Log the COMMANDLINE_ARGS sent to Forge

This reverts commit f737921eaa, reversing
changes made to b12a6b9537.

.gitignore

@@ -3,4 +3,6 @@ installer
 installer.tar
 dist
 .idea/*
-node_modules/*
+node_modules/*
+.tmp1
+.tmp2

CHANGES.md

@@ -1,5 +1,33 @@
 # What's new?
 
+## v3.5 (preview)
+### Major Changes
+- **Chroma** - support for the Chroma model, including quantized bnb and nf4 models.
+- **Flux** - full support for the Flux model, including quantized bnb and nf4 models.
+- **LyCORIS** - including `LoCon`, `Hada`, `IA3` and `Lokr`.
+- **11 new samplers** - `DDIM CFG++`, `DPM Fast`, `DPM++ 2m SDE Heun`, `DPM++ 3M SDE`, `Restart`, `Heun PP2`, `IPNDM`, `IPNDM_V`, `LCM`, `[Forge] Flux Realistic`, `[Forge] Flux Realistic (Slow)`.
+- **15 new schedulers** - `Uniform`, `Karras`, `Exponential`, `Polyexponential`, `SGM Uniform`, `KL Optimal`, `Align Your Steps`, `Normal`, `DDIM`, `Beta`, `Turbo`, `Align Your Steps GITS`, `Align Your Steps 11`, `Align Your Steps 32`.
+- **42 new Controlnet filters, and support for lots of new ControlNet models** (including QR ControlNets).
+- **5 upscalers** - `SwinIR`, `ScuNET`, `Nearest`, `Lanczos`, `ESRGAN`.
+- **Faster than v3.0**
+- **Major rewrite of the code** - We've switched to `Forge WebUI` under the hood, which brings a lot of new features, faster image generation, and support for all the extensions in the Forge/Automatic1111 community. This allows Easy Diffusion to stay up-to-date with the latest features, and focus on making the UI and installation experience even easier.
+
+v3.5 is currently an optional upgrade, and you can switch between the v3.0 (diffusers) engine and the v3.5 (webui) engine using the `Settings` tab in the UI.
+
+### Detailed changelog
+* 3.5.9 - 18 Jul 2025 - Stability fix for the Forge backend. Prevents unused Forge processes from hanging around even after closing Easy Diffusion.
+* 3.5.8 - 14 Jul 2025 - Support custom Text Encoders and Flux VAEs in the UI.
+* 3.5.7 - 27 Jun 2025 - Support for the Chroma model. Update Forge to the latest commit.
+* 3.5.6 - 17 Feb 2025 - Fix broken model merging.
+* 3.5.5 - 10 Feb 2025 - (Internal code change) Use `torchruntime` for installing torch/torchvision, instead of custom logic. This supports a lot more GPUs on various platforms, and was built using Easy Diffusion's torch-installation code.
+* 3.5.4 - 8 Feb 2025 - Fix a bug where the inpainting mask wasn't resized to the image size when using the WebUI/v3.5 backend. Thanks @AvidGameFan for their help in investigating and fixing this!
+* 3.5.3 - 6 Feb 2025 - (Internal code change) Remove hardcoded references to `torch.cuda`, and replace with torchruntime's device utilities.
+* 3.5.2 - 28 Jan 2025 - Fix for accidental jailbreak when using conda with WebUI - fixes the `type not subscriptable` error when using WebUI.
+* 3.5.2 - 28 Jan 2025 - Fix a bug affecting older versions of Easy Diffusion, which tried to upgrade to an incompatible version of PyTorch.
+* 3.5.2 - 4 Jan 2025 - Replace the use of WMIC (deprecated) with a powershell call.
+* 3.5.1 - 17 Dec 2024 - Update Forge to the latest commit.
+* 3.5.0 - 11 Oct 2024 - **Preview release** of the new v3.5 engine, powered by Forge WebUI (a fork of Automatic1111). This enables Flux, SD3, LyCORIS and lots of new features, while using the same familiar Easy Diffusion interface.
+
 ## v3.0
 ### Major Changes
 - **ControlNet** - Full support for ControlNet, with native integration of the common ControlNet models. Just select a control image, then choose the ControlNet filter/model and run. No additional configuration or download necessary. Supports custom ControlNets as well.
@@ -17,6 +45,24 @@
 - **Major rewrite of the code** - We've switched to using diffusers under-the-hood, which allows us to release new features faster, and focus on making the UI and installer even easier to use.
 
 ### Detailed changelog
+* 3.0.13 - 10 Feb 2025 - (Internal code change) Use `torchruntime` for installing torch/torchvision, instead of custom logic. This supports a lot more GPUs on various platforms, and was built using Easy Diffusion's torch-installation code.
+* 3.0.12 - 6 Feb 2025 - (Internal code change) Remove hardcoded references to `torch.cuda`, and replace with torchruntime's device utilities.
+* 3.0.11 - 4 Jan 2025 - Replace the use of WMIC (deprecated) with a powershell call.
+* 3.0.10 - 11 Oct 2024 - **Major Update** - An option to upgrade to v3.5, which enables Flux, Stable Diffusion 3, LyCORIS models and lots more.
+* 3.0.9 - 28 May 2024 - Slider for controlling the strength of controlnets.
+* 3.0.8 - 27 May 2024 - SDXL ControlNets for Img2Img and Inpainting.
+* 3.0.7 - 11 Dec 2023 - Setting to enable/disable VAE tiling (in the Image Settings panel). Sometimes VAE tiling reduces the quality of the image, so this setting will help control that.
+* 3.0.6 - 18 Sep 2023 - Add thumbnails to embeddings from the UI, using the new `Upload Thumbnail` button in the Embeddings popup. Thanks @JeLuf.
+* 3.0.6 - 15 Sep 2023 - Fix broken embeddings dialog when LoRA information couldn't be fetched.
+* 3.0.6 - 14 Sep 2023 - UI for adding notes to LoRA files (to help you remember which prompts to use). Also added a button to automatically fetch prompts from Civitai for a LoRA file, using the `Import from Civitai` button. Thanks @JeLuf.
+* 3.0.5 - 2 Sep 2023 - Support SDXL ControlNets.
+* 3.0.4 - 1 Sep 2023 - Fix incorrect metadata generated for embeddings, when the exact word doesn't match the case, or is part of a larger word.
+* 3.0.4 - 1 Sep 2023 - Simplify the installation for AMD users on Linux. Thanks @JeLuf.
+* 3.0.4 - 1 Sep 2023 - Allow using a different folder for models. This is useful if you want to share a models folder across different software, or on a different drive. You can change this path in the Settings tab.
+* 3.0.3 - 31 Aug 2023 - Auto-save images to disk (if enabled by the user) when upscaling/fixing using the buttons on the image.
+* 3.0.3 - 30 Aug 2023 - Allow loading NovelAI-based custom models.
+* 3.0.3 - 30 Aug 2023 - Fix broken VAE tiling. This allows you to create larger images with lesser VRAM usage.
+* 3.0.3 - 30 Aug 2023 - Allow blocking NSFW images using a server-side config. This prevents the browser from generating NSFW images or changing the config. Open `config.yaml` in a text editor (e.g. Notepad), and add `block_nsfw: true` at the end, and save the file.
 * 3.0.2 - 29 Aug 2023 - Fixed incorrect matching of embeddings from prompts.
 * 3.0.2 - 24 Aug 2023 - Fix broken seamless tiling.
 * 3.0.2 - 23 Aug 2023 - Fix styling on mobile devices.

CONTRIBUTING.md

@@ -47,3 +47,5 @@ Build the Windows installer using Windows, and the Linux installer using Linux.
 
 1. Run `build.bat` or `./build.sh` depending on whether you're in Windows or Linux.
 2. Make a new GitHub release and upload the Windows and Linux installer builds created inside the `dist` folder.
+
+For NSIS (on Windows), you need to have these plugins in the `nsis/Plugins` folder: `amd64-unicode`, `x86-ansi`, `x86-unicode`

NSIS/nsisconf.nsh

@@ -1 +0,0 @@
-!define EXISTING_INSTALLATION_DIR "D:\path\to\installed\easy-diffusion"

NSIS/sdui.nsi

@@ -3,13 +3,14 @@
 Target amd64-unicode
 Unicode True
 SetCompressor /FINAL lzma
+SetCompressorDictSize 64
 RequestExecutionLevel user
 !AddPluginDir /amd64-unicode "."
 ; HM NIS Edit Wizard helper defines
 !define PRODUCT_NAME "Easy Diffusion"
-!define PRODUCT_VERSION "2.5"
+!define PRODUCT_VERSION "3.0"
 !define PRODUCT_PUBLISHER "cmdr2 and contributors"
-!define PRODUCT_WEB_SITE "https://stable-diffusion-ui.github.io"
+!define PRODUCT_WEB_SITE "https://easydiffusion.github.io"
 !define PRODUCT_DIR_REGKEY "Software\Microsoft\Easy Diffusion\App Paths\installer.exe"
 
 ; MUI 1.67 compatible ------
@@ -165,9 +166,9 @@ FunctionEnd
 ; MUI Settings
 ;---------------------------------------------------------------------------------------------------------
 !define MUI_ABORTWARNING
-!define MUI_ICON "cyborg_flower_girl.ico"
+!define MUI_ICON "${EXISTING_INSTALLATION_DIR}\installer_files\cyborg_flower_girl.ico"
 
-!define MUI_WELCOMEFINISHPAGE_BITMAP "cyborg_flower_girl.bmp"
+!define MUI_WELCOMEFINISHPAGE_BITMAP "${EXISTING_INSTALLATION_DIR}\installer_files\cyborg_flower_girl.bmp"
 
 ; Welcome page
 !define MUI_WELCOMEPAGE_TEXT "This installer will guide you through the installation of Easy Diffusion.$\n$\n\
@@ -176,8 +177,8 @@ Click Next to continue."
 Page custom MediaPackDialog
 
 ; License page
-!insertmacro MUI_PAGE_LICENSE "..\LICENSE"
-!insertmacro MUI_PAGE_LICENSE "..\CreativeML Open RAIL-M License"
+!insertmacro MUI_PAGE_LICENSE "${EXISTING_INSTALLATION_DIR}\LICENSE"
+!insertmacro MUI_PAGE_LICENSE "${EXISTING_INSTALLATION_DIR}\CreativeML Open RAIL-M License"
 ; Directory page
 !define MUI_PAGE_CUSTOMFUNCTION_LEAVE "DirectoryLeave"
 !insertmacro MUI_PAGE_DIRECTORY
@@ -210,29 +211,33 @@ ShowInstDetails show
 ; List of files to be installed
 Section "MainSection" SEC01
   SetOutPath "$INSTDIR"
-  File "..\CreativeML Open RAIL-M License"
-  File "..\How to install and run.txt"
-  File "..\LICENSE"
-  File "..\scripts\Start Stable Diffusion UI.cmd"
+  File "${EXISTING_INSTALLATION_DIR}\CreativeML Open RAIL-M License"
+  File "${EXISTING_INSTALLATION_DIR}\How to install and run.txt"
+  File "${EXISTING_INSTALLATION_DIR}\LICENSE"
+  File "${EXISTING_INSTALLATION_DIR}\Start Stable Diffusion UI.cmd"
   File /r "${EXISTING_INSTALLATION_DIR}\installer_files"
-  File /r "${EXISTING_INSTALLATION_DIR}\profile"
   File /r "${EXISTING_INSTALLATION_DIR}\sd-ui-files"
-  SetOutPath "$INSTDIR\installer_files"
-  File "cyborg_flower_girl.ico"
+
   SetOutPath "$INSTDIR\scripts"
   File "${EXISTING_INSTALLATION_DIR}\scripts\install_status.txt"
-  File "..\scripts\on_env_start.bat"
+  File "${EXISTING_INSTALLATION_DIR}\scripts\on_env_start.bat"
   File "C:\windows\system32\curl.exe"
-  CreateDirectory "$INSTDIR\models"
+  File "${EXISTING_INSTALLATION_DIR}\scripts\config.yaml.sample"
+
   CreateDirectory "$INSTDIR\models\stable-diffusion"
   CreateDirectory "$INSTDIR\models\gfpgan"
   CreateDirectory "$INSTDIR\models\realesrgan"
   CreateDirectory "$INSTDIR\models\vae"
+
+  CreateDirectory "$INSTDIR\profile\.cache\huggingface\hub"
+  SetOutPath "$INSTDIR\profile\.cache\huggingface\hub"
+  File /r /x pytorch_model.bin "${EXISTING_INSTALLATION_DIR}\profile\.cache\huggingface\hub\models--openai--clip-vit-large-patch14"
+
   CreateDirectory "$SMPROGRAMS\Easy Diffusion"
   CreateShortCut "$SMPROGRAMS\Easy Diffusion\Easy Diffusion.lnk" "$INSTDIR\Start Stable Diffusion UI.cmd" "" "$INSTDIR\installer_files\cyborg_flower_girl.ico"
 
   DetailPrint 'Downloading the Stable Diffusion 1.4 model...'
-  NScurl::http get "https://huggingface.co/CompVis/stable-diffusion-v-1-4-original/resolve/main/sd-v1-4.ckpt" "$INSTDIR\models\stable-diffusion\sd-v1-4.ckpt" /CANCEL /INSIST /END
+  NScurl::http get "https://github.com/easydiffusion/sdkit-test-data/releases/download/assets/sd-v1-4.safetensors" "$INSTDIR\models\stable-diffusion\sd-v1-4.safetensors" /CANCEL /INSIST /END
 
   DetailPrint 'Downloading the GFPGAN model...'
   NScurl::http get "https://github.com/TencentARC/GFPGAN/releases/download/v1.3.4/GFPGANv1.4.pth" "$INSTDIR\models\gfpgan\GFPGANv1.4.pth" /CANCEL /INSIST /END

README.md

@@ -3,7 +3,9 @@
 
 Does not require technical knowledge, does not require pre-installed software. 1-click install, powerful features, friendly community.
 
-[Installation guide](#installation) | [Troubleshooting guide](https://github.com/easydiffusion/easydiffusion/wiki/Troubleshooting) | <sub>[![Discord Server](https://img.shields.io/discord/1014774730907209781?label=Discord)](https://discord.com/invite/u9yhsFmEkB)</sub> <sup>(for support queries, and development discussions)</sup>
+️‍🔥🎉 **New!** Support for SDXL, ControlNet, multiple LoRA files, embeddings (and a lot more) have been added!
+
+[Installation guide](#installation) | [Troubleshooting guide](https://github.com/easydiffusion/easydiffusion/wiki/Troubleshooting) | [User guide](https://github.com/easydiffusion/easydiffusion/wiki) | <sub>[![Discord Server](https://img.shields.io/discord/1014774730907209781?label=Discord)](https://discord.com/invite/u9yhsFmEkB)</sub> <sup>(for support queries, and development discussions)</sup>
 
 ---
 ![262597678-11089485-2514-4a11-88fb-c3acc81fc9ec](https://github.com/easydiffusion/easydiffusion/assets/844287/050b5e15-e909-45bf-8162-a38234830e38)
@@ -102,7 +104,7 @@ Just delete the `EasyDiffusion` folder to uninstall all the downloaded packages.
 - **Auto scan for malicious models**: Uses picklescan to prevent malicious models.
 - **Safetensors support**: Support loading models in the safetensor format, for improved safety.
 - **Auto-updater**: Gets you the latest improvements and bug-fixes to a rapidly evolving project.
-- **Developer Console**: A developer-mode for those who want to modify their Stable Diffusion code, and edit the conda environment.
+- **Developer Console**: A developer-mode for those who want to modify their Stable Diffusion code, modify packages, and edit the conda environment.
 
 **(and a lot more)**
 

build.bat

@@ -1,48 +1,78 @@
 @echo off
+setlocal enabledelayedexpansion
 
 @echo "Hi there, what you are running is meant for the developers of this project, not for users." & echo.
-@echo "If you only want to use the Stable Diffusion UI, you've downloaded the wrong file."
+@echo "If you only want to use Easy Diffusion, you've downloaded the wrong file."
 @echo "Please download and follow the instructions at https://github.com/easydiffusion/easydiffusion#installation" & echo.
 @echo "If you are actually a developer of this project, please type Y and press enter" & echo.
 
 set /p answer=Are you a developer of this project (Y/N)?
 if /i "%answer:~,1%" NEQ "Y" exit /b
 
-mkdir dist\win\stable-diffusion-ui\scripts
-@REM mkdir dist\linux-mac\stable-diffusion-ui\scripts
+@rem verify dependencies
+call makensis /VERSION >.tmp1 2>.tmp2
+if "!ERRORLEVEL!" NEQ "0" (
+    echo makensis.exe not found! Download it from https://sourceforge.net/projects/nsisbi/files/ and set it on the PATH variable.
+    pause
+    exit
+)
 
-@rem copy the installer files for Windows
+set /p OUT_DIR=Output folder path (will create the installer files inside this, e.g. F:\EasyDiffusion): 
 
-copy scripts\on_env_start.bat dist\win\stable-diffusion-ui\scripts\
-copy scripts\bootstrap.bat dist\win\stable-diffusion-ui\scripts\
-copy scripts\config.yaml.sample dist\win\stable-diffusion-ui\scripts\config.yaml
-copy "scripts\Start Stable Diffusion UI.cmd" dist\win\stable-diffusion-ui\
-copy LICENSE dist\win\stable-diffusion-ui\
-copy "CreativeML Open RAIL-M License" dist\win\stable-diffusion-ui\
-copy "How to install and run.txt" dist\win\stable-diffusion-ui\
-echo. > dist\win\stable-diffusion-ui\scripts\install_status.txt
+mkdir "%OUT_DIR%\scripts"
+mkdir "%OUT_DIR%\installer_files"
 
-@rem copy the installer files for Linux and Mac
+set BASE_DIR=%cd%
 
-@REM copy scripts\on_env_start.sh dist\linux-mac\stable-diffusion-ui\scripts\
-@REM copy scripts\bootstrap.sh dist\linux-mac\stable-diffusion-ui\scripts\
-@REM copy scripts\start.sh dist\linux-mac\stable-diffusion-ui\
-@REM copy LICENSE dist\linux-mac\stable-diffusion-ui\
-@REM copy "CreativeML Open RAIL-M License" dist\linux-mac\stable-diffusion-ui\
-@REM copy "How to install and run.txt" dist\linux-mac\stable-diffusion-ui\
-@REM echo. > dist\linux-mac\stable-diffusion-ui\scripts\install_status.txt
+@rem STEP 1: copy the installer files for Windows
 
-@rem make the zip
-
-cd dist\win
-call powershell Compress-Archive -Path stable-diffusion-ui -DestinationPath ..\stable-diffusion-ui-windows.zip
-cd ..\..
-
-@REM cd dist\linux-mac
-@REM call powershell Compress-Archive -Path stable-diffusion-ui -DestinationPath ..\stable-diffusion-ui-linux.zip
-@REM call powershell Compress-Archive -Path stable-diffusion-ui -DestinationPath ..\stable-diffusion-ui-mac.zip
-@REM cd ..\..
-
-echo "Build ready. Upload the zip files inside the 'dist' folder."
+copy "%BASE_DIR%\scripts\on_env_start.bat" "%OUT_DIR%\scripts\"
+copy "%BASE_DIR%\scripts\config.yaml.sample" "%OUT_DIR%\scripts\config.yaml.sample"
+copy "%BASE_DIR%\scripts\Start Stable Diffusion UI.cmd" "%OUT_DIR%\"
+copy "%BASE_DIR%\LICENSE" "%OUT_DIR%\"
+copy "%BASE_DIR%\CreativeML Open RAIL-M License" "%OUT_DIR%\"
+copy "%BASE_DIR%\How to install and run.txt" "%OUT_DIR%\"
+copy "%BASE_DIR%\NSIS\cyborg_flower_girl.ico" "%OUT_DIR%\installer_files\"
+copy "%BASE_DIR%\NSIS\cyborg_flower_girl.bmp" "%OUT_DIR%\installer_files\"
+echo. > "%OUT_DIR%\scripts\install_status.txt"
 
+echo ----
+echo Basic files ready. Verify the files in %OUT_DIR%, then press Enter to initialize the environment, or close to quit.
+echo ----
 pause
+
+@rem STEP 2: Initialize the environment with git, python and conda
+
+cd /d "%OUT_DIR%\"
+call "%BASE_DIR%\scripts\bootstrap.bat"
+
+echo ----
+echo Environment ready. Verify the environment, then press Enter to download the necessary packages, or close to quit.
+echo ----
+pause
+
+@rem STEP 3: Download the packages and create a working installation
+
+cd /d "%OUT_DIR%\"
+start "Install Easy Diffusion" /D "%OUT_DIR%" "Start Stable Diffusion UI.cmd"
+
+echo ----
+echo Installation in progress (in a new window). Once complete, verify the installation, then press Enter to create an installer from these files, or close to quit.
+echo ----
+pause
+
+@rem STEP 4: Build the installer from a working installation
+
+cd /d "%OUT_DIR%\"
+
+echo ^^!define EXISTING_INSTALLATION_DIR "%OUT_DIR%" > nsisconf.nsh
+call makensis /NOCD /V4 "%BASE_DIR%\NSIS\sdui.nsi"
+
+echo ----
+if "!ERRORLEVEL!" EQU "0" (
+    echo Installer built successfully at %OUT_DIR%
+) else (
+    echo Installer failed to build at %OUT_DIR%
+)
+echo ----
+pause

build.sh

@@ -1,7 +1,7 @@
 #!/bin/bash
 
 printf "Hi there, what you are running is meant for the developers of this project, not for users.\n\n"
-printf "If you only want to use the Stable Diffusion UI, you've downloaded the wrong file.\n"
+printf "If you only want to use Easy Diffusion, you've downloaded the wrong file.\n"
 printf "Please download and follow the instructions at https://github.com/easydiffusion/easydiffusion#installation \n\n"
 printf "If you are actually a developer of this project, please type Y and press enter\n\n"
 
@@ -11,40 +11,30 @@ case $yn in
     * ) exit;;
 esac
 
-# mkdir -p dist/win/stable-diffusion-ui/scripts
-mkdir -p dist/linux-mac/stable-diffusion-ui/scripts
-
-# copy the installer files for Windows
-
-# cp scripts/on_env_start.bat dist/win/stable-diffusion-ui/scripts/
-# cp scripts/bootstrap.bat dist/win/stable-diffusion-ui/scripts/
-# cp "scripts/Start Stable Diffusion UI.cmd" dist/win/stable-diffusion-ui/
-# cp LICENSE dist/win/stable-diffusion-ui/
-# cp "CreativeML Open RAIL-M License" dist/win/stable-diffusion-ui/
-# cp "How to install and run.txt" dist/win/stable-diffusion-ui/
-# echo "" > dist/win/stable-diffusion-ui/scripts/install_status.txt
+mkdir -p dist/linux-mac/easy-diffusion/scripts
 
 # copy the installer files for Linux and Mac
 
-cp scripts/on_env_start.sh dist/linux-mac/stable-diffusion-ui/scripts/
-cp scripts/bootstrap.sh dist/linux-mac/stable-diffusion-ui/scripts/
-cp scripts/functions.sh dist/linux-mac/stable-diffusion-ui/scripts/
-cp scripts/config.yaml.sample dist/linux-mac/stable-diffusion-ui/scripts/config.yaml
-cp scripts/start.sh dist/linux-mac/stable-diffusion-ui/
-cp LICENSE dist/linux-mac/stable-diffusion-ui/
-cp "CreativeML Open RAIL-M License" dist/linux-mac/stable-diffusion-ui/
-cp "How to install and run.txt" dist/linux-mac/stable-diffusion-ui/
-echo "" > dist/linux-mac/stable-diffusion-ui/scripts/install_status.txt
+cp scripts/on_env_start.sh dist/linux-mac/easy-diffusion/scripts/
+cp scripts/bootstrap.sh dist/linux-mac/easy-diffusion/scripts/
+cp scripts/functions.sh dist/linux-mac/easy-diffusion/scripts/
+cp scripts/config.yaml.sample dist/linux-mac/easy-diffusion/scripts/config.yaml.sample
+cp scripts/start.sh dist/linux-mac/easy-diffusion/
+cp LICENSE dist/linux-mac/easy-diffusion/
+cp "CreativeML Open RAIL-M License" dist/linux-mac/easy-diffusion/
+cp "How to install and run.txt" dist/linux-mac/easy-diffusion/
+echo "" > dist/linux-mac/easy-diffusion/scripts/install_status.txt
+
+# set the permissions
+chmod u+x dist/linux-mac/easy-diffusion/scripts/on_env_start.sh
+chmod u+x dist/linux-mac/easy-diffusion/scripts/bootstrap.sh
+chmod u+x dist/linux-mac/easy-diffusion/start.sh
 
 # make the zip
 
-# cd dist/win
-# zip -r ../stable-diffusion-ui-windows.zip stable-diffusion-ui
-# cd ../..
-
 cd dist/linux-mac
-zip -r ../stable-diffusion-ui-linux.zip stable-diffusion-ui
-zip -r ../stable-diffusion-ui-mac.zip stable-diffusion-ui
+zip -r ../Easy-Diffusion-Linux.zip easy-diffusion
+zip -r ../Easy-Diffusion-Mac.zip easy-diffusion
 cd ../..
 
 echo "Build ready. Upload the zip files inside the 'dist' folder."

scripts/Developer Console.cmd

@@ -4,7 +4,7 @@ echo "Opening Stable Diffusion UI - Developer Console.." & echo.
 
 cd /d %~dp0
 
-set PATH=C:\Windows\System32;%PATH%
+set PATH=C:\Windows\System32;C:\Windows\System32\WindowsPowerShell\v1.0;%PATH%
 
 @rem set legacy and new installer's PATH, if they exist
 if exist "installer" set PATH=%cd%\installer;%cd%\installer\Library\bin;%cd%\installer\Scripts;%cd%\installer\Library\usr\bin;%PATH%
@@ -26,18 +26,23 @@ call conda --version
 echo.
 echo COMSPEC=%COMSPEC%
 echo.
+powershell -Command "(Get-WmiObject Win32_VideoController | Select-Object Name, AdapterRAM, DriverDate, DriverVersion)"
 
 @rem activate the legacy environment (if present) and set PYTHONPATH
 if exist "installer_files\env" (
     set PYTHONPATH=%cd%\installer_files\env\lib\site-packages
+    set PYTHON=%cd%\installer_files\env\python.exe
+    echo PYTHON=%PYTHON%
 )
 if exist "stable-diffusion\env" (
     call conda activate .\stable-diffusion\env
     set PYTHONPATH=%cd%\stable-diffusion\env\lib\site-packages
+    set PYTHON=%cd%\stable-diffusion\env\python.exe
+    echo PYTHON=%PYTHON%
 )
 
-call where python
-call python --version
+@REM call where python
+call "%PYTHON%" --version
 
 echo PYTHONPATH=%PYTHONPATH%
 

scripts/Start Stable Diffusion UI.cmd

@@ -3,7 +3,7 @@
 cd /d %~dp0
 echo Install dir: %~dp0
 
-set PATH=C:\Windows\System32;%PATH%
+set PATH=C:\Windows\System32;C:\Windows\System32\WindowsPowerShell\v1.0;%PATH%
 set PYTHONHOME=
 
 if exist "on_sd_start.bat" (
@@ -15,7 +15,7 @@ if exist "on_sd_start.bat" (
     echo download. This will not work.
     echo.
     echo Recommended: Please close this window and download the installer from
-    echo https://stable-diffusion-ui.github.io/docs/installation/
+    echo https://easydiffusion.github.io/docs/installation/
     echo.
     echo ================================================================================
     echo.
@@ -39,6 +39,7 @@ call where conda
 call conda --version
 echo .
 echo COMSPEC=%COMSPEC%
+powershell -Command "(Get-WmiObject Win32_VideoController | Select-Object Name, AdapterRAM, DriverDate, DriverVersion)"
 
 @rem Download the rest of the installer and UI
 call scripts\on_env_start.bat

scripts/bootstrap.bat

@@ -14,6 +14,8 @@ set LEGACY_INSTALL_ENV_DIR=%cd%\installer
 set MICROMAMBA_DOWNLOAD_URL=https://github.com/easydiffusion/easydiffusion/releases/download/v1.1/micromamba.exe
 set umamba_exists=F
 
+set PYTHONHOME=
+
 set OLD_APPDATA=%APPDATA%
 set OLD_USERPROFILE=%USERPROFILE%
 set APPDATA=%cd%\installer_files\appdata
@@ -22,15 +24,12 @@ set USERPROFILE=%cd%\profile
 @rem figure out whether git and conda needs to be installed
 if exist "%INSTALL_ENV_DIR%" set PATH=%INSTALL_ENV_DIR%;%INSTALL_ENV_DIR%\Library\bin;%INSTALL_ENV_DIR%\Scripts;%INSTALL_ENV_DIR%\Library\usr\bin;%PATH%
 
-set PACKAGES_TO_INSTALL=
+set PACKAGES_TO_INSTALL=git python=3.9
 
 if not exist "%LEGACY_INSTALL_ENV_DIR%\etc\profile.d\conda.sh" (
-    if not exist "%INSTALL_ENV_DIR%\etc\profile.d\conda.sh" set PACKAGES_TO_INSTALL=%PACKAGES_TO_INSTALL% conda python=3.8.5
+    if not exist "%INSTALL_ENV_DIR%\etc\profile.d\conda.sh" set PACKAGES_TO_INSTALL=%PACKAGES_TO_INSTALL% conda
 )
 
-call git --version >.tmp1 2>.tmp2
-if "!ERRORLEVEL!" NEQ "0" set PACKAGES_TO_INSTALL=%PACKAGES_TO_INSTALL% git
-
 call "%MAMBA_ROOT_PREFIX%\micromamba.exe" --version >.tmp1 2>.tmp2
 if "!ERRORLEVEL!" EQU "0" set umamba_exists=T
 

scripts/bootstrap.sh

@@ -46,7 +46,7 @@ if [ -e "$INSTALL_ENV_DIR" ]; then export PATH="$INSTALL_ENV_DIR/bin:$PATH"; fi
 
 PACKAGES_TO_INSTALL=""
 
-if [ ! -e "$LEGACY_INSTALL_ENV_DIR/etc/profile.d/conda.sh" ] && [ ! -e "$INSTALL_ENV_DIR/etc/profile.d/conda.sh" ]; then PACKAGES_TO_INSTALL="$PACKAGES_TO_INSTALL conda python=3.8.5"; fi
+if [ ! -e "$LEGACY_INSTALL_ENV_DIR/etc/profile.d/conda.sh" ] && [ ! -e "$INSTALL_ENV_DIR/etc/profile.d/conda.sh" ]; then PACKAGES_TO_INSTALL="$PACKAGES_TO_INSTALL conda python=3.9"; fi
 if ! hash "git" &>/dev/null; then PACKAGES_TO_INSTALL="$PACKAGES_TO_INSTALL git"; fi
 
 if "$MAMBA_ROOT_PREFIX/micromamba" --version &>/dev/null; then umamba_exists="T"; fi

scripts/check_modules.py

@@ -8,28 +8,42 @@ a custom index URL depending on the platform.
 
 """
 
-import os
+import os, sys
 from importlib.metadata import version as pkg_version
 import platform
 import traceback
+import shutil
+from pathlib import Path
+from pprint import pprint
+import re
+import torchruntime
+from torchruntime.device_db import get_gpus
 
 os_name = platform.system()
 
 modules_to_check = {
-    "torch": ("1.11.0", "1.13.1", "2.0.0"),
-    "torchvision": ("0.12.0", "0.14.1", "0.15.1"),
-    "sdkit": "2.0.3",
-    "stable-diffusion-sdkit": "2.1.4",
+    "setuptools": "69.5.1",
+    # "sdkit": "2.0.15.6", # checked later
+    # "diffusers": "0.21.4", # checked later
+    "stable-diffusion-sdkit": "2.1.5",
     "rich": "12.6.0",
     "uvicorn": "0.19.0",
-    "fastapi": "0.85.1",
+    "fastapi": "0.115.6",
     "pycloudflared": "0.2.0",
     "ruamel.yaml": "0.17.21",
     "sqlalchemy": "2.0.19",
     "python-multipart": "0.0.6",
-    # "xformers": "0.0.16",
+    "onnxruntime": "1.19.2",
+    "huggingface-hub": "0.21.4",
+    "wandb": "0.17.2",
+    # "torchruntime": "1.16.2",
+    "torchsde": "0.2.6",
+    "basicsr": "1.4.2",
+    "gfpgan": "1.3.8",
 }
-modules_to_log = ["torch", "torchvision", "sdkit", "stable-diffusion-sdkit"]
+modules_to_log = ["torchruntime", "torch", "torchvision", "sdkit", "stable-diffusion-sdkit", "diffusers"]
+
+BLACKWELL_DEVICES = re.compile(r"\b(?:5060|5070|5080|5090)\b")
 
 
 def version(module_name: str) -> str:
@@ -39,54 +53,60 @@ def version(module_name: str) -> str:
         return None
 
 
-def install(module_name: str, module_version: str):
-    if module_name == "xformers" and (os_name == "Darwin" or is_amd_on_linux()):
-        return
-
-    index_url = None
-    if module_name in ("torch", "torchvision"):
-        module_version, index_url = apply_torch_install_overrides(module_version)
-
-    if is_amd_on_linux():  # hack until AMD works properly on torch 2.0 (avoids black images on some cards)
-        if module_name == "torch":
-            module_version = "1.13.1+rocm5.2"
-        elif module_name == "torchvision":
-            module_version = "0.14.1+rocm5.2"
-    elif os_name == "Darwin":
-        if module_name == "torch":
-            module_version = "1.13.1"
-        elif module_name == "torchvision":
-            module_version = "0.14.1"
-
-    install_cmd = f"python -m pip install --upgrade {module_name}=={module_version}"
+def install(module_name: str, module_version: str, index_url=None):
+    install_cmd = f'"{sys.executable}" -m pip install --upgrade {module_name}=={module_version}'
 
     if index_url:
         install_cmd += f" --index-url {index_url}"
     if module_name == "sdkit" and version("sdkit") is not None:
         install_cmd += " -q"
+    if module_name in ("basicsr", "gfpgan"):
+        install_cmd += " --use-pep517"  # potential fix for https://github.com/easydiffusion/easydiffusion/issues/1942
 
     print(">", install_cmd)
     os.system(install_cmd)
 
 
-def init():
+def update_modules():
+    if version("torch") is None:
+        torchruntime.install(["torch", "torchvision"])
+    else:
+        torch_version_str = version("torch")
+        torch_version = version_str_to_tuple(torch_version_str)
+        is_cpu_torch = "+" not in torch_version_str
+        print(f"Current torch version: {torch_version} ({torch_version_str})")
+        if torch_version < (2, 7) or is_cpu_torch:
+            gpu_infos = get_gpus()
+            device_names = set(gpu.device_name for gpu in gpu_infos)
+            if any(BLACKWELL_DEVICES.search(device_name) for device_name in device_names):
+                if sys.version_info < (3, 9):
+                    print(
+                        "\n###################################\n"
+                        "NVIDIA 50xx series of graphics cards detected!\n\n"
+                        "To use this graphics card, please install the latest version of Easy Diffusion from: https://github.com/easydiffusion/easydiffusion#installation"
+                        "\n###################################\n"
+                    )
+                    sys.exit()
+                else:
+                    print("Upgrading torch to support NVIDIA 50xx series of graphics cards")
+                    torchruntime.install(["--force", "--upgrade", "torch", "torchvision"])
+
     for module_name, allowed_versions in modules_to_check.items():
-        if os.path.exists(f"../src/{module_name}"):
+        if os.path.exists(f"src/{module_name}"):
             print(f"Skipping {module_name} update, since it's in developer/editable mode")
             continue
 
         allowed_versions, latest_version = get_allowed_versions(module_name, allowed_versions)
 
-        requires_install = False
-        if module_name in ("torch", "torchvision"):
-            if version(module_name) is None:  # allow any torch version
-                requires_install = True
-            elif os_name == "Darwin" and (  # force mac to downgrade from torch 2.0
-                version("torch").startswith("2.") or version("torchvision").startswith("0.15.")
-            ):
-                requires_install = True
-        elif version(module_name) not in allowed_versions:
-            requires_install = True
+        if module_name == "setuptools":
+            if os_name == "Windows":
+                allowed_versions = ("59.8.0",)
+                latest_version = "59.8.0"
+            else:
+                allowed_versions = ("69.5.1",)
+                latest_version = "69.5.1"
+
+        requires_install = version(module_name) not in allowed_versions
 
         if requires_install:
             try:
@@ -94,60 +114,129 @@ def init():
             except:
                 traceback.print_exc()
                 fail(module_name)
+            else:
+                if version(module_name) != latest_version:
+                    print(
+                        f"WARNING! Tried to install {module_name}=={latest_version}, but the version is still {version(module_name)}!"
+                    )
 
-        if module_name in modules_to_log:
-            print(f"{module_name}: {version(module_name)}")
+    # different sdkit versions, with the corresponding diffusers
+    #  if sdkit is 2.0.15.x (or lower), then diffusers should be restricted to 0.21.4 (see below for the reason)
+    #  otherwise use the current sdkit version (with the corresponding diffusers version)
+
+    expected_sdkit_version_str = "2.0.22.9"
+    expected_diffusers_version_str = "0.28.2"
+
+    legacy_sdkit_version_str = "2.0.15.18"
+    legacy_diffusers_version_str = "0.21.4"
+
+    sdkit_version_str = version("sdkit")
+    if sdkit_version_str is None:  # first install
+        _install("sdkit", expected_sdkit_version_str)
+        _install("diffusers", expected_diffusers_version_str)
+    else:
+        sdkit_version = version_str_to_tuple(sdkit_version_str)
+        legacy_sdkit_version = version_str_to_tuple(legacy_sdkit_version_str)
+
+        if sdkit_version[:3] <= legacy_sdkit_version[:3]:
+            # stick to diffusers 0.21.4, since it preserves torch 0.11+ compatibility.
+            # upgrading beyond this will result in a 2+ GB download of torch on older installations
+            #  and a time-consuming chain of small package updates due to huggingface_hub upgrade.
+            # for now, the user will need to explicitly upgrade to a newer sdkit, to break this ceiling.
+
+            install_pkg_if_necessary("sdkit", legacy_sdkit_version_str)
+            install_pkg_if_necessary("diffusers", legacy_diffusers_version_str)
+        else:
+            torch_version = version_str_to_tuple(version("torch"))
+            if torch_version < (1, 13):
+                # install the gpu-compatible torch (if necessary), instead of the default CPU-only one
+                # from the diffusers dependency chain
+                torchruntime.install(["--upgrade", "torch", "torchvision"])
+
+            install_pkg_if_necessary("sdkit", expected_sdkit_version_str)
+            install_pkg_if_necessary("diffusers", expected_diffusers_version_str)
+
+    # hotfix accelerate
+    accelerate_version = version("accelerate")
+    if accelerate_version is None:
+        install("accelerate", "0.23.0")
+    else:
+        accelerate_version = accelerate_version.split(".")
+        accelerate_version = tuple(map(int, accelerate_version))
+        if accelerate_version < (0, 23):
+            install("accelerate", "0.23.0")
+
+    # hotfix - 29 May 2024. sdkit has stopped pulling its dependencies for some reason
+    # temporarily dumping sdkit's requirements here:
+    if os_name != "Windows":
+        sdkit_deps = [
+            "gfpgan",
+            "piexif",
+            "realesrgan",
+            "requests",
+            "picklescan",
+            "safetensors==0.3.3",
+            "k-diffusion==0.0.12",
+            "compel==2.0.1",
+            "controlnet-aux==0.0.6",
+            "invisible-watermark==0.2.0",  # required for SD XL
+        ]
+
+        for mod in sdkit_deps:
+            mod_name = mod
+            mod_force_version_str = None
+            if "==" in mod:
+                mod_name, mod_force_version_str = mod.split("==")
+
+            curr_mod_version_str = version(mod_name)
+            if curr_mod_version_str is None:
+                _install(mod_name, mod_force_version_str)
+            elif mod_force_version_str is not None:
+                curr_mod_version = version_str_to_tuple(curr_mod_version_str)
+                mod_force_version = version_str_to_tuple(mod_force_version_str)
+
+                if curr_mod_version != mod_force_version:
+                    _install(mod_name, mod_force_version_str)
+
+    for module_name in modules_to_log:
+        print(f"{module_name}: {version(module_name)}")
+
+
+def _install(module_name, module_version=None):
+    if module_version is None:
+        install_cmd = f'"{sys.executable}" -m pip install {module_name}'
+    else:
+        install_cmd = f'"{sys.executable}" -m pip install --upgrade {module_name}=={module_version}'
+
+    print(">", install_cmd)
+    os.system(install_cmd)
+
+
+def install_pkg_if_necessary(pkg_name, required_version):
+    if os.path.exists(f"src/{pkg_name}"):
+        print(f"Skipping {pkg_name} update, since it's in developer/editable mode")
+        return
+
+    pkg_version = version(pkg_name)
+    if pkg_version != required_version:
+        _install(pkg_name, required_version)
+
+
+def version_str_to_tuple(ver_str):
+    ver_str = ver_str.split("+")[0]
+    ver_str = re.sub("[^0-9.]", "", ver_str)
+    ver = ver_str.split(".")
+    return tuple(map(int, ver))
 
 
 ### utilities
-
-
 def get_allowed_versions(module_name: str, allowed_versions: tuple):
     allowed_versions = (allowed_versions,) if isinstance(allowed_versions, str) else allowed_versions
     latest_version = allowed_versions[-1]
 
-    if module_name in ("torch", "torchvision"):
-        allowed_versions = include_cuda_versions(allowed_versions)
-
     return allowed_versions, latest_version
 
 
-def apply_torch_install_overrides(module_version: str):
-    index_url = None
-    if os_name == "Windows":
-        module_version += "+cu117"
-        index_url = "https://download.pytorch.org/whl/cu117"
-    elif is_amd_on_linux():
-        index_url = "https://download.pytorch.org/whl/rocm5.2"
-
-    return module_version, index_url
-
-
-def include_cuda_versions(module_versions: tuple) -> tuple:
-    "Adds CUDA-specific versions to the list of allowed version numbers"
-
-    allowed_versions = tuple(module_versions)
-    allowed_versions += tuple(f"{v}+cu116" for v in module_versions)
-    allowed_versions += tuple(f"{v}+cu117" for v in module_versions)
-    allowed_versions += tuple(f"{v}+rocm5.2" for v in module_versions)
-    allowed_versions += tuple(f"{v}+rocm5.4.2" for v in module_versions)
-
-    return allowed_versions
-
-
-def is_amd_on_linux():
-    if os_name == "Linux":
-        try:
-            with open("/proc/bus/pci/devices", "r") as f:
-                device_info = f.read()
-                if "amdgpu" in device_info and "nvidia" not in device_info:
-                    return True
-        except:
-            return False
-
-    return False
-
-
 def fail(module_name):
     print(
         f"""Error installing {module_name}. Sorry about that, please try to:
@@ -160,6 +249,100 @@ Thanks!"""
     exit(1)
 
 
-### start
+### Launcher
 
-init()
+
+def get_config():
+    config_directory = os.path.dirname(__file__)  # this will be "scripts"
+    config_yaml = os.path.abspath(os.path.join(config_directory, "..", "config.yaml"))
+    config_json = os.path.join(config_directory, "config.json")
+
+    config = None
+
+    # migrate the old config yaml location
+    config_legacy_yaml = os.path.join(config_directory, "config.yaml")
+    if os.path.isfile(config_legacy_yaml):
+        shutil.move(config_legacy_yaml, config_yaml)
+
+    if os.path.isfile(config_yaml):
+        from ruamel.yaml import YAML
+
+        yaml = YAML(typ="safe")
+        with open(config_yaml, "r") as configfile:
+            try:
+                config = yaml.load(configfile)
+            except Exception as e:
+                print(e, file=sys.stderr)
+    elif os.path.isfile(config_json):
+        import json
+
+        with open(config_json, "r") as configfile:
+            try:
+                config = json.load(configfile)
+            except Exception as e:
+                print(e, file=sys.stderr)
+
+    if config is None:
+        config = {}
+    return config
+
+
+def launch_uvicorn():
+    config = get_config()
+
+    pprint(config)
+
+    with open("scripts/install_status.txt", "a") as f:
+        f.write("sd_weights_downloaded\n")
+        f.write("sd_install_complete\n")
+
+    print("\n\nEasy Diffusion installation complete, starting the server!\n\n")
+
+    torchruntime.configure()
+    if hasattr(torchruntime, "info"):
+        torchruntime.info()
+
+    if os_name == "Windows":
+        os.environ["PYTHONPATH"] = str(Path(os.environ["INSTALL_ENV_DIR"], "lib", "site-packages"))
+    else:
+        os.environ["PYTHONPATH"] = str(Path(os.environ["INSTALL_ENV_DIR"], "lib", "python3.8", "site-packages"))
+    os.environ["SD_UI_PATH"] = str(Path(Path.cwd(), "ui"))
+
+    print(f"PYTHONPATH={os.environ['PYTHONPATH']}")
+    print(f"Python:  {shutil.which('python')}")
+    print(f"Version: {platform. python_version()}")
+
+    bind_ip = "127.0.0.1"
+    listen_port = 9000
+    if "net" in config:
+        print("Checking network settings")
+        if "listen_port" in config["net"]:
+            listen_port = config["net"]["listen_port"]
+            print("Set listen port to ", listen_port)
+        if "listen_to_network" in config["net"] and config["net"]["listen_to_network"] == True:
+            if "bind_ip" in config["net"]:
+                bind_ip = config["net"]["bind_ip"]
+            else:
+                bind_ip = "0.0.0.0"
+            print("Set bind_ip to ", bind_ip)
+
+    os.chdir("stable-diffusion")
+
+    print("\nLaunching uvicorn\n")
+
+    import uvicorn
+
+    uvicorn.run(
+        "main:server_api",
+        port=listen_port,
+        log_level="error",
+        app_dir=os.environ["SD_UI_PATH"],
+        host=bind_ip,
+        access_log=False,
+    )
+
+
+update_modules()
+
+if len(sys.argv) > 1 and sys.argv[1] == "--launch-uvicorn":
+    launch_uvicorn()

scripts/get_config.py

@@ -6,6 +6,7 @@ import shutil
 # The config file is in the same directory as this script
 config_directory = os.path.dirname(__file__)
 config_yaml = os.path.join(config_directory, "..", "config.yaml")
+config_yaml = os.path.abspath(config_yaml)
 config_json = os.path.join(config_directory, "config.json")
 
 parser = argparse.ArgumentParser(description='Get values from config file')

scripts/on_env_start.bat

@@ -26,19 +26,19 @@ if "%update_branch%"=="" (
     set update_branch=main
 )
 
-@>nul findstr /m "conda_sd_ui_deps_installed" scripts\install_status.txt
-@if "%ERRORLEVEL%" NEQ "0" (
-    for /f "tokens=*" %%a in ('python -c "import os; parts = os.getcwd().split(os.path.sep); print(len(parts))"') do if "%%a" NEQ "2" (
-        echo. & echo "!!!! WARNING !!!!" & echo.
-        echo "Your 'stable-diffusion-ui' folder is at %cd%" & echo.
-        echo "The 'stable-diffusion-ui' folder needs to be at the top of your drive, for e.g. 'C:\stable-diffusion-ui' or 'D:\stable-diffusion-ui' etc."
-        echo "Not placing this folder at the top of a drive can cause errors on some computers."
-        echo. & echo "Recommended: Please close this window and move the 'stable-diffusion-ui' folder to the top of a drive. For e.g. 'C:\stable-diffusion-ui'. Then run the installer again." & echo.
-        echo "Not Recommended: If you're sure that you want to install at the current location, please press any key to continue." & echo.
+@REM @>nul findstr /m "sd_install_complete" scripts\install_status.txt
+@REM @if "%ERRORLEVEL%" NEQ "0" (
+@REM     for /f "tokens=*" %%a in ('python -c "import os; parts = os.getcwd().split(os.path.sep); print(len(parts))"') do if "%%a" NEQ "2" (
+@REM         echo. & echo "!!!! WARNING !!!!" & echo.
+@REM         echo "Your 'stable-diffusion-ui' folder is at %cd%" & echo.
+@REM         echo "The 'stable-diffusion-ui' folder needs to be at the top of your drive, for e.g. 'C:\stable-diffusion-ui' or 'D:\stable-diffusion-ui' etc."
+@REM         echo "Not placing this folder at the top of a drive can cause errors on some computers."
+@REM         echo. & echo "Recommended: Please close this window and move the 'stable-diffusion-ui' folder to the top of a drive. For e.g. 'C:\stable-diffusion-ui'. Then run the installer again." & echo.
+@REM         echo "Not Recommended: If you're sure that you want to install at the current location, please press any key to continue." & echo.
 
