diff --git a/.eslintrc.js b/.eslintrc.js index cf8397695e1..2e7258f6b13 100644 --- a/.eslintrc.js +++ b/.eslintrc.js @@ -78,6 +78,8 @@ module.exports = { //extraNetworks.js requestGet: "readonly", popup: "readonly", + // profilerVisualization.js + createVisualizationTable: "readonly", // from python localization: "readonly", // progrssbar.js @@ -86,8 +88,6 @@ module.exports = { // imageviewer.js modalPrevImage: "readonly", modalNextImage: "readonly", - // token-counters.js - setupTokenCounters: "readonly", // localStorage.js localSet: "readonly", localGet: "readonly", diff --git a/.github/workflows/run_tests.yaml b/.github/workflows/run_tests.yaml index 3dafaf8dcfc..f42e4758e63 100644 --- a/.github/workflows/run_tests.yaml +++ b/.github/workflows/run_tests.yaml @@ -20,6 +20,12 @@ jobs: cache-dependency-path: | **/requirements*txt launch.py + - name: Cache models + id: cache-models + uses: actions/cache@v3 + with: + path: models + key: "2023-12-30" - name: Install test dependencies run: pip install wait-for-it -r requirements-test.txt env: @@ -33,6 +39,8 @@ jobs: TORCH_INDEX_URL: https://download.pytorch.org/whl/cpu WEBUI_LAUNCH_LIVE_OUTPUT: "1" PYTHONUNBUFFERED: "1" + - name: Print installed packages + run: pip freeze - name: Start test server run: > python -m coverage run @@ -49,7 +57,7 @@ jobs: 2>&1 | tee output.txt & - name: Run tests run: | - wait-for-it --service 127.0.0.1:7860 -t 600 + wait-for-it --service 127.0.0.1:7860 -t 20 python -m pytest -vv --junitxml=test/results.xml --cov . --cov-report=xml --verify-base-url test - name: Kill test server if: always() diff --git a/.gitignore b/.gitignore index 09734267ff5..6790e9ee728 100644 --- a/.gitignore +++ b/.gitignore @@ -37,3 +37,4 @@ notification.mp3 /node_modules /package-lock.json /.coverage* +/test/test_outputs diff --git a/CHANGELOG.md b/CHANGELOG.md index 67429bbff0f..0df47801ba4 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,3 +1,134 @@ +## 1.8.0-RC + +### Features: +* Update torch to version 2.1.2 +* Soft Inpainting ([#14208](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14208)) +* FP8 support ([#14031](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14031), [#14327](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14327)) +* Support for SDXL-Inpaint Model ([#14390](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14390)) +* Use Spandrel for upscaling and face restoration architectures ([#14425](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14425), [#14467](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14467), [#14473](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14473), [#14474](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14474), [#14477](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14477), [#14476](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14476), [#14484](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14484), [#14500](https://github.com/AUTOMATIC1111/stable-difusion-webui/pull/14500), [#14501](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14501), [#14504](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14504), [#14524](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14524), [#14809](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14809)) +* Automatic backwards version compatibility (when loading infotexts from old images with program version specified, will add compatibility settings) +* Implement zero terminal SNR noise schedule option (**[SEED BREAKING CHANGE](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Seed-breaking-changes#180-dev-170-225-2024-01-01---zero-terminal-snr-noise-schedule-option)**, [#14145](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14145), [#14979](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14979)) +* Add a [✨] button to run hires fix on selected image in the gallery (with help from [#14598](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14598), [#14626](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14626), [#14728](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14728)) +* [Separate assets repository](https://github.com/AUTOMATIC1111/stable-diffusion-webui-assets); serve fonts locally rather than from google's servers +* Official LCM Sampler Support ([#14583](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14583)) +* Add support for DAT upscaler models ([#14690](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14690), [#15039](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/15039)) +* Extra Networks Tree View ([#14588](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14588), [#14900](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14900)) +* NPU Support ([#14801](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14801)) +* Prompt comments support + +### Minor: +* Allow pasting in WIDTHxHEIGHT strings into the width/height fields ([#14296](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14296)) +* add option: Live preview in full page image viewer ([#14230](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14230), [#14307](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14307)) +* Add keyboard shortcuts for generate/skip/interrupt ([#14269](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14269)) +* Better TCMALLOC support on different platforms ([#14227](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14227), [#14883](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14883), [#14910](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14910)) +* Lora not found warning ([#14464](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14464)) +* Adding negative prompts to Loras in extra networks ([#14475](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14475)) +* xyz_grid: allow varying the seed along an axis separate from axis options ([#12180](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/12180)) +* option to convert VAE to bfloat16 (implementation of [#9295](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/9295)) +* Better IPEX support ([#14229](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14229), [#14353](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14353), [#14559](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14559), [#14562](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14562), [#14597](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14597)) +* Option to interrupt after current generation rather than immediately ([#13653](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/13653), [#14659](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14659)) +* Fullscreen Preview control fading/disable ([#14291](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14291)) +* Finer settings freezing control ([#13789](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/13789)) +* Increase Upscaler Limits ([#14589](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14589)) +* Adjust brush size with hotkeys ([#14638](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14638)) +* Add checkpoint info to csv log file when saving images ([#14663](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14663)) +* Make more columns resizable ([#14740](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14740), [#14884](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14884)) +* Add an option to not overlay original image for inpainting for #14727 +* Add Pad conds v0 option to support same generation with DDIM as before 1.6.0 +* Add "Interrupting..." placeholder. +* Button for refresh extensions list ([#14857](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14857)) +* Add an option to disable normalization after calculating emphasis. ([#14874](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14874)) +* When counting tokens, also include enabled styles (can be disabled in settings to revert to previous behavior) +* Configuration for the [📂] button for image gallery ([#14947](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14947)) +* Support inference with LyCORIS BOFT networks ([#14871](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14871), [#14973](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14973)) +* support resizable columns for touch (tablets) ([#15002](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/15002)) + +### Extensions and API: +* Removed packages from requirements: basicsr, gfpgan, realesrgan; as well as their dependencies: absl-py, addict, beautifulsoup4, future, gdown, grpcio, importlib-metadata, lmdb, lpips, Markdown, platformdirs, PySocks, soupsieve, tb-nightly, tensorboard-data-server, tomli, Werkzeug, yapf, zipp, soupsieve +* Enable task ids for API ([#14314](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14314)) +* add override_settings support for infotext API +* rename generation_parameters_copypaste module to infotext_utils +* prevent crash due to Script __init__ exception ([#14407](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14407)) +* Bump numpy to 1.26.2 ([#14471](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14471)) +* Add utility to inspect a model's dtype/device ([#14478](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14478)) +* Implement general forward method for all method in built-in lora ext ([#14547](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14547)) +* Execute model_loaded_callback after moving to target device ([#14563](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14563)) +* Add self to CFGDenoiserParams ([#14573](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14573)) +* Allow TLS with API only mode (--nowebui) ([#14593](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14593)) +* New callback: postprocess_image_after_composite ([#14657](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14657)) +* modules/api/api.py: add api endpoint to refresh embeddings list ([#14715](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14715)) +* set_named_arg ([#14773](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14773)) +* add before_token_counter callback and use it for prompt comments +* ResizeHandleRow - allow overridden column scale parameter ([#15004](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/15004)) + +### Performance +* Massive performance improvement for extra networks directories with a huge number of files in them in an attempt to tackle #14507 ([#14528](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14528)) +* Reduce unnecessary re-indexing extra networks directory ([#14512](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14512)) +* Avoid unnecessary `isfile`/`exists` calls ([#14527](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14527)) + +### Bug Fixes: +* fix multiple bugs related to styles multi-file support ([#14203](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14203), [#14276](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14276), [#14707](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14707)) +* Lora fixes ([#14300](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14300), [#14237](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14237), [#14546](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14546), [#14726](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14726)) +* Re-add setting lost as part of e294e46 ([#14266](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14266)) +* fix extras caption BLIP ([#14330](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14330)) +* include infotext into saved init image for img2img ([#14452](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14452)) +* xyz grid handle axis_type is None ([#14394](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14394)) +* Update Added (Fixed) IPV6 Functionality When there is No Webui Argument Passed webui.py ([#14354](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14354)) +* fix API thread safe issues of txt2img and img2img ([#14421](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14421)) +* handle selectable script_index is None ([#14487](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14487)) +* handle config.json failed to load ([#14525](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14525), [#14767](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14767)) +* paste infotext cast int as float ([#14523](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14523)) +* Ensure GRADIO_ANALYTICS_ENABLED is set early enough ([#14537](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14537)) +* Fix logging configuration again ([#14538](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14538)) +* Handle CondFunc exception when resolving attributes ([#14560](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14560)) +* Fix extras big batch crashes ([#14699](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14699)) +* Fix using wrong model caused by alias ([#14655](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14655)) +* Add # to the invalid_filename_chars list ([#14640](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14640)) +* Fix extension check for requirements ([#14639](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14639)) +* Fix tab indexes are reset after restart UI ([#14637](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14637)) +* Fix nested manual cast ([#14689](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14689)) +* Keep postprocessing upscale selected tab after restart ([#14702](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14702)) +* XYZ grid: filter out blank vals when axis is int or float type (like int axis seed) ([#14754](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14754)) +* fix CLIP Interrogator topN regex ([#14775](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14775)) +* Fix dtype error in MHA layer/change dtype checking mechanism for manual cast ([#14791](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14791)) +* catch load style.csv error ([#14814](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14814)) +* fix error when editing extra networks card +* fix extra networks metadata failing to work properly when you create the .json file with metadata for the first time. +* util.walk_files extensions case insensitive ([#14879](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14879)) +* if extensions page not loaded, prevent apply ([#14873](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14873)) +* call the right function for token counter in img2img +* Fix the bugs that search/reload will disappear when using other ExtraNetworks extensions ([#14939](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14939)) +* Gracefully handle mtime read exception from cache ([#14933](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14933)) +* Only trigger interrupt on `Esc` when interrupt button visible ([#14932](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14932)) +* Disable prompt token counters option actually disables token counting rather than just hiding results. +* avoid double upscaling in inpaint ([#14966](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14966)) +* Fix #14591 using translated content to do categories mapping ([#14995](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14995)) +* fix: the `split_threshold` parameter does not work when running Split oversized images ([#15006](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/15006)) +* Fix resize-handle for mobile ([#15010](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/15010), [#15065](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/15065)) + +### Other: +* Assign id for "extra_options". Replace numeric field with slider. ([#14270](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14270)) +* change state dict comparison to ref compare ([#14216](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14216)) +* Bump torch-rocm to 5.6/5.7 ([#14293](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14293)) +* Base output path off data path ([#14446](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14446)) +* reorder training preprocessing modules in extras tab ([#14367](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14367)) +* Remove `cleanup_models` code ([#14472](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14472)) +* only rewrite ui-config when there is change ([#14352](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14352)) +* Fix lint issue from 501993eb ([#14495](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14495)) +* Update README.md ([#14548](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14548)) +* hires button, fix seeds () +* Logging: set formatter correctly for fallback logger too ([#14618](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14618)) +* Read generation info from infotexts rather than json for internal needs (save, extract seed from generated pic) ([#14645](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14645)) +* improve get_crop_region ([#14709](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14709)) +* Bump safetensors' version to 0.4.2 ([#14782](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14782)) +* add tooltip create_submit_box ([#14803](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14803)) +* extensions tab table row hover highlight ([#14885](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14885)) +* Always add timestamp to displayed image ([#14890](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14890)) +* Added core.filemode=false so doesn't track changes in file permission… ([#14930](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14930)) +* Normalize command-line argument paths ([#14934](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14934), [#15035](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/15035)) +* Use original App Title in progress bar ([#14916](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14916)) +* register_tmp_file also for mtime ([#15012](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/15012)) + ## 1.7.0 ### Features: @@ -40,7 +171,8 @@ * infotext updates: add option to disregard certain infotext fields, add option to not include VAE in infotext, add explanation to infotext settings page, move some options to infotext settings page * add FP32 fallback support on sd_vae_approx ([#14046](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14046)) * support XYZ scripts / split hires path from unet ([#14126](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14126)) -* allow use of mutiple styles csv files ([#14125](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14125)) +* allow use of multiple styles csv files ([#14125](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14125)) +* make extra network card description plaintext by default, with an option (Treat card description as HTML) to re-enable HTML as it was (originally by [#13241](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/13241)) ### Extensions and API: * update gradio to 3.41.2 @@ -176,7 +308,7 @@ * new samplers: Restart, DPM++ 2M SDE Exponential, DPM++ 2M SDE Heun, DPM++ 2M SDE Heun Karras, DPM++ 2M SDE Heun Exponential, DPM++ 3M SDE, DPM++ 3M SDE Karras, DPM++ 3M SDE Exponential ([#12300](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/12300), [#12519](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/12519), [#12542](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/12542)) * rework DDIM, PLMS, UniPC to use CFG denoiser same as in k-diffusion samplers: * makes all of them work with img2img - * makes prompt composition posssible (AND) + * makes prompt composition possible (AND) * makes them available for SDXL * always show extra networks tabs in the UI ([#11808](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/11808)) * use less RAM when creating models ([#11958](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/11958), [#12599](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/12599)) @@ -352,7 +484,7 @@ * user metadata system for custom networks * extended Lora metadata editor: set activation text, default weight, view tags, training info * Lora extension rework to include other types of networks (all that were previously handled by LyCORIS extension) - * show github stars for extenstions + * show github stars for extensions * img2img batch mode can read extra stuff from png info * img2img batch works with subdirectories * hotkeys to move prompt elements: alt+left/right @@ -571,7 +703,7 @@ * do not wait for Stable Diffusion model to load at startup * add filename patterns: `[denoising]` * directory hiding for extra networks: dirs starting with `.` will hide their cards on extra network tabs unless specifically searched for - * LoRA: for the `<...>` text in prompt, use name of LoRA that is in the metdata of the file, if present, instead of filename (both can be used to activate LoRA) + * LoRA: for the `<...>` text in prompt, use name of LoRA that is in the metadata of the file, if present, instead of filename (both can be used to activate LoRA) * LoRA: read infotext params from kohya-ss's extension parameters if they are present and if his extension is not active * LoRA: fix some LoRAs not working (ones that have 3x3 convolution layer) * LoRA: add an option to use old method of applying LoRAs (producing same results as with kohya-ss) @@ -601,7 +733,7 @@ * fix gamepad navigation * make the lightbox fullscreen image function properly * fix squished thumbnails in extras tab - * keep "search" filter for extra networks when user refreshes the tab (previously it showed everthing after you refreshed) + * keep "search" filter for extra networks when user refreshes the tab (previously it showed everything after you refreshed) * fix webui showing the same image if you configure the generation to always save results into same file * fix bug with upscalers not working properly * fix MPS on PyTorch 2.0.1, Intel Macs @@ -619,7 +751,7 @@ * switch to PyTorch 2.0.0 (except for AMD GPUs) * visual improvements to custom code scripts * add filename patterns: `[clip_skip]`, `[hasprompt<>]`, `[batch_number]`, `[generation_number]` - * add support for saving init images in img2img, and record their hashes in infotext for reproducability + * add support for saving init images in img2img, and record their hashes in infotext for reproducibility * automatically select current word when adjusting weight with ctrl+up/down * add dropdowns for X/Y/Z plot * add setting: Stable Diffusion/Random number generator source: makes it possible to make images generated from a given manual seed consistent across different GPUs diff --git a/README.md b/README.md index 9f9f33b1295..f4cfcf29008 100644 --- a/README.md +++ b/README.md @@ -1,5 +1,5 @@ # Stable Diffusion web UI -A browser interface based on Gradio library for Stable Diffusion. +A web interface for Stable Diffusion, implemented using Gradio library. ![](screenshot.png) @@ -151,11 +151,12 @@ Licenses for borrowed code can be found in `Settings -> Licenses` screen, and al - Stable Diffusion - https://github.com/Stability-AI/stablediffusion, https://github.com/CompVis/taming-transformers - k-diffusion - https://github.com/crowsonkb/k-diffusion.git -- GFPGAN - https://github.com/TencentARC/GFPGAN.git -- CodeFormer - https://github.com/sczhou/CodeFormer -- ESRGAN - https://github.com/xinntao/ESRGAN -- SwinIR - https://github.com/JingyunLiang/SwinIR -- Swin2SR - https://github.com/mv-lab/swin2sr +- Spandrel - https://github.com/chaiNNer-org/spandrel implementing + - GFPGAN - https://github.com/TencentARC/GFPGAN.git + - CodeFormer - https://github.com/sczhou/CodeFormer + - ESRGAN - https://github.com/xinntao/ESRGAN + - SwinIR - https://github.com/JingyunLiang/SwinIR + - Swin2SR - https://github.com/mv-lab/swin2sr - LDSR - https://github.com/Hafiidz/latent-diffusion - MiDaS - https://github.com/isl-org/MiDaS - Ideas for optimizations - https://github.com/basujindal/stable-diffusion diff --git a/_typos.toml b/_typos.toml new file mode 100644 index 00000000000..1c63fe70331 --- /dev/null +++ b/_typos.toml @@ -0,0 +1,5 @@ +[default.extend-words] +# Part of "RGBa" (Pillow's pre-multiplied alpha RGB mode) +Ba = "Ba" +# HSA is something AMD uses for their GPUs +HSA = "HSA" diff --git a/configs/sd_xl_inpaint.yaml b/configs/sd_xl_inpaint.yaml new file mode 100644 index 00000000000..3bad372186f --- /dev/null +++ b/configs/sd_xl_inpaint.yaml @@ -0,0 +1,98 @@ +model: + target: sgm.models.diffusion.DiffusionEngine + params: + scale_factor: 0.13025 + disable_first_stage_autocast: True + + denoiser_config: + target: sgm.modules.diffusionmodules.denoiser.DiscreteDenoiser + params: + num_idx: 1000 + + weighting_config: + target: sgm.modules.diffusionmodules.denoiser_weighting.EpsWeighting + scaling_config: + target: sgm.modules.diffusionmodules.denoiser_scaling.EpsScaling + discretization_config: + target: sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization + + network_config: + target: sgm.modules.diffusionmodules.openaimodel.UNetModel + params: + adm_in_channels: 2816 + num_classes: sequential + use_checkpoint: True + in_channels: 9 + out_channels: 4 + model_channels: 320 + attention_resolutions: [4, 2] + num_res_blocks: 2 + channel_mult: [1, 2, 4] + num_head_channels: 64 + use_spatial_transformer: True + use_linear_in_transformer: True + transformer_depth: [1, 2, 10] # note: the first is unused (due to attn_res starting at 2) 32, 16, 8 --> 64, 32, 16 + context_dim: 2048 + spatial_transformer_attn_type: softmax-xformers + legacy: False + + conditioner_config: + target: sgm.modules.GeneralConditioner + params: + emb_models: + # crossattn cond + - is_trainable: False + input_key: txt + target: sgm.modules.encoders.modules.FrozenCLIPEmbedder + params: + layer: hidden + layer_idx: 11 + # crossattn and vector cond + - is_trainable: False + input_key: txt + target: sgm.modules.encoders.modules.FrozenOpenCLIPEmbedder2 + params: + arch: ViT-bigG-14 + version: laion2b_s39b_b160k + freeze: True + layer: penultimate + always_return_pooled: True + legacy: False + # vector cond + - is_trainable: False + input_key: original_size_as_tuple + target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND + params: + outdim: 256 # multiplied by two + # vector cond + - is_trainable: False + input_key: crop_coords_top_left + target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND + params: + outdim: 256 # multiplied by two + # vector cond + - is_trainable: False + input_key: target_size_as_tuple + target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND + params: + outdim: 256 # multiplied by two + + first_stage_config: + target: sgm.models.autoencoder.AutoencoderKLInferenceWrapper + params: + embed_dim: 4 + monitor: val/rec_loss + ddconfig: + attn_type: vanilla-xformers + double_z: true + z_channels: 4 + resolution: 256 + in_channels: 3 + out_ch: 3 + ch: 128 + ch_mult: [1, 2, 4, 4] + num_res_blocks: 2 + attn_resolutions: [] + dropout: 0.0 + lossconfig: + target: torch.nn.Identity diff --git a/extensions-builtin/LDSR/sd_hijack_ddpm_v1.py b/extensions-builtin/LDSR/sd_hijack_ddpm_v1.py index 04adc5eb2cf..9a1e0778f24 100644 --- a/extensions-builtin/LDSR/sd_hijack_ddpm_v1.py +++ b/extensions-builtin/LDSR/sd_hijack_ddpm_v1.py @@ -301,7 +301,7 @@ def p_losses(self, x_start, t, noise=None): elif self.parameterization == "x0": target = x_start else: - raise NotImplementedError(f"Paramterization {self.parameterization} not yet supported") + raise NotImplementedError(f"Parameterization {self.parameterization} not yet supported") loss = self.get_loss(model_out, target, mean=False).mean(dim=[1, 2, 3]) @@ -880,7 +880,7 @@ def forward(self, x, c, *args, **kwargs): def apply_model(self, x_noisy, t, cond, return_ids=False): if isinstance(cond, dict): - # hybrid case, cond is exptected to be a dict + # hybrid case, cond is expected to be a dict pass else: if not isinstance(cond, list): @@ -916,7 +916,7 @@ def apply_model(self, x_noisy, t, cond, return_ids=False): cond_list = [{c_key: [c[:, :, :, :, i]]} for i in range(c.shape[-1])] elif self.cond_stage_key == 'coordinates_bbox': - assert 'original_image_size' in self.split_input_params, 'BoudingBoxRescaling is missing original_image_size' + assert 'original_image_size' in self.split_input_params, 'BoundingBoxRescaling is missing original_image_size' # assuming padding of unfold is always 0 and its dilation is always 1 n_patches_per_row = int((w - ks[0]) / stride[0] + 1) @@ -926,7 +926,7 @@ def apply_model(self, x_noisy, t, cond, return_ids=False): num_downs = self.first_stage_model.encoder.num_resolutions - 1 rescale_latent = 2 ** (num_downs) - # get top left postions of patches as conforming for the bbbox tokenizer, therefore we + # get top left positions of patches as conforming for the bbbox tokenizer, therefore we # need to rescale the tl patch coordinates to be in between (0,1) tl_patch_coordinates = [(rescale_latent * stride[0] * (patch_nr % n_patches_per_row) / full_img_w, rescale_latent * stride[1] * (patch_nr // n_patches_per_row) / full_img_h) diff --git a/extensions-builtin/Lora/lyco_helpers.py b/extensions-builtin/Lora/lyco_helpers.py index 1679a0ce633..6f134d54eb1 100644 --- a/extensions-builtin/Lora/lyco_helpers.py +++ b/extensions-builtin/Lora/lyco_helpers.py @@ -30,7 +30,7 @@ def factorization(dimension: int, factor:int=-1) -> tuple[int, int]: In LoRA with Kroneckor Product, first value is a value for weight scale. secon value is a value for weight. - Becuase of non-commutative property, A⊗B ≠ B⊗A. Meaning of two matrices is slightly different. + Because of non-commutative property, A⊗B ≠ B⊗A. Meaning of two matrices is slightly different. examples) factor diff --git a/extensions-builtin/Lora/network.py b/extensions-builtin/Lora/network.py index 6021fd8de0f..b8fd91941c8 100644 --- a/extensions-builtin/Lora/network.py +++ b/extensions-builtin/Lora/network.py @@ -3,6 +3,9 @@ from collections import namedtuple import enum +import torch.nn as nn +import torch.nn.functional as F + from modules import sd_models, cache, errors, hashes, shared NetworkWeights = namedtuple('NetworkWeights', ['network_key', 'sd_key', 'w', 'sd_module']) @@ -115,6 +118,29 @@ def __init__(self, net: Network, weights: NetworkWeights): if hasattr(self.sd_module, 'weight'): self.shape = self.sd_module.weight.shape + self.ops = None + self.extra_kwargs = {} + if isinstance(self.sd_module, nn.Conv2d): + self.ops = F.conv2d + self.extra_kwargs = { + 'stride': self.sd_module.stride, + 'padding': self.sd_module.padding + } + elif isinstance(self.sd_module, nn.Linear): + self.ops = F.linear + elif isinstance(self.sd_module, nn.LayerNorm): + self.ops = F.layer_norm + self.extra_kwargs = { + 'normalized_shape': self.sd_module.normalized_shape, + 'eps': self.sd_module.eps + } + elif isinstance(self.sd_module, nn.GroupNorm): + self.ops = F.group_norm + self.extra_kwargs = { + 'num_groups': self.sd_module.num_groups, + 'eps': self.sd_module.eps + } + self.dim = None self.bias = weights.w.get("bias") self.alpha = weights.w["alpha"].item() if "alpha" in weights.w else None @@ -137,7 +163,7 @@ def calc_scale(self): def finalize_updown(self, updown, orig_weight, output_shape, ex_bias=None): if self.bias is not None: updown = updown.reshape(self.bias.shape) - updown += self.bias.to(orig_weight.device, dtype=orig_weight.dtype) + updown += self.bias.to(orig_weight.device, dtype=updown.dtype) updown = updown.reshape(output_shape) if len(output_shape) == 4: @@ -155,5 +181,10 @@ def calc_updown(self, target): raise NotImplementedError() def forward(self, x, y): - raise NotImplementedError() + """A general forward implementation for all modules""" + if self.ops is None: + raise NotImplementedError() + else: + updown, ex_bias = self.calc_updown(self.sd_module.weight) + return y + self.ops(x, weight=updown, bias=ex_bias, **self.extra_kwargs) diff --git a/extensions-builtin/Lora/network_full.py b/extensions-builtin/Lora/network_full.py index bf6930e96c0..f221c95f3b5 100644 --- a/extensions-builtin/Lora/network_full.py +++ b/extensions-builtin/Lora/network_full.py @@ -18,9 +18,9 @@ def __init__(self, net: network.Network, weights: network.NetworkWeights): def calc_updown(self, orig_weight): output_shape = self.weight.shape - updown = self.weight.to(orig_weight.device, dtype=orig_weight.dtype) + updown = self.weight.to(orig_weight.device) if self.ex_bias is not None: - ex_bias = self.ex_bias.to(orig_weight.device, dtype=orig_weight.dtype) + ex_bias = self.ex_bias.to(orig_weight.device) else: ex_bias = None diff --git a/extensions-builtin/Lora/network_glora.py b/extensions-builtin/Lora/network_glora.py index 492d487078d..efe5c6814fa 100644 --- a/extensions-builtin/Lora/network_glora.py +++ b/extensions-builtin/Lora/network_glora.py @@ -22,12 +22,12 @@ def __init__(self, net: network.Network, weights: network.NetworkWeights): self.w2b = weights.w["b2.weight"] def calc_updown(self, orig_weight): - w1a = self.w1a.to(orig_weight.device, dtype=orig_weight.dtype) - w1b = self.w1b.to(orig_weight.device, dtype=orig_weight.dtype) - w2a = self.w2a.to(orig_weight.device, dtype=orig_weight.dtype) - w2b = self.w2b.to(orig_weight.device, dtype=orig_weight.dtype) + w1a = self.w1a.to(orig_weight.device) + w1b = self.w1b.to(orig_weight.device) + w2a = self.w2a.to(orig_weight.device) + w2b = self.w2b.to(orig_weight.device) output_shape = [w1a.size(0), w1b.size(1)] - updown = ((w2b @ w1b) + ((orig_weight @ w2a) @ w1a)) + updown = ((w2b @ w1b) + ((orig_weight.to(dtype = w1a.dtype) @ w2a) @ w1a)) return self.finalize_updown(updown, orig_weight, output_shape) diff --git a/extensions-builtin/Lora/network_hada.py b/extensions-builtin/Lora/network_hada.py index 5fcb0695fbb..d95a0fd18e3 100644 --- a/extensions-builtin/Lora/network_hada.py +++ b/extensions-builtin/Lora/network_hada.py @@ -27,16 +27,16 @@ def __init__(self, net: network.Network, weights: network.NetworkWeights): self.t2 = weights.w.get("hada_t2") def calc_updown(self, orig_weight): - w1a = self.w1a.to(orig_weight.device, dtype=orig_weight.dtype) - w1b = self.w1b.to(orig_weight.device, dtype=orig_weight.dtype) - w2a = self.w2a.to(orig_weight.device, dtype=orig_weight.dtype) - w2b = self.w2b.to(orig_weight.device, dtype=orig_weight.dtype) + w1a = self.w1a.to(orig_weight.device) + w1b = self.w1b.to(orig_weight.device) + w2a = self.w2a.to(orig_weight.device) + w2b = self.w2b.to(orig_weight.device) output_shape = [w1a.size(0), w1b.size(1)] if self.t1 is not None: output_shape = [w1a.size(1), w1b.size(1)] - t1 = self.t1.to(orig_weight.device, dtype=orig_weight.dtype) + t1 = self.t1.to(orig_weight.device) updown1 = lyco_helpers.make_weight_cp(t1, w1a, w1b) output_shape += t1.shape[2:] else: @@ -45,7 +45,7 @@ def calc_updown(self, orig_weight): updown1 = lyco_helpers.rebuild_conventional(w1a, w1b, output_shape) if self.t2 is not None: - t2 = self.t2.to(orig_weight.device, dtype=orig_weight.dtype) + t2 = self.t2.to(orig_weight.device) updown2 = lyco_helpers.make_weight_cp(t2, w2a, w2b) else: updown2 = lyco_helpers.rebuild_conventional(w2a, w2b, output_shape) diff --git a/extensions-builtin/Lora/network_ia3.py b/extensions-builtin/Lora/network_ia3.py index 7edc4249791..96faeaf3ede 100644 --- a/extensions-builtin/Lora/network_ia3.py +++ b/extensions-builtin/Lora/network_ia3.py @@ -17,7 +17,7 @@ def __init__(self, net: network.Network, weights: network.NetworkWeights): self.on_input = weights.w["on_input"].item() def calc_updown(self, orig_weight): - w = self.w.to(orig_weight.device, dtype=orig_weight.dtype) + w = self.w.to(orig_weight.device) output_shape = [w.size(0), orig_weight.size(1)] if self.on_input: diff --git a/extensions-builtin/Lora/network_lokr.py b/extensions-builtin/Lora/network_lokr.py index 340acdab3d0..fcdaeafd896 100644 --- a/extensions-builtin/Lora/network_lokr.py +++ b/extensions-builtin/Lora/network_lokr.py @@ -37,22 +37,22 @@ def __init__(self, net: network.Network, weights: network.NetworkWeights): def calc_updown(self, orig_weight): if self.w1 is not None: - w1 = self.w1.to(orig_weight.device, dtype=orig_weight.dtype) + w1 = self.w1.to(orig_weight.device) else: - w1a = self.w1a.to(orig_weight.device, dtype=orig_weight.dtype) - w1b = self.w1b.to(orig_weight.device, dtype=orig_weight.dtype) + w1a = self.w1a.to(orig_weight.device) + w1b = self.w1b.to(orig_weight.device) w1 = w1a @ w1b if self.w2 is not None: - w2 = self.w2.to(orig_weight.device, dtype=orig_weight.dtype) + w2 = self.w2.to(orig_weight.device) elif self.t2 is None: - w2a = self.w2a.to(orig_weight.device, dtype=orig_weight.dtype) - w2b = self.w2b.to(orig_weight.device, dtype=orig_weight.dtype) + w2a = self.w2a.to(orig_weight.device) + w2b = self.w2b.to(orig_weight.device) w2 = w2a @ w2b else: - t2 = self.t2.to(orig_weight.device, dtype=orig_weight.dtype) - w2a = self.w2a.to(orig_weight.device, dtype=orig_weight.dtype) - w2b = self.w2b.to(orig_weight.device, dtype=orig_weight.dtype) + t2 = self.t2.to(orig_weight.device) + w2a = self.w2a.to(orig_weight.device) + w2b = self.w2b.to(orig_weight.device) w2 = lyco_helpers.make_weight_cp(t2, w2a, w2b) output_shape = [w1.size(0) * w2.size(0), w1.size(1) * w2.size(1)] diff --git a/extensions-builtin/Lora/network_lora.py b/extensions-builtin/Lora/network_lora.py index 26c0a72c237..4cc4029510f 100644 --- a/extensions-builtin/Lora/network_lora.py +++ b/extensions-builtin/Lora/network_lora.py @@ -61,13 +61,13 @@ def create_module(self, weights, key, none_ok=False): return module def calc_updown(self, orig_weight): - up = self.up_model.weight.to(orig_weight.device, dtype=orig_weight.dtype) - down = self.down_model.weight.to(orig_weight.device, dtype=orig_weight.dtype) + up = self.up_model.weight.to(orig_weight.device) + down = self.down_model.weight.to(orig_weight.device) output_shape = [up.size(0), down.size(1)] if self.mid_model is not None: # cp-decomposition - mid = self.mid_model.weight.to(orig_weight.device, dtype=orig_weight.dtype) + mid = self.mid_model.weight.to(orig_weight.device) updown = lyco_helpers.rebuild_cp_decomposition(up, down, mid) output_shape += mid.shape[2:] else: diff --git a/extensions-builtin/Lora/network_norm.py b/extensions-builtin/Lora/network_norm.py index ce450158068..d25afcbb928 100644 --- a/extensions-builtin/Lora/network_norm.py +++ b/extensions-builtin/Lora/network_norm.py @@ -18,10 +18,10 @@ def __init__(self, net: network.Network, weights: network.NetworkWeights): def calc_updown(self, orig_weight): output_shape = self.w_norm.shape - updown = self.w_norm.to(orig_weight.device, dtype=orig_weight.dtype) + updown = self.w_norm.to(orig_weight.device) if self.b_norm is not None: - ex_bias = self.b_norm.to(orig_weight.device, dtype=orig_weight.dtype) + ex_bias = self.b_norm.to(orig_weight.device) else: ex_bias = None diff --git a/extensions-builtin/Lora/network_oft.py b/extensions-builtin/Lora/network_oft.py index fa647020f0a..7821a8a7dbf 100644 --- a/extensions-builtin/Lora/network_oft.py +++ b/extensions-builtin/Lora/network_oft.py @@ -1,6 +1,5 @@ import torch import network -from lyco_helpers import factorization from einops import rearrange @@ -22,20 +21,28 @@ def __init__(self, net: network.Network, weights: network.NetworkWeights): self.org_module: list[torch.Module] = [self.sd_module] self.scale = 1.0 + self.is_R = False + self.is_boft = False - # kohya-ss + # kohya-ss/New LyCORIS OFT/BOFT if "oft_blocks" in weights.w.keys(): - self.is_kohya = True self.oft_blocks = weights.w["oft_blocks"] # (num_blocks, block_size, block_size) - self.alpha = weights.w["alpha"] # alpha is constraint + self.alpha = weights.w.get("alpha", None) # alpha is constraint self.dim = self.oft_blocks.shape[0] # lora dim - # LyCORIS + # Old LyCORIS OFT elif "oft_diag" in weights.w.keys(): - self.is_kohya = False + self.is_R = True self.oft_blocks = weights.w["oft_diag"] # self.alpha is unused self.dim = self.oft_blocks.shape[1] # (num_blocks, block_size, block_size) + # LyCORIS BOFT + if self.oft_blocks.dim() == 4: + self.is_boft = True + self.rescale = weights.w.get('rescale', None) + if self.rescale is not None: + self.rescale = self.rescale.reshape(-1, *[1]*(self.org_module[0].weight.dim() - 1)) + is_linear = type(self.sd_module) in [torch.nn.Linear, torch.nn.modules.linear.NonDynamicallyQuantizableLinear] is_conv = type(self.sd_module) in [torch.nn.Conv2d] is_other_linear = type(self.sd_module) in [torch.nn.MultiheadAttention] # unsupported @@ -47,36 +54,65 @@ def __init__(self, net: network.Network, weights: network.NetworkWeights): elif is_other_linear: self.out_dim = self.sd_module.embed_dim - if self.is_kohya: - self.constraint = self.alpha * self.out_dim - self.num_blocks = self.dim - self.block_size = self.out_dim // self.dim - else: + self.num_blocks = self.dim + self.block_size = self.out_dim // self.dim + self.constraint = (0 if self.alpha is None else self.alpha) * self.out_dim + if self.is_R: self.constraint = None - self.block_size, self.num_blocks = factorization(self.out_dim, self.dim) + self.block_size = self.dim + self.num_blocks = self.out_dim // self.dim + elif self.is_boft: + self.boft_m = self.oft_blocks.shape[0] + self.num_blocks = self.oft_blocks.shape[1] + self.block_size = self.oft_blocks.shape[2] + self.boft_b = self.block_size def calc_updown(self, orig_weight): - oft_blocks = self.oft_blocks.to(orig_weight.device, dtype=orig_weight.dtype) - eye = torch.eye(self.block_size, device=self.oft_blocks.device) - - if self.is_kohya: - block_Q = oft_blocks - oft_blocks.transpose(1, 2) # ensure skew-symmetric orthogonal matrix - norm_Q = torch.norm(block_Q.flatten()) - new_norm_Q = torch.clamp(norm_Q, max=self.constraint) - block_Q = block_Q * ((new_norm_Q + 1e-8) / (norm_Q + 1e-8)) + oft_blocks = self.oft_blocks.to(orig_weight.device) + eye = torch.eye(self.block_size, device=oft_blocks.device) + + if not self.is_R: + block_Q = oft_blocks - oft_blocks.transpose(-1, -2) # ensure skew-symmetric orthogonal matrix + if self.constraint != 0: + norm_Q = torch.norm(block_Q.flatten()) + new_norm_Q = torch.clamp(norm_Q, max=self.constraint.to(oft_blocks.device)) + block_Q = block_Q * ((new_norm_Q + 1e-8) / (norm_Q + 1e-8)) oft_blocks = torch.matmul(eye + block_Q, (eye - block_Q).float().inverse()) - R = oft_blocks.to(orig_weight.device, dtype=orig_weight.dtype) - - # This errors out for MultiheadAttention, might need to be handled up-stream - merged_weight = rearrange(orig_weight, '(k n) ... -> k n ...', k=self.num_blocks, n=self.block_size) - merged_weight = torch.einsum( - 'k n m, k n ... -> k m ...', - R, - merged_weight - ) - merged_weight = rearrange(merged_weight, 'k m ... -> (k m) ...') - - updown = merged_weight.to(orig_weight.device, dtype=orig_weight.dtype) - orig_weight + R = oft_blocks.to(orig_weight.device) + + if not self.is_boft: + # This errors out for MultiheadAttention, might need to be handled up-stream + merged_weight = rearrange(orig_weight, '(k n) ... -> k n ...', k=self.num_blocks, n=self.block_size) + merged_weight = torch.einsum( + 'k n m, k n ... -> k m ...', + R, + merged_weight + ) + merged_weight = rearrange(merged_weight, 'k m ... -> (k m) ...') + else: + # TODO: determine correct value for scale + scale = 1.0 + m = self.boft_m + b = self.boft_b + r_b = b // 2 + inp = orig_weight + for i in range(m): + bi = R[i] # b_num, b_size, b_size + if i == 0: + # Apply multiplier/scale and rescale into first weight + bi = bi * scale + (1 - scale) * eye + inp = rearrange(inp, "(c g k) ... -> (c k g) ...", g=2, k=2**i * r_b) + inp = rearrange(inp, "(d b) ... -> d b ...", b=b) + inp = torch.einsum("b i j, b j ... -> b i ...", bi, inp) + inp = rearrange(inp, "d b ... -> (d b) ...") + inp = rearrange(inp, "(c k g) ... -> (c g k) ...", g=2, k=2**i * r_b) + merged_weight = inp + + # Rescale mechanism + if self.rescale is not None: + merged_weight = self.rescale.to(merged_weight) * merged_weight + + updown = merged_weight.to(orig_weight.device) - orig_weight.to(merged_weight.dtype) output_shape = orig_weight.shape return self.finalize_updown(updown, orig_weight, output_shape) diff --git a/extensions-builtin/Lora/networks.py b/extensions-builtin/Lora/networks.py index 629bf85376d..04bd19117a5 100644 --- a/extensions-builtin/Lora/networks.py +++ b/extensions-builtin/Lora/networks.py @@ -1,3 +1,4 @@ +import gradio as gr import logging import os import re @@ -259,11 +260,11 @@ def load_networks(names, te_multipliers=None, unet_multipliers=None, dyn_dims=No loaded_networks.clear() - networks_on_disk = [available_network_aliases.get(name, None) for name in names] + networks_on_disk = [available_networks.get(name, None) if name.lower() in forbidden_network_aliases else available_network_aliases.get(name, None) for name in names] if any(x is None for x in networks_on_disk): list_available_networks() - networks_on_disk = [available_network_aliases.get(name, None) for name in names] + networks_on_disk = [available_networks.get(name, None) if name.lower() in forbidden_network_aliases else available_network_aliases.get(name, None) for name in names] failed_to_load_networks = [] @@ -314,7 +315,12 @@ def load_networks(names, te_multipliers=None, unet_multipliers=None, dyn_dims=No emb_db.skipped_embeddings[name] = embedding if failed_to_load_networks: - sd_hijack.model_hijack.comments.append("Networks not found: " + ", ".join(failed_to_load_networks)) + lora_not_found_message = f'Lora not found: {", ".join(failed_to_load_networks)}' + sd_hijack.model_hijack.comments.append(lora_not_found_message) + if shared.opts.lora_not_found_warning_console: + print(f'\n{lora_not_found_message}\n') + if shared.opts.lora_not_found_gradio_warning: + gr.Warning(lora_not_found_message) purge_networks_from_memory() @@ -349,7 +355,7 @@ def network_apply_weights(self: Union[torch.nn.Conv2d, torch.nn.Linear, torch.nn """ Applies the currently selected set of networks to the weights of torch layer self. If weights already have this particular set of networks applied, does nothing. - If not, restores orginal weights from backup and alters weights according to networks. + If not, restores original weights from backup and alters weights according to networks. """ network_layer_name = getattr(self, 'network_layer_name', None) @@ -389,18 +395,26 @@ def network_apply_weights(self: Union[torch.nn.Conv2d, torch.nn.Linear, torch.nn if module is not None and hasattr(self, 'weight'): try: with torch.no_grad(): - updown, ex_bias = module.calc_updown(self.weight) + if getattr(self, 'fp16_weight', None) is None: + weight = self.weight + bias = self.bias + else: + weight = self.fp16_weight.clone().to(self.weight.device) + bias = getattr(self, 'fp16_bias', None) + if bias is not None: + bias = bias.clone().to(self.bias.device) + updown, ex_bias = module.calc_updown(weight) - if len(self.weight.shape) == 4 and self.weight.shape[1] == 9: + if len(weight.shape) == 4 and weight.shape[1] == 9: # inpainting model. zero pad updown to make channel[1] 4 to 9 updown = torch.nn.functional.pad(updown, (0, 0, 0, 0, 0, 5)) - self.weight += updown + self.weight.copy_((weight.to(dtype=updown.dtype) + updown).to(dtype=self.weight.dtype)) if ex_bias is not None and hasattr(self, 'bias'): if self.bias is None: - self.bias = torch.nn.Parameter(ex_bias) + self.bias = torch.nn.Parameter(ex_bias).to(self.weight.dtype) else: - self.bias += ex_bias + self.bias.copy_((bias + ex_bias).to(dtype=self.bias.dtype)) except RuntimeError as e: logging.debug(f"Network {net.name} layer {network_layer_name}: {e}") extra_network_lora.errors[net.name] = extra_network_lora.errors.get(net.name, 0) + 1 @@ -444,23 +458,23 @@ def network_apply_weights(self: Union[torch.nn.Conv2d, torch.nn.Linear, torch.nn self.network_current_names = wanted_names -def network_forward(module, input, original_forward): +def network_forward(org_module, input, original_forward): """ Old way of applying Lora by executing operations during layer's forward. Stacking many loras this way results in big performance degradation. """ if len(loaded_networks) == 0: - return original_forward(module, input) + return original_forward(org_module, input) input = devices.cond_cast_unet(input) - network_restore_weights_from_backup(module) - network_reset_cached_weight(module) + network_restore_weights_from_backup(org_module) + network_reset_cached_weight(org_module) - y = original_forward(module, input) + y = original_forward(org_module, input) - network_layer_name = getattr(module, 'network_layer_name', None) + network_layer_name = getattr(org_module, 'network_layer_name', None) for lora in loaded_networks: module = lora.modules.get(network_layer_name, None) if module is None: diff --git a/extensions-builtin/Lora/preload.py b/extensions-builtin/Lora/preload.py index 50961be33d7..52fab29b08b 100644 --- a/extensions-builtin/Lora/preload.py +++ b/extensions-builtin/Lora/preload.py @@ -1,7 +1,8 @@ import os from modules import paths +from modules.paths_internal import normalized_filepath def preload(parser): - parser.add_argument("--lora-dir", type=str, help="Path to directory with Lora networks.", default=os.path.join(paths.models_path, 'Lora')) - parser.add_argument("--lyco-dir-backcompat", type=str, help="Path to directory with LyCORIS networks (for backawards compatibility; can also use --lyco-dir).", default=os.path.join(paths.models_path, 'LyCORIS')) + parser.add_argument("--lora-dir", type=normalized_filepath, help="Path to directory with Lora networks.", default=os.path.join(paths.models_path, 'Lora')) + parser.add_argument("--lyco-dir-backcompat", type=normalized_filepath, help="Path to directory with LyCORIS networks (for backawards compatibility; can also use --lyco-dir).", default=os.path.join(paths.models_path, 'LyCORIS')) diff --git a/extensions-builtin/Lora/scripts/lora_script.py b/extensions-builtin/Lora/scripts/lora_script.py index ef23968c563..1518f7e5c89 100644 --- a/extensions-builtin/Lora/scripts/lora_script.py +++ b/extensions-builtin/Lora/scripts/lora_script.py @@ -39,6 +39,8 @@ def before_ui(): "lora_show_all": shared.OptionInfo(False, "Always show all networks on the Lora page").info("otherwise, those detected as for incompatible version of Stable Diffusion will be hidden"), "lora_hide_unknown_for_versions": shared.OptionInfo([], "Hide networks of unknown versions for model versions", gr.CheckboxGroup, {"choices": ["SD1", "SD2", "SDXL"]}), "lora_in_memory_limit": shared.OptionInfo(0, "Number of Lora networks to keep cached in memory", gr.Number, {"precision": 0}), + "lora_not_found_warning_console": shared.OptionInfo(False, "Lora not found warning in console"), + "lora_not_found_gradio_warning": shared.OptionInfo(False, "Lora not found warning popup in webui"), })) diff --git a/extensions-builtin/Lora/ui_edit_user_metadata.py b/extensions-builtin/Lora/ui_edit_user_metadata.py index c7011909055..3160aecfa38 100644 --- a/extensions-builtin/Lora/ui_edit_user_metadata.py +++ b/extensions-builtin/Lora/ui_edit_user_metadata.py @@ -54,12 +54,13 @@ def __init__(self, ui, tabname, page): self.slider_preferred_weight = None self.edit_notes = None - def save_lora_user_metadata(self, name, desc, sd_version, activation_text, preferred_weight, notes): + def save_lora_user_metadata(self, name, desc, sd_version, activation_text, preferred_weight, negative_text, notes): user_metadata = self.get_user_metadata(name) user_metadata["description"] = desc user_metadata["sd version"] = sd_version user_metadata["activation text"] = activation_text user_metadata["preferred weight"] = preferred_weight + user_metadata["negative text"] = negative_text user_metadata["notes"] = notes self.write_user_metadata(name, user_metadata) @@ -127,6 +128,7 @@ def put_values_into_components(self, name): gr.HighlightedText.update(value=gradio_tags, visible=True if tags else False), user_metadata.get('activation text', ''), float(user_metadata.get('preferred weight', 0.0)), + user_metadata.get('negative text', ''), gr.update(visible=True if tags else False), gr.update(value=self.generate_random_prompt_from_tags(tags), visible=True if tags else False), ] @@ -162,7 +164,7 @@ def create_editor(self): self.taginfo = gr.HighlightedText(label="Training dataset tags") self.edit_activation_text = gr.Text(label='Activation text', info="Will be added to prompt along with Lora") self.slider_preferred_weight = gr.Slider(label='Preferred weight', info="Set to 0 to disable", minimum=0.0, maximum=2.0, step=0.01) - + self.edit_negative_text = gr.Text(label='Negative prompt', info="Will be added to negative prompts") with gr.Row() as row_random_prompt: with gr.Column(scale=8): random_prompt = gr.Textbox(label='Random prompt', lines=4, max_lines=4, interactive=False) @@ -198,6 +200,7 @@ def select_tag(activation_text, evt: gr.SelectData): self.taginfo, self.edit_activation_text, self.slider_preferred_weight, + self.edit_negative_text, row_random_prompt, random_prompt, ] @@ -211,7 +214,9 @@ def select_tag(activation_text, evt: gr.SelectData): self.select_sd_version, self.edit_activation_text, self.slider_preferred_weight, + self.edit_negative_text, self.edit_notes, ] + self.setup_save_handler(self.button_save, self.save_lora_user_metadata, edited_components) diff --git a/extensions-builtin/Lora/ui_extra_networks_lora.py b/extensions-builtin/Lora/ui_extra_networks_lora.py index df02c663b12..66d15dd0507 100644 --- a/extensions-builtin/Lora/ui_extra_networks_lora.py +++ b/extensions-builtin/Lora/ui_extra_networks_lora.py @@ -24,13 +24,16 @@ def create_item(self, name, index=None, enable_filter=True): alias = lora_on_disk.get_alias() + search_terms = [self.search_terms_from_path(lora_on_disk.filename)] + if lora_on_disk.hash: + search_terms.append(lora_on_disk.hash) item = { "name": name, "filename": lora_on_disk.filename, "shorthash": lora_on_disk.shorthash, "preview": self.find_preview(path), "description": self.find_description(path), - "search_term": self.search_terms_from_path(lora_on_disk.filename) + " " + (lora_on_disk.hash or ""), + "search_terms": search_terms, "local_preview": f"{path}.{shared.opts.samples_format}", "metadata": lora_on_disk.metadata, "sort_keys": {'default': index, **self.get_sort_keys(lora_on_disk.filename)}, @@ -45,6 +48,11 @@ def create_item(self, name, index=None, enable_filter=True): if activation_text: item["prompt"] += " + " + quote_js(" " + activation_text) + negative_prompt = item["user_metadata"].get("negative text") + item["negative_prompt"] = quote_js("") + if negative_prompt: + item["negative_prompt"] = quote_js('(' + negative_prompt + ':1)') + sd_version = item["user_metadata"].get("sd version") if sd_version in network.SdVersion.__members__: item["sd_version"] = sd_version diff --git a/extensions-builtin/ScuNET/scripts/scunet_model.py b/extensions-builtin/ScuNET/scripts/scunet_model.py index 167d2f64b8e..fe5e5a19265 100644 --- a/extensions-builtin/ScuNET/scripts/scunet_model.py +++ b/extensions-builtin/ScuNET/scripts/scunet_model.py @@ -1,16 +1,9 @@ import sys import PIL.Image -import numpy as np -import torch -from tqdm import tqdm import modules.upscaler -from modules import devices, modelloader, script_callbacks, errors -from scunet_model_arch import SCUNet - -from modules.modelloader import load_file_from_url -from modules.shared import opts +from modules import devices, errors, modelloader, script_callbacks, shared, upscaler_utils class UpscalerScuNET(modules.upscaler.Upscaler): @@ -42,100 +35,37 @@ def __init__(self, dirname): scalers.append(scaler_data2) self.scalers = scalers - @staticmethod - @torch.no_grad() - def tiled_inference(img, model): - # test the image tile by tile - h, w = img.shape[2:] - tile = opts.SCUNET_tile - tile_overlap = opts.SCUNET_tile_overlap - if tile == 0: - return model(img) - - device = devices.get_device_for('scunet') - assert tile % 8 == 0, "tile size should be a multiple of window_size" - sf = 1 - - stride = tile - tile_overlap - h_idx_list = list(range(0, h - tile, stride)) + [h - tile] - w_idx_list = list(range(0, w - tile, stride)) + [w - tile] - E = torch.zeros(1, 3, h * sf, w * sf, dtype=img.dtype, device=device) - W = torch.zeros_like(E, dtype=devices.dtype, device=device) - - with tqdm(total=len(h_idx_list) * len(w_idx_list), desc="ScuNET tiles") as pbar: - for h_idx in h_idx_list: - - for w_idx in w_idx_list: - - in_patch = img[..., h_idx: h_idx + tile, w_idx: w_idx + tile] - - out_patch = model(in_patch) - out_patch_mask = torch.ones_like(out_patch) - - E[ - ..., h_idx * sf: (h_idx + tile) * sf, w_idx * sf: (w_idx + tile) * sf - ].add_(out_patch) - W[ - ..., h_idx * sf: (h_idx + tile) * sf, w_idx * sf: (w_idx + tile) * sf - ].add_(out_patch_mask) - pbar.update(1) - output = E.div_(W) - - return output - def do_upscale(self, img: PIL.Image.Image, selected_file): - devices.torch_gc() - try: model = self.load_model(selected_file) except Exception as e: print(f"ScuNET: Unable to load model from {selected_file}: {e}", file=sys.stderr) return img - device = devices.get_device_for('scunet') - tile = opts.SCUNET_tile - h, w = img.height, img.width - np_img = np.array(img) - np_img = np_img[:, :, ::-1] # RGB to BGR - np_img = np_img.transpose((2, 0, 1)) / 255 # HWC to CHW - torch_img = torch.from_numpy(np_img).float().unsqueeze(0).to(device) # type: ignore - - if tile > h or tile > w: - _img = torch.zeros(1, 3, max(h, tile), max(w, tile), dtype=torch_img.dtype, device=torch_img.device) - _img[:, :, :h, :w] = torch_img # pad image - torch_img = _img - - torch_output = self.tiled_inference(torch_img, model).squeeze(0) - torch_output = torch_output[:, :h * 1, :w * 1] # remove padding, if any - np_output: np.ndarray = torch_output.float().cpu().clamp_(0, 1).numpy() - del torch_img, torch_output + img = upscaler_utils.upscale_2( + img, + model, + tile_size=shared.opts.SCUNET_tile, + tile_overlap=shared.opts.SCUNET_tile_overlap, + scale=1, # ScuNET is a denoising model, not an upscaler + desc='ScuNET', + ) devices.torch_gc() - - output = np_output.transpose((1, 2, 0)) # CHW to HWC - output = output[:, :, ::-1] # BGR to RGB - return PIL.Image.fromarray((output * 255).astype(np.uint8)) + return img def load_model(self, path: str): device = devices.get_device_for('scunet') if path.startswith("http"): # TODO: this doesn't use `path` at all? - filename = load_file_from_url(self.model_url, model_dir=self.model_download_path, file_name=f"{self.name}.pth") + filename = modelloader.load_file_from_url(self.model_url, model_dir=self.model_download_path, file_name=f"{self.name}.pth") else: filename = path - model = SCUNet(in_nc=3, config=[4, 4, 4, 4, 4, 4, 4], dim=64) - model.load_state_dict(torch.load(filename), strict=True) - model.eval() - for _, v in model.named_parameters(): - v.requires_grad = False - model = model.to(device) - - return model + return modelloader.load_spandrel_model(filename, device=device, expected_architecture='SCUNet') def on_ui_settings(): import gradio as gr - from modules import shared shared.opts.add_option("SCUNET_tile", shared.OptionInfo(256, "Tile size for SCUNET upscalers.", gr.Slider, {"minimum": 0, "maximum": 512, "step": 16}, section=('upscaling', "Upscaling")).info("0 = no tiling")) shared.opts.add_option("SCUNET_tile_overlap", shared.OptionInfo(8, "Tile overlap for SCUNET upscalers.", gr.Slider, {"minimum": 0, "maximum": 64, "step": 1}, section=('upscaling', "Upscaling")).info("Low values = visible seam")) diff --git a/extensions-builtin/ScuNET/scunet_model_arch.py b/extensions-builtin/ScuNET/scunet_model_arch.py deleted file mode 100644 index b51a880629b..00000000000 --- a/extensions-builtin/ScuNET/scunet_model_arch.py +++ /dev/null @@ -1,268 +0,0 @@ -# -*- coding: utf-8 -*- -import numpy as np -import torch -import torch.nn as nn -from einops import rearrange -from einops.layers.torch import Rearrange -from timm.models.layers import trunc_normal_, DropPath - - -class WMSA(nn.Module): - """ Self-attention module in Swin Transformer - """ - - def __init__(self, input_dim, output_dim, head_dim, window_size, type): - super(WMSA, self).__init__() - self.input_dim = input_dim - self.output_dim = output_dim - self.head_dim = head_dim - self.scale = self.head_dim ** -0.5 - self.n_heads = input_dim // head_dim - self.window_size = window_size - self.type = type - self.embedding_layer = nn.Linear(self.input_dim, 3 * self.input_dim, bias=True) - - self.relative_position_params = nn.Parameter( - torch.zeros((2 * window_size - 1) * (2 * window_size - 1), self.n_heads)) - - self.linear = nn.Linear(self.input_dim, self.output_dim) - - trunc_normal_(self.relative_position_params, std=.02) - self.relative_position_params = torch.nn.Parameter( - self.relative_position_params.view(2 * window_size - 1, 2 * window_size - 1, self.n_heads).transpose(1, - 2).transpose( - 0, 1)) - - def generate_mask(self, h, w, p, shift): - """ generating the mask of SW-MSA - Args: - shift: shift parameters in CyclicShift. - Returns: - attn_mask: should be (1 1 w p p), - """ - # supporting square. - attn_mask = torch.zeros(h, w, p, p, p, p, dtype=torch.bool, device=self.relative_position_params.device) - if self.type == 'W': - return attn_mask - - s = p - shift - attn_mask[-1, :, :s, :, s:, :] = True - attn_mask[-1, :, s:, :, :s, :] = True - attn_mask[:, -1, :, :s, :, s:] = True - attn_mask[:, -1, :, s:, :, :s] = True - attn_mask = rearrange(attn_mask, 'w1 w2 p1 p2 p3 p4 -> 1 1 (w1 w2) (p1 p2) (p3 p4)') - return attn_mask - - def forward(self, x): - """ Forward pass of Window Multi-head Self-attention module. - Args: - x: input tensor with shape of [b h w c]; - attn_mask: attention mask, fill -inf where the value is True; - Returns: - output: tensor shape [b h w c] - """ - if self.type != 'W': - x = torch.roll(x, shifts=(-(self.window_size // 2), -(self.window_size // 2)), dims=(1, 2)) - - x = rearrange(x, 'b (w1 p1) (w2 p2) c -> b w1 w2 p1 p2 c', p1=self.window_size, p2=self.window_size) - h_windows = x.size(1) - w_windows = x.size(2) - # square validation - # assert h_windows == w_windows - - x = rearrange(x, 'b w1 w2 p1 p2 c -> b (w1 w2) (p1 p2) c', p1=self.window_size, p2=self.window_size) - qkv = self.embedding_layer(x) - q, k, v = rearrange(qkv, 'b nw np (threeh c) -> threeh b nw np c', c=self.head_dim).chunk(3, dim=0) - sim = torch.einsum('hbwpc,hbwqc->hbwpq', q, k) * self.scale - # Adding learnable relative embedding - sim = sim + rearrange(self.relative_embedding(), 'h p q -> h 1 1 p q') - # Using Attn Mask to distinguish different subwindows. - if self.type != 'W': - attn_mask = self.generate_mask(h_windows, w_windows, self.window_size, shift=self.window_size // 2) - sim = sim.masked_fill_(attn_mask, float("-inf")) - - probs = nn.functional.softmax(sim, dim=-1) - output = torch.einsum('hbwij,hbwjc->hbwic', probs, v) - output = rearrange(output, 'h b w p c -> b w p (h c)') - output = self.linear(output) - output = rearrange(output, 'b (w1 w2) (p1 p2) c -> b (w1 p1) (w2 p2) c', w1=h_windows, p1=self.window_size) - - if self.type != 'W': - output = torch.roll(output, shifts=(self.window_size // 2, self.window_size // 2), dims=(1, 2)) - - return output - - def relative_embedding(self): - cord = torch.tensor(np.array([[i, j] for i in range(self.window_size) for j in range(self.window_size)])) - relation = cord[:, None, :] - cord[None, :, :] + self.window_size - 1 - # negative is allowed - return self.relative_position_params[:, relation[:, :, 0].long(), relation[:, :, 1].long()] - - -class Block(nn.Module): - def __init__(self, input_dim, output_dim, head_dim, window_size, drop_path, type='W', input_resolution=None): - """ SwinTransformer Block - """ - super(Block, self).__init__() - self.input_dim = input_dim - self.output_dim = output_dim - assert type in ['W', 'SW'] - self.type = type - if input_resolution <= window_size: - self.type = 'W' - - self.ln1 = nn.LayerNorm(input_dim) - self.msa = WMSA(input_dim, input_dim, head_dim, window_size, self.type) - self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity() - self.ln2 = nn.LayerNorm(input_dim) - self.mlp = nn.Sequential( - nn.Linear(input_dim, 4 * input_dim), - nn.GELU(), - nn.Linear(4 * input_dim, output_dim), - ) - - def forward(self, x): - x = x + self.drop_path(self.msa(self.ln1(x))) - x = x + self.drop_path(self.mlp(self.ln2(x))) - return x - - -class ConvTransBlock(nn.Module): - def __init__(self, conv_dim, trans_dim, head_dim, window_size, drop_path, type='W', input_resolution=None): - """ SwinTransformer and Conv Block - """ - super(ConvTransBlock, self).__init__() - self.conv_dim = conv_dim - self.trans_dim = trans_dim - self.head_dim = head_dim - self.window_size = window_size - self.drop_path = drop_path - self.type = type - self.input_resolution = input_resolution - - assert self.type in ['W', 'SW'] - if self.input_resolution <= self.window_size: - self.type = 'W' - - self.trans_block = Block(self.trans_dim, self.trans_dim, self.head_dim, self.window_size, self.drop_path, - self.type, self.input_resolution) - self.conv1_1 = nn.Conv2d(self.conv_dim + self.trans_dim, self.conv_dim + self.trans_dim, 1, 1, 0, bias=True) - self.conv1_2 = nn.Conv2d(self.conv_dim + self.trans_dim, self.conv_dim + self.trans_dim, 1, 1, 0, bias=True) - - self.conv_block = nn.Sequential( - nn.Conv2d(self.conv_dim, self.conv_dim, 3, 1, 1, bias=False), - nn.ReLU(True), - nn.Conv2d(self.conv_dim, self.conv_dim, 3, 1, 1, bias=False) - ) - - def forward(self, x): - conv_x, trans_x = torch.split(self.conv1_1(x), (self.conv_dim, self.trans_dim), dim=1) - conv_x = self.conv_block(conv_x) + conv_x - trans_x = Rearrange('b c h w -> b h w c')(trans_x) - trans_x = self.trans_block(trans_x) - trans_x = Rearrange('b h w c -> b c h w')(trans_x) - res = self.conv1_2(torch.cat((conv_x, trans_x), dim=1)) - x = x + res - - return x - - -class SCUNet(nn.Module): - # def __init__(self, in_nc=3, config=[2, 2, 2, 2, 2, 2, 2], dim=64, drop_path_rate=0.0, input_resolution=256): - def __init__(self, in_nc=3, config=None, dim=64, drop_path_rate=0.0, input_resolution=256): - super(SCUNet, self).