT2RandomPromptMaker](#gpt2randompromptmaker)
@@ -39,7 +40,9 @@ To use a community workflow, download the `.json` node graph file and load it in
     + [Match Histogram](#match-histogram)
     + [Metadata-Linked](#metadata-linked-nodes)
     + [Negative Image](#negative-image)
-    + [Nightmare Promptgen](#nightmare-promptgen)    
+    + [Nightmare Promptgen](#nightmare-promptgen)
+    + [Ollama](#ollama-node)
+    + [One Button Prompt](#one-button-prompt)
     + [Oobabooga](#oobabooga)
     + [Prompt Tools](#prompt-tools)
     + [Remote Image](#remote-image)
@@ -79,7 +82,7 @@ Note: These are inherited from the core nodes so any update to the core nodes sh
 
 **Example Usage:**
 </br>
-<img src="https://github.com/skunkworxdark/autostereogram_nodes/blob/main/images/spider.png" width="200" /> -> <img src="https://github.com/skunkworxdark/autostereogram_nodes/blob/main/images/spider-depth.png" width="200" /> -> <img src="https://github.com/skunkworxdark/autostereogram_nodes/raw/main/images/spider-dots.png" width="200" /> <img src="https://github.com/skunkworxdark/autostereogram_nodes/raw/main/images/spider-pattern.png" width="200" />
+<img src="https://raw.githubusercontent.com/skunkworxdark/autostereogram_nodes/refs/heads/main/images/spider.png" width="200" /> -> <img src="https://raw.githubusercontent.com/skunkworxdark/autostereogram_nodes/refs/heads/main/images/spider-depth.png" width="200" /> -> <img src="https://raw.githubusercontent.com/skunkworxdark/autostereogram_nodes/refs/heads/main/images/spider-dots.png" width="200" /> <img src="https://raw.githubusercontent.com/skunkworxdark/autostereogram_nodes/refs/heads/main/images/spider-pattern.png" width="200" />
 
 --------------------------------
 ### Average Images
@@ -140,6 +143,17 @@ To be imported, an .obj must use triangulated meshes, so make sure to enable tha
 **Example Usage:**
 </br><img src="https://raw.githubusercontent.com/dwringer/depth-from-obj-node/main/depth_from_obj_usage.jpg" width="500" />
 
+--------------------------------
+### Enhance Detail
+
+**Description:** A single node that can enhance the detail in an image. Increase or decrease details in an image using a guided filter (as opposed to the typical Gaussian blur used by most sharpening filters.) Based on the `Enhance Detail` ComfyUI node from  https://github.com/spacepxl/ComfyUI-Image-Filters
+
+**Node Link:** https://github.com/skunkworxdark/enhance-detail-node
+
+**Example Usage:**
+</br>
+<img src="https://raw.githubusercontent.com/skunkworxdark/enhance-detail-node/refs/heads/main/images/Comparison.png" />
+
 --------------------------------
 ### Film Grain
 
@@ -306,7 +320,7 @@ View:
 **Node Link:** https://github.com/helix4u/load_video_frame
 
 **Output Example:** 
-<img src="https://raw.githubusercontent.com/helix4u/load_video_frame/main/_git_assets/testmp4_embed_converted.gif" width="500" />
+<img src="https://raw.githubusercontent.com/helix4u/load_video_frame/refs/heads/main/_git_assets/dance1736978273.gif" width="500" />
 
 --------------------------------
 ### Make 3D
@@ -347,7 +361,7 @@ See full docs here: https://github.com/skunkworxdark/Prompt-tools-nodes/edit/mai
 
 **Output Examples** 
 
-<img src="https://github.com/skunkworxdark/match_histogram/assets/21961335/ed12f329-a0ef-444a-9bae-129ed60d6097" width="300" />
+<img src="https://github.com/skunkworxdark/match_histogram/assets/21961335/ed12f329-a0ef-444a-9bae-129ed60d6097" />
 
 --------------------------------
 ### Metadata Linked Nodes
@@ -389,6 +403,34 @@ View:
 
 **Node Link:** [https://github.com/gogurtenjoyer/nightmare-promptgen](https://github.com/gogurtenjoyer/nightmare-promptgen)
 
+--------------------------------
+### Ollama Node
+
+**Description:** Uses Ollama API to expand text prompts for text-to-image generation using local LLMs. Works great for expanding basic prompts into detailed natural language prompts for Flux. Also provides a toggle to unload the LLM model immediately after expanding, to free up VRAM for Invoke to continue the image generation workflow.
+
+**Node Link:** https://github.com/Jonseed/Ollama-Node
+
+**Example Node Graph:**  https://github.com/Jonseed/Ollama-Node/blob/main/Ollama-Node-Flux-example.json
+
+**View:** 
+
+![ollama node](https://raw.githubusercontent.com/Jonseed/Ollama-Node/a3e7cdc55e394cb89c1ea7ed54e106c212c85e8c/ollama-node-screenshot.png)
+
+--------------------------------
+### One Button Prompt
+
+<img src="https://raw.githubusercontent.com/AIrjen/OneButtonPrompt_X_InvokeAI/refs/heads/main/images/background.png" width="800" />
+
+**Description:** an extensive suite of auto prompt generation and prompt helper nodes based on extensive logic. Get creative with the best prompt generator in the world. 
+
+The main node generates interesting prompts based on a set of parameters. There are also some additional nodes such as Auto Negative Prompt, One Button Artify, Create Prompt Variant and other cool prompt toys to play around with.
+
+**Node Link:** [https://github.com/AIrjen/OneButtonPrompt_X_InvokeAI](https://github.com/AIrjen/OneButtonPrompt_X_InvokeAI)
+
+**Nodes:**
+
+<img src="https://raw.githubusercontent.com/AIrjen/OneButtonPrompt_X_InvokeAI/refs/heads/main/images/OBP_nodes_invokeai.png" width="800" />
+
 --------------------------------
 ### Oobabooga
 
@@ -440,7 +482,7 @@ See full docs here: https://github.com/skunkworxdark/Prompt-tools-nodes/edit/mai
 
 **Workflow Examples** 
 
-<img src="https://github.com/skunkworxdark/prompt-tools/blob/main/images/CSVToIndexStringNode.png" width="300" />
+<img src="https://raw.githubusercontent.com/skunkworxdark/prompt-tools/refs/heads/main/images/CSVToIndexStringNode.png"/>
 
 --------------------------------
 ### Remote Image
@@ -578,7 +620,7 @@ See full docs here: https://github.com/skunkworxdark/XYGrid_nodes/edit/main/READ
 
 **Output Examples** 
 
-<img src="https://github.com/skunkworxdark/XYGrid_nodes/blob/main/images/collage.png" width="300" />
+<img src="https://raw.githubusercontent.com/skunkworxdark/XYGrid_nodes/refs/heads/main/images/collage.png" />
 
 
 --------------------------------

flake.lock

@@ -2,11 +2,11 @@
   "nodes": {
     "nixpkgs": {
       "locked": {
-        "lastModified": 1690630721,
-        "narHash": "sha256-Y04onHyBQT4Erfr2fc82dbJTfXGYrf4V0ysLUYnPOP8=",
+        "lastModified": 1727955264,
+        "narHash": "sha256-lrd+7mmb5NauRoMa8+J1jFKYVa+rc8aq2qc9+CxPDKc=",
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "d2b52322f35597c62abf56de91b0236746b2a03d",
+        "rev": "71cd616696bd199ef18de62524f3df3ffe8b9333",
         "type": "github"
       },
       "original": {

flake.nix

@@ -34,7 +34,7 @@
             cudaPackages.cudnn
             cudaPackages.cuda_nvrtc
             cudatoolkit
-            pkgconfig
+            pkg-config
             libconfig
             cmake
             blas
@@ -66,7 +66,7 @@
             black
 
             # Frontend.
-            yarn
+            pnpm_8
             nodejs
           ];
           LD_LIBRARY_PATH = pkgs.lib.makeLibraryPath buildInputs;

installer/install.sh.in

@@ -12,7 +12,7 @@ MINIMUM_PYTHON_VERSION=3.10.0
 MAXIMUM_PYTHON_VERSION=3.11.100
 PYTHON=""
 for candidate in python3.11 python3.10 python3 python ; do
-    if ppath=`which $candidate`; then
+    if ppath=`which $candidate 2>/dev/null`; then
         # when using `pyenv`, the executable for an inactive Python version will exist but will not be operational
         # we check that this found executable can actually run
         if [ $($candidate --version &>/dev/null; echo ${PIPESTATUS}) -gt 0 ]; then continue; fi
@@ -30,10 +30,11 @@ done
 if [ -z "$PYTHON" ]; then
     echo "A suitable Python interpreter could not be found"
     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
 
+echo "For the best user experience we suggest enlarging or maximizing this window now."
+
 exec $PYTHON ./lib/main.py ${@}
 read -p "Press any key to exit"

installer/lib/installer.py

@@ -245,6 +245,9 @@ class InvokeAiInstance:
 
         pip = local[self.pip]
 
+        # Uninstall xformers if it is present; the correct version of it will be reinstalled if needed
+        _ = pip["uninstall", "-yqq", "xformers"] & FG
+
         pipeline = pip[
             "install",
             "--require-virtualenv",
@@ -282,12 +285,6 @@ class InvokeAiInstance:
             shutil.copy(src, dest)
             os.chmod(dest, 0o0755)
 
-    def update(self):
-        pass
-
-    def remove(self):
-        pass
-
 
 ### Utility functions ###
 
@@ -402,7 +399,7 @@ def get_torch_source() -> Tuple[str | None, str | None]:
     :rtype: list
     """
 
-    from messages import select_gpu
+    from messages import GpuType, select_gpu
 
     # device can be one of: "cuda", "rocm", "cpu", "cuda_and_dml, autodetect"
     device = select_gpu()
@@ -412,16 +409,22 @@ def get_torch_source() -> Tuple[str | None, str | None]:
     url = None
     optional_modules: str | None = None
     if OS == "Linux":
-        if device.value == "rocm":
-            url = "https://download.pytorch.org/whl/rocm5.6"
-        elif device.value == "cpu":
+        if device == GpuType.ROCM:
+            url = "https://download.pytorch.org/whl/rocm6.1"
+        elif device == GpuType.CPU:
             url = "https://download.pytorch.org/whl/cpu"
-        elif device.value == "cuda":
-            # CUDA uses the default PyPi index
+        elif device == GpuType.CUDA:
+            url = "https://download.pytorch.org/whl/cu124"
+            optional_modules = "[onnx-cuda]"
+        elif device == GpuType.CUDA_WITH_XFORMERS:
+            url = "https://download.pytorch.org/whl/cu124"
             optional_modules = "[xformers,onnx-cuda]"
     elif OS == "Windows":
-        if device.value == "cuda":
-            url = "https://download.pytorch.org/whl/cu121"
+        if device == GpuType.CUDA:
+            url = "https://download.pytorch.org/whl/cu124"
+            optional_modules = "[onnx-cuda]"
+        elif device == GpuType.CUDA_WITH_XFORMERS:
+            url = "https://download.pytorch.org/whl/cu124"
             optional_modules = "[xformers,onnx-cuda]"
         elif device.value == "cpu":
             # CPU  uses the default PyPi index, no optional modules

installer/lib/messages.py

@@ -206,6 +206,7 @@ def dest_path(dest: Optional[str | Path] = None) -> Path | None:
 
 
 class GpuType(Enum):
+    CUDA_WITH_XFORMERS = "xformers"
     CUDA = "cuda"
     ROCM = "rocm"
     CPU = "cpu"
@@ -221,11 +222,15 @@ def select_gpu() -> GpuType:
         return GpuType.CPU
 
     nvidia = (
-        "an [gold1 b]NVIDIA[/] GPU (using CUDA™)",
+        "an [gold1 b]NVIDIA[/] RTX 3060 or newer GPU using CUDA",
         GpuType.CUDA,
     )
+    vintage_nvidia = (
+        "an [gold1 b]NVIDIA[/] RTX 20xx or older GPU using CUDA+xFormers",
+        GpuType.CUDA_WITH_XFORMERS,
+    )
     amd = (
-        "an [gold1 b]AMD[/] GPU (using ROCm™)",
+        "an [gold1 b]AMD[/] GPU using ROCm",
         GpuType.ROCM,
     )
     cpu = (
@@ -235,14 +240,13 @@ def select_gpu() -> GpuType:
 
     options = []
     if OS == "Windows":
-        options = [nvidia, cpu]
+        options = [nvidia, vintage_nvidia, cpu]
     if OS == "Linux":
-        options = [nvidia, amd, cpu]
+        options = [nvidia, vintage_nvidia, amd, cpu]
     elif OS == "Darwin":
         options = [cpu]
 
     if len(options) == 1:
-        print(f'Your platform [gold1]{OS}-{ARCH}[/] only supports the "{options[0][1]}" driver. Proceeding with that.')
         return options[0][1]
 
     options = {str(i): opt for i, opt in enumerate(options, 1)}
@@ -255,7 +259,7 @@ def select_gpu() -> GpuType:
                     [
                         f"Detected the [gold1]{OS}-{ARCH}[/] platform",
                         "",
-                        "See [deep_sky_blue1]https://invoke-ai.github.io/InvokeAI/#system[/] to ensure your system meets the minimum requirements.",
+                        "See [deep_sky_blue1]https://invoke-ai.github.io/InvokeAI/installation/requirements/[/] to ensure your system meets the minimum requirements.",
                         "",
                         "[red3]🠶[/] [b]Your GPU drivers must be correctly installed before using InvokeAI![/] [red3]🠴[/]",
                     ]

installer/templates/invoke.sh.in

@@ -68,7 +68,7 @@ do_line_input() {
     printf "2: Open the developer console\n"
     printf "3: Command-line help\n"
     printf "Q: Quit\n\n"
-    printf "To update, download and run the installer from https://github.com/invoke-ai/InvokeAI/releases/latest.\n\n"
+    printf "To update, download and run the installer from https://github.com/invoke-ai/InvokeAI/releases/latest\n\n"
     read -p "Please enter 1-4, Q: [1] " yn
     choice=${yn:='1'}
     do_choice $choice

invokeai/app/api/routers/app_info.py

@@ -40,6 +40,8 @@ class AppVersion(BaseModel):
 
     version: str = Field(description="App version")
 
+    highlights: Optional[list[str]] = Field(default=None, description="Highlights of release")
+
 
 class AppDependencyVersions(BaseModel):
     """App depencency Versions Response"""

invokeai/app/api/routers/boards.py

@@ -5,9 +5,10 @@ from fastapi.routing import APIRouter
 from pydantic import BaseModel, Field
 
 from invokeai.app.api.dependencies import ApiDependencies
-from invokeai.app.services.board_records.board_records_common import BoardChanges
+from invokeai.app.services.board_records.board_records_common import BoardChanges, BoardRecordOrderBy
 from invokeai.app.services.boards.boards_common import BoardDTO
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 
 boards_router = APIRouter(prefix="/v1/boards", tags=["boards"])
 
@@ -115,6 +116,8 @@ async def delete_board(
     response_model=Union[OffsetPaginatedResults[BoardDTO], list[BoardDTO]],
 )
 async def list_boards(
+    order_by: BoardRecordOrderBy = Query(default=BoardRecordOrderBy.CreatedAt, description="The attribute to order by"),
+    direction: SQLiteDirection = Query(default=SQLiteDirection.Descending, description="The direction to order by"),
     all: Optional[bool] = Query(default=None, description="Whether to list all boards"),
     offset: Optional[int] = Query(default=None, description="The page offset"),
     limit: Optional[int] = Query(default=None, description="The number of boards per page"),
@@ -122,9 +125,9 @@ async def list_boards(
 ) -> Union[OffsetPaginatedResults[BoardDTO], list[BoardDTO]]:
     """Gets a list of boards"""
     if all:
-        return ApiDependencies.invoker.services.boards.get_all(include_archived)
+        return ApiDependencies.invoker.services.boards.get_all(order_by, direction, include_archived)
     elif offset is not None and limit is not None:
-        return ApiDependencies.invoker.services.boards.get_many(offset, limit, include_archived)
+        return ApiDependencies.invoker.services.boards.get_many(order_by, direction, offset, limit, include_archived)
     else:
         raise HTTPException(
             status_code=400,

invokeai/app/api/routers/model_manager.py

@@ -1,6 +1,7 @@
 # Copyright (c) 2023 Lincoln D. Stein
 """FastAPI route for model configuration records."""
 
+import contextlib
 import io
 import pathlib
 import shutil
@@ -10,6 +11,7 @@ from enum import Enum
 from tempfile import TemporaryDirectory
 from typing import List, Optional, Type
 
+import huggingface_hub
 from fastapi import Body, Path, Query, Response, UploadFile
 from fastapi.responses import FileResponse, HTMLResponse
 from fastapi.routing import APIRouter
@@ -27,6 +29,7 @@ from invokeai.app.services.model_records import (
     ModelRecordChanges,
     UnknownModelException,
 )
+from invokeai.app.util.suppress_output import SuppressOutput
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     BaseModelType,
@@ -38,7 +41,12 @@ from invokeai.backend.model_manager.load.model_cache.model_cache_base import Cac
 from invokeai.backend.model_manager.metadata.fetch.huggingface import HuggingFaceMetadataFetch
 from invokeai.backend.model_manager.metadata.metadata_base import ModelMetadataWithFiles, UnknownMetadataException
 from invokeai.backend.model_manager.search import ModelSearch
-from invokeai.backend.model_manager.starter_models import STARTER_MODELS, StarterModel, StarterModelWithoutDependencies
+from invokeai.backend.model_manager.starter_models import (
+    STARTER_BUNDLES,
+    STARTER_MODELS,
+    StarterModel,
+    StarterModelWithoutDependencies,
+)
 
 model_manager_router = APIRouter(prefix="/v2/models", tags=["model_manager"])
 
@@ -792,22 +800,52 @@ async def convert_model(
     return new_config
 
 
-@model_manager_router.get("/starter_models", operation_id="get_starter_models", response_model=list[StarterModel])
-async def get_starter_models() -> list[StarterModel]:
+class StarterModelResponse(BaseModel):
+    starter_models: list[StarterModel]
+    starter_bundles: dict[str, list[StarterModel]]
+
+
+def get_is_installed(
+    starter_model: StarterModel | StarterModelWithoutDependencies, installed_models: list[AnyModelConfig]
+) -> bool:
+    for model in installed_models:
+        if model.source == starter_model.source:
+            return True
+        if (
+            (model.name == starter_model.name or model.name in starter_model.previous_names)
+            and model.base == starter_model.base
+            and model.type == starter_model.type
+        ):
+            return True
+    return False
+
+
+@model_manager_router.get("/starter_models", operation_id="get_starter_models", response_model=StarterModelResponse)
+async def get_starter_models() -> StarterModelResponse:
     installed_models = ApiDependencies.invoker.services.model_manager.store.search_by_attr()
-    installed_model_sources = {m.source for m in installed_models}
     starter_models = deepcopy(STARTER_MODELS)
+    starter_bundles = deepcopy(STARTER_BUNDLES)
     for model in starter_models:
-        if model.source in installed_model_sources:
-            model.is_installed = True
+        model.is_installed = get_is_installed(model, installed_models)
         # Remove already-installed dependencies
         missing_deps: list[StarterModelWithoutDependencies] = []
+
         for dep in model.dependencies or []:
-            if dep.source not in installed_model_sources:
+            if not get_is_installed(dep, installed_models):
                 missing_deps.append(dep)
         model.dependencies = missing_deps
 
-    return starter_models
+    for bundle in starter_bundles.values():
+        for model in bundle:
+            model.is_installed = get_is_installed(model, installed_models)
+            # Remove already-installed dependencies
+            missing_deps: list[StarterModelWithoutDependencies] = []
+            for dep in model.dependencies or []:
+                if not get_is_installed(dep, installed_models):
+                    missing_deps.append(dep)
+            model.dependencies = missing_deps
+
+    return StarterModelResponse(starter_models=starter_models, starter_bundles=starter_bundles)
 
 
 @model_manager_router.get(
@@ -888,3 +926,51 @@ async def get_stats() -> Optional[CacheStats]:
     """Return performance statistics on the model manager's RAM cache. Will return null if no models have been loaded."""
 
     return ApiDependencies.invoker.services.model_manager.load.ram_cache.stats
+
+
+class HFTokenStatus(str, Enum):
+    VALID = "valid"
+    INVALID = "invalid"
+    UNKNOWN = "unknown"
+
+
+class HFTokenHelper:
+    @classmethod
+    def get_status(cls) -> HFTokenStatus:
+        try:
+            if huggingface_hub.get_token_permission(huggingface_hub.get_token()):
+                # Valid token!
+                return HFTokenStatus.VALID
+            # No token set
+            return HFTokenStatus.INVALID
+        except Exception:
+            return HFTokenStatus.UNKNOWN
+
+    @classmethod
+    def set_token(cls, token: str) -> HFTokenStatus:
+        with SuppressOutput(), contextlib.suppress(Exception):
+            huggingface_hub.login(token=token, add_to_git_credential=False)
+        return cls.get_status()
+
+
+@model_manager_router.get("/hf_login", operation_id="get_hf_login_status", response_model=HFTokenStatus)
+async def get_hf_login_status() -> HFTokenStatus:
+    token_status = HFTokenHelper.get_status()
+
+    if token_status is HFTokenStatus.UNKNOWN:
+        ApiDependencies.invoker.services.logger.warning("Unable to verify HF token")
+
+    return token_status
+
+
+@model_manager_router.post("/hf_login", operation_id="do_hf_login", response_model=HFTokenStatus)
+async def do_hf_login(
+    token: str = Body(description="Hugging Face token to use for login", embed=True),
+) -> HFTokenStatus:
+    HFTokenHelper.set_token(token)
+    token_status = HFTokenHelper.get_status()
+
+    if token_status is HFTokenStatus.UNKNOWN:
+        ApiDependencies.invoker.services.logger.warning("Unable to verify HF token")
+
+    return token_status

invokeai/app/api/routers/workflows.py

@@ -83,7 +83,7 @@ async def create_workflow(
 )
 async def list_workflows(
     page: int = Query(default=0, description="The page to get"),
-    per_page: int = Query(default=10, description="The number of workflows per page"),
+    per_page: Optional[int] = Query(default=None, description="The number of workflows per page"),
     order_by: WorkflowRecordOrderBy = Query(
         default=WorkflowRecordOrderBy.Name, description="The attribute to order by"
     ),
@@ -93,5 +93,5 @@ async def list_workflows(
 ) -> PaginatedResults[WorkflowRecordListItemDTO]:
     """Gets a page of workflows"""
     return ApiDependencies.invoker.services.workflow_records.get_many(
-        page=page, per_page=per_page, order_by=order_by, direction=direction, query=query, category=category
+        order_by=order_by, direction=direction, page=page, per_page=per_page, query=query, category=category
     )

invokeai/app/api_app.py

@@ -7,13 +7,14 @@ from pathlib import Path
 
 import torch
 import uvicorn
-from fastapi import FastAPI
+from fastapi import FastAPI, Request
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.middleware.gzip import GZipMiddleware
 from fastapi.openapi.docs import get_redoc_html, get_swagger_ui_html
-from fastapi.responses import HTMLResponse
+from fastapi.responses import HTMLResponse, RedirectResponse
 from fastapi_events.handlers.local import local_handler
 from fastapi_events.middleware import EventHandlerASGIMiddleware
+from starlette.middleware.base import BaseHTTPMiddleware, RequestResponseEndpoint
 from torch.backends.mps import is_available as is_mps_available
 
 # for PyCharm:
@@ -78,6 +79,29 @@ app = FastAPI(
     lifespan=lifespan,
 )
 
+
+class RedirectRootWithQueryStringMiddleware(BaseHTTPMiddleware):
+    """When a request is made to the root path with a query string, redirect to the root path without the query string.
+
+    For example, to force a Gradio app to use dark mode, users may append `?__theme=dark` to the URL. Their browser may
+    have this query string saved in history or a bookmark, so when the user navigates to `http://127.0.0.1:9090/`, the
+    browser takes them to `http://127.0.0.1:9090/?__theme=dark`.
+
+    This breaks the static file serving in the UI, so we redirect the user to the root path without the query string.
+    """
+
+    async def dispatch(self, request: Request, call_next: RequestResponseEndpoint):
+        if request.url.path == "/" and request.url.query:
+            return RedirectResponse(url="/")
+
+        response = await call_next(request)
+        return response
+
+
+# Add the middleware
+app.add_middleware(RedirectRootWithQueryStringMiddleware)
+
+
 # Add event handler
 event_handler_id: int = id(app)
 app.add_middleware(

invokeai/app/invocations/baseinvocation.py

@@ -4,6 +4,7 @@ from __future__ import annotations
 
 import inspect
 import re
+import sys
 import warnings
 from abc import ABC, abstractmethod
 from enum import Enum
@@ -192,12 +193,19 @@ class BaseInvocation(ABC, BaseModel):
         """Gets a pydantc TypeAdapter for the union of all invocation types."""
         if not cls._typeadapter or cls._typeadapter_needs_update:
             AnyInvocation = TypeAliasType(
-                "AnyInvocation", Annotated[Union[tuple(cls._invocation_classes)], Field(discriminator="type")]
+                "AnyInvocation", Annotated[Union[tuple(cls.get_invocations())], Field(discriminator="type")]
             )
             cls._typeadapter = TypeAdapter(AnyInvocation)
             cls._typeadapter_needs_update = False
         return cls._typeadapter
 
+    @classmethod
+    def invalidate_typeadapter(cls) -> None:
+        """Invalidates the typeadapter, forcing it to be rebuilt on next access. If the invocation allowlist or
+        denylist is changed, this should be called to ensure the typeadapter is updated and validation respects
+        the updated allowlist and denylist."""
+        cls._typeadapter_needs_update = True
+
     @classmethod
     def get_invocations(cls) -> Iterable[BaseInvocation]:
         """Gets all invocations, respecting the allowlist and denylist."""
@@ -479,6 +487,26 @@ def invocation(
             title="type", default=invocation_type, json_schema_extra={"field_kind": FieldKind.NodeAttribute}
         )
 
+        # Validate the `invoke()` method is implemented
+        if "invoke" in cls.__abstractmethods__:
+            raise ValueError(f'Invocation "{invocation_type}" must implement the "invoke" method')
+
+        # And validate that `invoke()` returns a subclass of `BaseInvocationOutput
+        invoke_return_annotation = signature(cls.invoke).return_annotation
+
+        try:
+            # TODO(psyche): If `invoke()` is not defined, `return_annotation` ends up as the string "BaseInvocationOutput"
+            # instead of the class `BaseInvocationOutput`. This may be a pydantic bug: https://github.com/pydantic/pydantic/issues/7978
+            if isinstance(invoke_return_annotation, str):
+                invoke_return_annotation = getattr(sys.modules[cls.__module__], invoke_return_annotation)
+
+            assert invoke_return_annotation is not BaseInvocationOutput
+            assert issubclass(invoke_return_annotation, BaseInvocationOutput)
+        except Exception:
+            raise ValueError(
+                f'Invocation "{invocation_type}" must have a return annotation of a subclass of BaseInvocationOutput (got "{invoke_return_annotation}")'
+            )
+
         docstring = cls.__doc__
         cls = create_model(
             cls.__qualname__,

invokeai/app/invocations/compel.py

@@ -95,6 +95,7 @@ class CompelInvocation(BaseInvocation):
                 ti_manager,
             ),
         ):
+            context.util.signal_progress("Building conditioning")
             assert isinstance(text_encoder, CLIPTextModel)
             assert isinstance(tokenizer, CLIPTokenizer)
             compel = Compel(
@@ -191,6 +192,7 @@ class SDXLPromptInvocationBase:
                 ti_manager,
             ),
         ):
+            context.util.signal_progress("Building conditioning")
             assert isinstance(text_encoder, (CLIPTextModel, CLIPTextModelWithProjection))
             assert isinstance(tokenizer, CLIPTokenizer)
 

invokeai/app/invocations/create_denoise_mask.py

@@ -65,6 +65,7 @@ class CreateDenoiseMaskInvocation(BaseInvocation):
             img_mask = tv_resize(mask, image_tensor.shape[-2:], T.InterpolationMode.BILINEAR, antialias=False)
             masked_image = image_tensor * torch.where(img_mask < 0.5, 0.0, 1.0)
             # TODO:
+            context.util.signal_progress("Running VAE encoder")
             masked_latents = ImageToLatentsInvocation.vae_encode(vae_info, self.fp32, self.tiled, masked_image.clone())
 
             masked_latents_name = context.tensors.save(tensor=masked_latents)

invokeai/app/invocations/create_gradient_mask.py

@@ -131,6 +131,7 @@ class CreateGradientMaskInvocation(BaseInvocation):
                     image_tensor = image_tensor.unsqueeze(0)
                 img_mask = tv_resize(mask, image_tensor.shape[-2:], T.InterpolationMode.BILINEAR, antialias=False)
                 masked_image = image_tensor * torch.where(img_mask < 0.5, 0.0, 1.0)
+                context.util.signal_progress("Running VAE encoder")
                 masked_latents = ImageToLatentsInvocation.vae_encode(
                     vae_info, self.fp32, self.tiled, masked_image.clone()
                 )

invokeai/app/invocations/denoise_latents.py

@@ -13,6 +13,7 @@ from diffusers.models.unets.unet_2d_condition import UNet2DConditionModel
 from diffusers.schedulers.scheduling_dpmsolver_sde import DPMSolverSDEScheduler
 from diffusers.schedulers.scheduling_tcd import TCDScheduler
 from diffusers.schedulers.scheduling_utils import SchedulerMixin as Scheduler
+from PIL import Image
 from pydantic import field_validator
 from torchvision.transforms.functional import resize as tv_resize
 from transformers import CLIPVisionModelWithProjection
@@ -510,6 +511,7 @@ class DenoiseLatentsInvocation(BaseInvocation):
         context: InvocationContext,
         t2i_adapters: Optional[Union[T2IAdapterField, list[T2IAdapterField]]],
         ext_manager: ExtensionsManager,
+        bgr_mode: bool = False,
     ) -> None:
         if t2i_adapters is None:
             return
@@ -519,6 +521,10 @@ class DenoiseLatentsInvocation(BaseInvocation):
             t2i_adapters = [t2i_adapters]
 
         for t2i_adapter_field in t2i_adapters:
+            image = context.images.get_pil(t2i_adapter_field.image.image_name)
+            if bgr_mode:  # SDXL t2i trained on cv2's BGR outputs, but PIL won't convert straight to BGR
+                r, g, b = image.split()
+                image = Image.merge("RGB", (b, g, r))
             ext_manager.add_extension(
                 T2IAdapterExt(
                     node_context=context,
@@ -547,7 +553,9 @@ class DenoiseLatentsInvocation(BaseInvocation):
                 if not isinstance(single_ipa_image_fields, list):
                     single_ipa_image_fields = [single_ipa_image_fields]
 
-                single_ipa_images = [context.images.get_pil(image.image_name) for image in single_ipa_image_fields]
+                single_ipa_images = [
+                    context.images.get_pil(image.image_name, mode="RGB") for image in single_ipa_image_fields
+                ]
                 with image_encoder_model_info as image_encoder_model:
                     assert isinstance(image_encoder_model, CLIPVisionModelWithProjection)
                     # Get image embeddings from CLIP and ImageProjModel.
@@ -614,13 +622,17 @@ class DenoiseLatentsInvocation(BaseInvocation):
         for t2i_adapter_field in t2i_adapter:
             t2i_adapter_model_config = context.models.get_config(t2i_adapter_field.t2i_adapter_model.key)
             t2i_adapter_loaded_model = context.models.load(t2i_adapter_field.t2i_adapter_model)
-            image = context.images.get_pil(t2i_adapter_field.image.image_name)
+            image = context.images.get_pil(t2i_adapter_field.image.image_name, mode="RGB")
 
             # The max_unet_downscale is the maximum amount that the UNet model downscales the latent image internally.
             if t2i_adapter_model_config.base == BaseModelType.StableDiffusion1:
                 max_unet_downscale = 8
             elif t2i_adapter_model_config.base == BaseModelType.StableDiffusionXL:
                 max_unet_downscale = 4
+
+                # SDXL adapters are trained on cv2's BGR outputs
+                r, g, b = image.split()
+                image = Image.merge("RGB", (b, g, r))
             else:
                 raise ValueError(f"Unexpected T2I-Adapter base model type: '{t2i_adapter_model_config.base}'.")
 
@@ -628,29 +640,39 @@ class DenoiseLatentsInvocation(BaseInvocation):
             with t2i_adapter_loaded_model as t2i_adapter_model:
                 total_downscale_factor = t2i_adapter_model.total_downscale_factor
 
-                # Resize the T2I-Adapter input image.
-                # We select the resize dimensions so that after the T2I-Adapter's total_downscale_factor is applied, the
-                # result will match the latent image's dimensions after max_unet_downscale is applied.
-                t2i_input_height = latents_shape[2] // max_unet_downscale * total_downscale_factor
-                t2i_input_width = latents_shape[3] // max_unet_downscale * total_downscale_factor
-
                 # Note: We have hard-coded `do_classifier_free_guidance=False`. This is because we only want to prepare
                 # a single image. If CFG is enabled, we will duplicate the resultant tensor after applying the
                 # T2I-Adapter model.
                 #
                 # Note: We re-use the `prepare_control_image(...)` from ControlNet for T2I-Adapter, because it has many
                 # of the same requirements (e.g. preserving binary masks during resize).
+
+                # Assuming fixed dimensional scaling of LATENT_SCALE_FACTOR.
+                _, _, latent_height, latent_width = latents_shape
+                control_height_resize = latent_height * LATENT_SCALE_FACTOR
+                control_width_resize = latent_width * LATENT_SCALE_FACTOR
                 t2i_image = prepare_control_image(
                     image=image,
                     do_classifier_free_guidance=False,
-                    width=t2i_input_width,
-                    height=t2i_input_height,
+                    width=control_width_resize,
+                    height=control_height_resize,
                     num_channels=t2i_adapter_model.config["in_channels"],  # mypy treats this as a FrozenDict
                     device=t2i_adapter_model.device,
                     dtype=t2i_adapter_model.dtype,
                     resize_mode=t2i_adapter_field.resize_mode,
                 )
 
+                # Resize the T2I-Adapter input image.
+                # We select the resize dimensions so that after the T2I-Adapter's total_downscale_factor is applied, the
+                # result will match the latent image's dimensions after max_unet_downscale is applied.
+                # We crop the image to this size so that the positions match the input image on non-standard resolutions
+                t2i_input_height = latents_shape[2] // max_unet_downscale * total_downscale_factor
+                t2i_input_width = latents_shape[3] // max_unet_downscale * total_downscale_factor
+                if t2i_image.shape[2] > t2i_input_height or t2i_image.shape[3] > t2i_input_width:
+                    t2i_image = t2i_image[
+                        :, :, : min(t2i_image.shape[2], t2i_input_height), : min(t2i_image.shape[3], t2i_input_width)
+                    ]
+
                 adapter_state = t2i_adapter_model(t2i_image)
 
             if do_classifier_free_guidance:
@@ -898,7 +920,8 @@ class DenoiseLatentsInvocation(BaseInvocation):
             #    ext = extension_field.to_extension(exit_stack, context, ext_manager)
             #    ext_manager.add_extension(ext)
             self.parse_controlnet_field(exit_stack, context, self.control, ext_manager)
-            self.parse_t2i_adapter_field(exit_stack, context, self.t2i_adapter, ext_manager)
+            bgr_mode = self.unet.unet.base == BaseModelType.StableDiffusionXL
+            self.parse_t2i_adapter_field(exit_stack, context, self.t2i_adapter, ext_manager, bgr_mode)
 
             # ext: t2i/ip adapter
             ext_manager.run_callback(ExtensionCallbackType.SETUP, denoise_ctx)

invokeai/app/invocations/fields.py

@@ -41,6 +41,7 @@ class UIType(str, Enum, metaclass=MetaEnum):
     # region Model Field Types
     MainModel = "MainModelField"
     FluxMainModel = "FluxMainModelField"
+    SD3MainModel = "SD3MainModelField"
     SDXLMainModel = "SDXLMainModelField"
     SDXLRefinerModel = "SDXLRefinerModelField"
     ONNXModel = "ONNXModelField"
@@ -52,6 +53,8 @@ class UIType(str, Enum, metaclass=MetaEnum):
     T2IAdapterModel = "T2IAdapterModelField"
     T5EncoderModel = "T5EncoderModelField"
     CLIPEmbedModel = "CLIPEmbedModelField"
+    CLIPLEmbedModel = "CLIPLEmbedModelField"
+    CLIPGEmbedModel = "CLIPGEmbedModelField"
     SpandrelImageToImageModel = "SpandrelImageToImageModelField"
     # endregion
 
@@ -131,8 +134,10 @@ class FieldDescriptions:
     clip = "CLIP (tokenizer, text encoder, LoRAs) and skipped layer count"
     t5_encoder = "T5 tokenizer and text encoder"
     clip_embed_model = "CLIP Embed loader"
+    clip_g_model = "CLIP-G Embed loader"
     unet = "UNet (scheduler, LoRAs)"
     transformer = "Transformer"
+    mmditx = "MMDiTX"
     vae = "VAE"
     cond = "Conditioning tensor"
     controlnet_model = "ControlNet model to load"
@@ -140,6 +145,7 @@ class FieldDescriptions:
     lora_model = "LoRA model to load"
     main_model = "Main model (UNet, VAE, CLIP) to load"
     flux_model = "Flux model (Transformer) to load"
+    sd3_model = "SD3 model (MMDiTX) to load"
     sdxl_main_model = "SDXL Main model (UNet, VAE, CLIP1, CLIP2) to load"
     sdxl_refiner_model = "SDXL Refiner Main Modde (UNet, VAE, CLIP2) to load"
     onnx_main_model = "ONNX Main model (UNet, VAE, CLIP) to load"
@@ -192,6 +198,7 @@ class FieldDescriptions:
     freeu_s2 = 'Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to mitigate the "oversmoothing effect" in the enhanced denoising process.'
     freeu_b1 = "Scaling factor for stage 1 to amplify the contributions of backbone features."
     freeu_b2 = "Scaling factor for stage 2 to amplify the contributions of backbone features."
+    instantx_control_mode = "The control mode for InstantX ControlNet union models. Ignored for other ControlNet models. The standard mapping is: canny (0), tile (1), depth (2), blur (3), pose (4), gray (5), low quality (6). Negative values will be treated as 'None'."
 
 
 class ImageField(BaseModel):
@@ -245,6 +252,12 @@ class FluxConditioningField(BaseModel):
     conditioning_name: str = Field(description="The name of conditioning tensor")
 