-        pause
-    )
-)
+@REM         pause
+@REM     )
+@REM )
 
 @>nul findstr /m "sd_ui_git_cloned" scripts\install_status.txt
 @if "%ERRORLEVEL%" EQU "0" (
@@ -71,6 +71,7 @@ if "%update_branch%"=="" (
 @copy sd-ui-files\scripts\check_modules.py scripts\ /Y
 @copy sd-ui-files\scripts\get_config.py scripts\ /Y
 @copy sd-ui-files\scripts\config.yaml.sample scripts\ /Y
+@copy sd-ui-files\scripts\webui_console.py scripts\ /Y
 @copy "sd-ui-files\scripts\Start Stable Diffusion UI.cmd" . /Y
 @copy "sd-ui-files\scripts\Developer Console.cmd" . /Y
 

scripts/on_env_start.sh

@@ -54,6 +54,7 @@ cp sd-ui-files/scripts/bootstrap.sh scripts/
 cp sd-ui-files/scripts/check_modules.py scripts/
 cp sd-ui-files/scripts/get_config.py scripts/
 cp sd-ui-files/scripts/config.yaml.sample scripts/
+cp sd-ui-files/scripts/webui_console.py scripts/
 cp sd-ui-files/scripts/start.sh .
 cp sd-ui-files/scripts/developer_console.sh .
 cp sd-ui-files/scripts/functions.sh scripts/

scripts/on_sd_start.bat

@@ -7,6 +7,7 @@
 @copy sd-ui-files\scripts\check_modules.py scripts\ /Y
 @copy sd-ui-files\scripts\get_config.py scripts\ /Y
 @copy sd-ui-files\scripts\config.yaml.sample scripts\ /Y
+@copy sd-ui-files\scripts\webui_console.py scripts\ /Y
 
 if exist "%cd%\profile" (
     set HF_HOME=%cd%\profile\.cache\huggingface
@@ -34,6 +35,7 @@ call conda activate
 
 @REM remove the old version of the dev console script, if it's still present
 if exist "Open Developer Console.cmd" del "Open Developer Console.cmd"
+if exist "ui\plugins\ui\merge.plugin.js" del "ui\plugins\ui\merge.plugin.js"
 
 @rem create the stable-diffusion folder, to work with legacy installations
 if not exist "stable-diffusion" mkdir stable-diffusion
@@ -52,73 +54,30 @@ if exist ldm rename ldm ldm-old
 if not exist "%INSTALL_ENV_DIR%\DLLs\libssl-1_1-x64.dll"    copy "%INSTALL_ENV_DIR%\Library\bin\libssl-1_1-x64.dll"    "%INSTALL_ENV_DIR%\DLLs\"
 if not exist "%INSTALL_ENV_DIR%\DLLs\libcrypto-1_1-x64.dll" copy "%INSTALL_ENV_DIR%\Library\bin\libcrypto-1_1-x64.dll" "%INSTALL_ENV_DIR%\DLLs\"
 
+cd ..
+
+@rem set any overrides
+set HF_HUB_DISABLE_SYMLINKS_WARNING=true
+
 @rem install or upgrade the required modules
 set PATH=C:\Windows\System32;%PATH%
 
 @REM prevent from using packages from the user's home directory, to avoid conflicts
 set PYTHONNOUSERSITE=1
-set PYTHONPATH=%INSTALL_ENV_DIR%\lib\site-packages
-
-@rem Download the required packages
-call python ..\scripts\check_modules.py
-if "%ERRORLEVEL%" NEQ "0" (
-    pause
-    exit /b
-)
-
-call WHERE uvicorn > .tmp
-@>nul findstr /m "uvicorn" .tmp
-@if "%ERRORLEVEL%" NEQ "0" (
-    @echo. & echo "UI packages not found! Sorry about that, please try to:" & echo "  1. Run this installer again." & echo "  2. If that doesn't fix it, please try the common troubleshooting steps at https://github.com/easydiffusion/easydiffusion/wiki/Troubleshooting" & echo "  3. If those steps don't help, please copy *all* the error messages in this window, and ask the community at https://discord.com/invite/u9yhsFmEkB" & echo "  4. If that doesn't solve the problem, please file an issue at https://github.com/easydiffusion/easydiffusion/issues" & echo "Thanks!" & echo.
-    pause
-    exit /b
-)
-
-@>nul findstr /m "conda_sd_ui_deps_installed" ..\scripts\install_status.txt
-@if "%ERRORLEVEL%" NEQ "0" (
-    @echo conda_sd_ui_deps_installed >> ..\scripts\install_status.txt
-)
-
-@>nul findstr /m "sd_install_complete" ..\scripts\install_status.txt
-@if "%ERRORLEVEL%" NEQ "0" (
-    @echo sd_weights_downloaded >> ..\scripts\install_status.txt
-    @echo sd_install_complete >> ..\scripts\install_status.txt
-)
-
-@echo. & echo "Easy Diffusion installation complete! Starting the server!" & echo.
-
-@set SD_DIR=%cd%
-
 set PYTHONPATH=%INSTALL_ENV_DIR%\lib\site-packages
 echo PYTHONPATH=%PYTHONPATH%
 
-call where python
-call python --version
+set PYTHON=%INSTALL_ENV_DIR%\python.exe
+echo PYTHON=%PYTHON%
 
-@cd ..
-@set SD_UI_PATH=%cd%\ui
+@rem Download the required packages
+@REM call where python
+call "%PYTHON%" --version
 
-@FOR /F "tokens=* USEBACKQ" %%F IN (`python scripts\get_config.py --default=9000 net listen_port`) DO (
-    @SET ED_BIND_PORT=%%F
-)
+@rem this is outside check_modules.py to ensure that the required version of torchruntime is present
+call "%PYTHON%" -m pip install -q "torchruntime>=1.19.1"
 
-@FOR /F "tokens=* USEBACKQ" %%F IN (`python scripts\get_config.py --default=False net listen_to_network`) DO (
-    if "%%F" EQU "True" (
-        @FOR /F "tokens=* USEBACKQ" %%G IN (`python scripts\get_config.py --default=0.0.0.0 net bind_ip`) DO (
-            @SET ED_BIND_IP=%%G
-        )
-    ) else (
-        @SET ED_BIND_IP=127.0.0.1
-    )
-)
+call "%PYTHON%" scripts\check_modules.py --launch-uvicorn
+pause
+exit /b
 
-
-@cd stable-diffusion
-
-@rem set any overrides
-set HF_HUB_DISABLE_SYMLINKS_WARNING=true
-
-@python -m uvicorn main:server_api --app-dir "%SD_UI_PATH%" --port %ED_BIND_PORT% --host %ED_BIND_IP% --log-level error
-
-
-@pause

scripts/on_sd_start.sh

@@ -6,20 +6,29 @@ cp sd-ui-files/scripts/bootstrap.sh scripts/
 cp sd-ui-files/scripts/check_modules.py scripts/
 cp sd-ui-files/scripts/get_config.py scripts/
 cp sd-ui-files/scripts/config.yaml.sample scripts/
+cp sd-ui-files/scripts/webui_console.py scripts/
+     
 
 source ./scripts/functions.sh
 
 # activate the installer env
-CONDA_BASEPATH=$(conda info --base)
+export CONDA_BASEPATH=$(conda info --base)
 source "$CONDA_BASEPATH/etc/profile.d/conda.sh" # avoids the 'shell not initialized' error
 
 conda activate || fail "Failed to activate conda"
 
+# hack to fix conda 4.14 on older installations
+cp $CONDA_BASEPATH/condabin/conda $CONDA_BASEPATH/bin/conda
+
 # remove the old version of the dev console script, if it's still present
 if [ -e "open_dev_console.sh" ]; then
     rm "open_dev_console.sh"
 fi
 
+if [ -e "ui/plugins/ui/merge.plugin.js" ]; then
+    rm "ui/plugins/ui/merge.plugin.js"
+fi
+
 # set the correct installer path (current vs legacy)
 if [ -e "installer_files/env" ]; then
     export INSTALL_ENV_DIR="$(pwd)/installer_files/env"
@@ -41,45 +50,11 @@ fi
 if [ -e "src" ]; then mv src src-old; fi
 if [ -e "ldm" ]; then mv ldm ldm-old; fi
 
-# Download the required packages
-if ! python ../scripts/check_modules.py; then
-    read -p "Press any key to continue"
-    exit 1
-fi
-
-if ! command -v uvicorn &> /dev/null; then
-    fail "UI packages not found!"
-fi
-
-if [ `grep -c sd_install_complete ../scripts/install_status.txt` -gt "0" ]; then
-    echo sd_weights_downloaded >> ../scripts/install_status.txt
-    echo sd_install_complete >> ../scripts/install_status.txt
-fi
-
-printf "\n\nEasy Diffusion installation complete, starting the server!\n\n"
-
-SD_PATH=`pwd`
-
-export PYTORCH_ENABLE_MPS_FALLBACK=1
-export PYTHONPATH="$INSTALL_ENV_DIR/lib/python3.8/site-packages"
-echo "PYTHONPATH=$PYTHONPATH"
-
-which python
-python --version
+# this is outside check_modules.py to ensure that the required version of torchruntime is present
+python -m pip install -q "torchruntime>=1.19.1"
 
 cd ..
-export SD_UI_PATH=`pwd`/ui
-export ED_BIND_PORT="$( python scripts/get_config.py --default=9000 net listen_port )"
-case "$( python scripts/get_config.py --default=False net listen_to_network )" in
-    "True")
-        export ED_BIND_IP=$( python scripts/get_config.py --default=0.0.0.0 net bind_ip)
-        ;;
-    "False")
-        export ED_BIND_IP=127.0.0.1
-        ;;
-esac
-cd stable-diffusion
-
-uvicorn main:server_api --app-dir "$SD_UI_PATH" --port "$ED_BIND_PORT" --host "$ED_BIND_IP" --log-level error
+# Download the required packages
+python scripts/check_modules.py --launch-uvicorn
 
 read -p "Press any key to continue"

scripts/start.sh

@@ -11,7 +11,7 @@ if [ -f "on_sd_start.bat" ]; then
     echo download. This will not work.
     echo
     echo Recommended: Please close this window and download the installer from
-    echo https://stable-diffusion-ui.github.io/docs/installation/
+    echo https://easydiffusion.github.io/docs/installation/
     echo
     echo ================================================================================
     echo

scripts/webui_console.py

@@ -0,0 +1,101 @@
+import os
+import platform
+import subprocess
+
+
+def configure_env(dir):
+    env_entries = {
+        "PATH": [
+            f"{dir}",
+            f"{dir}/bin",
+            f"{dir}/Library/bin",
+            f"{dir}/Scripts",
+            f"{dir}/usr/bin",
+        ],
+        "PYTHONPATH": [
+            f"{dir}",
+            f"{dir}/lib/site-packages",
+            f"{dir}/lib/python3.10/site-packages",
+        ],
+        "PYTHONHOME": [],
+        "PY_LIBS": [
+            f"{dir}/Scripts/Lib",
+            f"{dir}/Scripts/Lib/site-packages",
+            f"{dir}/lib",
+            f"{dir}/lib/python3.10/site-packages",
+        ],
+        "PY_PIP": [f"{dir}/Scripts", f"{dir}/bin"],
+    }
+
+    if platform.system() == "Windows":
+        env_entries["PATH"].append("C:/Windows/System32")
+        env_entries["PATH"].append("C:/Windows/System32/wbem")
+        env_entries["PYTHONNOUSERSITE"] = ["1"]
+        env_entries["PYTHON"] = [f"{dir}/python"]
+        env_entries["GIT"] = [f"{dir}/Library/bin/git"]
+    else:
+        env_entries["PATH"].append("/bin")
+        env_entries["PATH"].append("/usr/bin")
+        env_entries["PATH"].append("/usr/sbin")
+        env_entries["PYTHONNOUSERSITE"] = ["y"]
+        env_entries["PYTHON"] = [f"{dir}/bin/python"]
+        env_entries["GIT"] = [f"{dir}/bin/git"]
+
+    env = {}
+    for key, paths in env_entries.items():
+        paths = [p.replace("/", os.path.sep) for p in paths]
+        paths = os.pathsep.join(paths)
+
+        os.environ[key] = paths
+
+    return env
+
+
+def print_env_info():
+    which_cmd = "where" if platform.system() == "Windows" else "which"
+
+    python = "python"
+
+    def locate_python():
+        nonlocal python
+
+        python = subprocess.getoutput(f"{which_cmd} python")
+        python = python.split("\n")
+        python = python[0].strip()
+        print("python: ", python)
+
+    locate_python()
+
+    def run(cmd):
+        with subprocess.Popen(cmd) as p:
+            p.wait()
+
+    run([which_cmd, "git"])
+    run(["git", "--version"])
+    run([which_cmd, "python"])
+    run([python, "--version"])
+
+    print(f"PATH={os.environ['PATH']}")
+
+    # if platform.system() == "Windows":
+    #     print(f"COMSPEC={os.environ['COMSPEC']}")
+    #     print("")
+    #     run("wmic path win32_VideoController get name,AdapterRAM,DriverDate,DriverVersion".split(" "))
+
+    print(f"PYTHONPATH={os.environ['PYTHONPATH']}")
+    print("")
+
+
+def open_dev_shell():
+    if platform.system() == "Windows":
+        subprocess.Popen("cmd").communicate()
+    else:
+        subprocess.Popen("bash").communicate()
+
+
+if __name__ == "__main__":
+    env_dir = os.path.abspath(os.path.join("webui", "system"))
+
+    configure_env(env_dir)
+    print_env_info()
+    open_dev_shell()

ui/easydiffusion/app.py

@@ -11,7 +11,7 @@ from ruamel.yaml import YAML
 import urllib
 import warnings
 
-from easydiffusion import task_manager
+from easydiffusion import task_manager, backend_manager
 from easydiffusion.utils import log
 from rich.logging import RichHandler
 from rich.console import Console
@@ -36,11 +36,10 @@ ROOT_DIR = os.path.abspath(os.path.join(SD_DIR, ".."))
 
 SD_UI_DIR = os.getenv("SD_UI_PATH", None)
 
-CONFIG_DIR = os.path.abspath(os.path.join(SD_UI_DIR, "..", "scripts"))
-MODELS_DIR = os.path.abspath(os.path.join(SD_DIR, "..", "models"))
-BUCKET_DIR = os.path.abspath(os.path.join(SD_DIR, "..", "bucket"))
+CONFIG_DIR = os.path.abspath(os.path.join(ROOT_DIR, "scripts"))
+BUCKET_DIR = os.path.abspath(os.path.join(ROOT_DIR, "bucket"))
 
-USER_PLUGINS_DIR = os.path.abspath(os.path.join(SD_DIR, "..", "plugins"))
+USER_PLUGINS_DIR = os.path.abspath(os.path.join(ROOT_DIR, "plugins"))
 CORE_PLUGINS_DIR = os.path.abspath(os.path.join(SD_UI_DIR, "plugins"))
 
 USER_UI_PLUGINS_DIR = os.path.join(USER_PLUGINS_DIR, "ui")
@@ -55,13 +54,12 @@ OUTPUT_DIRNAME = "Stable Diffusion UI"  # in the user's home folder
 PRESERVE_CONFIG_VARS = ["FORCE_FULL_PRECISION"]
 TASK_TTL = 15 * 60  # Discard last session's task timeout
 APP_CONFIG_DEFAULTS = {
-    # auto: selects the cuda device with the most free memory, cuda: use the currently active cuda device.
-    "render_devices": "auto",  # valid entries: 'auto', 'cpu' or 'cuda:N' (where N is a GPU index)
+    "render_devices": "auto",
     "update_branch": "main",
     "ui": {
         "open_browser_on_start": True,
     },
-    "test_diffusers": True,
+    "backend": "ed_diffusers",
 }
 
 IMAGE_EXTENSIONS = [
@@ -78,7 +76,7 @@ IMAGE_EXTENSIONS = [
     ".avif",
     ".svg",
 ]
-CUSTOM_MODIFIERS_DIR = os.path.abspath(os.path.join(SD_DIR, "..", "modifiers"))
+CUSTOM_MODIFIERS_DIR = os.path.abspath(os.path.join(ROOT_DIR, "modifiers"))
 CUSTOM_MODIFIERS_PORTRAIT_EXTENSIONS = [
     ".portrait",
     "_portrait",
@@ -92,14 +90,25 @@ CUSTOM_MODIFIERS_LANDSCAPE_EXTENSIONS = [
     "-landscape",
 ]
 
+MODELS_DIR = os.path.abspath(os.path.join(ROOT_DIR, "models"))
+
 
 def init():
+    global MODELS_DIR
+
     os.makedirs(USER_UI_PLUGINS_DIR, exist_ok=True)
     os.makedirs(USER_SERVER_PLUGINS_DIR, exist_ok=True)
 
     # https://pytorch.org/docs/stable/storage.html
     warnings.filterwarnings("ignore", category=UserWarning, message="TypedStorage is deprecated")
 
+    config = getConfig()
+    config_models_dir = config.get("models_dir", None)
+    if config_models_dir is not None and config_models_dir != "":
+        MODELS_DIR = config_models_dir
+
+    backend_manager.start_backend()
+
 
 def init_render_threads():
     load_server_plugins()
@@ -109,6 +118,7 @@ def init_render_threads():
 
 def getConfig(default_val=APP_CONFIG_DEFAULTS):
     config_yaml_path = os.path.join(CONFIG_DIR, "..", "config.yaml")
+    config_yaml_path = os.path.abspath(config_yaml_path)
 
     # migrate the old config yaml location
     config_legacy_yaml = os.path.join(CONFIG_DIR, "config.yaml")
@@ -116,9 +126,9 @@ def getConfig(default_val=APP_CONFIG_DEFAULTS):
         shutil.move(config_legacy_yaml, config_yaml_path)
 
     def set_config_on_startup(config: dict):
-        if getConfig.__test_diffusers_on_startup is None:
-            getConfig.__test_diffusers_on_startup = config.get("test_diffusers", True)
-        config["config_on_startup"] = {"test_diffusers": getConfig.__test_diffusers_on_startup}
+        if getConfig.__use_backend_on_startup is None:
+            getConfig.__use_backend_on_startup = config.get("backend", "ed_diffusers")
+        config["config_on_startup"] = {"backend": getConfig.__use_backend_on_startup}
 
     if os.path.isfile(config_yaml_path):
         try:
@@ -136,6 +146,15 @@ def getConfig(default_val=APP_CONFIG_DEFAULTS):
                 else:
                     config["net"]["listen_to_network"] = True
 
+            if "backend" not in config:
+                if "use_v3_engine" in config:
+                    config["backend"] = "ed_diffusers" if config["use_v3_engine"] else "ed_classic"
+                else:
+                    config["backend"] = "ed_diffusers"
+                    # this default will need to be smarter when WebUI becomes the main backend, but needs to maintain backwards
+                    # compatibility with existing ED 3.0 installations that haven't opted into the WebUI backend, and haven't
+                    # set a "use_v3_engine" flag in their config
+
             set_config_on_startup(config)
 
             return config
@@ -166,12 +185,15 @@ def getConfig(default_val=APP_CONFIG_DEFAULTS):
             return default_val
 
 
-getConfig.__test_diffusers_on_startup = None
+getConfig.__use_backend_on_startup = None
 
 
 def setConfig(config):
+    global MODELS_DIR
+
     try:  # config.yaml
         config_yaml_path = os.path.join(CONFIG_DIR, "..", "config.yaml")
+        config_yaml_path = os.path.abspath(config_yaml_path)
         yaml = YAML()
 
         if not hasattr(config, "_yaml_comment"):
@@ -205,6 +227,9 @@ def setConfig(config):
     except:
         log.error(traceback.format_exc())
 
+    if config.get("models_dir"):
+        MODELS_DIR = config["models_dir"]
+
 
 def save_to_config(ckpt_model_name, vae_model_name, hypernetwork_model_name, vram_usage_level):
     config = getConfig()
@@ -293,28 +318,43 @@ def getIPConfig():
 
 
 def open_browser():
+    from easydiffusion.backend_manager import backend
+
     config = getConfig()
     ui = config.get("ui", {})
     net = config.get("net", {})
     port = net.get("listen_port", 9000)
 
-    if ui.get("open_browser_on_start", True):
-        import webbrowser
+    if backend.is_installed():
+        if ui.get("open_browser_on_start", True):
+            import webbrowser
 
-        log.info("Opening browser..")
+            log.info("Opening browser..")
 
-        webbrowser.open(f"http://localhost:{port}")
+            webbrowser.open(f"http://localhost:{port}")
 
-    Console().print(
-        Panel(
-            "\n"
-            + "[white]Easy Diffusion is ready to serve requests.\n\n"
-            + "A new browser tab should have been opened by now.\n"
-            + f"If not, please open your web browser and navigate to [bold yellow underline]http://localhost:{port}/\n",
-            title="Easy Diffusion is ready",
-            style="bold yellow on blue",
+        Console().print(
+            Panel(
+                "\n"
+                + "[white]Easy Diffusion is ready to serve requests.\n\n"
+                + "A new browser tab should have been opened by now.\n"
+                + f"If not, please open your web browser and navigate to [bold yellow underline]http://localhost:{port}/\n",
+                title="Easy Diffusion is ready",
+                style="bold yellow on blue",
+            )
+        )
+    else:
+        backend_name = config["backend"]
+        Console().print(
+            Panel(
+                "\n"
+                + f"[white]Backend: {backend_name} is still installing..\n\n"
+                + "A new browser tab will open automatically after it finishes.\n"
+                + f"If it does not, please open your web browser and navigate to [bold yellow underline]http://localhost:{port}/\n",
+                title=f"Backend engine is installing",
+                style="bold yellow on blue",
+            )
         )
-    )
 
 
 def fail_and_die(fail_type: str, data: str):

ui/easydiffusion/backend_manager.py

@@ -0,0 +1,105 @@
+import os
+import ast
+import sys
+import importlib.util
+import traceback
+
+from easydiffusion.utils import log
+
+backend = None
+curr_backend_name = None
+
+
+def is_valid_backend(file_path):
+    with open(file_path, "r", encoding="utf-8") as file:
+        node = ast.parse(file.read())
+
+    # Check for presence of a dictionary named 'ed_info'
+    for item in node.body:
+        if isinstance(item, ast.Assign):
+            for target in item.targets:
+                if isinstance(target, ast.Name) and target.id == "ed_info":
+                    return True
+    return False
+
+
+def find_valid_backends(root_dir) -> dict:
+    backends_path = os.path.join(root_dir, "backends")
+    valid_backends = {}
+
+    if not os.path.exists(backends_path):
+        return valid_backends
+
+    for item in os.listdir(backends_path):
+        item_path = os.path.join(backends_path, item)
+
+        if os.path.isdir(item_path):
+            init_file = os.path.join(item_path, "__init__.py")
+            if os.path.exists(init_file) and is_valid_backend(init_file):
+                valid_backends[item] = item_path
+        elif item.endswith(".py"):
+            if is_valid_backend(item_path):
+                backend_name = os.path.splitext(item)[0]  # strip the .py extension
+                valid_backends[backend_name] = item_path
+
+    return valid_backends
+
+
+def load_backend_module(backend_name, backend_dict):
+    if backend_name not in backend_dict:
+        raise ValueError(f"Backend '{backend_name}' not found.")
+
+    module_path = backend_dict[backend_name]
+
+    mod_dir = os.path.dirname(module_path)
+
+    sys.path.insert(0, mod_dir)
+
+    # If it's a package (directory), add its parent directory to sys.path
+    if os.path.isdir(module_path):
+        module_path = os.path.join(module_path, "__init__.py")
+
+    spec = importlib.util.spec_from_file_location(backend_name, module_path)
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+
+    if mod_dir in sys.path:
+        sys.path.remove(mod_dir)
+
+    log.info(f"Loaded backend: {module}")
+
+    return module
+
+
+def start_backend():
+    global backend, curr_backend_name
+
+    from easydiffusion.app import getConfig, ROOT_DIR
+
+    curr_dir = os.path.dirname(__file__)
+
+    backends = find_valid_backends(curr_dir)
+    plugin_backends = find_valid_backends(ROOT_DIR)
+    backends.update(plugin_backends)
+
+    config = getConfig()
+    backend_name = config["backend"]
+
+    if backend_name not in backends:
+        raise RuntimeError(
+            f"Couldn't find the backend configured in config.yaml: {backend_name}. Please check the name!"
+        )
+
+    if backend is not None and backend_name != curr_backend_name:
+        try:
+            backend.stop_backend()
+        except:
+            log.exception(traceback.format_exc())
+
+    log.info(f"Loading backend: {backend_name}")
+    backend = load_backend_module(backend_name, backends)
+
+    try:
+        backend.start_backend()
+    except:
+        log.exception(traceback.format_exc())

ui/easydiffusion/backends/ed_classic.py

@@ -0,0 +1,28 @@
+from sdkit_common import (
+    start_backend,
+    stop_backend,
+    install_backend,
+    uninstall_backend,
+    is_installed,
+    create_sdkit_context,
+    ping,
+    load_model,
+    unload_model,
+    set_options,
+    generate_images,
+    filter_images,
+    get_url,
+    stop_rendering,
+    refresh_models,
+    list_controlnet_filters,
+)
+
+ed_info = {
+    "name": "Classic backend for Easy Diffusion v2",
+    "version": (1, 0, 0),
+    "type": "backend",
+}
+
+
+def create_context():
+    return create_sdkit_context(use_diffusers=False)

ui/easydiffusion/backends/ed_diffusers.py

@@ -0,0 +1,28 @@
+from sdkit_common import (
+    start_backend,
+    stop_backend,
+    install_backend,
+    uninstall_backend,
+    is_installed,
+    create_sdkit_context,
+    ping,
+    load_model,
+    unload_model,
+    set_options,
+    generate_images,
+    filter_images,
+    get_url,
+    stop_rendering,
+    refresh_models,
+    list_controlnet_filters,
+)
+
+ed_info = {
+    "name": "Diffusers Backend for Easy Diffusion v3",
+    "version": (1, 0, 0),
+    "type": "backend",
+}
+
+
+def create_context():
+    return create_sdkit_context(use_diffusers=True)

ui/easydiffusion/backends/sdkit_common.py

@@ -0,0 +1,246 @@
+from sdkit import Context
+
+from easydiffusion.types import UserInitiatedStop
+
+from sdkit.utils import (
+    diffusers_latent_samples_to_images,
+    gc,
+    img_to_base64_str,
+    latent_samples_to_images,
+)
+
+opts = {}
+
+
+def install_backend():
+    pass
+
+
+def start_backend():
+    print("Started sdkit backend")
+
+
+def stop_backend():
+    pass
+
+
+def uninstall_backend():
+    pass
+
+
+def is_installed():
+    return True
+
+
+def create_sdkit_context(use_diffusers):
+    c = Context()
+    c.test_diffusers = use_diffusers
+    return c
+
+
+def ping(timeout=1):
+    return True
+
+
+def load_model(context, model_type, **kwargs):
+    from sdkit.models import load_model
+
+    load_model(context, model_type, **kwargs)
+
+
+def unload_model(context, model_type, **kwargs):
+    from sdkit.models import unload_model
+
+    unload_model(context, model_type, **kwargs)
+
+
+def set_options(context, **kwargs):
+    if "vae_tiling" in kwargs and context.test_diffusers:
+        pipe = context.models["stable-diffusion"]["default"]
+        vae_tiling = kwargs["vae_tiling"]
+
+        if vae_tiling:
+            if hasattr(pipe, "enable_vae_tiling"):
+                pipe.enable_vae_tiling()
+        else:
+            if hasattr(pipe, "disable_vae_tiling"):
+                pipe.disable_vae_tiling()
+
+    for key in (
+        "output_format",
+        "output_quality",
+        "output_lossless",
+        "stream_image_progress",
+        "stream_image_progress_interval",
+    ):
+        if key in kwargs:
+            opts[key] = kwargs[key]
+
+
+def generate_images(
+    context: Context,
+    callback=None,
+    controlnet_filter=None,
+    distilled_guidance_scale: float = 3.5,
+    scheduler_name: str = "simple",
+    output_type="pil",
+    **req,
+):
+    from sdkit.generate import generate_images
+
+    if req["init_image"] is not None and not context.test_diffusers:
+        req["sampler_name"] = "ddim"
+
+    gc(context)
+
+    context.stop_processing = False
+
+    if req["control_image"] and controlnet_filter:
+        controlnet_filter = convert_ED_controlnet_filter_name(controlnet_filter)
+        req["control_image"] = filter_images(context, req["control_image"], controlnet_filter)[0]
+
+    callback = make_step_callback(context, callback)
+
+    try:
+        images = generate_images(context, callback=callback, **req)
+    except UserInitiatedStop:
+        images = []
+        if context.partial_x_samples is not None:
+            if context.test_diffusers:
+                images = diffusers_latent_samples_to_images(context, context.partial_x_samples)
+            else:
+                images = latent_samples_to_images(context, context.partial_x_samples)
+    finally:
+        if hasattr(context, "partial_x_samples") and context.partial_x_samples is not None:
+            if not context.test_diffusers:
+                del context.partial_x_samples
+            context.partial_x_samples = None
+
+    gc(context)
+
+    if output_type == "base64":
+        output_format = opts.get("output_format", "jpeg")
+        output_quality = opts.get("output_quality", 75)
+        output_lossless = opts.get("output_lossless", False)
+        images = [img_to_base64_str(img, output_format, output_quality, output_lossless) for img in images]
+
+    return images
+
+
+def filter_images(context: Context, images, filters, filter_params={}, input_type="pil"):
+    gc(context)
+
+    if "nsfw_checker" in filters:
+        filters.remove("nsfw_checker")  # handled by ED directly
+
+    if len(filters) == 0:
+        return images
+
+    images = _filter_images(context, images, filters, filter_params)
+
+    if input_type == "base64":
+        output_format = opts.get("output_format", "jpg")
+        output_quality = opts.get("output_quality", 75)
+        output_lossless = opts.get("output_lossless", False)
+        images = [img_to_base64_str(img, output_format, output_quality, output_lossless) for img in images]
+
+    return images
+
+
+def _filter_images(context, images, filters, filter_params={}):
+    from sdkit.filter import apply_filters
+
+    filters = filters if isinstance(filters, list) else [filters]
+    filters = convert_ED_controlnet_filter_name(filters)
+
+    for filter_name in filters:
+        params = filter_params.get(filter_name, {})
+
+        previous_state = before_filter(context, filter_name, params)
+
+        try:
+            images = apply_filters(context, filter_name, images, **params)
+        finally:
+            after_filter(context, filter_name, params, previous_state)
+
+    return images
+
+
+def before_filter(context, filter_name, filter_params):
+    if filter_name == "codeformer":
+        from easydiffusion.model_manager import DEFAULT_MODELS, resolve_model_to_use
+
+        default_realesrgan = DEFAULT_MODELS["realesrgan"][0]["file_name"]
+        prev_realesrgan_path = None
+
+        upscale_faces = filter_params.get("upscale_faces", False)
+        if upscale_faces and default_realesrgan not in context.model_paths["realesrgan"]:
+            prev_realesrgan_path = context.model_paths.get("realesrgan")
+            context.model_paths["realesrgan"] = resolve_model_to_use(default_realesrgan, "realesrgan")
+            load_model(context, "realesrgan")
+
+        return prev_realesrgan_path
+
+
+def after_filter(context, filter_name, filter_params, previous_state):
+    if filter_name == "codeformer":
+        prev_realesrgan_path = previous_state
+        if prev_realesrgan_path:
+            context.model_paths["realesrgan"] = prev_realesrgan_path
+            load_model(context, "realesrgan")
+
+
+def get_url():
+    pass
+
+
+def stop_rendering(context):
+    context.stop_processing = True
+
+
+def refresh_models():
+    pass
+
+
+def list_controlnet_filters():
+    from sdkit.models.model_loader.controlnet_filters import filters as cn_filters
+
+    return cn_filters
+
+
+def make_step_callback(context, callback):
+    def on_step(x_samples, i, *args):
+        stream_image_progress = opts.get("stream_image_progress", False)
+        stream_image_progress_interval = opts.get("stream_image_progress_interval", 3)
+
+        if context.test_diffusers:
+            context.partial_x_samples = (x_samples, args[0])
+        else:
+            context.partial_x_samples = x_samples
+
+        if stream_image_progress and stream_image_progress_interval > 0 and i % stream_image_progress_interval == 0:
+            if context.test_diffusers:
+                images = diffusers_latent_samples_to_images(context, context.partial_x_samples)
+            else:
+                images = latent_samples_to_images(context, context.partial_x_samples)
+        else:
+            images = None
+
+        if callback:
+            callback(images, i, *args)
+
+        if context.stop_processing:
+            raise UserInitiatedStop("User requested that we stop processing")
+
+    return on_step
+
+
+def convert_ED_controlnet_filter_name(filter):
+    def cn(n):
+        if n.startswith("controlnet_"):
+            return n[len("controlnet_") :]
+        return n
+
+    if isinstance(filter, list):
+        return [cn(f) for f in filter]
+    return cn(filter)

ui/easydiffusion/backends/webui/__init__.py

@@ -0,0 +1,457 @@
+import os
+import platform
+import subprocess
+import threading
+from threading import local
+import psutil
+import time
+import shutil
+import atexit
+
+from easydiffusion.app import ROOT_DIR, getConfig
+from easydiffusion.model_manager import get_model_dirs
+from easydiffusion.utils import log
+from torchruntime.utils import get_device, get_device_name, get_installed_torch_platform
+from sdkit.utils import is_cpu_device
+
+from . import impl
+from .impl import (
+    ping,
+    load_model,
+    unload_model,
+    flush_model_changes,
+    set_options,
+    generate_images,
+    filter_images,
+    get_url,
+    stop_rendering,
+    refresh_models,
+    list_controlnet_filters,
+)
+
+
+ed_info = {
+    "name": "WebUI backend for Easy Diffusion",
+    "version": (1, 0, 0),
+    "type": "backend",
+}
+
+WEBUI_REPO = "https://github.com/lllyasviel/stable-diffusion-webui-forge.git"
+WEBUI_COMMIT = "dfdcbab685e57677014f05a3309b48cc87383167"
+
+BACKEND_DIR = os.path.abspath(os.path.join(ROOT_DIR, "webui"))
+SYSTEM_DIR = os.path.join(BACKEND_DIR, "system")
+WEBUI_DIR = os.path.join(BACKEND_DIR, "webui")
+
+OS_NAME = platform.system()
+
+MODELS_TO_OVERRIDE = {
+    "stable-diffusion": "--ckpt-dir",
+    "vae": "--vae-dir",
+    "hypernetwork": "--hypernetwork-dir",
+    "gfpgan": "--gfpgan-models-path",
+    "realesrgan": "--realesrgan-models-path",
+    "lora": "--lora-dir",
+    "codeformer": "--codeformer-models-path",
+    "embeddings": "--embeddings-dir",
+    "controlnet": "--controlnet-dir",
+    "text-encoder": "--text-encoder-dir",
+}
+
+WEBUI_PATCHES = [
+    "forge_exception_leak_patch.patch",
+    "forge_model_crash_recovery.patch",
+    "forge_api_refresh_text_encoders.patch",
+    "forge_loader_force_gc.patch",
+    "forge_monitor_parent_process.patch",
+    "forge_disable_corrupted_model_renaming.patch",
+]
+
+backend_process = None
+conda = "conda"
+
+
+def locate_conda():
+    global conda
+
+    which = "where" if OS_NAME == "Windows" else "which"
+    conda = subprocess.getoutput(f"{which} conda")
+    conda = conda.split("\n")
+    conda = conda[0].strip()
+    print("conda: ", conda)
+
+
+locate_conda()
+
+
+def install_backend():
+    print("Installing the WebUI backend..")
+
+    # create the conda env
+    run([conda, "create", "-y", "--prefix", SYSTEM_DIR], cwd=ROOT_DIR)
+
+    print("Installing packages..")
+
+    # install python 3.10 and git in the conda env
+    run([conda, "install", "-y", "--prefix", SYSTEM_DIR, "-c", "conda-forge", "python=3.10", "git"], cwd=ROOT_DIR)
+
+    env = dict(os.environ)
+    env.update(get_env())
+
+    # print info
+    run_in_conda(["git", "--version"], cwd=ROOT_DIR, env=env)
+    run_in_conda(["python", "--version"], cwd=ROOT_DIR, env=env)
+
+    # clone webui
+    run_in_conda(["git", "clone", WEBUI_REPO, WEBUI_DIR], cwd=ROOT_DIR, env=env)
+
+    # install the appropriate version of torch using torchruntime
+    run_in_conda(["python", "-m", "pip", "install", "torchruntime"], cwd=WEBUI_DIR, env=env)
+    run_in_conda(["python", "-m", "torchruntime", "install", "torch", "torchvision"], cwd=WEBUI_DIR, env=env)
+
+
+def start_backend():
+    config = getConfig()
+    backend_config = config.get("backend_config", {})
+
+    log.info(f"Expected WebUI backend dir: {BACKEND_DIR}")
+
+    if not os.path.exists(BACKEND_DIR):
+        install_backend()
+
+    env = dict(os.environ)
+    env.update(get_env())
+
+    was_still_installing = not is_installed()
+
+    if backend_config.get("auto_update", True):
+        run_in_conda(["git", "add", "-A", "."], cwd=WEBUI_DIR, env=env)
+        run_in_conda(["git", "stash"], cwd=WEBUI_DIR, env=env)
+        run_in_conda(["git", "reset", "--hard"], cwd=WEBUI_DIR, env=env)
+        run_in_conda(["git", "fetch"], cwd=WEBUI_DIR, env=env)
+        run_in_conda(["git", "-c", "advice.detachedHead=false", "checkout", WEBUI_COMMIT], cwd=WEBUI_DIR, env=env)
+
+        # patch forge for various stability-related fixes
+        for patch in WEBUI_PATCHES:
+            patch_path = os.path.join(os.path.dirname(__file__), patch)
+            log.info(f"Applying WebUI patch: {patch_path}")
+            run_in_conda(["git", "apply", patch_path], cwd=WEBUI_DIR, env=env)
+
+    # workaround for the installations that broke out of conda and used ED's python 3.8 instead of WebUI conda's Py 3.10
+    run_in_conda(["python", "-m", "pip", "install", "-q", "--upgrade", "urllib3==2.2.3"], cwd=WEBUI_DIR, env=env)
+
+    # hack to prevent webui-macos-env.sh from overwriting the COMMANDLINE_ARGS env variable
+    mac_webui_file = os.path.join(WEBUI_DIR, "webui-macos-env.sh")
+    if os.path.exists(mac_webui_file):
+        os.remove(mac_webui_file)
+
+    impl.WEBUI_HOST = backend_config.get("host", "localhost")
+    impl.WEBUI_PORT = backend_config.get("port", "7860")
+
+    def restart_if_webui_dies_after_starting():
+        has_started = False
+
+        while True:
+            try:
+                impl.ping(timeout=30)
+
+                is_first_start = not has_started
+                has_started = True
+
+                if was_still_installing and is_first_start:
+                    ui = config.get("ui", {})
+                    net = config.get("net", {})
+                    port = net.get("listen_port", 9000)
+
+                    if ui.get("open_browser_on_start", True):
+                        import webbrowser
+
+                        log.info("Opening browser..")
+
+                        webbrowser.open(f"http://localhost:{port}")
+            except (TimeoutError, ConnectionError):
+                if has_started:  # process probably died
+                    print("######################## WebUI probably died. Restarting...")
+                    stop_backend()
+                    backend_thread = threading.Thread(target=target)
+                    backend_thread.start()
+                    break
+            except Exception:
+                import traceback
+
+                log.exception(traceback.format_exc())
+
+            time.sleep(1)
+
+    def target():
+        global backend_process
+
+        cmd = "webui.bat" if OS_NAME == "Windows" else "./webui.sh"
+
+        log.info(f"starting: {cmd} in {WEBUI_DIR}")
+        log.info(f"COMMANDLINE_ARGS: {env['COMMANDLINE_ARGS']}")
+
+        backend_process = run_in_conda([cmd], cwd=WEBUI_DIR, env=env, wait=False, output_prefix="[WebUI] ")
+
+        # atexit.register isn't 100% reliable, that's why we also use `forge_monitor_parent_process.patch`
+        # which causes Forge to kill itself if the parent pid passed to it is no longer valid.
+        atexit.register(backend_process.terminate)
+
+        restart_if_dead_thread = threading.Thread(target=restart_if_webui_dies_after_starting)
+        restart_if_dead_thread.start()
+
+        backend_process.wait()
+
+    backend_thread = threading.Thread(target=target)
+    backend_thread.start()
+
+    start_proxy()
+
+
+def start_proxy():
+    # proxy
+    from easydiffusion.server import server_api
+    from fastapi import FastAPI, Request
+    from fastapi.responses import Response
+    import json
+
+    URI_PREFIX = "/webui"
+
+    webui_proxy = FastAPI(root_path=f"{URI_PREFIX}", docs_url="/swagger")
+
+    @webui_proxy.get("{uri:path}")
+    def proxy_get(uri: str, req: Request):
+        if uri == "/openapi-proxy.json":
+            uri = "/openapi.json"
+
+        res = impl.webui_get(uri, headers=req.headers)
+
+        content = res.content
+        headers = dict(res.headers)
+
+        if uri == "/docs":
+            content = res.text.replace("url: '/openapi.json'", f"url: '{URI_PREFIX}/openapi-proxy.json'")
+        elif uri == "/openapi.json":
+            content = res.json()
+            content["paths"] = {f"{URI_PREFIX}{k}": v for k, v in content["paths"].items()}
+            content = json.dumps(content)
+
+        if isinstance(content, str):
+            content = bytes(content, encoding="utf-8")
+            headers["content-length"] = str(len(content))
+
+        # Return the same response back to the client
+        return Response(content=content, status_code=res.status_code, headers=headers)
+
+    @webui_proxy.post("{uri:path}")
+    async def proxy_post(uri: str, req: Request):
+        body = await req.body()
+        res = impl.webui_post(uri, data=body, headers=req.headers)
+
+        # Return the same response back to the client
+        return Response(content=res.content, status_code=res.status_code, headers=dict(res.headers))
+
+    server_api.mount(f"{URI_PREFIX}", webui_proxy)
+
+
+def stop_backend():
+    global backend_process
+
+    if backend_process:
+        try:
+            kill(backend_process.pid)
+        except:
+            pass
+
+    backend_process = None
+
+
+def uninstall_backend():
+    shutil.rmtree(BACKEND_DIR)
+
+
+def is_installed():
+    if not os.path.exists(BACKEND_DIR) or not os.path.exists(SYSTEM_DIR) or not os.path.exists(WEBUI_DIR):
+        return True
+
+    env = dict(os.environ)
+    env.update(get_env())
+
+    try:
+        out = check_output_in_conda(["python", "-m", "pip", "show", "torch"], env=env)
+        return "Version" in out.decode()
+    except subprocess.CalledProcessError:
+        pass
+
+    return False
+
+
+def read_output(pipe, prefix=""):
+    while True:
+        output = pipe.readline()
+        if output:
+            print(f"{prefix}{output.decode('utf-8')}", end="")
+        else:
+            break  # Pipe is closed, subprocess has likely exited
+
+
+def run(cmds: list, cwd=None, env=None, stream_output=True, wait=True, output_prefix=""):
+    p = subprocess.Popen(cmds, cwd=cwd, env=env, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
+    if stream_output:
+        output_thread = threading.Thread(target=read_output, args=(p.stdout, output_prefix))
+        output_thread.start()
+
+    if wait:
+        p.wait()
+
+    return p
+
+
+def run_in_conda(cmds: list, *args, **kwargs):
+    cmds = [conda, "run", "--no-capture-output", "--prefix", SYSTEM_DIR] + cmds
+    return run(cmds, *args, **kwargs)
+
+
+def check_output_in_conda(cmds: list, cwd=None, env=None):
+    cmds = [conda, "run", "--no-capture-output", "--prefix", SYSTEM_DIR] + cmds
+    return subprocess.check_output(cmds, cwd=cwd, env=env, stderr=subprocess.PIPE)
+
+
+def create_context():
+    context = local()
+
+    # temp hack, throws an attribute not found error otherwise
+    context.torch_device = get_device(0)
+    context.device = f"{context.torch_device.type}:{context.torch_device.index}"
+    context.half_precision = True
+    context.vram_usage_level = None
+
+    context.models = {}
+    context.model_paths = {}
+    context.model_configs = {}
+    context.device_name = None
+    context.vram_optimizations = set()
+    context.vram_usage_level = "balanced"
+    context.test_diffusers = False
+    context.enable_codeformer = False
+
+    return context
+
+
+def get_env():
+    dir = os.path.abspath(SYSTEM_DIR)
+
+    if not os.path.exists(dir):
+        raise RuntimeError("The system folder is missing!")
+
+    config = getConfig()
+    backend_config = config.get("backend_config", {})
+    models_dir = config.get("models_dir", os.path.join(ROOT_DIR, "models"))
+
+    model_path_args = get_model_path_args()
+
+    env_entries = {
+        "PATH": [
+            f"{dir}",
+            f"{dir}/bin",
+            f"{dir}/Library/bin",
+            f"{dir}/Scripts",
+            f"{dir}/usr/bin",
+        ],
+        "PYTHONPATH": [
+            f"{dir}",
+            f"{dir}/lib/site-packages",
+            f"{dir}/lib/python3.10/site-packages",
+        ],
+        "PYTHONHOME": [],
+        "PY_LIBS": [
+            f"{dir}/Scripts/Lib",
+            f"{dir}/Scripts/Lib/site-packages",
+            f"{dir}/lib",
+            f"{dir}/lib/python3.10/site-packages",
+        ],
+        "PY_PIP": [f"{dir}/Scripts", f"{dir}/bin"],
+        "PIP_INSTALLER_LOCATION": [],  # [f"{dir}/python/get-pip.py"],
+        "TRANSFORMERS_CACHE": [f"{dir}/transformers-cache"],
+        "HF_HUB_DISABLE_SYMLINKS_WARNING": ["true"],
+        "COMMANDLINE_ARGS": [
+            f'--api --models-dir "{models_dir}" {model_path_args} --skip-torch-cuda-test --disable-gpu-warning --port {impl.WEBUI_PORT}'
+        ],
+        "SKIP_VENV": ["1"],
+        "SD_WEBUI_RESTARTING": ["1"],
+    }
+
+    if OS_NAME == "Windows":
+        env_entries["PATH"].append("C:/Windows/System32")
+        env_entries["PATH"].append("C:/Windows/System32/wbem")
+        env_entries["PYTHONNOUSERSITE"] = ["1"]
+        env_entries["PYTHON"] = [f"{dir}/python"]
+        env_entries["GIT"] = [f"{dir}/Library/bin/git"]
+        env_entries["COMMANDLINE_ARGS"][0] += f" --parent-pid {os.getpid()}"
+    else:
+        env_entries["PATH"].append("/bin")
+        env_entries["PATH"].append("/usr/bin")
+        env_entries["PATH"].append("/usr/sbin")
+        env_entries["PYTHONNOUSERSITE"] = ["y"]
+        env_entries["PYTHON"] = [f"{dir}/bin/python"]
+        env_entries["GIT"] = [f"{dir}/bin/git"]
+        env_entries["venv_dir"] = ["-"]
+
+    if OS_NAME == "Darwin":
+        # based on https://github.com/lllyasviel/stable-diffusion-webui-forge/blob/e26abf87ecd1eefd9ab0a198eee56f9c643e4001/webui-macos-env.sh
+        # hack - have to define these here, otherwise webui-macos-env.sh will overwrite COMMANDLINE_ARGS
+        env_entries["COMMANDLINE_ARGS"][0] += " --upcast-sampling --no-half-vae --use-cpu interrogate"
+        env_entries["PYTORCH_ENABLE_MPS_FALLBACK"] = ["1"]
+
+        cpu_name = str(subprocess.check_output(["sysctl", "-n", "machdep.cpu.brand_string"]))
+        if "Intel" in cpu_name:
+            env_entries["TORCH_COMMAND"] = ["pip install torch==2.1.2 torchvision==0.16.2"]
+        else:
+            env_entries["TORCH_COMMAND"] = ["pip install torch==2.3.1 torchvision==0.18.1"]
+    else:
+        from easydiffusion.device_manager import needs_to_force_full_precision
+
+        torch_platform_name = get_installed_torch_platform()[0]
+
+        vram_usage_level = config.get("vram_usage_level", "balanced")
+        if config.get("render_devices", "auto") == "cpu" or is_cpu_device(torch_platform_name):
+            env_entries["COMMANDLINE_ARGS"][0] += " --always-cpu"
+        else:
+            device = get_device(0)
+            if needs_to_force_full_precision(get_device_name(device)):
+                env_entries["COMMANDLINE_ARGS"][0] += " --no-half --precision full"
+
+            if vram_usage_level == "low":
+                env_entries["COMMANDLINE_ARGS"][0] += " --always-low-vram"
+            elif vram_usage_level == "high":
+                env_entries["COMMANDLINE_ARGS"][0] += " --always-high-vram"
+
+    cli_args = backend_config.get("COMMANDLINE_ARGS")
+    if cli_args:
+        env_entries["COMMANDLINE_ARGS"][0] += " " + cli_args
+
+    env = {}
+    for key, paths in env_entries.items():
+        paths = [p.replace("/", os.path.sep) for p in paths]
+        paths = os.pathsep.join(paths)
+
+        env[key] = paths
+
+    return env
+
+
+# https://stackoverflow.com/a/25134985
+def kill(proc_pid):
+    process = psutil.Process(proc_pid)
+    for proc in process.children(recursive=True):
+        proc.kill()
+    process.kill()
+
+
+def get_model_path_args():
+    args = []
+    for model_type, flag in MODELS_TO_OVERRIDE.items():
+        model_dir = get_model_dirs(model_type)[0]
+        args.append(f'{flag} "{model_dir}"')
+
+    return " ".join(args)