__init__() - if config is None: - config = [2, 2, 2, 2, 2, 2, 2] - self.config = config - self.dim = dim - self.head_dim = 32 - self.window_size = 8 - - # drop path rate for each layer - dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(config))] - - self.m_head = [nn.Conv2d(in_nc, dim, 3, 1, 1, bias=False)] - - begin = 0 - self.m_down1 = [ConvTransBlock(dim // 2, dim // 2, self.head_dim, self.window_size, dpr[i + begin], - 'W' if not i % 2 else 'SW', input_resolution) - for i in range(config[0])] + \ - [nn.Conv2d(dim, 2 * dim, 2, 2, 0, bias=False)] - - begin += config[0] - self.m_down2 = [ConvTransBlock(dim, dim, self.head_dim, self.window_size, dpr[i + begin], - 'W' if not i % 2 else 'SW', input_resolution // 2) - for i in range(config[1])] + \ - [nn.Conv2d(2 * dim, 4 * dim, 2, 2, 0, bias=False)] - - begin += config[1] - self.m_down3 = [ConvTransBlock(2 * dim, 2 * dim, self.head_dim, self.window_size, dpr[i + begin], - 'W' if not i % 2 else 'SW', input_resolution // 4) - for i in range(config[2])] + \ - [nn.Conv2d(4 * dim, 8 * dim, 2, 2, 0, bias=False)] - - begin += config[2] - self.m_body = [ConvTransBlock(4 * dim, 4 * dim, self.head_dim, self.window_size, dpr[i + begin], - 'W' if not i % 2 else 'SW', input_resolution // 8) - for i in range(config[3])] - - begin += config[3] - self.m_up3 = [nn.ConvTranspose2d(8 * dim, 4 * dim, 2, 2, 0, bias=False), ] + \ - [ConvTransBlock(2 * dim, 2 * dim, self.head_dim, self.window_size, dpr[i + begin], - 'W' if not i % 2 else 'SW', input_resolution // 4) - for i in range(config[4])] - - begin += config[4] - self.m_up2 = [nn.ConvTranspose2d(4 * dim, 2 * dim, 2, 2, 0, bias=False), ] + \ - [ConvTransBlock(dim, dim, self.head_dim, self.window_size, dpr[i + begin], - 'W' if not i % 2 else 'SW', input_resolution // 2) - for i in range(config[5])] - - begin += config[5] - self.m_up1 = [nn.ConvTranspose2d(2 * dim, dim, 2, 2, 0, bias=False), ] + \ - [ConvTransBlock(dim // 2, dim // 2, self.head_dim, self.window_size, dpr[i + begin], - 'W' if not i % 2 else 'SW', input_resolution) - for i in range(config[6])] - - self.m_tail = [nn.Conv2d(dim, in_nc, 3, 1, 1, bias=False)] - - self.m_head = nn.Sequential(*self.m_head) - self.m_down1 = nn.Sequential(*self.m_down1) - self.m_down2 = nn.Sequential(*self.m_down2) - self.m_down3 = nn.Sequential(*self.m_down3) - self.m_body = nn.Sequential(*self.m_body) - self.m_up3 = nn.Sequential(*self.m_up3) - self.m_up2 = nn.Sequential(*self.m_up2) - self.m_up1 = nn.Sequential(*self.m_up1) - self.m_tail = nn.Sequential(*self.m_tail) - # self.apply(self._init_weights) - - def forward(self, x0): - - h, w = x0.size()[-2:] - paddingBottom = int(np.ceil(h / 64) * 64 - h) - paddingRight = int(np.ceil(w / 64) * 64 - w) - x0 = nn.ReplicationPad2d((0, paddingRight, 0, paddingBottom))(x0) - - x1 = self.m_head(x0) - x2 = self.m_down1(x1) - x3 = self.m_down2(x2) - x4 = self.m_down3(x3) - x = self.m_body(x4) - x = self.m_up3(x + x4) - x = self.m_up2(x + x3) - x = self.m_up1(x + x2) - x = self.m_tail(x + x1) - - x = x[..., :h, :w] - - return x - - def _init_weights(self, m): - if isinstance(m, nn.Linear): - trunc_normal_(m.weight, std=.02) - if m.bias is not None: - nn.init.constant_(m.bias, 0) - elif isinstance(m, nn.LayerNorm): - nn.init.constant_(m.bias, 0) - nn.init.constant_(m.weight, 1.0) diff --git a/extensions-builtin/SwinIR/scripts/swinir_model.py b/extensions-builtin/SwinIR/scripts/swinir_model.py index ae0d0e6a8ea..16bf9b792fc 100644 --- a/extensions-builtin/SwinIR/scripts/swinir_model.py +++ b/extensions-builtin/SwinIR/scripts/swinir_model.py @@ -1,20 +1,15 @@ +import logging import sys -import platform -import numpy as np import torch from PIL import Image -from tqdm import tqdm -from modules import modelloader, devices, script_callbacks, shared -from modules.shared import opts, state -from swinir_model_arch import SwinIR -from swinir_model_arch_v2 import Swin2SR +from modules import devices, modelloader, script_callbacks, shared, upscaler_utils from modules.upscaler import Upscaler, UpscalerData SWINIR_MODEL_URL = "https://github.com/JingyunLiang/SwinIR/releases/download/v0.0/003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN.pth" -device_swinir = devices.get_device_for('swinir') +logger = logging.getLogger(__name__) class UpscalerSwinIR(Upscaler): @@ -37,26 +32,28 @@ def __init__(self, dirname): scalers.append(model_data) self.scalers = scalers - def do_upscale(self, img, model_file): - use_compile = hasattr(opts, 'SWIN_torch_compile') and opts.SWIN_torch_compile \ - and int(torch.__version__.split('.')[0]) >= 2 and platform.system() != "Windows" - current_config = (model_file, opts.SWIN_tile) + def do_upscale(self, img: Image.Image, model_file: str) -> Image.Image: + current_config = (model_file, shared.opts.SWIN_tile) - if use_compile and self._cached_model_config == current_config: + if self._cached_model_config == current_config: model = self._cached_model else: - self._cached_model = None try: model = self.load_model(model_file) except Exception as e: print(f"Failed loading SwinIR model {model_file}: {e}", file=sys.stderr) return img - model = model.to(device_swinir, dtype=devices.dtype) - if use_compile: - model = torch.compile(model) - self._cached_model = model - self._cached_model_config = current_config - img = upscale(img, model) + self._cached_model = model + self._cached_model_config = current_config + + img = upscaler_utils.upscale_2( + img, + model, + tile_size=shared.opts.SWIN_tile, + tile_overlap=shared.opts.SWIN_tile_overlap, + scale=model.scale, + desc="SwinIR", + ) devices.torch_gc() return img @@ -69,115 +66,22 @@ def load_model(self, path, scale=4): ) else: filename = path - if filename.endswith(".v2.pth"): - model = Swin2SR( - upscale=scale, - in_chans=3, - img_size=64, - window_size=8, - img_range=1.0, - depths=[6, 6, 6, 6, 6, 6], - embed_dim=180, - num_heads=[6, 6, 6, 6, 6, 6], - mlp_ratio=2, - upsampler="nearest+conv", - resi_connection="1conv", - ) - params = None - else: - model = SwinIR( - upscale=scale, - in_chans=3, - img_size=64, - window_size=8, - img_range=1.0, - depths=[6, 6, 6, 6, 6, 6, 6, 6, 6], - embed_dim=240, - num_heads=[8, 8, 8, 8, 8, 8, 8, 8, 8], - mlp_ratio=2, - upsampler="nearest+conv", - resi_connection="3conv", - ) - params = "params_ema" - pretrained_model = torch.load(filename) - if params is not None: - model.load_state_dict(pretrained_model[params], strict=True) - else: - model.load_state_dict(pretrained_model, strict=True) - return model - - -def upscale( - img, - model, - tile=None, - tile_overlap=None, - window_size=8, - scale=4, -): - tile = tile or opts.SWIN_tile - tile_overlap = tile_overlap or opts.SWIN_tile_overlap - - - img = np.array(img) - img = img[:, :, ::-1] - img = np.moveaxis(img, 2, 0) / 255 - img = torch.from_numpy(img).float() - img = img.unsqueeze(0).to(device_swinir, dtype=devices.dtype) - with torch.no_grad(), devices.autocast(): - _, _, h_old, w_old = img.size() - h_pad = (h_old // window_size + 1) * window_size - h_old - w_pad = (w_old // window_size + 1) * window_size - w_old - img = torch.cat([img, torch.flip(img, [2])], 2)[:, :, : h_old + h_pad, :] - img = torch.cat([img, torch.flip(img, [3])], 3)[:, :, :, : w_old + w_pad] - output = inference(img, model, tile, tile_overlap, window_size, scale) - output = output[..., : h_old * scale, : w_old * scale] - output = output.data.squeeze().float().cpu().clamp_(0, 1).numpy() - if output.ndim == 3: - output = np.transpose( - output[[2, 1, 0], :, :], (1, 2, 0) - ) # CHW-RGB to HCW-BGR - output = (output * 255.0).round().astype(np.uint8) # float32 to uint8 - return Image.fromarray(output, "RGB") - - -def inference(img, model, tile, tile_overlap, window_size, scale): - # test the image tile by tile - b, c, h, w = img.size() - tile = min(tile, h, w) - assert tile % window_size == 0, "tile size should be a multiple of window_size" - sf = scale - - stride = tile - tile_overlap - h_idx_list = list(range(0, h - tile, stride)) + [h - tile] - w_idx_list = list(range(0, w - tile, stride)) + [w - tile] - E = torch.zeros(b, c, h * sf, w * sf, dtype=devices.dtype, device=device_swinir).type_as(img) - W = torch.zeros_like(E, dtype=devices.dtype, device=device_swinir) - - with tqdm(total=len(h_idx_list) * len(w_idx_list), desc="SwinIR tiles") as pbar: - for h_idx in h_idx_list: - if state.interrupted or state.skipped: - break - - for w_idx in w_idx_list: - if state.interrupted or state.skipped: - break - - in_patch = img[..., h_idx: h_idx + tile, w_idx: w_idx + tile] - out_patch = model(in_patch) - out_patch_mask = torch.ones_like(out_patch) - - E[ - ..., h_idx * sf: (h_idx + tile) * sf, w_idx * sf: (w_idx + tile) * sf - ].add_(out_patch) - W[ - ..., h_idx * sf: (h_idx + tile) * sf, w_idx * sf: (w_idx + tile) * sf - ].add_(out_patch_mask) - pbar.update(1) - output = E.div_(W) - - return output + model_descriptor = modelloader.load_spandrel_model( + filename, + device=self._get_device(), + prefer_half=(devices.dtype == torch.float16), + expected_architecture="SwinIR", + ) + if getattr(shared.opts, 'SWIN_torch_compile', False): + try: + model_descriptor.model.compile() + except Exception: + logger.warning("Failed to compile SwinIR model, fallback to JIT", exc_info=True) + return model_descriptor + + def _get_device(self): + return devices.get_device_for('swinir') def on_ui_settings(): @@ -185,8 +89,7 @@ def on_ui_settings(): shared.opts.add_option("SWIN_tile", shared.OptionInfo(192, "Tile size for all SwinIR.", gr.Slider, {"minimum": 16, "maximum": 512, "step": 16}, section=('upscaling', "Upscaling"))) shared.opts.add_option("SWIN_tile_overlap", shared.OptionInfo(8, "Tile overlap, in pixels for SwinIR. Low values = visible seam.", gr.Slider, {"minimum": 0, "maximum": 48, "step": 1}, section=('upscaling', "Upscaling"))) - if int(torch.__version__.split('.')[0]) >= 2 and platform.system() != "Windows": # torch.compile() require pytorch 2.0 or above, and not on Windows - shared.opts.add_option("SWIN_torch_compile", shared.OptionInfo(False, "Use torch.compile to accelerate SwinIR.", gr.Checkbox, {"interactive": True}, section=('upscaling', "Upscaling")).info("Takes longer on first run")) + shared.opts.add_option("SWIN_torch_compile", shared.OptionInfo(False, "Use torch.compile to accelerate SwinIR.", gr.Checkbox, {"interactive": True}, section=('upscaling', "Upscaling")).info("Takes longer on first run")) script_callbacks.on_ui_settings(on_ui_settings) diff --git a/extensions-builtin/SwinIR/swinir_model_arch.py b/extensions-builtin/SwinIR/swinir_model_arch.py deleted file mode 100644 index 93b9327473a..00000000000 --- a/extensions-builtin/SwinIR/swinir_model_arch.py +++ /dev/null @@ -1,867 +0,0 @@ -# ----------------------------------------------------------------------------------- -# SwinIR: Image Restoration Using Swin Transformer, https://arxiv.org/abs/2108.10257 -# Originally Written by Ze Liu, Modified by Jingyun Liang. -# ----------------------------------------------------------------------------------- - -import math -import torch -import torch.nn as nn -import torch.nn.functional as F -import torch.utils.checkpoint as checkpoint -from timm.models.layers import DropPath, to_2tuple, trunc_normal_ - - -class Mlp(nn.Module): - def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.): - super().__init__() - out_features = out_features or in_features - hidden_features = hidden_features or in_features - self.fc1 = nn.Linear(in_features, hidden_features) - self.act = act_layer() - self.fc2 = nn.Linear(hidden_features, out_features) - self.drop = nn.Dropout(drop) - - def forward(self, x): - x = self.fc1(x) - x = self.act(x) - x = self.drop(x) - x = self.fc2(x) - x = self.drop(x) - return x - - -def window_partition(x, window_size): - """ - Args: - x: (B, H, W, C) - window_size (int): window size - - Returns: - windows: (num_windows*B, window_size, window_size, C) - """ - B, H, W, C = x.shape - x = x.view(B, H // window_size, window_size, W // window_size, window_size, C) - windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C) - return windows - - -def window_reverse(windows, window_size, H, W): - """ - Args: - windows: (num_windows*B, window_size, window_size, C) - window_size (int): Window size - H (int): Height of image - W (int): Width of image - - Returns: - x: (B, H, W, C) - """ - B = int(windows.shape[0] / (H * W / window_size / window_size)) - x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1) - x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1) - return x - - -class WindowAttention(nn.Module): - r""" Window based multi-head self attention (W-MSA) module with relative position bias. - It supports both of shifted and non-shifted window. - - Args: - dim (int): Number of input channels. - window_size (tuple[int]): The height and width of the window. - num_heads (int): Number of attention heads. - qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True - qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set - attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0 - proj_drop (float, optional): Dropout ratio of output. Default: 0.0 - """ - - def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, attn_drop=0., proj_drop=0.): - - super().__init__() - self.dim = dim - self.window_size = window_size # Wh, Ww - self.num_heads = num_heads - head_dim = dim // num_heads - self.scale = qk_scale or head_dim ** -0.5 - - # define a parameter table of relative position bias - self.relative_position_bias_table = nn.Parameter( - torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads)) # 2*Wh-1 * 2*Ww-1, nH - - # get pair-wise relative position index for each token inside the window - coords_h = torch.arange(self.window_size[0]) - coords_w = torch.arange(self.window_size[1]) - coords = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, Wh, Ww - coords_flatten = torch.flatten(coords, 1) # 2, Wh*Ww - relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :] # 2, Wh*Ww, Wh*Ww - relative_coords = relative_coords.permute(1, 2, 0).contiguous() # Wh*Ww, Wh*Ww, 2 - relative_coords[:, :, 0] += self.window_size[0] - 1 # shift to start from 0 - relative_coords[:, :, 1] += self.window_size[1] - 1 - relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1 - relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww - self.register_buffer("relative_position_index", relative_position_index) - - self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) - self.attn_drop = nn.Dropout(attn_drop) - self.proj = nn.Linear(dim, dim) - - self.proj_drop = nn.Dropout(proj_drop) - - trunc_normal_(self.relative_position_bias_table, std=.02) - self.softmax = nn.Softmax(dim=-1) - - def forward(self, x, mask=None): - """ - Args: - x: input features with shape of (num_windows*B, N, C) - mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None - """ - B_, N, C = x.shape - qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4) - q, k, v = qkv[0], qkv[1], qkv[2] # make torchscript happy (cannot use tensor as tuple) - - q = q * self.scale - attn = (q @ k.transpose(-2, -1)) - - relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view( - self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1) # Wh*Ww,Wh*Ww,nH - relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous() # nH, Wh*Ww, Wh*Ww - attn = attn + relative_position_bias.unsqueeze(0) - - if mask is not None: - nW = mask.shape[0] - attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0) - attn = attn.view(-1, self.num_heads, N, N) - attn = self.softmax(attn) - else: - attn = self.softmax(attn) - - attn = self.attn_drop(attn) - - x = (attn @ v).transpose(1, 2).reshape(B_, N, C) - x = self.proj(x) - x = self.proj_drop(x) - return x - - def extra_repr(self) -> str: - return f'dim={self.dim}, window_size={self.window_size}, num_heads={self.num_heads}' - - def flops(self, N): - # calculate flops for 1 window with token length of N - flops = 0 - # qkv = self.qkv(x) - flops += N * self.dim * 3 * self.dim - # attn = (q @ k.transpose(-2, -1)) - flops += self.num_heads * N * (self.dim // self.num_heads) * N - # x = (attn @ v) - flops += self.num_heads * N * N * (self.dim // self.num_heads) - # x = self.proj(x) - flops += N * self.dim * self.dim - return flops - - -class SwinTransformerBlock(nn.Module): - r""" Swin Transformer Block. - - Args: - dim (int): Number of input channels. - input_resolution (tuple[int]): Input resolution. - num_heads (int): Number of attention heads. - window_size (int): Window size. - shift_size (int): Shift size for SW-MSA. - mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. - qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True - qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. - drop (float, optional): Dropout rate. Default: 0.0 - attn_drop (float, optional): Attention dropout rate. Default: 0.0 - drop_path (float, optional): Stochastic depth rate. Default: 0.0 - act_layer (nn.Module, optional): Activation layer. Default: nn.GELU - norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm - """ - - def __init__(self, dim, input_resolution, num_heads, window_size=7, shift_size=0, - mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0., - act_layer=nn.GELU, norm_layer=nn.LayerNorm): - super().