 
+class SD3ConditioningField(BaseModel):
+    """A conditioning tensor primitive value"""
+
+    conditioning_name: str = Field(description="The name of conditioning tensor")
+
+
 class ConditioningField(BaseModel):
     """A conditioning tensor primitive value"""
 

invokeai/app/invocations/flux_controlnet.py

@@ -0,0 +1,99 @@
+from pydantic import BaseModel, Field, field_validator, model_validator
+
+from invokeai.app.invocations.baseinvocation import (
+    BaseInvocation,
+    BaseInvocationOutput,
+    Classification,
+    invocation,
+    invocation_output,
+)
+from invokeai.app.invocations.fields import FieldDescriptions, ImageField, InputField, OutputField, UIType
+from invokeai.app.invocations.model import ModelIdentifierField
+from invokeai.app.invocations.util import validate_begin_end_step, validate_weights
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.app.util.controlnet_utils import CONTROLNET_RESIZE_VALUES
+
+
+class FluxControlNetField(BaseModel):
+    image: ImageField = Field(description="The control image")
+    control_model: ModelIdentifierField = Field(description="The ControlNet model to use")
+    control_weight: float | list[float] = Field(default=1, description="The weight given to the ControlNet")
+    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)"
+    )
+    resize_mode: CONTROLNET_RESIZE_VALUES = Field(default="just_resize", description="The resize mode to use")
+    instantx_control_mode: int | None = Field(default=-1, description=FieldDescriptions.instantx_control_mode)
+
+    @field_validator("control_weight")
+    @classmethod
+    def validate_control_weight(cls, v: float | list[float]) -> float | list[float]:
+        validate_weights(v)
+        return v
+
+    @model_validator(mode="after")
+    def validate_begin_end_step_percent(self):
+        validate_begin_end_step(self.begin_step_percent, self.end_step_percent)
+        return self
+
+
+@invocation_output("flux_controlnet_output")
+class FluxControlNetOutput(BaseInvocationOutput):
+    """FLUX ControlNet info"""
+
+    control: FluxControlNetField = OutputField(description=FieldDescriptions.control)
+
+
+@invocation(
+    "flux_controlnet",
+    title="FLUX ControlNet",
+    tags=["controlnet", "flux"],
+    category="controlnet",
+    version="1.0.0",
+    classification=Classification.Prototype,
+)
+class FluxControlNetInvocation(BaseInvocation):
+    """Collect FLUX ControlNet info to pass to other nodes."""
+
+    image: ImageField = InputField(description="The control image")
+    control_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.controlnet_model, ui_type=UIType.ControlNetModel
+    )
+    control_weight: float | list[float] = InputField(
+        default=1.0, ge=-1, le=2, description="The weight given to the ControlNet"
+    )
+    begin_step_percent: float = InputField(
+        default=0, ge=0, le=1, description="When the ControlNet is first applied (% of total steps)"
+    )
+    end_step_percent: float = InputField(
+        default=1, ge=0, le=1, description="When the ControlNet is last applied (% of total steps)"
+    )
+    resize_mode: CONTROLNET_RESIZE_VALUES = InputField(default="just_resize", description="The resize mode used")
+    # Note: We default to -1 instead of None, because in the workflow editor UI None is not currently supported.
+    instantx_control_mode: int | None = InputField(default=-1, description=FieldDescriptions.instantx_control_mode)
+
+    @field_validator("control_weight")
+    @classmethod
+    def validate_control_weight(cls, v: float | list[float]) -> float | list[float]:
+        validate_weights(v)
+        return v
+
+    @model_validator(mode="after")
+    def validate_begin_end_step_percent(self):
+        validate_begin_end_step(self.begin_step_percent, self.end_step_percent)
+        return self
+
+    def invoke(self, context: InvocationContext) -> FluxControlNetOutput:
+        return FluxControlNetOutput(
+            control=FluxControlNetField(
+                image=self.image,
+                control_model=self.control_model,
+                control_weight=self.control_weight,
+                begin_step_percent=self.begin_step_percent,
+                end_step_percent=self.end_step_percent,
+                resize_mode=self.resize_mode,
+                instantx_control_mode=self.instantx_control_mode,
+            ),
+        )

invokeai/app/invocations/flux_denoise.py

@@ -1,35 +1,48 @@
 from contextlib import ExitStack
 from typing import Callable, Iterator, Optional, Tuple
 
+import numpy as np
+import numpy.typing as npt
 import torch
 import torchvision.transforms as tv_transforms
 from torchvision.transforms.functional import resize as tv_resize
+from transformers import CLIPImageProcessor, CLIPVisionModelWithProjection
 
 from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
 from invokeai.app.invocations.fields import (
     DenoiseMaskField,
     FieldDescriptions,
     FluxConditioningField,
+    ImageField,
     Input,
     InputField,
     LatentsField,
     WithBoard,
     WithMetadata,
 )
-from invokeai.app.invocations.model import TransformerField
+from invokeai.app.invocations.flux_controlnet import FluxControlNetField
+from invokeai.app.invocations.ip_adapter import IPAdapterField
+from invokeai.app.invocations.model import TransformerField, VAEField
 from invokeai.app.invocations.primitives import LatentsOutput
 from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.flux.controlnet.instantx_controlnet_flux import InstantXControlNetFlux
+from invokeai.backend.flux.controlnet.xlabs_controlnet_flux import XLabsControlNetFlux
 from invokeai.backend.flux.denoise import denoise
+from invokeai.backend.flux.extensions.inpaint_extension import InpaintExtension
+from invokeai.backend.flux.extensions.instantx_controlnet_extension import InstantXControlNetExtension
+from invokeai.backend.flux.extensions.xlabs_controlnet_extension import XLabsControlNetExtension
+from invokeai.backend.flux.extensions.xlabs_ip_adapter_extension import XLabsIPAdapterExtension
+from invokeai.backend.flux.ip_adapter.xlabs_ip_adapter_flux import XlabsIpAdapterFlux
 from invokeai.backend.flux.model import Flux
 from invokeai.backend.flux.sampling_utils import (
-    clip_timestep_schedule,
+    clip_timestep_schedule_fractional,
     generate_img_ids,
     get_noise,
     get_schedule,
     pack,
     unpack,
 )
-from invokeai.backend.flux.trajectory_guidance_extension import TrajectoryGuidanceExtension
+from invokeai.backend.lora.conversions.flux_lora_constants import FLUX_LORA_TRANSFORMER_PREFIX
 from invokeai.backend.lora.lora_model_raw import LoRAModelRaw
 from invokeai.backend.lora.lora_patcher import LoRAPatcher
 from invokeai.backend.model_manager.config import ModelFormat
@@ -43,7 +56,7 @@ from invokeai.backend.util.devices import TorchDevice
     title="FLUX Denoise",
     tags=["image", "flux"],
     category="image",
-    version="2.1.0",
+    version="3.2.1",
     classification=Classification.Prototype,
 )
 class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
@@ -68,12 +81,7 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         description=FieldDescriptions.denoising_start,
     )
     denoising_end: float = InputField(default=1.0, ge=0, le=1, description=FieldDescriptions.denoising_end)
-    trajectory_guidance_strength: float = InputField(
-        default=0.0,
-        ge=0.0,
-        le=1.0,
-        description="Value indicating how strongly to guide the denoising process towards the initial latents (during image-to-image). Range [0, 1]. A value of 0.0 is equivalent to vanilla image-to-image. A value of 1.0 will guide the denoising process very close to the original latents.",
-    )
+    add_noise: bool = InputField(default=True, description="Add noise based on denoising start.")
     transformer: TransformerField = InputField(
         description=FieldDescriptions.flux_model,
         input=Input.Connection,
@@ -82,6 +90,24 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
     positive_text_conditioning: FluxConditioningField = InputField(
         description=FieldDescriptions.positive_cond, input=Input.Connection
     )
+    negative_text_conditioning: FluxConditioningField | None = InputField(
+        default=None,
+        description="Negative conditioning tensor. Can be None if cfg_scale is 1.0.",
+        input=Input.Connection,
+    )
+    cfg_scale: float | list[float] = InputField(default=1.0, description=FieldDescriptions.cfg_scale, title="CFG Scale")
+    cfg_scale_start_step: int = InputField(
+        default=0,
+        title="CFG Scale Start Step",
+        description="Index of the first step to apply cfg_scale. Negative indices count backwards from the "
+        + "the last step (e.g. a value of -1 refers to the final step).",
+    )
+    cfg_scale_end_step: int = InputField(
+        default=-1,
+        title="CFG Scale End Step",
+        description="Index of the last step to apply cfg_scale. Negative indices count backwards from the "
+        + "last step (e.g. a value of -1 refers to the final step).",
+    )
     width: int = InputField(default=1024, multiple_of=16, description="Width of the generated image.")
     height: int = InputField(default=1024, multiple_of=16, description="Height of the generated image.")
     num_steps: int = InputField(
@@ -92,6 +118,18 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         description="The guidance strength. Higher values adhere more strictly to the prompt, and will produce less diverse images. FLUX dev only, ignored for schnell.",
     )
     seed: int = InputField(default=0, description="Randomness seed for reproducibility.")
+    control: FluxControlNetField | list[FluxControlNetField] | None = InputField(
+        default=None, input=Input.Connection, description="ControlNet models."
+    )
+    controlnet_vae: VAEField | None = InputField(
+        default=None,
+        description=FieldDescriptions.vae,
+        input=Input.Connection,
+    )
+
+    ip_adapter: IPAdapterField | list[IPAdapterField] | None = InputField(
+        description=FieldDescriptions.ip_adapter, title="IP-Adapter", default=None, input=Input.Connection
+    )
 
     @torch.no_grad()
     def invoke(self, context: InvocationContext) -> LatentsOutput:
@@ -101,6 +139,19 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         name = context.tensors.save(tensor=latents)
         return LatentsOutput.build(latents_name=name, latents=latents, seed=None)
 
+    def _load_text_conditioning(
+        self, context: InvocationContext, conditioning_name: str, dtype: torch.dtype
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        # Load the conditioning data.
+        cond_data = context.conditioning.load(conditioning_name)
+        assert len(cond_data.conditionings) == 1
+        flux_conditioning = cond_data.conditionings[0]
+        assert isinstance(flux_conditioning, FLUXConditioningInfo)
+        flux_conditioning = flux_conditioning.to(dtype=dtype)
+        t5_embeddings = flux_conditioning.t5_embeds
+        clip_embeddings = flux_conditioning.clip_embeds
+        return t5_embeddings, clip_embeddings
+
     def _run_diffusion(
         self,
         context: InvocationContext,
@@ -108,13 +159,15 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         inference_dtype = torch.bfloat16
 
         # Load the conditioning data.
-        cond_data = context.conditioning.load(self.positive_text_conditioning.conditioning_name)
-        assert len(cond_data.conditionings) == 1
-        flux_conditioning = cond_data.conditionings[0]
-        assert isinstance(flux_conditioning, FLUXConditioningInfo)
-        flux_conditioning = flux_conditioning.to(dtype=inference_dtype)
-        t5_embeddings = flux_conditioning.t5_embeds
-        clip_embeddings = flux_conditioning.clip_embeds
+        pos_t5_embeddings, pos_clip_embeddings = self._load_text_conditioning(
+            context, self.positive_text_conditioning.conditioning_name, inference_dtype
+        )
+        neg_t5_embeddings: torch.Tensor | None = None
+        neg_clip_embeddings: torch.Tensor | None = None
+        if self.negative_text_conditioning is not None:
+            neg_t5_embeddings, neg_clip_embeddings = self._load_text_conditioning(
+                context, self.negative_text_conditioning.conditioning_name, inference_dtype
+            )
 
         # Load the input latents, if provided.
         init_latents = context.tensors.load(self.latents.latents_name) if self.latents else None
@@ -143,7 +196,7 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         )
 
         # Clip the timesteps schedule based on denoising_start and denoising_end.
-        timesteps = clip_timestep_schedule(timesteps, self.denoising_start, self.denoising_end)
+        timesteps = clip_timestep_schedule_fractional(timesteps, self.denoising_start, self.denoising_end)
 
         # Prepare input latent image.
         if init_latents is not None:
@@ -155,9 +208,12 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
                     "to be poor. Consider using a FLUX dev model instead."
                 )
 
-            # Noise the orig_latents by the appropriate amount for the first timestep.
-            t_0 = timesteps[0]
-            x = t_0 * noise + (1.0 - t_0) * init_latents
+            if self.add_noise:
+                # Noise the orig_latents by the appropriate amount for the first timestep.
+                t_0 = timesteps[0]
+                x = t_0 * noise + (1.0 - t_0) * init_latents
+            else:
+                x = init_latents
         else:
             # init_latents are not provided, so we are not doing image-to-image (i.e. we are starting from pure noise).
             if self.denoising_start > 1e-5:
@@ -172,11 +228,19 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
 
         inpaint_mask = self._prep_inpaint_mask(context, x)
 
-        b, _c, h, w = x.shape
-        img_ids = generate_img_ids(h=h, w=w, batch_size=b, device=x.device, dtype=x.dtype)
+        b, _c, latent_h, latent_w = x.shape
+        img_ids = generate_img_ids(h=latent_h, w=latent_w, batch_size=b, device=x.device, dtype=x.dtype)
 
-        bs, t5_seq_len, _ = t5_embeddings.shape
-        txt_ids = torch.zeros(bs, t5_seq_len, 3, dtype=inference_dtype, device=TorchDevice.choose_torch_device())
+        pos_bs, pos_t5_seq_len, _ = pos_t5_embeddings.shape
+        pos_txt_ids = torch.zeros(
+            pos_bs, pos_t5_seq_len, 3, dtype=inference_dtype, device=TorchDevice.choose_torch_device()
+        )
+        neg_txt_ids: torch.Tensor | None = None
+        if neg_t5_embeddings is not None:
+            neg_bs, neg_t5_seq_len, _ = neg_t5_embeddings.shape
+            neg_txt_ids = torch.zeros(
+                neg_bs, neg_t5_seq_len, 3, dtype=inference_dtype, device=TorchDevice.choose_torch_device()
+            )
 
         # Pack all latent tensors.
         init_latents = pack(init_latents) if init_latents is not None else None
@@ -187,21 +251,46 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         # Now that we have 'packed' the latent tensors, verify that we calculated the image_seq_len correctly.
         assert image_seq_len == x.shape[1]
 
-        # Prepare trajectory guidance extension.
-        traj_guidance_extension: TrajectoryGuidanceExtension | None = None
-        if init_latents is not None:
-            traj_guidance_extension = TrajectoryGuidanceExtension(
+        # Prepare inpaint extension.
+        inpaint_extension: InpaintExtension | None = None
+        if inpaint_mask is not None:
+            assert init_latents is not None
+            inpaint_extension = InpaintExtension(
                 init_latents=init_latents,
                 inpaint_mask=inpaint_mask,
-                trajectory_guidance_strength=self.trajectory_guidance_strength,
+                noise=noise,
             )
 
-        with (
-            transformer_info.model_on_device() as (cached_weights, transformer),
-            ExitStack() as exit_stack,
-        ):
-            assert isinstance(transformer, Flux)
+        # Compute the IP-Adapter image prompt clip embeddings.
+        # We do this before loading other models to minimize peak memory.
+        # TODO(ryand): We should really do this in a separate invocation to benefit from caching.
+        ip_adapter_fields = self._normalize_ip_adapter_fields()
+        pos_image_prompt_clip_embeds, neg_image_prompt_clip_embeds = self._prep_ip_adapter_image_prompt_clip_embeds(
+            ip_adapter_fields, context
+        )
 
+        cfg_scale = self.prep_cfg_scale(
+            cfg_scale=self.cfg_scale,
+            timesteps=timesteps,
+            cfg_scale_start_step=self.cfg_scale_start_step,
+            cfg_scale_end_step=self.cfg_scale_end_step,
+        )
+
+        with ExitStack() as exit_stack:
+            # Prepare ControlNet extensions.
+            # Note: We do this before loading the transformer model to minimize peak memory (see implementation).
+            controlnet_extensions = self._prep_controlnet_extensions(
+                context=context,
+                exit_stack=exit_stack,
+                latent_height=latent_h,
+                latent_width=latent_w,
+                dtype=inference_dtype,
+                device=x.device,
+            )
+
+            # Load the transformer model.
+            (cached_weights, transformer) = exit_stack.enter_context(transformer_info.model_on_device())
+            assert isinstance(transformer, Flux)
             config = transformer_info.config
             assert config is not None
 
@@ -213,40 +302,110 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
                     LoRAPatcher.apply_lora_patches(
                         model=transformer,
                         patches=self._lora_iterator(context),
-                        prefix="",
+                        prefix=FLUX_LORA_TRANSFORMER_PREFIX,
                         cached_weights=cached_weights,
                     )
                 )
-            elif config.format in [ModelFormat.BnbQuantizedLlmInt8b, ModelFormat.BnbQuantizednf4b]:
+            elif config.format in [
+                ModelFormat.BnbQuantizedLlmInt8b,
+                ModelFormat.BnbQuantizednf4b,
+                ModelFormat.GGUFQuantized,
+            ]:
                 # The model is quantized, so apply the LoRA weights as sidecar layers. This results in slower inference,
                 # than directly patching the weights, but is agnostic to the quantization format.
                 exit_stack.enter_context(
                     LoRAPatcher.apply_lora_sidecar_patches(
                         model=transformer,
                         patches=self._lora_iterator(context),
-                        prefix="",
+                        prefix=FLUX_LORA_TRANSFORMER_PREFIX,
                         dtype=inference_dtype,
                     )
                 )
             else:
                 raise ValueError(f"Unsupported model format: {config.format}")
 
+            # Prepare IP-Adapter extensions.
+            pos_ip_adapter_extensions, neg_ip_adapter_extensions = self._prep_ip_adapter_extensions(
+                pos_image_prompt_clip_embeds=pos_image_prompt_clip_embeds,
+                neg_image_prompt_clip_embeds=neg_image_prompt_clip_embeds,
+                ip_adapter_fields=ip_adapter_fields,
+                context=context,
+                exit_stack=exit_stack,
+                dtype=inference_dtype,
+            )
+
             x = denoise(
                 model=transformer,
                 img=x,
                 img_ids=img_ids,
-                txt=t5_embeddings,
-                txt_ids=txt_ids,
-                vec=clip_embeddings,
+                txt=pos_t5_embeddings,
+                txt_ids=pos_txt_ids,
+                vec=pos_clip_embeddings,
+                neg_txt=neg_t5_embeddings,
+                neg_txt_ids=neg_txt_ids,
+                neg_vec=neg_clip_embeddings,
                 timesteps=timesteps,
                 step_callback=self._build_step_callback(context),
                 guidance=self.guidance,
-                traj_guidance_extension=traj_guidance_extension,
+                cfg_scale=cfg_scale,
+                inpaint_extension=inpaint_extension,
+                controlnet_extensions=controlnet_extensions,
+                pos_ip_adapter_extensions=pos_ip_adapter_extensions,
+                neg_ip_adapter_extensions=neg_ip_adapter_extensions,
             )
 
         x = unpack(x.float(), self.height, self.width)
         return x
 
+    @classmethod
+    def prep_cfg_scale(
+        cls, cfg_scale: float | list[float], timesteps: list[float], cfg_scale_start_step: int, cfg_scale_end_step: int
+    ) -> list[float]:
+        """Prepare the cfg_scale schedule.
+
+        - Clips the cfg_scale schedule based on cfg_scale_start_step and cfg_scale_end_step.
+        - If cfg_scale is a list, then it is assumed to be a schedule and is returned as-is.
+        - If cfg_scale is a scalar, then a linear schedule is created from cfg_scale_start_step to cfg_scale_end_step.
+        """
+        # num_steps is the number of denoising steps, which is one less than the number of timesteps.
+        num_steps = len(timesteps) - 1
+
+        # Normalize cfg_scale to a list if it is a scalar.
+        cfg_scale_list: list[float]
+        if isinstance(cfg_scale, float):
+            cfg_scale_list = [cfg_scale] * num_steps
+        elif isinstance(cfg_scale, list):
+            cfg_scale_list = cfg_scale
+        else:
+            raise ValueError(f"Unsupported cfg_scale type: {type(cfg_scale)}")
+        assert len(cfg_scale_list) == num_steps
+
+        # Handle negative indices for cfg_scale_start_step and cfg_scale_end_step.
+        start_step_index = cfg_scale_start_step
+        if start_step_index < 0:
+            start_step_index = num_steps + start_step_index
+        end_step_index = cfg_scale_end_step
+        if end_step_index < 0:
+            end_step_index = num_steps + end_step_index
+
+        # Validate the start and end step indices.
+        if not (0 <= start_step_index < num_steps):
+            raise ValueError(f"Invalid cfg_scale_start_step. Out of range: {cfg_scale_start_step}.")
+        if not (0 <= end_step_index < num_steps):
+            raise ValueError(f"Invalid cfg_scale_end_step. Out of range: {cfg_scale_end_step}.")
+        if start_step_index > end_step_index:
+            raise ValueError(
+                f"cfg_scale_start_step ({cfg_scale_start_step}) must be before cfg_scale_end_step "
+                + f"({cfg_scale_end_step})."
+            )
+
+        # Set values outside the start and end step indices to 1.0. This is equivalent to disabling cfg_scale for those
+        # steps.
+        clipped_cfg_scale = [1.0] * num_steps
+        clipped_cfg_scale[start_step_index : end_step_index + 1] = cfg_scale_list[start_step_index : end_step_index + 1]
+
+        return clipped_cfg_scale
+
     def _prep_inpaint_mask(self, context: InvocationContext, latents: torch.Tensor) -> torch.Tensor | None:
         """Prepare the inpaint mask.
 
@@ -288,6 +447,210 @@ class FluxDenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
         # `latents`.
         return mask.expand_as(latents)
 
+    def _prep_controlnet_extensions(
+        self,
+        context: InvocationContext,
+        exit_stack: ExitStack,
+        latent_height: int,
+        latent_width: int,
+        dtype: torch.dtype,
+        device: torch.device,
+    ) -> list[XLabsControlNetExtension | InstantXControlNetExtension]:
+        # Normalize the controlnet input to list[ControlField].
+        controlnets: list[FluxControlNetField]
+        if self.control is None:
+            controlnets = []
+        elif isinstance(self.control, FluxControlNetField):
+            controlnets = [self.control]
+        elif isinstance(self.control, list):
+            controlnets = self.control
+        else:
+            raise ValueError(f"Unsupported controlnet type: {type(self.control)}")
+
+        # TODO(ryand): Add a field to the model config so that we can distinguish between XLabs and InstantX ControlNets
+        # before loading the models. Then make sure that all VAE encoding is done before loading the ControlNets to
+        # minimize peak memory.
+
+        # First, load the ControlNet models so that we can determine the ControlNet types.
+        controlnet_models = [context.models.load(controlnet.control_model) for controlnet in controlnets]
+
+        # Calculate the controlnet conditioning tensors.
+        # We do this before loading the ControlNet models because it may require running the VAE, and we are trying to
+        # keep peak memory down.
+        controlnet_conds: list[torch.Tensor] = []
+        for controlnet, controlnet_model in zip(controlnets, controlnet_models, strict=True):
+            image = context.images.get_pil(controlnet.image.image_name)
+            if isinstance(controlnet_model.model, InstantXControlNetFlux):
+                if self.controlnet_vae is None:
+                    raise ValueError("A ControlNet VAE is required when using an InstantX FLUX ControlNet.")
+                vae_info = context.models.load(self.controlnet_vae.vae)
+                controlnet_conds.append(
+                    InstantXControlNetExtension.prepare_controlnet_cond(
+                        controlnet_image=image,
+                        vae_info=vae_info,
+                        latent_height=latent_height,
+                        latent_width=latent_width,
+                        dtype=dtype,
+                        device=device,
+                        resize_mode=controlnet.resize_mode,
+                    )
+                )
+            elif isinstance(controlnet_model.model, XLabsControlNetFlux):
+                controlnet_conds.append(
+                    XLabsControlNetExtension.prepare_controlnet_cond(
+                        controlnet_image=image,
+                        latent_height=latent_height,
+                        latent_width=latent_width,
+                        dtype=dtype,
+                        device=device,
+                        resize_mode=controlnet.resize_mode,
+                    )
+                )
+
+        # Finally, load the ControlNet models and initialize the ControlNet extensions.
+        controlnet_extensions: list[XLabsControlNetExtension | InstantXControlNetExtension] = []
+        for controlnet, controlnet_cond, controlnet_model in zip(
+            controlnets, controlnet_conds, controlnet_models, strict=True
+        ):
+            model = exit_stack.enter_context(controlnet_model)
+
+            if isinstance(model, XLabsControlNetFlux):
+                controlnet_extensions.append(
+                    XLabsControlNetExtension(
+                        model=model,
+                        controlnet_cond=controlnet_cond,
+                        weight=controlnet.control_weight,
+                        begin_step_percent=controlnet.begin_step_percent,
+                        end_step_percent=controlnet.end_step_percent,
+                    )
+                )
+            elif isinstance(model, InstantXControlNetFlux):
+                instantx_control_mode: torch.Tensor | None = None
+                if controlnet.instantx_control_mode is not None and controlnet.instantx_control_mode >= 0:
+                    instantx_control_mode = torch.tensor(controlnet.instantx_control_mode, dtype=torch.long)
+                    instantx_control_mode = instantx_control_mode.reshape([-1, 1])
+
+                controlnet_extensions.append(
+                    InstantXControlNetExtension(
+                        model=model,
+                        controlnet_cond=controlnet_cond,
+                        instantx_control_mode=instantx_control_mode,
+                        weight=controlnet.control_weight,
+                        begin_step_percent=controlnet.begin_step_percent,
+                        end_step_percent=controlnet.end_step_percent,
+                    )
+                )
+            else:
+                raise ValueError(f"Unsupported ControlNet model type: {type(model)}")
+
+        return controlnet_extensions
+
+    def _normalize_ip_adapter_fields(self) -> list[IPAdapterField]:
+        if self.ip_adapter is None:
+            return []
+        elif isinstance(self.ip_adapter, IPAdapterField):
+            return [self.ip_adapter]
+        elif isinstance(self.ip_adapter, list):
+            return self.ip_adapter
+        else:
+            raise ValueError(f"Unsupported IP-Adapter type: {type(self.ip_adapter)}")
+
+    def _prep_ip_adapter_image_prompt_clip_embeds(
+        self,
+        ip_adapter_fields: list[IPAdapterField],
+        context: InvocationContext,
+    ) -> tuple[list[torch.Tensor], list[torch.Tensor]]:
+        """Run the IPAdapter CLIPVisionModel, returning image prompt embeddings."""
+        clip_image_processor = CLIPImageProcessor()
+
+        pos_image_prompt_clip_embeds: list[torch.Tensor] = []
+        neg_image_prompt_clip_embeds: list[torch.Tensor] = []
+        for ip_adapter_field in ip_adapter_fields:
+            # `ip_adapter_field.image` could be a list or a single ImageField. Normalize to a list here.
+            ipa_image_fields: list[ImageField]
+            if isinstance(ip_adapter_field.image, ImageField):
+                ipa_image_fields = [ip_adapter_field.image]
+            elif isinstance(ip_adapter_field.image, list):
+                ipa_image_fields = ip_adapter_field.image
+            else:
+                raise ValueError(f"Unsupported IP-Adapter image type: {type(ip_adapter_field.image)}")
+
+            if len(ipa_image_fields) != 1:
+                raise ValueError(
+                    f"FLUX IP-Adapter only supports a single image prompt (received {len(ipa_image_fields)})."
+                )
+
+            ipa_images = [context.images.get_pil(image.image_name, mode="RGB") for image in ipa_image_fields]
+
+            pos_images: list[npt.NDArray[np.uint8]] = []
+            neg_images: list[npt.NDArray[np.uint8]] = []
+            for ipa_image in ipa_images:
+                assert ipa_image.mode == "RGB"
+                pos_image = np.array(ipa_image)
+                # We use a black image as the negative image prompt for parity with
+                # https://github.com/XLabs-AI/x-flux-comfyui/blob/45c834727dd2141aebc505ae4b01f193a8414e38/nodes.py#L592-L593
+                # An alternative scheme would be to apply zeros_like() after calling the clip_image_processor.
+                neg_image = np.zeros_like(pos_image)
+                pos_images.append(pos_image)
+                neg_images.append(neg_image)
+
+            with context.models.load(ip_adapter_field.image_encoder_model) as image_encoder_model:
+                assert isinstance(image_encoder_model, CLIPVisionModelWithProjection)
+
+                clip_image: torch.Tensor = clip_image_processor(images=pos_images, return_tensors="pt").pixel_values
+                clip_image = clip_image.to(device=image_encoder_model.device, dtype=image_encoder_model.dtype)
+                pos_clip_image_embeds = image_encoder_model(clip_image).image_embeds
+
+                clip_image = clip_image_processor(images=neg_images, return_tensors="pt").pixel_values
+                clip_image = clip_image.to(device=image_encoder_model.device, dtype=image_encoder_model.dtype)
+                neg_clip_image_embeds = image_encoder_model(clip_image).image_embeds
+
+            pos_image_prompt_clip_embeds.append(pos_clip_image_embeds)
+            neg_image_prompt_clip_embeds.append(neg_clip_image_embeds)
+
+        return pos_image_prompt_clip_embeds, neg_image_prompt_clip_embeds
+
+    def _prep_ip_adapter_extensions(
+        self,
+        ip_adapter_fields: list[IPAdapterField],
+        pos_image_prompt_clip_embeds: list[torch.Tensor],
+        neg_image_prompt_clip_embeds: list[torch.Tensor],
+        context: InvocationContext,
+        exit_stack: ExitStack,
+        dtype: torch.dtype,
+    ) -> tuple[list[XLabsIPAdapterExtension], list[XLabsIPAdapterExtension]]:
+        pos_ip_adapter_extensions: list[XLabsIPAdapterExtension] = []
+        neg_ip_adapter_extensions: list[XLabsIPAdapterExtension] = []
+        for ip_adapter_field, pos_image_prompt_clip_embed, neg_image_prompt_clip_embed in zip(
+            ip_adapter_fields, pos_image_prompt_clip_embeds, neg_image_prompt_clip_embeds, strict=True
+        ):
+            ip_adapter_model = exit_stack.enter_context(context.models.load(ip_adapter_field.ip_adapter_model))
+            assert isinstance(ip_adapter_model, XlabsIpAdapterFlux)
+            ip_adapter_model = ip_adapter_model.to(dtype=dtype)
+            if ip_adapter_field.mask is not None:
+                raise ValueError("IP-Adapter masks are not yet supported in Flux.")
+            ip_adapter_extension = XLabsIPAdapterExtension(
+                model=ip_adapter_model,
+                image_prompt_clip_embed=pos_image_prompt_clip_embed,
+                weight=ip_adapter_field.weight,
+                begin_step_percent=ip_adapter_field.begin_step_percent,
+                end_step_percent=ip_adapter_field.end_step_percent,
+            )
+            ip_adapter_extension.run_image_proj(dtype=dtype)
+            pos_ip_adapter_extensions.append(ip_adapter_extension)
+
+            ip_adapter_extension = XLabsIPAdapterExtension(
+                model=ip_adapter_model,
+                image_prompt_clip_embed=neg_image_prompt_clip_embed,
+                weight=ip_adapter_field.weight,
+                begin_step_percent=ip_adapter_field.begin_step_percent,
+                end_step_percent=ip_adapter_field.end_step_percent,
+            )
+            ip_adapter_extension.run_image_proj(dtype=dtype)
+            neg_ip_adapter_extensions.append(ip_adapter_extension)
+
+        return pos_ip_adapter_extensions, neg_ip_adapter_extensions
+
     def _lora_iterator(self, context: InvocationContext) -> Iterator[Tuple[LoRAModelRaw, float]]:
         for lora in self.transformer.loras:
             lora_info = context.models.load(lora.lora)

invokeai/app/invocations/flux_ip_adapter.py

@@ -0,0 +1,89 @@
+from builtins import float
+from typing import List, Literal, Union
+
+from pydantic import field_validator, model_validator
+from typing_extensions import Self
+
+from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.fields import InputField, UIType
+from invokeai.app.invocations.ip_adapter import (
+    CLIP_VISION_MODEL_MAP,
+    IPAdapterField,
+    IPAdapterInvocation,
+    IPAdapterOutput,
+)
+from invokeai.app.invocations.model import ModelIdentifierField
+from invokeai.app.invocations.primitives import ImageField
+from invokeai.app.invocations.util import validate_begin_end_step, validate_weights
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.model_manager.config import (
+    IPAdapterCheckpointConfig,
+    IPAdapterInvokeAIConfig,
+)
+
+
+@invocation(
+    "flux_ip_adapter",
+    title="FLUX IP-Adapter",
+    tags=["ip_adapter", "control"],
+    category="ip_adapter",
+    version="1.0.0",
+    classification=Classification.Prototype,
+)
+class FluxIPAdapterInvocation(BaseInvocation):
+    """Collects FLUX IP-Adapter info to pass to other nodes."""
+
+    # FLUXIPAdapterInvocation is based closely on IPAdapterInvocation, but with some unsupported features removed.
+
+    image: ImageField = InputField(description="The IP-Adapter image prompt(s).")
+    ip_adapter_model: ModelIdentifierField = InputField(
+        description="The IP-Adapter model.", title="IP-Adapter Model", ui_type=UIType.IPAdapterModel
+    )
+    # Currently, the only known ViT model used by FLUX IP-Adapters is ViT-L.
+    clip_vision_model: Literal["ViT-L"] = InputField(description="CLIP Vision model to use.", default="ViT-L")
+    weight: Union[float, List[float]] = InputField(
+        default=1, description="The weight given to the IP-Adapter", title="Weight"
+    )
+    begin_step_percent: float = InputField(
+        default=0, ge=0, le=1, description="When the IP-Adapter is first applied (% of total steps)"
+    )
+    end_step_percent: float = InputField(
+        default=1, ge=0, le=1, description="When the IP-Adapter is last applied (% of total steps)"
+    )
+
+    @field_validator("weight")
+    @classmethod
+    def validate_ip_adapter_weight(cls, v: float) -> float:
+        validate_weights(v)
+        return v
+
+    @model_validator(mode="after")
+    def validate_begin_end_step_percent(self) -> Self:
+        validate_begin_end_step(self.begin_step_percent, self.end_step_percent)
+        return self
+
+    def invoke(self, context: InvocationContext) -> IPAdapterOutput:
+        # Lookup the CLIP Vision encoder that is intended to be used with the IP-Adapter model.
+        ip_adapter_info = context.models.get_config(self.ip_adapter_model.key)
+        assert isinstance(ip_adapter_info, (IPAdapterInvokeAIConfig, IPAdapterCheckpointConfig))
+
+        # Note: There is a IPAdapterInvokeAIConfig.image_encoder_model_id field, but it isn't trustworthy.
+        image_encoder_starter_model = CLIP_VISION_MODEL_MAP[self.clip_vision_model]
+        image_encoder_model_id = image_encoder_starter_model.source
+        image_encoder_model_name = image_encoder_starter_model.name
+        image_encoder_model = IPAdapterInvocation.get_clip_image_encoder(
+            context, image_encoder_model_id, image_encoder_model_name
+        )
+
+        return IPAdapterOutput(
+            ip_adapter=IPAdapterField(
+                image=self.image,
+                ip_adapter_model=self.ip_adapter_model,
+                image_encoder_model=ModelIdentifierField.from_config(image_encoder_model),
+                weight=self.weight,
+                target_blocks=[],  # target_blocks is currently unused for FLUX IP-Adapters.
+                begin_step_percent=self.begin_step_percent,
+                end_step_percent=self.end_step_percent,
+                mask=None,  # mask is currently unused for FLUX IP-Adapters.
+            ),
+        )

invokeai/app/invocations/flux_lora_loader.py

@@ -8,7 +8,7 @@ from invokeai.app.invocations.baseinvocation import (
     invocation_output,
 )
 from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIType
-from invokeai.app.invocations.model import LoRAField, ModelIdentifierField, TransformerField
+from invokeai.app.invocations.model import CLIPField, LoRAField, ModelIdentifierField, TransformerField
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.model_manager.config import BaseModelType
 
@@ -20,6 +20,7 @@ class FluxLoRALoaderOutput(BaseInvocationOutput):
     transformer: Optional[TransformerField] = OutputField(
         default=None, description=FieldDescriptions.transformer, title="FLUX Transformer"
     )
+    clip: Optional[CLIPField] = OutputField(default=None, description=FieldDescriptions.clip, title="CLIP")
 
 
 @invocation(
@@ -27,21 +28,28 @@ class FluxLoRALoaderOutput(BaseInvocationOutput):
     title="FLUX LoRA",
     tags=["lora", "model", "flux"],
     category="model",
-    version="1.0.0",
+    version="1.1.0",
     classification=Classification.Prototype,
 )
 class FluxLoRALoaderInvocation(BaseInvocation):
-    """Apply a LoRA model to a FLUX transformer."""
+    """Apply a LoRA model to a FLUX transformer and/or text encoder."""
 
     lora: ModelIdentifierField = InputField(
         description=FieldDescriptions.lora_model, title="LoRA", ui_type=UIType.LoRAModel
     )
     weight: float = InputField(default=0.75, description=FieldDescriptions.lora_weight)
-    transformer: TransformerField = InputField(
+    transformer: TransformerField | None = InputField(
+        default=None,
         description=FieldDescriptions.transformer,
         input=Input.Connection,
         title="FLUX Transformer",
     )
+    clip: CLIPField | None = InputField(
+        default=None,
+        title="CLIP",
+        description=FieldDescriptions.clip,
+        input=Input.Connection,
+    )
 
     def invoke(self, context: InvocationContext) -> FluxLoRALoaderOutput:
         lora_key = self.lora.key
@@ -49,18 +57,33 @@ class FluxLoRALoaderInvocation(BaseInvocation):
         if not context.models.exists(lora_key):
             raise ValueError(f"Unknown lora: {lora_key}!")
 
-        if any(lora.lora.key == lora_key for lora in self.transformer.loras):
+        # Check for existing LoRAs with the same key.
+        if self.transformer and any(lora.lora.key == lora_key for lora in self.transformer.loras):
             raise ValueError(f'LoRA "{lora_key}" already applied to transformer.')
+        if self.clip and any(lora.lora.key == lora_key for lora in self.clip.loras):
+            raise ValueError(f'LoRA "{lora_key}" already applied to CLIP encoder.')
 