ui/easydiffusion/backends/webui/forge_api_refresh_text_encoders.patch

@@ -0,0 +1,25 @@
+diff --git a/modules/api/api.py b/modules/api/api.py
+index 9754be03..26d9eb9b 100644
+--- a/modules/api/api.py
++++ b/modules/api/api.py
+@@ -234,6 +234,7 @@ class Api:
+         self.add_api_route("/sdapi/v1/refresh-embeddings", self.refresh_embeddings, methods=["POST"])
+         self.add_api_route("/sdapi/v1/refresh-checkpoints", self.refresh_checkpoints, methods=["POST"])
+         self.add_api_route("/sdapi/v1/refresh-vae", self.refresh_vae, methods=["POST"])
++        self.add_api_route("/sdapi/v1/refresh-vae-and-text-encoders", self.refresh_vae_and_text_encoders, methods=["POST"])
+         self.add_api_route("/sdapi/v1/create/embedding", self.create_embedding, methods=["POST"], response_model=models.CreateResponse)
+         self.add_api_route("/sdapi/v1/create/hypernetwork", self.create_hypernetwork, methods=["POST"], response_model=models.CreateResponse)
+         self.add_api_route("/sdapi/v1/memory", self.get_memory, methods=["GET"], response_model=models.MemoryResponse)
+@@ -779,6 +780,12 @@ class Api:
+         with self.queue_lock:
+             shared_items.refresh_vae_list()
+ 
++    def refresh_vae_and_text_encoders(self):
++        from modules_forge.main_entry import refresh_models
++
++        with self.queue_lock:
++            refresh_models()
++
+     def create_embedding(self, args: dict):
+         try:
+             shared.state.begin(job="create_embedding")

ui/easydiffusion/backends/webui/forge_disable_corrupted_model_renaming.patch

@@ -0,0 +1,35 @@
+diff --git a/modules_forge/patch_basic.py b/modules_forge/patch_basic.py
+index 822e2838..5893efad 100644
+--- a/modules_forge/patch_basic.py
++++ b/modules_forge/patch_basic.py
+@@ -39,18 +39,18 @@ def build_loaded(module, loader_name):
+         except Exception as e:
+             result = None
+             exp = str(e) + '\n'
+-            for path in list(args) + list(kwargs.values()):
+-                if isinstance(path, str):
+-                    if os.path.exists(path):
+-                        exp += f'File corrupted: {path} \n'
+-                        corrupted_backup_file = path + '.corrupted'
+-                        if os.path.exists(corrupted_backup_file):
+-                            os.remove(corrupted_backup_file)
+-                        os.replace(path, corrupted_backup_file)
+-                        if os.path.exists(path):
+-                            os.remove(path)
+-                        exp += f'Forge has tried to move the corrupted file to {corrupted_backup_file} \n'
+-                        exp += f'You may try again now and Forge will download models again. \n'
++            # for path in list(args) + list(kwargs.values()):
++            #     if isinstance(path, str):
++            #         if os.path.exists(path):
++            #             exp += f'File corrupted: {path} \n'
++            #             corrupted_backup_file = path + '.corrupted'
++            #             if os.path.exists(corrupted_backup_file):
++            #                 os.remove(corrupted_backup_file)
++            #             os.replace(path, corrupted_backup_file)
++            #             if os.path.exists(path):
++            #                 os.remove(path)
++            #             exp += f'Forge has tried to move the corrupted file to {corrupted_backup_file} \n'
++            #             exp += f'You may try again now and Forge will download models again. \n'
+             raise ValueError(exp)
+         return result
+ 

ui/easydiffusion/backends/webui/forge_exception_leak_patch.patch

@@ -0,0 +1,20 @@
+diff --git a/modules_forge/main_thread.py b/modules_forge/main_thread.py
+index eb3e7889..0cfcadd1 100644
+--- a/modules_forge/main_thread.py
++++ b/modules_forge/main_thread.py
+@@ -33,8 +33,8 @@ class Task:
+         except Exception as e:
+             traceback.print_exc()
+             print(e)
+-            self.exception = e
+-            last_exception = e
++            self.exception = str(e)
++            last_exception = str(e)
+ 
+ 
+ def loop():
+@@ -74,4 +74,3 @@ def run_and_wait_result(func, *args, **kwargs):
+             with lock:
+                 finished_list.remove(finished_task)
+             return finished_task.result
+-

ui/easydiffusion/backends/webui/forge_loader_force_gc.patch

@@ -0,0 +1,27 @@
+diff --git a/backend/loader.py b/backend/loader.py
+index a090adab..53fdb26d 100644
+--- a/backend/loader.py
++++ b/backend/loader.py
+@@ -1,4 +1,5 @@
+ import os
++import gc
+ import torch
+ import logging
+ import importlib
+@@ -447,6 +448,8 @@ def preprocess_state_dict(sd):
+ 
+ 
+ def split_state_dict(sd, additional_state_dicts: list = None):
++    gc.collect()
++
+     sd = load_torch_file(sd)
+     sd = preprocess_state_dict(sd)
+     guess = huggingface_guess.guess(sd)
+@@ -456,6 +459,7 @@ def split_state_dict(sd, additional_state_dicts: list = None):
+             asd = load_torch_file(asd)
+             sd = replace_state_dict(sd, asd, guess)
+             del asd
++            gc.collect()
+ 
+     guess.clip_target = guess.clip_target(sd)
+     guess.model_type = guess.model_type(sd)

ui/easydiffusion/backends/webui/forge_model_crash_recovery.patch

@@ -0,0 +1,12 @@
+diff --git a/modules/sd_models.py b/modules/sd_models.py
+index 76940ecc..b07d84b6 100644
+--- a/modules/sd_models.py
++++ b/modules/sd_models.py
+@@ -482,6 +482,7 @@ def forge_model_reload():
+ 
+     if model_data.sd_model:
+         model_data.sd_model = None
++        model_data.forge_hash = None
+         memory_management.unload_all_models()
+         memory_management.soft_empty_cache()
+         gc.collect()

ui/easydiffusion/backends/webui/forge_monitor_parent_process.patch

@@ -0,0 +1,85 @@
+diff --git a/launch.py b/launch.py
+index c0568c7b..3919f7dd 100644
+--- a/launch.py
++++ b/launch.py
+@@ -2,6 +2,7 @@
+ # faulthandler.enable()
+ 
+ from modules import launch_utils
++from modules import parent_process_monitor
+ 
+ args = launch_utils.args
+ python = launch_utils.python
+@@ -28,6 +29,10 @@ start = launch_utils.start
+ 
+ 
+ def main():
++    if args.parent_pid != -1:
++        print(f"Monitoring parent process for termination. Parent PID: {args.parent_pid}")
++        parent_process_monitor.start_monitor_thread(args.parent_pid)
++
+     if args.dump_sysinfo:
+         filename = launch_utils.dump_sysinfo()
+ 
+diff --git a/modules/cmd_args.py b/modules/cmd_args.py
+index fcd8a50f..7f684bec 100644
+--- a/modules/cmd_args.py
++++ b/modules/cmd_args.py
+@@ -148,3 +148,6 @@ parser.add_argument(
+     help="Path to directory with annotator model directories",
+     default=None,
+ )
++
++# Easy Diffusion arguments
++parser.add_argument("--parent-pid", type=int, default=-1, help='parent process id, if running webui as a sub-process')
+diff --git a/modules/parent_process_monitor.py b/modules/parent_process_monitor.py
+new file mode 100644
+index 00000000..cc3e2049
+--- /dev/null
++++ b/modules/parent_process_monitor.py
+@@ -0,0 +1,45 @@
++# monitors and kills itself when the parent process dies. required when running Forge as a sub-process.
++# modified version of https://stackoverflow.com/a/23587108
++
++import sys
++import os
++import threading
++import platform
++import time
++
++
++def _monitor_parent_posix(parent_pid):
++    print(f"Monitoring parent pid: {parent_pid}")
++    while True:
++        if os.getppid() != parent_pid:
++            os._exit(0)
++        time.sleep(1)
++
++
++def _monitor_parent_windows(parent_pid):
++    from ctypes import WinDLL, WinError
++    from ctypes.wintypes import DWORD, BOOL, HANDLE
++
++    SYNCHRONIZE = 0x00100000  # Magic value from http://msdn.microsoft.com/en-us/library/ms684880.aspx
++    kernel32 = WinDLL("kernel32.dll")
++    kernel32.OpenProcess.argtypes = (DWORD, BOOL, DWORD)
++    kernel32.OpenProcess.restype = HANDLE
++
++    handle = kernel32.OpenProcess(SYNCHRONIZE, False, parent_pid)
++    if not handle:
++        raise WinError()
++
++    # Wait until parent exits
++    from ctypes import windll
++
++    print(f"Monitoring parent pid: {parent_pid}")
++    windll.kernel32.WaitForSingleObject(handle, -1)
++    os._exit(0)
++
++
++def start_monitor_thread(parent_pid):
++    if platform.system() == "Windows":
++        t = threading.Thread(target=_monitor_parent_windows, args=(parent_pid,), daemon=True)
++    else:
++        t = threading.Thread(target=_monitor_parent_posix, args=(parent_pid,), daemon=True)
++    t.start()

ui/easydiffusion/backends/webui/impl.py

@@ -0,0 +1,661 @@
+import os
+import requests
+from requests.exceptions import ConnectTimeout, ConnectionError, ReadTimeout
+from typing import Union, List
+from threading import local as Context
+from threading import Thread
+import uuid
+import time
+from copy import deepcopy
+
+from sdkit.utils import base64_str_to_img, img_to_base64_str, log
+
+WEBUI_HOST = "localhost"
+WEBUI_PORT = "7860"
+
+DEFAULT_WEBUI_OPTIONS = {
+    "show_progress_every_n_steps": 3,
+    "show_progress_grid": True,
+    "live_previews_enable": False,
+    "forge_additional_modules": [],
+}
+
+
+webui_opts: dict = None
+
+
+curr_models = {
+    "stable-diffusion": None,
+    "vae": None,
+    "text-encoder": None,
+}
+
+
+def set_options(context, **kwargs):
+    changed_opts = {}
+
+    opts_mapping = {
+        "stream_image_progress": ("live_previews_enable", bool),
+        "stream_image_progress_interval": ("show_progress_every_n_steps", int),
+        "clip_skip": ("CLIP_stop_at_last_layers", int),
+        "clip_skip_sdxl": ("sdxl_clip_l_skip", bool),
+        "output_format": ("samples_format", str),
+    }
+
+    for ed_key, webui_key in opts_mapping.items():
+        webui_key, webui_type = webui_key
+
+        if ed_key in kwargs and (webui_opts is None or webui_opts.get(webui_key, False) != webui_type(kwargs[ed_key])):
+            changed_opts[webui_key] = webui_type(kwargs[ed_key])
+
+    if changed_opts:
+        changed_opts["sd_model_checkpoint"] = curr_models["stable-diffusion"]
+
+        print(f"Got options: {kwargs}. Sending options: {changed_opts}")
+
+        try:
+            res = webui_post("/sdapi/v1/options", json=changed_opts)
+            if res.status_code != 200:
+                raise Exception(res.text)
+
+            webui_opts.update(changed_opts)
+        except Exception as e:
+            print(f"Error setting options: {e}")
+
+
+def ping(timeout=1):
+    "timeout (in seconds)"
+
+    global webui_opts
+
+    try:
+        res = webui_get("/internal/ping", timeout=timeout)
+
+        if res.status_code != 200:
+            raise ConnectTimeout(res.text)
+
+        if webui_opts is None:
+            try:
+                res = webui_post("/sdapi/v1/options", json=DEFAULT_WEBUI_OPTIONS)
+                if res.status_code != 200:
+                    raise Exception(res.text)
+            except Exception as e:
+                print(f"Error setting options: {e}")
+
+            try:
+                res = webui_get("/sdapi/v1/options")
+                if res.status_code != 200:
+                    raise Exception(res.text)
+
+                webui_opts = res.json()
+            except Exception as e:
+                print(f"Error getting options: {e}")
+
+        return True
+    except (ConnectTimeout, ConnectionError, ReadTimeout) as e:
+        raise TimeoutError(e)
+
+
+def load_model(context, model_type, **kwargs):
+    from easydiffusion.app import ROOT_DIR, getConfig
+
+    config = getConfig()
+    models_dir = config.get("models_dir", os.path.join(ROOT_DIR, "models"))
+
+    model_path = context.model_paths[model_type]
+
+    if model_type == "stable-diffusion":
+        base_dir = os.path.join(models_dir, model_type)
+        model_path = os.path.relpath(model_path, base_dir)
+
+    # print(f"load model: {model_type=} {model_path=} {curr_models=}")
+    curr_models[model_type] = model_path
+
+
+def unload_model(context, model_type, **kwargs):
+    # print(f"unload model: {model_type=} {curr_models=}")
+    curr_models[model_type] = None
+
+
+def flush_model_changes(context):
+    if webui_opts is None:
+        print("Server not ready, can't set the model")
+        return
+
+    modules = []
+    for model_type in ("vae", "text-encoder"):
+        if curr_models[model_type]:
+            model_paths = curr_models[model_type]
+            model_paths = [model_paths] if not isinstance(model_paths, list) else model_paths
+            modules += model_paths
+
+    opts = {"sd_model_checkpoint": curr_models["stable-diffusion"], "forge_additional_modules": modules}
+
+    print("Setting backend models", opts)
+
+    try:
+        res = webui_post("/sdapi/v1/options", json=opts)
+        print("got res", res.status_code)
+        if res.status_code != 200:
+            raise Exception(res.text)
+    except Exception as e:
+        raise RuntimeError(
+            f"The engine failed to set the required options. Please check the logs in the command line window for more details."
+        )
+
+
+def generate_images(
+    context: Context,
+    prompt: str = "",
+    negative_prompt: str = "",
+    seed: int = 42,
+    width: int = 512,
+    height: int = 512,
+    num_outputs: int = 1,
+    num_inference_steps: int = 25,
+    guidance_scale: float = 7.5,
+    distilled_guidance_scale: float = 3.5,
+    init_image=None,
+    init_image_mask=None,
+    control_image=None,
+    control_alpha=1.0,
+    controlnet_filter=None,
+    prompt_strength: float = 0.8,
+    preserve_init_image_color_profile=False,
+    strict_mask_border=False,
+    sampler_name: str = "euler_a",
+    scheduler_name: str = "simple",
+    hypernetwork_strength: float = 0,
+    tiling=None,
+    lora_alpha: Union[float, List[float]] = 0,
+    sampler_params={},
+    callback=None,
+    output_type="pil",
+):
+
+    task_id = str(uuid.uuid4())
+
+    sampler_name = convert_ED_sampler_names(sampler_name)
+    controlnet_filter = convert_ED_controlnet_filter_name(controlnet_filter)
+
+    cmd = {
+        "force_task_id": task_id,
+        "prompt": prompt,
+        "negative_prompt": negative_prompt,
+        "sampler_name": sampler_name,
+        "scheduler": scheduler_name,
+        "steps": num_inference_steps,
+        "seed": seed,
+        "cfg_scale": guidance_scale,
+        "distilled_cfg_scale": distilled_guidance_scale,
+        "batch_size": num_outputs,
+        "width": width,
+        "height": height,
+    }
+
+    if init_image:
+        cmd["init_images"] = [init_image]
+        cmd["denoising_strength"] = prompt_strength
+    if init_image_mask:
+        cmd["mask"] = init_image_mask if isinstance(init_image_mask, str) else img_to_base64_str(init_image_mask)
+        cmd["include_init_images"] = True
+        cmd["inpainting_fill"] = 1
+        cmd["initial_noise_multiplier"] = 1
+        cmd["inpaint_full_res"] = 1
+
+    if context.model_paths.get("lora"):
+        lora_model = context.model_paths["lora"]
+        lora_model = lora_model if isinstance(lora_model, list) else [lora_model]
+        lora_alpha = lora_alpha if isinstance(lora_alpha, list) else [lora_alpha]
+
+        for lora, alpha in zip(lora_model, lora_alpha):
+            lora = os.path.basename(lora)
+            lora = os.path.splitext(lora)[0]
+            cmd["prompt"] += f" <lora:{lora}:{alpha}>"
+
+    if controlnet_filter and control_image and context.model_paths.get("controlnet"):
+        controlnet_model = context.model_paths["controlnet"]
+
+        model_hash = auto1111_hash(controlnet_model)
+        controlnet_model = os.path.basename(controlnet_model)
+        controlnet_model = os.path.splitext(controlnet_model)[0]
+        print(f"setting controlnet model: {controlnet_model}")
+        controlnet_model = f"{controlnet_model} [{model_hash}]"
+
+        cmd["alwayson_scripts"] = {
+            "controlnet": {
+                "args": [
+                    {
+                        "image": control_image,
+                        "weight": control_alpha,
+                        "module": controlnet_filter,
+                        "model": controlnet_model,
+                        "resize_mode": "Crop and Resize",
+                        "threshold_a": 50,
+                        "threshold_b": 130,
+                    }
+                ]
+            }
+        }
+
+    operation_to_apply = "img2img" if init_image else "txt2img"
+
+    stream_image_progress = webui_opts.get("live_previews_enable", False)
+
+    progress_thread = Thread(
+        target=image_progress_thread, args=(task_id, callback, stream_image_progress, num_outputs, num_inference_steps)
+    )
+    progress_thread.start()
+
+    print(f"task id: {task_id}")
+    print_request(operation_to_apply, cmd)
+
+    res = webui_post(f"/sdapi/v1/{operation_to_apply}", json=cmd)
+    if res.status_code == 200:
+        res = res.json()
+    else:
+        if res.status_code == 500:
+            res = res.json()
+            log.error(f"Server error: {res}")
+            raise Exception(f"{res['message']}. Please check the logs in the command-line window for more details.")
+
+        raise Exception(
+            f"HTTP Status {res.status_code}. The engine failed while generating this image. Please check the logs in the command-line window for more details."
+        )
+
+    import json
+
+    print(json.loads(res["info"])["infotexts"])
+
+    images = res["images"]
+    if output_type == "pil":
+        images = [base64_str_to_img(img) for img in images]
+    elif output_type == "base64":
+        images = [base64_buffer_to_base64_img(img) for img in images]
+
+    return images
+
+
+def filter_images(context: Context, images, filters, filter_params={}, input_type="pil"):
+    """
+    * context: Context
+    * images: str or PIL.Image or list of str/PIL.Image - image to filter. if a string is passed, it needs to be a base64-encoded image
+    * filters: filter_type (string) or list of strings
+    * filter_params: dict
+
+    returns: [PIL.Image] - list of filtered images
+    """
+    images = images if isinstance(images, list) else [images]
+    filters = filters if isinstance(filters, list) else [filters]
+
+    if "nsfw_checker" in filters:
+        filters.remove("nsfw_checker")  # handled by ED directly
+
+    args = {}
+    controlnet_filters = []
+
+    print(filter_params)
+
+    for filter_name in filters:
+        params = filter_params.get(filter_name, {})
+
+        if filter_name == "gfpgan":
+            args["gfpgan_visibility"] = 1
+
+        if filter_name in ("realesrgan", "esrgan_4x", "lanczos", "nearest", "scunet", "swinir"):
+            args["upscaler_1"] = params.get("upscaler", "RealESRGAN_x4plus")
+            args["upscaling_resize"] = params.get("scale", 4)
+
+            if args["upscaler_1"] == "RealESRGAN_x4plus":
+                args["upscaler_1"] = "R-ESRGAN 4x+"
+            elif args["upscaler_1"] == "RealESRGAN_x4plus_anime_6B":
+                args["upscaler_1"] = "R-ESRGAN 4x+ Anime6B"
+
+        if filter_name == "codeformer":
+            args["codeformer_visibility"] = 1
+            args["codeformer_weight"] = params.get("codeformer_fidelity", 0.5)
+
+        if filter_name.startswith("controlnet_"):
+            filter_name = convert_ED_controlnet_filter_name(filter_name)
+            controlnet_filters.append(filter_name)
+
+    print(f"filtering {len(images)} images with {args}. {controlnet_filters=}")
+
+    if len(filters) > len(controlnet_filters):
+        filtered_images = extra_batch_images(images, input_type=input_type, **args)
+    else:
+        filtered_images = images
+
+    for filter_name in controlnet_filters:
+        filtered_images = controlnet_filter(filtered_images, module=filter_name, input_type=input_type)
+
+    return filtered_images
+
+
+def get_url():
+    return f"//{WEBUI_HOST}:{WEBUI_PORT}/?__theme=dark"
+
+
+def stop_rendering(context):
+    try:
+        res = webui_post("/sdapi/v1/interrupt")
+        if res.status_code != 200:
+            raise Exception(res.text)
+    except Exception as e:
+        print(f"Error interrupting webui: {e}")
+
+
+def refresh_models():
+    def make_refresh_call(type):
+        try:
+            webui_post(f"/sdapi/v1/refresh-{type}")
+        except:
+            pass
+
+    try:
+        for type in ("checkpoints", "vae-and-text-encoders"):
+            t = Thread(target=make_refresh_call, args=(type,))
+            t.start()
+    except Exception as e:
+        print(f"Error refreshing models: {e}")
+
+
+def list_controlnet_filters():
+    return [
+        "openpose",
+        "openpose_face",
+        "openpose_faceonly",
+        "openpose_hand",
+        "openpose_full",
+        "animal_openpose",
+        "densepose_parula (black bg & blue torso)",
+        "densepose (pruple bg & purple torso)",
+        "dw_openpose_full",
+        "mediapipe_face",
+        "instant_id_face_keypoints",
+        "InsightFace+CLIP-H (IPAdapter)",
+        "InsightFace (InstantID)",
+        "canny",
+        "mlsd",
+        "scribble_hed",
+        "scribble_hedsafe",
+        "scribble_pidinet",
+        "scribble_pidsafe",
+        "scribble_xdog",
+        "softedge_hed",
+        "softedge_hedsafe",
+        "softedge_pidinet",
+        "softedge_pidsafe",
+        "softedge_teed",
+        "normal_bae",
+        "depth_midas",
+        "normal_midas",
+        "depth_zoe",
+        "depth_leres",
+        "depth_leres++",
+        "depth_anything_v2",
+        "depth_anything",
+        "depth_hand_refiner",
+        "depth_marigold",
+        "lineart_coarse",
+        "lineart_realistic",
+        "lineart_anime",
+        "lineart_standard (from white bg & black line)",
+        "lineart_anime_denoise",
+        "reference_adain",
+        "reference_only",
+        "reference_adain+attn",
+        "tile_colorfix",
+        "tile_resample",
+        "tile_colorfix+sharp",
+        "CLIP-ViT-H (IPAdapter)",
+        "CLIP-G (Revision)",
+        "CLIP-G (Revision ignore prompt)",
+        "CLIP-ViT-bigG (IPAdapter)",
+        "InsightFace+CLIP-H (IPAdapter)",
+        "inpaint_only",
+        "inpaint_only+lama",
+        "inpaint_global_harmonious",
+        "seg_ufade20k",
+        "seg_ofade20k",
+        "seg_anime_face",
+        "seg_ofcoco",
+        "shuffle",
+        "segment",
+        "invert (from white bg & black line)",
+        "threshold",
+        "t2ia_sketch_pidi",
+        "t2ia_color_grid",
+        "recolor_intensity",
+        "recolor_luminance",
+        "blur_gaussian",
+    ]
+
+
+def controlnet_filter(images, module="none", processor_res=512, threshold_a=64, threshold_b=64, input_type="pil"):
+    if input_type == "pil":
+        images = [img_to_base64_str(x) for x in images]
+
+    payload = {
+        "controlnet_module": module,
+        "controlnet_input_images": images,
+        "controlnet_processor_res": processor_res,
+        "controlnet_threshold_a": threshold_a,
+        "controlnet_threshold_b": threshold_b,
+    }
+    res = webui_post("/controlnet/detect", json=payload)
+    res = res.json()
+    filtered_images = res["images"]
+
+    if input_type == "pil":
+        filtered_images = [base64_str_to_img(img) for img in filtered_images]
+    elif input_type == "base64":
+        filtered_images = [base64_buffer_to_base64_img(img) for img in filtered_images]
+
+    return filtered_images
+
+
+def image_progress_thread(task_id, callback, stream_image_progress, total_images, total_steps):
+    from PIL import Image
+
+    last_preview_id = -1
+
+    EMPTY_IMAGE = Image.new("RGB", (1, 1))
+
+    while True:
+        res = webui_post(
+            f"/internal/progress",
+            json={"id_task": task_id, "live_preview": stream_image_progress, "id_live_preview": last_preview_id},
+        )
+        if res.status_code == 200:
+            res = res.json()
+        else:
+            raise RuntimeError(f"Unexpected progress response. Status code: {res.status_code}. Res: {res.text}")
+
+        last_preview_id = res["id_live_preview"]
+
+        if res["progress"] is not None:
+            step_num = int(res["progress"] * total_steps)
+
+            if res["live_preview"] is not None:
+                img = res["live_preview"]
+                img = base64_str_to_img(img)
+                images = [EMPTY_IMAGE] * total_images
+                images[0] = img
+            else:
+                images = None
+
+            callback(images, step_num)
+
+        if res["completed"] == True:
+            print("Complete!")
+            break
+
+        time.sleep(0.5)
+
+
+def webui_get(uri, *args, **kwargs):
+    url = f"http://{WEBUI_HOST}:{WEBUI_PORT}{uri}"
+    return requests.get(url, *args, **kwargs)
+
+
+def webui_post(uri, *args, **kwargs):
+    url = f"http://{WEBUI_HOST}:{WEBUI_PORT}{uri}"
+    return requests.post(url, *args, **kwargs)
+
+
+def print_request(operation_to_apply, args):
+    args = deepcopy(args)
+    if "init_images" in args:
+        args["init_images"] = ["img" for _ in args["init_images"]]
+    if "mask" in args:
+        args["mask"] = "mask_img"
+
+    controlnet_args = args.get("alwayson_scripts", {}).get("controlnet", {}).get("args", [])
+    if controlnet_args:
+        controlnet_args[0]["image"] = "control_image"
+
+    print(f"operation: {operation_to_apply}, args: {args}")
+
+
+def auto1111_hash(file_path):
+    import hashlib
+
+    with open(file_path, "rb") as f:
+        f.seek(0x100000)
+        b = f.read(0x10000)
+        return hashlib.sha256(b).hexdigest()[:8]
+
+
+def extra_batch_images(
+    images,  # list of PIL images
+    name_list=None,  # list of image names
+    resize_mode=0,
+    show_extras_results=True,
+    gfpgan_visibility=0,
+    codeformer_visibility=0,
+    codeformer_weight=0,
+    upscaling_resize=2,
+    upscaling_resize_w=512,
+    upscaling_resize_h=512,
+    upscaling_crop=True,
+    upscaler_1="None",
+    upscaler_2="None",
+    extras_upscaler_2_visibility=0,
+    upscale_first=False,
+    use_async=False,
+    input_type="pil",
+):
+    if name_list is not None:
+        if len(name_list) != len(images):
+            raise RuntimeError("len(images) != len(name_list)")
+    else:
+        name_list = [f"image{i + 1:05}" for i in range(len(images))]
+
+    if input_type == "pil":
+        images = [img_to_base64_str(x) for x in images]
+
+    image_list = []
+    for name, image in zip(name_list, images):
+        image_list.append({"data": image, "name": name})
+
+    payload = {
+        "resize_mode": resize_mode,
+        "show_extras_results": show_extras_results,
+        "gfpgan_visibility": gfpgan_visibility,
+        "codeformer_visibility": codeformer_visibility,
+        "codeformer_weight": codeformer_weight,
+        "upscaling_resize": upscaling_resize,
+        "upscaling_resize_w": upscaling_resize_w,
+        "upscaling_resize_h": upscaling_resize_h,
+        "upscaling_crop": upscaling_crop,
+        "upscaler_1": upscaler_1,
+        "upscaler_2": upscaler_2,
+        "extras_upscaler_2_visibility": extras_upscaler_2_visibility,
+        "upscale_first": upscale_first,
+        "imageList": image_list,
+    }
+
+    res = webui_post("/sdapi/v1/extra-batch-images", json=payload)
+    if res.status_code == 200:
+        res = res.json()
+    else:
+        raise Exception(
+            "The engine failed while filtering this image. Please check the logs in the command-line window for more details."
+        )
+
+    images = res["images"]
+
+    if input_type == "pil":
+        images = [base64_str_to_img(img) for img in images]
+    elif input_type == "base64":
+        images = [base64_buffer_to_base64_img(img) for img in images]
+
+    return images
+
+
+def base64_buffer_to_base64_img(img):
+    output_format = webui_opts.get("samples_format", "jpeg")
+    mime_type = f"image/{output_format.lower()}"
+    return f"data:{mime_type};base64," + img
+
+
+def convert_ED_sampler_names(sampler_name):
+    name_mapping = {
+        "dpmpp_2m": "DPM++ 2M",
+        "dpmpp_sde": "DPM++ SDE",
+        "dpmpp_2m_sde": "DPM++ 2M SDE",
+        "dpmpp_2m_sde_heun": "DPM++ 2M SDE Heun",
+        "dpmpp_2s_a": "DPM++ 2S a",
+        "dpmpp_3m_sde": "DPM++ 3M SDE",
+        "euler_a": "Euler a",
+        "euler": "Euler",
+        "lms": "LMS",
+        "heun": "Heun",
+        "dpm2": "DPM2",
+        "dpm2_a": "DPM2 a",
+        "dpm_fast": "DPM fast",
+        "dpm_adaptive": "DPM adaptive",
+        "restart": "Restart",
+        "heun_pp2": "HeunPP2",
+        "ipndm": "IPNDM",
+        "ipndm_v": "IPNDM_V",
+        "deis": "DEIS",
+        "ddim": "DDIM",
+        "ddim_cfgpp": "DDIM CFG++",
+        "plms": "PLMS",
+        "unipc": "UniPC",
+        "lcm": "LCM",
+        "ddpm": "DDPM",
+        "forge_flux_realistic": "[Forge] Flux Realistic",
+        "forge_flux_realistic_slow": "[Forge] Flux Realistic (Slow)",
+        # deprecated samplers in 3.5
+        "dpm_solver_stability": None,
+        "unipc_snr": None,
+        "unipc_tu": None,
+        "unipc_snr_2": None,
+        "unipc_tu_2": None,
+        "unipc_tq": None,
+    }
+    return name_mapping.get(sampler_name)
+
+
+def convert_ED_controlnet_filter_name(filter):
+    if filter is None:
+        return None
+
+    def cn(n):
+        if n.startswith("controlnet_"):
+            return n[len("controlnet_") :]
+        return n
+
+    mapping = {
+        "controlnet_scribble_hedsafe": None,
+        "controlnet_scribble_pidsafe": None,
+        "controlnet_softedge_pidsafe": "controlnet_softedge_pidisafe",
+        "controlnet_normal_bae": "controlnet_normalbae",
+        "controlnet_segment": None,
+    }
+    if isinstance(filter, list):
+        return [cn(mapping.get(f, f)) for f in filter]
+    return cn(mapping.get(filter, filter))

ui/easydiffusion/bucket_manager.py

@@ -55,8 +55,13 @@ def init():
             return bucketfiles
 
         else:
-            bucket_id = crud.get_bucket_by_path(db, path).id
+            bucket = crud.get_bucket_by_path(db, path)
+            if bucket == None:
+                raise HTTPException(status_code=404, detail="Bucket not found")
+            bucket_id = bucket.id
             bucketfile = db.query(models.BucketFile).filter(models.BucketFile.bucket_id == bucket_id, models.BucketFile.filename == filename).first()
+            if bucketfile == None:
+                raise HTTPException(status_code=404, detail="File not found")
 
             suffix = get_suffix_from_filename(filename)
 

ui/easydiffusion/device_manager.py

@@ -6,6 +6,15 @@ import traceback
 import torch
 from easydiffusion.utils import log
 
+from torchruntime.utils import (
+    get_installed_torch_platform,
+    get_device,
+    get_device_count,
+    get_device_name,
+    SUPPORTED_BACKENDS,
+)
+from sdkit.utils import mem_get_info, is_cpu_device, has_half_precision_bug
+
 """
 Set `FORCE_FULL_PRECISION` in the environment variables, or in `config.bat`/`config.sh` to set full precision (i.e. float32).
 Otherwise the models will load at half-precision (i.e. float16).
@@ -22,33 +31,15 @@ mem_free_threshold = 0
 
 def get_device_delta(render_devices, active_devices):
     """
-    render_devices: 'cpu', or 'auto', or 'mps' or ['cuda:N'...]
-    active_devices: ['cpu', 'mps', 'cuda:N'...]
+    render_devices: 'auto' or backends listed in `torchruntime.utils.SUPPORTED_BACKENDS`
+    active_devices: [backends listed in `torchruntime.utils.SUPPORTED_BACKENDS`]
     """
 
-    if render_devices in ("cpu", "auto", "mps"):
-        render_devices = [render_devices]
-    elif render_devices is not None:
-        if isinstance(render_devices, str):
-            render_devices = [render_devices]
-        if isinstance(render_devices, list) and len(render_devices) > 0:
-            render_devices = list(filter(lambda x: x.startswith("cuda:") or x == "mps", render_devices))
-            if len(render_devices) == 0:
-                raise Exception(
-                    'Invalid render_devices value in config.json. Valid: {"render_devices": ["cuda:0", "cuda:1"...]}, or {"render_devices": "cpu"} or {"render_devices": "mps"} or {"render_devices": "auto"}'
-                )
+    render_devices = render_devices or "auto"
+    render_devices = [render_devices] if isinstance(render_devices, str) else render_devices
 
-            render_devices = list(filter(lambda x: is_device_compatible(x), render_devices))
-            if len(render_devices) == 0:
-                raise Exception(
-                    "Sorry, none of the render_devices configured in config.json are compatible with Stable Diffusion"
-                )
-        else:
-            raise Exception(
-                'Invalid render_devices value in config.json. Valid: {"render_devices": ["cuda:0", "cuda:1"...]}, or {"render_devices": "cpu"} or {"render_devices": "auto"}'
-            )
-    else:
-        render_devices = ["auto"]
+    # check for backend support
+    validate_render_devices(render_devices)
 
     if "auto" in render_devices:
         render_devices = auto_pick_devices(active_devices)
@@ -64,47 +55,39 @@ def get_device_delta(render_devices, active_devices):
     return devices_to_start, devices_to_stop
 
 
-def is_mps_available():
-    return (
-        platform.system() == "Darwin"
-        and hasattr(torch.backends, "mps")
-        and torch.backends.mps.is_available()
-        and torch.backends.mps.is_built()
-    )
+def validate_render_devices(render_devices):
+    supported_backends = ("auto",) + SUPPORTED_BACKENDS
+    unsupported_render_devices = [d for d in render_devices if not d.lower().startswith(supported_backends)]
 
-
-def is_cuda_available():
-    return torch.cuda.is_available()
+    if unsupported_render_devices:
+        raise ValueError(
+            f"Invalid render devices in config: {unsupported_render_devices}. Valid render devices: {supported_backends}"
+        )
 
 
 def auto_pick_devices(currently_active_devices):
     global mem_free_threshold
 
-    if is_mps_available():
-        return ["mps"]
+    torch_platform_name = get_installed_torch_platform()[0]
 
-    if not is_cuda_available():
-        return ["cpu"]
-
-    device_count = torch.cuda.device_count()
-    if device_count == 1:
-        return ["cuda:0"] if is_device_compatible("cuda:0") else ["cpu"]
+    if is_cpu_device(torch_platform_name):
+        return [torch_platform_name]
 
+    device_count = get_device_count()
     log.debug("Autoselecting GPU. Using most free memory.")
     devices = []
-    for device in range(device_count):
-        device = f"cuda:{device}"
-        if not is_device_compatible(device):
-            continue
+    for device_id in range(device_count):
+        device_id = f"{torch_platform_name}:{device_id}" if device_count > 1 else torch_platform_name
+        device = get_device(device_id)
 
-        mem_free, mem_total = torch.cuda.mem_get_info(device)
+        mem_free, mem_total = mem_get_info(device)
         mem_free /= float(10**9)
         mem_total /= float(10**9)
-        device_name = torch.cuda.get_device_name(device)
+        device_name = get_device_name(device)
         log.debug(
-            f"{device} detected: {device_name} - Memory (free/total): {round(mem_free, 2)}Gb / {round(mem_total, 2)}Gb"
+            f"{device_id} detected: {device_name} - Memory (free/total): {round(mem_free, 2)}Gb / {round(mem_total, 2)}Gb"
         )
-        devices.append({"device": device, "device_name": device_name, "mem_free": mem_free})
+        devices.append({"device": device_id, "device_name": device_name, "mem_free": mem_free})
 
     devices.sort(key=lambda x: x["mem_free"], reverse=True)
     max_mem_free = devices[0]["mem_free"]
@@ -117,69 +100,45 @@ def auto_pick_devices(currently_active_devices):
     #    always be very low (since their VRAM contains the model).
     #    These already-running devices probably aren't terrible, since they were picked in the past.
     #    Worst case, the user can restart the program and that'll get rid of them.
-    devices = list(
-        filter(
-            (lambda x: x["mem_free"] > mem_free_threshold or x["device"] in currently_active_devices),
-            devices,
-        )
-    )
-    devices = list(map(lambda x: x["device"], devices))
+    devices = [
+        x["device"] for x in devices if x["mem_free"] >= mem_free_threshold or x["device"] in currently_active_devices
+    ]
     return devices
 