__init__() - self.dim = dim - self.input_resolution = input_resolution - self.num_heads = num_heads - self.window_size = window_size - self.shift_size = shift_size - self.mlp_ratio = mlp_ratio - if min(self.input_resolution) <= self.window_size: - # if window size is larger than input resolution, we don't partition windows - self.shift_size = 0 - self.window_size = min(self.input_resolution) - assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size" - - self.norm1 = norm_layer(dim) - self.attn = WindowAttention( - dim, window_size=to_2tuple(self.window_size), num_heads=num_heads, - qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop) - - self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity() - self.norm2 = norm_layer(dim) - mlp_hidden_dim = int(dim * mlp_ratio) - self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop) - - if self.shift_size > 0: - attn_mask = self.calculate_mask(self.input_resolution) - else: - attn_mask = None - - self.register_buffer("attn_mask", attn_mask) - - def calculate_mask(self, x_size): - # calculate attention mask for SW-MSA - H, W = x_size - img_mask = torch.zeros((1, H, W, 1)) # 1 H W 1 - h_slices = (slice(0, -self.window_size), - slice(-self.window_size, -self.shift_size), - slice(-self.shift_size, None)) - w_slices = (slice(0, -self.window_size), - slice(-self.window_size, -self.shift_size), - slice(-self.shift_size, None)) - cnt = 0 - for h in h_slices: - for w in w_slices: - img_mask[:, h, w, :] = cnt - cnt += 1 - - mask_windows = window_partition(img_mask, self.window_size) # nW, window_size, window_size, 1 - mask_windows = mask_windows.view(-1, self.window_size * self.window_size) - attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2) - attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0)) - - return attn_mask - - def forward(self, x, x_size): - H, W = x_size - B, L, C = x.shape - # assert L == H * W, "input feature has wrong size" - - shortcut = x - x = self.norm1(x) - x = x.view(B, H, W, C) - - # cyclic shift - if self.shift_size > 0: - shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2)) - else: - shifted_x = x - - # partition windows - x_windows = window_partition(shifted_x, self.window_size) # nW*B, window_size, window_size, C - x_windows = x_windows.view(-1, self.window_size * self.window_size, C) # nW*B, window_size*window_size, C - - # W-MSA/SW-MSA (to be compatible for testing on images whose shapes are the multiple of window size - if self.input_resolution == x_size: - attn_windows = self.attn(x_windows, mask=self.attn_mask) # nW*B, window_size*window_size, C - else: - attn_windows = self.attn(x_windows, mask=self.calculate_mask(x_size).to(x.device)) - - # merge windows - attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C) - shifted_x = window_reverse(attn_windows, self.window_size, H, W) # B H' W' C - - # reverse cyclic shift - if self.shift_size > 0: - x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2)) - else: - x = shifted_x - x = x.view(B, H * W, C) - - # FFN - x = shortcut + self.drop_path(x) - x = x + self.drop_path(self.mlp(self.norm2(x))) - - return x - - def extra_repr(self) -> str: - return f"dim={self.dim}, input_resolution={self.input_resolution}, num_heads={self.num_heads}, " \ - f"window_size={self.window_size}, shift_size={self.shift_size}, mlp_ratio={self.mlp_ratio}" - - def flops(self): - flops = 0 - H, W = self.input_resolution - # norm1 - flops += self.dim * H * W - # W-MSA/SW-MSA - nW = H * W / self.window_size / self.window_size - flops += nW * self.attn.flops(self.window_size * self.window_size) - # mlp - flops += 2 * H * W * self.dim * self.dim * self.mlp_ratio - # norm2 - flops += self.dim * H * W - return flops - - -class PatchMerging(nn.Module): - r""" Patch Merging Layer. - - Args: - input_resolution (tuple[int]): Resolution of input feature. - dim (int): Number of input channels. - norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm - """ - - def __init__(self, input_resolution, dim, norm_layer=nn.LayerNorm): - super().__init__() - self.input_resolution = input_resolution - self.dim = dim - self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False) - self.norm = norm_layer(4 * dim) - - def forward(self, x): - """ - x: B, H*W, C - """ - H, W = self.input_resolution - B, L, C = x.shape - assert L == H * W, "input feature has wrong size" - assert H % 2 == 0 and W % 2 == 0, f"x size ({H}*{W}) are not even." - - x = x.view(B, H, W, C) - - x0 = x[:, 0::2, 0::2, :] # B H/2 W/2 C - x1 = x[:, 1::2, 0::2, :] # B H/2 W/2 C - x2 = x[:, 0::2, 1::2, :] # B H/2 W/2 C - x3 = x[:, 1::2, 1::2, :] # B H/2 W/2 C - x = torch.cat([x0, x1, x2, x3], -1) # B H/2 W/2 4*C - x = x.view(B, -1, 4 * C) # B H/2*W/2 4*C - - x = self.norm(x) - x = self.reduction(x) - - return x - - def extra_repr(self) -> str: - return f"input_resolution={self.input_resolution}, dim={self.dim}" - - def flops(self): - H, W = self.input_resolution - flops = H * W * self.dim - flops += (H // 2) * (W // 2) * 4 * self.dim * 2 * self.dim - return flops - - -class BasicLayer(nn.Module): - """ A basic Swin Transformer layer for one stage. - - Args: - dim (int): Number of input channels. - input_resolution (tuple[int]): Input resolution. - depth (int): Number of blocks. - num_heads (int): Number of attention heads. - window_size (int): Local window size. - mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. - qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True - qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. - drop (float, optional): Dropout rate. Default: 0.0 - attn_drop (float, optional): Attention dropout rate. Default: 0.0 - drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 - norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm - downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None - use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. - """ - - def __init__(self, dim, input_resolution, depth, num_heads, window_size, - mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., - drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False): - - super().__init__() - self.dim = dim - self.input_resolution = input_resolution - self.depth = depth - self.use_checkpoint = use_checkpoint - - # build blocks - self.blocks = nn.ModuleList([ - SwinTransformerBlock(dim=dim, input_resolution=input_resolution, - num_heads=num_heads, window_size=window_size, - shift_size=0 if (i % 2 == 0) else window_size // 2, - mlp_ratio=mlp_ratio, - qkv_bias=qkv_bias, qk_scale=qk_scale, - drop=drop, attn_drop=attn_drop, - drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path, - norm_layer=norm_layer) - for i in range(depth)]) - - # patch merging layer - if downsample is not None: - self.downsample = downsample(input_resolution, dim=dim, norm_layer=norm_layer) - else: - self.downsample = None - - def forward(self, x, x_size): - for blk in self.blocks: - if self.use_checkpoint: - x = checkpoint.checkpoint(blk, x, x_size) - else: - x = blk(x, x_size) - if self.downsample is not None: - x = self.downsample(x) - return x - - def extra_repr(self) -> str: - return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}" - - def flops(self): - flops = 0 - for blk in self.blocks: - flops += blk.flops() - if self.downsample is not None: - flops += self.downsample.flops() - return flops - - -class RSTB(nn.Module): - """Residual Swin Transformer Block (RSTB). - - Args: - dim (int): Number of input channels. - input_resolution (tuple[int]): Input resolution. - depth (int): Number of blocks. - num_heads (int): Number of attention heads. - window_size (int): Local window size. - mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. - qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True - qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set. - drop (float, optional): Dropout rate. Default: 0.0 - attn_drop (float, optional): Attention dropout rate. Default: 0.0 - drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 - norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm - downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None - use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. - img_size: Input image size. - patch_size: Patch size. - resi_connection: The convolutional block before residual connection. - """ - - def __init__(self, dim, input_resolution, depth, num_heads, window_size, - mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., - drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False, - img_size=224, patch_size=4, resi_connection='1conv'): - super(RSTB, self).__init__() - - self.dim = dim - self.input_resolution = input_resolution - - self.residual_group = BasicLayer(dim=dim, - input_resolution=input_resolution, - depth=depth, - num_heads=num_heads, - window_size=window_size, - mlp_ratio=mlp_ratio, - qkv_bias=qkv_bias, qk_scale=qk_scale, - drop=drop, attn_drop=attn_drop, - drop_path=drop_path, - norm_layer=norm_layer, - downsample=downsample, - use_checkpoint=use_checkpoint) - - if resi_connection == '1conv': - self.conv = nn.Conv2d(dim, dim, 3, 1, 1) - elif resi_connection == '3conv': - # to save parameters and memory - self.conv = nn.Sequential(nn.Conv2d(dim, dim // 4, 3, 1, 1), nn.LeakyReLU(negative_slope=0.2, inplace=True), - nn.Conv2d(dim // 4, dim // 4, 1, 1, 0), - nn.LeakyReLU(negative_slope=0.2, inplace=True), - nn.Conv2d(dim // 4, dim, 3, 1, 1)) - - self.patch_embed = PatchEmbed( - img_size=img_size, patch_size=patch_size, in_chans=0, embed_dim=dim, - norm_layer=None) - - self.patch_unembed = PatchUnEmbed( - img_size=img_size, patch_size=patch_size, in_chans=0, embed_dim=dim, - norm_layer=None) - - def forward(self, x, x_size): - return self.patch_embed(self.conv(self.patch_unembed(self.residual_group(x, x_size), x_size))) + x - - def flops(self): - flops = 0 - flops += self.residual_group.flops() - H, W = self.input_resolution - flops += H * W * self.dim * self.dim * 9 - flops += self.patch_embed.flops() - flops += self.patch_unembed.flops() - - return flops - - -class PatchEmbed(nn.Module): - r""" Image to Patch Embedding - - Args: - img_size (int): Image size. Default: 224. - patch_size (int): Patch token size. Default: 4. - in_chans (int): Number of input image channels. Default: 3. - embed_dim (int): Number of linear projection output channels. Default: 96. - norm_layer (nn.Module, optional): Normalization layer. Default: None - """ - - def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None): - super().__init__() - img_size = to_2tuple(img_size) - patch_size = to_2tuple(patch_size) - patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]] - self.img_size = img_size - self.patch_size = patch_size - self.patches_resolution = patches_resolution - self.num_patches = patches_resolution[0] * patches_resolution[1] - - self.in_chans = in_chans - self.embed_dim = embed_dim - - if norm_layer is not None: - self.norm = norm_layer(embed_dim) - else: - self.norm = None - - def forward(self, x): - x = x.flatten(2).transpose(1, 2) # B Ph*Pw C - if self.norm is not None: - x = self.norm(x) - return x - - def flops(self): - flops = 0 - H, W = self.img_size - if self.norm is not None: - flops += H * W * self.embed_dim - return flops - - -class PatchUnEmbed(nn.Module): - r""" Image to Patch Unembedding - - Args: - img_size (int): Image size. Default: 224. - patch_size (int): Patch token size. Default: 4. - in_chans (int): Number of input image channels. Default: 3. - embed_dim (int): Number of linear projection output channels. Default: 96. - norm_layer (nn.Module, optional): Normalization layer. Default: None - """ - - def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None): - super().__init__() - img_size = to_2tuple(img_size) - patch_size = to_2tuple(patch_size) - patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]] - self.img_size = img_size - self.patch_size = patch_size - self.patches_resolution = patches_resolution - self.num_patches = patches_resolution[0] * patches_resolution[1] - - self.in_chans = in_chans - self.embed_dim = embed_dim - - def forward(self, x, x_size): - B, HW, C = x.shape - x = x.transpose(1, 2).view(B, self.embed_dim, x_size[0], x_size[1]) # B Ph*Pw C - return x - - def flops(self): - flops = 0 - return flops - - -class Upsample(nn.Sequential): - """Upsample module. - - Args: - scale (int): Scale factor. Supported scales: 2^n and 3. - num_feat (int): Channel number of intermediate features. - """ - - def __init__(self, scale, num_feat): - m = [] - if (scale & (scale - 1)) == 0: # scale = 2^n - for _ in range(int(math.log(scale, 2))): - m.append(nn.Conv2d(num_feat, 4 * num_feat, 3, 1, 1)) - m.append(nn.PixelShuffle(2)) - elif scale == 3: - m.append(nn.Conv2d(num_feat, 9 * num_feat, 3, 1, 1)) - m.append(nn.PixelShuffle(3)) - else: - raise ValueError(f'scale {scale} is not supported. ' 'Supported scales: 2^n and 3.') - super(Upsample, self).__init__(*m) - - -class UpsampleOneStep(nn.Sequential): - """UpsampleOneStep module (the difference with Upsample is that it always only has 1conv + 1pixelshuffle) - Used in lightweight SR to save parameters. - - Args: - scale (int): Scale factor. Supported scales: 2^n and 3. - num_feat (int): Channel number of intermediate features. - - """ - - def __init__(self, scale, num_feat, num_out_ch, input_resolution=None): - self.num_feat = num_feat - self.input_resolution = input_resolution - m = [] - m.append(nn.Conv2d(num_feat, (scale ** 2) * num_out_ch, 3, 1, 1)) - m.append(nn.PixelShuffle(scale)) - super(UpsampleOneStep, self).__init__(*m) - - def flops(self): - H, W = self.input_resolution - flops = H * W * self.num_feat * 3 * 9 - return flops - - -class SwinIR(nn.Module): - r""" SwinIR - A PyTorch impl of : `SwinIR: Image Restoration Using Swin Transformer`, based on Swin Transformer. - - Args: - img_size (int | tuple(int)): Input image size. Default 64 - patch_size (int | tuple(int)): Patch size. Default: 1 - in_chans (int): Number of input image channels. Default: 3 - embed_dim (int): Patch embedding dimension. Default: 96 - depths (tuple(int)): Depth of each Swin Transformer layer. - num_heads (tuple(int)): Number of attention heads in different layers. - window_size (int): Window size. Default: 7 - mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4 - qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True - qk_scale (float): Override default qk scale of head_dim ** -0.5 if set. Default: None - drop_rate (float): Dropout rate. Default: 0 - attn_drop_rate (float): Attention dropout rate. Default: 0 - drop_path_rate (float): Stochastic depth rate. Default: 0.1 - norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm. - ape (bool): If True, add absolute position embedding to the patch embedding. Default: False - patch_norm (bool): If True, add normalization after patch embedding. Default: True - use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False - upscale: Upscale factor. 2/3/4/8 for image SR, 1 for denoising and compress artifact reduction - img_range: Image range. 1. or 255. - upsampler: The reconstruction reconstruction module. 'pixelshuffle'/'pixelshuffledirect'/'nearest+conv'/None - resi_connection: The convolutional block before residual connection. '1conv'/'3conv' - """ - - def __init__(self, img_size=64, patch_size=1, in_chans=3, - embed_dim=96, depths=(6, 6, 6, 6), num_heads=(6, 6, 6, 6), - window_size=7, mlp_ratio=4., qkv_bias=True, qk_scale=None, - drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1, - norm_layer=nn.LayerNorm, ape=False, patch_norm=True, - use_checkpoint=False, upscale=2, img_range=1., upsampler='', resi_connection='1conv', - **kwargs): - super(SwinIR, self).