-        transformer = self.transformer.model_copy(deep=True)
-        transformer.loras.append(
-            LoRAField(
-                lora=self.lora,
-                weight=self.weight,
+        output = FluxLoRALoaderOutput()
+
+        # Attach LoRA layers to the models.
+        if self.transformer is not None:
+            output.transformer = self.transformer.model_copy(deep=True)
+            output.transformer.loras.append(
+                LoRAField(
+                    lora=self.lora,
+                    weight=self.weight,
+                )
+            )
+        if self.clip is not None:
+            output.clip = self.clip.model_copy(deep=True)
+            output.clip.loras.append(
+                LoRAField(
+                    lora=self.lora,
+                    weight=self.weight,
+                )
             )
-        )
 
-        return FluxLoRALoaderOutput(transformer=transformer)
+        return output
 
 
 @invocation(
@@ -68,7 +91,7 @@ class FluxLoRALoaderInvocation(BaseInvocation):
     title="FLUX LoRA Collection Loader",
     tags=["lora", "model", "flux"],
     category="model",
-    version="1.0.0",
+    version="1.1.0",
     classification=Classification.Prototype,
 )
 class FLUXLoRACollectionLoader(BaseInvocation):
@@ -84,6 +107,12 @@ class FLUXLoRACollectionLoader(BaseInvocation):
         input=Input.Connection,
         title="Transformer",
     )
+    clip: CLIPField | None = InputField(
+        default=None,
+        title="CLIP",
+        description=FieldDescriptions.clip,
+        input=Input.Connection,
+    )
 
     def invoke(self, context: InvocationContext) -> FluxLoRALoaderOutput:
         output = FluxLoRALoaderOutput()
@@ -106,4 +135,9 @@ class FLUXLoRACollectionLoader(BaseInvocation):
                     output.transformer = self.transformer.model_copy(deep=True)
                 output.transformer.loras.append(lora)
 
+            if self.clip is not None:
+                if output.clip is None:
+                    output.clip = self.clip.model_copy(deep=True)
+                output.clip.loras.append(lora)
+
         return output

invokeai/app/invocations/flux_model_loader.py

@@ -0,0 +1,89 @@
+from typing import Literal
+
+from invokeai.app.invocations.baseinvocation import (
+    BaseInvocation,
+    BaseInvocationOutput,
+    Classification,
+    invocation,
+    invocation_output,
+)
+from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIType
+from invokeai.app.invocations.model import CLIPField, ModelIdentifierField, T5EncoderField, TransformerField, VAEField
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.flux.util import max_seq_lengths
+from invokeai.backend.model_manager.config import (
+    CheckpointConfigBase,
+    SubModelType,
+)
+
+
+@invocation_output("flux_model_loader_output")
+class FluxModelLoaderOutput(BaseInvocationOutput):
+    """Flux base model loader output"""
+
+    transformer: TransformerField = OutputField(description=FieldDescriptions.transformer, title="Transformer")
+    clip: CLIPField = OutputField(description=FieldDescriptions.clip, title="CLIP")
+    t5_encoder: T5EncoderField = OutputField(description=FieldDescriptions.t5_encoder, title="T5 Encoder")
+    vae: VAEField = OutputField(description=FieldDescriptions.vae, title="VAE")
+    max_seq_len: Literal[256, 512] = OutputField(
+        description="The max sequence length to used for the T5 encoder. (256 for schnell transformer, 512 for dev transformer)",
+        title="Max Seq Length",
+    )
+
+
+@invocation(
+    "flux_model_loader",
+    title="Flux Main Model",
+    tags=["model", "flux"],
+    category="model",
+    version="1.0.4",
+    classification=Classification.Prototype,
+)
+class FluxModelLoaderInvocation(BaseInvocation):
+    """Loads a flux base model, outputting its submodels."""
+
+    model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.flux_model,
+        ui_type=UIType.FluxMainModel,
+        input=Input.Direct,
+    )
+
+    t5_encoder_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.t5_encoder, ui_type=UIType.T5EncoderModel, input=Input.Direct, title="T5 Encoder"
+    )
+
+    clip_embed_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.clip_embed_model,
+        ui_type=UIType.CLIPEmbedModel,
+        input=Input.Direct,
+        title="CLIP Embed",
+    )
+
+    vae_model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.vae_model, ui_type=UIType.FluxVAEModel, title="VAE"
+    )
+
+    def invoke(self, context: InvocationContext) -> FluxModelLoaderOutput:
+        for key in [self.model.key, self.t5_encoder_model.key, self.clip_embed_model.key, self.vae_model.key]:
+            if not context.models.exists(key):
+                raise ValueError(f"Unknown model: {key}")
+
+        transformer = self.model.model_copy(update={"submodel_type": SubModelType.Transformer})
+        vae = self.vae_model.model_copy(update={"submodel_type": SubModelType.VAE})
+
+        tokenizer = self.clip_embed_model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
+        clip_encoder = self.clip_embed_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
+
+        tokenizer2 = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.Tokenizer2})
+        t5_encoder = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.TextEncoder2})
+
+        transformer_config = context.models.get_config(transformer)
+        assert isinstance(transformer_config, CheckpointConfigBase)
+
+        return FluxModelLoaderOutput(
+            transformer=TransformerField(transformer=transformer, loras=[]),
+            clip=CLIPField(tokenizer=tokenizer, text_encoder=clip_encoder, loras=[], skipped_layers=0),
+            t5_encoder=T5EncoderField(tokenizer=tokenizer2, text_encoder=t5_encoder),
+            vae=VAEField(vae=vae),
+            max_seq_len=max_seq_lengths[transformer_config.config_path],
+        )

invokeai/app/invocations/flux_text_encoder.py

@@ -1,4 +1,5 @@
-from typing import Literal
+from contextlib import ExitStack
+from typing import Iterator, Literal, Tuple
 
 import torch
 from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5Tokenizer
@@ -9,6 +10,10 @@ from invokeai.app.invocations.model import CLIPField, T5EncoderField
 from invokeai.app.invocations.primitives import FluxConditioningOutput
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.flux.modules.conditioner import HFEncoder
+from invokeai.backend.lora.conversions.flux_lora_constants import FLUX_LORA_CLIP_PREFIX
+from invokeai.backend.lora.lora_model_raw import LoRAModelRaw
+from invokeai.backend.lora.lora_patcher import LoRAPatcher
+from invokeai.backend.model_manager.config import ModelFormat
 from invokeai.backend.stable_diffusion.diffusion.conditioning_data import ConditioningFieldData, FLUXConditioningInfo
 
 
@@ -17,7 +22,7 @@ from invokeai.backend.stable_diffusion.diffusion.conditioning_data import Condit
     title="FLUX Text Encoding",
     tags=["prompt", "conditioning", "flux"],
     category="conditioning",
-    version="1.0.0",
+    version="1.1.0",
     classification=Classification.Prototype,
 )
 class FluxTextEncoderInvocation(BaseInvocation):
@@ -66,6 +71,7 @@ class FluxTextEncoderInvocation(BaseInvocation):
 
             t5_encoder = HFEncoder(t5_text_encoder, t5_tokenizer, False, self.t5_max_seq_len)
 
+            context.util.signal_progress("Running T5 encoder")
             prompt_embeds = t5_encoder(prompt)
 
         assert isinstance(prompt_embeds, torch.Tensor)
@@ -78,15 +84,43 @@ class FluxTextEncoderInvocation(BaseInvocation):
         prompt = [self.prompt]
 
         with (
-            clip_text_encoder_info as clip_text_encoder,
+            clip_text_encoder_info.model_on_device() as (cached_weights, clip_text_encoder),
             clip_tokenizer_info as clip_tokenizer,
+            ExitStack() as exit_stack,
         ):
             assert isinstance(clip_text_encoder, CLIPTextModel)
             assert isinstance(clip_tokenizer, CLIPTokenizer)
 
+            clip_text_encoder_config = clip_text_encoder_info.config
+            assert clip_text_encoder_config is not None
+
+            # Apply LoRA models to the CLIP encoder.
+            # Note: We apply the LoRA after the transformer has been moved to its target device for faster patching.
+            if clip_text_encoder_config.format in [ModelFormat.Diffusers]:
+                # The model is non-quantized, so we can apply the LoRA weights directly into the model.
+                exit_stack.enter_context(
+                    LoRAPatcher.apply_lora_patches(
+                        model=clip_text_encoder,
+                        patches=self._clip_lora_iterator(context),
+                        prefix=FLUX_LORA_CLIP_PREFIX,
+                        cached_weights=cached_weights,
+                    )
+                )
+            else:
+                # There are currently no supported CLIP quantized models. Add support here if needed.
+                raise ValueError(f"Unsupported model format: {clip_text_encoder_config.format}")
+
             clip_encoder = HFEncoder(clip_text_encoder, clip_tokenizer, True, 77)
 
+            context.util.signal_progress("Running CLIP encoder")
             pooled_prompt_embeds = clip_encoder(prompt)
 
         assert isinstance(pooled_prompt_embeds, torch.Tensor)
         return pooled_prompt_embeds
+
+    def _clip_lora_iterator(self, context: InvocationContext) -> Iterator[Tuple[LoRAModelRaw, float]]:
+        for lora in self.clip.loras:
+            lora_info = context.models.load(lora.lora)
+            assert isinstance(lora_info.model, LoRAModelRaw)
+            yield (lora_info.model, lora.weight)
+            del lora_info

invokeai/app/invocations/flux_vae_decode.py

@@ -41,7 +41,8 @@ class FluxVaeDecodeInvocation(BaseInvocation, WithMetadata, WithBoard):
     def _vae_decode(self, vae_info: LoadedModel, latents: torch.Tensor) -> Image.Image:
         with vae_info as vae:
             assert isinstance(vae, AutoEncoder)
-            latents = latents.to(device=TorchDevice.choose_torch_device(), dtype=TorchDevice.choose_torch_dtype())
+            vae_dtype = next(iter(vae.parameters())).dtype
+            latents = latents.to(device=TorchDevice.choose_torch_device(), dtype=vae_dtype)
             img = vae.decode(latents)
 
         img = img.clamp(-1, 1)
@@ -53,6 +54,7 @@ class FluxVaeDecodeInvocation(BaseInvocation, WithMetadata, WithBoard):
     def invoke(self, context: InvocationContext) -> ImageOutput:
         latents = context.tensors.load(self.latents.latents_name)
         vae_info = context.models.load(self.vae.vae)
+        context.util.signal_progress("Running VAE")
         image = self._vae_decode(vae_info=vae_info, latents=latents)
 
         TorchDevice.empty_cache()

invokeai/app/invocations/flux_vae_encode.py

@@ -44,9 +44,8 @@ class FluxVaeEncodeInvocation(BaseInvocation):
         generator = torch.Generator(device=TorchDevice.choose_torch_device()).manual_seed(0)
         with vae_info as vae:
             assert isinstance(vae, AutoEncoder)
-            image_tensor = image_tensor.to(
-                device=TorchDevice.choose_torch_device(), dtype=TorchDevice.choose_torch_dtype()
-            )
+            vae_dtype = next(iter(vae.parameters())).dtype
+            image_tensor = image_tensor.to(device=TorchDevice.choose_torch_device(), dtype=vae_dtype)
             latents = vae.encode(image_tensor, sample=True, generator=generator)
             return latents
 
@@ -60,6 +59,7 @@ class FluxVaeEncodeInvocation(BaseInvocation):
         if image_tensor.dim() == 3:
             image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
 
+        context.util.signal_progress("Running VAE")
         latents = self.vae_encode(vae_info=vae_info, image_tensor=image_tensor)
 
         latents = latents.to("cpu")

invokeai/app/invocations/image_to_latents.py

@@ -117,6 +117,7 @@ class ImageToLatentsInvocation(BaseInvocation):
         if image_tensor.dim() == 3:
             image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
 
+        context.util.signal_progress("Running VAE encoder")
         latents = self.vae_encode(
             vae_info=vae_info, upcast=self.fp32, tiled=self.tiled, image_tensor=image_tensor, tile_size=self.tile_size
         )

invokeai/app/invocations/ip_adapter.py

@@ -9,6 +9,7 @@ from invokeai.app.invocations.fields import FieldDescriptions, InputField, Outpu
 from invokeai.app.invocations.model import ModelIdentifierField
 from invokeai.app.invocations.primitives import ImageField
 from invokeai.app.invocations.util import validate_begin_end_step, validate_weights
+from invokeai.app.services.model_records.model_records_base import ModelRecordChanges
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
@@ -17,6 +18,12 @@ from invokeai.backend.model_manager.config import (
     IPAdapterInvokeAIConfig,
     ModelType,
 )
+from invokeai.backend.model_manager.starter_models import (
+    StarterModel,
+    clip_vit_l_image_encoder,
+    ip_adapter_sd_image_encoder,
+    ip_adapter_sdxl_image_encoder,
+)
 
 
 class IPAdapterField(BaseModel):
@@ -55,10 +62,14 @@ class IPAdapterOutput(BaseInvocationOutput):
     ip_adapter: IPAdapterField = OutputField(description=FieldDescriptions.ip_adapter, title="IP-Adapter")
 
 
-CLIP_VISION_MODEL_MAP = {"ViT-H": "ip_adapter_sd_image_encoder", "ViT-G": "ip_adapter_sdxl_image_encoder"}
+CLIP_VISION_MODEL_MAP: dict[Literal["ViT-L", "ViT-H", "ViT-G"], StarterModel] = {
+    "ViT-L": clip_vit_l_image_encoder,
+    "ViT-H": ip_adapter_sd_image_encoder,
+    "ViT-G": ip_adapter_sdxl_image_encoder,
+}
 
 
-@invocation("ip_adapter", title="IP-Adapter", tags=["ip_adapter", "control"], category="ip_adapter", version="1.4.1")
+@invocation("ip_adapter", title="IP-Adapter", tags=["ip_adapter", "control"], category="ip_adapter", version="1.5.0")
 class IPAdapterInvocation(BaseInvocation):
     """Collects IP-Adapter info to pass to other nodes."""
 
@@ -70,7 +81,7 @@ class IPAdapterInvocation(BaseInvocation):
         ui_order=-1,
         ui_type=UIType.IPAdapterModel,
     )
-    clip_vision_model: Literal["ViT-H", "ViT-G"] = InputField(
+    clip_vision_model: Literal["ViT-H", "ViT-G", "ViT-L"] = InputField(
         description="CLIP Vision model to use. Overrides model settings. Mandatory for checkpoint models.",
         default="ViT-H",
         ui_order=2,
@@ -111,9 +122,11 @@ class IPAdapterInvocation(BaseInvocation):
             image_encoder_model_id = ip_adapter_info.image_encoder_model_id
             image_encoder_model_name = image_encoder_model_id.split("/")[-1].strip()
         else:
-            image_encoder_model_name = CLIP_VISION_MODEL_MAP[self.clip_vision_model]
+            image_encoder_starter_model = CLIP_VISION_MODEL_MAP[self.clip_vision_model]
+            image_encoder_model_id = image_encoder_starter_model.source
+            image_encoder_model_name = image_encoder_starter_model.name
 
-        image_encoder_model = self._get_image_encoder(context, image_encoder_model_name)
+        image_encoder_model = self.get_clip_image_encoder(context, image_encoder_model_id, image_encoder_model_name)
 
         if self.method == "style":
             if ip_adapter_info.base == "sd-1":
@@ -147,7 +160,10 @@ class IPAdapterInvocation(BaseInvocation):
             ),
         )
 
-    def _get_image_encoder(self, context: InvocationContext, image_encoder_model_name: str) -> AnyModelConfig:
+    @classmethod
+    def get_clip_image_encoder(
+        cls, context: InvocationContext, image_encoder_model_id: str, image_encoder_model_name: str
+    ) -> AnyModelConfig:
         image_encoder_models = context.models.search_by_attrs(
             name=image_encoder_model_name, base=BaseModelType.Any, type=ModelType.CLIPVision
         )
@@ -159,7 +175,11 @@ class IPAdapterInvocation(BaseInvocation):
             )
 
             installer = context._services.model_manager.install
-            job = installer.heuristic_import(f"InvokeAI/{image_encoder_model_name}")
+            # Note: We hard-code the type to CLIPVision here because if the model contains both a CLIPVision and a
+            # CLIPText model, the probe may treat it as a CLIPText model.
+            job = installer.heuristic_import(
+                image_encoder_model_id, ModelRecordChanges(name=image_encoder_model_name, type=ModelType.CLIPVision)
+            )
             installer.wait_for_job(job, timeout=600)  # Wait for up to 10 minutes
             image_encoder_models = context.models.search_by_attrs(
                 name=image_encoder_model_name, base=BaseModelType.Any, type=ModelType.CLIPVision

invokeai/app/invocations/latents_to_image.py

@@ -60,6 +60,7 @@ class LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
         vae_info = context.models.load(self.vae.vae)
         assert isinstance(vae_info.model, (AutoencoderKL, AutoencoderTiny))
         with SeamlessExt.static_patch_model(vae_info.model, self.vae.seamless_axes), vae_info as vae:
+            context.util.signal_progress("Running VAE decoder")
             assert isinstance(vae, (AutoencoderKL, AutoencoderTiny))
             latents = latents.to(vae.device)
             if self.fp32:

invokeai/app/invocations/mask.py

@@ -5,6 +5,7 @@ from PIL import Image
 from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, InvocationContext, invocation
 from invokeai.app.invocations.fields import ImageField, InputField, TensorField, WithBoard, WithMetadata
 from invokeai.app.invocations.primitives import ImageOutput, MaskOutput
+from invokeai.backend.image_util.util import pil_to_np
 
 
 @invocation(
@@ -148,3 +149,55 @@ class MaskTensorToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
         mask_pil = Image.fromarray(mask_np, mode="L")
         image_dto = context.images.save(image=mask_pil)
         return ImageOutput.build(image_dto)
+
+
+@invocation(
+    "apply_tensor_mask_to_image",
+    title="Apply Tensor Mask to Image",
+    tags=["mask"],
+    category="mask",
+    version="1.0.0",
+)
+class ApplyMaskTensorToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
+    """Applies a tensor mask to an image.
+
+    The image is converted to RGBA and the mask is applied to the alpha channel."""
+
+    mask: TensorField = InputField(description="The mask tensor to apply.")
+    image: ImageField = InputField(description="The image to apply the mask to.")
+    invert: bool = InputField(default=False, description="Whether to invert the mask.")
+
+    def invoke(self, context: InvocationContext) -> ImageOutput:
+        image = context.images.get_pil(self.image.image_name, mode="RGBA")
+        mask = context.tensors.load(self.mask.tensor_name)
+
+        # Squeeze the channel dimension if it exists.
+        if mask.dim() == 3:
+            mask = mask.squeeze(0)
+
+        # Ensure that the mask is binary.
+        if mask.dtype != torch.bool:
+            mask = mask > 0.5
+        mask_np = (mask.float() * 255).byte().cpu().numpy().astype(np.uint8)
+
+        if self.invert:
+            mask_np = 255 - mask_np
+
+        # Apply the mask only to the alpha channel where the original alpha is non-zero. This preserves the original
+        # image's transparency - else the transparent regions would end up as opaque black.
+
+        # Separate the image into R, G, B, and A channels
+        image_np = pil_to_np(image)
+        r, g, b, a = np.split(image_np, 4, axis=-1)
+
+        # Apply the mask to the alpha channel
+        new_alpha = np.where(a.squeeze() > 0, mask_np, a.squeeze())
+
+        # Stack the RGB channels with the modified alpha
+        masked_image_np = np.dstack([r.squeeze(), g.squeeze(), b.squeeze(), new_alpha])
+
+        # Convert back to an image (RGBA)
+        masked_image = Image.fromarray(masked_image_np.astype(np.uint8), "RGBA")
+        image_dto = context.images.save(image=masked_image)
+
+        return ImageOutput.build(image_dto)

invokeai/app/invocations/metadata.py

@@ -40,7 +40,7 @@ class IPAdapterMetadataField(BaseModel):
 
     image: ImageField = Field(description="The IP-Adapter image prompt.")
     ip_adapter_model: ModelIdentifierField = Field(description="The IP-Adapter model.")
-    clip_vision_model: Literal["ViT-H", "ViT-G"] = Field(description="The CLIP Vision model")
+    clip_vision_model: Literal["ViT-L", "ViT-H", "ViT-G"] = Field(description="The CLIP Vision model")
     method: Literal["full", "style", "composition"] = Field(description="Method to apply IP Weights with")
     weight: Union[float, list[float]] = Field(description="The weight given to the IP-Adapter")
     begin_step_percent: float = Field(description="When the IP-Adapter is first applied (% of total steps)")
@@ -147,6 +147,10 @@ GENERATION_MODES = Literal[
     "flux_img2img",
     "flux_inpaint",
     "flux_outpaint",
+    "sd3_txt2img",
+    "sd3_img2img",
+    "sd3_inpaint",
+    "sd3_outpaint",
 ]
 
 

invokeai/app/invocations/model.py

@@ -1,5 +1,5 @@
 import copy
-from typing import List, Literal, Optional
+from typing import List, Optional
 
 from pydantic import BaseModel, Field
 
@@ -13,11 +13,9 @@ from invokeai.app.invocations.baseinvocation import (
 from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIType
 from invokeai.app.services.shared.invocation_context import InvocationContext
 from invokeai.app.shared.models import FreeUConfig
-from invokeai.backend.flux.util import max_seq_lengths
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     BaseModelType,
-    CheckpointConfigBase,
     ModelType,
     SubModelType,
 )
@@ -139,78 +137,6 @@ class ModelIdentifierInvocation(BaseInvocation):
         return ModelIdentifierOutput(model=self.model)
 
 
-@invocation_output("flux_model_loader_output")
-class FluxModelLoaderOutput(BaseInvocationOutput):
-    """Flux base model loader output"""
-
-    transformer: TransformerField = OutputField(description=FieldDescriptions.transformer, title="Transformer")
-    clip: CLIPField = OutputField(description=FieldDescriptions.clip, title="CLIP")
-    t5_encoder: T5EncoderField = OutputField(description=FieldDescriptions.t5_encoder, title="T5 Encoder")
-    vae: VAEField = OutputField(description=FieldDescriptions.vae, title="VAE")
-    max_seq_len: Literal[256, 512] = OutputField(
-        description="The max sequence length to used for the T5 encoder. (256 for schnell transformer, 512 for dev transformer)",
-        title="Max Seq Length",
-    )
-
-
-@invocation(
-    "flux_model_loader",
-    title="Flux Main Model",
-    tags=["model", "flux"],
-    category="model",
-    version="1.0.4",
-    classification=Classification.Prototype,
-)
-class FluxModelLoaderInvocation(BaseInvocation):
-    """Loads a flux base model, outputting its submodels."""
-
-    model: ModelIdentifierField = InputField(
-        description=FieldDescriptions.flux_model,
-        ui_type=UIType.FluxMainModel,
-        input=Input.Direct,
-    )
-
-    t5_encoder_model: ModelIdentifierField = InputField(
-        description=FieldDescriptions.t5_encoder, ui_type=UIType.T5EncoderModel, input=Input.Direct, title="T5 Encoder"
-    )
-
-    clip_embed_model: ModelIdentifierField = InputField(
-        description=FieldDescriptions.clip_embed_model,
-        ui_type=UIType.CLIPEmbedModel,
-        input=Input.Direct,
-        title="CLIP Embed",
-    )
-
-    vae_model: ModelIdentifierField = InputField(
-        description=FieldDescriptions.vae_model, ui_type=UIType.FluxVAEModel, title="VAE"
-    )
-
-    def invoke(self, context: InvocationContext) -> FluxModelLoaderOutput:
-        for key in [self.model.key, self.t5_encoder_model.key, self.clip_embed_model.key, self.vae_model.key]:
-            if not context.models.exists(key):
-                raise ValueError(f"Unknown model: {key}")
-
-        transformer = self.model.model_copy(update={"submodel_type": SubModelType.Transformer})
-        vae = self.vae_model.model_copy(update={"submodel_type": SubModelType.VAE})
-
-        tokenizer = self.clip_embed_model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
-        clip_encoder = self.clip_embed_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
-
-        tokenizer2 = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.Tokenizer2})
-        t5_encoder = self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.TextEncoder2})
-
-        transformer_config = context.models.get_config(transformer)
-        assert isinstance(transformer_config, CheckpointConfigBase)
-
-        return FluxModelLoaderOutput(
-            transformer=TransformerField(transformer=transformer, loras=[]),
-            clip=CLIPField(tokenizer=tokenizer, text_encoder=clip_encoder, loras=[], skipped_layers=0),
-            t5_encoder=T5EncoderField(tokenizer=tokenizer2, text_encoder=t5_encoder),
-            vae=VAEField(vae=vae),
-            max_seq_len=max_seq_lengths[transformer_config.config_path],
-        )
-
-
 @invocation(
     "main_model_loader",
     title="Main Model",

invokeai/app/invocations/primitives.py

@@ -18,6 +18,7 @@ from invokeai.app.invocations.fields import (
     InputField,
     LatentsField,
     OutputField,
+    SD3ConditioningField,
     TensorField,
     UIComponent,
 )
@@ -426,6 +427,17 @@ class FluxConditioningOutput(BaseInvocationOutput):
         return cls(conditioning=FluxConditioningField(conditioning_name=conditioning_name))
 
 
+@invocation_output("sd3_conditioning_output")
+class SD3ConditioningOutput(BaseInvocationOutput):
+    """Base class for nodes that output a single SD3 conditioning tensor"""
+
+    conditioning: SD3ConditioningField = OutputField(description=FieldDescriptions.cond)
+
+    @classmethod
+    def build(cls, conditioning_name: str) -> "SD3ConditioningOutput":
+        return cls(conditioning=SD3ConditioningField(conditioning_name=conditioning_name))
+
+
 @invocation_output("conditioning_output")
 class ConditioningOutput(BaseInvocationOutput):
     """Base class for nodes that output a single conditioning tensor"""

invokeai/app/invocations/sd3_denoise.py

@@ -0,0 +1,338 @@
+from typing import Callable, Optional, Tuple
+
+import torch
+import torchvision.transforms as tv_transforms
+from diffusers.models.transformers.transformer_sd3 import SD3Transformer2DModel
+from torchvision.transforms.functional import resize as tv_resize
+from tqdm import tqdm
+
+from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
+from invokeai.app.invocations.fields import (
+    DenoiseMaskField,
+    FieldDescriptions,
+    Input,
+    InputField,
+    LatentsField,
+    SD3ConditioningField,
+    WithBoard,
+    WithMetadata,
+)
+from invokeai.app.invocations.model import TransformerField
+from invokeai.app.invocations.primitives import LatentsOutput
+from invokeai.app.invocations.sd3_text_encoder import SD3_T5_MAX_SEQ_LEN
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.flux.sampling_utils import clip_timestep_schedule_fractional
+from invokeai.backend.model_manager.config import BaseModelType
+from invokeai.backend.sd3.extensions.inpaint_extension import InpaintExtension
+from invokeai.backend.stable_diffusion.diffusers_pipeline import PipelineIntermediateState
+from invokeai.backend.stable_diffusion.diffusion.conditioning_data import SD3ConditioningInfo
+from invokeai.backend.util.devices import TorchDevice
+
+
+@invocation(
+    "sd3_denoise",
+    title="SD3 Denoise",
+    tags=["image", "sd3"],
+    category="image",
+    version="1.1.0",
+    classification=Classification.Prototype,
+)
+class SD3DenoiseInvocation(BaseInvocation, WithMetadata, WithBoard):
+    """Run denoising process with a SD3 model."""
+
+    # If latents is provided, this means we are doing image-to-image.
+    latents: Optional[LatentsField] = InputField(
+        default=None, description=FieldDescriptions.latents, input=Input.Connection
+    )
+    # denoise_mask is used for image-to-image inpainting. Only the masked region is modified.
+    denoise_mask: Optional[DenoiseMaskField] = InputField(
+        default=None, description=FieldDescriptions.denoise_mask, input=Input.Connection
+    )
+    denoising_start: float = InputField(default=0.0, ge=0, le=1, description=FieldDescriptions.denoising_start)
+    denoising_end: float = InputField(default=1.0, ge=0, le=1, description=FieldDescriptions.denoising_end)
+    transformer: TransformerField = InputField(
+        description=FieldDescriptions.sd3_model, input=Input.Connection, title="Transformer"
+    )
+    positive_conditioning: SD3ConditioningField = InputField(
+        description=FieldDescriptions.positive_cond, input=Input.Connection
+    )
+    negative_conditioning: SD3ConditioningField = InputField(
+        description=FieldDescriptions.negative_cond, input=Input.Connection
+    )
+    cfg_scale: float | list[float] = InputField(default=3.5, description=FieldDescriptions.cfg_scale, title="CFG Scale")
+    width: int = InputField(default=1024, multiple_of=16, description="Width of the generated image.")
+    height: int = InputField(default=1024, multiple_of=16, description="Height of the generated image.")
+    steps: int = InputField(default=10, gt=0, description=FieldDescriptions.steps)
+    seed: int = InputField(default=0, description="Randomness seed for reproducibility.")
+
+    @torch.no_grad()
+    def invoke(self, context: InvocationContext) -> LatentsOutput:
+        latents = self._run_diffusion(context)
+        latents = latents.detach().to("cpu")
+
+        name = context.tensors.save(tensor=latents)
+        return LatentsOutput.build(latents_name=name, latents=latents, seed=None)
+
+    def _prep_inpaint_mask(self, context: InvocationContext, latents: torch.Tensor) -> torch.Tensor | None:
+        """Prepare the inpaint mask.
+        - Loads the mask
+        - Resizes if necessary
+        - Casts to same device/dtype as latents
+
+        Args:
+            context (InvocationContext): The invocation context, for loading the inpaint mask.
+            latents (torch.Tensor): A latent image tensor. Used to determine the target shape, device, and dtype for the
+                inpaint mask.
+
+        Returns:
+            torch.Tensor | None: Inpaint mask. Values of 0.0 represent the regions to be fully denoised, and 1.0
+                represent the regions to be preserved.
+        """
+        if self.denoise_mask is None:
+            return None
+        mask = context.tensors.load(self.denoise_mask.mask_name)
+
+        # The input denoise_mask contains values in [0, 1], where 0.0 represents the regions to be fully denoised, and
+        # 1.0 represents the regions to be preserved.
+        # We invert the mask so that the regions to be preserved are 0.0 and the regions to be denoised are 1.0.
+        mask = 1.0 - mask
+
+        _, _, latent_height, latent_width = latents.shape
+        mask = tv_resize(
+            img=mask,
+            size=[latent_height, latent_width],
+            interpolation=tv_transforms.InterpolationMode.BILINEAR,
+            antialias=False,
+        )
+
+        mask = mask.to(device=latents.device, dtype=latents.dtype)
+        return mask
+
+    def _load_text_conditioning(
+        self,
+        context: InvocationContext,
+        conditioning_name: str,
+        joint_attention_dim: int,
+        dtype: torch.dtype,
+        device: torch.device,
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        # Load the conditioning data.
+        cond_data = context.conditioning.load(conditioning_name)
+        assert len(cond_data.conditionings) == 1
+        sd3_conditioning = cond_data.conditionings[0]
+        assert isinstance(sd3_conditioning, SD3ConditioningInfo)
+        sd3_conditioning = sd3_conditioning.to(dtype=dtype, device=device)
+
+        t5_embeds = sd3_conditioning.t5_embeds
+        if t5_embeds is None:
+            t5_embeds = torch.zeros(
+                (1, SD3_T5_MAX_SEQ_LEN, joint_attention_dim),
+                device=device,
+                dtype=dtype,
+            )
+
+        clip_prompt_embeds = torch.cat([sd3_conditioning.clip_l_embeds, sd3_conditioning.clip_g_embeds], dim=-1)
+        clip_prompt_embeds = torch.nn.functional.pad(
+            clip_prompt_embeds, (0, t5_embeds.shape[-1] - clip_prompt_embeds.shape[-1])
+        )
+
+        prompt_embeds = torch.cat([clip_prompt_embeds, t5_embeds], dim=-2)
+        pooled_prompt_embeds = torch.cat(
+            [sd3_conditioning.clip_l_pooled_embeds, sd3_conditioning.clip_g_pooled_embeds], dim=-1
+        )
+
+        return prompt_embeds, pooled_prompt_embeds
+
+    def _get_noise(
+        self,
+        num_samples: int,
+        num_channels_latents: int,
+        height: int,
+        width: int,
+        dtype: torch.dtype,
+        device: torch.device,
+        seed: int,
+    ) -> torch.Tensor:
+        # We always generate noise on the same device and dtype then cast to ensure consistency across devices/dtypes.
+        rand_device = "cpu"
+        rand_dtype = torch.float16
+
+        return torch.randn(
+            num_samples,
+            num_channels_latents,
+            int(height) // LATENT_SCALE_FACTOR,
+            int(width) // LATENT_SCALE_FACTOR,
+            device=rand_device,
+            dtype=rand_dtype,
+            generator=torch.Generator(device=rand_device).manual_seed(seed),
+        ).to(device=device, dtype=dtype)
+
+    def _prepare_cfg_scale(self, num_timesteps: int) -> list[float]:
+        """Prepare the CFG scale list.
+
+        Args:
+            num_timesteps (int): The number of timesteps in the scheduler. Could be different from num_steps depending
+            on the scheduler used (e.g. higher order schedulers).
+
+        Returns:
+            list[float]: _description_
+        """
+        if isinstance(self.cfg_scale, float):
+            cfg_scale = [self.cfg_scale] * num_timesteps
+        elif isinstance(self.cfg_scale, list):
+            assert len(self.cfg_scale) == num_timesteps
+            cfg_scale = self.cfg_scale
+        else:
+            raise ValueError(f"Invalid CFG scale type: {type(self.cfg_scale)}")
+
+        return cfg_scale
+
+    def _run_diffusion(
+        self,
+        context: InvocationContext,
+    ):
+        inference_dtype = TorchDevice.choose_torch_dtype()
+        device = TorchDevice.choose_torch_device()
+
+        transformer_info = context.models.load(self.transformer.transformer)
+
+        # Load/process the conditioning data.
+        # TODO(ryand): Make CFG optional.
+        do_classifier_free_guidance = True
+        pos_prompt_embeds, pos_pooled_prompt_embeds = self._load_text_conditioning(
+            context=context,
+            conditioning_name=self.positive_conditioning.conditioning_name,
+            joint_attention_dim=transformer_info.model.config.joint_attention_dim,
+            dtype=inference_dtype,
+            device=device,
+        )
+        neg_prompt_embeds, neg_pooled_prompt_embeds = self._load_text_conditioning(
+            context=context,
+            conditioning_name=self.negative_conditioning.conditioning_name,
+            joint_attention_dim=transformer_info.model.config.joint_attention_dim,
+            dtype=inference_dtype,
+            device=device,
+        )
+        # TODO(ryand): Support both sequential and batched CFG inference.
+        prompt_embeds = torch.cat([neg_prompt_embeds, pos_prompt_embeds], dim=0)
+        pooled_prompt_embeds = torch.cat([neg_pooled_prompt_embeds, pos_pooled_prompt_embeds], dim=0)
+
+        # Prepare the timestep schedule.
+        # We add an extra step to the end to account for the final timestep of 0.0.
+        timesteps: list[float] = torch.linspace(1, 0, self.steps + 1).tolist()
+        # Clip the timesteps schedule based on denoising_start and denoising_end.
+        timesteps = clip_timestep_schedule_fractional(timesteps, self.denoising_start, self.denoising_end)
+        total_steps = len(timesteps) - 1
+
+        # Prepare the CFG scale list.
+        cfg_scale = self._prepare_cfg_scale(total_steps)
+
+        # Load the input latents, if provided.
+        init_latents = context.tensors.load(self.latents.latents_name) if self.latents else None
+        if init_latents is not None:
+            init_latents = init_latents.to(device=device, dtype=inference_dtype)
+
+        # Generate initial latent noise.
+        num_channels_latents = transformer_info.model.config.in_channels
+        assert isinstance(num_channels_latents, int)
+        noise = self._get_noise(
+            num_samples=1,
+            num_channels_latents=num_channels_latents,
+            height=self.height,
+            width=self.width,
+            dtype=inference_dtype,
+            device=device,
+            seed=self.seed,
+        )
+
+        # Prepare input latent image.
+        if init_latents is not None:
+            # Noise the init_latents by the appropriate amount for the first timestep.
+            t_0 = timesteps[0]
+            latents = t_0 * noise + (1.0 - t_0) * init_latents
+        else:
+            # init_latents are not provided, so we are not doing image-to-image (i.e. we are starting from pure noise).
+            if self.denoising_start > 1e-5:
+                raise ValueError("denoising_start should be 0 when initial latents are not provided.")
+            latents = noise
+
+        # If len(timesteps) == 1, then short-circuit. We are just noising the input latents, but not taking any
+        # denoising steps.
+        if len(timesteps) <= 1:
+            return latents
+
+        # Prepare inpaint extension.
+        inpaint_mask = self._prep_inpaint_mask(context, latents)
+        inpaint_extension: InpaintExtension | None = None
+        if inpaint_mask is not None:
+            assert init_latents is not None
+            inpaint_extension = InpaintExtension(
+                init_latents=init_latents,
+                inpaint_mask=inpaint_mask,
+                noise=noise,
+            )
+
+        step_callback = self._build_step_callback(context)
+
+        step_callback(
+            PipelineIntermediateState(
+                step=0,
+                order=1,
+                total_steps=total_steps,
+                timestep=int(timesteps[0]),
+                latents=latents,
+            ),
+        )
+
+        with transformer_info.model_on_device() as (cached_weights, transformer):
+            assert isinstance(transformer, SD3Transformer2DModel)
+
+            # 6. Denoising loop
+            for step_idx, (t_curr, t_prev) in tqdm(list(enumerate(zip(timesteps[:-1], timesteps[1:], strict=True)))):
+                # Expand the latents if we are doing CFG.
+                latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents
+                # Expand the timestep to match the latent model input.
+                # Multiply by 1000 to match the default FlowMatchEulerDiscreteScheduler num_train_timesteps.
+                timestep = torch.tensor([t_curr * 1000], device=device).expand(latent_model_input.shape[0])
+
+                noise_pred = transformer(
+                    hidden_states=latent_model_input,
+                    timestep=timestep,
+                    encoder_hidden_states=prompt_embeds,
+                    pooled_projections=pooled_prompt_embeds,
+                    joint_attention_kwargs=None,
+                    return_dict=False,
+                )[0]
+
+                # Apply CFG.
+                if do_classifier_free_guidance:
+                    noise_pred_uncond, noise_pred_cond = noise_pred.chunk(2)
+                    noise_pred = noise_pred_uncond + cfg_scale[step_idx] * (noise_pred_cond - noise_pred_uncond)
+
+                # Compute the previous noisy sample x_t -> x_t-1.
+                latents_dtype = latents.dtype
+                latents = latents.to(dtype=torch.float32)
+                latents = latents + (t_prev - t_curr) * noise_pred
+                latents = latents.to(dtype=latents_dtype)
+
+                if inpaint_extension is not None:
+                    latents = inpaint_extension.merge_intermediate_latents_with_init_latents(latents, t_prev)
+
+                step_callback(
+                    PipelineIntermediateState(
+                        step=step_idx + 1,
+                        order=1,
+                        total_steps=total_steps,
+                        timestep=int(t_curr),
+                        latents=latents,
+                    ),
+                )
+
+        return latents
+
+    def _build_step_callback(self, context: InvocationContext) -> Callable[[PipelineIntermediateState], None]:
+        def step_callback(state: PipelineIntermediateState) -> None:
+            context.util.sd_step_callback(state, BaseModelType.StableDiffusion3)
+
+        return step_callback

invokeai/app/invocations/sd3_image_to_latents.py

@@ -0,0 +1,65 @@
+import einops
+import torch
+from diffusers.models.autoencoders.autoencoder_kl import AutoencoderKL
+
+from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.fields import (
+    FieldDescriptions,
+    ImageField,
+    Input,
+    InputField,
+    WithBoard,
+    WithMetadata,
+)
+from invokeai.app.invocations.model import VAEField
+from invokeai.app.invocations.primitives import LatentsOutput
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.model_manager.load.load_base import LoadedModel
+from invokeai.backend.stable_diffusion.diffusers_pipeline import image_resized_to_grid_as_tensor
+
+
+@invocation(
+    "sd3_i2l",
+    title="SD3 Image to Latents",
+    tags=["image", "latents", "vae", "i2l", "sd3"],
+    category="image",
+    version="1.0.0",
+    classification=Classification.Prototype,
+)
+class SD3ImageToLatentsInvocation(BaseInvocation, WithMetadata, WithBoard):
+    """Generates latents from an image."""
+
+    image: ImageField = InputField(description="The image to encode")
+    vae: VAEField = InputField(description=FieldDescriptions.vae, input=Input.Connection)
+
+    @staticmethod
+    def vae_encode(vae_info: LoadedModel, image_tensor: torch.Tensor) -> torch.Tensor:
+        with vae_info as vae:
+            assert isinstance(vae, AutoencoderKL)
+
+            vae.disable_tiling()
+
+            image_tensor = image_tensor.to(device=vae.device, dtype=vae.dtype)
+            with torch.inference_mode():
+                image_tensor_dist = vae.encode(image_tensor).latent_dist
+                # TODO: Use seed to make sampling reproducible.
+                latents: torch.Tensor = image_tensor_dist.sample().to(dtype=vae.dtype)
+
+            latents = vae.config.scaling_factor * latents
+
+        return latents
+
+    @torch.no_grad()
+    def invoke(self, context: InvocationContext) -> LatentsOutput:
+        image = context.images.get_pil(self.image.image_name)
+
+        image_tensor = image_resized_to_grid_as_tensor(image.convert("RGB"))
+        if image_tensor.dim() == 3:
+            image_tensor = einops.rearrange(image_tensor, "c h w -> 1 c h w")
+
+        vae_info = context.models.load(self.vae.vae)
+        latents = self.vae_encode(vae_info=vae_info, image_tensor=image_tensor)
+
+        latents = latents.to("cpu")
+        name = context.tensors.save(tensor=latents)
+        return LatentsOutput.build(latents_name=name, latents=latents, seed=None)