 
-def device_init(context, device):
-    """
-    This function assumes the 'device' has already been verified to be compatible.
-    `get_device_delta()` has already filtered out incompatible devices.
-    """
+def device_init(context, device_id):
+    context.device = device_id
 
-    validate_device_id(device, log_prefix="device_init")
-
-    if "cuda" not in device:
-        context.device = device
+    if is_cpu_device(context.torch_device):
         context.device_name = get_processor_name()
         context.half_precision = False
-        log.debug(f"Render device available as {context.device_name}")
-        return
+    else:
+        context.device_name = get_device_name(context.torch_device)
 
-    context.device_name = torch.cuda.get_device_name(device)
-    context.device = device
+        # Some graphics cards have bugs in their firmware that prevent image generation at half precision
+        if needs_to_force_full_precision(context.device_name):
+            log.warn(f"forcing full precision on this GPU, to avoid corrupted images. GPU: {context.device_name}")
+            context.half_precision = False
 
-    # Force full precision on 1660 and 1650 NVIDIA cards to avoid creating green images
-    if needs_to_force_full_precision(context):
-        log.warn(f"forcing full precision on this GPU, to avoid green images. GPU detected: {context.device_name}")
-        # Apply force_full_precision now before models are loaded.
-        context.half_precision = False
-
-    log.info(f'Setting {device} as active, with precision: {"half" if context.half_precision else "full"}')
-    torch.cuda.device(device)
+    log.info(f'Setting {device_id} as active, with precision: {"half" if context.half_precision else "full"}')
 
 
-def needs_to_force_full_precision(context):
+def needs_to_force_full_precision(device_name):
     if "FORCE_FULL_PRECISION" in os.environ:
         return True
 
-    device_name = context.device_name.lower()
-    return (
-        ("nvidia" in device_name or "geforce" in device_name or "quadro" in device_name)
-        and (
-            " 1660" in device_name
-            or " 1650" in device_name
-            or " 1630" in device_name
-            or " t400" in device_name
-            or " t550" in device_name
-            or " t600" in device_name
-            or " t1000" in device_name
-            or " t1200" in device_name
-            or " t2000" in device_name
-        )
-    ) or ("tesla k40m" in device_name)
+    return has_half_precision_bug(device_name.lower())
 
 
 def get_max_vram_usage_level(device):
-    if "cuda" in device:
-        _, mem_total = torch.cuda.mem_get_info(device)
-    else:
+    "Expects a torch.device as the argument"
+
+    if is_cpu_device(device):
+        return "high"
+
+    _, mem_total = mem_get_info(device)
+
+    if mem_total < 0.001:  # probably a torch platform without a mem_get_info() implementation
         return "high"
 
     mem_total /= float(10**9)
@@ -191,51 +150,6 @@ def get_max_vram_usage_level(device):
     return "high"
 
 
-def validate_device_id(device, log_prefix=""):
-    def is_valid():
-        if not isinstance(device, str):
-            return False
-        if device == "cpu" or device == "mps":
-            return True
-        if not device.startswith("cuda:") or not device[5:].isnumeric():
-            return False
-        return True
-
-    if not is_valid():
-        raise EnvironmentError(
-            f"{log_prefix}: device id should be 'cpu', 'mps', or 'cuda:N' (where N is an integer index for the GPU). Got: {device}"
-        )
-
-
-def is_device_compatible(device):
-    """
-    Returns True/False, and prints any compatibility errors
-    """
-    # static variable "history".
-    is_device_compatible.history = getattr(is_device_compatible, "history", {})
-    try:
-        validate_device_id(device, log_prefix="is_device_compatible")
-    except:
-        log.error(str(e))
-        return False
-
-    if device in ("cpu", "mps"):
-        return True
-    # Memory check
-    try:
-        _, mem_total = torch.cuda.mem_get_info(device)
-        mem_total /= float(10**9)
-        if mem_total < 1.9:
-            if is_device_compatible.history.get(device) == None:
-                log.warn(f"GPU {device} with less than 2 GB of VRAM is not compatible with Stable Diffusion")
-                is_device_compatible.history[device] = 1
-            return False
-    except RuntimeError as e:
-        log.error(str(e))
-        return False
-    return True
-
-
 def get_processor_name():
     try:
         import subprocess
@@ -243,7 +157,8 @@ def get_processor_name():
         if platform.system() == "Windows":
             return platform.processor()
         elif platform.system() == "Darwin":
-            os.environ["PATH"] = os.environ["PATH"] + os.pathsep + "/usr/sbin"
+            if "/usr/sbin" not in os.environ["PATH"].split(os.pathsep):
+                os.environ["PATH"] = os.environ["PATH"] + os.pathsep + "/usr/sbin"
             command = "sysctl -n machdep.cpu.brand_string"
             return subprocess.check_output(command, shell=True).decode().strip()
         elif platform.system() == "Linux":

ui/easydiffusion/easydb/schemas.py

@@ -33,4 +33,3 @@ class Bucket(BucketBase):
 
     class Config:
         orm_mode = True
-

ui/easydiffusion/model_manager.py

@@ -3,13 +3,13 @@ import shutil
 from glob import glob
 import traceback
 from typing import Union
+from os import path
 
 from easydiffusion import app
 from easydiffusion.types import ModelsData
 from easydiffusion.utils import log
 from sdkit import Context
-from sdkit.models import load_model, scan_model, unload_model, download_model, get_model_info_from_db
-from sdkit.models.model_loader.controlnet_filters import filters as cn_filters
+from sdkit.models import scan_model, download_model, get_model_info_from_db
 from sdkit.utils import hash_file_quick
 from sdkit.models.model_loader.embeddings import get_embedding_token
 
@@ -23,21 +23,23 @@ KNOWN_MODEL_TYPES = [
     "codeformer",
     "embeddings",
     "controlnet",
+    "text-encoder",
 ]
 MODEL_EXTENSIONS = {
-    "stable-diffusion": [".ckpt", ".safetensors"],
-    "vae": [".vae.pt", ".ckpt", ".safetensors"],
-    "hypernetwork": [".pt", ".safetensors"],
+    "stable-diffusion": [".ckpt", ".safetensors", ".sft", ".gguf"],
+    "vae": [".vae.pt", ".ckpt", ".safetensors", ".sft", ".gguf"],
+    "hypernetwork": [".pt", ".safetensors", ".sft"],
     "gfpgan": [".pth"],
     "realesrgan": [".pth"],
-    "lora": [".ckpt", ".safetensors", ".pt"],
+    "lora": [".ckpt", ".safetensors", ".sft", ".pt"],
     "codeformer": [".pth"],
-    "embeddings": [".pt", ".bin", ".safetensors"],
-    "controlnet": [".pth", ".safetensors"],
+    "embeddings": [".pt", ".bin", ".safetensors", ".sft"],
+    "controlnet": [".pth", ".safetensors", ".sft"],
+    "text-encoder": [".safetensors", ".sft", ".gguf"],
 }
 DEFAULT_MODELS = {
     "stable-diffusion": [
-        {"file_name": "sd-v1-4.ckpt", "model_id": "1.4"},
+        {"file_name": "sd-v1-5.safetensors", "model_id": "1.5-pruned-emaonly-fp16"},
     ],
     "gfpgan": [
         {"file_name": "GFPGANv1.4.pth", "model_id": "1.4"},
@@ -51,6 +53,17 @@ DEFAULT_MODELS = {
     ],
 }
 MODELS_TO_LOAD_ON_START = ["stable-diffusion", "vae", "hypernetwork", "lora"]
+ALTERNATE_FOLDER_NAMES = {  # for WebUI compatibility
+    "stable-diffusion": "Stable-diffusion",
+    "vae": "VAE",
+    "hypernetwork": "hypernetworks",
+    "codeformer": "Codeformer",
+    "gfpgan": "GFPGAN",
+    "realesrgan": "RealESRGAN",
+    "lora": "Lora",
+    "controlnet": "ControlNet",
+    "text-encoder": "text_encoder",
+}
 
 known_models = {}
 
@@ -63,6 +76,7 @@ def init():
 
 def load_default_models(context: Context):
     from easydiffusion import runtime
+    from easydiffusion.backend_manager import backend
 
     runtime.set_vram_optimizations(context)
 
@@ -70,13 +84,13 @@ def load_default_models(context: Context):
     for model_type in MODELS_TO_LOAD_ON_START:
         context.model_paths[model_type] = resolve_model_to_use(model_type=model_type, fail_if_not_found=False)
         try:
-            load_model(
+            backend.load_model(
                 context,
                 model_type,
                 scan_model=context.model_paths[model_type] != None
                 and not context.model_paths[model_type].endswith(".safetensors"),
             )
-            if model_type in context.model_load_errors:
+            if hasattr(context, "model_load_errors") and model_type in context.model_load_errors:
                 del context.model_load_errors[model_type]
         except Exception as e:
             log.error(f"[red]Error while loading {model_type} model: {context.model_paths[model_type]}[/red]")
@@ -88,13 +102,17 @@ def load_default_models(context: Context):
                 log.exception(e)
             del context.model_paths[model_type]
 
+            if not hasattr(context, "model_load_errors"):
+                context.model_load_errors = {}
             context.model_load_errors[model_type] = str(e)  # storing the entire Exception can lead to memory leaks
 
 
 def unload_all(context: Context):
+    from easydiffusion.backend_manager import backend
+
     for model_type in KNOWN_MODEL_TYPES:
-        unload_model(context, model_type)
-        if model_type in context.model_load_errors:
+        backend.unload_model(context, model_type)
+        if hasattr(context, "model_load_errors") and model_type in context.model_load_errors:
             del context.model_load_errors[model_type]
 
 
@@ -119,33 +137,33 @@ def resolve_model_to_use_single(model_name: str = None, model_type: str = None,
     default_models = DEFAULT_MODELS.get(model_type, [])
     config = app.getConfig()
 
-    model_dir = os.path.join(app.MODELS_DIR, model_type)
     if not model_name:  # When None try user configured model.
         # config = getConfig()
         if "model" in config and model_type in config["model"]:
             model_name = config["model"][model_type]
 
-    if model_name:
-        # Check models directory
-        model_path = os.path.join(model_dir, model_name)
-        if os.path.exists(model_path):
-            return model_path
-        for model_extension in model_extensions:
-            if os.path.exists(model_path + model_extension):
-                return model_path + model_extension
-            if os.path.exists(model_name + model_extension):
-                return os.path.abspath(model_name + model_extension)
+    for model_dir in get_model_dirs(model_type):
+        if model_name:
+            # Check models directory
+            model_path = os.path.join(model_dir, model_name)
+            if os.path.exists(model_path):
+                return model_path
+            for model_extension in model_extensions:
+                if os.path.exists(model_path + model_extension):
+                    return model_path + model_extension
+                if os.path.exists(model_name + model_extension):
+                    return os.path.abspath(model_name + model_extension)
 
-    # Can't find requested model, check the default paths.
-    if model_type == "stable-diffusion" and not fail_if_not_found:
-        for default_model in default_models:
-            default_model_path = os.path.join(model_dir, default_model["file_name"])
-            if os.path.exists(default_model_path):
-                if model_name is not None:
-                    log.warn(
-                        f"Could not find the configured custom model {model_name}. Using the default one: {default_model_path}"
-                    )
-                return default_model_path
+        # Can't find requested model, check the default paths.
+        if model_type == "stable-diffusion" and not fail_if_not_found:
+            for default_model in default_models:
+                default_model_path = os.path.join(model_dir, default_model["file_name"])
+                if os.path.exists(default_model_path):
+                    if model_name is not None:
+                        log.warn(
+                            f"Could not find the configured custom model {model_name}. Using the default one: {default_model_path}"
+                        )
+                    return default_model_path
 
     if model_name and fail_if_not_found:
         raise FileNotFoundError(
@@ -154,6 +172,8 @@ def resolve_model_to_use_single(model_name: str = None, model_type: str = None,
 
 
 def reload_models_if_necessary(context: Context, models_data: ModelsData, models_to_force_reload: list = []):
+    from easydiffusion.backend_manager import backend
+
     models_to_reload = {
         model_type: path
         for model_type, path in models_data.model_paths.items()
@@ -175,39 +195,68 @@ def reload_models_if_necessary(context: Context, models_data: ModelsData, models
     for model_type, model_path_in_req in models_to_reload.items():
         context.model_paths[model_type] = model_path_in_req
 
-        action_fn = unload_model if context.model_paths[model_type] is None else load_model
+        action_fn = backend.unload_model if context.model_paths[model_type] is None else backend.load_model
         extra_params = models_data.model_params.get(model_type, {})
         try:
             action_fn(context, model_type, scan_model=False, **extra_params)  # we've scanned them already
-            if model_type in context.model_load_errors:
+            if hasattr(context, "model_load_errors") and model_type in context.model_load_errors:
                 del context.model_load_errors[model_type]
         except Exception as e:
             log.exception(e)
-            if action_fn == load_model:
+            if action_fn == backend.load_model:
+                if not hasattr(context, "model_load_errors"):
+                    context.model_load_errors = {}
                 context.model_load_errors[model_type] = str(e)  # storing the entire Exception can lead to memory leaks
 
+    if hasattr(backend, "flush_model_changes"):
+        backend.flush_model_changes(context)
+
 
 def resolve_model_paths(models_data: ModelsData):
+    from easydiffusion.backend_manager import backend
+
+    cn_filters = backend.list_controlnet_filters()
+
     model_paths = models_data.model_paths
+    skip_models = cn_filters + [
+        "latent_upscaler",
+        "nsfw_checker",
+        "esrgan_4x",
+        "lanczos",
+        "nearest",
+        "scunet",
+        "swinir",
+    ]
+
     for model_type in model_paths:
-        skip_models = cn_filters + ["latent_upscaler", "nsfw_checker"]
         if model_type in skip_models:  # doesn't use model paths
             continue
-        if model_type == "codeformer":
-            download_if_necessary("codeformer", "codeformer.pth", "codeformer-0.1.0")
-        elif model_type == "controlnet":
-            model_id = model_paths[model_type]
-            model_info = get_model_info_from_db(model_type=model_type, model_id=model_id)
-            if model_info:
-                filename = model_info.get("url", "").split("/")[-1]
-                download_if_necessary("controlnet", filename, model_id, skip_if_others_exist=False)
+
+        if model_type in ("vae", "codeformer", "controlnet", "text-encoder") and model_paths[model_type]:
+            model_ids = model_paths[model_type]
+            model_ids = model_ids if isinstance(model_ids, list) else [model_ids]
+
+            new_model_paths = []
+
+            for model_id in model_ids:
+                # log.info(f"Checking for {model_id=}")
+                model_info = get_model_info_from_db(model_type=model_type, model_id=model_id)
+                if model_info:
+                    filename = model_info.get("url", "").split("/")[-1]
+                    download_if_necessary(model_type, filename, model_id, skip_if_others_exist=False)
+
+                    new_model_paths.append(path.splitext(filename)[0])
+                else:  # not in the model db, probably a regular file
+                    new_model_paths.append(model_id)
+
+            model_paths[model_type] = new_model_paths
 
         model_paths[model_type] = resolve_model_to_use(model_paths[model_type], model_type=model_type)
 
 
 def fail_if_models_did_not_load(context: Context):
     for model_type in KNOWN_MODEL_TYPES:
-        if model_type in context.model_load_errors:
+        if hasattr(context, "model_load_errors") and model_type in context.model_load_errors:
             e = context.model_load_errors[model_type]
             raise Exception(f"Could not load the {model_type} model! Reason: " + e)
 
@@ -225,16 +274,31 @@ def download_default_models_if_necessary():
 
 
 def download_if_necessary(model_type: str, file_name: str, model_id: str, skip_if_others_exist=True):
-    model_path = os.path.join(app.MODELS_DIR, model_type, file_name)
+    from easydiffusion.backend_manager import backend
+
     expected_hash = get_model_info_from_db(model_type=model_type, model_id=model_id)["quick_hash"]
-
     other_models_exist = any_model_exists(model_type) and skip_if_others_exist
-    known_model_exists = os.path.exists(model_path)
-    known_model_is_corrupt = known_model_exists and hash_file_quick(model_path) != expected_hash
 
-    if known_model_is_corrupt or (not other_models_exist and not known_model_exists):
-        print("> download", model_type, model_id)
-        download_model(model_type, model_id, download_base_dir=app.MODELS_DIR, download_config_if_available=False)
+    for model_dir in get_model_dirs(model_type):
+        model_path = os.path.join(model_dir, file_name)
+
+        known_model_exists = os.path.exists(model_path)
+        known_model_is_corrupt = known_model_exists and hash_file_quick(model_path) != expected_hash
+
+        needs_download = known_model_is_corrupt or (not other_models_exist and not known_model_exists)
+
+        # log.info(f"{model_path=} {needs_download=}")
+        # if known_model_exists:
+        #     log.info(f"{expected_hash=} {hash_file_quick(model_path)=}")
+        # log.info(f"{known_model_is_corrupt=} {other_models_exist=} {known_model_exists=}")
+
+        if not needs_download:
+            return
+
+    print("> download", model_type, model_id)
+    download_model(model_type, model_id, download_base_dir=app.MODELS_DIR, download_config_if_available=False)
+
+    backend.refresh_models()
 
 
 def migrate_legacy_model_location():
@@ -245,34 +309,55 @@ def migrate_legacy_model_location():
             file_name = model["file_name"]
             legacy_path = os.path.join(app.SD_DIR, file_name)
             if os.path.exists(legacy_path):
-                shutil.move(legacy_path, os.path.join(app.MODELS_DIR, model_type, file_name))
+                model_dir = get_model_dirs(model_type)[0]
+                shutil.move(legacy_path, os.path.join(model_dir, file_name))
 
 
 def any_model_exists(model_type: str) -> bool:
     extensions = MODEL_EXTENSIONS.get(model_type, [])
-    for ext in extensions:
-        if any(glob(f"{app.MODELS_DIR}/{model_type}/**/*{ext}", recursive=True)):
-            return True
+    for model_dir in get_model_dirs(model_type):
+        for ext in extensions:
+            if any(glob(f"{model_dir}/**/*{ext}", recursive=True)):
+                return True
 
     return False
 
 
 def make_model_folders():
     for model_type in KNOWN_MODEL_TYPES:
-        model_dir_path = os.path.join(app.MODELS_DIR, model_type)
+        model_dir_path = get_model_dirs(model_type)[0]
 
-        os.makedirs(model_dir_path, exist_ok=True)
+        try:
+            os.makedirs(model_dir_path, exist_ok=True)
+        except Exception as e:
+            from rich.console import Console
+            from rich.panel import Panel
+
+            Console().print(
+                Panel(
+                    "\n"
+                    + f"Error while creating the models directory: '{model_dir_path}'\n"
+                    + f"Error: {e}\n\n"
+                    + f"[white]Check the 'models_dir:' line in the file '{os.path.join(app.ROOT_DIR, 'config.yaml')}'.[/white]\n",
+                    title="Fatal Error starting Easy Diffusion",
+                    style="bold yellow on red",
+                )
+            )
+            input("Press Enter to terminate...")
+            exit(1)
 
         help_file_name = f"Place your {model_type} model files here.txt"
         help_file_contents = f'Supported extensions: {" or ".join(MODEL_EXTENSIONS.get(model_type))}'
-
-        with open(os.path.join(model_dir_path, help_file_name), "w", encoding="utf-8") as f:
-            f.write(help_file_contents)
+        try:
+            with open(os.path.join(model_dir_path, help_file_name), "w", encoding="utf-8") as f:
+                f.write(help_file_contents)
+        except Exception as e:
+            log.exception(e)
 
 
 def is_malicious_model(file_path):
     try:
-        if file_path.endswith(".safetensors"):
+        if file_path.endswith((".safetensors", ".sft", ".gguf")):
             return False
         scan_result = scan_model(file_path)
         if scan_result.issues_count > 0 or scan_result.infected_files > 0:
@@ -303,28 +388,37 @@ def is_malicious_model(file_path):
 
 
 def getModels(scan_for_malicious: bool = True):
+    from easydiffusion.backend_manager import backend
+
+    backend.refresh_models()
+
     models = {
         "options": {
-            "stable-diffusion": [{"sd-v1-4": "SD 1.4"}],
-            "vae": [],
+            "stable-diffusion": [],
+            "vae": [{"ae": "ae (Flux VAE fp16)"}],
             "hypernetwork": [],
             "lora": [],
             "codeformer": [{"codeformer": "CodeFormer"}],
             "embeddings": [],
             "controlnet": [
-                {"control_v11p_sd15_canny": "Canny (*)"},
-                {"control_v11p_sd15_openpose": "OpenPose (*)"},
-                {"control_v11p_sd15_normalbae": "Normal BAE (*)"},
-                {"control_v11f1p_sd15_depth": "Depth (*)"},
-                {"control_v11p_sd15_scribble": "Scribble"},
-                {"control_v11p_sd15_softedge": "Soft Edge"},
-                {"control_v11p_sd15_inpaint": "Inpaint"},
-                {"control_v11p_sd15_lineart": "Line Art"},
-                {"control_v11p_sd15s2_lineart_anime": "Line Art Anime"},
-                {"control_v11p_sd15_mlsd": "Straight Lines"},
-                {"control_v11p_sd15_seg": "Segment"},
-                {"control_v11e_sd15_shuffle": "Shuffle"},
-                {"control_v11f1e_sd15_tile": "Tile"},
+                # {"control_v11p_sd15_canny": "Canny (*)"},
+                # {"control_v11p_sd15_openpose": "OpenPose (*)"},
+                # {"control_v11p_sd15_normalbae": "Normal BAE (*)"},
+                # {"control_v11f1p_sd15_depth": "Depth (*)"},
+                # {"control_v11p_sd15_scribble": "Scribble"},
+                # {"control_v11p_sd15_softedge": "Soft Edge"},
+                # {"control_v11p_sd15_inpaint": "Inpaint"},
+                # {"control_v11p_sd15_lineart": "Line Art"},
+                # {"control_v11p_sd15s2_lineart_anime": "Line Art Anime"},
+                # {"control_v11p_sd15_mlsd": "Straight Lines"},
+                # {"control_v11p_sd15_seg": "Segment"},
+                # {"control_v11e_sd15_shuffle": "Shuffle"},
+                # {"control_v11f1e_sd15_tile": "Tile"},
+            ],
+            "text-encoder": [
+                {"t5xxl_fp16": "T5 XXL fp16"},
+                {"clip_l": "CLIP L"},
+                {"clip_g": "CLIP G"},
             ],
         },
     }
@@ -339,6 +433,9 @@ def getModels(scan_for_malicious: bool = True):
 
         tree = list(default_entries)
 
+        if not os.path.exists(directory):
+            return tree
+
         for entry in sorted(
             os.scandir(directory),
             key=lambda entry: (entry.is_file() == directoriesFirst, entry.name.lower()),
@@ -361,6 +458,8 @@ def getModels(scan_for_malicious: bool = True):
                 model_id = entry.name[: -len(matching_suffix)]
                 if callable(nameFilter):
                     model_id = nameFilter(model_id)
+                    if model_id is None:
+                        continue
 
                 model_exists = False
                 for m in tree:  # allows default "named" models, like CodeFormer and known ControlNet models
@@ -381,17 +480,18 @@ def getModels(scan_for_malicious: bool = True):
         nonlocal models_scanned
 
         model_extensions = MODEL_EXTENSIONS.get(model_type, [])
-        models_dir = os.path.join(app.MODELS_DIR, model_type)
-        if not os.path.exists(models_dir):
-            os.makedirs(models_dir)
+        models_dirs = get_model_dirs(model_type)
+        if not os.path.exists(models_dirs[0]):
+            os.makedirs(models_dirs[0])
 
-        try:
-            default_tree = models["options"].get(model_type, [])
-            models["options"][model_type] = scan_directory(
-                models_dir, model_extensions, default_entries=default_tree, nameFilter=nameFilter
-            )
-        except MaliciousModelException as e:
-            models["scan-error"] = str(e)
+        for model_dir in models_dirs:
+            try:
+                default_tree = models["options"].get(model_type, [])
+                models["options"][model_type] = scan_directory(
+                    model_dir, model_extensions, default_entries=default_tree, nameFilter=nameFilter
+                )
+            except MaliciousModelException as e:
+                models["scan-error"] = str(e)
 
     if scan_for_malicious:
         log.info(f"[green]Scanning all model folders for models...[/]")
@@ -399,12 +499,30 @@ def getModels(scan_for_malicious: bool = True):
     listModels(model_type="stable-diffusion")
     listModels(model_type="vae")
     listModels(model_type="hypernetwork")
-    listModels(model_type="gfpgan")
+    listModels(model_type="gfpgan", nameFilter=lambda x: (x if "gfpgan" in x.lower() else None))
     listModels(model_type="lora")
     listModels(model_type="embeddings", nameFilter=get_embedding_token)
     listModels(model_type="controlnet")
+    listModels(model_type="text-encoder")
 
     if scan_for_malicious and models_scanned > 0:
         log.info(f"[green]Scanned {models_scanned} models. Nothing infected[/]")
 
     return models
+
+
+def get_model_dirs(model_type: str, base_dir=None):
+    "Returns the possible model directory paths for the given model type. Mainly used for WebUI compatibility"
+
+    if base_dir is None:
+        base_dir = app.MODELS_DIR
+
+    dirs = [os.path.join(base_dir, model_type)]
+
+    if model_type in ALTERNATE_FOLDER_NAMES:
+        alt_dir = ALTERNATE_FOLDER_NAMES[model_type]
+        alt_dir = os.path.join(base_dir, alt_dir)
+        if os.path.exists(alt_dir) and os.path.isdir(alt_dir):
+            dirs.append(alt_dir)
+
+    return dirs

ui/easydiffusion/package_manager.py

@@ -3,18 +3,17 @@ import os
 import platform
 from importlib.metadata import version as pkg_version
 
-from sdkit.utils import log
-
 from easydiffusion import app
 
-# future home of scripts/check_modules.py
+# was meant to be a rewrite of scripts/check_modules.py
+# but probably dead for now
 
 manifest = {
     "tensorrt": {
         "install": [
-            "nvidia-cudnn --pre --extra-index-url=https://pypi.nvidia.com --trusted-host pypi.nvidia.com",
-            "tensorrt-libs --pre --extra-index-url=https://pypi.nvidia.com --trusted-host pypi.nvidia.com",
-            "tensorrt --pre --extra-index-url=https://pypi.nvidia.com --trusted-host pypi.nvidia.com",
+            "wheel",
+            "nvidia-cudnn-cu11==8.9.4.25",
+            "tensorrt==9.0.0.post11.dev1 --pre --extra-index-url=https://pypi.nvidia.com --trusted-host pypi.nvidia.com",
         ],
         "uninstall": ["tensorrt"],
         # TODO also uninstall tensorrt-libs and nvidia-cudnn, but do it upon restarting (avoid 'file in use' error)
@@ -50,6 +49,8 @@ def is_installed(module_name) -> bool:
 
 
 def install(module_name):
+    from easydiffusion.utils import log
+
     if is_installed(module_name):
         log.info(f"{module_name} has already been installed!")
         return
@@ -79,6 +80,8 @@ def install(module_name):
 
 
 def uninstall(module_name):
+    from easydiffusion.utils import log
+
     if not is_installed(module_name):
         log.info(f"{module_name} hasn't been installed!")
         return

ui/easydiffusion/runtime.py

@@ -1,4 +1,5 @@
 """
+(OUTDATED DOC)
 A runtime that runs on a specific device (in a thread).
 
 It can run various tasks like image generation, image filtering, model merge etc by using that thread-local context.
@@ -6,45 +7,38 @@ It can run various tasks like image generation, image filtering, model merge etc
 This creates an `sdkit.Context` that's bound to the device specified while calling the `init()` function.
 """
 
-from easydiffusion import device_manager
-from easydiffusion.utils import log
-from sdkit import Context
-from sdkit.utils import get_device_usage
-
-context = Context()  # thread-local
-"""
-runtime data (bound locally to this thread), for e.g. device, references to loaded models, optimization flags etc
-"""
+context = None
 
 
 def init(device):
     """
     Initializes the fields that will be bound to this runtime's context, and sets the current torch device
     """
+
+    global context
+
+    from easydiffusion import device_manager
+    from easydiffusion.backend_manager import backend
+    from easydiffusion.app import getConfig
+
+    context = backend.create_context()
+
     context.stop_processing = False
     context.temp_images = {}
     context.partial_x_samples = None
     context.model_load_errors = {}
     context.enable_codeformer = True
 
-    from easydiffusion import app
-
-    app_config = app.getConfig()
-    context.test_diffusers = app_config.get("test_diffusers", True)
-
-    log.info("Device usage during initialization:")
-    get_device_usage(device, log_info=True, process_usage_only=False)
-
     device_manager.device_init(context, device)
 
 
-def set_vram_optimizations(context: Context):
-    from easydiffusion import app
+def set_vram_optimizations(context):
+    from easydiffusion.app import getConfig
 
-    config = app.getConfig()
+    config = getConfig()
     vram_usage_level = config.get("vram_usage_level", "balanced")
 
-    if vram_usage_level != context.vram_usage_level:
+    if hasattr(context, "vram_usage_level") and vram_usage_level != context.vram_usage_level:
         context.vram_usage_level = vram_usage_level
         return True
 

ui/easydiffusion/server.py

@@ -2,6 +2,7 @@
 Notes:
     async endpoints always run on the main thread. Without they run on the thread pool.
 """
+
 import datetime
 import mimetypes
 import os
@@ -15,9 +16,12 @@ from easydiffusion.types import (
     FilterImageRequest,
     MergeRequest,
     TaskData,
+    RenderTaskData,
     ModelsData,
     OutputFormatData,
+    SaveToDiskData,
     convert_legacy_render_req_to_new,
+    convert_legacy_controlnet_filter_name,
 )
 from easydiffusion.utils import log
 from fastapi import FastAPI, HTTPException
@@ -36,6 +40,7 @@ NOCACHE_HEADERS = {
     "Pragma": "no-cache",
     "Expires": "0",
 }
+PROTECTED_CONFIG_KEYS = ("block_nsfw",)  # can't change these via the HTTP API
 
 
 class NoCacheStaticFiles(StaticFiles):
@@ -63,7 +68,10 @@ class SetAppConfigRequest(BaseModel, extra=Extra.allow):
     ui_open_browser_on_start: bool = None
     listen_to_network: bool = None
     listen_port: int = None
-    test_diffusers: bool = True
+    use_v3_engine: bool = True
+    backend: str = "ed_diffusers"
+    models_dir: str = None
+    vram_usage_level: str = "balanced"
 
 
 def init():
@@ -151,6 +159,12 @@ def init():
     def shutdown_event():  # Signal render thread to close on shutdown
         task_manager.current_state_error = SystemExit("Application shutting down.")
 
+    @server_api.on_event("startup")
+    def start_event():
+        from easydiffusion.app import open_browser
+
+        open_browser()
+
 
 # API implementations
 def set_app_config_internal(req: SetAppConfigRequest):
@@ -172,10 +186,13 @@ def set_app_config_internal(req: SetAppConfigRequest):
             config["net"] = {}
         config["net"]["listen_port"] = int(req.listen_port)
 
-    config["test_diffusers"] = req.test_diffusers
+    config["use_v3_engine"] = req.backend == "ed_diffusers"
+    config["backend"] = req.backend
+    config["models_dir"] = req.models_dir
+    config["vram_usage_level"] = req.vram_usage_level
 
     for property, property_value in req.dict().items():
-        if property_value is not None and property not in req.__fields__:
+        if property_value is not None and property not in req.__fields__ and property not in PROTECTED_CONFIG_KEYS:
             config[property] = property_value
 
     try:
@@ -191,11 +208,13 @@ def set_app_config_internal(req: SetAppConfigRequest):
 
 
 def update_render_devices_in_config(config, render_devices):
-    if render_devices not in ("cpu", "auto") and not render_devices.startswith("cuda:"):
-        raise HTTPException(status_code=400, detail=f"Invalid render device requested: {render_devices}")
+    from easydiffusion.device_manager import validate_render_devices
 
-    if render_devices.startswith("cuda:"):
+    try:
         render_devices = render_devices.split(",")
+        validate_render_devices(render_devices)
+    except ValueError as e:
+        raise HTTPException(status_code=400, detail=str(e))
 
     config["render_devices"] = render_devices
 
@@ -204,8 +223,15 @@ def read_web_data_internal(key: str = None, **kwargs):
     if not key:  # /get without parameters, stable-diffusion easter egg.
         raise HTTPException(status_code=418, detail="StableDiffusion is drawing a teapot!")  # HTTP418 I'm a teapot
     elif key == "app_config":
-        return JSONResponse(app.getConfig(), headers=NOCACHE_HEADERS)
+        config = app.getConfig()
+
+        if "models_dir" not in config:
+            config["models_dir"] = app.MODELS_DIR
+
+        return JSONResponse(config, headers=NOCACHE_HEADERS)
     elif key == "system_info":
+        from easydiffusion.backend_manager import backend
+
         config = app.getConfig()
 
         output_dir = config.get("force_save_path", os.path.join(os.path.expanduser("~"), app.OUTPUT_DIRNAME))
@@ -215,6 +241,8 @@ def read_web_data_internal(key: str = None, **kwargs):
             "hosts": app.getIPConfig(),
             "default_output_dir": output_dir,
             "enforce_output_dir": ("force_save_path" in config),
+            "enforce_output_metadata": ("force_save_metadata" in config),
+            "backend_url": backend.get_url(),
         }
         system_info["devices"]["config"] = config.get("render_devices", "auto")
         return JSONResponse(system_info, headers=NOCACHE_HEADERS)
@@ -261,14 +289,15 @@ def render_internal(req: dict):
 
         # separate out the request data into rendering and task-specific data
         render_req: GenerateImageRequest = GenerateImageRequest.parse_obj(req)
-        task_data: TaskData = TaskData.parse_obj(req)
+        task_data: RenderTaskData = RenderTaskData.parse_obj(req)
         models_data: ModelsData = ModelsData.parse_obj(req)
         output_format: OutputFormatData = OutputFormatData.parse_obj(req)
+        save_data: SaveToDiskData = SaveToDiskData.parse_obj(req)
 
         # Overwrite user specified save path
         config = app.getConfig()
         if "force_save_path" in config:
-            task_data.save_to_disk_path = config["force_save_path"]
+            save_data.save_to_disk_path = config["force_save_path"]
 
         render_req.init_image_mask = req.get("mask")  # hack: will rename this in the HTTP API in a future revision
 
@@ -280,7 +309,7 @@ def render_internal(req: dict):
         )
 
         # enqueue the task
-        task = RenderTask(render_req, task_data, models_data, output_format)
+        task = RenderTask(render_req, task_data, models_data, output_format, save_data)
         return enqueue_task(task)
     except HTTPException as e:
         raise e
@@ -291,13 +320,23 @@ def render_internal(req: dict):
 
 def filter_internal(req: dict):
     try:
-        session_id = req.get("session_id", "session")
         filter_req: FilterImageRequest = FilterImageRequest.parse_obj(req)
+        task_data: TaskData = TaskData.parse_obj(req)
         models_data: ModelsData = ModelsData.parse_obj(req)
         output_format: OutputFormatData = OutputFormatData.parse_obj(req)
+        save_data: SaveToDiskData = SaveToDiskData.parse_obj(req)
+
+        filter_req.filter = convert_legacy_controlnet_filter_name(filter_req.filter)
+
+        for model_name in ("realesrgan", "esrgan_4x", "lanczos", "nearest", "scunet", "swinir"):
+            if models_data.model_paths.get(model_name):
+                if model_name not in filter_req.filter_params:
+                    filter_req.filter_params[model_name] = {}
+
+                filter_req.filter_params[model_name]["upscaler"] = models_data.model_paths[model_name]
 
         # enqueue the task
-        task = FilterTask(filter_req, session_id, models_data, output_format)
+        task = FilterTask(filter_req, task_data, models_data, output_format, save_data)
         return enqueue_task(task)
     except HTTPException as e:
         raise e
@@ -329,15 +368,13 @@ def model_merge_internal(req: dict):
 
         mergeReq: MergeRequest = MergeRequest.parse_obj(req)
 
+        sd_model_dir = model_manager.get_model_dirs("stable-diffusion")[0]
+
         merge_models(
             model_manager.resolve_model_to_use(mergeReq.model0, "stable-diffusion"),
             model_manager.resolve_model_to_use(mergeReq.model1, "stable-diffusion"),
             mergeReq.ratio,
-            os.path.join(
-                app.MODELS_DIR,
-                "stable-diffusion",
-                filename_regex.sub("_", mergeReq.out_path),
-            ),
+            os.path.join(sd_model_dir, filename_regex.sub("_", mergeReq.out_path)),
             mergeReq.use_fp16,
         )
         return JSONResponse({"status": "OK"}, headers=NOCACHE_HEADERS)
@@ -456,8 +493,8 @@ def modify_package_internal(package_name: str, req: dict):
         log.error(traceback.format_exc())
         return HTTPException(status_code=500, detail=str(e))
 
+
 def get_sha256_internal(obj_path):
-    import hashlib
     from easydiffusion.utils import sha256sum
 
     path = obj_path.split("/")
@@ -477,4 +514,3 @@ def get_sha256_internal(obj_path):
         log.error(str(e))
         log.error(traceback.format_exc())
         return HTTPException(status_code=500, detail=str(e))
-

ui/easydiffusion/task_manager.py

@@ -4,6 +4,7 @@ Notes:
     Use weak_thread_data to store all other data using weak keys.
     This will allow for garbage collection after the thread dies.
 """
+
 import json
 import traceback
 
@@ -19,7 +20,9 @@ import torch
 from easydiffusion import device_manager
 from easydiffusion.tasks import Task
 from easydiffusion.utils import log
-from sdkit.utils import gc
+
+from torchruntime.utils import get_device_count, get_device, get_device_name, get_installed_torch_platform
+from sdkit.utils import is_cpu_device, mem_get_info
 
 THREAD_NAME_PREFIX = ""
 ERR_LOCK_FAILED = " failed to acquire lock within timeout."
@@ -233,6 +236,8 @@ def thread_render(device):
     global current_state, current_state_error
 
     from easydiffusion import model_manager, runtime
+    from easydiffusion.backend_manager import backend
+    from requests import ConnectionError
 
     try:
         runtime.init(device)
@@ -244,8 +249,17 @@ def thread_render(device):
         }
 
         current_state = ServerStates.LoadingModel
-        model_manager.load_default_models(runtime.context)
 
+        while True:
+            try:
+                if backend.ping(timeout=1):
+                    break
+
+                time.sleep(1)
+            except (TimeoutError, ConnectionError):
+                time.sleep(1)
+
+        model_manager.load_default_models(runtime.context)
         current_state = ServerStates.Online
     except Exception as e:
         log.error(traceback.format_exc())
@@ -291,7 +305,6 @@ def thread_render(device):
             task.buffer_queue.put(json.dumps(task.response))
             log.error(traceback.format_exc())
         finally:
-            gc(runtime.context)
             task.lock.release()
 
         keep_task_alive(task)
@@ -329,34 +342,33 @@ def get_devices():
         "active": {},
     }
 
-    def get_device_info(device):
-        if device in ("cpu", "mps"):
+    def get_device_info(device_id):
+        if is_cpu_device(device_id):
             return {"name": device_manager.get_processor_name()}
 
-        mem_free, mem_total = torch.cuda.mem_get_info(device)
+        device = get_device(device_id)
+
+        mem_free, mem_total = mem_get_info(device)
         mem_free /= float(10**9)
         mem_total /= float(10**9)
 
         return {
-            "name": torch.cuda.get_device_name(device),
+            "name": get_device_name(device),
             "mem_free": mem_free,
             "mem_total": mem_total,
             "max_vram_usage_level": device_manager.get_max_vram_usage_level(device),
         }
 
     # list the compatible devices
-    cuda_count = torch.cuda.device_count()
-    for device in range(cuda_count):
-        device = f"cuda:{device}"
-        if not device_manager.is_device_compatible(device):
-            continue
+    torch_platform_name = get_installed_torch_platform()[0]
+    device_count = get_device_count()
+    for device_id in range(device_count):
+        device_id = f"{torch_platform_name}:{device_id}" if device_count > 1 else torch_platform_name
 
-        devices["all"].update({device: get_device_info(device)})
+        devices["all"].update({device_id: get_device_info(device_id)})
 
-    if device_manager.is_mps_available():
-        devices["all"].update({"mps": get_device_info("mps")})
-
-    devices["all"].update({"cpu": get_device_info("cpu")})
+    if torch_platform_name != "cpu":
+        devices["all"].update({"cpu": get_device_info("cpu")})
 
     # list the activated devices
     if not manager_lock.acquire(blocking=True, timeout=LOCK_TIMEOUT):
@@ -368,8 +380,8 @@ def get_devices():
             weak_data = weak_thread_data.get(rthread)
             if not weak_data or not "device" in weak_data or not "device_name" in weak_data:
                 continue
-            device = weak_data["device"]
-            devices["active"].update({device: get_device_info(device)})
+            device_id = weak_data["device"]
+            devices["active"].update({device_id: get_device_info(device_id)})
     finally:
         manager_lock.release()
 
@@ -427,12 +439,6 @@ def start_render_thread(device):
 
 
 def stop_render_thread(device):
-    try:
-        device_manager.validate_device_id(device, log_prefix="stop_render_thread")
-    except:
-        log.error(traceback.format_exc())
-        return False
-
     if not manager_lock.acquire(blocking=True, timeout=LOCK_TIMEOUT):
         raise Exception("stop_render_thread" + ERR_LOCK_FAILED)
     log.info(f"Stopping Rendering Thread on device: {device}")

ui/easydiffusion/tasks/filter_images.py

@@ -1,12 +1,24 @@
+import os
 import json
 import pprint
+import time
 
-from sdkit.filter import apply_filters
-from sdkit.models import load_model
-from sdkit.utils import img_to_base64_str, get_image, log
+from numpy import base_repr
+
+from sdkit.utils import img_to_base64_str, log, save_images, base64_str_to_img
 
 from easydiffusion import model_manager, runtime
-from easydiffusion.types import FilterImageRequest, FilterImageResponse, ModelsData, OutputFormatData
+from easydiffusion.types import (
+    FilterImageRequest,
+    FilterImageResponse,
+    ModelsData,
+    OutputFormatData,
+    SaveToDiskData,
+    TaskData,
+    GenerateImageRequest,
+)
+from easydiffusion.utils import filter_nsfw
+from easydiffusion.utils.save_utils import format_folder_name
 
 from .task import Task
 
@@ -15,17 +27,28 @@ class FilterTask(Task):
     "For applying filters to input images"
 
     def __init__(
-        self, req: FilterImageRequest, session_id: str, models_data: ModelsData, output_format: OutputFormatData
+        self,
+        req: FilterImageRequest,
+        task_data: TaskData,
+        models_data: ModelsData,
+        output_format: OutputFormatData,
+        save_data: SaveToDiskData,
     ):
-        super().__init__(session_id)
+        super().__init__(task_data.session_id)
+
+        task_data.request_id = self.id
 
         self.request = req
+        self.task_data = task_data
         self.models_data = models_data
         self.output_format = output_format
+        self.save_data = save_data
 
         # convert to multi-filter format, if necessary
         if isinstance(req.filter, str):
-            req.filter_params = {req.filter: req.filter_params}
+            if req.filter not in req.filter_params:
+                req.filter_params = {req.filter: req.filter_params}
+
             req.filter = [req.filter]
 
         if not isinstance(req.image, list):
@@ -34,82 +57,73 @@ class FilterTask(Task):
     def run(self):
         "Runs the image filtering task on the assigned thread"
 
+        from easydiffusion import app
+        from easydiffusion.backend_manager import backend
+
         context = runtime.context
 
         model_manager.resolve_model_paths(self.models_data)
         model_manager.reload_models_if_necessary(context, self.models_data)
         model_manager.fail_if_models_did_not_load(context)
 
-        print_task_info(self.request, self.models_data, self.output_format)
+        print_task_info(self.request, self.models_data, self.output_format, self.save_data)
 