__init__() - num_in_ch = in_chans - num_out_ch = in_chans - num_feat = 64 - self.img_range = img_range - if in_chans == 3: - rgb_mean = (0.4488, 0.4371, 0.4040) - self.mean = torch.Tensor(rgb_mean).view(1, 3, 1, 1) - else: - self.mean = torch.zeros(1, 1, 1, 1) - self.upscale = upscale - self.upsampler = upsampler - self.window_size = window_size - - ##################################################################################################### - ################################### 1, shallow feature extraction ################################### - self.conv_first = nn.Conv2d(num_in_ch, embed_dim, 3, 1, 1) - - ##################################################################################################### - ################################### 2, deep feature extraction ###################################### - self.num_layers = len(depths) - self.embed_dim = embed_dim - self.ape = ape - self.patch_norm = patch_norm - self.num_features = embed_dim - self.mlp_ratio = mlp_ratio - - # split image into non-overlapping patches - self.patch_embed = PatchEmbed( - img_size=img_size, patch_size=patch_size, in_chans=embed_dim, embed_dim=embed_dim, - norm_layer=norm_layer if self.patch_norm else None) - num_patches = self.patch_embed.num_patches - patches_resolution = self.patch_embed.patches_resolution - self.patches_resolution = patches_resolution - - # merge non-overlapping patches into image - self.patch_unembed = PatchUnEmbed( - img_size=img_size, patch_size=patch_size, in_chans=embed_dim, embed_dim=embed_dim, - norm_layer=norm_layer if self.patch_norm else None) - - # absolute position embedding - if self.ape: - self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim)) - trunc_normal_(self.absolute_pos_embed, std=.02) - - self.pos_drop = nn.Dropout(p=drop_rate) - - # stochastic depth - dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))] # stochastic depth decay rule - - # build Residual Swin Transformer blocks (RSTB) - self.layers = nn.ModuleList() - for i_layer in range(self.num_layers): - layer = RSTB(dim=embed_dim, - input_resolution=(patches_resolution[0], - patches_resolution[1]), - depth=depths[i_layer], - num_heads=num_heads[i_layer], - window_size=window_size, - mlp_ratio=self.mlp_ratio, - qkv_bias=qkv_bias, qk_scale=qk_scale, - drop=drop_rate, attn_drop=attn_drop_rate, - drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])], # no impact on SR results - norm_layer=norm_layer, - downsample=None, - use_checkpoint=use_checkpoint, - img_size=img_size, - patch_size=patch_size, - resi_connection=resi_connection - - ) - self.layers.append(layer) - self.norm = norm_layer(self.num_features) - - # build the last conv layer in deep feature extraction - if resi_connection == '1conv': - self.conv_after_body = nn.Conv2d(embed_dim, embed_dim, 3, 1, 1) - elif resi_connection == '3conv': - # to save parameters and memory - self.conv_after_body = nn.Sequential(nn.Conv2d(embed_dim, embed_dim // 4, 3, 1, 1), - nn.LeakyReLU(negative_slope=0.2, inplace=True), - nn.Conv2d(embed_dim // 4, embed_dim // 4, 1, 1, 0), - nn.LeakyReLU(negative_slope=0.2, inplace=True), - nn.Conv2d(embed_dim // 4, embed_dim, 3, 1, 1)) - - ##################################################################################################### - ################################ 3, high quality image reconstruction ################################ - if self.upsampler == 'pixelshuffle': - # for classical SR - self.conv_before_upsample = nn.Sequential(nn.Conv2d(embed_dim, num_feat, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.upsample = Upsample(upscale, num_feat) - self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) - elif self.upsampler == 'pixelshuffledirect': - # for lightweight SR (to save parameters) - self.upsample = UpsampleOneStep(upscale, embed_dim, num_out_ch, - (patches_resolution[0], patches_resolution[1])) - elif self.upsampler == 'nearest+conv': - # for real-world SR (less artifacts) - self.conv_before_upsample = nn.Sequential(nn.Conv2d(embed_dim, num_feat, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.conv_up1 = nn.Conv2d(num_feat, num_feat, 3, 1, 1) - if self.upscale == 4: - self.conv_up2 = nn.Conv2d(num_feat, num_feat, 3, 1, 1) - self.conv_hr = nn.Conv2d(num_feat, num_feat, 3, 1, 1) - self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) - self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) - else: - # for image denoising and JPEG compression artifact reduction - self.conv_last = nn.Conv2d(embed_dim, num_out_ch, 3, 1, 1) - - self.apply(self._init_weights) - - def _init_weights(self, m): - if isinstance(m, nn.Linear): - trunc_normal_(m.weight, std=.02) - if isinstance(m, nn.Linear) and m.bias is not None: - nn.init.constant_(m.bias, 0) - elif isinstance(m, nn.LayerNorm): - nn.init.constant_(m.bias, 0) - nn.init.constant_(m.weight, 1.0) - - @torch.jit.ignore - def no_weight_decay(self): - return {'absolute_pos_embed'} - - @torch.jit.ignore - def no_weight_decay_keywords(self): - return {'relative_position_bias_table'} - - def check_image_size(self, x): - _, _, h, w = x.size() - mod_pad_h = (self.window_size - h % self.window_size) % self.window_size - mod_pad_w = (self.window_size - w % self.window_size) % self.window_size - x = F.pad(x, (0, mod_pad_w, 0, mod_pad_h), 'reflect') - return x - - def forward_features(self, x): - x_size = (x.shape[2], x.shape[3]) - x = self.patch_embed(x) - if self.ape: - x = x + self.absolute_pos_embed - x = self.pos_drop(x) - - for layer in self.layers: - x = layer(x, x_size) - - x = self.norm(x) # B L C - x = self.patch_unembed(x, x_size) - - return x - - def forward(self, x): - H, W = x.shape[2:] - x = self.check_image_size(x) - - self.mean = self.mean.type_as(x) - x = (x - self.mean) * self.img_range - - if self.upsampler == 'pixelshuffle': - # for classical SR - x = self.conv_first(x) - x = self.conv_after_body(self.forward_features(x)) + x - x = self.conv_before_upsample(x) - x = self.conv_last(self.upsample(x)) - elif self.upsampler == 'pixelshuffledirect': - # for lightweight SR - x = self.conv_first(x) - x = self.conv_after_body(self.forward_features(x)) + x - x = self.upsample(x) - elif self.upsampler == 'nearest+conv': - # for real-world SR - x = self.conv_first(x) - x = self.conv_after_body(self.forward_features(x)) + x - x = self.conv_before_upsample(x) - x = self.lrelu(self.conv_up1(torch.nn.functional.interpolate(x, scale_factor=2, mode='nearest'))) - if self.upscale == 4: - x = self.lrelu(self.conv_up2(torch.nn.functional.interpolate(x, scale_factor=2, mode='nearest'))) - x = self.conv_last(self.lrelu(self.conv_hr(x))) - else: - # for image denoising and JPEG compression artifact reduction - x_first = self.conv_first(x) - res = self.conv_after_body(self.forward_features(x_first)) + x_first - x = x + self.conv_last(res) - - x = x / self.img_range + self.mean - - return x[:, :, :H*self.upscale, :W*self.upscale] - - def flops(self): - flops = 0 - H, W = self.patches_resolution - flops += H * W * 3 * self.embed_dim * 9 - flops += self.patch_embed.flops() - for layer in self.layers: - flops += layer.flops() - flops += H * W * 3 * self.embed_dim * self.embed_dim - flops += self.upsample.flops() - return flops - - -if __name__ == '__main__': - upscale = 4 - window_size = 8 - height = (1024 // upscale // window_size + 1) * window_size - width = (720 // upscale // window_size + 1) * window_size - model = SwinIR(upscale=2, img_size=(height, width), - window_size=window_size, img_range=1., depths=[6, 6, 6, 6], - embed_dim=60, num_heads=[6, 6, 6, 6], mlp_ratio=2, upsampler='pixelshuffledirect') - print(model) - print(height, width, model.flops() / 1e9) - - x = torch.randn((1, 3, height, width)) - x = model(x) - print(x.shape) diff --git a/extensions-builtin/SwinIR/swinir_model_arch_v2.py b/extensions-builtin/SwinIR/swinir_model_arch_v2.py deleted file mode 100644 index dad22cca29e..00000000000 --- a/extensions-builtin/SwinIR/swinir_model_arch_v2.py +++ /dev/null @@ -1,1017 +0,0 @@ -# ----------------------------------------------------------------------------------- -# Swin2SR: Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration, https://arxiv.org/abs/ -# Written by Conde and Choi et al. -# ----------------------------------------------------------------------------------- - -import math -import numpy as np -import torch -import torch.nn as nn -import torch.nn.functional as F -import torch.utils.checkpoint as checkpoint -from timm.models.layers import DropPath, to_2tuple, trunc_normal_ - - -class Mlp(nn.Module): - def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.): - super().__init__() - out_features = out_features or in_features - hidden_features = hidden_features or in_features - self.fc1 = nn.Linear(in_features, hidden_features) - self.act = act_layer() - self.fc2 = nn.Linear(hidden_features, out_features) - self.drop = nn.Dropout(drop) - - def forward(self, x): - x = self.fc1(x) - x = self.act(x) - x = self.drop(x) - x = self.fc2(x) - x = self.drop(x) - return x - - -def window_partition(x, window_size): - """ - Args: - x: (B, H, W, C) - window_size (int): window size - Returns: - windows: (num_windows*B, window_size, window_size, C) - """ - B, H, W, C = x.shape - x = x.view(B, H // window_size, window_size, W // window_size, window_size, C) - windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C) - return windows - - -def window_reverse(windows, window_size, H, W): - """ - Args: - windows: (num_windows*B, window_size, window_size, C) - window_size (int): Window size - H (int): Height of image - W (int): Width of image - Returns: - x: (B, H, W, C) - """ - B = int(windows.shape[0] / (H * W / window_size / window_size)) - x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1) - x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1) - return x - -class WindowAttention(nn.Module): - r""" Window based multi-head self attention (W-MSA) module with relative position bias. - It supports both of shifted and non-shifted window. - Args: - dim (int): Number of input channels. - window_size (tuple[int]): The height and width of the window. - num_heads (int): Number of attention heads. - qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True - attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0 - proj_drop (float, optional): Dropout ratio of output. Default: 0.0 - pretrained_window_size (tuple[int]): The height and width of the window in pre-training. - """ - - def __init__(self, dim, window_size, num_heads, qkv_bias=True, attn_drop=0., proj_drop=0., - pretrained_window_size=(0, 0)): - - super().__init__() - self.dim = dim - self.window_size = window_size # Wh, Ww - self.pretrained_window_size = pretrained_window_size - self.num_heads = num_heads - - self.logit_scale = nn.Parameter(torch.log(10 * torch.ones((num_heads, 1, 1))), requires_grad=True) - - # mlp to generate continuous relative position bias - self.cpb_mlp = nn.Sequential(nn.Linear(2, 512, bias=True), - nn.ReLU(inplace=True), - nn.Linear(512, num_heads, bias=False)) - - # get relative_coords_table - relative_coords_h = torch.arange(-(self.window_size[0] - 1), self.window_size[0], dtype=torch.float32) - relative_coords_w = torch.arange(-(self.window_size[1] - 1), self.window_size[1], dtype=torch.float32) - relative_coords_table = torch.stack( - torch.meshgrid([relative_coords_h, - relative_coords_w])).permute(1, 2, 0).contiguous().unsqueeze(0) # 1, 2*Wh-1, 2*Ww-1, 2 - if pretrained_window_size[0] > 0: - relative_coords_table[:, :, :, 0] /= (pretrained_window_size[0] - 1) - relative_coords_table[:, :, :, 1] /= (pretrained_window_size[1] - 1) - else: - relative_coords_table[:, :, :, 0] /= (self.window_size[0] - 1) - relative_coords_table[:, :, :, 1] /= (self.window_size[1] - 1) - relative_coords_table *= 8 # normalize to -8, 8 - relative_coords_table = torch.sign(relative_coords_table) * torch.log2( - torch.abs(relative_coords_table) + 1.0) / np.log2(8) - - self.register_buffer("relative_coords_table", relative_coords_table) - - # get pair-wise relative position index for each token inside the window - coords_h = torch.arange(self.window_size[0]) - coords_w = torch.arange(self.window_size[1]) - coords = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, Wh, Ww - coords_flatten = torch.flatten(coords, 1) # 2, Wh*Ww - relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :] # 2, Wh*Ww, Wh*Ww - relative_coords = relative_coords.permute(1, 2, 0).contiguous() # Wh*Ww, Wh*Ww, 2 - relative_coords[:, :, 0] += self.window_size[0] - 1 # shift to start from 0 - relative_coords[:, :, 1] += self.window_size[1] - 1 - relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1 - relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww - self.register_buffer("relative_position_index", relative_position_index) - - self.qkv = nn.Linear(dim, dim * 3, bias=False) - if qkv_bias: - self.q_bias = nn.Parameter(torch.zeros(dim)) - self.v_bias = nn.Parameter(torch.zeros(dim)) - else: - self.q_bias = None - self.v_bias = None - self.attn_drop = nn.Dropout(attn_drop) - self.proj = nn.Linear(dim, dim) - self.proj_drop = nn.Dropout(proj_drop) - self.softmax = nn.Softmax(dim=-1) - - def forward(self, x, mask=None): - """ - Args: - x: input features with shape of (num_windows*B, N, C) - mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None - """ - B_, N, C = x.shape - qkv_bias = None - if self.q_bias is not None: - qkv_bias = torch.cat((self.q_bias, torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias)) - qkv = F.linear(input=x, weight=self.qkv.weight, bias=qkv_bias) - qkv = qkv.reshape(B_, N, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4) - q, k, v = qkv[0], qkv[1], qkv[2] # make torchscript happy (cannot use tensor as tuple) - - # cosine attention - attn = (F.normalize(q, dim=-1) @ F.normalize(k, dim=-1).transpose(-2, -1)) - logit_scale = torch.clamp(self.logit_scale, max=torch.log(torch.tensor(1. / 0.01)).to(self.logit_scale.device)).exp() - attn = attn * logit_scale - - relative_position_bias_table = self.cpb_mlp(self.relative_coords_table).view(-1, self.num_heads) - relative_position_bias = relative_position_bias_table[self.relative_position_index.view(-1)].view( - self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1) # Wh*Ww,Wh*Ww,nH - relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous() # nH, Wh*Ww, Wh*Ww - relative_position_bias = 16 * torch.sigmoid(relative_position_bias) - attn = attn + relative_position_bias.unsqueeze(0) - - if mask is not None: - nW = mask.shape[0] - attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0) - attn = attn.view(-1, self.num_heads, N, N) - attn = self.softmax(attn) - else: - attn = self.softmax(attn) - - attn = self.attn_drop(attn) - - x = (attn @ v).transpose(1, 2).reshape(B_, N, C) - x = self.proj(x) - x = self.proj_drop(x) - return x - - def extra_repr(self) -> str: - return f'dim={self.dim}, window_size={self.window_size}, ' \ - f'pretrained_window_size={self.pretrained_window_size}, num_heads={self.num_heads}' - - def flops(self, N): - # calculate flops for 1 window with token length of N - flops = 0 - # qkv = self.qkv(x) - flops += N * self.dim * 3 * self.dim - # attn = (q @ k.transpose(-2, -1)) - flops += self.num_heads * N * (self.dim // self.num_heads) * N - # x = (attn @ v) - flops += self.num_heads * N * N * (self.dim // self.num_heads) - # x = self.proj(x) - flops += N * self.dim * self.dim - return flops - -class SwinTransformerBlock(nn.Module): - r""" Swin Transformer Block. - Args: - dim (int): Number of input channels. - input_resolution (tuple[int]): Input resulotion. - num_heads (int): Number of attention heads. - window_size (int): Window size. - shift_size (int): Shift size for SW-MSA. - mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. - qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True - drop (float, optional): Dropout rate. Default: 0.0 - attn_drop (float, optional): Attention dropout rate. Default: 0.0 - drop_path (float, optional): Stochastic depth rate. Default: 0.0 - act_layer (nn.Module, optional): Activation layer. Default: nn.GELU - norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm - pretrained_window_size (int): Window size in pre-training. - """ - - def __init__(self, dim, input_resolution, num_heads, window_size=7, shift_size=0, - mlp_ratio=4., qkv_bias=True, drop=0., attn_drop=0., drop_path=0., - act_layer=nn.GELU, norm_layer=nn.LayerNorm, pretrained_window_size=0): - super().