invokeai/app/invocations/sd3_latents_to_image.py

@@ -0,0 +1,74 @@
+from contextlib import nullcontext
+
+import torch
+from diffusers.models.autoencoders.autoencoder_kl import AutoencoderKL
+from einops import rearrange
+from PIL import Image
+
+from invokeai.app.invocations.baseinvocation import BaseInvocation, invocation
+from invokeai.app.invocations.fields import (
+    FieldDescriptions,
+    Input,
+    InputField,
+    LatentsField,
+    WithBoard,
+    WithMetadata,
+)
+from invokeai.app.invocations.model import VAEField
+from invokeai.app.invocations.primitives import ImageOutput
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.stable_diffusion.extensions.seamless import SeamlessExt
+from invokeai.backend.util.devices import TorchDevice
+
+
+@invocation(
+    "sd3_l2i",
+    title="SD3 Latents to Image",
+    tags=["latents", "image", "vae", "l2i", "sd3"],
+    category="latents",
+    version="1.3.0",
+)
+class SD3LatentsToImageInvocation(BaseInvocation, WithMetadata, WithBoard):
+    """Generates an image from latents."""
+
+    latents: LatentsField = InputField(
+        description=FieldDescriptions.latents,
+        input=Input.Connection,
+    )
+    vae: VAEField = InputField(
+        description=FieldDescriptions.vae,
+        input=Input.Connection,
+    )
+
+    @torch.no_grad()
+    def invoke(self, context: InvocationContext) -> ImageOutput:
+        latents = context.tensors.load(self.latents.latents_name)
+
+        vae_info = context.models.load(self.vae.vae)
+        assert isinstance(vae_info.model, (AutoencoderKL))
+        with SeamlessExt.static_patch_model(vae_info.model, self.vae.seamless_axes), vae_info as vae:
+            context.util.signal_progress("Running VAE")
+            assert isinstance(vae, (AutoencoderKL))
+            latents = latents.to(vae.device)
+
+            vae.disable_tiling()
+
+            tiling_context = nullcontext()
+
+            # clear memory as vae decode can request a lot
+            TorchDevice.empty_cache()
+
+            with torch.inference_mode(), tiling_context:
+                # copied from diffusers pipeline
+                latents = latents / vae.config.scaling_factor
+                img = vae.decode(latents, return_dict=False)[0]
+
+            img = img.clamp(-1, 1)
+            img = rearrange(img[0], "c h w -> h w c")  # noqa: F821
+            img_pil = Image.fromarray((127.5 * (img + 1.0)).byte().cpu().numpy())
+
+        TorchDevice.empty_cache()
+
+        image_dto = context.images.save(image=img_pil)
+
+        return ImageOutput.build(image_dto)

invokeai/app/invocations/sd3_model_loader.py

@@ -0,0 +1,108 @@
+from typing import Optional
+
+from invokeai.app.invocations.baseinvocation import (
+    BaseInvocation,
+    BaseInvocationOutput,
+    Classification,
+    invocation,
+    invocation_output,
+)
+from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField, OutputField, UIType
+from invokeai.app.invocations.model import CLIPField, ModelIdentifierField, T5EncoderField, TransformerField, VAEField
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.model_manager.config import SubModelType
+
+
+@invocation_output("sd3_model_loader_output")
+class Sd3ModelLoaderOutput(BaseInvocationOutput):
+    """SD3 base model loader output."""
+
+    transformer: TransformerField = OutputField(description=FieldDescriptions.transformer, title="Transformer")
+    clip_l: CLIPField = OutputField(description=FieldDescriptions.clip, title="CLIP L")
+    clip_g: CLIPField = OutputField(description=FieldDescriptions.clip, title="CLIP G")
+    t5_encoder: T5EncoderField = OutputField(description=FieldDescriptions.t5_encoder, title="T5 Encoder")
+    vae: VAEField = OutputField(description=FieldDescriptions.vae, title="VAE")
+
+
+@invocation(
+    "sd3_model_loader",
+    title="SD3 Main Model",
+    tags=["model", "sd3"],
+    category="model",
+    version="1.0.0",
+    classification=Classification.Prototype,
+)
+class Sd3ModelLoaderInvocation(BaseInvocation):
+    """Loads a SD3 base model, outputting its submodels."""
+
+    model: ModelIdentifierField = InputField(
+        description=FieldDescriptions.sd3_model,
+        ui_type=UIType.SD3MainModel,
+        input=Input.Direct,
+    )
+
+    t5_encoder_model: Optional[ModelIdentifierField] = InputField(
+        description=FieldDescriptions.t5_encoder,
+        ui_type=UIType.T5EncoderModel,
+        input=Input.Direct,
+        title="T5 Encoder",
+        default=None,
+    )
+
+    clip_l_model: Optional[ModelIdentifierField] = InputField(
+        description=FieldDescriptions.clip_embed_model,
+        ui_type=UIType.CLIPLEmbedModel,
+        input=Input.Direct,
+        title="CLIP L Encoder",
+        default=None,
+    )
+
+    clip_g_model: Optional[ModelIdentifierField] = InputField(
+        description=FieldDescriptions.clip_g_model,
+        ui_type=UIType.CLIPGEmbedModel,
+        input=Input.Direct,
+        title="CLIP G Encoder",
+        default=None,
+    )
+
+    vae_model: Optional[ModelIdentifierField] = InputField(
+        description=FieldDescriptions.vae_model, ui_type=UIType.VAEModel, title="VAE", default=None
+    )
+
+    def invoke(self, context: InvocationContext) -> Sd3ModelLoaderOutput:
+        transformer = self.model.model_copy(update={"submodel_type": SubModelType.Transformer})
+        vae = (
+            self.vae_model.model_copy(update={"submodel_type": SubModelType.VAE})
+            if self.vae_model
+            else self.model.model_copy(update={"submodel_type": SubModelType.VAE})
+        )
+        tokenizer_l = self.model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
+        clip_encoder_l = (
+            self.clip_l_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
+            if self.clip_l_model
+            else self.model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
+        )
+        tokenizer_g = self.model.model_copy(update={"submodel_type": SubModelType.Tokenizer2})
+        clip_encoder_g = (
+            self.clip_g_model.model_copy(update={"submodel_type": SubModelType.TextEncoder2})
+            if self.clip_g_model
+            else self.model.model_copy(update={"submodel_type": SubModelType.TextEncoder2})
+        )
+        tokenizer_t5 = (
+            self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.Tokenizer3})
+            if self.t5_encoder_model
+            else self.model.model_copy(update={"submodel_type": SubModelType.Tokenizer3})
+        )
+        t5_encoder = (
+            self.t5_encoder_model.model_copy(update={"submodel_type": SubModelType.TextEncoder3})
+            if self.t5_encoder_model
+            else self.model.model_copy(update={"submodel_type": SubModelType.TextEncoder3})
+        )
+
+        return Sd3ModelLoaderOutput(
+            transformer=TransformerField(transformer=transformer, loras=[]),
+            clip_l=CLIPField(tokenizer=tokenizer_l, text_encoder=clip_encoder_l, loras=[], skipped_layers=0),
+            clip_g=CLIPField(tokenizer=tokenizer_g, text_encoder=clip_encoder_g, loras=[], skipped_layers=0),
+            t5_encoder=T5EncoderField(tokenizer=tokenizer_t5, text_encoder=t5_encoder),
+            vae=VAEField(vae=vae),
+        )

invokeai/app/invocations/sd3_text_encoder.py

@@ -0,0 +1,201 @@
+from contextlib import ExitStack
+from typing import Iterator, Tuple
+
+import torch
+from transformers import (
+    CLIPTextModel,
+    CLIPTextModelWithProjection,
+    CLIPTokenizer,
+    T5EncoderModel,
+    T5Tokenizer,
+    T5TokenizerFast,
+)
+
+from invokeai.app.invocations.baseinvocation import BaseInvocation, Classification, invocation
+from invokeai.app.invocations.fields import FieldDescriptions, Input, InputField
+from invokeai.app.invocations.model import CLIPField, T5EncoderField
+from invokeai.app.invocations.primitives import SD3ConditioningOutput
+from invokeai.app.services.shared.invocation_context import InvocationContext
+from invokeai.backend.lora.conversions.flux_lora_constants import FLUX_LORA_CLIP_PREFIX
+from invokeai.backend.lora.lora_model_raw import LoRAModelRaw
+from invokeai.backend.lora.lora_patcher import LoRAPatcher
+from invokeai.backend.model_manager.config import ModelFormat
+from invokeai.backend.stable_diffusion.diffusion.conditioning_data import ConditioningFieldData, SD3ConditioningInfo
+
+# The SD3 T5 Max Sequence Length set based on the default in diffusers.
+SD3_T5_MAX_SEQ_LEN = 256
+
+
+@invocation(
+    "sd3_text_encoder",
+    title="SD3 Text Encoding",
+    tags=["prompt", "conditioning", "sd3"],
+    category="conditioning",
+    version="1.0.0",
+    classification=Classification.Prototype,
+)
+class Sd3TextEncoderInvocation(BaseInvocation):
+    """Encodes and preps a prompt for a SD3 image."""
+
+    clip_l: CLIPField = InputField(
+        title="CLIP L",
+        description=FieldDescriptions.clip,
+        input=Input.Connection,
+    )
+    clip_g: CLIPField = InputField(
+        title="CLIP G",
+        description=FieldDescriptions.clip,
+        input=Input.Connection,
+    )
+
+    # The SD3 models were trained with text encoder dropout, so the T5 encoder can be omitted to save time/memory.
+    t5_encoder: T5EncoderField | None = InputField(
+        title="T5Encoder",
+        default=None,
+        description=FieldDescriptions.t5_encoder,
+        input=Input.Connection,
+    )
+    prompt: str = InputField(description="Text prompt to encode.")
+
+    @torch.no_grad()
+    def invoke(self, context: InvocationContext) -> SD3ConditioningOutput:
+        # Note: The text encoding model are run in separate functions to ensure that all model references are locally
+        # scoped. This ensures that earlier models can be freed and gc'd before loading later models (if necessary).
+
+        clip_l_embeddings, clip_l_pooled_embeddings = self._clip_encode(context, self.clip_l)
+        clip_g_embeddings, clip_g_pooled_embeddings = self._clip_encode(context, self.clip_g)
+
+        t5_embeddings: torch.Tensor | None = None
+        if self.t5_encoder is not None:
+            t5_embeddings = self._t5_encode(context, SD3_T5_MAX_SEQ_LEN)
+
+        conditioning_data = ConditioningFieldData(
+            conditionings=[
+                SD3ConditioningInfo(
+                    clip_l_embeds=clip_l_embeddings,
+                    clip_l_pooled_embeds=clip_l_pooled_embeddings,
+                    clip_g_embeds=clip_g_embeddings,
+                    clip_g_pooled_embeds=clip_g_pooled_embeddings,
+                    t5_embeds=t5_embeddings,
+                )
+            ]
+        )
+
+        conditioning_name = context.conditioning.save(conditioning_data)
+        return SD3ConditioningOutput.build(conditioning_name)
+
+    def _t5_encode(self, context: InvocationContext, max_seq_len: int) -> torch.Tensor:
+        assert self.t5_encoder is not None
+        t5_tokenizer_info = context.models.load(self.t5_encoder.tokenizer)
+        t5_text_encoder_info = context.models.load(self.t5_encoder.text_encoder)
+
+        prompt = [self.prompt]
+
+        with (
+            t5_text_encoder_info as t5_text_encoder,
+            t5_tokenizer_info as t5_tokenizer,
+        ):
+            context.util.signal_progress("Running T5 encoder")
+            assert isinstance(t5_text_encoder, T5EncoderModel)
+            assert isinstance(t5_tokenizer, (T5Tokenizer, T5TokenizerFast))
+
+            text_inputs = t5_tokenizer(
+                prompt,
+                padding="max_length",
+                max_length=max_seq_len,
+                truncation=True,
+                add_special_tokens=True,
+                return_tensors="pt",
+            )
+            text_input_ids = text_inputs.input_ids
+            untruncated_ids = t5_tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+            assert isinstance(text_input_ids, torch.Tensor)
+            assert isinstance(untruncated_ids, torch.Tensor)
+            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
+                text_input_ids, untruncated_ids
+            ):
+                removed_text = t5_tokenizer.batch_decode(untruncated_ids[:, max_seq_len - 1 : -1])
+                context.logger.warning(
+                    "The following part of your input was truncated because `max_sequence_length` is set to "
+                    f" {max_seq_len} tokens: {removed_text}"
+                )
+
+            prompt_embeds = t5_text_encoder(text_input_ids.to(t5_text_encoder.device))[0]
+
+        assert isinstance(prompt_embeds, torch.Tensor)
+        return prompt_embeds
+
+    def _clip_encode(
+        self, context: InvocationContext, clip_model: CLIPField, tokenizer_max_length: int = 77
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        clip_tokenizer_info = context.models.load(clip_model.tokenizer)
+        clip_text_encoder_info = context.models.load(clip_model.text_encoder)
+
+        prompt = [self.prompt]
+
+        with (
+            clip_text_encoder_info.model_on_device() as (cached_weights, clip_text_encoder),
+            clip_tokenizer_info as clip_tokenizer,
+            ExitStack() as exit_stack,
+        ):
+            context.util.signal_progress("Running CLIP encoder")
+            assert isinstance(clip_text_encoder, (CLIPTextModel, CLIPTextModelWithProjection))
+            assert isinstance(clip_tokenizer, CLIPTokenizer)
+
+            clip_text_encoder_config = clip_text_encoder_info.config
+            assert clip_text_encoder_config is not None
+
+            # Apply LoRA models to the CLIP encoder.
+            # Note: We apply the LoRA after the transformer has been moved to its target device for faster patching.
+            if clip_text_encoder_config.format in [ModelFormat.Diffusers]:
+                # The model is non-quantized, so we can apply the LoRA weights directly into the model.
+                exit_stack.enter_context(
+                    LoRAPatcher.apply_lora_patches(
+                        model=clip_text_encoder,
+                        patches=self._clip_lora_iterator(context, clip_model),
+                        prefix=FLUX_LORA_CLIP_PREFIX,
+                        cached_weights=cached_weights,
+                    )
+                )
+            else:
+                # There are currently no supported CLIP quantized models. Add support here if needed.
+                raise ValueError(f"Unsupported model format: {clip_text_encoder_config.format}")
+
+            clip_text_encoder = clip_text_encoder.eval().requires_grad_(False)
+
+            text_inputs = clip_tokenizer(
+                prompt,
+                padding="max_length",
+                max_length=tokenizer_max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+
+            text_input_ids = text_inputs.input_ids
+            untruncated_ids = clip_tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+            assert isinstance(text_input_ids, torch.Tensor)
+            assert isinstance(untruncated_ids, torch.Tensor)
+            if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
+                text_input_ids, untruncated_ids
+            ):
+                removed_text = clip_tokenizer.batch_decode(untruncated_ids[:, tokenizer_max_length - 1 : -1])
+                context.logger.warning(
+                    "The following part of your input was truncated because CLIP can only handle sequences up to"
+                    f" {tokenizer_max_length} tokens: {removed_text}"
+                )
+            prompt_embeds = clip_text_encoder(
+                input_ids=text_input_ids.to(clip_text_encoder.device), output_hidden_states=True
+            )
+            pooled_prompt_embeds = prompt_embeds[0]
+            prompt_embeds = prompt_embeds.hidden_states[-2]
+
+            return prompt_embeds, pooled_prompt_embeds
+
+    def _clip_lora_iterator(
+        self, context: InvocationContext, clip_model: CLIPField
+    ) -> Iterator[Tuple[LoRAModelRaw, float]]:
+        for lora in clip_model.loras:
+            lora_info = context.models.load(lora.lora)
+            assert isinstance(lora_info.model, LoRAModelRaw)
+            yield (lora_info.model, lora.weight)
+            del lora_info

invokeai/app/invocations/segment_anything.py

@@ -1,9 +1,11 @@
+from enum import Enum
 from pathlib import Path
 from typing import Literal
 
 import numpy as np
 import torch
 from PIL import Image
+from pydantic import BaseModel, Field
 from transformers import AutoModelForMaskGeneration, AutoProcessor
 from transformers.models.sam import SamModel
 from transformers.models.sam.processing_sam import SamProcessor
@@ -23,12 +25,31 @@ SEGMENT_ANYTHING_MODEL_IDS: dict[SegmentAnythingModelKey, str] = {
 }
 
 
+class SAMPointLabel(Enum):
+    negative = -1
+    neutral = 0
+    positive = 1
+
+
+class SAMPoint(BaseModel):
+    x: int = Field(..., description="The x-coordinate of the point")
+    y: int = Field(..., description="The y-coordinate of the point")
+    label: SAMPointLabel = Field(..., description="The label of the point")
+
+
+class SAMPointsField(BaseModel):
+    points: list[SAMPoint] = Field(..., description="The points of the object")
+
+    def to_list(self) -> list[list[int]]:
+        return [[point.x, point.y, point.label.value] for point in self.points]
+
+
 @invocation(
     "segment_anything",
     title="Segment Anything",
     tags=["prompt", "segmentation"],
     category="segmentation",
-    version="1.0.0",
+    version="1.1.0",
 )
 class SegmentAnythingInvocation(BaseInvocation):
     """Runs a Segment Anything Model."""
@@ -40,7 +61,13 @@ class SegmentAnythingInvocation(BaseInvocation):
 
     model: SegmentAnythingModelKey = InputField(description="The Segment Anything model to use.")
     image: ImageField = InputField(description="The image to segment.")
-    bounding_boxes: list[BoundingBoxField] = InputField(description="The bounding boxes to prompt the SAM model with.")
+    bounding_boxes: list[BoundingBoxField] | None = InputField(
+        default=None, description="The bounding boxes to prompt the SAM model with."
+    )
+    point_lists: list[SAMPointsField] | None = InputField(
+        default=None,
+        description="The list of point lists to prompt the SAM model with. Each list of points represents a single object.",
+    )
     apply_polygon_refinement: bool = InputField(
         description="Whether to apply polygon refinement to the masks. This will smooth the edges of the masks slightly and ensure that each mask consists of a single closed polygon (before merging).",
         default=True,
@@ -55,7 +82,12 @@ class SegmentAnythingInvocation(BaseInvocation):
         # The models expect a 3-channel RGB image.
         image_pil = context.images.get_pil(self.image.image_name, mode="RGB")
 
-        if len(self.bounding_boxes) == 0:
+        if self.point_lists is not None and self.bounding_boxes is not None:
+            raise ValueError("Only one of point_lists or bounding_box can be provided.")
+
+        if (not self.bounding_boxes or len(self.bounding_boxes) == 0) and (
+            not self.point_lists or len(self.point_lists) == 0
+        ):
             combined_mask = torch.zeros(image_pil.size[::-1], dtype=torch.bool)
         else:
             masks = self._segment(context=context, image=image_pil)
@@ -83,14 +115,13 @@ class SegmentAnythingInvocation(BaseInvocation):
         assert isinstance(sam_processor, SamProcessor)
         return SegmentAnythingPipeline(sam_model=sam_model, sam_processor=sam_processor)
 
-    def _segment(
-        self,
-        context: InvocationContext,
-        image: Image.Image,
-    ) -> list[torch.Tensor]:
+    def _segment(self, context: InvocationContext, image: Image.Image) -> list[torch.Tensor]:
         """Use Segment Anything (SAM) to generate masks given an image + a set of bounding boxes."""
         # Convert the bounding boxes to the SAM input format.
-        sam_bounding_boxes = [[bb.x_min, bb.y_min, bb.x_max, bb.y_max] for bb in self.bounding_boxes]
+        sam_bounding_boxes = (
+            [[bb.x_min, bb.y_min, bb.x_max, bb.y_max] for bb in self.bounding_boxes] if self.bounding_boxes else None
+        )
+        sam_points = [p.to_list() for p in self.point_lists] if self.point_lists else None
 
         with (
             context.models.load_remote_model(
@@ -98,7 +129,7 @@ class SegmentAnythingInvocation(BaseInvocation):
             ) as sam_pipeline,
         ):
             assert isinstance(sam_pipeline, SegmentAnythingPipeline)
-            masks = sam_pipeline.segment(image=image, bounding_boxes=sam_bounding_boxes)
+            masks = sam_pipeline.segment(image=image, bounding_boxes=sam_bounding_boxes, point_lists=sam_points)
 
         masks = self._process_masks(masks)
         if self.apply_polygon_refinement:
@@ -141,9 +172,10 @@ class SegmentAnythingInvocation(BaseInvocation):
 
         return masks
 
-    def _filter_masks(self, masks: list[torch.Tensor], bounding_boxes: list[BoundingBoxField]) -> list[torch.Tensor]:
+    def _filter_masks(
+        self, masks: list[torch.Tensor], bounding_boxes: list[BoundingBoxField] | None
+    ) -> list[torch.Tensor]:
         """Filter the detected masks based on the specified mask filter."""
-        assert len(masks) == len(bounding_boxes)
 
         if self.mask_filter == "all":
             return masks
@@ -151,6 +183,10 @@ class SegmentAnythingInvocation(BaseInvocation):
             # Find the largest mask.
             return [max(masks, key=lambda x: float(x.sum()))]
         elif self.mask_filter == "highest_box_score":
+            assert (
+                bounding_boxes is not None
+            ), "Bounding boxes must be provided to use the 'highest_box_score' mask filter."
+            assert len(masks) == len(bounding_boxes)
             # Find the index of the bounding box with the highest score.
             # Note that we fallback to -1.0 if the score is None. This is mainly to satisfy the type checker. In most
             # cases the scores should all be non-None when using this filtering mode. That being said, -1.0 is a

invokeai/app/services/board_records/board_records_base.py

@@ -1,7 +1,8 @@
 from abc import ABC, abstractmethod
 
-from invokeai.app.services.board_records.board_records_common import BoardChanges, BoardRecord
+from invokeai.app.services.board_records.board_records_common import BoardChanges, BoardRecord, BoardRecordOrderBy
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 
 
 class BoardRecordStorageBase(ABC):
@@ -39,12 +40,19 @@ class BoardRecordStorageBase(ABC):
 
     @abstractmethod
     def get_many(
-        self, offset: int = 0, limit: int = 10, include_archived: bool = False
+        self,
+        order_by: BoardRecordOrderBy,
+        direction: SQLiteDirection,
+        offset: int = 0,
+        limit: int = 10,
+        include_archived: bool = False,
     ) -> OffsetPaginatedResults[BoardRecord]:
         """Gets many board records."""
         pass
 
     @abstractmethod
-    def get_all(self, include_archived: bool = False) -> list[BoardRecord]:
+    def get_all(
+        self, order_by: BoardRecordOrderBy, direction: SQLiteDirection, include_archived: bool = False
+    ) -> list[BoardRecord]:
         """Gets all board records."""
         pass

invokeai/app/services/board_records/board_records_common.py

@@ -1,8 +1,10 @@
 from datetime import datetime
+from enum import Enum
 from typing import Optional, Union
 
 from pydantic import BaseModel, Field
 
+from invokeai.app.util.metaenum import MetaEnum
 from invokeai.app.util.misc import get_iso_timestamp
 from invokeai.app.util.model_exclude_null import BaseModelExcludeNull
 
@@ -60,6 +62,13 @@ class BoardChanges(BaseModel, extra="forbid"):
     archived: Optional[bool] = Field(default=None, description="Whether or not the board is archived")
 
 
+class BoardRecordOrderBy(str, Enum, metaclass=MetaEnum):
+    """The order by options for board records"""
+
+    CreatedAt = "created_at"
+    Name = "board_name"
+
+
 class BoardRecordNotFoundException(Exception):
     """Raised when an board record is not found."""
 

invokeai/app/services/board_records/board_records_sqlite.py

@@ -8,10 +8,12 @@ from invokeai.app.services.board_records.board_records_common import (
     BoardRecord,
     BoardRecordDeleteException,
     BoardRecordNotFoundException,
+    BoardRecordOrderBy,
     BoardRecordSaveException,
     deserialize_board_record,
 )
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 from invokeai.app.services.shared.sqlite.sqlite_database import SqliteDatabase
 from invokeai.app.util.misc import uuid_string
 
@@ -144,7 +146,12 @@ class SqliteBoardRecordStorage(BoardRecordStorageBase):
         return self.get(board_id)
 
     def get_many(
-        self, offset: int = 0, limit: int = 10, include_archived: bool = False
+        self,
+        order_by: BoardRecordOrderBy,
+        direction: SQLiteDirection,
+        offset: int = 0,
+        limit: int = 10,
+        include_archived: bool = False,
     ) -> OffsetPaginatedResults[BoardRecord]:
         try:
             self._lock.acquire()
@@ -154,17 +161,16 @@ class SqliteBoardRecordStorage(BoardRecordStorageBase):
                 SELECT *
                 FROM boards
                 {archived_filter}
-                ORDER BY created_at DESC
+                ORDER BY {order_by} {direction}
                 LIMIT ? OFFSET ?;
             """
 
             # Determine archived filter condition
-            if include_archived:
-                archived_filter = ""
-            else:
-                archived_filter = "WHERE archived = 0"
+            archived_filter = "" if include_archived else "WHERE archived = 0"
 
-            final_query = base_query.format(archived_filter=archived_filter)
+            final_query = base_query.format(
+                archived_filter=archived_filter, order_by=order_by.value, direction=direction.value
+            )
 
             # Execute query to fetch boards
             self._cursor.execute(final_query, (limit, offset))
@@ -198,23 +204,32 @@ class SqliteBoardRecordStorage(BoardRecordStorageBase):
         finally:
             self._lock.release()
 
-    def get_all(self, include_archived: bool = False) -> list[BoardRecord]:
+    def get_all(
+        self, order_by: BoardRecordOrderBy, direction: SQLiteDirection, include_archived: bool = False
+    ) -> list[BoardRecord]:
         try:
             self._lock.acquire()
 
-            base_query = """
-                SELECT *
-                FROM boards
-                {archived_filter}
-                ORDER BY created_at DESC
-            """
-
-            if include_archived:
-                archived_filter = ""
+            if order_by == BoardRecordOrderBy.Name:
+                base_query = """
+                    SELECT *
+                    FROM boards
+                    {archived_filter}
+                    ORDER BY LOWER(board_name) {direction}
+                """
             else:
-                archived_filter = "WHERE archived = 0"
+                base_query = """
+                    SELECT *
+                    FROM boards
+                    {archived_filter}
+                    ORDER BY {order_by} {direction}
+                """
 
-            final_query = base_query.format(archived_filter=archived_filter)
+            archived_filter = "" if include_archived else "WHERE archived = 0"
+
+            final_query = base_query.format(
+                archived_filter=archived_filter, order_by=order_by.value, direction=direction.value
+            )
 
             self._cursor.execute(final_query)
 

invokeai/app/services/boards/boards_base.py

@@ -1,8 +1,9 @@
 from abc import ABC, abstractmethod
 
-from invokeai.app.services.board_records.board_records_common import BoardChanges
+from invokeai.app.services.board_records.board_records_common import BoardChanges, BoardRecordOrderBy
 from invokeai.app.services.boards.boards_common import BoardDTO
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 
 
 class BoardServiceABC(ABC):
@@ -43,12 +44,19 @@ class BoardServiceABC(ABC):
 
     @abstractmethod
     def get_many(
-        self, offset: int = 0, limit: int = 10, include_archived: bool = False
+        self,
+        order_by: BoardRecordOrderBy,
+        direction: SQLiteDirection,
+        offset: int = 0,
+        limit: int = 10,
+        include_archived: bool = False,
     ) -> OffsetPaginatedResults[BoardDTO]:
         """Gets many boards."""
         pass
 
     @abstractmethod
-    def get_all(self, include_archived: bool = False) -> list[BoardDTO]:
+    def get_all(
+        self, order_by: BoardRecordOrderBy, direction: SQLiteDirection, include_archived: bool = False
+    ) -> list[BoardDTO]:
         """Gets all boards."""
         pass

invokeai/app/services/boards/boards_default.py

@@ -1,8 +1,9 @@
-from invokeai.app.services.board_records.board_records_common import BoardChanges
+from invokeai.app.services.board_records.board_records_common import BoardChanges, BoardRecordOrderBy
 from invokeai.app.services.boards.boards_base import BoardServiceABC
 from invokeai.app.services.boards.boards_common import BoardDTO, board_record_to_dto
 from invokeai.app.services.invoker import Invoker
 from invokeai.app.services.shared.pagination import OffsetPaginatedResults
+from invokeai.app.services.shared.sqlite.sqlite_common import SQLiteDirection
 
 
 class BoardService(BoardServiceABC):
@@ -47,9 +48,16 @@ class BoardService(BoardServiceABC):
         self.__invoker.services.board_records.delete(board_id)
 
     def get_many(
-        self, offset: int = 0, limit: int = 10, include_archived: bool = False
+        self,
+        order_by: BoardRecordOrderBy,
+        direction: SQLiteDirection,
+        offset: int = 0,
+        limit: int = 10,
+        include_archived: bool = False,
     ) -> OffsetPaginatedResults[BoardDTO]:
-        board_records = self.__invoker.services.board_records.get_many(offset, limit, include_archived)
+        board_records = self.__invoker.services.board_records.get_many(
+            order_by, direction, offset, limit, include_archived
+        )
         board_dtos = []
         for r in board_records.items:
             cover_image = self.__invoker.services.image_records.get_most_recent_image_for_board(r.board_id)
@@ -63,8 +71,10 @@ class BoardService(BoardServiceABC):
 
         return OffsetPaginatedResults[BoardDTO](items=board_dtos, offset=offset, limit=limit, total=len(board_dtos))
 
-    def get_all(self, include_archived: bool = False) -> list[BoardDTO]:
-        board_records = self.__invoker.services.board_records.get_all(include_archived)
+    def get_all(
+        self, order_by: BoardRecordOrderBy, direction: SQLiteDirection, include_archived: bool = False
+    ) -> list[BoardDTO]:
+        board_records = self.__invoker.services.board_records.get_all(order_by, direction, include_archived)
         board_dtos = []
         for r in board_records:
             cover_image = self.__invoker.services.image_records.get_most_recent_image_for_board(r.board_id)

invokeai/app/services/config/config_default.py

@@ -250,13 +250,13 @@ class InvokeAIAppConfig(BaseSettings):
             )
 
             if as_example:
-                file.write(
-                    "# This is an example file with default and example settings. Use the values here as a baseline.\n\n"
-                )
+                file.write("# This is an example file with default and example settings.\n")
+                file.write("# You should not copy this whole file into your config.\n")
+                file.write("# Only add the settings you need to change to your config file.\n\n")
             file.write("# Internal metadata - do not edit:\n")
             file.write(yaml.dump(meta_dict, sort_keys=False))
             file.write("\n")
-            file.write("# Put user settings here - see https://invoke-ai.github.io/InvokeAI/features/CONFIGURATION/:\n")
+            file.write("# Put user settings here - see https://invoke-ai.github.io/InvokeAI/configuration/:\n")
             if len(config_dict) > 0:
                 file.write(yaml.dump(config_dict, sort_keys=False))
 

invokeai/app/services/image_files/image_files_disk.py

@@ -110,15 +110,26 @@ class DiskImageFileStorage(ImageFileStorageBase):
         except Exception as e:
             raise ImageFileDeleteException from e
 
-    # 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
+        base_folder = self.__thumbnails_folder if thumbnail else self.__output_folder
+        filename = get_thumbnail_name(image_name) if thumbnail else image_name
 
-        if thumbnail:
-            thumbnail_name = get_thumbnail_name(image_name)
-            path = self.__thumbnails_folder / thumbnail_name
+        # Strip any path information from the filename
+        basename = Path(filename).name
 
-        return path
+        if basename != filename:
+            raise ValueError("Invalid image name, potential directory traversal detected")
+
+        image_path = base_folder / basename
+
+        # Ensure the image path is within the base folder to prevent directory traversal
+        resolved_base = base_folder.resolve()
+        resolved_image_path = image_path.resolve()
+
+        if not resolved_image_path.is_relative_to(resolved_base):
+            raise ValueError("Image path outside outputs folder, potential directory traversal detected")
+
+        return resolved_image_path
 
     def validate_path(self, path: Union[str, Path]) -> bool:
         """Validates the path given for an image or thumbnail."""

invokeai/app/services/model_records/model_records_base.py

@@ -15,6 +15,7 @@ from invokeai.app.util.model_exclude_null import BaseModelExcludeNull
 from invokeai.backend.model_manager.config import (
     AnyModelConfig,
     BaseModelType,
+    ClipVariantType,
     ControlAdapterDefaultSettings,
     MainModelDefaultSettings,
     ModelFormat,
@@ -85,7 +86,7 @@ class ModelRecordChanges(BaseModelExcludeNull):
 
     # Checkpoint-specific changes
     # TODO(MM2): Should we expose these? Feels footgun-y...
-    variant: Optional[ModelVariantType] = Field(description="The variant of the model.", default=None)
+    variant: Optional[ModelVariantType | ClipVariantType] = Field(description="The variant of the model.", default=None)
     prediction_type: Optional[SchedulerPredictionType] = Field(
         description="The prediction type of the model.", default=None
     )

invokeai/app/services/shared/invocation_context.py

@@ -1,3 +1,4 @@
+from copy import deepcopy
 from dataclasses import dataclass
 from pathlib import Path
 from typing import TYPE_CHECKING, Callable, Optional, Union
@@ -159,6 +160,10 @@ class LoggerInterface(InvocationContextInterface):
 
 
 class ImagesInterface(InvocationContextInterface):
+    def __init__(self, services: InvocationServices, data: InvocationContextData, util: "UtilInterface") -> None:
+        super().__init__(services, data)
+        self._util = util
+
     def save(
         self,
         image: Image,
@@ -185,6 +190,8 @@ class ImagesInterface(InvocationContextInterface):
             The saved image DTO.
         """
 
+        self._util.signal_progress("Saving image")
+
         # If `metadata` is provided directly, use that. Else, use the metadata provided by `WithMetadata`, falling back to None.
         metadata_ = None
         if metadata:
@@ -221,7 +228,7 @@ class ImagesInterface(InvocationContextInterface):
         )
 
     def get_pil(self, image_name: str, mode: IMAGE_MODES | None = None) -> Image:
-        """Gets an image as a PIL Image object.
+        """Gets an image as a PIL Image object. This method returns a copy of the image.
 
         Args:
             image_name: The name of the image to get.
@@ -233,11 +240,15 @@ class ImagesInterface(InvocationContextInterface):
         image = self._services.images.get_pil_image(image_name)
         if mode and mode != image.mode:
             try:
+                # convert makes a copy!
                 image = image.convert(mode)
             except ValueError:
                 self._services.logger.warning(
                     f"Could not convert image from {image.mode} to {mode}. Using original mode instead."
                 )
+        else:
+            # copy the image to prevent the user from modifying the original
+            image = image.copy()
         return image
 
     def get_metadata(self, image_name: str) -> Optional[MetadataField]:
@@ -290,15 +301,15 @@ class TensorsInterface(InvocationContextInterface):
         return name
 
     def load(self, name: str) -> Tensor:
-        """Loads a tensor by name.
+        """Loads a tensor by name. This method returns a copy of the tensor.
 
         Args:
             name: The name of the tensor to load.
 
         Returns:
-            The loaded tensor.
+            The tensor.
         """
-        return self._services.tensors.load(name)
+        return self._services.tensors.load(name).clone()
 
 
 class ConditioningInterface(InvocationContextInterface):
@@ -316,21 +327,25 @@ class ConditioningInterface(InvocationContextInterface):
         return name
 
     def load(self, name: str) -> ConditioningFieldData:
-        """Loads conditioning data by name.
+        """Loads conditioning data by name. This method returns a copy of the conditioning data.
 
         Args:
             name: The name of the conditioning data to load.
 
         Returns:
-            The loaded conditioning data.
+            The conditioning data.
         """
 
-        return self._services.conditioning.load(name)
+        return deepcopy(self._services.conditioning.load(name))
 
 
 class ModelsInterface(InvocationContextInterface):
     """Common API for loading, downloading and managing models."""
 
+    def __init__(self, services: InvocationServices, data: InvocationContextData, util: "UtilInterface") -> None:
+        super().__init__(services, data)
+        self._util = util
+
     def exists(self, identifier: Union[str, "ModelIdentifierField"]) -> bool:
         """Check if a model exists.
 