-        if isinstance(self.request.image, list):
-            images = [get_image(img) for img in self.request.image]
-        else:
-            images = get_image(self.request.image)
-
-        images = filter_images(context, images, self.request.filter, self.request.filter_params)
+        has_nsfw_filter = "nsfw_filter" in self.request.filter
 
         output_format = self.output_format
-        images = [
-            img_to_base64_str(
-                img, output_format.output_format, output_format.output_quality, output_format.output_lossless
+
+        backend.set_options(
+            context,
+            output_format=output_format.output_format,
+            output_quality=output_format.output_quality,
+            output_lossless=output_format.output_lossless,
+        )
+
+        images = backend.filter_images(
+            context, self.request.image, self.request.filter, self.request.filter_params, input_type="base64"
+        )
+
+        if has_nsfw_filter:
+            images = filter_nsfw(images)
+
+        if self.save_data.save_to_disk_path is not None:
+            app_config = app.getConfig()
+            folder_format = app_config.get("folder_format", "$id")
+
+            dummy_req = GenerateImageRequest()
+            img_id = base_repr(int(time.time() * 10000), 36)[-7:]  # Base 36 conversion, 0-9, A-Z
+
+            save_dir_path = os.path.join(
+                self.save_data.save_to_disk_path, format_folder_name(folder_format, dummy_req, self.task_data)
+            )
+            images_pil = [base64_str_to_img(img) for img in images]
+            save_images(
+                images_pil,
+                save_dir_path,
+                file_name=img_id,
+                output_format=output_format.output_format,
+                output_quality=output_format.output_quality,
+                output_lossless=output_format.output_lossless,
             )
-            for img in images
-        ]
 
         res = FilterImageResponse(self.request, self.models_data, images=images)
         res = res.json()
         self.buffer_queue.put(json.dumps(res))
+
         log.info("Filter task completed")
 
         self.response = res
 
 
-def filter_images(context, images, filters, filter_params={}):
-    filters = filters if isinstance(filters, list) else [filters]
-
-    for filter_name in filters:
-        params = filter_params.get(filter_name, {})
-
-        previous_state = before_filter(context, filter_name, params)
-
-        try:
-            images = apply_filters(context, filter_name, images, **params)
-        finally:
-            after_filter(context, filter_name, params, previous_state)
-
-    return images
-
-
-def before_filter(context, filter_name, filter_params):
-    if filter_name == "codeformer":
-        from easydiffusion.model_manager import DEFAULT_MODELS, resolve_model_to_use
-
-        default_realesrgan = DEFAULT_MODELS["realesrgan"][0]["file_name"]
-        prev_realesrgan_path = None
-
-        upscale_faces = filter_params.get("upscale_faces", False)
-        if upscale_faces and default_realesrgan not in context.model_paths["realesrgan"]:
-            prev_realesrgan_path = context.model_paths.get("realesrgan")
-            context.model_paths["realesrgan"] = resolve_model_to_use(default_realesrgan, "realesrgan")
-            load_model(context, "realesrgan")
-
-        return prev_realesrgan_path
-
-
-def after_filter(context, filter_name, filter_params, previous_state):
-    if filter_name == "codeformer":
-        prev_realesrgan_path = previous_state
-        if prev_realesrgan_path:
-            context.model_paths["realesrgan"] = prev_realesrgan_path
-            load_model(context, "realesrgan")
-
-
-def print_task_info(req: FilterImageRequest, models_data: ModelsData, output_format: OutputFormatData):
+def print_task_info(
+    req: FilterImageRequest, models_data: ModelsData, output_format: OutputFormatData, save_data: SaveToDiskData
+):
     req_str = pprint.pformat({"filter": req.filter, "filter_params": req.filter_params}).replace("[", "\[")
     models_data = pprint.pformat(models_data.dict()).replace("[", "\[")
     output_format = pprint.pformat(output_format.dict()).replace("[", "\[")
+    save_data = pprint.pformat(save_data.dict()).replace("[", "\[")
 
     log.info(f"request: {req_str}")
     log.info(f"models data: {models_data}")
     log.info(f"output format: {output_format}")
+    log.info(f"save data: {save_data}")

ui/easydiffusion/tasks/render_images.py

@@ -2,49 +2,59 @@ import json
 import pprint
 import queue
 import time
+from PIL import Image
 
 from easydiffusion import model_manager, runtime
-from easydiffusion.types import GenerateImageRequest, ModelsData, OutputFormatData
+from easydiffusion.types import GenerateImageRequest, ModelsData, OutputFormatData, SaveToDiskData
 from easydiffusion.types import Image as ResponseImage
-from easydiffusion.types import GenerateImageResponse, TaskData, UserInitiatedStop
-from easydiffusion.utils import get_printable_request, log, save_images_to_disk
-from sdkit.generate import generate_images
+from easydiffusion.types import GenerateImageResponse, RenderTaskData
+from easydiffusion.utils import get_printable_request, log, save_images_to_disk, filter_nsfw
 from sdkit.utils import (
-    diffusers_latent_samples_to_images,
-    gc,
     img_to_base64_str,
+    base64_str_to_img,
     img_to_buffer,
-    latent_samples_to_images,
     resize_img,
     get_image,
     log,
 )
 
 from .task import Task
-from .filter_images import filter_images
 
 
 class RenderTask(Task):
     "For image generation"
 
     def __init__(
-        self, req: GenerateImageRequest, task_data: TaskData, models_data: ModelsData, output_format: OutputFormatData
+        self,
+        req: GenerateImageRequest,
+        task_data: RenderTaskData,
+        models_data: ModelsData,
+        output_format: OutputFormatData,
+        save_data: SaveToDiskData,
     ):
         super().__init__(task_data.session_id)
 
         task_data.request_id = self.id
-        self.render_request: GenerateImageRequest = req  # Initial Request
-        self.task_data: TaskData = task_data
+
+        self.render_request = req  # Initial Request
+        self.task_data = task_data
         self.models_data = models_data
         self.output_format = output_format
+        self.save_data = save_data
+
         self.temp_images: list = [None] * req.num_outputs * (1 if task_data.show_only_filtered_image else 2)
 
     def run(self):
         "Runs the image generation task on the assigned thread"
 
-        from easydiffusion import task_manager
+        from easydiffusion import task_manager, app
+        from easydiffusion.backend_manager import backend
 
         context = runtime.context
+        config = app.getConfig()
+
+        if config.get("block_nsfw", False):  # override if set on the server
+            self.task_data.block_nsfw = True
 
         def step_callback():
             task_manager.keep_task_alive(self)
@@ -53,7 +63,7 @@ class RenderTask(Task):
             if isinstance(task_manager.current_state_error, (SystemExit, StopAsyncIteration)) or isinstance(
                 self.error, StopAsyncIteration
             ):
-                context.stop_processing = True
+                backend.stop_rendering(context)
                 if isinstance(task_manager.current_state_error, StopAsyncIteration):
                     self.error = task_manager.current_state_error
                     task_manager.current_state_error = None
@@ -63,11 +73,7 @@ class RenderTask(Task):
         model_manager.resolve_model_paths(self.models_data)
 
         models_to_force_reload = []
-        if (
-            runtime.set_vram_optimizations(context)
-            or self.has_param_changed(context, "clip_skip")
-            or self.trt_needs_reload(context)
-        ):
+        if runtime.set_vram_optimizations(context) or self.has_param_changed(context, "clip_skip"):
             models_to_force_reload.append("stable-diffusion")
 
         model_manager.reload_models_if_necessary(context, self.models_data, models_to_force_reload)
@@ -80,13 +86,15 @@ class RenderTask(Task):
             self.task_data,
             self.models_data,
             self.output_format,
+            self.save_data,
             self.buffer_queue,
             self.temp_images,
             step_callback,
+            self,
         )
 
     def has_param_changed(self, context, param_name):
-        if not context.test_diffusers:
+        if not getattr(context, "test_diffusers", False):
             return False
         if "stable-diffusion" not in context.models or "params" not in context.models["stable-diffusion"]:
             return True
@@ -95,49 +103,36 @@ class RenderTask(Task):
         new_val = self.models_data.model_params.get("stable-diffusion", {}).get(param_name, False)
         return model["params"].get(param_name) != new_val
 
-    def trt_needs_reload(self, context):
-        if not context.test_diffusers:
-            return False
-        if "stable-diffusion" not in context.models or "params" not in context.models["stable-diffusion"]:
-            return True
-
-        model = context.models["stable-diffusion"]
-
-        # curr_convert_to_trt = model["params"].get("convert_to_tensorrt")
-        new_convert_to_trt = self.models_data.model_params.get("stable-diffusion", {}).get("convert_to_tensorrt", False)
-
-        pipe = model["default"]
-        is_trt_loaded = hasattr(pipe.unet, "_allocate_trt_buffers") or hasattr(
-            pipe.unet, "_allocate_trt_buffers_backup"
-        )
-        if new_convert_to_trt and not is_trt_loaded:
-            return True
-
-        curr_build_config = model["params"].get("trt_build_config")
-        new_build_config = self.models_data.model_params.get("stable-diffusion", {}).get("trt_build_config", {})
-
-        return new_convert_to_trt and curr_build_config != new_build_config
-
 
 def make_images(
     context,
     req: GenerateImageRequest,
-    task_data: TaskData,
+    task_data: RenderTaskData,
     models_data: ModelsData,
     output_format: OutputFormatData,
+    save_data: SaveToDiskData,
     data_queue: queue.Queue,
     task_temp_images: list,
     step_callback,
+    task,
 ):
-    context.stop_processing = False
-    print_task_info(req, task_data, models_data, output_format)
+    print_task_info(req, task_data, models_data, output_format, save_data)
 
     images, seeds = make_images_internal(
-        context, req, task_data, models_data, output_format, data_queue, task_temp_images, step_callback
+        context,
+        req,
+        task_data,
+        models_data,
+        output_format,
+        save_data,
+        data_queue,
+        task_temp_images,
+        step_callback,
+        task,
     )
 
     res = GenerateImageResponse(
-        req, task_data, models_data, output_format, images=construct_response(images, seeds, output_format)
+        req, task_data, models_data, output_format, save_data, images=construct_response(images, seeds, output_format)
     )
     res = res.json()
     data_queue.put(json.dumps(res))
@@ -147,48 +142,72 @@ def make_images(
 
 
 def print_task_info(
-    req: GenerateImageRequest, task_data: TaskData, models_data: ModelsData, output_format: OutputFormatData
+    req: GenerateImageRequest,
+    task_data: RenderTaskData,
+    models_data: ModelsData,
+    output_format: OutputFormatData,
+    save_data: SaveToDiskData,
 ):
-    req_str = pprint.pformat(get_printable_request(req, task_data, output_format)).replace("[", "\[")
+    req_str = pprint.pformat(get_printable_request(req, task_data, models_data, output_format, save_data)).replace(
+        "[", "\["
+    )
     task_str = pprint.pformat(task_data.dict()).replace("[", "\[")
     models_data = pprint.pformat(models_data.dict()).replace("[", "\[")
     output_format = pprint.pformat(output_format.dict()).replace("[", "\[")
+    save_data = pprint.pformat(save_data.dict()).replace("[", "\[")
 
     log.info(f"request: {req_str}")
     log.info(f"task data: {task_str}")
-    # log.info(f"models data: {models_data}")
+    log.info(f"models data: {models_data}")
     log.info(f"output format: {output_format}")
+    log.info(f"save data: {save_data}")
 
 
 def make_images_internal(
     context,
     req: GenerateImageRequest,
-    task_data: TaskData,
+    task_data: RenderTaskData,
     models_data: ModelsData,
     output_format: OutputFormatData,
+    save_data: SaveToDiskData,
     data_queue: queue.Queue,
     task_temp_images: list,
     step_callback,
+    task,
 ):
-    images, user_stopped = generate_images_internal(
+    from easydiffusion.backend_manager import backend
+
+    # prep the nsfw_filter
+    if task_data.block_nsfw:
+        filter_nsfw([Image.new("RGB", (1, 1))])  # hack - ensures that the model is available
+
+    images = generate_images_internal(
         context,
         req,
         task_data,
         models_data,
+        output_format,
         data_queue,
         task_temp_images,
         step_callback,
         task_data.stream_image_progress,
         task_data.stream_image_progress_interval,
     )
-
-    gc(context)
+    user_stopped = isinstance(task.error, StopAsyncIteration)
 
     filters, filter_params = task_data.filters, task_data.filter_params
-    filtered_images = filter_images(context, images, filters, filter_params) if not user_stopped else images
+    if len(filters) > 0 and not user_stopped:
+        filtered_images = backend.filter_images(context, images, filters, filter_params, input_type="base64")
+    else:
+        filtered_images = images
 
-    if task_data.save_to_disk_path is not None:
-        save_images_to_disk(images, filtered_images, req, task_data, output_format)
+    if task_data.block_nsfw:
+        filtered_images = filter_nsfw(filtered_images)
+
+    if save_data.save_to_disk_path is not None:
+        images_pil = [base64_str_to_img(img) for img in images]
+        filtered_images_pil = [base64_str_to_img(img) for img in filtered_images]
+        save_images_to_disk(images_pil, filtered_images_pil, req, task_data, models_data, output_format, save_data)
 
     seeds = [*range(req.seed, req.seed + len(images))]
     if task_data.show_only_filtered_image or filtered_images is images:
@@ -200,148 +219,96 @@ def make_images_internal(
 def generate_images_internal(
     context,
     req: GenerateImageRequest,
-    task_data: TaskData,
+    task_data: RenderTaskData,
     models_data: ModelsData,
+    output_format: OutputFormatData,
     data_queue: queue.Queue,
     task_temp_images: list,
     step_callback,
     stream_image_progress: bool,
     stream_image_progress_interval: int,
 ):
-    context.temp_images.clear()
+    from easydiffusion.backend_manager import backend
 
-    callback = make_step_callback(
+    callback = make_step_callback(context, req, task_data, data_queue, task_temp_images, step_callback)
+
+    req.width, req.height = map(lambda x: x - x % 8, (req.width, req.height))  # clamp to 8
+
+    if req.control_image and task_data.control_filter_to_apply:
+        req.controlnet_filter = task_data.control_filter_to_apply
+
+    if req.init_image is not None and int(req.num_inference_steps * req.prompt_strength) == 0:
+        req.prompt_strength = 1 / req.num_inference_steps if req.num_inference_steps > 0 else 1
+
+    if req.init_image_mask:
+        req.init_image_mask = get_image(req.init_image_mask)
+        req.init_image_mask = resize_img(req.init_image_mask.convert("RGB"), req.width, req.height, clamp_to_8=True)
+
+    backend.set_options(
         context,
-        req,
-        task_data,
-        data_queue,
-        task_temp_images,
-        step_callback,
-        stream_image_progress,
-        stream_image_progress_interval,
+        output_format=output_format.output_format,
+        output_quality=output_format.output_quality,
+        output_lossless=output_format.output_lossless,
+        vae_tiling=task_data.enable_vae_tiling,
+        stream_image_progress=stream_image_progress,
+        stream_image_progress_interval=stream_image_progress_interval,
+        clip_skip=2 if task_data.clip_skip else 1,
     )
 
-    try:
-        if req.init_image is not None and not context.test_diffusers:
-            req.sampler_name = "ddim"
+    images = backend.generate_images(context, callback=callback, output_type="base64", **req.dict())
 
-        req.width, req.height = map(lambda x: x - x % 8, (req.width, req.height))  # clamp to 8
-
-        if req.control_image and task_data.control_filter_to_apply:
-            req.control_image = get_image(req.control_image)
-            req.control_image = resize_img(req.control_image.convert("RGB"), req.width, req.height, clamp_to_8=True)
-            req.control_image = filter_images(context, req.control_image, task_data.control_filter_to_apply)[0]
-
-        if req.init_image is not None and int(req.num_inference_steps * req.prompt_strength) == 0:
-            req.prompt_strength = 1 / req.num_inference_steps if req.num_inference_steps > 0 else 1
-
-        if context.test_diffusers:
-            pipe = context.models["stable-diffusion"]["default"]
-            if hasattr(pipe.unet, "_allocate_trt_buffers_backup"):
-                setattr(pipe.unet, "_allocate_trt_buffers", pipe.unet._allocate_trt_buffers_backup)
-                delattr(pipe.unet, "_allocate_trt_buffers_backup")
-
-            if hasattr(pipe.unet, "_allocate_trt_buffers"):
-                convert_to_trt = models_data.model_params["stable-diffusion"].get("convert_to_tensorrt", False)
-                if convert_to_trt:
-                    pipe.unet.forward = pipe.unet._trt_forward
-                    # pipe.vae.decoder.forward = pipe.vae.decoder._trt_forward
-                    log.info(f"Setting unet.forward to TensorRT")
-                else:
-                    log.info(f"Not using TensorRT for unet.forward")
-                    pipe.unet.forward = pipe.unet._non_trt_forward
-                    # pipe.vae.decoder.forward = pipe.vae.decoder._non_trt_forward
-                    setattr(pipe.unet, "_allocate_trt_buffers_backup", pipe.unet._allocate_trt_buffers)
-                    delattr(pipe.unet, "_allocate_trt_buffers")
-
-        images = generate_images(context, callback=callback, **req.dict())
-        user_stopped = False
-    except UserInitiatedStop:
-        images = []
-        user_stopped = True
-        if context.partial_x_samples is not None:
-            if context.test_diffusers:
-                images = diffusers_latent_samples_to_images(context, context.partial_x_samples)
-            else:
-                images = latent_samples_to_images(context, context.partial_x_samples)
-    finally:
-        if hasattr(context, "partial_x_samples") and context.partial_x_samples is not None:
-            if not context.test_diffusers:
-                del context.partial_x_samples
-            context.partial_x_samples = None
-
-    return images, user_stopped
+    return images
 
 
 def construct_response(images: list, seeds: list, output_format: OutputFormatData):
-    return [
-        ResponseImage(
-            data=img_to_base64_str(
-                img,
-                output_format.output_format,
-                output_format.output_quality,
-                output_format.output_lossless,
-            ),
-            seed=seed,
-        )
-        for img, seed in zip(images, seeds)
-    ]
+    return [ResponseImage(data=img, seed=seed) for img, seed in zip(images, seeds)]
 
 
 def make_step_callback(
     context,
     req: GenerateImageRequest,
-    task_data: TaskData,
+    task_data: RenderTaskData,
     data_queue: queue.Queue,
     task_temp_images: list,
     step_callback,
-    stream_image_progress: bool,
-    stream_image_progress_interval: int,
 ):
+    from easydiffusion.backend_manager import backend
+
     n_steps = req.num_inference_steps if req.init_image is None else int(req.num_inference_steps * req.prompt_strength)
     last_callback_time = -1
 
-    def update_temp_img(x_samples, task_temp_images: list):
+    def update_temp_img(images, task_temp_images: list):
         partial_images = []
 
-        if context.test_diffusers:
-            images = diffusers_latent_samples_to_images(context, x_samples)
-        else:
-            images = latent_samples_to_images(context, x_samples)
+        if images is None:
+            return []
 
         if task_data.block_nsfw:
-            images = filter_images(context, images, "nsfw_checker")
+            images = filter_nsfw(images, print_log=False)
 
         for i, img in enumerate(images):
+            img = img.convert("RGB")
+            img = resize_img(img, req.width, req.height)
             buf = img_to_buffer(img, output_format="JPEG")
 
-            context.temp_images[f"{task_data.request_id}/{i}"] = buf
             task_temp_images[i] = buf
             partial_images.append({"path": f"/image/tmp/{task_data.request_id}/{i}"})
         del images
         return partial_images
 
-    def on_image_step(x_samples, i, *args):
+    def on_image_step(images, i, *args):
         nonlocal last_callback_time
 
-        if context.test_diffusers:
-            context.partial_x_samples = (x_samples, args[0])
-        else:
-            context.partial_x_samples = x_samples
-
         step_time = time.time() - last_callback_time if last_callback_time != -1 else -1
         last_callback_time = time.time()
 
         progress = {"step": i, "step_time": step_time, "total_steps": n_steps}
 
-        if stream_image_progress and stream_image_progress_interval > 0 and i % stream_image_progress_interval == 0:
-            progress["output"] = update_temp_img(context.partial_x_samples, task_temp_images)
+        if images is not None:
+            progress["output"] = update_temp_img(images, task_temp_images)
 
         data_queue.put(json.dumps(progress))
 
         step_callback()
 
-        if context.stop_processing:
-            raise UserInitiatedStop("User requested that we stop processing")
-
     return on_image_step

ui/easydiffusion/types.py

@@ -14,19 +14,22 @@ class GenerateImageRequest(BaseModel):
     num_outputs: int = 1
     num_inference_steps: int = 50
     guidance_scale: float = 7.5
+    distilled_guidance_scale: float = 3.5
 
     init_image: Any = None
     init_image_mask: Any = None
     control_image: Any = None
     control_alpha: Union[float, List[float]] = None
+    controlnet_filter: str = None
     prompt_strength: float = 0.8
-    preserve_init_image_color_profile = False
-    strict_mask_border = False
+    preserve_init_image_color_profile: bool = False
+    strict_mask_border: bool = False
 
     sampler_name: str = None  # "ddim", "plms", "heun", "euler", "euler_a", "dpm2", "dpm2_a", "lms"
+    scheduler_name: str = None
     hypernetwork_strength: float = 0
     lora_alpha: Union[float, List[float]] = 0
-    tiling: str = "none"  # "none", "x", "y", "xy"
+    tiling: str = None  # None, "x", "y", "xy"
 
 
 class FilterImageRequest(BaseModel):
@@ -58,10 +61,17 @@ class OutputFormatData(BaseModel):
     output_lossless: bool = False
 
 
+class SaveToDiskData(BaseModel):
+    save_to_disk_path: str = None
+    metadata_output_format: str = "txt"  # or "json"
+
+
 class TaskData(BaseModel):
     request_id: str = None
     session_id: str = "session"
-    save_to_disk_path: str = None
+
+
+class RenderTaskData(TaskData):
     vram_usage_level: str = "balanced"  # or "low" or "medium"
 
     use_face_correction: Union[str, List[str]] = None  # or "GFPGANv1.3"
@@ -70,6 +80,7 @@ class TaskData(BaseModel):
     latent_upscaler_steps: int = 10
     use_stable_diffusion_model: Union[str, List[str]] = "sd-v1-4"
     use_vae_model: Union[str, List[str]] = None
+    use_text_encoder_model: Union[str, List[str]] = None
     use_hypernetwork_model: Union[str, List[str]] = None
     use_lora_model: Union[str, List[str]] = None
     use_controlnet_model: Union[str, List[str]] = None
@@ -77,10 +88,10 @@ class TaskData(BaseModel):
     filters: List[str] = []
     filter_params: Dict[str, Dict[str, Any]] = {}
     control_filter_to_apply: Union[str, List[str]] = None
+    enable_vae_tiling: bool = True
 
     show_only_filtered_image: bool = False
     block_nsfw: bool = False
-    metadata_output_format: str = "txt"  # or "json"
     stream_image_progress: bool = False
     stream_image_progress_interval: int = 5
     clip_skip: bool = False
@@ -93,7 +104,7 @@ class MergeRequest(BaseModel):
     model1: str = None
     ratio: float = None
     out_path: str = "mix"
-    use_fp16 = True
+    use_fp16: bool = True
 
 
 class Image:
@@ -126,12 +137,14 @@ class GenerateImageResponse:
         task_data: TaskData,
         models_data: ModelsData,
         output_format: OutputFormatData,
+        save_data: SaveToDiskData,
         images: list,
     ):
         self.render_request = render_request
         self.task_data = task_data
         self.models_data = models_data
         self.output_format = output_format
+        self.save_data = save_data
         self.images = images
 
     def json(self):
@@ -141,6 +154,7 @@ class GenerateImageResponse:
 
         task_data = self.task_data.dict()
         task_data.update(self.output_format.dict())
+        task_data.update(self.save_data.dict())
 
         res = {
             "status": "succeeded",
@@ -198,27 +212,25 @@ def convert_legacy_render_req_to_new(old_req: dict):
     # move the model info
     model_paths["stable-diffusion"] = old_req.get("use_stable_diffusion_model")
     model_paths["vae"] = old_req.get("use_vae_model")
+    model_paths["text-encoder"] = old_req.get("use_text_encoder_model")
     model_paths["hypernetwork"] = old_req.get("use_hypernetwork_model")
     model_paths["lora"] = old_req.get("use_lora_model")
     model_paths["controlnet"] = old_req.get("use_controlnet_model")
     model_paths["embeddings"] = old_req.get("use_embeddings_model")
 
-    model_paths["gfpgan"] = old_req.get("use_face_correction", "")
-    model_paths["gfpgan"] = model_paths["gfpgan"] if "gfpgan" in model_paths["gfpgan"].lower() else None
+    ## ensure that the model name is in the model path
+    for model_name in ("gfpgan", "codeformer"):
+        model_paths[model_name] = old_req.get("use_face_correction", "")
+        model_paths[model_name] = model_paths[model_name] if model_name in model_paths[model_name].lower() else None
 
-    model_paths["codeformer"] = old_req.get("use_face_correction", "")
-    model_paths["codeformer"] = model_paths["codeformer"] if "codeformer" in model_paths["codeformer"].lower() else None
+    for model_name in ("realesrgan", "latent_upscaler", "esrgan_4x", "lanczos", "nearest", "scunet", "swinir"):
+        model_paths[model_name] = old_req.get("use_upscale", "")
+        model_paths[model_name] = model_paths[model_name] if model_name in model_paths[model_name].lower() else None
 
-    model_paths["realesrgan"] = old_req.get("use_upscale", "")
-    model_paths["realesrgan"] = model_paths["realesrgan"] if "realesrgan" in model_paths["realesrgan"].lower() else None
-
-    model_paths["latent_upscaler"] = old_req.get("use_upscale", "")
-    model_paths["latent_upscaler"] = (
-        model_paths["latent_upscaler"] if "latent_upscaler" in model_paths["latent_upscaler"].lower() else None
-    )
     if "control_filter_to_apply" in old_req:
         filter_model = old_req["control_filter_to_apply"]
         model_paths[filter_model] = filter_model
+        old_req["control_filter_to_apply"] = convert_legacy_controlnet_filter_name(old_req["control_filter_to_apply"])
 
     if old_req.get("block_nsfw"):
         model_paths["nsfw_checker"] = "nsfw_checker"
@@ -234,8 +246,12 @@ def convert_legacy_render_req_to_new(old_req: dict):
         }
 
     # move the filter params
-    if model_paths["realesrgan"]:
-        filter_params["realesrgan"] = {"scale": int(old_req.get("upscale_amount", 4))}
+    for model_name in ("realesrgan", "esrgan_4x", "lanczos", "nearest", "scunet", "swinir"):
+        if model_paths[model_name]:
+            filter_params[model_name] = {
+                "upscaler": model_paths[model_name],
+                "scale": int(old_req.get("upscale_amount", 4)),
+            }
     if model_paths["latent_upscaler"]:
         filter_params["latent_upscaler"] = {
             "prompt": old_req["prompt"],
@@ -254,14 +270,31 @@ def convert_legacy_render_req_to_new(old_req: dict):
     if old_req.get("block_nsfw"):
         filters.append("nsfw_checker")
 
-    if model_paths["codeformer"]:
-        filters.append("codeformer")
-    elif model_paths["gfpgan"]:
-        filters.append("gfpgan")
+    for model_name in ("gfpgan", "codeformer"):
+        if model_paths[model_name]:
+            filters.append(model_name)
+            break
 
-    if model_paths["realesrgan"]:
-        filters.append("realesrgan")
-    elif model_paths["latent_upscaler"]:
-        filters.append("latent_upscaler")
+    for model_name in ("realesrgan", "latent_upscaler", "esrgan_4x", "lanczos", "nearest", "scunet", "swinir"):
+        if model_paths[model_name]:
+            filters.append(model_name)
+            break
 
     return new_req
+
+
+def convert_legacy_controlnet_filter_name(filter):
+    from easydiffusion.backend_manager import backend
+
+    if filter is None:
+        return None
+
+    controlnet_filter_names = backend.list_controlnet_filters()
+
+    def apply(f):
+        return f"controlnet_{f}" if f in controlnet_filter_names else f
+
+    if isinstance(filter, list):
+        return [apply(f) for f in filter]
+
+    return apply(filter)

ui/easydiffusion/utils/__init__.py

@@ -1,4 +1,5 @@
 import logging
+import hashlib
 
 log = logging.getLogger("easydiffusion")
 
@@ -6,6 +7,8 @@ from .save_utils import (
     save_images_to_disk,
     get_printable_request,
 )
+from .nsfw_checker import filter_nsfw
+
 
 def sha256sum(filename):
     sha256 = hashlib.sha256()
@@ -17,4 +20,3 @@ def sha256sum(filename):
             sha256.update(data)
 
     return sha256.hexdigest()
-

ui/easydiffusion/utils/nsfw_checker.py

@@ -0,0 +1,80 @@
+# possibly move this to sdkit in the future
+import os
+
+# mirror of https://huggingface.co/AdamCodd/vit-base-nsfw-detector/blob/main/onnx/model_quantized.onnx
+NSFW_MODEL_URL = (
+    "https://github.com/easydiffusion/sdkit-test-data/releases/download/assets/vit-base-nsfw-detector-quantized.onnx"
+)
+MODEL_HASH_QUICK = "220123559305b1b07b7a0894c3471e34dccd090d71cdf337dd8012f9e40d6c28"
+
+nsfw_check_model = None
+
+
+def filter_nsfw(images, blur_radius: float = 75, print_log=True):
+    global nsfw_check_model
+
+    from easydiffusion.model_manager import get_model_dirs
+    from sdkit.utils import base64_str_to_img, img_to_base64_str, download_file, log, hash_file_quick
+
+    import onnxruntime as ort
+    from PIL import ImageFilter
+    import numpy as np
+
+    if nsfw_check_model is None:
+        model_dir = get_model_dirs("nsfw-checker")[0]
+        model_path = os.path.join(model_dir, "vit-base-nsfw-detector-quantized.onnx")
+
+        os.makedirs(model_dir, exist_ok=True)
+
+        if not os.path.exists(model_path) or hash_file_quick(model_path) != MODEL_HASH_QUICK:
+            download_file(NSFW_MODEL_URL, model_path)
+
+        nsfw_check_model = ort.InferenceSession(model_path, providers=["CPUExecutionProvider"])
+
+    # Preprocess the input image
+    def preprocess_image(img):
+        img = img.convert("RGB")
+
+        # config based on based on https://huggingface.co/AdamCodd/vit-base-nsfw-detector/blob/main/onnx/preprocessor_config.json
+        # Resize the image
+        img = img.resize((384, 384))
+
+        # Normalize the image
+        img = np.array(img) / 255.0  # Scale pixel values to [0, 1]
+        mean = np.array([0.5, 0.5, 0.5])
+        std = np.array([0.5, 0.5, 0.5])
+        img = (img - mean) / std
+
+        # Transpose to match input shape (batch_size, channels, height, width)
+        img = np.transpose(img, (2, 0, 1)).astype(np.float32)
+
+        # Add batch dimension
+        img = np.expand_dims(img, axis=0)
+
+        return img
+
+    # Run inference
+    input_name = nsfw_check_model.get_inputs()[0].name
+    output_name = nsfw_check_model.get_outputs()[0].name
+
+    if print_log:
+        log.info("Running NSFW checker (onnx)")
+
+    results = []
+    for img in images:
+        is_base64 = isinstance(img, str)
+
+        input_img = base64_str_to_img(img) if is_base64 else img
+
+        result = nsfw_check_model.run([output_name], {input_name: preprocess_image(input_img)})
+        is_nsfw = [np.argmax(arr) == 1 for arr in result][0]
+
+        if is_nsfw:
+            output_img = input_img.filter(ImageFilter.GaussianBlur(blur_radius))
+            output_img = img_to_base64_str(output_img) if is_base64 else output_img
+        else:
+            output_img = img
+
+        results.append(output_img)
+
+    return results

ui/easydiffusion/utils/save_utils.py

@@ -7,7 +7,14 @@ from datetime import datetime
 from functools import reduce
 
 from easydiffusion import app
-from easydiffusion.types import GenerateImageRequest, TaskData, OutputFormatData
+from easydiffusion.types import (
+    GenerateImageRequest,
+    TaskData,
+    RenderTaskData,
+    OutputFormatData,
+    SaveToDiskData,
+    ModelsData,
+)
 from numpy import base_repr
 from sdkit.utils import save_dicts, save_images
 from sdkit.models.model_loader.embeddings import get_embedding_token
@@ -24,12 +31,15 @@ TASK_TEXT_MAPPING = {
     "clip_skip": "Clip Skip",
     "use_controlnet_model": "ControlNet model",
     "control_filter_to_apply": "ControlNet Filter",
+    "control_alpha": "ControlNet Strength",
     "use_vae_model": "VAE model",
     "sampler_name": "Sampler",
+    "scheduler_name": "Scheduler",
     "width": "Width",
     "height": "Height",
     "num_inference_steps": "Steps",
     "guidance_scale": "Guidance Scale",
+    "distilled_guidance_scale": "Distilled Guidance",
     "prompt_strength": "Prompt Strength",
     "use_lora_model": "LoRA model",
     "lora_alpha": "LoRA Strength",
@@ -95,7 +105,7 @@ def format_folder_name(format: str, req: GenerateImageRequest, task_data: TaskDa
 def format_file_name(
     format: str,
     req: GenerateImageRequest,
-    task_data: TaskData,
+    task_data: RenderTaskData,
     now: float,
     batch_file_number: int,
     folder_img_number: ImageNumber,
@@ -118,13 +128,19 @@ def format_file_name(
 
 
 def save_images_to_disk(
-    images: list, filtered_images: list, req: GenerateImageRequest, task_data: TaskData, output_format: OutputFormatData
+    images: list,
+    filtered_images: list,
+    req: GenerateImageRequest,
+    task_data: RenderTaskData,
+    models_data: ModelsData,
+    output_format: OutputFormatData,
+    save_data: SaveToDiskData,
 ):
     now = time.time()
     app_config = app.getConfig()
     folder_format = app_config.get("folder_format", "$id")
-    save_dir_path = os.path.join(task_data.save_to_disk_path, format_folder_name(folder_format, req, task_data))
-    metadata_entries = get_metadata_entries_for_request(req, task_data, output_format)
+    save_dir_path = os.path.join(save_data.save_to_disk_path, format_folder_name(folder_format, req, task_data))
+    metadata_entries = get_metadata_entries_for_request(req, task_data, models_data, output_format, save_data)
     file_number = calculate_img_number(save_dir_path, task_data)
     make_filename = make_filename_callback(
         app_config.get("filename_format", "$p_$tsb64"),
@@ -143,8 +159,8 @@ def save_images_to_disk(
             output_quality=output_format.output_quality,
             output_lossless=output_format.output_lossless,
         )
-        if task_data.metadata_output_format:
-            for metadata_output_format in task_data.metadata_output_format.split(","):
+        if save_data.metadata_output_format:
+            for metadata_output_format in save_data.metadata_output_format.split(","):
                 if metadata_output_format.lower() in ["json", "txt", "embed"]:
                     save_dicts(
                         metadata_entries,
@@ -179,8 +195,8 @@ def save_images_to_disk(
             output_quality=output_format.output_quality,
             output_lossless=output_format.output_lossless,
         )
-        if task_data.metadata_output_format:
-            for metadata_output_format in task_data.metadata_output_format.split(","):
+        if save_data.metadata_output_format:
+            for metadata_output_format in save_data.metadata_output_format.split(","):
                 if metadata_output_format.lower() in ["json", "txt", "embed"]:
                     save_dicts(
                         metadata_entries,
@@ -191,11 +207,17 @@ def save_images_to_disk(
                     )
 
 
-def get_metadata_entries_for_request(req: GenerateImageRequest, task_data: TaskData, output_format: OutputFormatData):
-    metadata = get_printable_request(req, task_data, output_format)
+def get_metadata_entries_for_request(
+    req: GenerateImageRequest,
+    task_data: RenderTaskData,
+    models_data: ModelsData,
+    output_format: OutputFormatData,
+    save_data: SaveToDiskData,
+):
+    metadata = get_printable_request(req, task_data, models_data, output_format, save_data)
 
     # if text, format it in the text format expected by the UI
-    is_txt_format = task_data.metadata_output_format and "txt" in task_data.metadata_output_format.lower().split(",")
+    is_txt_format = save_data.metadata_output_format and "txt" in save_data.metadata_output_format.lower().split(",")
     if is_txt_format:
 
         def format_value(value):
@@ -214,13 +236,20 @@ def get_metadata_entries_for_request(req: GenerateImageRequest, task_data: TaskD
     return entries
 
 
-def get_printable_request(req: GenerateImageRequest, task_data: TaskData, output_format: OutputFormatData):
+def get_printable_request(
+    req: GenerateImageRequest,
+    task_data: RenderTaskData,
+    models_data: ModelsData,
+    output_format: OutputFormatData,
+    save_data: SaveToDiskData,
+):
     req_metadata = req.dict()
     task_data_metadata = task_data.dict()
     task_data_metadata.update(output_format.dict())
+    task_data_metadata.update(save_data.dict())
 
     app_config = app.getConfig()
-    using_diffusers = app_config.get("test_diffusers", True)
+    using_diffusers = app_config.get("backend", "ed_diffusers") in ("ed_diffusers", "webui")
 
     # Save the metadata in the order defined in TASK_TEXT_MAPPING
     metadata = {}
@@ -230,25 +259,11 @@ def get_printable_request(req: GenerateImageRequest, task_data: TaskData, output
         elif key in task_data_metadata:
             metadata[key] = task_data_metadata[key]
 
-        if key == "use_embeddings_model" and using_diffusers:
-            embeddings_extensions = {".pt", ".bin", ".safetensors"}
+        if key == "use_embeddings_model" and task_data_metadata[key] and using_diffusers:
+            embeddings_used = models_data.model_paths["embeddings"]
+            embeddings_used = embeddings_used if isinstance(embeddings_used, list) else [embeddings_used]
 
-            def scan_directory(directory_path: str):
-                used_embeddings = []
-                for entry in os.scandir(directory_path):
-                    if entry.is_file():
-                        # Check if the filename has the right extension
-                        if not any(map(lambda ext: entry.name.endswith(ext), embeddings_extensions)):
-                            continue
-                        embedding_name_regex = regex.compile(r"(^|[\s,])" + regex.escape(get_embedding_token(entry.name)) + r"([+-]*$|[\s,]|[+-]+[\s,])")
-                        if embedding_name_regex.search(req.prompt) or embedding_name_regex.search(req.negative_prompt):
-                            used_embeddings.append(entry.path)
-                    elif entry.is_dir():
-                        used_embeddings.extend(scan_directory(entry.path))
-                return used_embeddings
-
-            used_embeddings = scan_directory(os.path.join(app.MODELS_DIR, "embeddings"))
-            metadata["use_embeddings_model"] = used_embeddings if len(used_embeddings) > 0 else None
+            metadata["use_embeddings_model"] = embeddings_used if len(embeddings_used) > 0 else None
 
     # Clean up the metadata
     if req.init_image is None and "prompt_strength" in metadata:
@@ -269,7 +284,17 @@ def get_printable_request(req: GenerateImageRequest, task_data: TaskData, output
             del metadata[key]
     else:
         for key in (
-            x for x in ["use_lora_model", "lora_alpha", "clip_skip", "tiling", "latent_upscaler_steps", "use_controlnet_model", "control_filter_to_apply"] if x in metadata
+            x
+            for x in [
+                "use_lora_model",
+                "lora_alpha",
+                "clip_skip",
+                "tiling",
+                "latent_upscaler_steps",
+                "use_controlnet_model",
+                "control_filter_to_apply",
+            ]
+            if x in metadata
         ):
             del metadata[key]
 
@@ -279,7 +304,7 @@ def get_printable_request(req: GenerateImageRequest, task_data: TaskData, output
 def make_filename_callback(
     filename_format: str,
     req: GenerateImageRequest,
-    task_data: TaskData,
+    task_data: RenderTaskData,
     folder_img_number: int,
     suffix=None,
     now=None,
@@ -296,7 +321,7 @@ def make_filename_callback(
     return make_filename
 
 
-def _calculate_img_number(save_dir_path: str, task_data: TaskData):
+def _calculate_img_number(save_dir_path: str, task_data: RenderTaskData):
     def get_highest_img_number(accumulator: int, file: os.DirEntry) -> int:
         if not file.is_file:
             return accumulator
@@ -340,5 +365,5 @@ def _calculate_img_number(save_dir_path: str, task_data: TaskData):
 _calculate_img_number.session_img_numbers = {}
 