@@ -363,11 +378,15 @@ class ModelsInterface(InvocationContextInterface):
 
         if isinstance(identifier, str):
             model = self._services.model_manager.store.get_model(identifier)
-            return self._services.model_manager.load.load_model(model, submodel_type)
         else:
-            _submodel_type = submodel_type or identifier.submodel_type
+            submodel_type = submodel_type or identifier.submodel_type
             model = self._services.model_manager.store.get_model(identifier.key)
-            return self._services.model_manager.load.load_model(model, _submodel_type)
+
+        message = f"Loading model {model.name}"
+        if submodel_type:
+            message += f" ({submodel_type.value})"
+        self._util.signal_progress(message)
+        return self._services.model_manager.load.load_model(model, submodel_type)
 
     def load_by_attrs(
         self, name: str, base: BaseModelType, type: ModelType, submodel_type: Optional[SubModelType] = None
@@ -392,6 +411,10 @@ class ModelsInterface(InvocationContextInterface):
         if len(configs) > 1:
             raise ValueError(f"More than one model found with name {name}, base {base}, and type {type}")
 
+        message = f"Loading model {name}"
+        if submodel_type:
+            message += f" ({submodel_type.value})"
+        self._util.signal_progress(message)
         return self._services.model_manager.load.load_model(configs[0], submodel_type)
 
     def get_config(self, identifier: Union[str, "ModelIdentifierField"]) -> AnyModelConfig:
@@ -462,6 +485,7 @@ class ModelsInterface(InvocationContextInterface):
         Returns:
             Path to the downloaded model
         """
+        self._util.signal_progress(f"Downloading model {source}")
         return self._services.model_manager.install.download_and_cache_model(source=source)
 
     def load_local_model(
@@ -484,6 +508,8 @@ class ModelsInterface(InvocationContextInterface):
         Returns:
             A LoadedModelWithoutConfig object.
         """
+
+        self._util.signal_progress(f"Loading model {model_path.name}")
         return self._services.model_manager.load.load_model_from_path(model_path=model_path, loader=loader)
 
     def load_remote_model(
@@ -509,6 +535,8 @@ class ModelsInterface(InvocationContextInterface):
             A LoadedModelWithoutConfig object.
         """
         model_path = self._services.model_manager.install.download_and_cache_model(source=str(source))
+
+        self._util.signal_progress(f"Loading model {source}")
         return self._services.model_manager.load.load_model_from_path(model_path=model_path, loader=loader)
 
 
@@ -702,12 +730,12 @@ def build_invocation_context(
     """
 
     logger = LoggerInterface(services=services, data=data)
-    images = ImagesInterface(services=services, data=data)
     tensors = TensorsInterface(services=services, data=data)
-    models = ModelsInterface(services=services, data=data)
     config = ConfigInterface(services=services, data=data)
     util = UtilInterface(services=services, data=data, is_canceled=is_canceled)
     conditioning = ConditioningInterface(services=services, data=data)
+    models = ModelsInterface(services=services, data=data, util=util)
+    images = ImagesInterface(services=services, data=data, util=util)
     boards = BoardsInterface(services=services, data=data)
 
     ctx = InvocationContext(

invokeai/app/services/workflow_records/default_workflows/FLUX Image to Image.json

@@ -38,12 +38,12 @@
   },
   "nodes": [
     {
-      "id": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
+      "id": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
       "type": "invocation",
       "data": {
-        "id": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
+        "id": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
         "type": "flux_denoise",
-        "version": "2.1.0",
+        "version": "3.0.0",
         "label": "",
         "notes": "",
         "isOpen": true,
@@ -77,11 +77,6 @@
             "label": "",
             "value": 1
           },
-          "trajectory_guidance_strength": {
-            "name": "trajectory_guidance_strength",
-            "label": "",
-            "value": 0.0
-          },
           "transformer": {
             "name": "transformer",
             "label": ""
@@ -118,8 +113,8 @@
         }
       },
       "position": {
-        "x": 1159.584057771928,
-        "y": -175.90561201366845
+        "x": 1176.8139201354052,
+        "y": -244.36724863022368
       }
     },
     {
@@ -201,14 +196,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "Model (dev variant recommended for Image-to-Image)",
-            "value": {
-              "key": "b4990a6c-0899-48e9-969b-d6f3801acc6a",
-              "hash": "random:aad8f7bc19ce76541dfb394b62a30f77722542b66e48064a9f25453263b45fba",
-              "name": "FLUX Dev (Quantized)_2",
-              "base": "flux",
-              "type": "main"
-            }
+            "label": "Model (dev variant recommended for Image-to-Image)"
           },
           "t5_encoder_model": {
             "name": "t5_encoder_model",
@@ -314,67 +302,67 @@
         }
       },
       "position": {
-        "x": 725.834098928012,
-        "y": 496.2710031089931
+        "x": 750.4061458984118,
+        "y": 279.2179215371294
       }
     }
   ],
   "edges": [
     {
-      "id": "reactflow__edge-eebd7252-0bd8-401a-bb26-2b8bc64892falatents-7e5172eb-48c1-44db-a770-8fd83e1435d1latents",
+      "id": "reactflow__edge-cd367e62-2b45-4118-b4ba-7c33e2e0b370latents-7e5172eb-48c1-44db-a770-8fd83e1435d1latents",
       "type": "default",
-      "source": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
+      "source": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
       "target": "7e5172eb-48c1-44db-a770-8fd83e1435d1",
       "sourceHandle": "latents",
       "targetHandle": "latents"
     },
     {
-      "id": "reactflow__edge-f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90transformer-eebd7252-0bd8-401a-bb26-2b8bc64892fatransformer",
+      "id": "reactflow__edge-4754c534-a5f3-4ad0-9382-7887985e668cvalue-cd367e62-2b45-4118-b4ba-7c33e2e0b370seed",
       "type": "default",
-      "source": "f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90",
-      "target": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
-      "sourceHandle": "transformer",
-      "targetHandle": "transformer"
+      "source": "4754c534-a5f3-4ad0-9382-7887985e668c",
+      "target": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
+      "sourceHandle": "value",
+      "targetHandle": "seed"
     },
     {
-      "id": "reactflow__edge-01f674f8-b3d1-4df1-acac-6cb8e0bfb63cconditioning-eebd7252-0bd8-401a-bb26-2b8bc64892fapositive_text_conditioning",
-      "type": "default",
-      "source": "01f674f8-b3d1-4df1-acac-6cb8e0bfb63c",
-      "target": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
-      "sourceHandle": "conditioning",
-      "targetHandle": "positive_text_conditioning"
-    },
-    {
-      "id": "reactflow__edge-2981a67c-480f-4237-9384-26b68dbf912blatents-eebd7252-0bd8-401a-bb26-2b8bc64892falatents",
+      "id": "reactflow__edge-2981a67c-480f-4237-9384-26b68dbf912bheight-cd367e62-2b45-4118-b4ba-7c33e2e0b370height",
       "type": "default",
       "source": "2981a67c-480f-4237-9384-26b68dbf912b",
-      "target": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
-      "sourceHandle": "latents",
-      "targetHandle": "latents"
-    },
-    {
-      "id": "reactflow__edge-2981a67c-480f-4237-9384-26b68dbf912bwidth-eebd7252-0bd8-401a-bb26-2b8bc64892fawidth",
-      "type": "default",
-      "source": "2981a67c-480f-4237-9384-26b68dbf912b",
-      "target": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
-      "sourceHandle": "width",
-      "targetHandle": "width"
-    },
-    {
-      "id": "reactflow__edge-2981a67c-480f-4237-9384-26b68dbf912bheight-eebd7252-0bd8-401a-bb26-2b8bc64892faheight",
-      "type": "default",
-      "source": "2981a67c-480f-4237-9384-26b68dbf912b",
-      "target": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
+      "target": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
       "sourceHandle": "height",
       "targetHandle": "height"
     },
     {
-      "id": "reactflow__edge-4754c534-a5f3-4ad0-9382-7887985e668cvalue-eebd7252-0bd8-401a-bb26-2b8bc64892faseed",
+      "id": "reactflow__edge-2981a67c-480f-4237-9384-26b68dbf912bwidth-cd367e62-2b45-4118-b4ba-7c33e2e0b370width",
       "type": "default",
-      "source": "4754c534-a5f3-4ad0-9382-7887985e668c",
-      "target": "eebd7252-0bd8-401a-bb26-2b8bc64892fa",
-      "sourceHandle": "value",
-      "targetHandle": "seed"
+      "source": "2981a67c-480f-4237-9384-26b68dbf912b",
+      "target": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
+      "sourceHandle": "width",
+      "targetHandle": "width"
+    },
+    {
+      "id": "reactflow__edge-01f674f8-b3d1-4df1-acac-6cb8e0bfb63cconditioning-cd367e62-2b45-4118-b4ba-7c33e2e0b370positive_text_conditioning",
+      "type": "default",
+      "source": "01f674f8-b3d1-4df1-acac-6cb8e0bfb63c",
+      "target": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
+      "sourceHandle": "conditioning",
+      "targetHandle": "positive_text_conditioning"
+    },
+    {
+      "id": "reactflow__edge-f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90transformer-cd367e62-2b45-4118-b4ba-7c33e2e0b370transformer",
+      "type": "default",
+      "source": "f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90",
+      "target": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
+      "sourceHandle": "transformer",
+      "targetHandle": "transformer"
+    },
+    {
+      "id": "reactflow__edge-2981a67c-480f-4237-9384-26b68dbf912blatents-cd367e62-2b45-4118-b4ba-7c33e2e0b370latents",
+      "type": "default",
+      "source": "2981a67c-480f-4237-9384-26b68dbf912b",
+      "target": "cd367e62-2b45-4118-b4ba-7c33e2e0b370",
+      "sourceHandle": "latents",
+      "targetHandle": "latents"
     },
     {
       "id": "reactflow__edge-f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90vae-2981a67c-480f-4237-9384-26b68dbf912bvae",

invokeai/app/services/workflow_records/default_workflows/Flux Text to Image.json

@@ -34,12 +34,12 @@
   },
   "nodes": [
     {
-      "id": "4ecda92d-ee0e-45ca-aa35-6e9410ac39b9",
+      "id": "0940bc54-21fb-4346-bc68-fca5724c2747",
       "type": "invocation",
       "data": {
-        "id": "4ecda92d-ee0e-45ca-aa35-6e9410ac39b9",
+        "id": "0940bc54-21fb-4346-bc68-fca5724c2747",
         "type": "flux_denoise",
-        "version": "2.1.0",
+        "version": "3.0.0",
         "label": "",
         "notes": "",
         "isOpen": true,
@@ -61,7 +61,7 @@
           },
           "denoise_mask": {
             "name": "denoise_mask",
-            "label": ""
+            "label": "Denoise Mask"
           },
           "denoising_start": {
             "name": "denoising_start",
@@ -73,11 +73,6 @@
             "label": "",
             "value": 1
           },
-          "trajectory_guidance_strength": {
-            "name": "trajectory_guidance_strength",
-            "label": "",
-            "value": 0
-          },
           "transformer": {
             "name": "transformer",
             "label": ""
@@ -114,8 +109,8 @@
         }
       },
       "position": {
-        "x": 1161.0101524413685,
-        "y": -223.33548695623742
+        "x": 1180.8001377784371,
+        "y": -219.96908055568326
       }
     },
     {
@@ -169,14 +164,7 @@
         "inputs": {
           "model": {
             "name": "model",
-            "label": "",
-            "value": {
-              "key": "b4990a6c-0899-48e9-969b-d6f3801acc6a",
-              "hash": "random:aad8f7bc19ce76541dfb394b62a30f77722542b66e48064a9f25453263b45fba",
-              "name": "FLUX Dev (Quantized)_2",
-              "base": "flux",
-              "type": "main"
-            }
+            "label": ""
           },
           "t5_encoder_model": {
             "name": "t5_encoder_model",
@@ -289,36 +277,36 @@
   ],
   "edges": [
     {
-      "id": "reactflow__edge-f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90transformer-4ecda92d-ee0e-45ca-aa35-6e9410ac39b9transformer",
+      "id": "reactflow__edge-0940bc54-21fb-4346-bc68-fca5724c2747latents-7e5172eb-48c1-44db-a770-8fd83e1435d1latents",
       "type": "default",
-      "source": "f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90",
-      "target": "4ecda92d-ee0e-45ca-aa35-6e9410ac39b9",
-      "sourceHandle": "transformer",
-      "targetHandle": "transformer"
+      "source": "0940bc54-21fb-4346-bc68-fca5724c2747",
+      "target": "7e5172eb-48c1-44db-a770-8fd83e1435d1",
+      "sourceHandle": "latents",
+      "targetHandle": "latents"
     },
     {
-      "id": "reactflow__edge-01f674f8-b3d1-4df1-acac-6cb8e0bfb63cconditioning-4ecda92d-ee0e-45ca-aa35-6e9410ac39b9positive_text_conditioning",
-      "type": "default",
-      "source": "01f674f8-b3d1-4df1-acac-6cb8e0bfb63c",
-      "target": "4ecda92d-ee0e-45ca-aa35-6e9410ac39b9",
-      "sourceHandle": "conditioning",
-      "targetHandle": "positive_text_conditioning"
-    },
-    {
-      "id": "reactflow__edge-4754c534-a5f3-4ad0-9382-7887985e668cvalue-4ecda92d-ee0e-45ca-aa35-6e9410ac39b9seed",
+      "id": "reactflow__edge-4754c534-a5f3-4ad0-9382-7887985e668cvalue-0940bc54-21fb-4346-bc68-fca5724c2747seed",
       "type": "default",
       "source": "4754c534-a5f3-4ad0-9382-7887985e668c",
-      "target": "4ecda92d-ee0e-45ca-aa35-6e9410ac39b9",
+      "target": "0940bc54-21fb-4346-bc68-fca5724c2747",
       "sourceHandle": "value",
       "targetHandle": "seed"
     },
     {
-      "id": "reactflow__edge-4ecda92d-ee0e-45ca-aa35-6e9410ac39b9latents-7e5172eb-48c1-44db-a770-8fd83e1435d1latents",
+      "id": "reactflow__edge-01f674f8-b3d1-4df1-acac-6cb8e0bfb63cconditioning-0940bc54-21fb-4346-bc68-fca5724c2747positive_text_conditioning",
       "type": "default",
-      "source": "4ecda92d-ee0e-45ca-aa35-6e9410ac39b9",
-      "target": "7e5172eb-48c1-44db-a770-8fd83e1435d1",
-      "sourceHandle": "latents",
-      "targetHandle": "latents"
+      "source": "01f674f8-b3d1-4df1-acac-6cb8e0bfb63c",
+      "target": "0940bc54-21fb-4346-bc68-fca5724c2747",
+      "sourceHandle": "conditioning",
+      "targetHandle": "positive_text_conditioning"
+    },
+    {
+      "id": "reactflow__edge-f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90transformer-0940bc54-21fb-4346-bc68-fca5724c2747transformer",
+      "type": "default",
+      "source": "f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90",
+      "target": "0940bc54-21fb-4346-bc68-fca5724c2747",
+      "sourceHandle": "transformer",
+      "targetHandle": "transformer"
     },
     {
       "id": "reactflow__edge-f8d9d7c8-9ed7-4bd7-9e42-ab0e89bfac90vae-7e5172eb-48c1-44db-a770-8fd83e1435d1vae",

invokeai/app/services/workflow_records/default_workflows/SD3.5 Text to Image.json

@@ -0,0 +1,382 @@
+{
+    "name": "SD3.5 Text to Image",
+   "author": "InvokeAI",
+    "description": "Sample text to image workflow for Stable Diffusion 3.5",
+    "version": "1.0.0",
+    "contact": "invoke@invoke.ai",
+    "tags": "text2image, SD3.5, default",
+  "notes": "",
+    "exposedFields": [
+      {
+        "nodeId": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "fieldName": "model"
+      },
+      {
+        "nodeId": "e17d34e7-6ed1-493c-9a85-4fcd291cb084",
+        "fieldName": "prompt"
+      }
+    ],
+    "meta": {
+      "version": "3.0.0",
+      "category": "default"
+    },
+    "id": "e3a51d6b-8208-4d6d-b187-fcfe8b32934c",
+    "nodes": [
+      {
+        "id": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "type": "invocation",
+        "data": {
+          "id": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+          "type": "sd3_model_loader",
+          "version": "1.0.0",
+          "label": "",
+          "notes": "",
+          "isOpen": true,
+          "isIntermediate": true,
+          "useCache": true,
+          "nodePack": "invokeai",
+          "inputs": {
+            "model": {
+              "name": "model",
+              "label": "",
+              "value": {
+                "key": "f7b20be9-92a8-4cfb-bca4-6c3b5535c10b",
+                "hash": "placeholder",
+                "name": "stable-diffusion-3.5-medium",
+                "base": "sd-3",
+                "type": "main"
+              }
+            },
+            "t5_encoder_model": {
+              "name": "t5_encoder_model",
+              "label": ""
+            },
+            "clip_l_model": {
+              "name": "clip_l_model",
+              "label": ""
+            },
+            "clip_g_model": {
+              "name": "clip_g_model",
+              "label": ""
+            },
+            "vae_model": {
+              "name": "vae_model",
+              "label": ""
+            }
+          }
+        },
+        "position": {
+          "x": -55.58689609637031,
+          "y": -111.53602444662268
+        }
+      },
+      {
+        "id": "f7e394ac-6394-4096-abcb-de0d346506b3",
+        "type": "invocation",
+        "data": {
+          "id": "f7e394ac-6394-4096-abcb-de0d346506b3",
+          "type": "rand_int",
+          "version": "1.0.1",
+          "label": "",
+          "notes": "",
+          "isOpen": true,
+          "isIntermediate": true,
+          "useCache": false,
+          "nodePack": "invokeai",
+          "inputs": {
+            "low": {
+              "name": "low",
+              "label": "",
+              "value": 0
+            },
+            "high": {
+              "name": "high",
+              "label": "",
+              "value": 2147483647
+            }
+          }
+        },
+        "position": {
+          "x": 470.45870147220353,
+          "y": 350.3141781644303
+        }
+      },
+      {
+        "id": "9eb72af0-dd9e-4ec5-ad87-d65e3c01f48b",
+        "type": "invocation",
+        "data": {
+          "id": "9eb72af0-dd9e-4ec5-ad87-d65e3c01f48b",
+          "type": "sd3_l2i",
+          "version": "1.3.0",
+          "label": "",
+          "notes": "",
+          "isOpen": true,
+          "isIntermediate": false,
+          "useCache": true,
+          "nodePack": "invokeai",
+          "inputs": {
+            "board": {
+              "name": "board",
+              "label": ""
+            },
+            "metadata": {
+              "name": "metadata",
+              "label": ""
+            },
+            "latents": {
+              "name": "latents",
+              "label": ""
+            },
+            "vae": {
+              "name": "vae",
+              "label": ""
+            }
+          }
+        },
+        "position": {
+          "x": 1192.3097009334897,
+          "y": -366.0994675072209
+        }
+      },
+      {
+        "id": "3b4f7f27-cfc0-4373-a009-99c5290d0cd6",
+        "type": "invocation",
+        "data": {
+          "id": "3b4f7f27-cfc0-4373-a009-99c5290d0cd6",
+          "type": "sd3_text_encoder",
+          "version": "1.0.0",
+          "label": "",
+          "notes": "",
+          "isOpen": true,
+          "isIntermediate": true,
+          "useCache": true,
+          "nodePack": "invokeai",
+          "inputs": {
+            "clip_l": {
+              "name": "clip_l",
+              "label": ""
+            },
+            "clip_g": {
+              "name": "clip_g",
+              "label": ""
+            },
+            "t5_encoder": {
+              "name": "t5_encoder",
+              "label": ""
+            },
+            "prompt": {
+              "name": "prompt",
+              "label": "",
+              "value": ""
+            }
+          }
+        },
+        "position": {
+          "x": 408.16054647924784,
+          "y": 65.06415352118786
+        }
+      },
+      {
+        "id": "e17d34e7-6ed1-493c-9a85-4fcd291cb084",
+        "type": "invocation",
+        "data": {
+          "id": "e17d34e7-6ed1-493c-9a85-4fcd291cb084",
+          "type": "sd3_text_encoder",
+          "version": "1.0.0",
+          "label": "",
+          "notes": "",
+          "isOpen": true,
+          "isIntermediate": true,
+          "useCache": true,
+          "nodePack": "invokeai",
+          "inputs": {
+            "clip_l": {
+              "name": "clip_l",
+              "label": ""
+            },
+            "clip_g": {
+              "name": "clip_g",
+              "label": ""
+            },
+            "t5_encoder": {
+              "name": "t5_encoder",
+              "label": ""
+            },
+            "prompt": {
+              "name": "prompt",
+              "label": "",
+              "value": ""
+            }
+          }
+        },
+        "position": {
+          "x": 378.9283412440941,
+          "y": -302.65777497352553
+        }
+      },
+      {
+        "id": "c7539f7b-7ac5-49b9-93eb-87ede611409f",
+        "type": "invocation",
+        "data": {
+          "id": "c7539f7b-7ac5-49b9-93eb-87ede611409f",
+          "type": "sd3_denoise",
+          "version": "1.0.0",
+          "label": "",
+          "notes": "",
+          "isOpen": true,
+          "isIntermediate": true,
+          "useCache": true,
+          "nodePack": "invokeai",
+          "inputs": {
+            "board": {
+              "name": "board",
+              "label": ""
+            },
+            "metadata": {
+              "name": "metadata",
+              "label": ""
+            },
+            "transformer": {
+              "name": "transformer",
+              "label": ""
+            },
+            "positive_conditioning": {
+              "name": "positive_conditioning",
+              "label": ""
+            },
+            "negative_conditioning": {
+              "name": "negative_conditioning",
+              "label": ""
+            },
+            "cfg_scale": {
+              "name": "cfg_scale",
+              "label": "",
+              "value": 3.5
+            },
+            "width": {
+              "name": "width",
+              "label": "",
+              "value": 1024
+            },
+            "height": {
+              "name": "height",
+              "label": "",
+              "value": 1024
+            },
+            "steps": {
+              "name": "steps",
+              "label": "",
+              "value": 30
+            },
+            "seed": {
+              "name": "seed",
+              "label": "",
+              "value": 0
+            }
+          }
+        },
+        "position": {
+          "x": 813.7814762740603,
+          "y": -142.20529727605867
+        }
+      }
+    ],
+    "edges": [
+      {
+        "id": "reactflow__edge-3f22f668-0e02-4fde-a2bb-c339586ceb4cvae-9eb72af0-dd9e-4ec5-ad87-d65e3c01f48bvae",
+        "type": "default",
+        "source": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "target": "9eb72af0-dd9e-4ec5-ad87-d65e3c01f48b",
+        "sourceHandle": "vae",
+        "targetHandle": "vae"
+      },
+      {
+        "id": "reactflow__edge-3f22f668-0e02-4fde-a2bb-c339586ceb4ct5_encoder-3b4f7f27-cfc0-4373-a009-99c5290d0cd6t5_encoder",
+        "type": "default",
+        "source": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "target": "3b4f7f27-cfc0-4373-a009-99c5290d0cd6",
+        "sourceHandle": "t5_encoder",
+        "targetHandle": "t5_encoder"
+      },
+      {
+        "id": "reactflow__edge-3f22f668-0e02-4fde-a2bb-c339586ceb4ct5_encoder-e17d34e7-6ed1-493c-9a85-4fcd291cb084t5_encoder",
+        "type": "default",
+        "source": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "target": "e17d34e7-6ed1-493c-9a85-4fcd291cb084",
+        "sourceHandle": "t5_encoder",
+        "targetHandle": "t5_encoder"
+      },
+      {
+        "id": "reactflow__edge-3f22f668-0e02-4fde-a2bb-c339586ceb4cclip_g-3b4f7f27-cfc0-4373-a009-99c5290d0cd6clip_g",
+        "type": "default",
+        "source": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "target": "3b4f7f27-cfc0-4373-a009-99c5290d0cd6",
+        "sourceHandle": "clip_g",
+        "targetHandle": "clip_g"
+      },
+      {
+        "id": "reactflow__edge-3f22f668-0e02-4fde-a2bb-c339586ceb4cclip_g-e17d34e7-6ed1-493c-9a85-4fcd291cb084clip_g",
+        "type": "default",
+        "source": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "target": "e17d34e7-6ed1-493c-9a85-4fcd291cb084",
+        "sourceHandle": "clip_g",
+        "targetHandle": "clip_g"
+      },
+      {
+        "id": "reactflow__edge-3f22f668-0e02-4fde-a2bb-c339586ceb4cclip_l-3b4f7f27-cfc0-4373-a009-99c5290d0cd6clip_l",
+        "type": "default",
+        "source": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "target": "3b4f7f27-cfc0-4373-a009-99c5290d0cd6",
+        "sourceHandle": "clip_l",
+        "targetHandle": "clip_l"
+      },
+      {
+        "id": "reactflow__edge-3f22f668-0e02-4fde-a2bb-c339586ceb4cclip_l-e17d34e7-6ed1-493c-9a85-4fcd291cb084clip_l",
+        "type": "default",
+        "source": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "target": "e17d34e7-6ed1-493c-9a85-4fcd291cb084",
+        "sourceHandle": "clip_l",
+        "targetHandle": "clip_l"
+      },
+      {
+        "id": "reactflow__edge-3f22f668-0e02-4fde-a2bb-c339586ceb4ctransformer-c7539f7b-7ac5-49b9-93eb-87ede611409ftransformer",
+        "type": "default",
+        "source": "3f22f668-0e02-4fde-a2bb-c339586ceb4c",
+        "target": "c7539f7b-7ac5-49b9-93eb-87ede611409f",
+        "sourceHandle": "transformer",
+        "targetHandle": "transformer"
+      },
+      {
+        "id": "reactflow__edge-f7e394ac-6394-4096-abcb-de0d346506b3value-c7539f7b-7ac5-49b9-93eb-87ede611409fseed",
+        "type": "default",
+        "source": "f7e394ac-6394-4096-abcb-de0d346506b3",
+        "target": "c7539f7b-7ac5-49b9-93eb-87ede611409f",
+        "sourceHandle": "value",
+        "targetHandle": "seed"
+      },
+      {
+        "id": "reactflow__edge-c7539f7b-7ac5-49b9-93eb-87ede611409flatents-9eb72af0-dd9e-4ec5-ad87-d65e3c01f48blatents",
+        "type": "default",
+        "source": "c7539f7b-7ac5-49b9-93eb-87ede611409f",
+        "target": "9eb72af0-dd9e-4ec5-ad87-d65e3c01f48b",
+        "sourceHandle": "latents",
+        "targetHandle": "latents"
+      },
+      {
+        "id": "reactflow__edge-e17d34e7-6ed1-493c-9a85-4fcd291cb084conditioning-c7539f7b-7ac5-49b9-93eb-87ede611409fpositive_conditioning",
+        "type": "default",
+        "source": "e17d34e7-6ed1-493c-9a85-4fcd291cb084",
+        "target": "c7539f7b-7ac5-49b9-93eb-87ede611409f",
+        "sourceHandle": "conditioning",
+        "targetHandle": "positive_conditioning"
+      },
+      {
+        "id": "reactflow__edge-3b4f7f27-cfc0-4373-a009-99c5290d0cd6conditioning-c7539f7b-7ac5-49b9-93eb-87ede611409fnegative_conditioning",
+        "type": "default",
+        "source": "3b4f7f27-cfc0-4373-a009-99c5290d0cd6",
+        "target": "c7539f7b-7ac5-49b9-93eb-87ede611409f",
+        "sourceHandle": "conditioning",
+        "targetHandle": "negative_conditioning"
+      }
+    ]
+  }

invokeai/app/services/workflow_records/workflow_records_base.py

@@ -39,11 +39,11 @@ class WorkflowRecordsStorageBase(ABC):
     @abstractmethod
     def get_many(
         self,
-        page: int,
-        per_page: int,
         order_by: WorkflowRecordOrderBy,
         direction: SQLiteDirection,
         category: WorkflowCategory,
+        page: int,
+        per_page: Optional[int],
         query: Optional[str],
     ) -> PaginatedResults[WorkflowRecordListItemDTO]:
         """Gets many workflows."""

invokeai/app/services/workflow_records/workflow_records_sqlite.py

@@ -125,11 +125,11 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
 
     def get_many(
         self,
-        page: int,
-        per_page: int,
         order_by: WorkflowRecordOrderBy,
         direction: SQLiteDirection,
         category: WorkflowCategory,
+        page: int = 0,
+        per_page: Optional[int] = None,
         query: Optional[str] = None,
     ) -> PaginatedResults[WorkflowRecordListItemDTO]:
         try:
@@ -153,6 +153,7 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
                 """
             main_params: list[int | str] = [category.value]
             count_params: list[int | str] = [category.value]
+
             stripped_query = query.strip() if query else None
             if stripped_query:
                 wildcard_query = "%" + stripped_query + "%"
@@ -161,20 +162,28 @@ class SqliteWorkflowRecordsStorage(WorkflowRecordsStorageBase):
                 main_params.extend([wildcard_query, wildcard_query])
                 count_params.extend([wildcard_query, wildcard_query])
 
-            main_query += f" ORDER BY {order_by.value} {direction.value} LIMIT ? OFFSET ?;"
-            main_params.extend([per_page, page * per_page])
+            main_query += f" ORDER BY {order_by.value} {direction.value}"
+
+            if per_page:
+                main_query += " LIMIT ? OFFSET ?"
+                main_params.extend([per_page, page * per_page])
+
             self._cursor.execute(main_query, main_params)
             rows = self._cursor.fetchall()
             workflows = [WorkflowRecordListItemDTOValidator.validate_python(dict(row)) for row in rows]
 
             self._cursor.execute(count_query, count_params)
             total = self._cursor.fetchone()[0]
-            pages = total // per_page + (total % per_page > 0)
+
+            if per_page:
+                pages = total // per_page + (total % per_page > 0)
+            else:
+                pages = 1  # If no pagination, there is only one page
 
             return PaginatedResults(
                 items=workflows,
                 page=page,
-                per_page=per_page,
+                per_page=per_page if per_page else total,
                 pages=pages,
                 total=total,
             )

invokeai/app/util/step_callback.py

@@ -34,6 +34,25 @@ SD1_5_LATENT_RGB_FACTORS = [
     [-0.1307, -0.1874, -0.7445],  # L4
 ]
 
+SD3_5_LATENT_RGB_FACTORS = [
+    [-0.05240681, 0.03251581, 0.0749016],
+    [-0.0580572, 0.00759826, 0.05729818],
+    [0.16144888, 0.01270368, -0.03768577],
+    [0.14418615, 0.08460266, 0.15941818],
+    [0.04894035, 0.0056485, -0.06686988],
+    [0.05187166, 0.19222395, 0.06261094],
+    [0.1539433, 0.04818359, 0.07103094],
+    [-0.08601796, 0.09013458, 0.10893912],
+    [-0.12398469, -0.06766567, 0.0033688],
+    [-0.0439737, 0.07825329, 0.02258823],
+    [0.03101129, 0.06382551, 0.07753657],
+    [-0.01315361, 0.08554491, -0.08772475],
+    [0.06464487, 0.05914605, 0.13262741],
+    [-0.07863674, -0.02261737, -0.12761454],
+    [-0.09923835, -0.08010759, -0.06264447],
+    [-0.03392309, -0.0804029, -0.06078822],
+]
+
 FLUX_LATENT_RGB_FACTORS = [
     [-0.0412, 0.0149, 0.0521],
     [0.0056, 0.0291, 0.0768],
@@ -110,6 +129,9 @@ def stable_diffusion_step_callback(
         sdxl_latent_rgb_factors = torch.tensor(SDXL_LATENT_RGB_FACTORS, dtype=sample.dtype, device=sample.device)
         sdxl_smooth_matrix = torch.tensor(SDXL_SMOOTH_MATRIX, dtype=sample.dtype, device=sample.device)
         image = sample_to_lowres_estimated_image(sample, sdxl_latent_rgb_factors, sdxl_smooth_matrix)
+    elif base_model == BaseModelType.StableDiffusion3:
+        sd3_latent_rgb_factors = torch.tensor(SD3_5_LATENT_RGB_FACTORS, dtype=sample.dtype, device=sample.device)
+        image = sample_to_lowres_estimated_image(sample, sd3_latent_rgb_factors)
     else:
         v1_5_latent_rgb_factors = torch.tensor(SD1_5_LATENT_RGB_FACTORS, dtype=sample.dtype, device=sample.device)
         image = sample_to_lowres_estimated_image(sample, v1_5_latent_rgb_factors)

invokeai/backend/flux/controlnet/controlnet_flux_output.py

@@ -0,0 +1,58 @@
+from dataclasses import dataclass
+
+import torch
+
+
+@dataclass
+class ControlNetFluxOutput:
+    single_block_residuals: list[torch.Tensor] | None
+    double_block_residuals: list[torch.Tensor] | None
+
+    def apply_weight(self, weight: float):
+        if self.single_block_residuals is not None:
+            for i in range(len(self.single_block_residuals)):
+                self.single_block_residuals[i] = self.single_block_residuals[i] * weight
+        if self.double_block_residuals is not None:
+            for i in range(len(self.double_block_residuals)):
+                self.double_block_residuals[i] = self.double_block_residuals[i] * weight
+
+
+def add_tensor_lists_elementwise(
+    list1: list[torch.Tensor] | None, list2: list[torch.Tensor] | None
+) -> list[torch.Tensor] | None:
+    """Add two tensor lists elementwise that could be None."""
+    if list1 is None and list2 is None:
+        return None
+    if list1 is None:
+        return list2
+    if list2 is None:
+        return list1
+
+    new_list: list[torch.Tensor] = []
+    for list1_tensor, list2_tensor in zip(list1, list2, strict=True):
+        new_list.append(list1_tensor + list2_tensor)
+    return new_list
+
+
+def add_controlnet_flux_outputs(
+    controlnet_output_1: ControlNetFluxOutput, controlnet_output_2: ControlNetFluxOutput
+) -> ControlNetFluxOutput:
+    return ControlNetFluxOutput(
+        single_block_residuals=add_tensor_lists_elementwise(
+            controlnet_output_1.single_block_residuals, controlnet_output_2.single_block_residuals
+        ),
+        double_block_residuals=add_tensor_lists_elementwise(
+            controlnet_output_1.double_block_residuals, controlnet_output_2.double_block_residuals
+        ),
+    )
+
+
+def sum_controlnet_flux_outputs(
+    controlnet_outputs: list[ControlNetFluxOutput],
+) -> ControlNetFluxOutput:
+    controlnet_output_sum = ControlNetFluxOutput(single_block_residuals=None, double_block_residuals=None)
+
+    for controlnet_output in controlnet_outputs:
+        controlnet_output_sum = add_controlnet_flux_outputs(controlnet_output_sum, controlnet_output)
+
+    return controlnet_output_sum

invokeai/backend/flux/controlnet/instantx_controlnet_flux.py

@@ -0,0 +1,180 @@
+# This file was initially copied from:
+# https://github.com/huggingface/diffusers/blob/99f608218caa069a2f16dcf9efab46959b15aec0/src/diffusers/models/controlnet_flux.py
+
+
+from dataclasses import dataclass
+
+import torch
+import torch.nn as nn
+
+from invokeai.backend.flux.controlnet.zero_module import zero_module
+from invokeai.backend.flux.model import FluxParams
+from invokeai.backend.flux.modules.layers import (
+    DoubleStreamBlock,
+    EmbedND,
+    MLPEmbedder,
+    SingleStreamBlock,
+    timestep_embedding,
+)
+
+
+@dataclass
+class InstantXControlNetFluxOutput:
+    controlnet_block_samples: list[torch.Tensor] | None
+    controlnet_single_block_samples: list[torch.Tensor] | None
+
+
+# NOTE(ryand): Mapping between diffusers FLUX transformer params and BFL FLUX transformer params:
+# - Diffusers: BFL
+# - in_channels: in_channels
+# - num_layers: depth
+# - num_single_layers: depth_single_blocks
+# - attention_head_dim: hidden_size // num_heads
+# - num_attention_heads: num_heads
+# - joint_attention_dim: context_in_dim
+# - pooled_projection_dim: vec_in_dim
+# - guidance_embeds: guidance_embed
+# - axes_dims_rope: axes_dim
+
+
+class InstantXControlNetFlux(torch.nn.Module):
+    def __init__(self, params: FluxParams, num_control_modes: int | None = None):
+        """
+        Args:
+            params (FluxParams): The parameters for the FLUX model.
+            num_control_modes (int | None, optional): The number of controlnet modes. If non-None, then the model is a
+                'union controlnet' model and expects a mode conditioning input at runtime.
+        """
+        super().__init__()
+
+        # The following modules mirror the base FLUX transformer model.
+        # -------------------------------------------------------------
+        self.params = params
+        self.in_channels = params.in_channels
+        self.out_channels = self.in_channels
+        if params.hidden_size % params.num_heads != 0:
+            raise ValueError(f"Hidden size {params.hidden_size} must be divisible by num_heads {params.num_heads}")
+        pe_dim = params.hidden_size // params.num_heads
+        if sum(params.axes_dim) != pe_dim:
+            raise ValueError(f"Got {params.axes_dim} but expected positional dim {pe_dim}")
+        self.hidden_size = params.hidden_size
+        self.num_heads = params.num_heads
+        self.pe_embedder = EmbedND(dim=pe_dim, theta=params.theta, axes_dim=params.axes_dim)
+        self.img_in = nn.Linear(self.in_channels, self.hidden_size, bias=True)
+        self.time_in = MLPEmbedder(in_dim=256, hidden_dim=self.hidden_size)
+        self.vector_in = MLPEmbedder(params.vec_in_dim, self.hidden_size)
+        self.guidance_in = (
+            MLPEmbedder(in_dim=256, hidden_dim=self.hidden_size) if params.guidance_embed else nn.Identity()
+        )
+        self.txt_in = nn.Linear(params.context_in_dim, self.hidden_size)
+
+        self.double_blocks = nn.ModuleList(
+            [
+                DoubleStreamBlock(
+                    self.hidden_size,
+                    self.num_heads,
+                    mlp_ratio=params.mlp_ratio,
+                    qkv_bias=params.qkv_bias,
+                )
+                for _ in range(params.depth)
+            ]
+        )
+
+        self.single_blocks = nn.ModuleList(
+            [
+                SingleStreamBlock(self.hidden_size, self.num_heads, mlp_ratio=params.mlp_ratio)
+                for _ in range(params.depth_single_blocks)
+            ]
+        )
+
+        # The following modules are specific to the ControlNet model.
+        # -----------------------------------------------------------
+        self.controlnet_blocks = nn.ModuleList([])
+        for _ in range(len(self.double_blocks)):
+            self.controlnet_blocks.append(zero_module(nn.Linear(self.hidden_size, self.hidden_size)))
+
+        self.controlnet_single_blocks = nn.ModuleList([])
+        for _ in range(len(self.single_blocks)):
+            self.controlnet_single_blocks.append(zero_module(nn.Linear(self.hidden_size, self.hidden_size)))
+
+        self.is_union = False
+        if num_control_modes is not None:
+            self.is_union = True
+            self.controlnet_mode_embedder = nn.Embedding(num_control_modes, self.hidden_size)
+
+        self.controlnet_x_embedder = zero_module(torch.nn.Linear(self.in_channels, self.hidden_size))
+
+    def forward(
+        self,
+        controlnet_cond: torch.Tensor,
+        controlnet_mode: torch.Tensor | None,
+        img: torch.Tensor,
+        img_ids: torch.Tensor,
+        txt: torch.Tensor,
+        txt_ids: torch.Tensor,
+        timesteps: torch.Tensor,
+        y: torch.Tensor,
+        guidance: torch.Tensor | None = None,
+    ) -> InstantXControlNetFluxOutput:
+        if img.ndim != 3 or txt.ndim != 3:
+            raise ValueError("Input img and txt tensors must have 3 dimensions.")
+
+        img = self.img_in(img)
+
+        # Add controlnet_cond embedding.
+        img = img + self.controlnet_x_embedder(controlnet_cond)
+
+        vec = self.time_in(timestep_embedding(timesteps, 256))
+        if self.params.guidance_embed:
+            if guidance is None:
+                raise ValueError("Didn't get guidance strength for guidance distilled model.")
+            vec = vec + self.guidance_in(timestep_embedding(guidance, 256))
+        vec = vec + self.vector_in(y)
+        txt = self.txt_in(txt)
+
+        # If this is a union ControlNet, then concat the control mode embedding to the T5 text embedding.
+        if self.is_union:
+            if controlnet_mode is None:
+                # We allow users to enter 'None' as the controlnet_mode if they don't want to worry about this input.
+                # We've chosen to use a zero-embedding in this case.
+                zero_index = torch.zeros([1, 1], dtype=torch.long, device=txt.device)
+                controlnet_mode_emb = torch.zeros_like(self.controlnet_mode_embedder(zero_index))
+            else:
+                controlnet_mode_emb = self.controlnet_mode_embedder(controlnet_mode)
+            txt = torch.cat([controlnet_mode_emb, txt], dim=1)
+            txt_ids = torch.cat([txt_ids[:, :1, :], txt_ids], dim=1)
+        else:
+            assert controlnet_mode is None
+
+        ids = torch.cat((txt_ids, img_ids), dim=1)
+        pe = self.pe_embedder(ids)
+
+        double_block_samples: list[torch.Tensor] = []
+        for block in self.double_blocks:
+            img, txt = block(img=img, txt=txt, vec=vec, pe=pe)
+            double_block_samples.append(img)
+
+        img = torch.cat((txt, img), 1)
+
+        single_block_samples: list[torch.Tensor] = []
+        for block in self.single_blocks:
+            img = block(img, vec=vec, pe=pe)
+            single_block_samples.append(img[:, txt.shape[1] :])
+
+        # ControlNet Block
+        controlnet_double_block_samples: list[torch.Tensor] = []
+        for double_block_sample, controlnet_block in zip(double_block_samples, self.controlnet_blocks, strict=True):
+            double_block_sample = controlnet_block(double_block_sample)
+            controlnet_double_block_samples.append(double_block_sample)
+
+        controlnet_single_block_samples: list[torch.Tensor] = []
+        for single_block_sample, controlnet_block in zip(
+            single_block_samples, self.controlnet_single_blocks, strict=True
+        ):
+            single_block_sample = controlnet_block(single_block_sample)
+            controlnet_single_block_samples.append(single_block_sample)
+
+        return InstantXControlNetFluxOutput(
+            controlnet_block_samples=controlnet_double_block_samples or None,
+            controlnet_single_block_samples=controlnet_single_block_samples or None,
+        )