 
-def calculate_img_number(save_dir_path: str, task_data: TaskData):
+def calculate_img_number(save_dir_path: str, task_data: RenderTaskData):
     return ImageNumber(lambda: _calculate_img_number(save_dir_path, task_data))

ui/index.html

@@ -35,7 +35,13 @@
             <h1>
                 <img id="logo_img" src="/media/images/icon-512x512.png" >
                 Easy Diffusion
-                <small><span id="version">v3.0.2</span> <span id="updateBranchLabel"></span></small>
+                <small>
+                    <span id="version">
+                        <span class="gated-feature" data-feature-keys="backend_ed_classic backend_ed_diffusers">v3.0.13</span>
+                        <span class="gated-feature" data-feature-keys="backend_webui">v3.5.9</span>
+                    </span> <span id="updateBranchLabel"></span>
+                    <div id="engine-logo" class="gated-feature" data-feature-keys="backend_webui">(Powered by <a id="backend-url" href="https://github.com/lllyasviel/stable-diffusion-webui-forge" target="_blank">Stable Diffusion WebUI Forge</a>)</div>
+                </small>
             </h1>
         </div>
         <div id="server-status">
@@ -73,7 +79,7 @@
                         </div>
                         <div id="prompt-toolbar-right" class="toolbar-right">
                             <button id="image-modifier-dropdown" class="tertiaryButton smallButton">+ Image Modifiers</button>
-                            <button id="embeddings-button" class="tertiaryButton smallButton displayNone">+ Embedding</button>
+                            <button id="embeddings-button" class="tertiaryButton smallButton gated-feature" data-feature-keys="backend_ed_diffusers backend_webui">+ Embedding</button>
                         </div>
                     </div>
                     <textarea id="prompt" class="col-free">a photograph of an astronaut riding a horse</textarea>
@@ -83,7 +89,7 @@
                         <a href="https://github.com/easydiffusion/easydiffusion/wiki/Writing-prompts#negative-prompts" target="_blank"><i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top">Click to learn more about Negative Prompts</span></i></a>
                         <small>(optional)</small>
                     </label>
-                    <button id="negative-embeddings-button" class="tertiaryButton smallButton displayNone">+ Negative Embedding</button>
+                    <button id="negative-embeddings-button" class="tertiaryButton smallButton gated-feature" data-feature-keys="backend_ed_diffusers backend_webui">+ Negative Embedding</button>
                     <div class="collapsible-content">
                         <textarea id="negative_prompt" name="negative_prompt" placeholder="list the things to remove from the image (e.g. fog, green)"></textarea>
                     </div>
@@ -155,11 +161,11 @@
                 <div id="editor-settings-entries" class="collapsible-content">
                     <div><table>
                     <tr><b class="settings-subheader">Image Settings</b></tr>
-                    <tr class="pl-5"><td><label for="seed">Seed:</label></td><td><input id="seed" name="seed" size="10" value="0" onkeypress="preventNonNumericalInput(event)"> <input id="random_seed" name="random_seed" type="checkbox" checked><label for="random_seed">Random</label></td></tr>
+                    <tr class="pl-5"><td><label for="seed">Seed:</label></td><td><input id="seed" name="seed" size="10" value="0" onkeypress="preventNonNumericalInput(event)" inputmode="numeric"> <input id="random_seed" name="random_seed" type="checkbox" checked><label for="random_seed">Random</label></td></tr>
                     <tr class="pl-5"><td><label for="num_outputs_total">Number of Images:</label></td>
-                        <td><input id="num_outputs_total" name="num_outputs_total" value="1" type="number" value="1" min="1" step="1" onkeypres"="preventNonNumericalInput(event)">
+                        <td><input id="num_outputs_total" name="num_outputs_total" value="1" type="number" value="1" min="1" step="1" onkeypres"="preventNonNumericalInput(event)" inputmode="numeric">
                         <label><small>(total)</small></label>
-                        <input id="num_outputs_parallel" name="num_outputs_parallel" value="1" type="number" value="1" min="1" step="1" onkeypress="preventNonNumericalInput(event)">
+                        <input id="num_outputs_parallel" name="num_outputs_parallel" value="1" type="number" value="1" min="1" step="1" onkeypress="preventNonNumericalInput(event)" inputmode="numeric">
                         <label id="num_outputs_parallel_label" for="num_outputs_parallel"><small>(in parallel)</small></label></td>
                     </tr>
                     <tr class="pl-5"><td><label for="stable_diffusion_model">Model:</label></td><td class="model-input">
@@ -174,14 +180,14 @@
                             <!-- <label><small>Takes upto 20 mins the first time</small></label> -->
                         </td>
                     </tr>
-                    <tr class="pl-5 displayNone" id="clip_skip_config">
+                    <tr class="pl-5 gated-feature" id="clip_skip_config" data-feature-keys="backend_ed_diffusers backend_webui">
                         <td><label for="clip_skip">Clip Skip:</label></td>
                         <td class="diffusers-restart-needed">
                             <input id="clip_skip" name="clip_skip" type="checkbox">
                             <a href="https://github.com/easydiffusion/easydiffusion/wiki/Clip-Skip" target="_blank"><i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Click to learn more about Clip Skip</span></i></a>
                         </td>
                     </tr>
-                    <tr id="controlnet_model_container" class="pl-5">
+                    <tr id="controlnet_model_container" class="pl-5 gated-feature" data-feature-keys="backend_ed_diffusers backend_webui">
                         <td><label for="controlnet_model">ControlNet Image:</label></td>
                         <td class="diffusers-restart-needed">
                             <div id="control_image_wrapper" class="preview_image_wrapper">
@@ -201,40 +207,94 @@
                                         <option value="openpose_faceonly">OpenPose face-only</option>
                                         <option value="openpose_hand">OpenPose hand</option>
                                         <option value="openpose_full">OpenPose full</option>
+                                        <option value="animal_openpose" class="gated-feature" data-feature-keys="backend_webui">animal_openpose</option>
+                                        <option value="densepose_parula (black bg & blue torso)" class="gated-feature" data-feature-keys="backend_webui">densepose_parula (black bg & blue torso)</option>
+                                        <option value="densepose (pruple bg & purple torso)" class="gated-feature" data-feature-keys="backend_webui">densepose (pruple bg & purple torso)</option>
+                                        <option value="dw_openpose_full" class="gated-feature" data-feature-keys="backend_webui">dw_openpose_full</option>
+                                        <option value="mediapipe_face" class="gated-feature" data-feature-keys="backend_webui">mediapipe_face</option>
+                                        <option value="instant_id_face_keypoints" class="gated-feature" data-feature-keys="backend_webui">instant_id_face_keypoints</option>
+                                        <option value="InsightFace+CLIP-H (IPAdapter)" class="gated-feature" data-feature-keys="backend_webui">InsightFace+CLIP-H (IPAdapter)</option>
+                                        <option value="InsightFace (InstantID)" class="gated-feature" data-feature-keys="backend_webui">InsightFace (InstantID)</option>
                                     </optgroup>
                                     <optgroup label="Outline">
                                         <option value="canny">Canny (*)</option>
                                         <option value="mlsd">Straight lines</option>
                                         <option value="scribble_hed">Scribble hed (*)</option>
-                                        <option value="scribble_hedsafe">Scribble hedsafe</option>
+                                        <option value="scribble_hedsafe" class="gated-feature" data-feature-keys="backend_diffusers">Scribble hedsafe</option>
                                         <option value="scribble_pidinet">Scribble pidinet</option>
-                                        <option value="scribble_pidsafe">Scribble pidsafe</option>
+                                        <option value="scribble_pidsafe" class="gated-feature" data-feature-keys="backend_diffusers">Scribble pidsafe</option>
+                                        <option value="scribble_xdog" class="gated-feature" data-feature-keys="backend_webui">scribble_xdog</option>
                                         <option value="softedge_hed">Softedge hed</option>
                                         <option value="softedge_hedsafe">Softedge hedsafe</option>
                                         <option value="softedge_pidinet">Softedge pidinet</option>
                                         <option value="softedge_pidsafe">Softedge pidsafe</option>
+                                        <option value="softedge_teed" class="gated-feature" data-feature-keys="backend_webui">softedge_teed</option>
                                     </optgroup>
                                     <optgroup label="Depth">
                                         <option value="normal_bae">Normal bae (*)</option>
                                         <option value="depth_midas">Depth midas</option>
+                                        <option value="normal_midas" class="gated-feature" data-feature-keys="backend_webui">normal_midas</option>
                                         <option value="depth_zoe">Depth zoe</option>
                                         <option value="depth_leres">Depth leres</option>
                                         <option value="depth_leres++">Depth leres++</option>
+                                        <option value="depth_anything_v2" class="gated-feature" data-feature-keys="backend_webui">depth_anything_v2</option>
+                                        <option value="depth_anything" class="gated-feature" data-feature-keys="backend_webui">depth_anything</option>
+                                        <option value="depth_hand_refiner" class="gated-feature" data-feature-keys="backend_webui">depth_hand_refiner</option>
+                                        <option value="depth_marigold" class="gated-feature" data-feature-keys="backend_webui">depth_marigold</option>
                                     </optgroup>
                                     <optgroup label="Line art">
                                         <option value="lineart_coarse">Lineart coarse</option>
                                         <option value="lineart_realistic">Lineart realistic</option>
                                         <option value="lineart_anime">Lineart anime</option>
+                                        <option value="lineart_standard (from white bg & black line)" class="gated-feature" data-feature-keys="backend_webui">lineart_standard (from white bg & black line)</option>
+                                        <option value="lineart_anime_denoise" class="gated-feature" data-feature-keys="backend_webui">lineart_anime_denoise</option>
+                                    </optgroup>
+                                    <optgroup label="Reference" class="gated-feature" data-feature-keys="backend_webui">
+                                        <option value="reference_adain">reference_adain</option>
+                                        <option value="reference_only">reference_only</option>
+                                        <option value="reference_adain+attn">reference_adain+attn</option>
+                                    </optgroup>
+                                    <optgroup label="Tile" class="gated-feature" data-feature-keys="backend_webui">
+                                        <option value="tile_colorfix">tile_colorfix</option>
+                                        <option value="tile_resample">tile_resample</option>
+                                        <option value="tile_colorfix+sharp">tile_colorfix+sharp</option>
+                                    </optgroup>
+                                    <optgroup label="CLIP (IPAdapter)" class="gated-feature" data-feature-keys="backend_webui">
+                                        <option value="CLIP-ViT-H (IPAdapter)">CLIP-ViT-H (IPAdapter)</option>
+                                        <option value="CLIP-G (Revision)">CLIP-G (Revision)</option>
+                                        <option value="CLIP-G (Revision ignore prompt)">CLIP-G (Revision ignore prompt)</option>
+                                        <option value="CLIP-ViT-bigG (IPAdapter)">CLIP-ViT-bigG (IPAdapter)</option>
+                                        <option value="InsightFace+CLIP-H (IPAdapter)">InsightFace+CLIP-H (IPAdapter)</option>
+                                    </optgroup>
+                                    <optgroup label="Inpaint" class="gated-feature" data-feature-keys="backend_webui">
+                                        <option value="inpaint_only">inpaint_only</option>
+                                        <option value="inpaint_only+lama">inpaint_only+lama</option>
+                                        <option value="inpaint_global_harmonious">inpaint_global_harmonious</option>
+                                    </optgroup>
+                                    <optgroup label="Segment" class="gated-feature" data-feature-keys="backend_webui">
+                                        <option value="seg_ufade20k">seg_ufade20k</option>
+                                        <option value="seg_ofade20k">seg_ofade20k</option>
+                                        <option value="seg_anime_face">seg_anime_face</option>
+                                        <option value="seg_ofcoco">seg_ofcoco</option>
                                     </optgroup>
                                     <optgroup label="Misc">
                                         <option value="shuffle">Shuffle</option>
-                                        <option value="segment">Segment</option>
+                                        <option value="segment" class="gated-feature" data-feature-keys="backend_diffusers">Segment</option>
+                                        <option value="invert (from white bg & black line)" class="gated-feature" data-feature-keys="backend_webui">invert (from white bg & black line)</option>
+                                        <option value="threshold" class="gated-feature" data-feature-keys="backend_webui">threshold</option>
+                                        <option value="t2ia_sketch_pidi" class="gated-feature" data-feature-keys="backend_webui">t2ia_sketch_pidi</option>
+                                        <option value="t2ia_color_grid" class="gated-feature" data-feature-keys="backend_webui">t2ia_color_grid</option>
+                                        <option value="recolor_intensity" class="gated-feature" data-feature-keys="backend_webui">recolor_intensity</option>
+                                        <option value="recolor_luminance" class="gated-feature" data-feature-keys="backend_webui">recolor_luminance</option>
+                                        <option value="blur_gaussian" class="gated-feature" data-feature-keys="backend_webui">blur_gaussian</option>
                                     </optgroup>
                                 </select>
                                 <br/>
                                 <label for="controlnet_model"><small>Model:</small></label> <input id="controlnet_model" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" />
+                                <!-- <br/>
+                                <label><small>Will download the necessary models, the first time.</small></label> -->
                                 <br/>
-                                <label><small>Will download the necessary models, the first time.</small></label>
+                                <label for="controlnet_alpha_slider"><small>Strength:</small></label> <input id="controlnet_alpha_slider" name="controlnet_alpha_slider" class="editor-slider" value="10" type="range" min="0" max="10"> <input id="controlnet_alpha" name="controlnet_alpha" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)" inputmode="decimal">
                             </div>
                         </td>
                     </tr>
@@ -242,32 +302,73 @@
                         <input id="vae_model" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" />
                         <a href="https://github.com/easydiffusion/easydiffusion/wiki/VAE-Variational-Auto-Encoder" target="_blank"><i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Click to learn more about VAEs</span></i></a>
                     </td></tr>
+                    <tr id="text_encoder_model_container" class="pl-5 gated-feature" data-feature-keys="backend_webui">
+                        <td>
+                            <label for="text_encoder_model">Text Encoder:</label>
+                        </td>
+                        <td>
+                            <div id="text_encoder_model" data-path=""></div>
+                        </td>
+                    </tr>
                     <tr id="samplerSelection" class="pl-5"><td><label for="sampler_name">Sampler:</label></td><td>
                         <select id="sampler_name" name="sampler_name">
                             <option value="plms">PLMS</option>
                             <option value="ddim">DDIM</option>
+                            <option value="ddim_cfgpp" class="gated-feature" data-feature-keys="backend_webui">DDIM CFG++</option>
                             <option value="heun">Heun</option>
                             <option value="euler">Euler</option>
                             <option value="euler_a" selected>Euler Ancestral</option>
                             <option value="dpm2">DPM2</option>
                             <option value="dpm2_a">DPM2 Ancestral</option>
+                            <option value="dpm_fast" class="gated-feature" data-feature-keys="backend_webui">DPM Fast</option>
+                            <option value="dpm_adaptive" class="gated-feature" data-feature-keys="backend_ed_classic backend_webui">DPM Adaptive</option>
                             <option value="lms">LMS</option>
-                            <option value="dpm_solver_stability">DPM Solver (Stability AI)</option>
+                            <option value="dpm_solver_stability" class="gated-feature" data-feature-keys="backend_ed_classic backend_ed_diffusers">DPM Solver (Stability AI)</option>
                             <option value="dpmpp_2s_a">DPM++ 2s Ancestral (Karras)</option>
                             <option value="dpmpp_2m">DPM++ 2m (Karras)</option>
-                            <option value="dpmpp_2m_sde" class="diffusers-only">DPM++ 2m SDE (Karras)</option>
+                            <option value="dpmpp_2m_sde" class="gated-feature" data-feature-keys="backend_ed_diffusers backend_webui">DPM++ 2m SDE</option>
+                            <option value="dpmpp_2m_sde_heun" class="gated-feature" data-feature-keys="backend_webui">DPM++ 2m SDE Heun</option>
+                            <option value="dpmpp_3m_sde" class="gated-feature" data-feature-keys="backend_webui">DPM++ 3M SDE</option>
                             <option value="dpmpp_sde">DPM++ SDE (Karras)</option>
-                            <option value="dpm_adaptive" class="k_diffusion-only">DPM Adaptive (Karras)</option>
-                            <option value="ddpm" class="diffusers-only">DDPM</option>
-                            <option value="deis" class="diffusers-only">DEIS</option>
-                            <option value="unipc_snr" class="k_diffusion-only">UniPC SNR</option>
-                            <option value="unipc_tu">UniPC TU</option>
-                            <option value="unipc_snr_2" class="k_diffusion-only">UniPC SNR 2</option>
-                            <option value="unipc_tu_2" class="k_diffusion-only">UniPC TU 2</option>
-                            <option value="unipc_tq" class="k_diffusion-only">UniPC TQ</option>
+                            <option value="restart" class="gated-feature" data-feature-keys="backend_webui">Restart</option>
+                            <option value="heun_pp2" class="gated-feature" data-feature-keys="backend_webui">Heun PP2</option>
+                            <option value="ipndm" class="gated-feature" data-feature-keys="backend_webui">IPNDM</option>
+                            <option value="ipndm_v" class="gated-feature" data-feature-keys="backend_webui">IPNDM_V</option>
+                            <option value="ddpm" class="gated-feature" data-feature-keys="backend_ed_diffusers backend_webui">DDPM</option>
+                            <option value="deis" class="gated-feature" data-feature-keys="backend_ed_diffusers backend_webui">DEIS</option>
+                            <option value="lcm" class="gated-feature" data-feature-keys="backend_webui">LCM</option>
+                            <option value="forge_flux_realistic" class="gated-feature" data-feature-keys="backend_webui">[Forge] Flux Realistic</option>
+                            <option value="forge_flux_realistic_slow" class="gated-feature" data-feature-keys="backend_webui">[Forge] Flux Realistic (Slow)</option>
+                            <option value="unipc_snr" class="gated-feature" data-feature-keys="backend_ed_classic">UniPC SNR</option>
+                            <option value="unipc_tu" class="gated-feature" data-feature-keys="backend_ed_classic backend_ed_diffusers">UniPC TU</option>
+                            <option value="unipc_snr_2" class="gated-feature" data-feature-keys="backend_ed_classic">UniPC SNR 2</option>
+                            <option value="unipc_tu_2" class="gated-feature" data-feature-keys="backend_ed_classic">UniPC TU 2</option>
+                            <option value="unipc_tq" class="gated-feature" data-feature-keys="backend_ed_classic">UniPC TQ</option>
                         </select>
                         <a href="https://github.com/easydiffusion/easydiffusion/wiki/How-to-Use#samplers" target="_blank"><i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Click to learn more about samplers</span></i></a>
                     </td></tr>
+                    <tr class="pl-5 warning-label displayNone" id="fluxSamplerWarning"><td>Tip:</td><td>This sampler does not work well with Flux!</td></tr>
+                    <tr id="schedulerSelection" class="pl-5 gated-feature" data-feature-keys="backend_webui"><td><label for="scheduler_name">Scheduler:</label></td><td>
+                        <select id="scheduler_name" name="scheduler_name">
+                            <option value="automatic">Automatic</option>
+                            <option value="uniform">Uniform</option>
+                            <option value="karras">Karras</option>
+                            <option value="exponential">Exponential</option>
+                            <option value="polyexponential">Polyexponential</option>
+                            <option value="sgm_uniform">SGM Uniform</option>
+                            <option value="kl_optimal">KL Optimal</option>
+                            <option value="simple" selected>Simple</option>
+                            <option value="normal">Normal</option>
+                            <option value="ddim">DDIM</option>
+                            <option value="beta">Beta</option>
+                            <option value="turbo">Turbo</option>
+                            <option value="align_your_steps">Align Your Steps</option>
+                            <option value="align_your_steps_GITS">Align Your Steps GITS</option>
+                            <option value="align_your_steps_11">Align Your Steps 11</option>
+                            <option value="align_your_steps_32">Align Your Steps 32</option>
+                        </select>
+                    </td></tr>
+                    <tr class="pl-5 warning-label displayNone" id="fluxSchedulerWarning"><td>Tip:</td><td>This scheduler does not work well with Flux!</td></tr>
                     <tr class="pl-5"><td><label>Image Size: </label></td><td id="image-size-options">
                         <select id="width" name="width" value="512">
                             <option value="128">128</option>
@@ -291,7 +392,9 @@
                             <option value="2048">2048</option>
                         </select>
                         <label id="widthLabel" for="width"><small><span>(width)</span></small></label>
-                        <span id="swap-width-height" class="clickable smallButton" style="margin-left: 2px; margin-right:2px;"><i class="fa-solid fa-right-left"><span class="simple-tooltip top-left"> Swap width and height </span></i></span>
+                        <div class="tooltip-container">
+                            <span id="swap-width-height" class="clickable smallButton" style="margin-left: 2px; margin-right:2px;"><i class="fa-solid fa-right-left"><span class="simple-tooltip top-left"> Swap width and height </span></i></span>
+                        </div>
                         <select id="height" name="height" value="512">
                             <option value="128">128</option>
                             <option value="192">192</option>
@@ -318,9 +421,9 @@
                             <span id="recent-resolutions-button" class="clickable"><i class="fa-solid fa-sliders"><span class="simple-tooltip top-left"> Advanced sizes </span></i></span>
                             <div id="recent-resolutions-popup" class="displayNone">
                                 <small>Custom size:</small><br>
-                                <input id="custom-width" name="custom-width" type="number" min="128" value="512" onkeypress="preventNonNumericalInput(event)">
+                                <input id="custom-width" name="custom-width" type="number" min="128" value="512" onkeypress="preventNonNumericalInput(event)" inputmode="numeric">
                                 &times;
-                                <input id="custom-height" name="custom-height" type="number" min="128" value="512" onkeypress="preventNonNumericalInput(event)"><br>
+                                <input id="custom-height" name="custom-height" type="number" min="128" value="512" onkeypress="preventNonNumericalInput(event)" inputmode="numeric"><br>
                                 <small>Resize:</small><br>
                                 <input id="resize-slider" name="resize-slider" class="editor-slider" value="1" type="range" min="0.4" max="2" step="0.005" style="width:100%;"><br>
                                 <div id="enlarge-buttons"><button data-factor="0.5" class="tertiaryButton smallButton">×0.5</button>&nbsp;<button data-factor="1.2" class="tertiaryButton smallButton">×1.2</button>&nbsp;<button data-factor="1.5" class="tertiaryButton smallButton">×1.5</button>&nbsp;<button data-factor="2" class="tertiaryButton smallButton">×2</button>&nbsp;<button data-factor="3" class="tertiaryButton smallButton">×3</button></div>
@@ -340,12 +443,15 @@
                                 </div>
                             </div>
                         </div>
-                        <div id="small_image_warning" class="displayNone">Small image sizes can cause bad image quality</div>
+                        <div id="small_image_warning" class="displayNone warning-label">Small image sizes can cause bad image quality</div>
                     </td></tr>
-                    <tr class="pl-5"><td><label for="num_inference_steps">Inference Steps:</label></td><td> <input id="num_inference_steps" name="num_inference_steps" type="number" min="1" step="1" style="width: 42pt" value="25" onkeypress="preventNonNumericalInput(event)"></td></tr>
-                    <tr class="pl-5"><td><label for="guidance_scale_slider">Guidance Scale:</label></td><td> <input id="guidance_scale_slider" name="guidance_scale_slider" class="editor-slider" value="75" type="range" min="11" max="500"> <input id="guidance_scale" name="guidance_scale" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)"></td></tr>
-                    <tr id="prompt_strength_container" class="pl-5"><td><label for="prompt_strength_slider">Prompt Strength:</label></td><td> <input id="prompt_strength_slider" name="prompt_strength_slider" class="editor-slider" value="80" type="range" min="0" max="99"> <input id="prompt_strength" name="prompt_strength" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)"><br/></td></tr>
-                    <tr id="lora_model_container" class="pl-5">
+                    <tr class="pl-5"><td><label for="num_inference_steps">Inference Steps:</label></td><td> <input id="num_inference_steps" name="num_inference_steps" type="number" min="1" step="1" style="width: 42pt" value="25" onkeypress="preventNonNumericalInput(event)" inputmode="numeric"></td></tr>
+                    <tr class="pl-5 warning-label displayNone" id="fluxSchedulerStepsWarning"><td>Tip:</td><td>This scheduler needs 15 steps or more</td></tr>
+                    <tr class="pl-5"><td><label for="guidance_scale_slider">Guidance Scale:</label></td><td> <input id="guidance_scale_slider" name="guidance_scale_slider" class="editor-slider" value="75" type="range" min="11" max="500"> <input id="guidance_scale" name="guidance_scale" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)" inputmode="decimal"></td></tr>
+                    <tr class="pl-5 displayNone warning-label" id="guidanceWarning"><td></td><td id="guidanceWarningText"></td></tr>
+                    <tr id="prompt_strength_container" class="pl-5"><td><label for="prompt_strength_slider">Prompt Strength:</label></td><td> <input id="prompt_strength_slider" name="prompt_strength_slider" class="editor-slider" value="80" type="range" min="0" max="99"> <input id="prompt_strength" name="prompt_strength" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)" inputmode="decimal"><br/></td></tr>
+                    <tr id="distilled_guidance_scale_container" class="pl-5 gated-feature" data-feature-keys="backend_webui"><td><label for="distilled_guidance_scale_slider">Distilled Guidance:</label></td><td> <input id="distilled_guidance_scale_slider" name="distilled_guidance_scale_slider" class="editor-slider" value="35" type="range" min="11" max="500"> <input id="distilled_guidance_scale" name="distilled_guidance_scale" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)" inputmode="decimal"></td></tr>
+                    <tr id="lora_model_container" class="pl-5 gated-feature" data-feature-keys="backend_ed_diffusers backend_webui">
                         <td>
                             <label for="lora_model">LoRA:</label>
                         </td>
@@ -353,14 +459,14 @@
                             <div id="lora_model" data-path=""></div>
                         </td>
                     </tr>
-                    <tr id="hypernetwork_model_container" class="pl-5"><td><label for="hypernetwork_model">Hypernetwork:</label></td><td>
+                    <tr id="hypernetwork_model_container" class="pl-5 gated-feature" data-feature-keys="backend_ed_classic"><td><label for="hypernetwork_model">Hypernetwork:</label></td><td>
                         <input id="hypernetwork_model" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" />
                     </td></tr>
-                    <tr id="hypernetwork_strength_container" class="pl-5">
+                    <tr id="hypernetwork_strength_container" class="pl-5 gated-feature" data-feature-keys="backend_ed_classic">
                         <td><label for="hypernetwork_strength_slider">Hypernetwork Strength:</label></td>
-                        <td> <input id="hypernetwork_strength_slider" name="hypernetwork_strength_slider" class="editor-slider" value="100" type="range" min="0" max="100"> <input id="hypernetwork_strength" name="hypernetwork_strength" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)"><br/></td>
+                        <td> <input id="hypernetwork_strength_slider" name="hypernetwork_strength_slider" class="editor-slider" value="100" type="range" min="0" max="100"> <input id="hypernetwork_strength" name="hypernetwork_strength" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)" inputmode="decimal"><br/></td>
                     </tr>
-                    <tr id="tiling_container" class="pl-5">
+                    <tr id="tiling_container" class="pl-5 gated-feature" data-feature-keys="backend_ed_diffusers">
                         <td><label for="tiling">Seamless Tiling:</label></td>
                         <td class="diffusers-restart-needed">
                             <select id="tiling" name="tiling">
@@ -383,8 +489,15 @@
                         </span>
                     </td></tr>
                     <tr class="pl-5" id="output_quality_row"><td><label for="output_quality">Image Quality:</label></td><td>
-                    <input id="output_quality_slider" name="output_quality" class="editor-slider" value="75" type="range" min="10" max="95"> <input id="output_quality" name="output_quality" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)">
+                    <input id="output_quality_slider" name="output_quality" class="editor-slider" value="75" type="range" min="10" max="95"> <input id="output_quality" name="output_quality" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)" inputmode="numeric">
                     </td></tr>
+                    <tr class="pl-5 gated-feature" data-feature-keys="backend_ed_diffusers">
+                        <td><label for="tiling">Enable VAE Tiling:</label></td>
+                        <td class="diffusers-restart-needed">
+                            <input id="enable_vae_tiling" name="enable_vae_tiling" type="checkbox" checked>
+                            <label><small>Optimizes memory for larger images</small></label>
+                        </td>
+                    </tr>
                     </table></div>
 
                     <div><ul>
@@ -393,8 +506,8 @@
                     <li class="pl-5" id="use_face_correction_container">
                         <input id="use_face_correction" name="use_face_correction" type="checkbox"> <label for="use_face_correction">Fix incorrect faces and eyes</label> <div style="display:inline-block;"><input id="gfpgan_model" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" /></div>
                         <table id="codeformer_settings" class="displayNone sub-settings">
-                            <tr class="pl-5"><td><label for="codeformer_fidelity_slider">Strength:</label></td><td><input id="codeformer_fidelity_slider" name="codeformer_fidelity_slider" class="editor-slider" value="5" type="range" min="0" max="10"> <input id="codeformer_fidelity" name="codeformer_fidelity" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)"></td></tr>
-                            <tr class="pl-5"><td><label for="codeformer_upscale_faces">Upscale Faces:</label></td><td><input id="codeformer_upscale_faces" name="codeformer_upscale_faces" type="checkbox" checked> <label><small>(improves the resolution of faces)</small></label></td></tr>
+                            <tr class="pl-5"><td><label for="codeformer_fidelity_slider">Strength:</label></td><td><input id="codeformer_fidelity_slider" name="codeformer_fidelity_slider" class="editor-slider" value="5" type="range" min="0" max="10"> <input id="codeformer_fidelity" name="codeformer_fidelity" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)" inputmode="decimal"></td></tr>
+                            <tr class="pl-5 gated-feature" data-feature-keys="backend_ed_diffusers"><td><label for="codeformer_upscale_faces">Upscale Faces:</label></td><td><input id="codeformer_upscale_faces" name="codeformer_upscale_faces" type="checkbox" checked> <label><small>(improves the resolution of faces)</small></label></td></tr>
                         </table>
                     </li>
                     <li class="pl-5">
@@ -407,10 +520,16 @@
                         <select id="upscale_model" name="upscale_model">
                             <option value="RealESRGAN_x4plus" selected>RealESRGAN_x4plus</option>
                             <option value="RealESRGAN_x4plus_anime_6B">RealESRGAN_x4plus_anime_6B</option>
-                            <option value="latent_upscaler">Latent Upscaler 2x</option>
+                            <option value="ESRGAN_4x" class="pl-5 gated-feature" data-feature-keys="backend_webui">ESRGAN_4x</option>
+                            <option value="Lanczos" class="pl-5 gated-feature" data-feature-keys="backend_webui">Lanczos</option>
+                            <option value="Nearest" class="pl-5 gated-feature" data-feature-keys="backend_webui">Nearest</option>
+                            <option value="ScuNET" class="pl-5 gated-feature" data-feature-keys="backend_webui">ScuNET GAN</option>
+                            <option value="ScuNET PSNR" class="pl-5 gated-feature" data-feature-keys="backend_webui">ScuNET PSNR</option>
+                            <option value="SwinIR_4x" class="pl-5 gated-feature" data-feature-keys="backend_webui">SwinIR 4x</option>
+                            <option value="latent_upscaler" class="pl-5 gated-feature" data-feature-keys="backend_ed_classic backend_ed_diffusers">Latent Upscaler 2x</option>
                         </select>
                         <table id="latent_upscaler_settings" class="displayNone sub-settings">
-                            <tr class="pl-5"><td><label for="latent_upscaler_steps_slider">Upscaling Steps:</label></td><td><input id="latent_upscaler_steps_slider" name="latent_upscaler_steps_slider" class="editor-slider" value="10" type="range" min="1" max="50"> <input id="latent_upscaler_steps" name="latent_upscaler_steps" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)"></td></tr>
+                            <tr class="pl-5"><td><label for="latent_upscaler_steps_slider">Upscaling Steps:</label></td><td><input id="latent_upscaler_steps_slider" name="latent_upscaler_steps_slider" class="editor-slider" value="10" type="range" min="1" max="50"> <input id="latent_upscaler_steps" name="latent_upscaler_steps" size="4" pattern="^[0-9\.]+$" onkeypress="preventNonNumericalInput(event)" inputmode="numeric"></td></tr>
                         </table>
                     </li>
                     <li class="pl-5"><input id="show_only_filtered_image" name="show_only_filtered_image" type="checkbox" checked> <label for="show_only_filtered_image">Show only the corrected/upscaled image</label></li>
@@ -455,14 +574,14 @@
                         <div class="dropdown-content">
                             <div class="dropdown-item">
                                 <input id="thumbnail_size" name="thumbnail_size" class="editor-slider" type="range" value="70" min="5" max="200" oninput="sliderUpdate(event)"> 
-                                <input id="thumbnail_size-input" name="thumbnail_size-input" size="3" value="70" pattern="^[0-9.]+$" onkeypress="preventNonNumericalInput(event)" oninput="sliderUpdate(event)">&nbsp;%
+                                <input id="thumbnail_size-input" name="thumbnail_size-input" size="3" value="70" pattern="^[0-9.]+$" onkeypress="preventNonNumericalInput(event)" oninput="sliderUpdate(event)" inputmode="numeric">&nbsp;%
                             </div>
                         </div>
                     </div>
                     <div class="clearfix" style="clear: both;"></div>
                 </div>
                 <div id="supportBanner" class="displayNone">
-                    If you found this project useful and want to help keep it alive, please consider <a href="https://ko-fi.com/easydiffusion" target="_blank">buying me a coffee</a> or <a href="https://www.patreon.com/EasyDiffusion" target="_blank">supporting me on Patreon</a> to help cover the cost of development and maintenance! Or even better, <a href="https://cmdr2.itch.io/easydiffusion" target="_blank">purchasing it at the full price</a>. Thank you for your support!
+                    If you found this project useful and want to help keep it alive, please consider <a href="https://ko-fi.com/easydiffusion" target="_blank">buying me a coffee</a> to help cover the cost of development and maintenance! Thanks for your support!
                 </div>
             </div>
         </div>
@@ -508,28 +627,44 @@
                 <div class="float-container">
                     <div class="float-child">
                         <h1>Help</h1>
-                        <ul id="help-links">
-                            <li><span class="help-section">Using the software</span>
+                        <div id="help-links">
+                            <h4><span class="help-section"><b>Basics</b></span></h4>
                             <ul>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/How-To-Use" target="_blank"><i class="fa-solid fa-book fa-fw"></i> How to use</a>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/UI-Overview" target="_blank"><i class="fa-solid fa-list fa-fw"></i> UI Overview</a>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Writing-Prompts" target="_blank"><i class="fa-solid fa-pen-to-square fa-fw"></i> Writing prompts</a>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Inpainting" target="_blank"><i class="fa-solid fa-paintbrush fa-fw"></i> Inpainting</a>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Run-on-Multiple-GPUs" target="_blank"><i class="fa-solid fa-paintbrush fa-fw"></i> Run on Multiple GPUs</a>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/How-To-Use" target="_blank">How to use</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Writing-Prompts" target="_blank">Writing prompts</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Image-Modifiers" target="_blank">Image Modifiers</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Inpainting" target="_blank">Inpainting</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Samplers" target="_blank">Samplers</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/UI-Overview" target="_blank">Summary of every UI option</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Troubleshooting" target="_blank">Common error messages (and solutions)</a></li>
                             </ul>
   
-                            <li><span class="help-section">Installation</span>
+                            <h4><span class="help-section"><b>Intermediate</b></span></h4>
                             <ul>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Troubleshooting" target="_blank"><i class="fa-solid fa-circle-question fa-fw"></i> Troubleshooting</a>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Custom-Models" target="_blank">Custom Models</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Prompt-Syntax" target="_blank">Prompt Syntax (weights, emphasis etc)</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/UI-Plugins" target="_blank">UI Plugins</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Embeddings" target="_blank">Embeddings</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/LoRA" target="_blank">LoRA</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/SDXL" target="_blank">SDXL</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/ControlNet" target="_blank">ControlNet</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Seamless-Tiling" target="_blank">Seamless Tiling</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/xFormers" target="_blank">xFormers</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/The-beta-channel" target="_blank">The beta channel</a></li>
                             </ul>
   
-                            <li><span class="help-section">Downloadable Content</span>
+                            <h4><span class="help-section"><b>Advanced topics</b></span></h4>
                             <ul>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Custom-Models" target="_blank"><i class="fa-solid fa-images fa-fw"></i> Custom Models</a>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/UI-Plugins" target="_blank"><i class="fa-solid fa-puzzle-piece fa-fw"></i> UI Plugins</a>
-                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/VAE-Variational-Auto-Encoder" target="_blank"><i class="fa-solid fa-hand-sparkles fa-fw"></i> VAE Variational Auto Encoder</a>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Run-on-Multiple-GPUs" target="_blank">Run on Multiple GPUs</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Model-Merging" target="_blank">Model Merging</a></li>
+                                <li> <a href="https://github.com/easydiffusion/easydiffusion/wiki/Custom-Modifiers" target="_blank">Custom Modifiers</a></li>
                             </ul>
-                        </ul>
+
+                            <h4><span class="help-section"><b>Misc</b></span></h4>
+                            <ul>
+                                <li> <a href="https://theally.notion.site/The-Definitive-Stable-Diffusion-Glossary-1d1e6d15059c41e6a6b4306b4ecd9df9" target="_blank">Glossary of Stable Diffusion related terms</a></li>
+                            </ul>
+                        </div>
                     </div>
   
                     <div class="float-child">
@@ -712,7 +847,7 @@
                 <span class="embeddings-action-text">Expand Categories</span>
             </button>
             <i class="fa-solid fa-magnifying-glass"></i>
-            <input id="embeddings-search-box" type="text" spellcheck="false" autocomplete="off" placeholder="Search...">
+            <input id="embeddings-search-box" type="text" spellcheck="false" autocomplete="off" placeholder="Search..." inputmode="search">
             <label for="embedding-card-size-selector"><small>Thumbnail Size:</small></label>
             <select id="embedding-card-size-selector" name="embedding-card-size-selector">
                 <option value="-2">0</option>
@@ -798,6 +933,8 @@
             <p>This license of this software forbids you from sharing any content that violates any laws, produce any harm to a person, disseminate any personal information that would be meant for harm, <br/>spread misinformation and target vulnerable groups. For the full list of restrictions please read <a href="https://github.com/easydiffusion/easydiffusion/blob/main/LICENSE" target="_blank">the license</a>.</p>
             <p>By using this software, you consent to the terms and conditions of the license.</p>
         </div>
+        <input id="test_diffusers" type="checkbox" style="display: none" checked /> <!-- for backwards compatibility -->
+        <input id="use_v3_engine" type="checkbox" style="display: none" checked /> <!-- for backwards compatibility -->
     </div>
 </div>
 </body>

ui/main.py

@@ -9,6 +9,3 @@ server.init()
 model_manager.init()
 app.init_render_threads()
 bucket_manager.init()
-
-# start the browser ui
-app.open_browser()

ui/media/css/auto-save.css

@@ -79,6 +79,7 @@
 }
 
 .parameters-table .fa-fire,
-.parameters-table .fa-bolt {
+.parameters-table .fa-bolt,
+.parameters-table .fa-robot {
     color: #F7630C;
 }

ui/media/css/main.css

@@ -36,6 +36,15 @@ code {
     transform: translateY(4px);
     cursor: pointer;
 }
+#engine-logo {
+    font-size: 8pt;
+    padding-left: 10pt;
+    color: var(--small-label-color);
+}
+#engine-logo a {
+    text-decoration: none;
+    /* color: var(--small-label-color); */
+}
 #prompt {
     width: 100%;
     height: 65pt;
@@ -541,7 +550,7 @@ div.img-preview img {
     position: relative;
     background: var(--background-color4);
     display: flex;
-    padding: 12px 0 0;
+    padding: 6px 0 0;
 }
 .tab .icon {
     padding-right: 4pt;
@@ -609,11 +618,18 @@ div.img-preview img {
     margin: auto;
     padding: 0px;
 }
+#help-links ul {
+    list-style-type: disc;
+    padding-left: 12pt;
+}
 #help-links li {
-    padding-bottom: 12pt;
+    padding-bottom: 6pt;
     display: block;
     font-size: 10pt;
 }
+#help-links ul li {
+    display: list-item;
+}
 #help-links li .fa-fw {
     padding-right: 2pt;
 }
@@ -650,6 +666,15 @@ div.img-preview img {
     display: block;
 }
 
+.gated-feature {
+    display: none;
+}
+
+.warning-label {
+    font-size: smaller;
+    color: var(--status-orange);
+}
+
 .display-settings {
     float: right;
     position: relative;
@@ -1207,6 +1232,12 @@ input::file-selector-button {
         visibility: visible;
     }
 }
+
+.tooltip-container {
+    display: inline-block;
+    position: relative;
+}
+
 .simple-tooltip.right {
     right: 0px;
     top: 50%;
@@ -1446,11 +1477,6 @@ div.top-right {
     margin-top: 6px;
 }
 
-#small_image_warning {
-    font-size: smaller;
-    color: var(--status-orange);
-}
-
 button#save-system-settings-btn {
     padding: 4pt 8pt;
 }
@@ -2026,4 +2052,4 @@ div#enlarge-buttons {
     border-radius: 4pt;
     padding-top: 6pt;
     color: var(--small-label-color);
-}
+}

ui/media/js/auto-save.js

@@ -16,10 +16,12 @@ const SETTINGS_IDS_LIST = [
     "clip_skip",
     "vae_model",
     "sampler_name",
+    "scheduler_name",
     "width",
     "height",
     "num_inference_steps",
     "guidance_scale",
+    "distilled_guidance_scale",
     "prompt_strength",
     "tiling",
     "output_format",
@@ -29,6 +31,8 @@ const SETTINGS_IDS_LIST = [
     "stream_image_progress",
     "use_face_correction",
     "gfpgan_model",
+    "codeformer_fidelity",
+    "codeformer_upscale_faces",
     "use_upscale",
     "upscale_amount",
     "latent_upscaler_steps",
@@ -56,6 +60,9 @@ const SETTINGS_IDS_LIST = [
     "extract_lora_from_prompt",
     "embedding-card-size-selector",
     "lora_model",
+    "text_encoder_model",
+    "enable_vae_tiling",
+    "controlnet_alpha",
 ]
 
 const IGNORE_BY_DEFAULT = ["prompt"]
@@ -176,23 +183,6 @@ function loadSettings() {
             }
         })
         CURRENTLY_LOADING_SETTINGS = false
-    } else if (localStorage.length < 2) {
-        // localStorage is too short for OldSettings
-        // So this is likely the first time Easy Diffusion is running.
-        // Initialize vram_usage_level based on the available VRAM
-        function initGPUProfile(event) {
-            if (
-                "detail" in event &&
-                "active" in event.detail &&
-                "cuda:0" in event.detail.active &&
-                event.detail.active["cuda:0"].mem_total < 4.5
-            ) {
-                vramUsageLevelField.value = "low"
-                vramUsageLevelField.dispatchEvent(new Event("change"))
-            }
-            document.removeEventListener("system_info_update", initGPUProfile)
-        }
-        document.addEventListener("system_info_update", initGPUProfile)
     } else {
         CURRENTLY_LOADING_SETTINGS = true
         tryLoadOldSettings()

ui/media/js/dnd.js

@@ -131,6 +131,15 @@ const TASK_MAPPING = {
         readUI: () => parseFloat(guidanceScaleField.value),
         parse: (val) => parseFloat(val),
     },
+    distilled_guidance_scale: {
+        name: "Distilled Guidance",
+        setUI: (distilled_guidance_scale) => {
+            distilledGuidanceScaleField.value = distilled_guidance_scale
+            updateDistilledGuidanceScaleSlider()
+        },
+        readUI: () => parseFloat(distilledGuidanceScaleField.value),
+        parse: (val) => parseFloat(val),
+    },
     prompt_strength: {
         name: "Prompt Strength",
         setUI: (prompt_strength) => {
@@ -242,6 +251,14 @@ const TASK_MAPPING = {
         readUI: () => samplerField.value,
         parse: (val) => val,
     },
+    scheduler_name: {
+        name: "Scheduler",
+        setUI: (scheduler_name) => {
+            schedulerField.value = scheduler_name
+        },
+        readUI: () => schedulerField.value,
+        parse: (val) => val,
+    },
     use_stable_diffusion_model: {
         name: "Stable Diffusion model",
         setUI: (use_stable_diffusion_model) => {
@@ -268,7 +285,11 @@ const TASK_MAPPING = {
     tiling: {
         name: "Tiling",
         setUI: (val) => {
-            tilingField.value = val
+            if (val === null || val === "None") {
+                tilingField.value = "none"
+            } else {
+                tilingField.value = val
+            }
         },
         readUI: () => tilingField.value,
         parse: (val) => val,
@@ -305,10 +326,21 @@ const TASK_MAPPING = {
         readUI: () => controlImageFilterField.value,
         parse: (val) => val,
     },
+    control_alpha: {
+        name: "ControlNet Strength",
+        setUI: (control_alpha) => {
+            control_alpha = control_alpha || 1.0
+            controlAlphaField.value = control_alpha
+            updateControlAlphaSlider()
+        },
+        readUI: () => parseFloat(controlAlphaField.value),
+        parse: (val) => val === null ? 1.0 : parseFloat(val),
+    },
     use_lora_model: {
         name: "LoRA model",
         setUI: (use_lora_model) => {
             let modelPaths = []
+            use_lora_model = use_lora_model === null ? "" : use_lora_model
             use_lora_model = Array.isArray(use_lora_model) ? use_lora_model : [use_lora_model]
             use_lora_model.forEach((m) => {
                 if (m.includes("models\\lora\\")) {
@@ -362,6 +394,45 @@ const TASK_MAPPING = {
             return val
         },
     },
+    use_text_encoder_model: {
+        name: "Text Encoder model",
+        setUI: (use_text_encoder_model) => {
+            let modelPaths = []
+            use_text_encoder_model = use_text_encoder_model === null ? "" : use_text_encoder_model
+            use_text_encoder_model = Array.isArray(use_text_encoder_model) ? use_text_encoder_model : [use_text_encoder_model]
+            use_text_encoder_model.forEach((m) => {
+                if (m.includes("models\\text-encoder\\")) {
+                    m = m.split("models\\text-encoder\\")[1]
+                } else if (m.includes("models\\\\text-encoder\\\\")) {
+                    m = m.split("models\\\\text-encoder\\\\")[1]
+                } else if (m.includes("models/text-encoder/")) {
+                    m = m.split("models/text-encoder/")[1]
+                }
+                m = m.replaceAll("\\\\", "/")
+                m = getModelPath(m, [".safetensors", ".sft"])
+                modelPaths.push(m)
+            })
+            textEncoderModelField.modelNames = modelPaths
+        },
+        readUI: () => {
+            return textEncoderModelField.modelNames
+        },
+        parse: (val) => {
+            val = !val || val === "None" ? "" : val
+            if (typeof val === "string" && val.includes(",")) {
+                val = val.split(",")
+                val = val.map((v) => v.trim())
+                val = val.map((v) => v.replaceAll("\\", "\\\\"))
+                val = val.map((v) => v.replaceAll('"', ""))
+                val = val.map((v) => v.replaceAll("'", ""))
+                val = val.map((v) => '"' + v + '"')
+                val = "[" + val + "]"
+                val = JSON.parse(val)
+            }
+            val = Array.isArray(val) ? val : [val]
+            return val
+        },
+    },
     use_hypernetwork_model: {
         name: "Hypernetwork model",
         setUI: (use_hypernetwork_model) => {
@@ -525,6 +596,11 @@ function restoreTaskToUI(task, fieldsToSkip) {
         // listen for inpainter loading event, which happens AFTER the main image loads (which reloads the inpai
         controlImagePreview.src = task.reqBody.control_image
     }
+
+    if ("use_controlnet_model" in task.reqBody && task.reqBody.use_controlnet_model && !("control_alpha" in task.reqBody)) {
+        controlAlphaField.value = 1.0
+        updateControlAlphaSlider()
+    }
 }
 function readUI() {
     const reqBody = {}
@@ -570,19 +646,24 @@ const TASK_TEXT_MAPPING = {
     seed: "Seed",
     num_inference_steps: "Steps",
     guidance_scale: "Guidance Scale",
+    distilled_guidance_scale: "Distilled Guidance",
     prompt_strength: "Prompt Strength",
     use_face_correction: "Use Face Correction",
     use_upscale: "Use Upscaling",
     upscale_amount: "Upscale By",
     sampler_name: "Sampler",
+    scheduler_name: "Scheduler",
     negative_prompt: "Negative Prompt",
     use_stable_diffusion_model: "Stable Diffusion model",
     use_hypernetwork_model: "Hypernetwork model",
     hypernetwork_strength: "Hypernetwork Strength",
     use_lora_model: "LoRA model",
     lora_alpha: "LoRA Strength",
+    use_text_encoder_model: "Text Encoder model",
     use_controlnet_model: "ControlNet model",
     control_filter_to_apply: "ControlNet Filter",
+    control_alpha: "ControlNet Strength",
+    tiling: "Seamless Tiling",
 }
 function parseTaskFromText(str) {
     const taskReqBody = {}

ui/media/js/main.js

@@ -12,8 +12,16 @@ const taskConfigSetup = {
         seed: { value: ({ seed }) => seed, label: "Seed" },
         dimensions: { value: ({ reqBody }) => `${reqBody?.width}x${reqBody?.height}`, label: "Dimensions" },
         sampler_name: "Sampler",
+        scheduler_name: {
+            label: "Scheduler",
+            visible: ({ reqBody }) => reqBody?.scheduler_name,
+        },
         num_inference_steps: "Inference Steps",
         guidance_scale: "Guidance Scale",
+        distilled_guidance_scale: {
+            label: "Distilled Guidance Scale",
+            visible: ({ reqBody }) => reqBody?.distilled_guidance_scale,
+        },
         use_stable_diffusion_model: "Model",
         clip_skip: {
             label: "Clip Skip",
@@ -22,7 +30,8 @@ const taskConfigSetup = {
         },
         tiling: {
             label: "Tiling",
-            visible: ({ reqBody }) => reqBody?.tiling != "none",
+            visible: ({ reqBody }) =>
+                reqBody?.tiling != "none" && reqBody?.tiling !== null && reqBody?.tiling !== undefined,
             value: ({ reqBody }) => reqBody?.tiling,
         },
         use_vae_model: {
@@ -45,11 +54,16 @@ const taskConfigSetup = {
             label: "Hypernetwork Strength",
             visible: ({ reqBody }) => !!reqBody?.use_hypernetwork_model,
         },
+        use_text_encoder_model: { label: "Text Encoder", visible: ({ reqBody }) => !!reqBody?.use_text_encoder_model },
         use_lora_model: { label: "Lora Model", visible: ({ reqBody }) => !!reqBody?.use_lora_model },
         lora_alpha: { label: "Lora Strength", visible: ({ reqBody }) => !!reqBody?.use_lora_model },
         preserve_init_image_color_profile: "Preserve Color Profile",
         strict_mask_border: "Strict Mask Border",
         use_controlnet_model: "ControlNet Model",
+        control_alpha: {
+            label: "ControlNet Strength",
+            visible: ({ reqBody }) => !!reqBody?.use_controlnet_model,
+        },
     },
     pluginTaskConfig: {},
     getCSSKey: (key) =>
@@ -71,6 +85,8 @@ let numOutputsParallelField = document.querySelector("#num_outputs_parallel")
 let numInferenceStepsField = document.querySelector("#num_inference_steps")
 let guidanceScaleSlider = document.querySelector("#guidance_scale_slider")
 let guidanceScaleField = document.querySelector("#guidance_scale")
+let distilledGuidanceScaleSlider = document.querySelector("#distilled_guidance_scale_slider")
+let distilledGuidanceScaleField = document.querySelector("#distilled_guidance_scale")
 let outputQualitySlider = document.querySelector("#output_quality_slider")
 let outputQualityField = document.querySelector("#output_quality")
 let outputQualityRow = document.querySelector("#output_quality_row")
@@ -98,6 +114,8 @@ let controlImagePreview = document.querySelector("#control_image_preview")
 let controlImageClearBtn = document.querySelector(".control_image_clear")
 let controlImageContainer = document.querySelector("#control_image_wrapper")
 let controlImageFilterField = document.querySelector("#control_image_filter")
+let controlAlphaSlider = document.querySelector("#controlnet_alpha_slider")
+let controlAlphaField = document.querySelector("#controlnet_alpha")
 let applyColorCorrectionField = document.querySelector("#apply_color_correction")
 let strictMaskBorderField = document.querySelector("#strict_mask_border")
 let colorCorrectionSetting = document.querySelector("#apply_color_correction_setting")
@@ -106,6 +124,8 @@ let promptStrengthSlider = document.querySelector("#prompt_strength_slider")
 let promptStrengthField = document.querySelector("#prompt_strength")
 let samplerField = document.querySelector("#sampler_name")
 let samplerSelectionContainer = document.querySelector("#samplerSelection")
+let schedulerField = document.querySelector("#scheduler_name")
+let schedulerSelectionContainer = document.querySelector("#schedulerSelection")
 let useFaceCorrectionField = document.querySelector("#use_face_correction")
 let gfpganModelField = new ModelDropdown(document.querySelector("#gfpgan_model"), ["gfpgan", "codeformer"], "", false)
 let useUpscalingField = document.querySelector("#use_upscale")
@@ -122,12 +142,14 @@ let tilingField = document.querySelector("#tiling")
 let controlnetModelField = new ModelDropdown(document.querySelector("#controlnet_model"), "controlnet", "None", false)
 let vaeModelField = new ModelDropdown(document.querySelector("#vae_model"), "vae", "None")
 let loraModelField = new MultiModelSelector(document.querySelector("#lora_model"), "lora", "LoRA", 0.5, 0.02)
+let textEncoderModelField = new MultiModelSelector(document.querySelector("#text_encoder_model"), "text-encoder", "Text Encoder", 0.5, 0.02, false)
 let hypernetworkModelField = new ModelDropdown(document.querySelector("#hypernetwork_model"), "hypernetwork", "None")
 let hypernetworkStrengthSlider = document.querySelector("#hypernetwork_strength_slider")
 let hypernetworkStrengthField = document.querySelector("#hypernetwork_strength")
 let outputFormatField = document.querySelector("#output_format")
 let outputLosslessField = document.querySelector("#output_lossless")
 let outputLosslessContainer = document.querySelector("#output_lossless_container")
+let enableVAETilingField = document.querySelector("#enable_vae_tiling")
 let blockNSFWField = document.querySelector("#block_nsfw")
 let showOnlyFilteredImageField = document.querySelector("#show_only_filtered_image")
 let updateBranchLabel = document.querySelector("#updateBranchLabel")
@@ -510,10 +532,10 @@ function showImages(reqBody, res, outputContainer, livePreview) {
                     { text: "Upscale", on_click: onUpscaleClick },
                     { text: "Fix Faces", on_click: onFixFacesClick },
                 ],
-                {
+                { 
                     text: "Use as Thumbnail",
                     on_click: onUseAsThumbnailClick,
-                    filter: (req, img) => "use_embeddings_model" in req,
+                    filter: (req, img) => "use_embeddings_model" in req || "use_lora_model" in req
                 },
             ]
 