invokeai/backend/flux/controlnet/state_dict_utils.py

@@ -0,0 +1,295 @@
+from typing import Any, Dict
+
+import torch
+
+from invokeai.backend.flux.model import FluxParams
+
+
+def is_state_dict_xlabs_controlnet(sd: Dict[str, Any]) -> bool:
+    """Is the state dict for an XLabs ControlNet model?
+
+    This is intended to be a reasonably high-precision detector, but it is not guaranteed to have perfect precision.
+    """
+    # If all of the expected keys are present, then this is very likely an XLabs ControlNet model.
+    expected_keys = {
+        "controlnet_blocks.0.bias",
+        "controlnet_blocks.0.weight",
+        "input_hint_block.0.bias",
+        "input_hint_block.0.weight",
+        "pos_embed_input.bias",
+        "pos_embed_input.weight",
+    }
+
+    if expected_keys.issubset(sd.keys()):
+        return True
+    return False
+
+
+def is_state_dict_instantx_controlnet(sd: Dict[str, Any]) -> bool:
+    """Is the state dict for an InstantX ControlNet model?
+
+    This is intended to be a reasonably high-precision detector, but it is not guaranteed to have perfect precision.
+    """
+    # If all of the expected keys are present, then this is very likely an InstantX ControlNet model.
+    expected_keys = {
+        "controlnet_blocks.0.bias",
+        "controlnet_blocks.0.weight",
+        "controlnet_x_embedder.bias",
+        "controlnet_x_embedder.weight",
+    }
+
+    if expected_keys.issubset(sd.keys()):
+        return True
+    return False
+
+
+def _fuse_weights(*t: torch.Tensor) -> torch.Tensor:
+    """Fuse weights along dimension 0.
+
+    Used to fuse q, k, v attention weights into a single qkv tensor when converting from diffusers to BFL format.
+    """
+    # TODO(ryand): Double check dim=0 is correct.
+    return torch.cat(t, dim=0)
+
+
+def _convert_flux_double_block_sd_from_diffusers_to_bfl_format(
+    sd: Dict[str, torch.Tensor], double_block_index: int
+) -> Dict[str, torch.Tensor]:
+    """Convert the state dict for a double block from diffusers format to BFL format."""
+    to_prefix = f"double_blocks.{double_block_index}"
+    from_prefix = f"transformer_blocks.{double_block_index}"
+
+    new_sd: dict[str, torch.Tensor] = {}
+
+    # Check one key to determine if this block exists.
+    if f"{from_prefix}.attn.add_q_proj.bias" not in sd:
+        return new_sd
+
+    # txt_attn.qkv
+    new_sd[f"{to_prefix}.txt_attn.qkv.bias"] = _fuse_weights(
+        sd.pop(f"{from_prefix}.attn.add_q_proj.bias"),
+        sd.pop(f"{from_prefix}.attn.add_k_proj.bias"),
+        sd.pop(f"{from_prefix}.attn.add_v_proj.bias"),
+    )
+    new_sd[f"{to_prefix}.txt_attn.qkv.weight"] = _fuse_weights(
+        sd.pop(f"{from_prefix}.attn.add_q_proj.weight"),
+        sd.pop(f"{from_prefix}.attn.add_k_proj.weight"),
+        sd.pop(f"{from_prefix}.attn.add_v_proj.weight"),
+    )
+
+    # img_attn.qkv
+    new_sd[f"{to_prefix}.img_attn.qkv.bias"] = _fuse_weights(
+        sd.pop(f"{from_prefix}.attn.to_q.bias"),
+        sd.pop(f"{from_prefix}.attn.to_k.bias"),
+        sd.pop(f"{from_prefix}.attn.to_v.bias"),
+    )
+    new_sd[f"{to_prefix}.img_attn.qkv.weight"] = _fuse_weights(
+        sd.pop(f"{from_prefix}.attn.to_q.weight"),
+        sd.pop(f"{from_prefix}.attn.to_k.weight"),
+        sd.pop(f"{from_prefix}.attn.to_v.weight"),
+    )
+
+    # Handle basic 1-to-1 key conversions.
+    key_map = {
+        # img_attn
+        "attn.norm_k.weight": "img_attn.norm.key_norm.scale",
+        "attn.norm_q.weight": "img_attn.norm.query_norm.scale",
+        "attn.to_out.0.weight": "img_attn.proj.weight",
+        "attn.to_out.0.bias": "img_attn.proj.bias",
+        # img_mlp
+        "ff.net.0.proj.weight": "img_mlp.0.weight",
+        "ff.net.0.proj.bias": "img_mlp.0.bias",
+        "ff.net.2.weight": "img_mlp.2.weight",
+        "ff.net.2.bias": "img_mlp.2.bias",
+        # img_mod
+        "norm1.linear.weight": "img_mod.lin.weight",
+        "norm1.linear.bias": "img_mod.lin.bias",
+        # txt_attn
+        "attn.norm_added_q.weight": "txt_attn.norm.query_norm.scale",
+        "attn.norm_added_k.weight": "txt_attn.norm.key_norm.scale",
+        "attn.to_add_out.weight": "txt_attn.proj.weight",
+        "attn.to_add_out.bias": "txt_attn.proj.bias",
+        # txt_mlp
+        "ff_context.net.0.proj.weight": "txt_mlp.0.weight",
+        "ff_context.net.0.proj.bias": "txt_mlp.0.bias",
+        "ff_context.net.2.weight": "txt_mlp.2.weight",
+        "ff_context.net.2.bias": "txt_mlp.2.bias",
+        # txt_mod
+        "norm1_context.linear.weight": "txt_mod.lin.weight",
+        "norm1_context.linear.bias": "txt_mod.lin.bias",
+    }
+    for from_key, to_key in key_map.items():
+        new_sd[f"{to_prefix}.{to_key}"] = sd.pop(f"{from_prefix}.{from_key}")
+
+    return new_sd
+
+
+def _convert_flux_single_block_sd_from_diffusers_to_bfl_format(
+    sd: Dict[str, torch.Tensor], single_block_index: int
+) -> Dict[str, torch.Tensor]:
+    """Convert the state dict for a single block from diffusers format to BFL format."""
+    to_prefix = f"single_blocks.{single_block_index}"
+    from_prefix = f"single_transformer_blocks.{single_block_index}"
+
+    new_sd: dict[str, torch.Tensor] = {}
+
+    # Check one key to determine if this block exists.
+    if f"{from_prefix}.attn.to_q.bias" not in sd:
+        return new_sd
+
+    # linear1 (qkv)
+    new_sd[f"{to_prefix}.linear1.bias"] = _fuse_weights(
+        sd.pop(f"{from_prefix}.attn.to_q.bias"),
+        sd.pop(f"{from_prefix}.attn.to_k.bias"),
+        sd.pop(f"{from_prefix}.attn.to_v.bias"),
+        sd.pop(f"{from_prefix}.proj_mlp.bias"),
+    )
+    new_sd[f"{to_prefix}.linear1.weight"] = _fuse_weights(
+        sd.pop(f"{from_prefix}.attn.to_q.weight"),
+        sd.pop(f"{from_prefix}.attn.to_k.weight"),
+        sd.pop(f"{from_prefix}.attn.to_v.weight"),
+        sd.pop(f"{from_prefix}.proj_mlp.weight"),
+    )
+
+    # Handle basic 1-to-1 key conversions.
+    key_map = {
+        # linear2
+        "proj_out.weight": "linear2.weight",
+        "proj_out.bias": "linear2.bias",
+        # modulation
+        "norm.linear.weight": "modulation.lin.weight",
+        "norm.linear.bias": "modulation.lin.bias",
+        # norm
+        "attn.norm_k.weight": "norm.key_norm.scale",
+        "attn.norm_q.weight": "norm.query_norm.scale",
+    }
+    for from_key, to_key in key_map.items():
+        new_sd[f"{to_prefix}.{to_key}"] = sd.pop(f"{from_prefix}.{from_key}")
+
+    return new_sd
+
+
+def convert_diffusers_instantx_state_dict_to_bfl_format(sd: Dict[str, torch.Tensor]) -> Dict[str, torch.Tensor]:
+    """Convert an InstantX ControlNet state dict to the format that can be loaded by our internal
+    InstantXControlNetFlux model.
+
+    The original InstantX ControlNet model was developed to be used in diffusers. We have ported the original
+    implementation to InstantXControlNetFlux to make it compatible with BFL-style models. This function converts the
+    original state dict to the format expected by InstantXControlNetFlux.
+    """
+    # Shallow copy sd so that we can pop keys from it without modifying the original.
+    sd = sd.copy()
+
+    new_sd: dict[str, torch.Tensor] = {}
+
+    # Handle basic 1-to-1 key conversions.
+    basic_key_map = {
+        # Base model keys.
+        # ----------------
+        # txt_in keys.
+        "context_embedder.bias": "txt_in.bias",
+        "context_embedder.weight": "txt_in.weight",
+        # guidance_in MLPEmbedder keys.
+        "time_text_embed.guidance_embedder.linear_1.bias": "guidance_in.in_layer.bias",
+        "time_text_embed.guidance_embedder.linear_1.weight": "guidance_in.in_layer.weight",
+        "time_text_embed.guidance_embedder.linear_2.bias": "guidance_in.out_layer.bias",
+        "time_text_embed.guidance_embedder.linear_2.weight": "guidance_in.out_layer.weight",
+        # vector_in MLPEmbedder keys.
+        "time_text_embed.text_embedder.linear_1.bias": "vector_in.in_layer.bias",
+        "time_text_embed.text_embedder.linear_1.weight": "vector_in.in_layer.weight",
+        "time_text_embed.text_embedder.linear_2.bias": "vector_in.out_layer.bias",
+        "time_text_embed.text_embedder.linear_2.weight": "vector_in.out_layer.weight",
+        # time_in MLPEmbedder keys.
+        "time_text_embed.timestep_embedder.linear_1.bias": "time_in.in_layer.bias",
+        "time_text_embed.timestep_embedder.linear_1.weight": "time_in.in_layer.weight",
+        "time_text_embed.timestep_embedder.linear_2.bias": "time_in.out_layer.bias",
+        "time_text_embed.timestep_embedder.linear_2.weight": "time_in.out_layer.weight",
+        # img_in keys.
+        "x_embedder.bias": "img_in.bias",
+        "x_embedder.weight": "img_in.weight",
+    }
+    for old_key, new_key in basic_key_map.items():
+        v = sd.pop(old_key, None)
+        if v is not None:
+            new_sd[new_key] = v
+
+    # Handle the double_blocks.
+    block_index = 0
+    while True:
+        converted_double_block_sd = _convert_flux_double_block_sd_from_diffusers_to_bfl_format(sd, block_index)
+        if len(converted_double_block_sd) == 0:
+            break
+        new_sd.update(converted_double_block_sd)
+        block_index += 1
+
+    # Handle the single_blocks.
+    block_index = 0
+    while True:
+        converted_singe_block_sd = _convert_flux_single_block_sd_from_diffusers_to_bfl_format(sd, block_index)
+        if len(converted_singe_block_sd) == 0:
+            break
+        new_sd.update(converted_singe_block_sd)
+        block_index += 1
+
+    # Transfer controlnet keys as-is.
+    for k in list(sd.keys()):
+        if k.startswith("controlnet_"):
+            new_sd[k] = sd.pop(k)
+
+    # Assert that all keys have been handled.
+    assert len(sd) == 0
+    return new_sd
+
+
+def infer_flux_params_from_state_dict(sd: Dict[str, torch.Tensor]) -> FluxParams:
+    """Infer the FluxParams from the shape of a FLUX state dict. When a model is distributed in diffusers format, this
+    information is all contained in the config.json file that accompanies the model. However, being apple to infer the
+    params from the state dict enables us to load models (e.g. an InstantX ControlNet) from a single weight file.
+    """
+    hidden_size = sd["img_in.weight"].shape[0]
+    mlp_hidden_dim = sd["double_blocks.0.img_mlp.0.weight"].shape[0]
+    # mlp_ratio is a float, but we treat it as an int here to avoid having to think about possible float precision
+    # issues. In practice, mlp_ratio is usually 4.
+    mlp_ratio = mlp_hidden_dim // hidden_size
+
+    head_dim = sd["double_blocks.0.img_attn.norm.query_norm.scale"].shape[0]
+    num_heads = hidden_size // head_dim
+
+    # Count the number of double blocks.
+    double_block_index = 0
+    while f"double_blocks.{double_block_index}.img_attn.qkv.weight" in sd:
+        double_block_index += 1
+
+    # Count the number of single blocks.
+    single_block_index = 0
+    while f"single_blocks.{single_block_index}.linear1.weight" in sd:
+        single_block_index += 1
+
+    return FluxParams(
+        in_channels=sd["img_in.weight"].shape[1],
+        vec_in_dim=sd["vector_in.in_layer.weight"].shape[1],
+        context_in_dim=sd["txt_in.weight"].shape[1],
+        hidden_size=hidden_size,
+        mlp_ratio=mlp_ratio,
+        num_heads=num_heads,
+        depth=double_block_index,
+        depth_single_blocks=single_block_index,
+        # axes_dim cannot be inferred from the state dict. The hard-coded value is correct for dev/schnell models.
+        axes_dim=[16, 56, 56],
+        # theta cannot be inferred from the state dict. The hard-coded value is correct for dev/schnell models.
+        theta=10_000,
+        qkv_bias="double_blocks.0.img_attn.qkv.bias" in sd,
+        guidance_embed="guidance_in.in_layer.weight" in sd,
+    )
+
+
+def infer_instantx_num_control_modes_from_state_dict(sd: Dict[str, torch.Tensor]) -> int | None:
+    """Infer the number of ControlNet Union modes from the shape of a InstantX ControlNet state dict.
+
+    Returns None if the model is not a ControlNet Union model. Otherwise returns the number of modes.
+    """
+    mode_embedder_key = "controlnet_mode_embedder.weight"
+    if mode_embedder_key not in sd:
+        return None
+
+    return sd[mode_embedder_key].shape[0]

invokeai/backend/flux/controlnet/xlabs_controlnet_flux.py

@@ -0,0 +1,130 @@
+# This file was initially based on:
+# https://github.com/XLabs-AI/x-flux/blob/47495425dbed499be1e8e5a6e52628b07349cba2/src/flux/controlnet.py
+
+
+from dataclasses import dataclass
+
+import torch
+from einops import rearrange
+
+from invokeai.backend.flux.controlnet.zero_module import zero_module
+from invokeai.backend.flux.model import FluxParams
+from invokeai.backend.flux.modules.layers import DoubleStreamBlock, EmbedND, MLPEmbedder, timestep_embedding
+
+
+@dataclass
+class XLabsControlNetFluxOutput:
+    controlnet_double_block_residuals: list[torch.Tensor] | None
+
+
+class XLabsControlNetFlux(torch.nn.Module):
+    """A ControlNet model for FLUX.
+
+    The architecture is very similar to the base FLUX model, with the following differences:
+    - A `controlnet_depth` parameter is passed to control the number of double_blocks that the ControlNet is applied to.
+      In order to keep the ControlNet small, this is typically much less than the depth of the base FLUX model.
+    - There is a set of `controlnet_blocks` that are applied to the output of each double_block.
+    """
+
+    def __init__(self, params: FluxParams, controlnet_depth: int = 2):
+        super().__init__()
+
+        self.params = params
+        self.in_channels = params.in_channels
+        self.out_channels = self.in_channels
+        if params.hidden_size % params.num_heads != 0:
+            raise ValueError(f"Hidden size {params.hidden_size} must be divisible by num_heads {params.num_heads}")
+        pe_dim = params.hidden_size // params.num_heads
+        if sum(params.axes_dim) != pe_dim:
+            raise ValueError(f"Got {params.axes_dim} but expected positional dim {pe_dim}")
+        self.hidden_size = params.hidden_size
+        self.num_heads = params.num_heads
+        self.pe_embedder = EmbedND(dim=pe_dim, theta=params.theta, axes_dim=params.axes_dim)
+        self.img_in = torch.nn.Linear(self.in_channels, self.hidden_size, bias=True)
+        self.time_in = MLPEmbedder(in_dim=256, hidden_dim=self.hidden_size)
+        self.vector_in = MLPEmbedder(params.vec_in_dim, self.hidden_size)
+        self.guidance_in = (
+            MLPEmbedder(in_dim=256, hidden_dim=self.hidden_size) if params.guidance_embed else torch.nn.Identity()
+        )
+        self.txt_in = torch.nn.Linear(params.context_in_dim, self.hidden_size)
+
+        self.double_blocks = torch.nn.ModuleList(
+            [
+                DoubleStreamBlock(
+                    self.hidden_size,
+                    self.num_heads,
+                    mlp_ratio=params.mlp_ratio,
+                    qkv_bias=params.qkv_bias,
+                )
+                for _ in range(controlnet_depth)
+            ]
+        )
+
+        # Add ControlNet blocks.
+        self.controlnet_blocks = torch.nn.ModuleList([])
+        for _ in range(controlnet_depth):
+            controlnet_block = torch.nn.Linear(self.hidden_size, self.hidden_size)
+            controlnet_block = zero_module(controlnet_block)
+            self.controlnet_blocks.append(controlnet_block)
+        self.pos_embed_input = torch.nn.Linear(self.in_channels, self.hidden_size, bias=True)
+        self.input_hint_block = torch.nn.Sequential(
+            torch.nn.Conv2d(3, 16, 3, padding=1),
+            torch.nn.SiLU(),
+            torch.nn.Conv2d(16, 16, 3, padding=1),
+            torch.nn.SiLU(),
+            torch.nn.Conv2d(16, 16, 3, padding=1, stride=2),
+            torch.nn.SiLU(),
+            torch.nn.Conv2d(16, 16, 3, padding=1),
+            torch.nn.SiLU(),
+            torch.nn.Conv2d(16, 16, 3, padding=1, stride=2),
+            torch.nn.SiLU(),
+            torch.nn.Conv2d(16, 16, 3, padding=1),
+            torch.nn.SiLU(),
+            torch.nn.Conv2d(16, 16, 3, padding=1, stride=2),
+            torch.nn.SiLU(),
+            zero_module(torch.nn.Conv2d(16, 16, 3, padding=1)),
+        )
+
+    def forward(
+        self,
+        img: torch.Tensor,
+        img_ids: torch.Tensor,
+        controlnet_cond: torch.Tensor,
+        txt: torch.Tensor,
+        txt_ids: torch.Tensor,
+        timesteps: torch.Tensor,
+        y: torch.Tensor,
+        guidance: torch.Tensor | None = None,
+    ) -> XLabsControlNetFluxOutput:
+        if img.ndim != 3 or txt.ndim != 3:
+            raise ValueError("Input img and txt tensors must have 3 dimensions.")
+
+        # running on sequences img
+        img = self.img_in(img)
+        controlnet_cond = self.input_hint_block(controlnet_cond)
+        controlnet_cond = rearrange(controlnet_cond, "b c (h ph) (w pw) -> b (h w) (c ph pw)", ph=2, pw=2)
+        controlnet_cond = self.pos_embed_input(controlnet_cond)
+        img = img + controlnet_cond
+        vec = self.time_in(timestep_embedding(timesteps, 256))
+        if self.params.guidance_embed:
+            if guidance is None:
+                raise ValueError("Didn't get guidance strength for guidance distilled model.")
+            vec = vec + self.guidance_in(timestep_embedding(guidance, 256))
+        vec = vec + self.vector_in(y)
+        txt = self.txt_in(txt)
+
+        ids = torch.cat((txt_ids, img_ids), dim=1)
+        pe = self.pe_embedder(ids)
+
+        block_res_samples: list[torch.Tensor] = []
+
+        for block in self.double_blocks:
+            img, txt = block(img=img, txt=txt, vec=vec, pe=pe)
+            block_res_samples.append(img)
+
+        controlnet_block_res_samples: list[torch.Tensor] = []
+        for block_res_sample, controlnet_block in zip(block_res_samples, self.controlnet_blocks, strict=True):
+            block_res_sample = controlnet_block(block_res_sample)
+            controlnet_block_res_samples.append(block_res_sample)
+
+        return XLabsControlNetFluxOutput(controlnet_double_block_residuals=controlnet_block_res_samples)

invokeai/backend/flux/controlnet/zero_module.py

@@ -0,0 +1,12 @@
+from typing import TypeVar
+
+import torch
+
+T = TypeVar("T", bound=torch.nn.Module)
+
+
+def zero_module(module: T) -> T:
+    """Initialize the parameters of a module to zero."""
+    for p in module.parameters():
+        torch.nn.init.zeros_(p)
+    return module

invokeai/backend/flux/custom_block_processor.py

@@ -0,0 +1,83 @@
+import einops
+import torch
+
+from invokeai.backend.flux.extensions.xlabs_ip_adapter_extension import XLabsIPAdapterExtension
+from invokeai.backend.flux.math import attention
+from invokeai.backend.flux.modules.layers import DoubleStreamBlock
+
+
+class CustomDoubleStreamBlockProcessor:
+    """A class containing a custom implementation of DoubleStreamBlock.forward() with additional features
+    (IP-Adapter, etc.).
+    """
+
+    @staticmethod
+    def _double_stream_block_forward(
+        block: DoubleStreamBlock, img: torch.Tensor, txt: torch.Tensor, vec: torch.Tensor, pe: torch.Tensor
+    ) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
+        """This function is a direct copy of DoubleStreamBlock.forward(), but it returns some of the intermediate
+        values.
+        """
+        img_mod1, img_mod2 = block.img_mod(vec)
+        txt_mod1, txt_mod2 = block.txt_mod(vec)
+
+        # prepare image for attention
+        img_modulated = block.img_norm1(img)
+        img_modulated = (1 + img_mod1.scale) * img_modulated + img_mod1.shift
+        img_qkv = block.img_attn.qkv(img_modulated)
+        img_q, img_k, img_v = einops.rearrange(img_qkv, "B L (K H D) -> K B H L D", K=3, H=block.num_heads)
+        img_q, img_k = block.img_attn.norm(img_q, img_k, img_v)
+
+        # prepare txt for attention
+        txt_modulated = block.txt_norm1(txt)
+        txt_modulated = (1 + txt_mod1.scale) * txt_modulated + txt_mod1.shift
+        txt_qkv = block.txt_attn.qkv(txt_modulated)
+        txt_q, txt_k, txt_v = einops.rearrange(txt_qkv, "B L (K H D) -> K B H L D", K=3, H=block.num_heads)
+        txt_q, txt_k = block.txt_attn.norm(txt_q, txt_k, txt_v)
+
+        # run actual attention
+        q = torch.cat((txt_q, img_q), dim=2)
+        k = torch.cat((txt_k, img_k), dim=2)
+        v = torch.cat((txt_v, img_v), dim=2)
+
+        attn = attention(q, k, v, pe=pe)
+        txt_attn, img_attn = attn[:, : txt.shape[1]], attn[:, txt.shape[1] :]
+
+        # calculate the img bloks
+        img = img + img_mod1.gate * block.img_attn.proj(img_attn)
+        img = img + img_mod2.gate * block.img_mlp((1 + img_mod2.scale) * block.img_norm2(img) + img_mod2.shift)
+
+        # calculate the txt bloks
+        txt = txt + txt_mod1.gate * block.txt_attn.proj(txt_attn)
+        txt = txt + txt_mod2.gate * block.txt_mlp((1 + txt_mod2.scale) * block.txt_norm2(txt) + txt_mod2.shift)
+        return img, txt, img_q
+
+    @staticmethod
+    def custom_double_block_forward(
+        timestep_index: int,
+        total_num_timesteps: int,
+        block_index: int,
+        block: DoubleStreamBlock,
+        img: torch.Tensor,
+        txt: torch.Tensor,
+        vec: torch.Tensor,
+        pe: torch.Tensor,
+        ip_adapter_extensions: list[XLabsIPAdapterExtension],
+    ) -> tuple[torch.Tensor, torch.Tensor]:
+        """A custom implementation of DoubleStreamBlock.forward() with additional features:
+        - IP-Adapter support
+        """
+        img, txt, img_q = CustomDoubleStreamBlockProcessor._double_stream_block_forward(block, img, txt, vec, pe)
+
+        # Apply IP-Adapter conditioning.
+        for ip_adapter_extension in ip_adapter_extensions:
+            img = ip_adapter_extension.run_ip_adapter(
+                timestep_index=timestep_index,
+                total_num_timesteps=total_num_timesteps,
+                block_index=block_index,
+                block=block,
+                img_q=img_q,
+                img=img,
+            )
+
+        return img, txt

invokeai/backend/flux/denoise.py

@@ -1,10 +1,15 @@
+import math
 from typing import Callable
 
 import torch
 from tqdm import tqdm
 
+from invokeai.backend.flux.controlnet.controlnet_flux_output import ControlNetFluxOutput, sum_controlnet_flux_outputs
+from invokeai.backend.flux.extensions.inpaint_extension import InpaintExtension
+from invokeai.backend.flux.extensions.instantx_controlnet_extension import InstantXControlNetExtension
+from invokeai.backend.flux.extensions.xlabs_controlnet_extension import XLabsControlNetExtension
+from invokeai.backend.flux.extensions.xlabs_ip_adapter_extension import XLabsIPAdapterExtension
 from invokeai.backend.flux.model import Flux
-from invokeai.backend.flux.trajectory_guidance_extension import TrajectoryGuidanceExtension
 from invokeai.backend.stable_diffusion.diffusers_pipeline import PipelineIntermediateState
 
 
@@ -13,14 +18,23 @@ def denoise(
     # model input
     img: torch.Tensor,
     img_ids: torch.Tensor,
+    # positive text conditioning
     txt: torch.Tensor,
     txt_ids: torch.Tensor,
     vec: torch.Tensor,
+    # negative text conditioning
+    neg_txt: torch.Tensor | None,
+    neg_txt_ids: torch.Tensor | None,
+    neg_vec: torch.Tensor | None,
     # sampling parameters
     timesteps: list[float],
     step_callback: Callable[[PipelineIntermediateState], None],
     guidance: float,
-    traj_guidance_extension: TrajectoryGuidanceExtension | None,  # noqa: F821
+    cfg_scale: list[float],
+    inpaint_extension: InpaintExtension | None,
+    controlnet_extensions: list[XLabsControlNetExtension | InstantXControlNetExtension],
+    pos_ip_adapter_extensions: list[XLabsIPAdapterExtension],
+    neg_ip_adapter_extensions: list[XLabsIPAdapterExtension],
 ):
     # step 0 is the initial state
     total_steps = len(timesteps) - 1
@@ -33,11 +47,34 @@ def denoise(
             latents=img,
         ),
     )
-    step = 1
     # guidance_vec is ignored for schnell.
     guidance_vec = torch.full((img.shape[0],), guidance, device=img.device, dtype=img.dtype)
-    for t_curr, t_prev in tqdm(list(zip(timesteps[:-1], timesteps[1:], strict=True))):
+    for step_index, (t_curr, t_prev) in tqdm(list(enumerate(zip(timesteps[:-1], timesteps[1:], strict=True)))):
         t_vec = torch.full((img.shape[0],), t_curr, dtype=img.dtype, device=img.device)
+
+        # Run ControlNet models.
+        controlnet_residuals: list[ControlNetFluxOutput] = []
+        for controlnet_extension in controlnet_extensions:
+            controlnet_residuals.append(
+                controlnet_extension.run_controlnet(
+                    timestep_index=step_index,
+                    total_num_timesteps=total_steps,
+                    img=img,
+                    img_ids=img_ids,
+                    txt=txt,
+                    txt_ids=txt_ids,
+                    y=vec,
+                    timesteps=t_vec,
+                    guidance=guidance_vec,
+                )
+            )
+
+        # Merge the ControlNet residuals from multiple ControlNets.
+        # TODO(ryand): We may want to calculate the sum just-in-time to keep peak memory low. Keep in mind, that the
+        # controlnet_residuals datastructure is efficient in that it likely contains multiple references to the same
+        # tensors. Calculating the sum materializes each tensor into its own instance.
+        merged_controlnet_residuals = sum_controlnet_flux_outputs(controlnet_residuals)
+
         pred = model(
             img=img,
             img_ids=img_ids,
@@ -46,25 +83,54 @@ def denoise(
             y=vec,
             timesteps=t_vec,
             guidance=guidance_vec,
+            timestep_index=step_index,
+            total_num_timesteps=total_steps,
+            controlnet_double_block_residuals=merged_controlnet_residuals.double_block_residuals,
+            controlnet_single_block_residuals=merged_controlnet_residuals.single_block_residuals,
+            ip_adapter_extensions=pos_ip_adapter_extensions,
         )
 
-        if traj_guidance_extension is not None:
-            pred = traj_guidance_extension.update_noise(
-                t_curr_latents=img, pred_noise=pred, t_curr=t_curr, t_prev=t_prev
+        step_cfg_scale = cfg_scale[step_index]
+
+        # If step_cfg_scale, is 1.0, then we don't need to run the negative prediction.
+        if not math.isclose(step_cfg_scale, 1.0):
+            # TODO(ryand): Add option to run positive and negative predictions in a single batch for better performance
+            # on systems with sufficient VRAM.
+
+            if neg_txt is None or neg_txt_ids is None or neg_vec is None:
+                raise ValueError("Negative text conditioning is required when cfg_scale is not 1.0.")
+
+            neg_pred = model(
+                img=img,
+                img_ids=img_ids,
+                txt=neg_txt,
+                txt_ids=neg_txt_ids,
+                y=neg_vec,
+                timesteps=t_vec,
+                guidance=guidance_vec,
+                timestep_index=step_index,
+                total_num_timesteps=total_steps,
+                controlnet_double_block_residuals=None,
+                controlnet_single_block_residuals=None,
+                ip_adapter_extensions=neg_ip_adapter_extensions,
             )
+            pred = neg_pred + step_cfg_scale * (pred - neg_pred)
 
         preview_img = img - t_curr * pred
         img = img + (t_prev - t_curr) * pred
 
+        if inpaint_extension is not None:
+            img = inpaint_extension.merge_intermediate_latents_with_init_latents(img, t_prev)
+            preview_img = inpaint_extension.merge_intermediate_latents_with_init_latents(preview_img, 0.0)
+
         step_callback(
             PipelineIntermediateState(
-                step=step,
+                step=step_index + 1,
                 order=1,
                 total_steps=total_steps,
                 timestep=int(t_curr),
                 latents=preview_img,
             ),
         )
-        step += 1
 
     return img

invokeai/backend/flux/extensions/base_controlnet_extension.py

@@ -0,0 +1,45 @@
+import math
+from abc import ABC, abstractmethod
+from typing import List, Union
+
+import torch
+
+from invokeai.backend.flux.controlnet.controlnet_flux_output import ControlNetFluxOutput
+
+
+class BaseControlNetExtension(ABC):
+    def __init__(
+        self,
+        weight: Union[float, List[float]],
+        begin_step_percent: float,
+        end_step_percent: float,
+    ):
+        self._weight = weight
+        self._begin_step_percent = begin_step_percent
+        self._end_step_percent = end_step_percent
+
+    def _get_weight(self, timestep_index: int, total_num_timesteps: int) -> float:
+        first_step = math.floor(self._begin_step_percent * total_num_timesteps)
+        last_step = math.ceil(self._end_step_percent * total_num_timesteps)
+
+        if timestep_index < first_step or timestep_index > last_step:
+            return 0.0
+
+        if isinstance(self._weight, list):
+            return self._weight[timestep_index]
+
+        return self._weight
+
+    @abstractmethod
+    def run_controlnet(
+        self,
+        timestep_index: int,
+        total_num_timesteps: int,
+        img: torch.Tensor,
+        img_ids: torch.Tensor,
+        txt: torch.Tensor,
+        txt_ids: torch.Tensor,
+        y: torch.Tensor,
+        timesteps: torch.Tensor,
+        guidance: torch.Tensor | None,
+    ) -> ControlNetFluxOutput: ...

invokeai/backend/flux/extensions/inpaint_extension.py

@@ -19,8 +19,26 @@ class InpaintExtension:
         self._inpaint_mask = inpaint_mask
         self._noise = noise
 
+    def _apply_mask_gradient_adjustment(self, t_prev: float) -> torch.Tensor:
+        """Applies inpaint mask gradient adjustment and returns the inpaint mask to be used at the current timestep."""
+        # As we progress through the denoising process, we promote gradient regions of the mask to have a full weight of
+        # 1.0. This helps to produce more coherent seams around the inpainted region. We experimented with a (small)
+        # number of promotion strategies (e.g. gradual promotion based on timestep), but found that a simple cutoff
+        # threshold worked well.
+        # We use a small epsilon to avoid any potential issues with floating point precision.
+        eps = 1e-4
+        mask_gradient_t_cutoff = 0.5
+        if t_prev > mask_gradient_t_cutoff:
+            # Early in the denoising process, use the inpaint mask as-is.
+            return self._inpaint_mask
+        else:
+            # After the cut-off, promote all non-zero mask values to 1.0.
+            mask = self._inpaint_mask.where(self._inpaint_mask <= (0.0 + eps), 1.0)
+
+        return mask
+
     def merge_intermediate_latents_with_init_latents(
-        self, intermediate_latents: torch.Tensor, timestep: float
+        self, intermediate_latents: torch.Tensor, t_prev: float
     ) -> torch.Tensor:
         """Merge the intermediate latents with the initial latents for the current timestep using the inpaint mask. I.e.
         update the intermediate latents to keep the regions that are not being inpainted on the correct noise
@@ -28,8 +46,10 @@ class InpaintExtension:
 
         This function should be called after each denoising step.
         """
+        mask = self._apply_mask_gradient_adjustment(t_prev)
+
         # Noise the init latents for the current timestep.
-        noised_init_latents = self._noise * timestep + (1.0 - timestep) * self._init_latents
+        noised_init_latents = self._noise * t_prev + (1.0 - t_prev) * self._init_latents
 
         # Merge the intermediate latents with the noised_init_latents using the inpaint_mask.
-        return intermediate_latents * self._inpaint_mask + noised_init_latents * (1.0 - self._inpaint_mask)
+        return intermediate_latents * mask + noised_init_latents * (1.0 - mask)

invokeai/backend/flux/extensions/instantx_controlnet_extension.py

@@ -0,0 +1,194 @@
+import math
+from typing import List, Union
+
+import torch
+from PIL.Image import Image
+
+from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
+from invokeai.app.invocations.flux_vae_encode import FluxVaeEncodeInvocation
+from invokeai.app.util.controlnet_utils import CONTROLNET_RESIZE_VALUES, prepare_control_image
+from invokeai.backend.flux.controlnet.controlnet_flux_output import ControlNetFluxOutput
+from invokeai.backend.flux.controlnet.instantx_controlnet_flux import (
+    InstantXControlNetFlux,
+    InstantXControlNetFluxOutput,
+)
+from invokeai.backend.flux.extensions.base_controlnet_extension import BaseControlNetExtension
+from invokeai.backend.flux.sampling_utils import pack
+from invokeai.backend.model_manager.load.load_base import LoadedModel
+
+
+class InstantXControlNetExtension(BaseControlNetExtension):
+    def __init__(
+        self,
+        model: InstantXControlNetFlux,
+        controlnet_cond: torch.Tensor,
+        instantx_control_mode: torch.Tensor | None,
+        weight: Union[float, List[float]],
+        begin_step_percent: float,
+        end_step_percent: float,
+    ):
+        super().__init__(
+            weight=weight,
+            begin_step_percent=begin_step_percent,
+            end_step_percent=end_step_percent,
+        )
+        self._model = model
+        # The VAE-encoded and 'packed' control image to pass to the ControlNet model.
+        self._controlnet_cond = controlnet_cond
+        # TODO(ryand): Should we define an enum for the instantx_control_mode? Is it likely to change for future models?
+        # The control mode for InstantX ControlNet union models.
+        # See the values defined here: https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Union#control-mode
+        # Expected shape: (batch_size, 1), Expected dtype: torch.long
+        # If None, a zero-embedding will be used.
+        self._instantx_control_mode = instantx_control_mode
+
+        # TODO(ryand): Pass in these params if a new base transformer / InstantX ControlNet pair get released.
+        self._flux_transformer_num_double_blocks = 19
+        self._flux_transformer_num_single_blocks = 38
+
+    @classmethod
+    def prepare_controlnet_cond(
+        cls,
+        controlnet_image: Image,
+        vae_info: LoadedModel,
+        latent_height: int,
+        latent_width: int,
+        dtype: torch.dtype,
+        device: torch.device,
+        resize_mode: CONTROLNET_RESIZE_VALUES,
+    ):
+        image_height = latent_height * LATENT_SCALE_FACTOR
+        image_width = latent_width * LATENT_SCALE_FACTOR
+
+        resized_controlnet_image = prepare_control_image(
+            image=controlnet_image,
+            do_classifier_free_guidance=False,
+            width=image_width,
+            height=image_height,
+            device=device,
+            dtype=dtype,
+            control_mode="balanced",
+            resize_mode=resize_mode,
+        )
+
+        # Shift the image from [0, 1] to [-1, 1].
+        resized_controlnet_image = resized_controlnet_image * 2 - 1
+
+        # Run VAE encoder.
+        controlnet_cond = FluxVaeEncodeInvocation.vae_encode(vae_info=vae_info, image_tensor=resized_controlnet_image)
+        controlnet_cond = pack(controlnet_cond)
+
+        return controlnet_cond
+
+    @classmethod
+    def from_controlnet_image(
+        cls,
+        model: InstantXControlNetFlux,
+        controlnet_image: Image,
+        instantx_control_mode: torch.Tensor | None,
+        vae_info: LoadedModel,
+        latent_height: int,
+        latent_width: int,
+        dtype: torch.dtype,
+        device: torch.device,
+        resize_mode: CONTROLNET_RESIZE_VALUES,
+        weight: Union[float, List[float]],
+        begin_step_percent: float,
+        end_step_percent: float,
+    ):
+        image_height = latent_height * LATENT_SCALE_FACTOR
+        image_width = latent_width * LATENT_SCALE_FACTOR
+
+        resized_controlnet_image = prepare_control_image(
+            image=controlnet_image,
+            do_classifier_free_guidance=False,
+            width=image_width,
+            height=image_height,
+            device=device,
+            dtype=dtype,
+            control_mode="balanced",
+            resize_mode=resize_mode,
+        )
+
+        # Shift the image from [0, 1] to [-1, 1].
+        resized_controlnet_image = resized_controlnet_image * 2 - 1
+
+        # Run VAE encoder.
+        controlnet_cond = FluxVaeEncodeInvocation.vae_encode(vae_info=vae_info, image_tensor=resized_controlnet_image)
+        controlnet_cond = pack(controlnet_cond)
+
+        return cls(
+            model=model,
+            controlnet_cond=controlnet_cond,
+            instantx_control_mode=instantx_control_mode,
+            weight=weight,
+            begin_step_percent=begin_step_percent,
+            end_step_percent=end_step_percent,
+        )
+
+    def _instantx_output_to_controlnet_output(
+        self, instantx_output: InstantXControlNetFluxOutput
+    ) -> ControlNetFluxOutput:
+        # The `interval_control` logic here is based on
+        # https://github.com/huggingface/diffusers/blob/31058cdaef63ca660a1a045281d156239fba8192/src/diffusers/models/transformers/transformer_flux.py#L507-L511
+
+        # Handle double block residuals.
+        double_block_residuals: list[torch.Tensor] = []
+        double_block_samples = instantx_output.controlnet_block_samples
+        if double_block_samples:
+            interval_control = self._flux_transformer_num_double_blocks / len(double_block_samples)
+            interval_control = int(math.ceil(interval_control))
+            for i in range(self._flux_transformer_num_double_blocks):
+                double_block_residuals.append(double_block_samples[i // interval_control])
+
+        # Handle single block residuals.
+        single_block_residuals: list[torch.Tensor] = []
+        single_block_samples = instantx_output.controlnet_single_block_samples
+        if single_block_samples:
+            interval_control = self._flux_transformer_num_single_blocks / len(single_block_samples)
+            interval_control = int(math.ceil(interval_control))
+            for i in range(self._flux_transformer_num_single_blocks):
+                single_block_residuals.append(single_block_samples[i // interval_control])
+
+        return ControlNetFluxOutput(
+            double_block_residuals=double_block_residuals or None,
+            single_block_residuals=single_block_residuals or None,
+        )
+
+    def run_controlnet(
+        self,
+        timestep_index: int,
+        total_num_timesteps: int,
+        img: torch.Tensor,
+        img_ids: torch.Tensor,
+        txt: torch.Tensor,
+        txt_ids: torch.Tensor,
+        y: torch.Tensor,
+        timesteps: torch.Tensor,
+        guidance: torch.Tensor | None,
+    ) -> ControlNetFluxOutput:
+        weight = self._get_weight(timestep_index=timestep_index, total_num_timesteps=total_num_timesteps)
+        if weight < 1e-6:
+            return ControlNetFluxOutput(single_block_residuals=None, double_block_residuals=None)
+
+        # Make sure inputs have correct device and dtype.
+        self._controlnet_cond = self._controlnet_cond.to(device=img.device, dtype=img.dtype)
+        self._instantx_control_mode = (
+            self._instantx_control_mode.to(device=img.device) if self._instantx_control_mode is not None else None
+        )
+
+        instantx_output: InstantXControlNetFluxOutput = self._model(
+            controlnet_cond=self._controlnet_cond,
+            controlnet_mode=self._instantx_control_mode,
+            img=img,
+            img_ids=img_ids,
+            txt=txt,
+            txt_ids=txt_ids,
+            timesteps=timesteps,
+            y=y,
+            guidance=guidance,
+        )
+
+        controlnet_output = self._instantx_output_to_controlnet_output(instantx_output)
+        controlnet_output.apply_weight(weight)
+        return controlnet_output

invokeai/backend/flux/extensions/xlabs_controlnet_extension.py

@@ -0,0 +1,150 @@
+from typing import List, Union
+
+import torch
+from PIL.Image import Image
+
+from invokeai.app.invocations.constants import LATENT_SCALE_FACTOR
+from invokeai.app.util.controlnet_utils import CONTROLNET_RESIZE_VALUES, prepare_control_image
+from invokeai.backend.flux.controlnet.controlnet_flux_output import ControlNetFluxOutput
+from invokeai.backend.flux.controlnet.xlabs_controlnet_flux import XLabsControlNetFlux, XLabsControlNetFluxOutput
+from invokeai.backend.flux.extensions.base_controlnet_extension import BaseControlNetExtension
+
+
+class XLabsControlNetExtension(BaseControlNetExtension):
+    def __init__(
+        self,
+        model: XLabsControlNetFlux,
+        controlnet_cond: torch.Tensor,
+        weight: Union[float, List[float]],
+        begin_step_percent: float,
+        end_step_percent: float,
+    ):
+        super().__init__(
+            weight=weight,
+            begin_step_percent=begin_step_percent,
+            end_step_percent=end_step_percent,
+        )
+
+        self._model = model
+        # _controlnet_cond is the control image passed to the ControlNet model.
+        # Pixel values are in the range [-1, 1]. Shape: (batch_size, 3, height, width).
+        self._controlnet_cond = controlnet_cond
+
+        # TODO(ryand): Pass in these params if a new base transformer / XLabs ControlNet pair get released.
+        self._flux_transformer_num_double_blocks = 19
+        self._flux_transformer_num_single_blocks = 38
+
+    @classmethod
+    def prepare_controlnet_cond(
+        cls,
+        controlnet_image: Image,
+        latent_height: int,
+        latent_width: int,
+        dtype: torch.dtype,
+        device: torch.device,
+        resize_mode: CONTROLNET_RESIZE_VALUES,
+    ):
+        image_height = latent_height * LATENT_SCALE_FACTOR
+        image_width = latent_width * LATENT_SCALE_FACTOR
+
+        controlnet_cond = prepare_control_image(
+            image=controlnet_image,
+            do_classifier_free_guidance=False,
+            width=image_width,
+            height=image_height,
+            device=device,
+            dtype=dtype,
+            control_mode="balanced",
+            resize_mode=resize_mode,
+        )
+
+        # Map pixel values from [0, 1] to [-1, 1].
+        controlnet_cond = controlnet_cond * 2 - 1
+
+        return controlnet_cond
+
+    @classmethod
+    def from_controlnet_image(
+        cls,
+        model: XLabsControlNetFlux,
+        controlnet_image: Image,
+        latent_height: int,
+        latent_width: int,
+        dtype: torch.dtype,
+        device: torch.device,
+        resize_mode: CONTROLNET_RESIZE_VALUES,
+        weight: Union[float, List[float]],
+        begin_step_percent: float,
+        end_step_percent: float,
+    ):
+        image_height = latent_height * LATENT_SCALE_FACTOR
+        image_width = latent_width * LATENT_SCALE_FACTOR
+
+        controlnet_cond = prepare_control_image(
+            image=controlnet_image,
+            do_classifier_free_guidance=False,
+            width=image_width,
+            height=image_height,
+            device=device,
+            dtype=dtype,
+            control_mode="balanced",
+            resize_mode=resize_mode,
+        )
+
+        # Map pixel values from [0, 1] to [-1, 1].
+        controlnet_cond = controlnet_cond * 2 - 1
+
+        return cls(
+            model=model,
+            controlnet_cond=controlnet_cond,
+            weight=weight,
+            begin_step_percent=begin_step_percent,
+            end_step_percent=end_step_percent,
+        )
+
+    def _xlabs_output_to_controlnet_output(self, xlabs_output: XLabsControlNetFluxOutput) -> ControlNetFluxOutput:
+        # The modulo index logic used here is based on:
+        # https://github.com/XLabs-AI/x-flux/blob/47495425dbed499be1e8e5a6e52628b07349cba2/src/flux/model.py#L198-L200
+
+        # Handle double block residuals.
+        double_block_residuals: list[torch.Tensor] = []
+        xlabs_double_block_residuals = xlabs_output.controlnet_double_block_residuals
+        if xlabs_double_block_residuals is not None:
+            for i in range(self._flux_transformer_num_double_blocks):
+                double_block_residuals.append(xlabs_double_block_residuals[i % len(xlabs_double_block_residuals)])
+
+        return ControlNetFluxOutput(
+            double_block_residuals=double_block_residuals,
+            single_block_residuals=None,
+        )
+
+    def run_controlnet(
+        self,
+        timestep_index: int,
+        total_num_timesteps: int,
+        img: torch.Tensor,
+        img_ids: torch.Tensor,
+        txt: torch.Tensor,
+        txt_ids: torch.Tensor,
+        y: torch.Tensor,
+        timesteps: torch.Tensor,
+        guidance: torch.Tensor | None,
+    ) -> ControlNetFluxOutput:
+        weight = self._get_weight(timestep_index=timestep_index, total_num_timesteps=total_num_timesteps)
+        if weight < 1e-6:
+            return ControlNetFluxOutput(single_block_residuals=None, double_block_residuals=None)
+
+        xlabs_output: XLabsControlNetFluxOutput = self._model(
+            img=img,
+            img_ids=img_ids,
+            controlnet_cond=self._controlnet_cond,
+            txt=txt,
+            txt_ids=txt_ids,
+            timesteps=timesteps,
+            y=y,
+            guidance=guidance,
+        )
+
+        controlnet_output = self._xlabs_output_to_controlnet_output(xlabs_output)
+        controlnet_output.apply_weight(weight)
+        return controlnet_output

invokeai/backend/flux/extensions/xlabs_ip_adapter_extension.py

@@ -0,0 +1,89 @@
+import math
+from typing import List, Union
+
+import einops
+import torch
+from PIL import Image
+from transformers import CLIPImageProcessor, CLIPVisionModelWithProjection
+
+from invokeai.backend.flux.ip_adapter.xlabs_ip_adapter_flux import XlabsIpAdapterFlux
+from invokeai.backend.flux.modules.layers import DoubleStreamBlock
+
+
+class XLabsIPAdapterExtension:
+    def __init__(
+        self,
+        model: XlabsIpAdapterFlux,
+        image_prompt_clip_embed: torch.Tensor,
+        weight: Union[float, List[float]],
+        begin_step_percent: float,
+        end_step_percent: float,
+    ):
+        self._model = model
+        self._image_prompt_clip_embed = image_prompt_clip_embed
+        self._weight = weight
+        self._begin_step_percent = begin_step_percent
+        self._end_step_percent = end_step_percent
+
+        self._image_proj: torch.Tensor | None = None
+
+    def _get_weight(self, timestep_index: int, total_num_timesteps: int) -> float:
+        first_step = math.floor(self._begin_step_percent * total_num_timesteps)
+        last_step = math.ceil(self._end_step_percent * total_num_timesteps)
+
+        if timestep_index < first_step or timestep_index > last_step:
+            return 0.0
+
+        if isinstance(self._weight, list):
+            return self._weight[timestep_index]
+
+        return self._weight
+
+    @staticmethod
+    def run_clip_image_encoder(
+        pil_image: List[Image.Image], image_encoder: CLIPVisionModelWithProjection
+    ) -> torch.Tensor:
+        clip_image_processor = CLIPImageProcessor()
+        clip_image: torch.Tensor = clip_image_processor(images=pil_image, return_tensors="pt").pixel_values
+        clip_image = clip_image.to(device=image_encoder.device, dtype=image_encoder.dtype)
+        clip_image_embeds = image_encoder(clip_image).image_embeds
+        return clip_image_embeds
+
+    def run_image_proj(self, dtype: torch.dtype):
+        image_prompt_clip_embed = self._image_prompt_clip_embed.to(dtype=dtype)
+        self._image_proj = self._model.image_proj(image_prompt_clip_embed)
+
+    def run_ip_adapter(
+        self,
+        timestep_index: int,
+        total_num_timesteps: int,
+        block_index: int,
+        block: DoubleStreamBlock,
+        img_q: torch.Tensor,
+        img: torch.Tensor,
+    ) -> torch.Tensor:
+        """The logic in this function is based on:
+        https://github.com/XLabs-AI/x-flux/blob/47495425dbed499be1e8e5a6e52628b07349cba2/src/flux/modules/layers.py#L245-L301
+        """
+        weight = self._get_weight(timestep_index=timestep_index, total_num_timesteps=total_num_timesteps)
+        if weight < 1e-6:
+            return img
+
+        ip_adapter_block = self._model.ip_adapter_double_blocks.double_blocks[block_index]
+
+        ip_key = ip_adapter_block.ip_adapter_double_stream_k_proj(self._image_proj)
+        ip_value = ip_adapter_block.ip_adapter_double_stream_v_proj(self._image_proj)
+
+        # Reshape projections for multi-head attention.
+        ip_key = einops.rearrange(ip_key, "B L (H D) -> B H L D", H=block.num_heads)
+        ip_value = einops.rearrange(ip_value, "B L (H D) -> B H L D", H=block.num_heads)
+
+        # Compute attention between IP projections and the latent query.
+        ip_attn = torch.nn.functional.scaled_dot_product_attention(
+            img_q, ip_key, ip_value, dropout_p=0.0, is_causal=False
+        )
+        ip_attn = einops.rearrange(ip_attn, "B H L D -> B L (H D)", H=block.num_heads)
+
+        img = img + weight * ip_attn
+
+        return img

invokeai/backend/flux/ip_adapter/ip_double_stream_block_processor.py

@@ -0,0 +1,93 @@
+# This file is based on:
+# https://github.com/XLabs-AI/x-flux/blob/47495425dbed499be1e8e5a6e52628b07349cba2/src/flux/modules/layers.py#L221
+import einops
+import torch
+
+from invokeai.backend.flux.math import attention
+from invokeai.backend.flux.modules.layers import DoubleStreamBlock
+
+
+class IPDoubleStreamBlockProcessor(torch.nn.Module):
+    """Attention processor for handling IP-adapter with double stream block."""
+
+    def __init__(self, context_dim: int, hidden_dim: int):
+        super().__init__()
+
+        # Ensure context_dim matches the dimension of image_proj
+        self.context_dim = context_dim
+        self.hidden_dim = hidden_dim
+
+        # Initialize projections for IP-adapter
+        self.ip_adapter_double_stream_k_proj = torch.nn.Linear(context_dim, hidden_dim, bias=True)
+        self.ip_adapter_double_stream_v_proj = torch.nn.Linear(context_dim, hidden_dim, bias=True)
+
+        torch.nn.init.zeros_(self.ip_adapter_double_stream_k_proj.weight)
+        torch.nn.init.zeros_(self.ip_adapter_double_stream_k_proj.bias)
+
+        torch.nn.init.zeros_(self.ip_adapter_double_stream_v_proj.weight)
+        torch.nn.init.zeros_(self.ip_adapter_double_stream_v_proj.bias)
+
+    def __call__(
+        self,
+        attn: DoubleStreamBlock,
+        img: torch.Tensor,
+        txt: torch.Tensor,
+        vec: torch.Tensor,
+        pe: torch.Tensor,
+        image_proj: torch.Tensor,
+        ip_scale: float = 1.0,
+    ):
+        # Prepare image for attention
+        img_mod1, img_mod2 = attn.img_mod(vec)
+        txt_mod1, txt_mod2 = attn.txt_mod(vec)
+
+        img_modulated = attn.img_norm1(img)
+        img_modulated = (1 + img_mod1.scale) * img_modulated + img_mod1.shift
+        img_qkv = attn.img_attn.qkv(img_modulated)
+        img_q, img_k, img_v = einops.rearrange(
+            img_qkv, "B L (K H D) -> K B H L D", K=3, H=attn.num_heads, D=attn.head_dim
+        )
+        img_q, img_k = attn.img_attn.norm(img_q, img_k, img_v)
+
+        txt_modulated = attn.txt_norm1(txt)
+        txt_modulated = (1 + txt_mod1.scale) * txt_modulated + txt_mod1.shift
+        txt_qkv = attn.txt_attn.qkv(txt_modulated)
+        txt_q, txt_k, txt_v = einops.rearrange(
+            txt_qkv, "B L (K H D) -> K B H L D", K=3, H=attn.num_heads, D=attn.head_dim
+        )
+        txt_q, txt_k = attn.txt_attn.norm(txt_q, txt_k, txt_v)
+
+        q = torch.cat((txt_q, img_q), dim=2)
+        k = torch.cat((txt_k, img_k), dim=2)
+        v = torch.cat((txt_v, img_v), dim=2)
+
+        attn1 = attention(q, k, v, pe=pe)
+        txt_attn, img_attn = attn1[:, : txt.shape[1]], attn1[:, txt.shape[1] :]
+
+        # print(f"txt_attn shape: {txt_attn.size()}")
+        # print(f"img_attn shape: {img_attn.size()}")
+
+        img = img + img_mod1.gate * attn.img_attn.proj(img_attn)
+        img = img + img_mod2.gate * attn.img_mlp((1 + img_mod2.scale) * attn.img_norm2(img) + img_mod2.shift)
+
+        txt = txt + txt_mod1.gate * attn.txt_attn.proj(txt_attn)
+        txt = txt + txt_mod2.gate * attn.txt_mlp((1 + txt_mod2.scale) * attn.txt_norm2(txt) + txt_mod2.shift)
+
+        # IP-adapter processing
+        ip_query = img_q  # latent sample query
+        ip_key = self.ip_adapter_double_stream_k_proj(image_proj)
+        ip_value = self.ip_adapter_double_stream_v_proj(image_proj)
+
+        # Reshape projections for multi-head attention
+        ip_key = einops.rearrange(ip_key, "B L (H D) -> B H L D", H=attn.num_heads, D=attn.head_dim)
+        ip_value = einops.rearrange(ip_value, "B L (H D) -> B H L D", H=attn.num_heads, D=attn.head_dim)
+
+        # Compute attention between IP projections and the latent query
+        ip_attention = torch.nn.functional.scaled_dot_product_attention(
+            ip_query, ip_key, ip_value, dropout_p=0.0, is_causal=False
+        )
+        ip_attention = einops.rearrange(ip_attention, "B H L D -> B L (H D)", H=attn.num_heads, D=attn.head_dim)
+
+        img = img + ip_scale * ip_attention
+
+        return img, txt

invokeai/backend/flux/ip_adapter/state_dict_utils.py

@@ -0,0 +1,50 @@
+from typing import Any, Dict
+
+import torch
+
+from invokeai.backend.flux.ip_adapter.xlabs_ip_adapter_flux import XlabsIpAdapterParams
+
+
+def is_state_dict_xlabs_ip_adapter(sd: Dict[str, Any]) -> bool:
+    """Is the state dict for an XLabs FLUX IP-Adapter model?
+
+    This is intended to be a reasonably high-precision detector, but it is not guaranteed to have perfect precision.
+    """
+    # If all of the expected keys are present, then this is very likely an XLabs IP-Adapter model.
+    expected_keys = {
+        "double_blocks.0.processor.ip_adapter_double_stream_k_proj.bias",
+        "double_blocks.0.processor.ip_adapter_double_stream_k_proj.weight",
+        "double_blocks.0.processor.ip_adapter_double_stream_v_proj.bias",
+        "double_blocks.0.processor.ip_adapter_double_stream_v_proj.weight",
+        "ip_adapter_proj_model.norm.bias",
+        "ip_adapter_proj_model.norm.weight",
+        "ip_adapter_proj_model.proj.bias",
+        "ip_adapter_proj_model.proj.weight",
+    }
+
+    if expected_keys.issubset(sd.keys()):
+        return True
+    return False
+
+
+def infer_xlabs_ip_adapter_params_from_state_dict(state_dict: dict[str, torch.Tensor]) -> XlabsIpAdapterParams:
+    num_double_blocks = 0
+    context_dim = 0
+    hidden_dim = 0
+
+    # Count the number of double blocks.
+    double_block_index = 0
+    while f"double_blocks.{double_block_index}.processor.ip_adapter_double_stream_k_proj.weight" in state_dict:
+        double_block_index += 1
+    num_double_blocks = double_block_index
+
+    hidden_dim = state_dict["double_blocks.0.processor.ip_adapter_double_stream_k_proj.weight"].shape[0]
+    context_dim = state_dict["double_blocks.0.processor.ip_adapter_double_stream_k_proj.weight"].shape[1]
+    clip_embeddings_dim = state_dict["ip_adapter_proj_model.proj.weight"].shape[1]
+
+    return XlabsIpAdapterParams(
+        num_double_blocks=num_double_blocks,
+        context_dim=context_dim,
+        hidden_dim=hidden_dim,
+        clip_embeddings_dim=clip_embeddings_dim,
+    )

invokeai/backend/flux/ip_adapter/xlabs_ip_adapter_flux.py

@@ -0,0 +1,67 @@
+from dataclasses import dataclass
+
+import torch
+
+from invokeai.backend.ip_adapter.ip_adapter import ImageProjModel
+
+
+class IPDoubleStreamBlock(torch.nn.Module):
+    def __init__(self, context_dim: int, hidden_dim: int):
+        super().__init__()
+
+        self.context_dim = context_dim
+        self.hidden_dim = hidden_dim
+
+        self.ip_adapter_double_stream_k_proj = torch.nn.Linear(context_dim, hidden_dim, bias=True)
+        self.ip_adapter_double_stream_v_proj = torch.nn.Linear(context_dim, hidden_dim, bias=True)
+
+
+class IPAdapterDoubleBlocks(torch.nn.Module):
+    def __init__(self, num_double_blocks: int, context_dim: int, hidden_dim: int):
+        super().__init__()
+        self.double_blocks = torch.nn.ModuleList(
+            [IPDoubleStreamBlock(context_dim, hidden_dim) for _ in range(num_double_blocks)]
+        )
+
+
+@dataclass
+class XlabsIpAdapterParams:
+    num_double_blocks: int
+    context_dim: int
+    hidden_dim: int
+
+    clip_embeddings_dim: int
+
+
+class XlabsIpAdapterFlux(torch.nn.Module):
+    def __init__(self, params: XlabsIpAdapterParams):
+        super().__init__()
+        self.image_proj = ImageProjModel(
+            cross_attention_dim=params.context_dim, clip_embeddings_dim=params.clip_embeddings_dim
+        )
+        self.ip_adapter_double_blocks = IPAdapterDoubleBlocks(
+            num_double_blocks=params.num_double_blocks, context_dim=params.context_dim, hidden_dim=params.hidden_dim
+        )
+
+    def load_xlabs_state_dict(self, state_dict: dict[str, torch.Tensor], assign: bool = False):
+        """We need this custom function to load state dicts rather than using .load_state_dict(...) because the model
+        structure does not match the state_dict structure.
+        """
+        # Split the state_dict into the image projection model and the double blocks.
+        image_proj_sd: dict[str, torch.Tensor] = {}
+        double_blocks_sd: dict[str, torch.Tensor] = {}
+        for k, v in state_dict.items():
+            if k.startswith("ip_adapter_proj_model."):
+                image_proj_sd[k] = v
+            elif k.startswith("double_blocks."):
+                double_blocks_sd[k] = v
+            else:
+                raise ValueError(f"Unexpected key: {k}")
+
+        # Initialize the image projection model.
+        image_proj_sd = {k.replace("ip_adapter_proj_model.", ""): v for k, v in image_proj_sd.items()}
+        self.image_proj.load_state_dict(image_proj_sd, assign=assign)
+
+        # Initialize the double blocks.
+        double_blocks_sd = {k.replace("processor.", ""): v for k, v in double_blocks_sd.items()}
+        self.ip_adapter_double_blocks.load_state_dict(double_blocks_sd, assign=assign)

invokeai/backend/flux/math.py

@@ -16,7 +16,10 @@ def attention(q: Tensor, k: Tensor, v: Tensor, pe: Tensor) -> Tensor:
 
 def rope(pos: Tensor, dim: int, theta: int) -> Tensor:
     assert dim % 2 == 0
-    scale = torch.arange(0, dim, 2, dtype=torch.float64, device=pos.device) / dim
+    scale = (
+        torch.arange(0, dim, 2, dtype=torch.float32 if pos.device.type == "mps" else torch.float64, device=pos.device)
+        / dim
+    )
     omega = 1.0 / (theta**scale)
     out = torch.einsum("...n,d->...nd", pos, omega)
     out = torch.stack([torch.cos(out), -torch.sin(out), torch.sin(out), torch.cos(out)], dim=-1)

invokeai/backend/flux/model.py

@@ -5,6 +5,8 @@ from dataclasses import dataclass
 import torch
 from torch import Tensor, nn
 
+from invokeai.backend.flux.custom_block_processor import CustomDoubleStreamBlockProcessor
+from invokeai.backend.flux.extensions.xlabs_ip_adapter_extension import XLabsIPAdapterExtension
 from invokeai.backend.flux.modules.layers import (
     DoubleStreamBlock,
     EmbedND,
@@ -87,7 +89,12 @@ class Flux(nn.Module):
         txt_ids: Tensor,
         timesteps: Tensor,
         y: Tensor,
-        guidance: Tensor | None = None,
+        guidance: Tensor | None,
+        timestep_index: int,
+        total_num_timesteps: int,
+        controlnet_double_block_residuals: list[Tensor] | None,
+        controlnet_single_block_residuals: list[Tensor] | None,
+        ip_adapter_extensions: list[XLabsIPAdapterExtension],
     ) -> Tensor:
         if img.ndim != 3 or txt.ndim != 3:
             raise ValueError("Input img and txt tensors must have 3 dimensions.")
@@ -105,12 +112,39 @@ class Flux(nn.Module):
         ids = torch.cat((txt_ids, img_ids), dim=1)
         pe = self.pe_embedder(ids)
 
-        for block in self.double_blocks:
-            img, txt = block(img=img, txt=txt, vec=vec, pe=pe)
+        # Validate double_block_residuals shape.
+        if controlnet_double_block_residuals is not None:
+            assert len(controlnet_double_block_residuals) == len(self.double_blocks)
+        for block_index, block in enumerate(self.double_blocks):
+            assert isinstance(block, DoubleStreamBlock)
+
+            img, txt = CustomDoubleStreamBlockProcessor.custom_double_block_forward(
+                timestep_index=timestep_index,
+                total_num_timesteps=total_num_timesteps,
+                block_index=block_index,
+                block=block,
+                img=img,
+                txt=txt,
+                vec=vec,
+                pe=pe,
+                ip_adapter_extensions=ip_adapter_extensions,
+            )
+
+            if controlnet_double_block_residuals is not None:
+                img += controlnet_double_block_residuals[block_index]
 
         img = torch.cat((txt, img), 1)
-        for block in self.single_blocks:
+
+        # Validate single_block_residuals shape.
+        if controlnet_single_block_residuals is not None:
+            assert len(controlnet_single_block_residuals) == len(self.single_blocks)
+
+        for block_index, block in enumerate(self.single_blocks):
             img = block(img, vec=vec, pe=pe)
+
+            if controlnet_single_block_residuals is not None:
+                img[:, txt.shape[1] :, ...] += controlnet_single_block_residuals[block_index]
+
         img = img[:, txt.shape[1] :, ...]
 
         img = self.final_layer(img, vec)  # (N, T, patch_size ** 2 * out_channels)

invokeai/backend/flux/sampling_utils.py

@@ -97,6 +97,46 @@ def clip_timestep_schedule(timesteps: list[float], denoising_start: float, denoi
     return clipped_timesteps
 
 
+def clip_timestep_schedule_fractional(
+    timesteps: list[float], denoising_start: float, denoising_end: float
+) -> list[float]:
+    """Clip the timestep schedule to the denoising range. Insert new timesteps to exactly match the desired denoising
+    range. (A fractional version of clip_timestep_schedule().)
+
+    Args:
+        timesteps (list[float]): The original timestep schedule: [1.0, ..., 0.0].
+        denoising_start (float): A value in [0, 1] specifying the start of the denoising process. E.g. a value of 0.2
+            would mean that the denoising process start at t=0.8.
+        denoising_end (float): A value in [0, 1] specifying the end of the denoising process. E.g. a value of 0.8 would
+            mean that the denoising process ends at t=0.2.
+
+    Returns:
+        list[float]: The clipped timestep schedule.
+    """
+    assert 0.0 <= denoising_start <= 1.0
+    assert 0.0 <= denoising_end <= 1.0
+    assert denoising_start <= denoising_end
+
+    t_start_val = 1.0 - denoising_start
+    t_end_val = 1.0 - denoising_end
+
+    t_start_idx = _find_last_index_ge_val(timesteps, t_start_val)
+    t_end_idx = _find_last_index_ge_val(timesteps, t_end_val)
+
+    clipped_timesteps = timesteps[t_start_idx : t_end_idx + 1]
+
+    # We know that clipped_timesteps[0] >= t_start_val. Replace clipped_timesteps[0] with t_start_val.
+    clipped_timesteps[0] = t_start_val
+
+    # We know that clipped_timesteps[-1] >= t_end_val. If clipped_timesteps[-1] > t_end_val, add another step to
+    # t_end_val.
+    eps = 1e-6
+    if clipped_timesteps[-1] > t_end_val + eps:
+        clipped_timesteps.append(t_end_val)
+
+    return clipped_timesteps
+
+
 def unpack(x: torch.Tensor, height: int, width: int) -> torch.Tensor:
     """Unpack flat array of patch embeddings to latent image."""
     return rearrange(
@@ -128,8 +168,17 @@ def generate_img_ids(h: int, w: int, batch_size: int, device: torch.device, dtyp
     Returns:
         torch.Tensor: Image position ids.
     """
+
+    if device.type == "mps":
+        orig_dtype = dtype
+        dtype = torch.float16
+
     img_ids = torch.zeros(h // 2, w // 2, 3, device=device, dtype=dtype)
     img_ids[..., 1] = img_ids[..., 1] + torch.arange(h // 2, device=device, dtype=dtype)[:, None]
     img_ids[..., 2] = img_ids[..., 2] + torch.arange(w // 2, device=device, dtype=dtype)[None, :]
     img_ids = repeat(img_ids, "h w c -> b (h w) c", b=batch_size)
+
+    if device.type == "mps":
+        img_ids.to(orig_dtype)
+
     return img_ids

invokeai/backend/flux/trajectory_guidance_extension.py

@@ -1,134 +0,0 @@
-import torch
-
-from invokeai.backend.util.build_line import build_line
-
-
-class TrajectoryGuidanceExtension:
-    """An implementation of trajectory guidance for FLUX.
-
-    What is trajectory guidance?
-    ----------------------------
-    With SD 1 and SDXL, the amount of change in image-to-image denoising is largely controlled by the denoising_start
-    parameter. Doing the same thing with the FLUX model does not work as well, because the FLUX model converges very
-    quickly (roughly time 1.0 to 0.9) to the structure of the final image. The result of this model characteristic is
-    that you typically get one of two outcomes:
-    1) a result that is very similar to the original image
-    2) a result that is very different from the original image, as though it was generated from the text prompt with
-       pure noise.
-
-    To address this issue with image-to-image workflows with FLUX, we employ the concept of trajectory guidance. The
-    idea is that in addition to controlling the denoising_start parameter (i.e. the amount of noise added to the
-    original image), we can also guide the denoising process to stay close to the trajectory that would reproduce the
-    original. By controlling the strength of the trajectory guidance throughout the denoising process, we can achieve
-    FLUX image-to-image behavior with the same level of control offered by SD1 and SDXL.
-
-    What is the trajectory_guidance_strength?
-    -----------------------------------------
-    In the limit, we could apply a different trajectory guidance 'strength' for every latent value in every timestep.
-    This would be impractical for a user, so instead we have engineered a strength schedule that is more convenient to
-    use. The `trajectory_guidance_strength` parameter is a single scalar value that maps to a schedule. The engineered
-    schedule is defined as:
-    1) An initial change_ratio at t=1.0.
-    2) A linear ramp up to change_ratio=1.0 at t = t_cutoff.
-    3) A constant change_ratio=1.0 after t = t_cutoff.
-    """
-
-    def __init__(
-        self, init_latents: torch.Tensor, inpaint_mask: torch.Tensor | None, trajectory_guidance_strength: float
-    ):
-        """Initialize TrajectoryGuidanceExtension.
-
-        Args:
-            init_latents (torch.Tensor): The initial latents (i.e. un-noised at timestep 0). In 'packed' format.
-            inpaint_mask (torch.Tensor | None): A mask specifying which elements to inpaint. Range [0, 1]. Values of 1
-                will be re-generated. Values of 0 will remain unchanged. Values between 0 and 1 can be used to blend the
-                inpainted region with the background. In 'packed' format. If None, will be treated as a mask of all 1s.
-            trajectory_guidance_strength (float): A value in [0, 1] specifying the strength of the trajectory guidance.
-                A value of 0.0 is equivalent to vanilla image-to-image. A value of 1.0 will guide the denoising process
-                very close to the original latents.
-        """
-        assert 0.0 <= trajectory_guidance_strength <= 1.0
-        self._init_latents = init_latents
-        if inpaint_mask is None:
-            # The inpaing mask is None, so we initialize a mask with a single value of 1.0.
-            # This value will be broadcasted and treated as a mask of all 1s.
-            self._inpaint_mask = torch.ones(1, device=init_latents.device, dtype=init_latents.dtype)
-        else:
-            self._inpaint_mask = inpaint_mask
-
-        # Calculate the params that define the trajectory guidance schedule.
-        # These mappings from trajectory_guidance_strength have no theoretical basis - they were tuned manually.
-        self._trajectory_guidance_strength = trajectory_guidance_strength
-        self._change_ratio_at_t_1 = build_line(x1=0.0, y1=1.0, x2=1.0, y2=0.0)(self._trajectory_guidance_strength)
-        self._change_ratio_at_cutoff = 1.0
-        self._t_cutoff = build_line(x1=0.0, y1=1.0, x2=1.0, y2=0.5)(self._trajectory_guidance_strength)
-
-    def _apply_mask_gradient_adjustment(self, t_prev: float) -> torch.Tensor:
-        """Applies inpaint mask gradient adjustment and returns the inpaint mask to be used at the current timestep."""
-        # As we progress through the denoising process, we promote gradient regions of the mask to have a full weight of
-        # 1.0. This helps to produce more coherent seams around the inpainted region. We experimented with a (small)
-        # number of promotion strategies (e.g. gradual promotion based on timestep), but found that a simple cutoff
-        # threshold worked well.
-        # We use a small epsilon to avoid any potential issues with floating point precision.
-        eps = 1e-4
-        mask_gradient_t_cutoff = 0.5
-        if t_prev > mask_gradient_t_cutoff:
-            # Early in the denoising process, use the inpaint mask as-is.
-            return self._inpaint_mask
-        else:
-            # After the cut-off, promote all non-zero mask values to 1.0.
-            mask = self._inpaint_mask.where(self._inpaint_mask <= (0.0 + eps), 1.0)
-
-        return mask
-
-    def _get_change_ratio(self, t_prev: float) -> float:
-        """Get the change_ratio for t_prev based on the change schedule."""
-        change_ratio = 1.0
-        if t_prev > self._t_cutoff:
-            # If we are before the cutoff, linearly interpolate between the change_ratio at t=1.0 and the change_ratio
-            # at the cutoff.
-            change_ratio = build_line(
-                x1=1.0, y1=self._change_ratio_at_t_1, x2=self._t_cutoff, y2=self._change_ratio_at_cutoff
-            )(t_prev)
-
-        # The change_ratio should be in the range [0, 1]. Assert that we didn't make any mistakes.
-        eps = 1e-5
-        assert 0.0 - eps <= change_ratio <= 1.0 + eps
-        return change_ratio
-
-    def update_noise(
-        self, t_curr_latents: torch.Tensor, pred_noise: torch.Tensor, t_curr: float, t_prev: float
-    ) -> torch.Tensor:
-        # Handle gradient cutoff.
-        mask = self._apply_mask_gradient_adjustment(t_prev)
-
-        mask = mask * self._get_change_ratio(t_prev)
-
-        # NOTE(ryand): During inpainting, it is common to guide the denoising process by noising the initial latents for
-        # the current timestep and then blending the predicted intermediate latents with the noised initial latents.
-        # For example:
-        # ```
-        # noised_init_latents = self._noise * t_prev + (1.0 - t_prev) * self._init_latents
-        # return t_prev_latents * self._inpaint_mask + noised_init_latents * (1.0 - self._inpaint_mask)
-        # ```
-        # Instead of guiding based on the noised initial latents, we have decided to guide based on the noise prediction
-        # that points towards the initial latents. The difference between these guidance strategies is minor, but
-        # qualitatively we found the latter to produce slightly better results. When change_ratio is 0.0 or 1.0 there is
-        # no difference between the two strategies.
-        #
-        # We experimented with a number of related guidance strategies, but not exhaustively. It's entirely possible
-        # that there's a much better way to do this.
-
-        # Calculate noise guidance
-        # What noise should the model have predicted at this timestep to step towards self._init_latents?
-        # Derivation:
-        # > t_prev_latents = t_curr_latents + (t_prev - t_curr) * pred_noise
-        # > t_0_latents = t_curr_latents + (0 - t_curr) * init_traj_noise
-        # > t_0_latents = t_curr_latents - t_curr * init_traj_noise
-        # > init_traj_noise = (t_curr_latents - t_0_latents) / t_curr)
-        init_traj_noise = (t_curr_latents - self._init_latents) / t_curr
-
-        # Blend the init_traj_noise with the pred_noise according to the inpaint mask and the trajectory guidance.
-        noise = pred_noise * mask + init_traj_noise * (1.0 - mask)
-
-        return noise

invokeai/backend/image_util/segment_anything/segment_anything_pipeline.py

@@ -1,4 +1,4 @@
-from typing import Optional
+from typing import Optional, TypeAlias
 
 import torch
 from PIL import Image
@@ -7,6 +7,14 @@ from transformers.models.sam.processing_sam import SamProcessor
 
 from invokeai.backend.raw_model import RawModel
 
+# Type aliases for the inputs to the SAM model.
+ListOfBoundingBoxes: TypeAlias = list[list[int]]
+"""A list of bounding boxes. Each bounding box is in the format [xmin, ymin, xmax, ymax]."""
+ListOfPoints: TypeAlias = list[list[int]]
+"""A list of points. Each point is in the format [x, y]."""
+ListOfPointLabels: TypeAlias = list[int]
+"""A list of SAM point labels. Each label is an integer where -1 is background, 0 is neutral, and 1 is foreground."""
+
 
 class SegmentAnythingPipeline(RawModel):
     """A wrapper class for the transformers SAM model and processor that makes it compatible with the model manager."""
@@ -27,20 +35,53 @@ class SegmentAnythingPipeline(RawModel):
 
         return calc_module_size(self._sam_model)
 
-    def segment(self, image: Image.Image, bounding_boxes: list[list[int]]) -> torch.Tensor:
+    def segment(
+        self,
+        image: Image.Image,
+        bounding_boxes: list[list[int]] | None = None,
+        point_lists: list[list[list[int]]] | None = None,
+    ) -> torch.Tensor:
         """Run the SAM model.
 
+        Either bounding_boxes or point_lists must be provided. If both are provided, bounding_boxes will be used and
+        point_lists will be ignored.
+
         Args:
             image (Image.Image): The image to segment.
             bounding_boxes (list[list[int]]): The bounding box prompts. Each bounding box is in the format
                 [xmin, ymin, xmax, ymax].
+            point_lists (list[list[list[int]]]): The points prompts. Each point is in the format [x, y, label].
+                `label` is an integer where -1 is background, 0 is neutral, and 1 is foreground.
 