@@ -603,6 +625,13 @@ function onUseAsInputClick(req, img) {
     initImagePreview.src = imgData
 
     maskSetting.checked = false
+
+    //Force the image settings size to match the input, as inpaint currently only works correctly
+    //if input image and generate sizes match.
+    addImageSizeOption(img.naturalWidth);
+    addImageSizeOption(img.naturalHeight);
+    widthField.value = img.naturalWidth;
+    heightField.value = img.naturalHeight;
 }
 
 function onUseForControlnetClick(req, img) {
@@ -680,7 +709,7 @@ function getAllModelNames(type) {
 
 // gets a flattened list of all models of a certain type. e.g. "path/subpath/modelname"
 // use the filter to search for all models having a certain name.
-function getAllModelPathes(type,filter="") {
+function getAllModelPathes(type, filter = "") {
     function f(tree, prefix) {
         if (tree == undefined) {
             return []
@@ -690,7 +719,7 @@ function getAllModelPathes(type,filter="") {
             if (typeof e == "object") {
                 result = result.concat(f(e[1], prefix + e[0] + "/"))
             } else {
-                if (filter=="" || e==filter) {
+                if (filter == "" || e == filter) {
                     result.push(prefix + e)
                 }
             }
@@ -700,7 +729,6 @@ function getAllModelPathes(type,filter="") {
     return f(modelsOptions[type], "")
 }
 
-
 function onUseAsThumbnailClick(req, img) {
     let scale = 1
     let targetWidth = img.naturalWidth
@@ -748,25 +776,45 @@ function onUseAsThumbnailClick(req, img) {
         onUseAsThumbnailClick.croppr.setImage(img.src)
     }
 
-    let embeddings = req.use_embeddings_model.map((e) => e.split("/").pop())
-    let LORA = []
+    useAsThumbSelect.innerHTML=""
 
-    if ("use_lora_model" in req) {
-        LORA = req.use_lora_model
+    if ("use_embeddings_model" in req) {
+        let embeddings = req.use_embeddings_model.map((e) => e.split("/").pop())
+
+        let embOptions = document.createElement("optgroup")
+        embOptions.label = "Embeddings"
+        embOptions.replaceChildren(
+            ...embeddings.map((e) => {
+                let option = document.createElement("option")
+                option.innerText = e
+                option.dataset["type"] = "embeddings"
+                return option
+            })
+        )
+        useAsThumbSelect.appendChild(embOptions)
     }
 
-    let optgroup = document.createElement("optgroup")
-    optgroup.label = "Embeddings"
-    optgroup.replaceChildren(
-        ...embeddings.map((e) => {
-            let option = document.createElement("option")
-            option.innerText = e
-            option.dataset["type"] = "embeddings"
-            return option
-        })
-    )
 
-    useAsThumbSelect.replaceChildren(optgroup)
+    if ("use_lora_model" in req) {
+        let LORA = req.use_lora_model
+        if (typeof LORA == "string") {
+            LORA = [LORA]
+        }
+        LORA = LORA.map((e) => e.split("/").pop())
+
+        let loraOptions = document.createElement("optgroup")
+        loraOptions.label = "LORA"
+        loraOptions.replaceChildren(
+            ...LORA.map((e) => {
+                let option = document.createElement("option")
+                option.innerText = e
+                option.dataset["type"] = "lora"
+                return option
+            })
+        )
+        useAsThumbSelect.appendChild(loraOptions)
+    }
+
     useAsThumbDialog.showModal()
     onUseAsThumbnailClick.scale = scale
 }
@@ -782,6 +830,50 @@ useAsThumbCancelBtn.addEventListener("click", () => {
     useAsThumbDialog.close()
 })
 
+const Bucket = {
+    upload(path, blob) {
+        const formData = new FormData()
+        formData.append("file", blob)
+        return fetch(`bucket/${path}`, {
+            method: "POST",
+            body: formData,
+        })
+    },
+
+    getImageAsDataURL(path) {
+        return fetch(`bucket/${path}`)
+            .then((response) => {
+                if (response.status == 200) {
+                    return response.blob()
+                } else {
+                    throw new Error("Bucket error")
+                }
+            })
+            .then((blob) => {
+                return new Promise((resolve, reject) => {
+                    const reader = new FileReader()
+                    reader.onload = () => resolve(reader.result)
+                    reader.onerror = reject
+                    reader.readAsDataURL(blob)
+                })
+            })
+    },
+
+    getList(path) {
+        return fetch(`bucket/${path}`)
+            .then((response) => (response.status == 200 ? response.json() : []))
+    },
+
+    store(path, data) {
+        return Bucket.upload(`${path}.json`, JSON.stringify(data))
+    },
+
+    retrieve(path) {
+        return fetch(`bucket/${path}.json`)
+            .then((response) => (response.status == 200 ? response.json() : null))
+    },
+}
+
 useAsThumbSaveBtn.addEventListener("click", (e) => {
     let scale = 1 / onUseAsThumbnailClick.scale
     let crop = onUseAsThumbnailClick.croppr.getValue()
@@ -793,22 +885,18 @@ useAsThumbSaveBtn.addEventListener("click", (e) => {
         .then((thumb) => fetch(thumb))
         .then((response) => response.blob())
         .then(async function(blob) {
-            const formData = new FormData()
-            formData.append("file", blob)
             let options = useAsThumbSelect.selectedOptions
             let promises = []
             for (let embedding of options) {
                 promises.push(
-                    fetch(`bucket/${profileName}/${embedding.dataset["type"]}/${embedding.value}.png`, {
-                        method: "POST",
-                        body: formData,
-                    })
+                    Bucket.upload(`${profileName}/${embedding.dataset["type"]}/${embedding.value}.png`, blob)
                 )
             }
             return Promise.all(promises)
         })
         .then(() => {
             useAsThumbDialog.close()
+            document.dispatchEvent(new CustomEvent("saveThumb", { detail: useAsThumbSelect.selectedOptions }))
         })
         .catch((error) => {
             console.error(error)
@@ -852,6 +940,10 @@ function applyInlineFilter(filterName, path, filterParams, img, statusText, tool
     }
     filterReq.model_paths[filterName] = path
 
+    if (saveToDiskField.checked && diskPathField.value.trim() !== "") {
+        filterReq.save_to_disk_path = diskPathField.value.trim()
+    }
+
     tools.spinnerStatus.innerText = statusText
     tools.spinner.classList.remove("displayNone")
 
@@ -910,7 +1002,20 @@ function onRedoFilter(req, img, e, tools) {
 function onUpscaleClick(req, img, e, tools) {
     let path = upscaleModelField.value
     let scale = parseInt(upscaleAmountField.value)
-    let filterName = path.toLowerCase().includes("realesrgan") ? "realesrgan" : "latent_upscaler"
+
+    let filterName = null
+    const FILTERS = ["realesrgan", "latent_upscaler", "esrgan_4x", "lanczos", "nearest", "scunet", "swinir"]
+    for (let idx in FILTERS) {
+        let f = FILTERS[idx]
+        if (path.toLowerCase().includes(f)) {
+            filterName = f
+            break
+        }
+    }
+
+    if (!filterName) {
+        return
+    }
     let statusText = "Upscaling by " + scale + "x using " + filterName
     applyInlineFilter(filterName, path, { scale: scale }, img, statusText, tools)
 }
@@ -967,6 +1072,9 @@ function makeImage() {
     if (guidanceScaleField.value == "") {
         guidanceScaleField.value = guidanceScaleSlider.value / 10
     }
+    if (distilledGuidanceScaleField.value == "") {
+        distilledGuidanceScaleField.value = distilledGuidanceScaleSlider.value / 10
+    }
     if (hypernetworkStrengthField.value == "") {
         hypernetworkStrengthField.value = hypernetworkStrengthSlider.value / 100
     }
@@ -1237,7 +1345,6 @@ function getCurrentUserRequest() {
             //render_device: undefined, // Set device affinity. Prefer this device, but wont activate.
             use_stable_diffusion_model: stableDiffusionModelField.value,
             clip_skip: clipSkipField.checked,
-            tiling: tilingField.value,
             use_vae_model: vaeModelField.value,
             stream_progress_updates: true,
             stream_image_progress: numOutputsTotal > 50 ? false : streamImageProgressField.checked,
@@ -1291,6 +1398,7 @@ function getCurrentUserRequest() {
         newTask.reqBody.hypernetwork_strength = parseFloat(hypernetworkStrengthField.value)
     }
     if (testDiffusers.checked) {
+        // lora
         let loraModelData = loraModelField.value
         let modelNames = loraModelData["modelNames"]
         let modelStrengths = loraModelData["modelWeights"]
@@ -1302,6 +1410,23 @@ function getCurrentUserRequest() {
             newTask.reqBody.use_lora_model = modelNames
             newTask.reqBody.lora_alpha = modelStrengths
         }
+
+        // text encoder
+        let textEncoderModelNames = textEncoderModelField.modelNames
+
+        if (textEncoderModelNames.length > 0) {
+            textEncoderModelNames = textEncoderModelNames.length == 1 ? textEncoderModelNames[0] : textEncoderModelNames
+
+            newTask.reqBody.use_text_encoder_model = textEncoderModelNames
+        } else {
+            newTask.reqBody.use_text_encoder_model = ""
+        }
+
+        // vae tiling
+        if (tilingField.value !== "none") {
+            newTask.reqBody.tiling = tilingField.value
+        }
+        newTask.reqBody.enable_vae_tiling = enableVAETilingField.checked
     }
     if (testDiffusers.checked && document.getElementById("toggle-tensorrt-install").innerHTML == "Uninstall") {
         // TRT is installed
@@ -1326,10 +1451,17 @@ function getCurrentUserRequest() {
     if (controlnetModelField.value !== "" && IMAGE_REGEX.test(controlImagePreview.src)) {
         newTask.reqBody.use_controlnet_model = controlnetModelField.value
         newTask.reqBody.control_image = controlImagePreview.src
+        newTask.reqBody.control_alpha = parseFloat(controlAlphaField.value)
         if (controlImageFilterField.value !== "") {
             newTask.reqBody.control_filter_to_apply = controlImageFilterField.value
         }
     }
+    if (stableDiffusionModelField.value.toLowerCase().includes("flux")) {
+        newTask.reqBody.distilled_guidance_scale = parseFloat(distilledGuidanceScaleField.value)
+    }
+    if (schedulerSelectionContainer.style.display !== "none") {
+        newTask.reqBody.scheduler_name = schedulerField.value
+    }
 
     return newTask
 }
@@ -1338,7 +1470,7 @@ function setEmbeddings(task) {
     let prompt = task.reqBody.prompt
     let negativePrompt = task.reqBody.negative_prompt
     let overallPrompt = (prompt + " " + negativePrompt).toLowerCase()
-    overallPrompt = overallPrompt.replaceAll(/[^a-z0-9\.]/g, " ") // only allow alpha-numeric and dots
+    overallPrompt = overallPrompt.replaceAll(/[^a-z0-9\-_\.]/g, " ") // only allow alpha-numeric, dots and hyphens
     overallPrompt = overallPrompt.split(" ")
 
     let embeddingsTree = modelsOptions["embeddings"]
@@ -1562,7 +1694,7 @@ function updateInitialText() {
 
         const countBeforeBanner = localStorage.getItem("countBeforeBanner") || 1
         if (countBeforeBanner <= 0) {
-            // supportBanner.classList.remove("displayNone")
+            supportBanner.classList.remove("displayNone")
         }
     }
 }
@@ -1769,36 +1901,162 @@ controlImagePreview.addEventListener("load", onControlnetModelChange)
 controlImagePreview.addEventListener("unload", onControlnetModelChange)
 onControlnetModelChange()
 
-function onControlImageFilterChange() {
-    let filterId = controlImageFilterField.value
-    if (filterId.includes("openpose")) {
-        controlnetModelField.value = "control_v11p_sd15_openpose"
-    } else if (filterId === "canny") {
-        controlnetModelField.value = "control_v11p_sd15_canny"
-    } else if (filterId === "mlsd") {
-        controlnetModelField.value = "control_v11p_sd15_mlsd"
-    } else if (filterId === "mlsd") {
-        controlnetModelField.value = "control_v11p_sd15_mlsd"
-    } else if (filterId.includes("scribble")) {
-        controlnetModelField.value = "control_v11p_sd15_scribble"
-    } else if (filterId.includes("softedge")) {
-        controlnetModelField.value = "control_v11p_sd15_softedge"
-    } else if (filterId === "normal_bae") {
-        controlnetModelField.value = "control_v11p_sd15_normalbae"
-    } else if (filterId.includes("depth")) {
-        controlnetModelField.value = "control_v11f1p_sd15_depth"
-    } else if (filterId === "lineart_anime") {
-        controlnetModelField.value = "control_v11p_sd15s2_lineart_anime"
-    } else if (filterId.includes("lineart")) {
-        controlnetModelField.value = "control_v11p_sd15_lineart"
-    } else if (filterId === "shuffle") {
-        controlnetModelField.value = "control_v11e_sd15_shuffle"
-    } else if (filterId === "segment") {
-        controlnetModelField.value = "control_v11p_sd15_seg"
+document.addEventListener("refreshModels", function() {
+    onFixFaceModelChange()
+    onControlnetModelChange()
+})
+
+// utilities for Flux and Chroma
+let sdModelField = document.querySelector("#stable_diffusion_model")
+
+// function checkAndSetDependentModels() {
+//     let sdModel = sdModelField.value.toLowerCase()
+//     let isFlux = sdModel.includes("flux")
+//     let isChroma = sdModel.includes("chroma")
+
+//     if (isFlux || isChroma) {
+//         vaeModelField.value = "ae"
+
+//         if (isFlux) {
+//             textEncoderModelField.modelNames = ["t5xxl_fp16", "clip_l"]
+//         } else {
+//             textEncoderModelField.modelNames = ["t5xxl_fp16"]
+//         }
+//     } else {
+//         if (vaeModelField.value == "ae") {
+//             vaeModelField.value = ""
+//         }
+//         textEncoderModelField.modelNames = []
+//     }
+// }
+// sdModelField.addEventListener("change", checkAndSetDependentModels)
+
+function checkGuidanceValue() {
+    let guidance = parseFloat(guidanceScaleField.value)
+    let guidanceWarning = document.querySelector("#guidanceWarning")
+    let guidanceWarningText = document.querySelector("#guidanceWarningText")
+    if (sdModelField.value.toLowerCase().includes("flux")) {
+        if (guidance > 1.5) {
+            guidanceWarningText.innerText = "Flux recommends a 'Guidance Scale' of 1"
+            guidanceWarning.classList.remove("displayNone")
+        } else {
+            guidanceWarning.classList.add("displayNone")
+        }
+    } else {
+        if (guidance < 2) {
+            guidanceWarningText.innerText = "A higher 'Guidance Scale' is recommended!"
+            guidanceWarning.classList.remove("displayNone")
+        } else {
+            guidanceWarning.classList.add("displayNone")
+        }
     }
 }
-controlImageFilterField.addEventListener("change", onControlImageFilterChange)
-onControlImageFilterChange()
+sdModelField.addEventListener("change", checkGuidanceValue)
+guidanceScaleField.addEventListener("change", checkGuidanceValue)
+guidanceScaleSlider.addEventListener("change", checkGuidanceValue)
+
+// disabling until we can detect flux models more reliably
+// function checkGuidanceScaleVisibility() {
+//     let guidanceScaleContainer = document.querySelector("#distilled_guidance_scale_container")
+//     if (sdModelField.value.toLowerCase().includes("flux")) {
+//         guidanceScaleContainer.classList.remove("displayNone")
+//     } else {
+//         guidanceScaleContainer.classList.add("displayNone")
+//     }
+// }
+// sdModelField.addEventListener("change", checkGuidanceScaleVisibility)
+
+function checkFluxSampler() {
+    let samplerWarning = document.querySelector("#fluxSamplerWarning")
+    if (sdModelField.value.toLowerCase().includes("flux")) {
+        if (samplerField.value == "euler_a") {
+            samplerWarning.classList.remove("displayNone")
+        } else {
+            samplerWarning.classList.add("displayNone")
+        }
+    } else {
+        samplerWarning.classList.add("displayNone")
+    }
+}
+
+function checkFluxScheduler() {
+    const badSchedulers = ["automatic", "uniform", "turbo", "align_your_steps", "align_your_steps_GITS", "align_your_steps_11", "align_your_steps_32"]
+
+    let schedulerWarning = document.querySelector("#fluxSchedulerWarning")
+    if (sdModelField.value.toLowerCase().includes("flux")) {
+        if (badSchedulers.includes(schedulerField.value)) {
+            schedulerWarning.classList.remove("displayNone")
+        } else {
+            schedulerWarning.classList.add("displayNone")
+        }
+    } else {
+        schedulerWarning.classList.add("displayNone")
+    }
+}
+
+function checkFluxSchedulerSteps() {
+    const problematicSchedulers = ["karras", "exponential", "polyexponential"]
+
+    let schedulerWarning = document.querySelector("#fluxSchedulerStepsWarning")
+    if (sdModelField.value.toLowerCase().includes("flux") && parseInt(numInferenceStepsField.value) < 15) {
+        if (problematicSchedulers.includes(schedulerField.value)) {
+            schedulerWarning.classList.remove("displayNone")
+        } else {
+            schedulerWarning.classList.add("displayNone")
+        }
+    } else {
+        schedulerWarning.classList.add("displayNone")
+    }
+}
+sdModelField.addEventListener("change", checkFluxSampler)
+samplerField.addEventListener("change", checkFluxSampler)
+
+sdModelField.addEventListener("change", checkFluxScheduler)
+schedulerField.addEventListener("change", checkFluxScheduler)
+
+sdModelField.addEventListener("change", checkFluxSchedulerSteps)
+schedulerField.addEventListener("change", checkFluxSchedulerSteps)
+numInferenceStepsField.addEventListener("change", checkFluxSchedulerSteps)
+
+document.addEventListener("refreshModels", function() {
+    // checkAndSetDependentModels()
+    checkGuidanceValue()
+    // checkGuidanceScaleVisibility()
+    checkFluxSampler()
+    checkFluxScheduler()
+    checkFluxSchedulerSteps()
+})
+
+// function onControlImageFilterChange() {
+//     let filterId = controlImageFilterField.value
+//     if (filterId.includes("openpose")) {
+//         controlnetModelField.value = "control_v11p_sd15_openpose"
+//     } else if (filterId === "canny") {
+//         controlnetModelField.value = "control_v11p_sd15_canny"
+//     } else if (filterId === "mlsd") {
+//         controlnetModelField.value = "control_v11p_sd15_mlsd"
+//     } else if (filterId === "mlsd") {
+//         controlnetModelField.value = "control_v11p_sd15_mlsd"
+//     } else if (filterId.includes("scribble")) {
+//         controlnetModelField.value = "control_v11p_sd15_scribble"
+//     } else if (filterId.includes("softedge")) {
+//         controlnetModelField.value = "control_v11p_sd15_softedge"
+//     } else if (filterId === "normal_bae") {
+//         controlnetModelField.value = "control_v11p_sd15_normalbae"
+//     } else if (filterId.includes("depth")) {
+//         controlnetModelField.value = "control_v11f1p_sd15_depth"
+//     } else if (filterId === "lineart_anime") {
+//         controlnetModelField.value = "control_v11p_sd15s2_lineart_anime"
+//     } else if (filterId.includes("lineart")) {
+//         controlnetModelField.value = "control_v11p_sd15_lineart"
+//     } else if (filterId === "shuffle") {
+//         controlnetModelField.value = "control_v11e_sd15_shuffle"
+//     } else if (filterId === "segment") {
+//         controlnetModelField.value = "control_v11p_sd15_seg"
+//     }
+// }
+// controlImageFilterField.addEventListener("change", onControlImageFilterChange)
+// onControlImageFilterChange()
 
 upscaleModelField.disabled = !useUpscalingField.checked
 upscaleAmountField.disabled = !useUpscalingField.checked
@@ -1897,6 +2155,27 @@ guidanceScaleSlider.addEventListener("input", updateGuidanceScale)
 guidanceScaleField.addEventListener("input", updateGuidanceScaleSlider)
 updateGuidanceScale()
 
+/********************* Distilled Guidance **************************/
+function updateDistilledGuidanceScale() {
+    distilledGuidanceScaleField.value = distilledGuidanceScaleSlider.value / 10
+    distilledGuidanceScaleField.dispatchEvent(new Event("change"))
+}
+
+function updateDistilledGuidanceScaleSlider() {
+    if (distilledGuidanceScaleField.value < 0) {
+        distilledGuidanceScaleField.value = 0
+    } else if (distilledGuidanceScaleField.value > 50) {
+        distilledGuidanceScaleField.value = 50
+    }
+
+    distilledGuidanceScaleSlider.value = distilledGuidanceScaleField.value * 10
+    distilledGuidanceScaleSlider.dispatchEvent(new Event("change"))
+}
+
+distilledGuidanceScaleSlider.addEventListener("input", updateDistilledGuidanceScale)
+distilledGuidanceScaleField.addEventListener("input", updateDistilledGuidanceScaleSlider)
+updateDistilledGuidanceScale()
+
 /********************* Prompt Strength *******************/
 function updatePromptStrength() {
     promptStrengthField.value = promptStrengthSlider.value / 100
@@ -1946,6 +2225,27 @@ function updateHypernetworkStrengthContainer() {
 hypernetworkModelField.addEventListener("change", updateHypernetworkStrengthContainer)
 updateHypernetworkStrengthContainer()
 
+/********************* Controlnet Alpha **************************/
+function updateControlAlpha() {
+    controlAlphaField.value = controlAlphaSlider.value / 10
+    controlAlphaField.dispatchEvent(new Event("change"))
+}
+
+function updateControlAlphaSlider() {
+    if (controlAlphaField.value < 0) {
+        controlAlphaField.value = 0
+    } else if (controlAlphaField.value > 10) {
+        controlAlphaField.value = 10
+    }
+
+    controlAlphaSlider.value = controlAlphaField.value * 10
+    controlAlphaSlider.dispatchEvent(new Event("change"))
+}
+
+controlAlphaSlider.addEventListener("input", updateControlAlpha)
+controlAlphaField.addEventListener("input", updateControlAlphaSlider)
+updateControlAlpha()
+
 /********************* JPEG/WEBP Quality **********************/
 function updateOutputQuality() {
     outputQualityField.value = 0 | outputQualitySlider.value
@@ -2355,20 +2655,10 @@ function loadThumbnailImageFromFile() {
 }
 
 function updateEmbeddingsList(filter = "") {
-    function html(model, iconlist = [], prefix = "", filter = "") {
+    function html(model, iconMap = {}, prefix = "", filter = "") {
         filter = filter.toLowerCase()
         let toplevel = document.createElement("div")
         let folders = document.createElement("div")
-        let embIcon = Object.assign(
-            {},
-            ...iconlist.map((x) => ({
-                [x
-                    .toLowerCase()
-                    .split(".")
-                    .slice(0, -1)
-                    .join(".")]: x,
-            }))
-        )
 
         let profileName = profileNameField.value
         model?.forEach((m) => {
@@ -2376,13 +2666,9 @@ function updateEmbeddingsList(filter = "") {
                 let token = m.toLowerCase()
                 if (token.search(filter) != -1) {
                     let button
-                    // if (iconlist.length==0) {
-                    //     button = document.createElement("button")
-                    //     button.innerText = m
-                    // } else {
                     let img = "/media/images/noimg.png"
-                    if (token in embIcon) {
-                        img = `/bucket/${profileName}/embeddings/${embIcon[token]}`
+                    if (token in iconMap) {
+                        img = `/bucket/${profileName}/${iconMap[token]}`
                     }
                     button = createModifierCard(m, [img, img], true)
                     // }
@@ -2391,7 +2677,7 @@ function updateEmbeddingsList(filter = "") {
                     toplevel.appendChild(button)
                 }
             } else {
-                let subdir = html(m[1], iconlist, prefix + m[0] + "/", filter)
+                let subdir = html(m[1], iconMap, prefix + m[0] + "/", filter)
                 if (typeof subdir == "object") {
                     let div1 = document.createElement("div")
                     let div2 = document.createElement("div")
@@ -2450,11 +2736,44 @@ function updateEmbeddingsList(filter = "") {
         </div>
     `
 
+    let loraTokens = []
     let profileName = profileNameField.value
-    fetch(`/bucket/${profileName}/embeddings/`)
-        .then((response) => (response.status == 200 ? response.json() : []))
-        .then(async function(iconlist) {
-            embeddingsList.replaceChildren(html(modelsOptions.embeddings, iconlist, "", filter))
+    let iconMap = {}
+
+    Bucket.getList(`${profileName}/embeddings/`)
+        .then((icons) => {
+            iconMap = Object.assign(
+                {},
+                ...icons.map((x) => ({
+                    [x
+                        .toLowerCase()
+                        .split(".")
+                        .slice(0, -1)
+                        .join(".")]: `embeddings/${x}`,
+                }))
+            )
+
+            return Bucket.getList(`${profileName}/lora/`)
+        })
+        .then(async function (icons) {
+            for (let lora of loraModelField.value.modelNames) {
+                let keywords = await getLoraKeywords(lora)
+                loraTokens = loraTokens.concat(keywords)
+                let loraname = lora.split("/").pop()
+
+                if (icons.includes(`${loraname}.png`)) {
+                    keywords.forEach((kw) => {
+                        iconMap[kw.toLowerCase()] = `lora/${loraname}.png`
+                        
+                    })
+                }
+            }
+
+            let tokenList = [...modelsOptions.embeddings]
+            if (loraTokens.length != 0) {
+                tokenList.unshift(['LORA Keywords', loraTokens])
+            }
+            embeddingsList.replaceChildren(html(tokenList, iconMap, "", filter))
             createCollapsibles(embeddingsList)
             if (filter != "") {
                 embeddingsExpandAll()

ui/media/js/multi-model-selector.js

@@ -10,6 +10,7 @@ class MultiModelSelector {
     root
     modelType
     modelNameFriendly
+    showWeights
     defaultWeight
     weightStep
 
@@ -35,13 +36,13 @@ class MultiModelSelector {
         if (typeof modelData !== "object") {
             throw new Error("Multi-model selector expects an object containing modelNames and modelWeights as keys!")
         }
-        if (!("modelNames" in modelData) || !("modelWeights" in modelData)) {
+        if (!("modelNames" in modelData) || (this.showWeights && !("modelWeights" in modelData))) {
             throw new Error("modelNames or modelWeights not present in the data passed to the multi-model selector")
         }
 
         let newModelNames = modelData["modelNames"]
         let newModelWeights = modelData["modelWeights"]
-        if (newModelNames.length !== newModelWeights.length) {
+        if (newModelWeights && newModelNames.length !== newModelWeights.length) {
             throw new Error("Need to pass an equal number of modelNames and modelWeights!")
         }
 
@@ -50,7 +51,7 @@ class MultiModelSelector {
         // the root of all this unholiness is because searchable-models automatically dispatches an update event
         // as soon as the value is updated via JS, which is against the DOM pattern of not dispatching an event automatically
         // unless the caller explicitly dispatches the event.
-        this.modelWeights = newModelWeights
+        this.modelWeights = newModelWeights || []
         this.modelNames = newModelNames
     }
     get disabled() {
@@ -91,10 +92,11 @@ class MultiModelSelector {
         }
     }
 
-    constructor(root, modelType, modelNameFriendly = undefined, defaultWeight = 0.5, weightStep = 0.02) {
+    constructor(root, modelType, modelNameFriendly = undefined, defaultWeight = 0.5, weightStep = 0.02, showWeights = true) {
         this.root = root
         this.modelType = modelType
         this.modelNameFriendly = modelNameFriendly || modelType
+        this.showWeights = showWeights
         this.defaultWeight = defaultWeight
         this.weightStep = weightStep
 
@@ -135,10 +137,13 @@ class MultiModelSelector {
 
         const modelElement = document.createElement("div")
         modelElement.className = "model_entry"
-        modelElement.innerHTML = `
-            <input id="${this.modelType}_${idx}" class="model_name model-filter" type="text" spellcheck="false" autocomplete="off" data-path="" />
-            <input class="model_weight" type="number" step="${this.weightStep}" value="${this.defaultWeight}" pattern="^-?[0-9]*\.?[0-9]*$" onkeypress="preventNonNumericalInput(event)">
-        `
+        let html = `<input id="${this.modelType}_${idx}" class="model_name model-filter" type="text" spellcheck="false" autocomplete="off" data-path="" />`
+
+        if (this.showWeights) {
+            html += `<input class="model_weight" type="number" step="${this.weightStep}" value="${this.defaultWeight}" pattern="^-?[0-9]*\.?[0-9]*$" onkeypress="preventNonNumericalInput(event)">`
+        }
+        modelElement.innerHTML = html
+
         this.modelContainer.appendChild(modelElement)
 
         let modelNameEl = modelElement.querySelector(".model_name")
@@ -160,8 +165,8 @@ class MultiModelSelector {
 
         modelNameEl.addEventListener("change", makeUpdateEvent("change"))
         modelNameEl.addEventListener("input", makeUpdateEvent("input"))
-        modelWeightEl.addEventListener("change", makeUpdateEvent("change"))
-        modelWeightEl.addEventListener("input", makeUpdateEvent("input"))
+        modelWeightEl?.addEventListener("change", makeUpdateEvent("change"))
+        modelWeightEl?.addEventListener("input", makeUpdateEvent("input"))
 
         let removeBtn = document.createElement("button")
         removeBtn.className = "remove_model_btn"
@@ -218,10 +223,14 @@ class MultiModelSelector {
     }
 
     get modelWeights() {
-        return this.getModelElements(true).map((e) => e.weight.value)
+        return this.getModelElements(true).map((e) => e.weight?.value)
     }
 
     set modelWeights(newModelWeights) {
+        if (!this.showWeights) {
+            return
+        }
+
         this.resizeEntryList(newModelWeights.length)
 
         if (newModelWeights.length === 0) {

ui/media/js/parameters.js

@@ -97,6 +97,17 @@ var PARAMETERS = [
             },
         ],
     },
+    {
+        id: "models_dir",
+        type: ParameterType.custom,
+        icon: "fa-folder-tree",
+        label: "Models Folder",
+        note: "Path to the 'models' folder. Please save and restart Easy Diffusion after changing this.",
+        saveInAppConfig: true,
+        render: (parameter) => {
+            return `<input id="${parameter.id}" name="${parameter.id}" size="30">`
+        },
+    },
     {
         id: "block_nsfw",
         type: ParameterType.checkbox,
@@ -150,6 +161,7 @@ var PARAMETERS = [
             "<b>Low:</b> slowest, recommended for GPUs with 3 to 4 GB memory",
         icon: "fa-forward",
         default: "balanced",
+        saveInAppConfig: true,
         options: [
             { value: "balanced", label: "Balanced" },
             { value: "high", label: "High" },
@@ -238,14 +250,19 @@ var PARAMETERS = [
         default: false,
     },
     {
-        id: "test_diffusers",
-        type: ParameterType.checkbox,
-        label: "Use the new v3 engine (diffusers)",
+        id: "backend",
+        type: ParameterType.select,
+        label: "Engine to use",
         note:
-            "Use our new v3 engine, with additional features like LoRA, ControlNet, SDXL, Embeddings, Tiling and lots more! Please press Save, then restart the program after changing this.",
-        icon: "fa-bolt",
-        default: true,
+            "Use our new v3.5 engine (Forge), with additional features like Flux, SD3, Lycoris and lots more! Please press Save, then restart the program after changing this.",
+        icon: "fa-robot",
         saveInAppConfig: true,
+        default: "ed_diffusers",
+        options: [
+            { value: "webui", label: "v3.5 (latest)" },
+            { value: "ed_diffusers", label: "v3.0" },
+            { value: "ed_classic", label: "v2.0" },
+        ],
     },
     {
         id: "cloudflare",
@@ -420,8 +437,10 @@ let listenPortField = document.querySelector("#listen_port")
 let useBetaChannelField = document.querySelector("#use_beta_channel")
 let uiOpenBrowserOnStartField = document.querySelector("#ui_open_browser_on_start")
 let confirmDangerousActionsField = document.querySelector("#confirm_dangerous_actions")
-let testDiffusers = document.querySelector("#test_diffusers")
+let testDiffusers = document.querySelector("#use_v3_engine")
+let backendEngine = document.querySelector("#backend")
 let profileNameField = document.querySelector("#profileName")
+let modelsDirField = document.querySelector("#models_dir")
 
 let saveSettingsBtn = document.querySelector("#save-system-settings-btn")
 
@@ -442,6 +461,23 @@ async function changeAppConfig(configDelta) {
     }
 }
 
+function getDefaultDisplay(element) {
+    const tag = element.tagName.toLowerCase();
+    const defaultDisplays = {
+        div: 'block',
+        span: 'inline',
+        p: 'block',
+        tr: 'table-row',
+        table: 'table',
+        li: 'list-item',
+        ul: 'block',
+        ol: 'block',
+        button: 'inline',
+        // Add more if needed
+    };
+    return defaultDisplays[tag] || 'block'; // Default to 'block' if not listed
+}
+
 async function getAppConfig() {
     try {
         let res = await fetch("/get/app_config")
@@ -463,15 +499,19 @@ async function getAppConfig() {
         if (config.net && config.net.listen_port !== undefined) {
             listenPortField.value = config.net.listen_port
         }
+        modelsDirField.value = config.models_dir
 
         let testDiffusersEnabled = true
-        if (config.test_diffusers === false) {
+        if (config.backend === "ed_classic") {
             testDiffusersEnabled = false
         }
         testDiffusers.checked = testDiffusersEnabled
+        backendEngine.value = config.backend
+        document.querySelector("#test_diffusers").checked = testDiffusers.checked // don't break plugins
+        document.querySelector("#use_v3_engine").checked = testDiffusers.checked // don't break plugins
 
         if (config.config_on_startup) {
-            if (config.config_on_startup?.test_diffusers) {
+            if (config.config_on_startup?.backend !== "ed_classic") {
                 document.body.classList.add("diffusers-enabled-on-startup")
                 document.body.classList.remove("diffusers-disabled-on-startup")
             } else {
@@ -480,35 +520,29 @@ async function getAppConfig() {
             }
         }
 
-        if (!testDiffusersEnabled) {
-            document.querySelector("#lora_model_container").style.display = "none"
-            document.querySelector("#tiling_container").style.display = "none"
-            document.querySelector("#controlnet_model_container").style.display = "none"
-            document.querySelector("#hypernetwork_model_container").style.display = ""
-            document.querySelector("#hypernetwork_strength_container").style.display = ""
-            document.querySelector("#negative-embeddings-button").style.display = "none"
-
-            document.querySelectorAll("#sampler_name option.diffusers-only").forEach((option) => {
-                option.style.display = "none"
-            })
+        if (config.backend === "ed_classic") {
             IMAGE_STEP_SIZE = 64
-            customWidthField.step = IMAGE_STEP_SIZE
-            customHeightField.step = IMAGE_STEP_SIZE
         } else {
-            document.querySelector("#lora_model_container").style.display = ""
-            document.querySelector("#tiling_container").style.display = ""
-            document.querySelector("#controlnet_model_container").style.display = ""
-            document.querySelector("#hypernetwork_model_container").style.display = "none"
-            document.querySelector("#hypernetwork_strength_container").style.display = "none"
-
-            document.querySelectorAll("#sampler_name option.k_diffusion-only").forEach((option) => {
-                option.style.display = "none"
-            })
-            document.querySelector("#clip_skip_config").classList.remove("displayNone")
-            document.querySelector("#embeddings-button").classList.remove("displayNone")
             IMAGE_STEP_SIZE = 8
-            customWidthField.step = IMAGE_STEP_SIZE
-            customHeightField.step = IMAGE_STEP_SIZE
+        }
+
+        customWidthField.step = IMAGE_STEP_SIZE
+        customHeightField.step = IMAGE_STEP_SIZE
+
+        const currentBackendKey = "backend_" + config.backend
+
+        document.querySelectorAll('.gated-feature').forEach((element) => {
+            const featureKeys = element.getAttribute('data-feature-keys').split(' ')
+
+            if (featureKeys.includes(currentBackendKey)) {
+                element.style.display = getDefaultDisplay(element)
+            } else {
+                element.style.display = 'none'
+            }
+        });
+
+        if (config.force_save_metadata) {
+            metadataOutputFormatField.value = config.force_save_metadata
         }
 
         console.log("get config status response", config)
@@ -624,7 +658,7 @@ function setDeviceInfo(devices) {
 
     function ID_TO_TEXT(d) {
         let info = devices.all[d]
-        if ("mem_free" in info && "mem_total" in info) {
+        if ("mem_free" in info && "mem_total" in info && info["mem_total"] > 0) {
             return `${info.name} <small>(${d}) (${info.mem_free.toFixed(1)}Gb free / ${info.mem_total.toFixed(
                 1
             )} Gb total)</small>`
@@ -722,12 +756,20 @@ async function getSystemInfo() {
             force = res["enforce_output_dir"]
             if (force == true) {
                 saveToDiskField.checked = true
-                metadataOutputFormatField.disabled = false
+                metadataOutputFormatField.disabled = res["enforce_output_metadata"]
+                diskPathField.disabled = true
             }
             saveToDiskField.disabled = force
-            diskPathField.disabled = force
+        } else {
+            diskPathField.disabled = !saveToDiskField.checked
+            metadataOutputFormatField.disabled = !saveToDiskField.checked
         }
         setDiskPath(res["default_output_dir"], force)
+
+        // backend info
+        if (res["backend_url"]) {
+            document.querySelector("#backend-url").setAttribute("href", res["backend_url"])
+        }
     } catch (e) {
         console.log("error fetching devices", e)
     }