         Returns:
             torch.Tensor: The segmentation masks. dtype: torch.bool. shape: [num_masks, channels, height, width].
         """
-        # Add batch dimension of 1 to the bounding boxes.
-        boxes = [bounding_boxes]
-        inputs = self._sam_processor(images=image, input_boxes=boxes, return_tensors="pt").to(self._sam_model.device)
+
+        # Prep the inputs:
+        # - Create a list of bounding boxes or points and labels.
+        # - Add a batch dimension of 1 to the inputs.
+        if bounding_boxes:
+            input_boxes: list[ListOfBoundingBoxes] | None = [bounding_boxes]
+            input_points: list[ListOfPoints] | None = None
+            input_labels: list[ListOfPointLabels] | None = None
+        elif point_lists:
+            input_boxes: list[ListOfBoundingBoxes] | None = None
+            input_points: list[ListOfPoints] | None = []
+            input_labels: list[ListOfPointLabels] | None = []
+            for point_list in point_lists:
+                input_points.append([[p[0], p[1]] for p in point_list])
+                input_labels.append([p[2] for p in point_list])
+
+        else:
+            raise ValueError("Either bounding_boxes or points and labels must be provided.")
+
+        inputs = self._sam_processor(
+            images=image,
+            input_boxes=input_boxes,
+            input_points=input_points,
+            input_labels=input_labels,
+            return_tensors="pt",
+        ).to(self._sam_model.device)
         outputs = self._sam_model(**inputs)
         masks = self._sam_processor.post_process_masks(
             masks=outputs.pred_masks,

invokeai/backend/lora/conversions/flux_diffusers_lora_conversion_utils.py

@@ -2,6 +2,7 @@ from typing import Dict
 
 import torch
 
+from invokeai.backend.lora.conversions.flux_lora_constants import FLUX_LORA_TRANSFORMER_PREFIX
 from invokeai.backend.lora.layers.any_lora_layer import AnyLoRALayer
 from invokeai.backend.lora.layers.concatenated_lora_layer import ConcatenatedLoRALayer
 from invokeai.backend.lora.layers.lora_layer import LoRALayer
@@ -44,8 +45,9 @@ def lora_model_from_flux_diffusers_state_dict(state_dict: Dict[str, torch.Tensor
     # Constants for FLUX.1
     num_double_layers = 19
     num_single_layers = 38
-    # inner_dim = 3072
-    # mlp_ratio = 4.0
+    hidden_size = 3072
+    mlp_ratio = 4.0
+    mlp_hidden_dim = int(hidden_size * mlp_ratio)
 
     layers: dict[str, AnyLoRALayer] = {}
 
@@ -61,30 +63,43 @@ def lora_model_from_flux_diffusers_state_dict(state_dict: Dict[str, torch.Tensor
             layers[dst_key] = LoRALayer.from_state_dict_values(values=value)
             assert len(src_layer_dict) == 0
 
-    def add_qkv_lora_layer_if_present(src_keys: list[str], dst_qkv_key: str) -> None:
+    def add_qkv_lora_layer_if_present(
+        src_keys: list[str],
+        src_weight_shapes: list[tuple[int, int]],
+        dst_qkv_key: str,
+        allow_missing_keys: bool = False,
+    ) -> None:
         """Handle the Q, K, V matrices for a transformer block. We need special handling because the diffusers format
         stores them in separate matrices, whereas the BFL format used internally by InvokeAI concatenates them.
         """
-        # We expect that either all src keys are present or none of them are. Verify this.
-        keys_present = [key in grouped_state_dict for key in src_keys]
-        assert all(keys_present) or not any(keys_present)
-
         # If none of the keys are present, return early.
+        keys_present = [key in grouped_state_dict for key in src_keys]
         if not any(keys_present):
             return
 
-        src_layer_dicts = [grouped_state_dict.pop(key) for key in src_keys]
         sub_layers: list[LoRALayer] = []
-        for src_layer_dict in src_layer_dicts:
-            values = {
-                "lora_down.weight": src_layer_dict.pop("lora_A.weight"),
-                "lora_up.weight": src_layer_dict.pop("lora_B.weight"),
-            }
-            if alpha is not None:
-                values["alpha"] = torch.tensor(alpha)
-            sub_layers.append(LoRALayer.from_state_dict_values(values=values))
-            assert len(src_layer_dict) == 0
-        layers[dst_qkv_key] = ConcatenatedLoRALayer(lora_layers=sub_layers, concat_axis=0)
+        for src_key, src_weight_shape in zip(src_keys, src_weight_shapes, strict=True):
+            src_layer_dict = grouped_state_dict.pop(src_key, None)
+            if src_layer_dict is not None:
+                values = {
+                    "lora_down.weight": src_layer_dict.pop("lora_A.weight"),
+                    "lora_up.weight": src_layer_dict.pop("lora_B.weight"),
+                }
+                if alpha is not None:
+                    values["alpha"] = torch.tensor(alpha)
+                assert values["lora_down.weight"].shape[1] == src_weight_shape[1]
+                assert values["lora_up.weight"].shape[0] == src_weight_shape[0]
+                sub_layers.append(LoRALayer.from_state_dict_values(values=values))
+                assert len(src_layer_dict) == 0
+            else:
+                if not allow_missing_keys:
+                    raise ValueError(f"Missing LoRA layer: '{src_key}'.")
+                values = {
+                    "lora_up.weight": torch.zeros((src_weight_shape[0], 1)),
+                    "lora_down.weight": torch.zeros((1, src_weight_shape[1])),
+                }
+                sub_layers.append(LoRALayer.from_state_dict_values(values=values))
+        layers[dst_qkv_key] = ConcatenatedLoRALayer(lora_layers=sub_layers)
 
     # time_text_embed.timestep_embedder -> time_in.
     add_lora_layer_if_present("time_text_embed.timestep_embedder.linear_1", "time_in.in_layer")
@@ -117,6 +132,7 @@ def lora_model_from_flux_diffusers_state_dict(state_dict: Dict[str, torch.Tensor
                 f"transformer_blocks.{i}.attn.to_k",
                 f"transformer_blocks.{i}.attn.to_v",
             ],
+            [(hidden_size, hidden_size), (hidden_size, hidden_size), (hidden_size, hidden_size)],
             f"double_blocks.{i}.img_attn.qkv",
         )
         add_qkv_lora_layer_if_present(
@@ -125,6 +141,7 @@ def lora_model_from_flux_diffusers_state_dict(state_dict: Dict[str, torch.Tensor
                 f"transformer_blocks.{i}.attn.add_k_proj",
                 f"transformer_blocks.{i}.attn.add_v_proj",
             ],
+            [(hidden_size, hidden_size), (hidden_size, hidden_size), (hidden_size, hidden_size)],
             f"double_blocks.{i}.txt_attn.qkv",
         )
 
@@ -174,7 +191,14 @@ def lora_model_from_flux_diffusers_state_dict(state_dict: Dict[str, torch.Tensor
                 f"single_transformer_blocks.{i}.attn.to_v",
                 f"single_transformer_blocks.{i}.proj_mlp",
             ],
+            [
+                (hidden_size, hidden_size),
+                (hidden_size, hidden_size),
+                (hidden_size, hidden_size),
+                (mlp_hidden_dim, hidden_size),
+            ],
             f"single_blocks.{i}.linear1",
+            allow_missing_keys=True,
         )
 
         # Output projections.
@@ -189,7 +213,9 @@ def lora_model_from_flux_diffusers_state_dict(state_dict: Dict[str, torch.Tensor
     # Assert that all keys were processed.
     assert len(grouped_state_dict) == 0
 
-    return LoRAModelRaw(layers=layers)
+    layers_with_prefix = {f"{FLUX_LORA_TRANSFORMER_PREFIX}{k}": v for k, v in layers.items()}
+
+    return LoRAModelRaw(layers=layers_with_prefix)
 
 
 def _group_by_layer(state_dict: Dict[str, torch.Tensor]) -> dict[str, dict[str, torch.Tensor]]:

invokeai/backend/lora/conversions/flux_kohya_lora_conversion_utils.py

@@ -3,18 +3,25 @@ from typing import Any, Dict, TypeVar
 
 import torch
 
+from invokeai.backend.lora.conversions.flux_lora_constants import FLUX_LORA_CLIP_PREFIX, FLUX_LORA_TRANSFORMER_PREFIX
 from invokeai.backend.lora.layers.any_lora_layer import AnyLoRALayer
 from invokeai.backend.lora.layers.utils import any_lora_layer_from_state_dict
 from invokeai.backend.lora.lora_model_raw import LoRAModelRaw
 
-# A regex pattern that matches all of the keys in the Kohya FLUX LoRA format.
+# A regex pattern that matches all of the transformer keys in the Kohya FLUX LoRA format.
 # Example keys:
 #   lora_unet_double_blocks_0_img_attn_proj.alpha
 #   lora_unet_double_blocks_0_img_attn_proj.lora_down.weight
 #   lora_unet_double_blocks_0_img_attn_proj.lora_up.weight
-FLUX_KOHYA_KEY_REGEX = (
+FLUX_KOHYA_TRANSFORMER_KEY_REGEX = (
     r"lora_unet_(\w+_blocks)_(\d+)_(img_attn|img_mlp|img_mod|txt_attn|txt_mlp|txt_mod|linear1|linear2|modulation)_?(.*)"
 )
+# A regex pattern that matches all of the CLIP keys in the Kohya FLUX LoRA format.
+# Example keys:
+#   lora_te1_text_model_encoder_layers_0_mlp_fc1.alpha
+#   lora_te1_text_model_encoder_layers_0_mlp_fc1.lora_down.weight
+#   lora_te1_text_model_encoder_layers_0_mlp_fc1.lora_up.weight
+FLUX_KOHYA_CLIP_KEY_REGEX = r"lora_te1_text_model_encoder_layers_(\d+)_(mlp|self_attn)_(\w+)\.?.*"
 
 
 def is_state_dict_likely_in_flux_kohya_format(state_dict: Dict[str, Any]) -> bool:
@@ -23,7 +30,10 @@ def is_state_dict_likely_in_flux_kohya_format(state_dict: Dict[str, Any]) -> boo
     This is intended to be a high-precision detector, but it is not guaranteed to have perfect precision. (A
     perfect-precision detector would require checking all keys against a whitelist and verifying tensor shapes.)
     """
-    return all(re.match(FLUX_KOHYA_KEY_REGEX, k) for k in state_dict.keys())
+    return all(
+        re.match(FLUX_KOHYA_TRANSFORMER_KEY_REGEX, k) or re.match(FLUX_KOHYA_CLIP_KEY_REGEX, k)
+        for k in state_dict.keys()
+    )
 
 
 def lora_model_from_flux_kohya_state_dict(state_dict: Dict[str, torch.Tensor]) -> LoRAModelRaw:
@@ -35,13 +45,27 @@ def lora_model_from_flux_kohya_state_dict(state_dict: Dict[str, torch.Tensor]) -
             grouped_state_dict[layer_name] = {}
         grouped_state_dict[layer_name][param_name] = value
 
-    # Convert the state dict to the InvokeAI format.
-    grouped_state_dict = convert_flux_kohya_state_dict_to_invoke_format(grouped_state_dict)
+    # Split the grouped state dict into transformer and CLIP state dicts.
+    transformer_grouped_sd: dict[str, dict[str, torch.Tensor]] = {}
+    clip_grouped_sd: dict[str, dict[str, torch.Tensor]] = {}
+    for layer_name, layer_state_dict in grouped_state_dict.items():
+        if layer_name.startswith("lora_unet"):
+            transformer_grouped_sd[layer_name] = layer_state_dict
+        elif layer_name.startswith("lora_te1"):
+            clip_grouped_sd[layer_name] = layer_state_dict
+        else:
+            raise ValueError(f"Layer '{layer_name}' does not match the expected pattern for FLUX LoRA weights.")
+
+    # Convert the state dicts to the InvokeAI format.
+    transformer_grouped_sd = _convert_flux_transformer_kohya_state_dict_to_invoke_format(transformer_grouped_sd)
+    clip_grouped_sd = _convert_flux_clip_kohya_state_dict_to_invoke_format(clip_grouped_sd)
 
     # Create LoRA layers.
     layers: dict[str, AnyLoRALayer] = {}
-    for layer_key, layer_state_dict in grouped_state_dict.items():
-        layers[layer_key] = any_lora_layer_from_state_dict(layer_state_dict)
+    for layer_key, layer_state_dict in transformer_grouped_sd.items():
+        layers[FLUX_LORA_TRANSFORMER_PREFIX + layer_key] = any_lora_layer_from_state_dict(layer_state_dict)
+    for layer_key, layer_state_dict in clip_grouped_sd.items():
+        layers[FLUX_LORA_CLIP_PREFIX + layer_key] = any_lora_layer_from_state_dict(layer_state_dict)
 
     # Create and return the LoRAModelRaw.
     return LoRAModelRaw(layers=layers)
@@ -50,16 +74,34 @@ def lora_model_from_flux_kohya_state_dict(state_dict: Dict[str, torch.Tensor]) -
 T = TypeVar("T")
 
 
-def convert_flux_kohya_state_dict_to_invoke_format(state_dict: Dict[str, T]) -> Dict[str, T]:
-    """Converts a state dict from the Kohya FLUX LoRA format to LoRA weight format used internally by InvokeAI.
+def _convert_flux_clip_kohya_state_dict_to_invoke_format(state_dict: Dict[str, T]) -> Dict[str, T]:
+    """Converts a CLIP LoRA state dict from the Kohya FLUX LoRA format to LoRA weight format used internally by
+    InvokeAI.
+
+    Example key conversions:
+
+    "lora_te1_text_model_encoder_layers_0_mlp_fc1" -> "text_model.encoder.layers.0.mlp.fc1",
+    "lora_te1_text_model_encoder_layers_0_self_attn_k_proj" -> "text_model.encoder.layers.0.self_attn.k_proj"
+    """
+    converted_sd: dict[str, T] = {}
+    for k, v in state_dict.items():
+        match = re.match(FLUX_KOHYA_CLIP_KEY_REGEX, k)
+        if match:
+            new_key = f"text_model.encoder.layers.{match.group(1)}.{match.group(2)}.{match.group(3)}"
+            converted_sd[new_key] = v
+        else:
+            raise ValueError(f"Key '{k}' does not match the expected pattern for FLUX LoRA weights.")
+
+    return converted_sd
+
+
+def _convert_flux_transformer_kohya_state_dict_to_invoke_format(state_dict: Dict[str, T]) -> Dict[str, T]:
+    """Converts a FLUX tranformer LoRA state dict from the Kohya FLUX LoRA format to LoRA weight format used internally
+    by InvokeAI.
 
     Example key conversions:
     "lora_unet_double_blocks_0_img_attn_proj" -> "double_blocks.0.img_attn.proj"
-    "lora_unet_double_blocks_0_img_attn_proj" -> "double_blocks.0.img_attn.proj"
-    "lora_unet_double_blocks_0_img_attn_proj" -> "double_blocks.0.img_attn.proj"
     "lora_unet_double_blocks_0_img_attn_qkv" -> "double_blocks.0.img_attn.qkv"
-    "lora_unet_double_blocks_0_img_attn_qkv" -> "double_blocks.0.img.attn.qkv"
-    "lora_unet_double_blocks_0_img_attn_qkv" -> "double_blocks.0.img.attn.qkv"
     """
 
     def replace_func(match: re.Match[str]) -> str:
@@ -70,9 +112,9 @@ def convert_flux_kohya_state_dict_to_invoke_format(state_dict: Dict[str, T]) ->
 
     converted_dict: dict[str, T] = {}
     for k, v in state_dict.items():
-        match = re.match(FLUX_KOHYA_KEY_REGEX, k)
+        match = re.match(FLUX_KOHYA_TRANSFORMER_KEY_REGEX, k)
         if match:
-            new_key = re.sub(FLUX_KOHYA_KEY_REGEX, replace_func, k)
+            new_key = re.sub(FLUX_KOHYA_TRANSFORMER_KEY_REGEX, replace_func, k)
             converted_dict[new_key] = v
         else:
             raise ValueError(f"Key '{k}' does not match the expected pattern for FLUX LoRA weights.")

invokeai/backend/lora/conversions/flux_lora_constants.py

@@ -0,0 +1,3 @@
+# Prefixes used to distinguish between transformer and CLIP text encoder keys in the FLUX InvokeAI LoRA format.
+FLUX_LORA_TRANSFORMER_PREFIX = "lora_transformer-"
+FLUX_LORA_CLIP_PREFIX = "lora_clip-"

invokeai/backend/model_manager/config.py

@@ -53,6 +53,7 @@ class BaseModelType(str, Enum):
     Any = "any"
     StableDiffusion1 = "sd-1"
     StableDiffusion2 = "sd-2"
+    StableDiffusion3 = "sd-3"
     StableDiffusionXL = "sdxl"
     StableDiffusionXLRefiner = "sdxl-refiner"
     Flux = "flux"
@@ -83,8 +84,10 @@ class SubModelType(str, Enum):
     Transformer = "transformer"
     TextEncoder = "text_encoder"
     TextEncoder2 = "text_encoder_2"
+    TextEncoder3 = "text_encoder_3"
     Tokenizer = "tokenizer"
     Tokenizer2 = "tokenizer_2"
+    Tokenizer3 = "tokenizer_3"
     VAE = "vae"
     VAEDecoder = "vae_decoder"
     VAEEncoder = "vae_encoder"
@@ -92,6 +95,13 @@ class SubModelType(str, Enum):
     SafetyChecker = "safety_checker"
 
 
+class ClipVariantType(str, Enum):
+    """Variant type."""
+
+    L = "large"
+    G = "gigantic"
+
+
 class ModelVariantType(str, Enum):
     """Variant type."""
 
@@ -114,6 +124,7 @@ class ModelFormat(str, Enum):
     T5Encoder = "t5_encoder"
     BnbQuantizedLlmInt8b = "bnb_quantized_int8b"
     BnbQuantizednf4b = "bnb_quantized_nf4b"
+    GGUFQuantized = "gguf_quantized"
 
 
 class SchedulerPredictionType(str, Enum):
@@ -146,6 +157,17 @@ class ModelSourceType(str, Enum):
 DEFAULTS_PRECISION = Literal["fp16", "fp32"]
 
 
+AnyVariant: TypeAlias = Union[ModelVariantType, ClipVariantType, None]
+
+
+class SubmodelDefinition(BaseModel):
+    path_or_prefix: str
+    model_type: ModelType
+    variant: AnyVariant = None
+
+    model_config = ConfigDict(protected_namespaces=())
+
+
 class MainModelDefaultSettings(BaseModel):
     vae: str | None = Field(default=None, description="Default VAE for this model (model key)")
     vae_precision: DEFAULTS_PRECISION | None = Field(default=None, description="Default VAE precision for this model")
@@ -157,6 +179,7 @@ class MainModelDefaultSettings(BaseModel):
     )
     width: int | None = Field(default=None, multiple_of=8, ge=64, description="Default width for this model")
     height: int | None = Field(default=None, multiple_of=8, ge=64, description="Default height for this model")
+    guidance: float | None = Field(default=None, ge=1, description="Default Guidance for this model")
 
     model_config = ConfigDict(extra="forbid")
 
@@ -191,12 +214,15 @@ class ModelConfigBase(BaseModel):
         schema["required"].extend(["key", "type", "format"])
 
     model_config = ConfigDict(validate_assignment=True, json_schema_extra=json_schema_extra)
+    submodels: Optional[Dict[SubModelType, SubmodelDefinition]] = Field(
+        description="Loadable submodels in this model", default=None
+    )
 
 
 class CheckpointConfigBase(ModelConfigBase):
     """Model config for checkpoint-style models."""
 
-    format: Literal[ModelFormat.Checkpoint, ModelFormat.BnbQuantizednf4b] = Field(
+    format: Literal[ModelFormat.Checkpoint, ModelFormat.BnbQuantizednf4b, ModelFormat.GGUFQuantized] = Field(
         description="Format of the provided checkpoint model", default=ModelFormat.Checkpoint
     )
     config_path: str = Field(description="path to the checkpoint model config file")
@@ -333,7 +359,7 @@ class MainConfigBase(ModelConfigBase):
     default_settings: Optional[MainModelDefaultSettings] = Field(
         description="Default settings for this model", default=None
     )
-    variant: ModelVariantType = ModelVariantType.Normal
+    variant: AnyVariant = ModelVariantType.Normal
 
 
 class MainCheckpointConfig(CheckpointConfigBase, MainConfigBase):
@@ -362,6 +388,21 @@ class MainBnbQuantized4bCheckpointConfig(CheckpointConfigBase, MainConfigBase):
         return Tag(f"{ModelType.Main.value}.{ModelFormat.BnbQuantizednf4b.value}")
 
 
+class MainGGUFCheckpointConfig(CheckpointConfigBase, MainConfigBase):
+    """Model config for main checkpoint models."""
+
+    prediction_type: SchedulerPredictionType = SchedulerPredictionType.Epsilon
+    upcast_attention: bool = False
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.format = ModelFormat.GGUFQuantized
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.Main.value}.{ModelFormat.GGUFQuantized.value}")
+
+
 class MainDiffusersConfig(DiffusersConfigBase, MainConfigBase):
     """Model config for main diffusers models."""
 
@@ -377,6 +418,8 @@ class IPAdapterBaseConfig(ModelConfigBase):
 class IPAdapterInvokeAIConfig(IPAdapterBaseConfig):
     """Model config for IP Adapter diffusers format models."""
 
+    # TODO(ryand): Should we deprecate this field? From what I can tell, it hasn't been probed correctly for a long
+    # time. Need to go through the history to make sure I'm understanding this fully.
     image_encoder_model_id: str
     format: Literal[ModelFormat.InvokeAI]
 
@@ -400,12 +443,33 @@ class CLIPEmbedDiffusersConfig(DiffusersConfigBase):
 
     type: Literal[ModelType.CLIPEmbed] = ModelType.CLIPEmbed
     format: Literal[ModelFormat.Diffusers] = ModelFormat.Diffusers
+    variant: ClipVariantType = ClipVariantType.L
 
     @staticmethod
     def get_tag() -> Tag:
         return Tag(f"{ModelType.CLIPEmbed.value}.{ModelFormat.Diffusers.value}")
 
 
+class CLIPGEmbedDiffusersConfig(CLIPEmbedDiffusersConfig):
+    """Model config for CLIP-G Embeddings."""
+
+    variant: ClipVariantType = ClipVariantType.G
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.CLIPEmbed.value}.{ModelFormat.Diffusers.value}.{ClipVariantType.G}")
+
+
+class CLIPLEmbedDiffusersConfig(CLIPEmbedDiffusersConfig):
+    """Model config for CLIP-L Embeddings."""
+
+    variant: ClipVariantType = ClipVariantType.L
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.CLIPEmbed.value}.{ModelFormat.Diffusers.value}.{ClipVariantType.L}")
+
+
 class CLIPVisionDiffusersConfig(DiffusersConfigBase):
     """Model config for CLIPVision."""
 
@@ -465,6 +529,7 @@ AnyModelConfig = Annotated[
         Annotated[MainDiffusersConfig, MainDiffusersConfig.get_tag()],
         Annotated[MainCheckpointConfig, MainCheckpointConfig.get_tag()],
         Annotated[MainBnbQuantized4bCheckpointConfig, MainBnbQuantized4bCheckpointConfig.get_tag()],
+        Annotated[MainGGUFCheckpointConfig, MainGGUFCheckpointConfig.get_tag()],
         Annotated[VAEDiffusersConfig, VAEDiffusersConfig.get_tag()],
         Annotated[VAECheckpointConfig, VAECheckpointConfig.get_tag()],
         Annotated[ControlNetDiffusersConfig, ControlNetDiffusersConfig.get_tag()],
@@ -481,6 +546,8 @@ AnyModelConfig = Annotated[
         Annotated[SpandrelImageToImageConfig, SpandrelImageToImageConfig.get_tag()],
         Annotated[CLIPVisionDiffusersConfig, CLIPVisionDiffusersConfig.get_tag()],
         Annotated[CLIPEmbedDiffusersConfig, CLIPEmbedDiffusersConfig.get_tag()],
+        Annotated[CLIPLEmbedDiffusersConfig, CLIPLEmbedDiffusersConfig.get_tag()],
+        Annotated[CLIPGEmbedDiffusersConfig, CLIPGEmbedDiffusersConfig.get_tag()],
     ],
     Discriminator(get_model_discriminator_value),
 ]

invokeai/backend/model_manager/load/load_default.py

@@ -35,6 +35,7 @@ class ModelLoader(ModelLoaderBase):
         self._logger = logger
         self._ram_cache = ram_cache
         self._torch_dtype = TorchDevice.choose_torch_dtype()
+        self._torch_device = TorchDevice.choose_torch_device()
 
     def load_model(self, model_config: AnyModelConfig, submodel_type: Optional[SubModelType] = None) -> LoadedModel:
         """

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

@@ -0,0 +1,41 @@
+from pathlib import Path
+from typing import Optional
+
+from transformers import CLIPVisionModelWithProjection
+
+from invokeai.backend.model_manager.config import (
+    AnyModel,
+    AnyModelConfig,
+    BaseModelType,
+    DiffusersConfigBase,
+    ModelFormat,
+    ModelType,
+    SubModelType,
+)
+from invokeai.backend.model_manager.load.load_default import ModelLoader
+from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
+
+
+@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.CLIPVision, format=ModelFormat.Diffusers)
+class ClipVisionLoader(ModelLoader):
+    """Class to load CLIPVision models."""
+
+    def _load_model(
+        self,
+        config: AnyModelConfig,
+        submodel_type: Optional[SubModelType] = None,
+    ) -> AnyModel:
+        if not isinstance(config, DiffusersConfigBase):
+            raise ValueError("Only DiffusersConfigBase models are currently supported here.")
+
+        if submodel_type is not None:
+            raise Exception("There are no submodels in CLIP Vision models.")
+
+        model_path = Path(config.path)
+
+        model = CLIPVisionModelWithProjection.from_pretrained(
+            model_path, torch_dtype=self._torch_dtype, local_files_only=True
+        )
+        assert isinstance(model, CLIPVisionModelWithProjection)
+
+        return model

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

@@ -8,17 +8,36 @@ from diffusers import ControlNetModel
 from invokeai.backend.model_manager import (
     AnyModel,
     AnyModelConfig,
+)
+from invokeai.backend.model_manager.config import (
     BaseModelType,
+    ControlNetCheckpointConfig,
     ModelFormat,
     ModelType,
+    SubModelType,
 )
-from invokeai.backend.model_manager.config import ControlNetCheckpointConfig, SubModelType
 from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
 from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
 
 
-@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.ControlNet, format=ModelFormat.Diffusers)
-@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.ControlNet, format=ModelFormat.Checkpoint)
+@ModelLoaderRegistry.register(
+    base=BaseModelType.StableDiffusion1, type=ModelType.ControlNet, format=ModelFormat.Diffusers
+)
+@ModelLoaderRegistry.register(
+    base=BaseModelType.StableDiffusion1, type=ModelType.ControlNet, format=ModelFormat.Checkpoint
+)
+@ModelLoaderRegistry.register(
+    base=BaseModelType.StableDiffusion2, type=ModelType.ControlNet, format=ModelFormat.Diffusers
+)
+@ModelLoaderRegistry.register(
+    base=BaseModelType.StableDiffusion2, type=ModelType.ControlNet, format=ModelFormat.Checkpoint
+)
+@ModelLoaderRegistry.register(
+    base=BaseModelType.StableDiffusionXL, type=ModelType.ControlNet, format=ModelFormat.Diffusers
+)
+@ModelLoaderRegistry.register(
+    base=BaseModelType.StableDiffusionXL, type=ModelType.ControlNet, format=ModelFormat.Checkpoint
+)
 class ControlNetLoader(GenericDiffusersLoader):
     """Class to load ControlNet models."""
 

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

@@ -10,6 +10,19 @@ from safetensors.torch import load_file
 from transformers import AutoConfig, AutoModelForTextEncoding, CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5Tokenizer
 
 from invokeai.app.services.config.config_default import get_config
+from invokeai.backend.flux.controlnet.instantx_controlnet_flux import InstantXControlNetFlux
+from invokeai.backend.flux.controlnet.state_dict_utils import (
+    convert_diffusers_instantx_state_dict_to_bfl_format,
+    infer_flux_params_from_state_dict,
+    infer_instantx_num_control_modes_from_state_dict,
+    is_state_dict_instantx_controlnet,
+    is_state_dict_xlabs_controlnet,
+)
+from invokeai.backend.flux.controlnet.xlabs_controlnet_flux import XLabsControlNetFlux
+from invokeai.backend.flux.ip_adapter.state_dict_utils import infer_xlabs_ip_adapter_params_from_state_dict
+from invokeai.backend.flux.ip_adapter.xlabs_ip_adapter_flux import (
+    XlabsIpAdapterFlux,
+)
 from invokeai.backend.flux.model import Flux
 from invokeai.backend.flux.modules.autoencoder import AutoEncoder
 from invokeai.backend.flux.util import ae_params, params
@@ -24,8 +37,12 @@ from invokeai.backend.model_manager import (
 from invokeai.backend.model_manager.config import (
     CheckpointConfigBase,
     CLIPEmbedDiffusersConfig,
+    ControlNetCheckpointConfig,
+    ControlNetDiffusersConfig,
+    IPAdapterCheckpointConfig,
     MainBnbQuantized4bCheckpointConfig,
     MainCheckpointConfig,
+    MainGGUFCheckpointConfig,
     T5EncoderBnbQuantizedLlmInt8bConfig,
     T5EncoderConfig,
     VAECheckpointConfig,
@@ -35,6 +52,8 @@ from invokeai.backend.model_manager.load.model_loader_registry import ModelLoade
 from invokeai.backend.model_manager.util.model_util import (
     convert_bundle_to_flux_transformer_checkpoint,
 )
+from invokeai.backend.quantization.gguf.loaders import gguf_sd_loader
+from invokeai.backend.quantization.gguf.utils import TORCH_COMPATIBLE_QTYPES
 from invokeai.backend.util.silence_warnings import SilenceWarnings
 
 try:
@@ -65,7 +84,15 @@ class FluxVAELoader(ModelLoader):
             model = AutoEncoder(ae_params[config.config_path])
             sd = load_file(model_path)
             model.load_state_dict(sd, assign=True)
-            model.to(dtype=self._torch_dtype)
+            # VAE is broken in float16, which mps defaults to
+            if self._torch_dtype == torch.float16:
+                try:
+                    vae_dtype = torch.tensor([1.0], dtype=torch.bfloat16, device=self._torch_device).dtype
+                except TypeError:
+                    vae_dtype = torch.float32
+            else:
+                vae_dtype = self._torch_dtype
+            model.to(vae_dtype)
 
         return model
 
@@ -109,9 +136,9 @@ class BnbQuantizedLlmInt8bCheckpointModel(ModelLoader):
                 "The bnb modules are not available. Please install bitsandbytes if available on your platform."
             )
         match submodel_type:
-            case SubModelType.Tokenizer2:
+            case SubModelType.Tokenizer2 | SubModelType.Tokenizer3:
                 return T5Tokenizer.from_pretrained(Path(config.path) / "tokenizer_2", max_length=512)
-            case SubModelType.TextEncoder2:
+            case SubModelType.TextEncoder2 | SubModelType.TextEncoder3:
                 te2_model_path = Path(config.path) / "text_encoder_2"
                 model_config = AutoConfig.from_pretrained(te2_model_path)
                 with accelerate.init_empty_weights():
@@ -153,10 +180,10 @@ class T5EncoderCheckpointModel(ModelLoader):
             raise ValueError("Only T5EncoderConfig models are currently supported here.")
 
         match submodel_type:
-            case SubModelType.Tokenizer2:
+            case SubModelType.Tokenizer2 | SubModelType.Tokenizer3:
                 return T5Tokenizer.from_pretrained(Path(config.path) / "tokenizer_2", max_length=512)
-            case SubModelType.TextEncoder2:
-                return T5EncoderModel.from_pretrained(Path(config.path) / "text_encoder_2")
+            case SubModelType.TextEncoder2 | SubModelType.TextEncoder3:
+                return T5EncoderModel.from_pretrained(Path(config.path) / "text_encoder_2", torch_dtype="auto")
 
         raise ValueError(
             f"Only Tokenizer and TextEncoder submodels are currently supported. Received: {submodel_type.value if submodel_type else 'None'}"
@@ -204,6 +231,52 @@ class FluxCheckpointModel(ModelLoader):
         return model
 
 
+@ModelLoaderRegistry.register(base=BaseModelType.Flux, type=ModelType.Main, format=ModelFormat.GGUFQuantized)
+class FluxGGUFCheckpointModel(ModelLoader):
+    """Class to load GGUF main models."""
+
+    def _load_model(
+        self,
+        config: AnyModelConfig,
+        submodel_type: Optional[SubModelType] = None,
+    ) -> AnyModel:
+        if not isinstance(config, CheckpointConfigBase):
+            raise ValueError("Only CheckpointConfigBase models are currently supported here.")
+
+        match submodel_type:
+            case SubModelType.Transformer:
+                return self._load_from_singlefile(config)
+
+        raise ValueError(
+            f"Only Transformer submodels are currently supported. Received: {submodel_type.value if submodel_type else 'None'}"
+        )
+
+    def _load_from_singlefile(
+        self,
+        config: AnyModelConfig,
+    ) -> AnyModel:
+        assert isinstance(config, MainGGUFCheckpointConfig)
+        model_path = Path(config.path)
+
+        with SilenceWarnings():
+            model = Flux(params[config.config_path])
+
+            # HACK(ryand): We shouldn't be hard-coding the compute_dtype here.
+            sd = gguf_sd_loader(model_path, compute_dtype=torch.bfloat16)
+
+        # HACK(ryand): There are some broken GGUF models in circulation that have the wrong shape for img_in.weight.
+        # We override the shape here to fix the issue.
+        # Example model with this issue (Q4_K_M): https://civitai.com/models/705823/ggufk-flux-unchained-km-quants
+        img_in_weight = sd.get("img_in.weight", None)
+        if img_in_weight is not None and img_in_weight._ggml_quantization_type in TORCH_COMPATIBLE_QTYPES:
+            expected_img_in_weight_shape = model.img_in.weight.shape
+            img_in_weight.quantized_data = img_in_weight.quantized_data.view(expected_img_in_weight_shape)
+            img_in_weight.tensor_shape = expected_img_in_weight_shape
+
+        model.load_state_dict(sd, assign=True)
+        return model
+
+
 @ModelLoaderRegistry.register(base=BaseModelType.Flux, type=ModelType.Main, format=ModelFormat.BnbQuantizednf4b)
 class FluxBnbQuantizednf4bCheckpointModel(ModelLoader):
     """Class to load main models."""
@@ -244,3 +317,74 @@ class FluxBnbQuantizednf4bCheckpointModel(ModelLoader):
                 sd = convert_bundle_to_flux_transformer_checkpoint(sd)
             model.load_state_dict(sd, assign=True)
         return model
+
+
+@ModelLoaderRegistry.register(base=BaseModelType.Flux, type=ModelType.ControlNet, format=ModelFormat.Checkpoint)
+@ModelLoaderRegistry.register(base=BaseModelType.Flux, type=ModelType.ControlNet, format=ModelFormat.Diffusers)
+class FluxControlnetModel(ModelLoader):
+    """Class to load FLUX ControlNet models."""
+
+    def _load_model(
+        self,
+        config: AnyModelConfig,
+        submodel_type: Optional[SubModelType] = None,
+    ) -> AnyModel:
+        if isinstance(config, ControlNetCheckpointConfig):
+            model_path = Path(config.path)
+        elif isinstance(config, ControlNetDiffusersConfig):
+            # If this is a diffusers directory, we simply ignore the config file and load from the weight file.
+            model_path = Path(config.path) / "diffusion_pytorch_model.safetensors"
+        else:
+            raise ValueError(f"Unexpected ControlNet model config type: {type(config)}")
+
+        sd = load_file(model_path)
+
+        # Detect the FLUX ControlNet model type from the state dict.
+        if is_state_dict_xlabs_controlnet(sd):
+            return self._load_xlabs_controlnet(sd)
+        elif is_state_dict_instantx_controlnet(sd):
+            return self._load_instantx_controlnet(sd)
+        else:
+            raise ValueError("Do not recognize the state dict as an XLabs or InstantX ControlNet model.")
+
+    def _load_xlabs_controlnet(self, sd: dict[str, torch.Tensor]) -> AnyModel:
+        with accelerate.init_empty_weights():
+            # HACK(ryand): Is it safe to assume dev here?
+            model = XLabsControlNetFlux(params["flux-dev"])
+
+        model.load_state_dict(sd, assign=True)
+        return model
+
+    def _load_instantx_controlnet(self, sd: dict[str, torch.Tensor]) -> AnyModel:
+        sd = convert_diffusers_instantx_state_dict_to_bfl_format(sd)
+        flux_params = infer_flux_params_from_state_dict(sd)
+        num_control_modes = infer_instantx_num_control_modes_from_state_dict(sd)
+
+        with accelerate.init_empty_weights():
+            model = InstantXControlNetFlux(flux_params, num_control_modes)
+
+        model.load_state_dict(sd, assign=True)
+        return model
+
+
+@ModelLoaderRegistry.register(base=BaseModelType.Flux, type=ModelType.IPAdapter, format=ModelFormat.Checkpoint)
+class FluxIpAdapterModel(ModelLoader):
+    """Class to load FLUX IP-Adapter models."""
+
+    def _load_model(
+        self,
+        config: AnyModelConfig,
+        submodel_type: Optional[SubModelType] = None,
+    ) -> AnyModel:
+        if not isinstance(config, IPAdapterCheckpointConfig):
+            raise ValueError(f"Unexpected model config type: {type(config)}.")
+
+        sd = load_file(Path(config.path))
+
+        params = infer_xlabs_ip_adapter_params_from_state_dict(sd)
+
+        with accelerate.init_empty_weights():
+            model = XlabsIpAdapterFlux(params=params)
+
+        model.load_xlabs_state_dict(sd, assign=True)
+        return model

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

@@ -22,7 +22,6 @@ from invokeai.backend.model_manager.load.load_default import ModelLoader
 from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
 
 
-@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.CLIPVision, format=ModelFormat.Diffusers)
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.T2IAdapter, format=ModelFormat.Diffusers)
 class GenericDiffusersLoader(ModelLoader):
     """Class to load simple diffusers models."""

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

@@ -42,6 +42,7 @@ VARIANT_TO_IN_CHANNEL_MAP = {
 @ModelLoaderRegistry.register(
     base=BaseModelType.StableDiffusionXLRefiner, type=ModelType.Main, format=ModelFormat.Diffusers
 )
+@ModelLoaderRegistry.register(base=BaseModelType.StableDiffusion3, type=ModelType.Main, format=ModelFormat.Diffusers)
 @ModelLoaderRegistry.register(base=BaseModelType.StableDiffusion1, type=ModelType.Main, format=ModelFormat.Checkpoint)
 @ModelLoaderRegistry.register(base=BaseModelType.StableDiffusion2, type=ModelType.Main, format=ModelFormat.Checkpoint)
 @ModelLoaderRegistry.register(base=BaseModelType.StableDiffusionXL, type=ModelType.Main, format=ModelFormat.Checkpoint)
@@ -51,13 +52,6 @@ VARIANT_TO_IN_CHANNEL_MAP = {
 class StableDiffusionDiffusersModel(GenericDiffusersLoader):
     """Class to load main models."""
 
-    model_base_to_model_type = {
-        BaseModelType.StableDiffusion1: "FrozenCLIPEmbedder",
-        BaseModelType.StableDiffusion2: "FrozenOpenCLIPEmbedder",
-        BaseModelType.StableDiffusionXL: "SDXL",
-        BaseModelType.StableDiffusionXLRefiner: "SDXL-Refiner",
-    }
-
     def _load_model(
         self,
         config: AnyModelConfig,
@@ -117,8 +111,6 @@ class StableDiffusionDiffusersModel(GenericDiffusersLoader):
             load_class = load_classes[config.base][config.variant]
         except KeyError as e:
             raise Exception(f"No diffusers pipeline known for base={config.base}, variant={config.variant}") from e
-        prediction_type = config.prediction_type.value
-        upcast_attention = config.upcast_attention
 
         # Without SilenceWarnings we get log messages like this:
         # site-packages/huggingface_hub/file_download.py:1132: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`.
@@ -129,13 +121,7 @@ class StableDiffusionDiffusersModel(GenericDiffusersLoader):
         # ['text_model.embeddings.position_ids']
 
         with SilenceWarnings():
-            pipeline = load_class.from_single_file(
-                config.path,
-                torch_dtype=self._torch_dtype,
-                prediction_type=prediction_type,
-                upcast_attention=upcast_attention,
-                load_safety_checker=False,
-            )
+            pipeline = load_class.from_single_file(config.path, torch_dtype=self._torch_dtype)
 
         if not submodel_type:
             return pipeline

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

@@ -20,7 +20,7 @@ from typing import Optional
 
 import requests
 from huggingface_hub import HfApi, configure_http_backend, hf_hub_url
-from huggingface_hub.utils._errors import RepositoryNotFoundError, RevisionNotFoundError
+from huggingface_hub.errors import RepositoryNotFoundError, RevisionNotFoundError
 from pydantic.networks import AnyHttpUrl
 from requests.sessions import Session