ui/plugins/ui/merge.plugin.js

@@ -1,454 +0,0 @@
-;(function() {
-    "use strict"
-
-    ///////////////////// Function section
-    function smoothstep(x) {
-        return x * x * (3 - 2 * x)
-    }
-
-    function smootherstep(x) {
-        return x * x * x * (x * (x * 6 - 15) + 10)
-    }
-
-    function smootheststep(x) {
-        let y = -20 * Math.pow(x, 7)
-        y += 70 * Math.pow(x, 6)
-        y -= 84 * Math.pow(x, 5)
-        y += 35 * Math.pow(x, 4)
-        return y
-    }
-    function getCurrentTime() {
-        const now = new Date()
-        let hours = now.getHours()
-        let minutes = now.getMinutes()
-        let seconds = now.getSeconds()
-
-        hours = hours < 10 ? `0${hours}` : hours
-        minutes = minutes < 10 ? `0${minutes}` : minutes
-        seconds = seconds < 10 ? `0${seconds}` : seconds
-
-        return `${hours}:${minutes}:${seconds}`
-    }
-
-    function addLogMessage(message) {
-        const logContainer = document.getElementById("merge-log")
-        logContainer.innerHTML += `<i>${getCurrentTime()}</i> ${message}<br>`
-
-        // Scroll to the bottom of the log
-        logContainer.scrollTop = logContainer.scrollHeight
-
-        document.querySelector("#merge-log-container").style.display = "block"
-    }
-
-    function addLogSeparator() {
-        const logContainer = document.getElementById("merge-log")
-        logContainer.innerHTML += "<hr>"
-
-        logContainer.scrollTop = logContainer.scrollHeight
-    }
-
-    function drawDiagram(fn) {
-        const SIZE = 300
-        const canvas = document.getElementById("merge-canvas")
-        canvas.height = canvas.width = SIZE
-        const ctx = canvas.getContext("2d")
-
-        // Draw coordinate system
-        ctx.scale(1, -1)
-        ctx.translate(0, -canvas.height)
-        ctx.lineWidth = 1
-        ctx.beginPath()
-
-        ctx.strokeStyle = "white"
-        ctx.moveTo(0, 0)
-        ctx.lineTo(0, SIZE)
-        ctx.lineTo(SIZE, SIZE)
-        ctx.lineTo(SIZE, 0)
-        ctx.lineTo(0, 0)
-        ctx.lineTo(SIZE, SIZE)
-        ctx.stroke()
-        ctx.beginPath()
-        ctx.setLineDash([1, 2])
-        const n = SIZE / 10
-        for (let i = n; i < SIZE; i += n) {
-            ctx.moveTo(0, i)
-            ctx.lineTo(SIZE, i)
-            ctx.moveTo(i, 0)
-            ctx.lineTo(i, SIZE)
-        }
-        ctx.stroke()
-        ctx.beginPath()
-        ctx.setLineDash([])
-        ctx.beginPath()
-        ctx.strokeStyle = "black"
-        ctx.lineWidth = 3
-        // Plot function
-        const numSamples = 20
-        for (let i = 0; i <= numSamples; i++) {
-            const x = i / numSamples
-            const y = fn(x)
-
-            const canvasX = x * SIZE
-            const canvasY = y * SIZE
-
-            if (i === 0) {
-                ctx.moveTo(canvasX, canvasY)
-            } else {
-                ctx.lineTo(canvasX, canvasY)
-            }
-        }
-        ctx.stroke()
-        // Plot alpha values (yellow boxes)
-        let start = parseFloat(document.querySelector("#merge-start").value)
-        let step = parseFloat(document.querySelector("#merge-step").value)
-        let iterations = document.querySelector("#merge-count").value >> 0
-        ctx.beginPath()
-        ctx.fillStyle = "yellow"
-        for (let i = 0; i < iterations; i++) {
-            const alpha = (start + i * step) / 100
-            const x = alpha * SIZE
-            const y = fn(alpha) * SIZE
-            if (x <= SIZE) {
-                ctx.rect(x - 3, y - 3, 6, 6)
-                ctx.fill()
-            } else {
-                ctx.strokeStyle = "red"
-                ctx.moveTo(0, 0)
-                ctx.lineTo(0, SIZE)
-                ctx.lineTo(SIZE, SIZE)
-                ctx.lineTo(SIZE, 0)
-                ctx.lineTo(0, 0)
-                ctx.lineTo(SIZE, SIZE)
-                ctx.stroke()
-                addLogMessage("<i>Warning: maximum ratio is &#8805; 100%</i>")
-            }
-        }
-    }
-
-    function updateChart() {
-        let fn = (x) => x
-        switch (document.querySelector("#merge-interpolation").value) {
-            case "SmoothStep":
-                fn = smoothstep
-                break
-            case "SmootherStep":
-                fn = smootherstep
-                break
-            case "SmoothestStep":
-                fn = smootheststep
-                break
-        }
-        drawDiagram(fn)
-    }
-    createTab({
-        id: "merge",
-        icon: "fa-code-merge",
-        label: "Merge models",
-        css: `
-        #tab-content-merge .tab-content-inner {
-            max-width: 100%;
-            padding: 10pt;
-        }
-        .merge-container {  
-            margin-left: 15%;
-            margin-right: 15%;
-            text-align: left;
-            display: inline-grid;
-            grid-template-columns: 1fr 1fr;
-            grid-template-rows: auto auto auto;
-            gap: 0px 0px;
-            grid-auto-flow: row;
-            grid-template-areas:
-              "merge-input merge-config"
-              "merge-buttons merge-buttons";
-        }
-        .merge-container p {
-            margin-top: 3pt;
-            margin-bottom: 3pt;
-        }
-        .merge-config .tab-content {
-            background: var(--background-color1);
-            border-radius: 3pt;
-        }
-        .merge-config .tab-content-inner {
-            text-align: left;
-        }
-
-        .merge-input { 
-            grid-area: merge-input; 
-            padding-left:1em;
-        }
-        .merge-config { 
-            grid-area: merge-config; 
-            padding:1em;
-        }
-        .merge-config input { 
-            margin-bottom: 3px; 
-        }
-        .merge-config select { 
-            margin-bottom: 3px; 
-        }
-        .merge-buttons { 
-            grid-area: merge-buttons; 
-            padding:1em; 
-            text-align: center;
-        }
-        #merge-button  { 
-            padding: 8px; 
-            width:20em; 
-        }
-        div#merge-log {
-            height:150px; 
-            overflow-x:hidden;
-            overflow-y:scroll;
-            background:var(--background-color1);
-            border-radius: 3pt;
-        }
-        div#merge-log i {
-            color: hsl(var(--accent-hue), 100%, calc(2*var(--accent-lightness))); 
-            font-family: monospace;
-        }
-        .disabled { 
-            background: var(--background-color4); 
-            color: var(--text-color); 
-        }
-        #merge-type-tabs {
-            border-bottom: 1px solid black;
-        }
-        #merge-log-container {
-            display: none;
-        }
-        .merge-container #merge-warning {
-            color: rgb(153, 153, 153);
-        }`,
-        content: `
-        <div class="merge-container panel-box">
-          <div class="merge-input">
-             <p><label for="#mergeModelA">Select Model A:</label></p>
-             <input id="mergeModelA" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" />
-             <p><label for="#mergeModelB">Select Model B:</label></p>
-             <input id="mergeModelB" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" />
-             <br/><br/>
-             <p id="merge-warning"><small><b>Important:</b> Please merge models of similar type.<br/>For e.g. <code>SD 1.4</code> models with only <code>SD 1.4/1.5</code> models,<br/><code>SD 2.0</code> with <code>SD 2.0</code>-type, and <code>SD 2.1</code> with <code>SD 2.1</code>-type models.</small></p>
-             <br/>
-             <table>
-                <tr>
-                    <td><label for="#merge-filename">Output file name:</label></td>
-                    <td><input id="merge-filename" size=24> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Base name of the output file.<br>Mix ratio and file suffix will be appended to this.</span></i></td>
-                </tr>
-                <tr>
-                    <td><label for="#merge-fp">Output precision:</label></td>
-                    <td><select id="merge-fp">
-                        <option value="fp16">fp16 (smaller file size)</option>
-                        <option value="fp32">fp32 (larger file size)</option>
-                    </select>
-                    <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Image generation uses fp16, so it's a good choice.<br>Use fp32 if you want to use the result models for more mixes</span></i>
-                    </td>
-                </tr>
-                <tr>
-                    <td><label for="#merge-format">Output file format:</label></td>
-                    <td><select id="merge-format">
-                        <option value="safetensors">Safetensors (recommended)</option>
-                        <option value="ckpt">CKPT/Pickle (legacy format)</option>
-                    </select>
-                    </td>
-                </tr>
-             </table>
-             <br/>
-             <div id="merge-log-container">
-                <p><label for="#merge-log">Log messages:</label></p>
-                <div id="merge-log"></div>
-             </div>
-          </div>
-          <div class="merge-config">
-             <div class="tab-container">
-                <span id="tab-merge-opts-single" class="tab active">
-                    <span>Make a single file</small></span>
-                </span>
-                <span id="tab-merge-opts-batch" class="tab">
-                    <span>Make multiple variations</small></span>
-                </span>
-             </div>
-             <div>
-                <div id="tab-content-merge-opts-single" class="tab-content active">
-                    <div class="tab-content-inner">
-                        <small>Saves a single merged model file, at the specified merge ratio.</small><br/><br/>
-                        <label for="#single-merge-ratio-slider">Merge ratio:</label>
-                        <input id="single-merge-ratio-slider" name="single-merge-ratio-slider" class="editor-slider" value="50" type="range" min="1" max="1000">
-                        <input id="single-merge-ratio" size=2 value="5">%
-                        <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Model A's contribution to the mix. The rest will be from Model B.</span></i>
-                    </div>
-                </div>
-                <div id="tab-content-merge-opts-batch" class="tab-content">
-                    <div class="tab-content-inner">
-                        <small>Saves multiple variations of the model, at different merge ratios.<br/>Each variation will be saved as a separate file.</small><br/><br/>
-                        <table>
-                            <tr><td><label for="#merge-count">Number of variations:</label></td>
-                                <td> <input id="merge-count" size=2 value="5"></td>
-                                <td> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Number of models to create</span></i></td></tr>
-                            <tr><td><label for="#merge-start">Starting merge ratio:</label></td>
-                                <td> <input id="merge-start" size=2 value="5">%</td>
-                                <td> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Smallest share of model A in the mix</span></i></td></tr>
-                            <tr><td><label for="#merge-step">Increment each step:</label></td>
-                                <td> <input id="merge-step" size=2 value="10">%</td>
-                                <td> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Share of model A added into the mix per step</span></i></td></tr>
-                            <tr><td><label for="#merge-interpolation">Interpolation model:</label></td>
-                                <td> <select id="merge-interpolation">
-                                    <option>Exact</option>
-                                    <option>SmoothStep</option>
-                                    <option>SmootherStep</option>
-                                    <option>SmoothestStep</option>
-                                </select></td>
-                                <td> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Sigmoid function to be applied to the model share before mixing</span></i></td></tr>
-                        </table>
-                        <br/>
-                        <small>Preview of variation ratios:</small><br/>
-                        <canvas id="merge-canvas" width="400" height="400"></canvas>
-                    </div>
-                </div>
-             </div>
-          </div>
-          <div class="merge-buttons">
-             <button id="merge-button" class="primaryButton">Merge models</button>
-          </div>
-        </div>`,
-        onOpen: ({ firstOpen }) => {
-            if (!firstOpen) {
-                return
-            }
-
-            const tabSettingsSingle = document.querySelector("#tab-merge-opts-single")
-            const tabSettingsBatch = document.querySelector("#tab-merge-opts-batch")
-            linkTabContents(tabSettingsSingle)
-            linkTabContents(tabSettingsBatch)
-
-            console.log("Activate")
-            let mergeModelAField = new ModelDropdown(document.querySelector("#mergeModelA"), "stable-diffusion")
-            let mergeModelBField = new ModelDropdown(document.querySelector("#mergeModelB"), "stable-diffusion")
-            updateChart()
-
-            // slider
-            const singleMergeRatioField = document.querySelector("#single-merge-ratio")
-            const singleMergeRatioSlider = document.querySelector("#single-merge-ratio-slider")
-
-            function updateSingleMergeRatio() {
-                singleMergeRatioField.value = singleMergeRatioSlider.value / 10
-                singleMergeRatioField.dispatchEvent(new Event("change"))
-            }
-
-            function updateSingleMergeRatioSlider() {
-                if (singleMergeRatioField.value < 0) {
-                    singleMergeRatioField.value = 0
-                } else if (singleMergeRatioField.value > 100) {
-                    singleMergeRatioField.value = 100
-                }
-
-                singleMergeRatioSlider.value = singleMergeRatioField.value * 10
-                singleMergeRatioSlider.dispatchEvent(new Event("change"))
-            }
-
-            singleMergeRatioSlider.addEventListener("input", updateSingleMergeRatio)
-            singleMergeRatioField.addEventListener("input", updateSingleMergeRatioSlider)
-            updateSingleMergeRatio()
-
-            document.querySelector(".merge-config").addEventListener("change", updateChart)
-
-            document.querySelector("#merge-button").addEventListener("click", async function(e) {
-                // Build request template
-                let model0 = mergeModelAField.value
-                let model1 = mergeModelBField.value
-                let request = { model0: model0, model1: model1 }
-                request["use_fp16"] = document.querySelector("#merge-fp").value == "fp16"
-                let iterations = document.querySelector("#merge-count").value >> 0
-                let start = parseFloat(document.querySelector("#merge-start").value)
-                let step = parseFloat(document.querySelector("#merge-step").value)
-
-                if (isTabActive(tabSettingsSingle)) {
-                    start = parseFloat(singleMergeRatioField.value)
-                    step = 0
-                    iterations = 1
-                    addLogMessage(`merge ratio = ${start}%`)
-                } else {
-                    addLogMessage(`start = ${start}%`)
-                    addLogMessage(`step  = ${step}%`)
-                }
-
-                if (start + (iterations - 1) * step >= 100) {
-                    addLogMessage("<i>Aborting: maximum ratio is &#8805; 100%</i>")
-                    addLogMessage("Reduce the number of variations or the step size")
-                    addLogSeparator()
-                    document.querySelector("#merge-count").focus()
-                    return
-                }
-
-                if (document.querySelector("#merge-filename").value == "") {
-                    addLogMessage("<i>Aborting: No output file name specified</i>")
-                    addLogSeparator()
-                    document.querySelector("#merge-filename").focus()
-                    return
-                }
-
-                // Disable merge button
-                e.target.disabled = true
-                e.target.classList.add("disabled")
-                let cursor = $("body").css("cursor")
-                let label = document.querySelector("#merge-button").innerHTML
-                $("body").css("cursor", "progress")
-                document.querySelector("#merge-button").innerHTML = "Merging models ..."
-
-                addLogMessage("Merging models")
-                addLogMessage("Model A: " + model0)
-                addLogMessage("Model B: " + model1)
-
-                // Batch main loop
-                for (let i = 0; i < iterations; i++) {
-                    let alpha = (start + i * step) / 100
-
-                    if (isTabActive(tabSettingsBatch)) {
-                        switch (document.querySelector("#merge-interpolation").value) {
-                            case "SmoothStep":
-                                alpha = smoothstep(alpha)
-                                break
-                            case "SmootherStep":
-                                alpha = smootherstep(alpha)
-                                break
-                            case "SmoothestStep":
-                                alpha = smootheststep(alpha)
-                                break
-                        }
-                    }
-                    addLogMessage(`merging batch job ${i + 1}/${iterations}, alpha = ${alpha.toFixed(5)}...`)
-
-                    request["out_path"] = document.querySelector("#merge-filename").value
-                    request["out_path"] += "-" + alpha.toFixed(5) + "." + document.querySelector("#merge-format").value
-                    addLogMessage(`&nbsp;&nbsp;filename: ${request["out_path"]}`)
-
-                    // sdkit documentation: "ratio - the ratio of the second model. 1 means only the second model will be used."
-                    request["ratio"] = 1-alpha 
-                    let res = await fetch("/model/merge", {
-                        method: "POST",
-                        headers: { "Content-Type": "application/json" },
-                        body: JSON.stringify(request),
-                    })
-                    const data = await res.json()
-                    addLogMessage(JSON.stringify(data))
-                }
-                addLogMessage(
-                    "<b>Done.</b> The models have been saved to your <tt>models/stable-diffusion</tt> folder."
-                )
-                addLogSeparator()
-                // Re-enable merge button
-                $("body").css("cursor", cursor)
-                document.querySelector("#merge-button").innerHTML = label
-                e.target.disabled = false
-                e.target.classList.remove("disabled")
-
-                // Update model list
-                stableDiffusionModelField.innerHTML = ""
-                vaeModelField.innerHTML = ""
-                hypernetworkModelField.innerHTML = ""
-                await getModels()
-            })
-        },
-    })
-})()

ui/plugins/ui/model-tools.plugin.js

@@ -0,0 +1,770 @@
+;(function() {
+    "use strict"
+
+    let mergeCSS = `
+        /*********** Main tab ***********/
+        .tab-centered {
+            justify-content: center;
+        }
+
+        #model-tool-tab-content {
+            background-color: var(--background-color3);
+        }
+
+        #model-tool-tab-content .tab-content-inner {
+            text-align: initial;
+        }
+
+        #model-tool-tab-bar .tab {
+            margin-bottom: 0px;
+            border-top-left-radius: var(--input-border-radius);
+            background-color: var(--background-color3);
+            padding: 6px 6px 0.8em 6px;
+        }
+        #tab-content-merge .tab-content-inner {
+            max-width: 100%;
+            padding: 10pt;
+        }
+
+        /*********** Merge UI ***********/
+        .merge-model-container {  
+            margin-left: 15%;
+            margin-right: 15%;
+            text-align: left;
+            display: inline-grid;
+            grid-template-columns: 1fr 1fr;
+            grid-template-rows: auto auto auto;
+            gap: 0px 0px;
+            grid-auto-flow: row;
+            grid-template-areas:
+              "merge-input merge-config"
+              "merge-buttons merge-buttons";
+        }
+        .merge-model-container p {
+            margin-top: 3pt;
+            margin-bottom: 3pt;
+        }
+        .merge-config .tab-content {
+            background: var(--background-color1);
+            border-radius: 3pt;
+        }
+        .merge-config .tab-content-inner {
+            text-align: left;
+        }
+
+        .merge-input { 
+            grid-area: merge-input; 
+            padding-left:1em;
+        }
+        .merge-config { 
+            grid-area: merge-config; 
+            padding:1em;
+        }
+        .merge-config input { 
+            margin-bottom: 3px; 
+        }
+        .merge-config select { 
+            margin-bottom: 3px; 
+        }
+        .merge-buttons { 
+            grid-area: merge-buttons; 
+            padding:1em; 
+            text-align: center;
+        }
+        #merge-button  { 
+            padding: 8px; 
+            width:20em; 
+        }
+        div#merge-log {
+            height:150px; 
+            overflow-x:hidden;
+            overflow-y:scroll;
+            background:var(--background-color1);
+            border-radius: 3pt;
+        }
+        div#merge-log i {
+            color: hsl(var(--accent-hue), 100%, calc(2*var(--accent-lightness))); 
+            font-family: monospace;
+        }
+        .disabled { 
+            background: var(--background-color4); 
+            color: var(--text-color); 
+        }
+        #merge-type-tabs {
+            border-bottom: 1px solid black;
+        }
+        #merge-log-container {
+            display: none;
+        }
+        .merge-model-container #merge-warning {
+            color: var(--small-label-color);
+        }
+
+        /*********** LORA UI ***********/
+        .lora-manager-grid {  
+            display: grid;
+            gap: 0px 8px;
+            grid-auto-flow: row;
+        }
+
+        @media screen and (min-width: 1501px) {
+            .lora-manager-grid textarea {
+                height:350px;
+            }
+
+            .lora-manager-grid {  
+                grid-template-columns: auto 1fr 1fr;
+                grid-template-rows: auto 1fr;
+                grid-template-areas:
+                    "selector  selector selector"
+                    "thumbnail keywords notes";
+            }
+        }
+
+        @media screen and (min-width: 1001px) and (max-width: 1500px) {
+            .lora-manager-grid textarea {
+                height:250px;
+            }
+
+            .lora-manager-grid {  
+                grid-template-columns: auto auto;
+                grid-template-rows: auto auto auto;
+                grid-template-areas:
+                    "selector  selector"
+                    "thumbnail keywords"
+                    "thumbnail notes";
+            }
+
+        }
+
+        @media screen and (max-width: 1000px) {
+            .lora-manager-grid textarea {
+                height:200px;
+            }
+
+            .lora-manager-grid {  
+                grid-template-columns: auto;
+                grid-template-rows: auto auto auto auto;
+                grid-template-areas:
+                    "selector"
+                    "keywords"
+                    "thumbnail"
+                    "notes";
+            }
+
+        }
+
+        .lora-manager-grid-selector { 
+            grid-area: selector; 
+            justify-self: start;
+        }
+
+        .lora-manager-grid-thumbnail { 
+            grid-area: thumbnail; 
+            justify-self: center;
+        }
+
+        .lora-manager-grid-keywords { 
+            grid-area: keywords; 
+        }
+
+        .lora-manager-grid-notes { 
+            grid-area: notes; 
+        }
+        
+        .lora-manager-grid p {
+            margin-bottom: 2px;
+        }
+
+
+        `
+
+    let mergeUI = `
+        <div class="merge-model-container panel-box">
+          <div class="merge-input">
+             <p><label for="#mergeModelA">Select Model A:</label></p>
+             <input id="mergeModelA" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" />
+             <p><label for="#mergeModelB">Select Model B:</label></p>
+             <input id="mergeModelB" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" />
+             <br/><br/>
+             <p id="merge-warning"><small><b>Important:</b> Please merge models of similar type.<br/>For e.g. <code>SD 1.4</code> models with only <code>SD 1.4/1.5</code> models,<br/><code>SD 2.0</code> with <code>SD 2.0</code>-type, and <code>SD 2.1</code> with <code>SD 2.1</code>-type models.</small></p>
+             <br/>
+             <table>
+                <tr>
+                    <td><label for="#merge-filename">Output file name:</label></td>
+                    <td><input id="merge-filename" size=24> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Base name of the output file.<br>Mix ratio and file suffix will be appended to this.</span></i></td>
+                </tr>
+                <tr>
+                    <td><label for="#merge-fp">Output precision:</label></td>
+                    <td><select id="merge-fp">
+                        <option value="fp16">fp16 (smaller file size)</option>
+                        <option value="fp32">fp32 (larger file size)</option>
+                    </select>
+                    <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Image generation uses fp16, so it's a good choice.<br>Use fp32 if you want to use the result models for more mixes</span></i>
+                    </td>
+                </tr>
+                <tr>
+                    <td><label for="#merge-format">Output file format:</label></td>
+                    <td><select id="merge-format">
+                        <option value="safetensors">Safetensors (recommended)</option>
+                        <option value="ckpt">CKPT/Pickle (legacy format)</option>
+                    </select>
+                    </td>
+                </tr>
+             </table>
+             <br/>
+             <div id="merge-log-container">
+                <p><label for="#merge-log">Log messages:</label></p>
+                <div id="merge-log"></div>
+             </div>
+          </div>
+          <div class="merge-config">
+             <div class="tab-container">
+                <span id="tab-merge-opts-single" class="tab active">
+                    <span>Make a single file</small></span>
+                </span>
+                <span id="tab-merge-opts-batch" class="tab">
+                    <span>Make multiple variations</small></span>
+                </span>
+             </div>
+             <div>
+                <div id="tab-content-merge-opts-single" class="tab-content active">
+                    <div class="tab-content-inner">
+                        <small>Saves a single merged model file, at the specified merge ratio.</small><br/><br/>
+                        <label for="#single-merge-ratio-slider">Merge ratio:</label>
+                        <input id="single-merge-ratio-slider" name="single-merge-ratio-slider" class="editor-slider" value="50" type="range" min="1" max="1000">
+                        <input id="single-merge-ratio" size=2 value="5">%
+                        <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Model A's contribution to the mix. The rest will be from Model B.</span></i>
+                    </div>
+                </div>
+                <div id="tab-content-merge-opts-batch" class="tab-content">
+                    <div class="tab-content-inner">
+                        <small>Saves multiple variations of the model, at different merge ratios.<br/>Each variation will be saved as a separate file.</small><br/><br/>
+                        <table>
+                            <tr><td><label for="#merge-count">Number of variations:</label></td>
+                                <td> <input id="merge-count" size=2 value="5"></td>
+                                <td> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Number of models to create</span></i></td></tr>
+                            <tr><td><label for="#merge-start">Starting merge ratio:</label></td>
+                                <td> <input id="merge-start" size=2 value="5">%</td>
+                                <td> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Smallest share of model A in the mix</span></i></td></tr>
+                            <tr><td><label for="#merge-step">Increment each step:</label></td>
+                                <td> <input id="merge-step" size=2 value="10">%</td>
+                                <td> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Share of model A added into the mix per step</span></i></td></tr>
+                            <tr><td><label for="#merge-interpolation">Interpolation model:</label></td>
+                                <td> <select id="merge-interpolation">
+                                    <option>Exact</option>
+                                    <option>SmoothStep</option>
+                                    <option>SmootherStep</option>
+                                    <option>SmoothestStep</option>
+                                </select></td>
+                                <td> <i class="fa-solid fa-circle-question help-btn"><span class="simple-tooltip top-left">Sigmoid function to be applied to the model share before mixing</span></i></td></tr>
+                        </table>
+                        <br/>
+                        <small>Preview of variation ratios:</small><br/>
+                        <canvas id="merge-canvas" width="400" height="400"></canvas>
+                    </div>
+                </div>
+             </div>
+          </div>
+          <div class="merge-buttons">
+             <button id="merge-button" class="primaryButton">Merge models</button>
+          </div>
+        </div>`
+
+        
+    let loraUI=`
+        <div class="panel-box lora-manager-grid">
+            <div class="lora-manager-grid-selector">
+                <label for="#loraModel">Select Lora:</label>
+                <input id="loraModel" type="text" spellcheck="false" autocomplete="off" class="model-filter" data-path="" />
+            </div>
+            <div class="lora-manager-grid-thumbnail">
+                <p style="height:2em;">Thumbnail:</p>
+                <div style="position:relative; height:256px; width:256px;background-color:#222;border-radius:1em;margin-bottom:1em;">
+                    <i id="lora-manager-image-placeholder" class="fa-regular fa-image" style="font-size:500%;color:#555;position:absolute; top: 50%; left: 50%; transform: translate(-50%,-50%);"></i>
+                    <img id="lora-manager-image" class="displayNone" style="border-radius:6px;max-height:256px;max-width:256px;"/>
+                </div>
+                <div style="text-align:center;">
+                    <button class="tertiaryButton" id="lora-manager-upload-button"><i class="fa-solid fa-upload"></i> Upload new thumbnail</button>
+                    <input id="lora-manager-upload-input" name="lora-manager-upload-input" type="file" class="displayNone">
+                    <!-- button class="tertiaryButton"><i class="fa-solid fa-trash-can"></i> Remove</button -->
+                </div>
+            </div>
+            <div class="lora-manager-grid-keywords">
+                <p style="height:2em;">Keywords:
+                <span style="float:right;margin-bottom:4px;"><button id="lora-keyword-from-civitai" class="tertiaryButton smallButton">Import from Civitai</button></span></p>
+                <textarea style="width:100%;resize:vertical;" id="lora-manager-keywords" placeholder="Put LORA specific keywords here..."></textarea>
+                <p style="color:var(--small-label-color);">
+                    <b>LORA model keywords</b> can be used via the <code>+&nbsp;Embeddings</code> button. They get added to the embedding 
+                    keyword menu when the LORA has been selected in the image settings.
+                </p>
+            </div>
+            <div class="lora-manager-grid-notes">
+                <p style="height:2em;">Notes:</p>
+                <textarea style="width:100%;resize:vertical;"  id="lora-manager-notes" placeholder="Place for things you want to remember..."></textarea>
+                <p id="civitai-section" class="displayNone">
+                    <b>Civitai model page:</b>
+                    <a id="civitai-model-page" target="_blank"></a>
+                </p>
+            </div>
+        </div>`
+
+    let tabHTML=`
+        <div id="model-tool-tab-bar" class="tab-container tab-centered">
+            <span id="tab-model-loraUI" class="tab active">
+                <span><i class="fa-solid fa-key"></i> Lora Keywords</small></span>
+            </span>
+            <span id="tab-model-mergeUI" class="tab">
+                <span><i class="fa-solid fa-code-merge"></i> Merge Models</small></span>
+            </span>
+        </div>
+        <div id="model-tool-tab-content" class="panel-box">
+            <div id="tab-content-model-loraUI" class="tab-content active">
+                <div class="tab-content-inner">
+                    ${loraUI}
+                </div>
+            </div>
+
+            <div id="tab-content-model-mergeUI" class="tab-content">
+                <div class="tab-content-inner">
+                    ${mergeUI}
+                </div>
+            </div>
+        </div>`
+
+
+    ///////////////////// Function section
+    function smoothstep(x) {
+        return x * x * (3 - 2 * x)
+    }
+
+    function smootherstep(x) {
+        return x * x * x * (x * (x * 6 - 15) + 10)
+    }
+
+    function smootheststep(x) {
+        let y = -20 * Math.pow(x, 7)
+        y += 70 * Math.pow(x, 6)
+        y -= 84 * Math.pow(x, 5)
+        y += 35 * Math.pow(x, 4)
+        return y
+    }
+    function getCurrentTime() {
+        const now = new Date()
+        let hours = now.getHours()
+        let minutes = now.getMinutes()
+        let seconds = now.getSeconds()
+
+        hours = hours < 10 ? `0${hours}` : hours
+        minutes = minutes < 10 ? `0${minutes}` : minutes
+        seconds = seconds < 10 ? `0${seconds}` : seconds
+
+        return `${hours}:${minutes}:${seconds}`
+    }
+
+    function addLogMessage(message) {
+        const logContainer = document.getElementById("merge-log")
+        logContainer.innerHTML += `<i>${getCurrentTime()}</i> ${message}<br>`
+
+        // Scroll to the bottom of the log
+        logContainer.scrollTop = logContainer.scrollHeight
+
+        document.querySelector("#merge-log-container").style.display = "block"
+    }
+
+    function addLogSeparator() {
+        const logContainer = document.getElementById("merge-log")
+        logContainer.innerHTML += "<hr>"
+
+        logContainer.scrollTop = logContainer.scrollHeight
+    }
+
+    function drawDiagram(fn) {
+        const SIZE = 300
+        const canvas = document.getElementById("merge-canvas")
+        canvas.height = canvas.width = SIZE
+        const ctx = canvas.getContext("2d")
+
+        // Draw coordinate system
+        ctx.scale(1, -1)
+        ctx.translate(0, -canvas.height)
+        ctx.lineWidth = 1
+        ctx.beginPath()
+
+        ctx.strokeStyle = "white"
+        ctx.moveTo(0, 0)
+        ctx.lineTo(0, SIZE)
+        ctx.lineTo(SIZE, SIZE)
+        ctx.lineTo(SIZE, 0)
+        ctx.lineTo(0, 0)
+        ctx.lineTo(SIZE, SIZE)
+        ctx.stroke()
+        ctx.beginPath()
+        ctx.setLineDash([1, 2])
+        const n = SIZE / 10
+        for (let i = n; i < SIZE; i += n) {
+            ctx.moveTo(0, i)
+            ctx.lineTo(SIZE, i)
+            ctx.moveTo(i, 0)
+            ctx.lineTo(i, SIZE)
+        }
+        ctx.stroke()
+        ctx.beginPath()
+        ctx.setLineDash([])
+        ctx.beginPath()
+        ctx.strokeStyle = "black"
+        ctx.lineWidth = 3
+        // Plot function
+        const numSamples = 20
+        for (let i = 0; i <= numSamples; i++) {
+            const x = i / numSamples
+            const y = fn(x)
+
+            const canvasX = x * SIZE
+            const canvasY = y * SIZE
+
+            if (i === 0) {
+                ctx.moveTo(canvasX, canvasY)
+            } else {
+                ctx.lineTo(canvasX, canvasY)
+            }
+        }
+        ctx.stroke()
+        // Plot alpha values (yellow boxes)
+        let start = parseFloat(document.querySelector("#merge-start").value)
+        let step = parseFloat(document.querySelector("#merge-step").value)
+        let iterations = document.querySelector("#merge-count").value >> 0
+        ctx.beginPath()
+        ctx.fillStyle = "yellow"
+        for (let i = 0; i < iterations; i++) {
+            const alpha = (start + i * step) / 100
+            const x = alpha * SIZE
+            const y = fn(alpha) * SIZE
+            if (x <= SIZE) {
+                ctx.rect(x - 3, y - 3, 6, 6)
+                ctx.fill()
+            } else {
+                ctx.strokeStyle = "red"
+                ctx.moveTo(0, 0)
+                ctx.lineTo(0, SIZE)
+                ctx.lineTo(SIZE, SIZE)
+                ctx.lineTo(SIZE, 0)
+                ctx.lineTo(0, 0)
+                ctx.lineTo(SIZE, SIZE)
+                ctx.stroke()
+                addLogMessage("<i>Warning: maximum ratio is &#8805; 100%</i>")
+            }
+        }
+    }
+
+    function updateChart() {
+        let fn = (x) => x
+        switch (document.querySelector("#merge-interpolation").value) {
+            case "SmoothStep":
+                fn = smoothstep
+                break
+            case "SmootherStep":
+                fn = smootherstep
+                break
+            case "SmoothestStep":
+                fn = smootheststep
+                break
+        }
+        drawDiagram(fn)
+    }
+
+    function initMergeUI() {
+        const tabSettingsSingle = document.querySelector("#tab-merge-opts-single")
+        const tabSettingsBatch = document.querySelector("#tab-merge-opts-batch")
+        linkTabContents(tabSettingsSingle)
+        linkTabContents(tabSettingsBatch)
+
+        let mergeModelAField = new ModelDropdown(document.querySelector("#mergeModelA"), "stable-diffusion")
+        let mergeModelBField = new ModelDropdown(document.querySelector("#mergeModelB"), "stable-diffusion")
+        updateChart()
+
+        // slider
+        const singleMergeRatioField = document.querySelector("#single-merge-ratio")
+        const singleMergeRatioSlider = document.querySelector("#single-merge-ratio-slider")
+
+        function updateSingleMergeRatio() {
+            singleMergeRatioField.value = singleMergeRatioSlider.value / 10
+            singleMergeRatioField.dispatchEvent(new Event("change"))
+        }
+
+        function updateSingleMergeRatioSlider() {
+            if (singleMergeRatioField.value < 0) {
+                singleMergeRatioField.value = 0
+            } else if (singleMergeRatioField.value > 100) {
+                singleMergeRatioField.value = 100
+            }
+
+            singleMergeRatioSlider.value = singleMergeRatioField.value * 10
+            singleMergeRatioSlider.dispatchEvent(new Event("change"))
+        }
+
+        singleMergeRatioSlider.addEventListener("input", updateSingleMergeRatio)
+        singleMergeRatioField.addEventListener("input", updateSingleMergeRatioSlider)
+        updateSingleMergeRatio()
+
+        document.querySelector(".merge-config").addEventListener("change", updateChart)
+
+        document.querySelector("#merge-button").addEventListener("click", async function(e) {
+            // Build request template
+            let model0 = mergeModelAField.value
+            let model1 = mergeModelBField.value
+            let request = { model0: model0, model1: model1 }
+            request["use_fp16"] = document.querySelector("#merge-fp").value == "fp16"
+            let iterations = document.querySelector("#merge-count").value >> 0
+            let start = parseFloat(document.querySelector("#merge-start").value)
+            let step = parseFloat(document.querySelector("#merge-step").value)
+
+            if (isTabActive(tabSettingsSingle)) {
+                start = parseFloat(singleMergeRatioField.value)
+                step = 0
+                iterations = 1
+                addLogMessage(`merge ratio = ${start}%`)
+            } else {
+                addLogMessage(`start = ${start}%`)
+                addLogMessage(`step  = ${step}%`)
+            }
+
+            if (start + (iterations - 1) * step >= 100) {
+                addLogMessage("<i>Aborting: maximum ratio is &#8805; 100%</i>")
+                addLogMessage("Reduce the number of variations or the step size")
+                addLogSeparator()
+                document.querySelector("#merge-count").focus()
+                return
+            }
+
+            if (document.querySelector("#merge-filename").value == "") {
+                addLogMessage("<i>Aborting: No output file name specified</i>")
+                addLogSeparator()
+                document.querySelector("#merge-filename").focus()
+                return
+            }
+
+            // Disable merge button
+            e.target.disabled = true
+            e.target.classList.add("disabled")
+            let cursor = $("body").css("cursor")
+            let label = document.querySelector("#merge-button").innerHTML
+            $("body").css("cursor", "progress")
+            document.querySelector("#merge-button").innerHTML = "Merging models ..."
+
+            addLogMessage("Merging models")
+            addLogMessage("Model A: " + model0)
+            addLogMessage("Model B: " + model1)
+
+            // Batch main loop
+            for (let i = 0; i < iterations; i++) {
+                let alpha = (start + i * step) / 100
+
+                if (isTabActive(tabSettingsBatch)) {
+                    switch (document.querySelector("#merge-interpolation").value) {
+                        case "SmoothStep":
+                            alpha = smoothstep(alpha)
+                            break
+                        case "SmootherStep":
+                            alpha = smootherstep(alpha)
+                            break
+                        case "SmoothestStep":
+                            alpha = smootheststep(alpha)
+                            break
+                    }
+                }
+                addLogMessage(`merging batch job ${i + 1}/${iterations}, alpha = ${alpha.toFixed(5)}...`)
+
+                request["out_path"] = document.querySelector("#merge-filename").value
+                request["out_path"] += "-" + alpha.toFixed(5) + "." + document.querySelector("#merge-format").value
+                addLogMessage(`&nbsp;&nbsp;filename: ${request["out_path"]}`)
+
+                // sdkit documentation: "ratio - the ratio of the second model. 1 means only the second model will be used."
+                request["ratio"] = 1-alpha 
+                let res = await fetch("/model/merge", {
+                    method: "POST",
+                    headers: { "Content-Type": "application/json" },
+                    body: JSON.stringify(request),
+                })
+                const data = await res.json()
+                addLogMessage(JSON.stringify(data))
+            }
+            addLogMessage(
+                "<b>Done.</b> The models have been saved to your <tt>models/stable-diffusion</tt> folder."
+            )
+            addLogSeparator()
+            // Re-enable merge button
+            $("body").css("cursor", cursor)
+            document.querySelector("#merge-button").innerHTML = label
+            e.target.disabled = false
+            e.target.classList.remove("disabled")
+
+            // Update model list
+            stableDiffusionModelField.innerHTML = ""
+            vaeModelField.innerHTML = ""
+            hypernetworkModelField.innerHTML = ""
+            await getModels()
+        })
+    }
+
+    const LoraUI = {
+        modelField: undefined,
+        keywordsField: undefined,
+        notesField: undefined,
+        civitaiImportBtn: undefined,
+        civitaiSecion: undefined,
+        civitaiAnchor: undefined,
+        image: undefined,
+        imagePlaceholder: undefined,
+
+        init() {
+            LoraUI.modelField = new ModelDropdown(document.querySelector("#loraModel"), "lora", "None")
+            LoraUI.keywordsField = document.querySelector("#lora-manager-keywords")
+            LoraUI.notesField = document.querySelector("#lora-manager-notes")
+            LoraUI.civitaiImportBtn = document.querySelector("#lora-keyword-from-civitai")
+            LoraUI.civitaiSection = document.querySelector("#civitai-section")
+            LoraUI.civitaiAnchor = document.querySelector("#civitai-model-page")
+            LoraUI.image = document.querySelector("#lora-manager-image")
+            LoraUI.imagePlaceholder = document.querySelector("#lora-manager-image-placeholder")
+            LoraUI.uploadBtn = document.querySelector("#lora-manager-upload-button")
+            LoraUI.uploadInput = document.querySelector("#lora-manager-upload-input")
+
+            LoraUI.modelField.addEventListener("change", LoraUI.updateFields)
+            LoraUI.keywordsField.addEventListener("focusout", LoraUI.saveInfos)
+            LoraUI.notesField.addEventListener("focusout", LoraUI.saveInfos)
+            LoraUI.civitaiImportBtn.addEventListener("click", LoraUI.importFromCivitai)
+
+            LoraUI.uploadBtn.addEventListener("click", (e) => LoraUI.uploadInput.click())
+            LoraUI.uploadInput.addEventListener("change", LoraUI.uploadLoraThumb)
+
+            document.addEventListener("saveThumb", LoraUI.updateFields)
+
+            LoraUI.updateFields()
+        },
+
+        uploadLoraThumb(e) {
+            console.log(e)
+            if (LoraUI.uploadInput.files.length === 0) {
+                return
+            }
+
+            let reader = new FileReader()
+            let file = LoraUI.uploadInput.files[0]
+
+            reader.addEventListener("load", (event) => {
+                let img = document.createElement("img")
+                img.src = reader.result
+                onUseAsThumbnailClick(
+                    {
+                        use_lora_model: LoraUI.modelField.value,
+                    },
+                    img
+                )
+            })
+
+            if (file) {
+                reader.readAsDataURL(file)
+            }
+        },
+
+        updateFields() {
+            document.getElementById("civitai-section").classList.add("displayNone")
+            Bucket.retrieve(`modelinfo/lora/${LoraUI.modelField.value}`)
+                .then((info) => {
+                    if (info == null) {
+                        LoraUI.keywordsField.value = ""
+                        LoraUI.notesField.value = ""
+                        LoraUI.hideCivitaiLink()
+                    } else {
+                        LoraUI.keywordsField.value = info.keywords.join("\n")
+                        LoraUI.notesField.value = info.notes
+                        if ("civitai" in info && info["civitai"] != null) {
+                            LoraUI.showCivitaiLink(info.civitai)
+                        } else {
+                            LoraUI.hideCivitaiLink()
+                        }
+                    }
+                })
+            Bucket.getImageAsDataURL(`${profileNameField.value}/lora/${LoraUI.modelField.value}.png`)
+                .then((data) => {
+                    LoraUI.image.src=data
+                    LoraUI.image.classList.remove("displayNone")
+                    LoraUI.imagePlaceholder.classList.add("displayNone")
+                })
+                .catch((error) => {
+                    LoraUI.image.classList.add("displayNone")
+                    LoraUI.imagePlaceholder.classList.remove("displayNone")
+                })
+        },
+
+        saveInfos() {
+            let info = {
+                keywords: LoraUI.keywordsField.value
+                            .split("\n")
+                            .filter((x) => (x != "")),
+                notes: LoraUI.notesField.value,
+                civitai: LoraUI.civitaiSection.checkVisibility() ? LoraUI.civitaiAnchor.href : null, 
+            }
+            Bucket.store(`modelinfo/lora/${LoraUI.modelField.value}`, info)
+        },
+
+        importFromCivitai() {
+            document.body.style["cursor"] = "progress"
+            fetch("/sha256/lora/"+LoraUI.modelField.value)
+                .then((result) => result.json())
+                .then((json) => fetch("https://civitai.com/api/v1/model-versions/by-hash/" + json.digest))
+                .then((result) => result.json())
+                .then((json) => {
+                    document.body.style["cursor"] = "default"
+                    if (json == null) {
+                        return
+                    }
+                    if ("trainedWords" in json) {
+                        LoraUI.keywordsField.value = json["trainedWords"].join("\n")
+                    } else {
+                        showToast("No keyword info found.")
+                    }
+                    if ("modelId" in json) {
+                        LoraUI.showCivitaiLink("https://civitai.com/models/" + json.modelId)
+                    } else {
+                        LoraUI.hideCivitaiLink()
+                    }
+
+                    LoraUI.saveInfos()
+                })
+        },
+
+        showCivitaiLink(href) {
+            LoraUI.civitaiSection.classList.remove("displayNone")
+            LoraUI.civitaiAnchor.href = href
+            LoraUI.civitaiAnchor.innerHTML = LoraUI.civitaiAnchor.href
+        },
+
+        hideCivitaiLink() {
+            LoraUI.civitaiSection.classList.add("displayNone")
+        }
+    }
+
+    createTab({
+        id: "merge",
+        icon: "fa-toolbox",
+        label: "Model tools",
+        css: mergeCSS,
+        content: tabHTML, 
+        onOpen: ({ firstOpen }) => {
+            if (!firstOpen) {
+                return
+            }
+            initMergeUI()
+            LoraUI.init()
+            const tabMergeUI = document.querySelector("#tab-model-mergeUI")
+            const tabLoraUI = document.querySelector("#tab-model-loraUI")
+            linkTabContents(tabMergeUI)
+            linkTabContents(tabLoraUI)
+        },
+    })
+})()
+async function getLoraKeywords(model) {
+    return Bucket.retrieve(`modelinfo/lora/${model}`)
+        .then((info) => info ? info.keywords : [])
+}

ui/plugins/ui/snow.plugin.js

@@ -0,0 +1,80 @@
+// christmas hack, courtesy: https://pajasevi.github.io/CSSnowflakes/
+
+;(function(){
+    "use strict";
+
+    function makeItSnow() {
+        const styleSheet = document.createElement("style")
+        styleSheet.textContent = `
+        /* customizable snowflake styling */
+        .snowflake {
+            color: #fff;
+            font-size: 1em;
+            font-family: Arial, sans-serif;
+            text-shadow: 0 0 5px #000;
+        }
+
+        .snowflake,.snowflake .inner{animation-iteration-count:infinite;animation-play-state:running}@keyframes snowflakes-fall{0%{transform:translateY(0)}100%{transform:translateY(110vh)}}@keyframes snowflakes-shake{0%,100%{transform:translateX(0)}50%{transform:translateX(80px)}}.snowflake{position:fixed;top:-10%;z-index:9999;-webkit-user-select:none;user-select:none;cursor:default;animation-name:snowflakes-shake;animation-duration:3s;animation-timing-function:ease-in-out}.snowflake .inner{animation-duration:10s;animation-name:snowflakes-fall;animation-timing-function:linear}.snowflake:nth-of-type(0){left:1%;animation-delay:0s}.snowflake:nth-of-type(0) .inner{animation-delay:0s}.snowflake:first-of-type{left:10%;animation-delay:1s}.snowflake:first-of-type .inner,.snowflake:nth-of-type(8) .inner{animation-delay:1s}.snowflake:nth-of-type(2){left:20%;animation-delay:.5s}.snowflake:nth-of-type(2) .inner,.snowflake:nth-of-type(6) .inner{animation-delay:6s}.snowflake:nth-of-type(3){left:30%;animation-delay:2s}.snowflake:nth-of-type(11) .inner,.snowflake:nth-of-type(3) .inner{animation-delay:4s}.snowflake:nth-of-type(4){left:40%;animation-delay:2s}.snowflake:nth-of-type(10) .inner,.snowflake:nth-of-type(4) .inner{animation-delay:2s}.snowflake:nth-of-type(5){left:50%;animation-delay:3s}.snowflake:nth-of-type(5) .inner{animation-delay:8s}.snowflake:nth-of-type(6){left:60%;animation-delay:2s}.snowflake:nth-of-type(7){left:70%;animation-delay:1s}.snowflake:nth-of-type(7) .inner{animation-delay:2.5s}.snowflake:nth-of-type(8){left:80%;animation-delay:0s}.snowflake:nth-of-type(9){left:90%;animation-delay:1.5s}.snowflake:nth-of-type(9) .inner{animation-delay:3s}.snowflake:nth-of-type(10){left:25%;animation-delay:0s}.snowflake:nth-of-type(11){left:65%;animation-delay:2.5s}
+        `
+        document.head.appendChild(styleSheet)
+
+        const snowflakes = document.createElement("div")
+        snowflakes.id = "snowflakes-container"
+        snowflakes.innerHTML = `
+            <div class="snowflakes" aria-hidden="true">
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+                <div class="snowflake">
+                <div class="inner">❅</div>
+                </div>
+            </div>`
+
+        document.body.appendChild(snowflakes)
+
+        const script = document.createElement("script")
+        script.innerHTML = `
+            $(document).ready(function() {
+                setTimeout(function() {
+                    $("#snowflakes-container").fadeOut("slow", function() {$(this).remove()})
+                }, 10 * 1000)
+            })
+        `
+        document.body.appendChild(script)
+    }
+
+    let date = new Date()
+    if (date.getMonth() === 11 && date.getDate() >= 12) {
+        makeItSnow()
+    }
+})()