__init__() - self.dim = dim - self.input_resolution = input_resolution - self.num_heads = num_heads - self.window_size = window_size - self.shift_size = shift_size - self.mlp_ratio = mlp_ratio - if min(self.input_resolution) <= self.window_size: - # if window size is larger than input resolution, we don't partition windows - self.shift_size = 0 - self.window_size = min(self.input_resolution) - assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size" - - self.norm1 = norm_layer(dim) - self.attn = WindowAttention( - dim, window_size=to_2tuple(self.window_size), num_heads=num_heads, - qkv_bias=qkv_bias, attn_drop=attn_drop, proj_drop=drop, - pretrained_window_size=to_2tuple(pretrained_window_size)) - - self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity() - self.norm2 = norm_layer(dim) - mlp_hidden_dim = int(dim * mlp_ratio) - self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop) - - if self.shift_size > 0: - attn_mask = self.calculate_mask(self.input_resolution) - else: - attn_mask = None - - self.register_buffer("attn_mask", attn_mask) - - def calculate_mask(self, x_size): - # calculate attention mask for SW-MSA - H, W = x_size - img_mask = torch.zeros((1, H, W, 1)) # 1 H W 1 - h_slices = (slice(0, -self.window_size), - slice(-self.window_size, -self.shift_size), - slice(-self.shift_size, None)) - w_slices = (slice(0, -self.window_size), - slice(-self.window_size, -self.shift_size), - slice(-self.shift_size, None)) - cnt = 0 - for h in h_slices: - for w in w_slices: - img_mask[:, h, w, :] = cnt - cnt += 1 - - mask_windows = window_partition(img_mask, self.window_size) # nW, window_size, window_size, 1 - mask_windows = mask_windows.view(-1, self.window_size * self.window_size) - attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2) - attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0)) - - return attn_mask - - def forward(self, x, x_size): - H, W = x_size - B, L, C = x.shape - #assert L == H * W, "input feature has wrong size" - - shortcut = x - x = x.view(B, H, W, C) - - # cyclic shift - if self.shift_size > 0: - shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2)) - else: - shifted_x = x - - # partition windows - x_windows = window_partition(shifted_x, self.window_size) # nW*B, window_size, window_size, C - x_windows = x_windows.view(-1, self.window_size * self.window_size, C) # nW*B, window_size*window_size, C - - # W-MSA/SW-MSA (to be compatible for testing on images whose shapes are the multiple of window size - if self.input_resolution == x_size: - attn_windows = self.attn(x_windows, mask=self.attn_mask) # nW*B, window_size*window_size, C - else: - attn_windows = self.attn(x_windows, mask=self.calculate_mask(x_size).to(x.device)) - - # merge windows - attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C) - shifted_x = window_reverse(attn_windows, self.window_size, H, W) # B H' W' C - - # reverse cyclic shift - if self.shift_size > 0: - x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2)) - else: - x = shifted_x - x = x.view(B, H * W, C) - x = shortcut + self.drop_path(self.norm1(x)) - - # FFN - x = x + self.drop_path(self.norm2(self.mlp(x))) - - return x - - def extra_repr(self) -> str: - return f"dim={self.dim}, input_resolution={self.input_resolution}, num_heads={self.num_heads}, " \ - f"window_size={self.window_size}, shift_size={self.shift_size}, mlp_ratio={self.mlp_ratio}" - - def flops(self): - flops = 0 - H, W = self.input_resolution - # norm1 - flops += self.dim * H * W - # W-MSA/SW-MSA - nW = H * W / self.window_size / self.window_size - flops += nW * self.attn.flops(self.window_size * self.window_size) - # mlp - flops += 2 * H * W * self.dim * self.dim * self.mlp_ratio - # norm2 - flops += self.dim * H * W - return flops - -class PatchMerging(nn.Module): - r""" Patch Merging Layer. - Args: - input_resolution (tuple[int]): Resolution of input feature. - dim (int): Number of input channels. - norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm - """ - - def __init__(self, input_resolution, dim, norm_layer=nn.LayerNorm): - super().__init__() - self.input_resolution = input_resolution - self.dim = dim - self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False) - self.norm = norm_layer(2 * dim) - - def forward(self, x): - """ - x: B, H*W, C - """ - H, W = self.input_resolution - B, L, C = x.shape - assert L == H * W, "input feature has wrong size" - assert H % 2 == 0 and W % 2 == 0, f"x size ({H}*{W}) are not even." - - x = x.view(B, H, W, C) - - x0 = x[:, 0::2, 0::2, :] # B H/2 W/2 C - x1 = x[:, 1::2, 0::2, :] # B H/2 W/2 C - x2 = x[:, 0::2, 1::2, :] # B H/2 W/2 C - x3 = x[:, 1::2, 1::2, :] # B H/2 W/2 C - x = torch.cat([x0, x1, x2, x3], -1) # B H/2 W/2 4*C - x = x.view(B, -1, 4 * C) # B H/2*W/2 4*C - - x = self.reduction(x) - x = self.norm(x) - - return x - - def extra_repr(self) -> str: - return f"input_resolution={self.input_resolution}, dim={self.dim}" - - def flops(self): - H, W = self.input_resolution - flops = (H // 2) * (W // 2) * 4 * self.dim * 2 * self.dim - flops += H * W * self.dim // 2 - return flops - -class BasicLayer(nn.Module): - """ A basic Swin Transformer layer for one stage. - Args: - dim (int): Number of input channels. - input_resolution (tuple[int]): Input resolution. - depth (int): Number of blocks. - num_heads (int): Number of attention heads. - window_size (int): Local window size. - mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. - qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True - drop (float, optional): Dropout rate. Default: 0.0 - attn_drop (float, optional): Attention dropout rate. Default: 0.0 - drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 - norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm - downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None - use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. - pretrained_window_size (int): Local window size in pre-training. - """ - - def __init__(self, dim, input_resolution, depth, num_heads, window_size, - mlp_ratio=4., qkv_bias=True, drop=0., attn_drop=0., - drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False, - pretrained_window_size=0): - - super().__init__() - self.dim = dim - self.input_resolution = input_resolution - self.depth = depth - self.use_checkpoint = use_checkpoint - - # build blocks - self.blocks = nn.ModuleList([ - SwinTransformerBlock(dim=dim, input_resolution=input_resolution, - num_heads=num_heads, window_size=window_size, - shift_size=0 if (i % 2 == 0) else window_size // 2, - mlp_ratio=mlp_ratio, - qkv_bias=qkv_bias, - drop=drop, attn_drop=attn_drop, - drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path, - norm_layer=norm_layer, - pretrained_window_size=pretrained_window_size) - for i in range(depth)]) - - # patch merging layer - if downsample is not None: - self.downsample = downsample(input_resolution, dim=dim, norm_layer=norm_layer) - else: - self.downsample = None - - def forward(self, x, x_size): - for blk in self.blocks: - if self.use_checkpoint: - x = checkpoint.checkpoint(blk, x, x_size) - else: - x = blk(x, x_size) - if self.downsample is not None: - x = self.downsample(x) - return x - - def extra_repr(self) -> str: - return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}" - - def flops(self): - flops = 0 - for blk in self.blocks: - flops += blk.flops() - if self.downsample is not None: - flops += self.downsample.flops() - return flops - - def _init_respostnorm(self): - for blk in self.blocks: - nn.init.constant_(blk.norm1.bias, 0) - nn.init.constant_(blk.norm1.weight, 0) - nn.init.constant_(blk.norm2.bias, 0) - nn.init.constant_(blk.norm2.weight, 0) - -class PatchEmbed(nn.Module): - r""" Image to Patch Embedding - Args: - img_size (int): Image size. Default: 224. - patch_size (int): Patch token size. Default: 4. - in_chans (int): Number of input image channels. Default: 3. - embed_dim (int): Number of linear projection output channels. Default: 96. - norm_layer (nn.Module, optional): Normalization layer. Default: None - """ - - def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None): - super().__init__() - img_size = to_2tuple(img_size) - patch_size = to_2tuple(patch_size) - patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]] - self.img_size = img_size - self.patch_size = patch_size - self.patches_resolution = patches_resolution - self.num_patches = patches_resolution[0] * patches_resolution[1] - - self.in_chans = in_chans - self.embed_dim = embed_dim - - self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size) - if norm_layer is not None: - self.norm = norm_layer(embed_dim) - else: - self.norm = None - - def forward(self, x): - B, C, H, W = x.shape - # FIXME look at relaxing size constraints - # assert H == self.img_size[0] and W == self.img_size[1], - # f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})." - x = self.proj(x).flatten(2).transpose(1, 2) # B Ph*Pw C - if self.norm is not None: - x = self.norm(x) - return x - - def flops(self): - Ho, Wo = self.patches_resolution - flops = Ho * Wo * self.embed_dim * self.in_chans * (self.patch_size[0] * self.patch_size[1]) - if self.norm is not None: - flops += Ho * Wo * self.embed_dim - return flops - -class RSTB(nn.Module): - """Residual Swin Transformer Block (RSTB). - - Args: - dim (int): Number of input channels. - input_resolution (tuple[int]): Input resolution. - depth (int): Number of blocks. - num_heads (int): Number of attention heads. - window_size (int): Local window size. - mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. - qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True - drop (float, optional): Dropout rate. Default: 0.0 - attn_drop (float, optional): Attention dropout rate. Default: 0.0 - drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 - norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm - downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None - use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. - img_size: Input image size. - patch_size: Patch size. - resi_connection: The convolutional block before residual connection. - """ - - def __init__(self, dim, input_resolution, depth, num_heads, window_size, - mlp_ratio=4., qkv_bias=True, drop=0., attn_drop=0., - drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False, - img_size=224, patch_size=4, resi_connection='1conv'): - super(RSTB, self).__init__() - - self.dim = dim - self.input_resolution = input_resolution - - self.residual_group = BasicLayer(dim=dim, - input_resolution=input_resolution, - depth=depth, - num_heads=num_heads, - window_size=window_size, - mlp_ratio=mlp_ratio, - qkv_bias=qkv_bias, - drop=drop, attn_drop=attn_drop, - drop_path=drop_path, - norm_layer=norm_layer, - downsample=downsample, - use_checkpoint=use_checkpoint) - - if resi_connection == '1conv': - self.conv = nn.Conv2d(dim, dim, 3, 1, 1) - elif resi_connection == '3conv': - # to save parameters and memory - self.conv = nn.Sequential(nn.Conv2d(dim, dim // 4, 3, 1, 1), nn.LeakyReLU(negative_slope=0.2, inplace=True), - nn.Conv2d(dim // 4, dim // 4, 1, 1, 0), - nn.LeakyReLU(negative_slope=0.2, inplace=True), - nn.Conv2d(dim // 4, dim, 3, 1, 1)) - - self.patch_embed = PatchEmbed( - img_size=img_size, patch_size=patch_size, in_chans=dim, embed_dim=dim, - norm_layer=None) - - self.patch_unembed = PatchUnEmbed( - img_size=img_size, patch_size=patch_size, in_chans=dim, embed_dim=dim, - norm_layer=None) - - def forward(self, x, x_size): - return self.patch_embed(self.conv(self.patch_unembed(self.residual_group(x, x_size), x_size))) + x - - def flops(self): - flops = 0 - flops += self.residual_group.flops() - H, W = self.input_resolution - flops += H * W * self.dim * self.dim * 9 - flops += self.patch_embed.flops() - flops += self.patch_unembed.flops() - - return flops - -class PatchUnEmbed(nn.Module): - r""" Image to Patch Unembedding - - Args: - img_size (int): Image size. Default: 224. - patch_size (int): Patch token size. Default: 4. - in_chans (int): Number of input image channels. Default: 3. - embed_dim (int): Number of linear projection output channels. Default: 96. - norm_layer (nn.Module, optional): Normalization layer. Default: None - """ - - def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None): - super().__init__() - img_size = to_2tuple(img_size) - patch_size = to_2tuple(patch_size) - patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]] - self.img_size = img_size - self.patch_size = patch_size - self.patches_resolution = patches_resolution - self.num_patches = patches_resolution[0] * patches_resolution[1] - - self.in_chans = in_chans - self.embed_dim = embed_dim - - def forward(self, x, x_size): - B, HW, C = x.shape - x = x.transpose(1, 2).view(B, self.embed_dim, x_size[0], x_size[1]) # B Ph*Pw C - return x - - def flops(self): - flops = 0 - return flops - - -class Upsample(nn.Sequential): - """Upsample module. - - Args: - scale (int): Scale factor. Supported scales: 2^n and 3. - num_feat (int): Channel number of intermediate features. - """ - - def __init__(self, scale, num_feat): - m = [] - if (scale & (scale - 1)) == 0: # scale = 2^n - for _ in range(int(math.log(scale, 2))): - m.append(nn.Conv2d(num_feat, 4 * num_feat, 3, 1, 1)) - m.append(nn.PixelShuffle(2)) - elif scale == 3: - m.append(nn.Conv2d(num_feat, 9 * num_feat, 3, 1, 1)) - m.append(nn.PixelShuffle(3)) - else: - raise ValueError(f'scale {scale} is not supported. ' 'Supported scales: 2^n and 3.') - super(Upsample, self).__init__(*m) - -class Upsample_hf(nn.Sequential): - """Upsample module. - - Args: - scale (int): Scale factor. Supported scales: 2^n and 3. - num_feat (int): Channel number of intermediate features. - """ - - def __init__(self, scale, num_feat): - m = [] - if (scale & (scale - 1)) == 0: # scale = 2^n - for _ in range(int(math.log(scale, 2))): - m.append(nn.Conv2d(num_feat, 4 * num_feat, 3, 1, 1)) - m.append(nn.PixelShuffle(2)) - elif scale == 3: - m.append(nn.Conv2d(num_feat, 9 * num_feat, 3, 1, 1)) - m.append(nn.PixelShuffle(3)) - else: - raise ValueError(f'scale {scale} is not supported. ' 'Supported scales: 2^n and 3.') - super(Upsample_hf, self).__init__(*m) - - -class UpsampleOneStep(nn.Sequential): - """UpsampleOneStep module (the difference with Upsample is that it always only has 1conv + 1pixelshuffle) - Used in lightweight SR to save parameters. - - Args: - scale (int): Scale factor. Supported scales: 2^n and 3. - num_feat (int): Channel number of intermediate features. - - """ - - def __init__(self, scale, num_feat, num_out_ch, input_resolution=None): - self.num_feat = num_feat - self.input_resolution = input_resolution - m = [] - m.append(nn.Conv2d(num_feat, (scale ** 2) * num_out_ch, 3, 1, 1)) - m.append(nn.PixelShuffle(scale)) - super(UpsampleOneStep, self).__init__(*m) - - def flops(self): - H, W = self.input_resolution - flops = H * W * self.num_feat * 3 * 9 - return flops - - - -class Swin2SR(nn.Module): - r""" Swin2SR - A PyTorch impl of : `Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration`. - - Args: - img_size (int | tuple(int)): Input image size. Default 64 - patch_size (int | tuple(int)): Patch size. Default: 1 - in_chans (int): Number of input image channels. Default: 3 - embed_dim (int): Patch embedding dimension. Default: 96 - depths (tuple(int)): Depth of each Swin Transformer layer. - num_heads (tuple(int)): Number of attention heads in different layers. - window_size (int): Window size. Default: 7 - mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4 - qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True - drop_rate (float): Dropout rate. Default: 0 - attn_drop_rate (float): Attention dropout rate. Default: 0 - drop_path_rate (float): Stochastic depth rate. Default: 0.1 - norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm. - ape (bool): If True, add absolute position embedding to the patch embedding. Default: False - patch_norm (bool): If True, add normalization after patch embedding. Default: True - use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False - upscale: Upscale factor. 2/3/4/8 for image SR, 1 for denoising and compress artifact reduction - img_range: Image range. 1. or 255. - upsampler: The reconstruction reconstruction module. 'pixelshuffle'/'pixelshuffledirect'/'nearest+conv'/None - resi_connection: The convolutional block before residual connection. '1conv'/'3conv' - """ - - def __init__(self, img_size=64, patch_size=1, in_chans=3, - embed_dim=96, depths=(6, 6, 6, 6), num_heads=(6, 6, 6, 6), - window_size=7, mlp_ratio=4., qkv_bias=True, - drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1, - norm_layer=nn.LayerNorm, ape=False, patch_norm=True, - use_checkpoint=False, upscale=2, img_range=1., upsampler='', resi_connection='1conv', - **kwargs): - super(Swin2SR, self).__init__() - num_in_ch = in_chans - num_out_ch = in_chans - num_feat = 64 - self.img_range = img_range - if in_chans == 3: - rgb_mean = (0.4488, 0.4371, 0.4040) - self.mean = torch.Tensor(rgb_mean).view(1, 3, 1, 1) - else: - self.mean = torch.zeros(1, 1, 1, 1) - self.upscale = upscale - self.upsampler = upsampler - self.window_size = window_size - - ##################################################################################################### - ################################### 1, shallow feature extraction ################################### - self.conv_first = nn.Conv2d(num_in_ch, embed_dim, 3, 1, 1) - - ##################################################################################################### - ################################### 2, deep feature extraction ###################################### - self.num_layers = len(depths) - self.embed_dim = embed_dim - self.ape = ape - self.patch_norm = patch_norm - self.num_features = embed_dim - self.mlp_ratio = mlp_ratio - - # split image into non-overlapping patches - self.patch_embed = PatchEmbed( - img_size=img_size, patch_size=patch_size, in_chans=embed_dim, embed_dim=embed_dim, - norm_layer=norm_layer if self.patch_norm else None) - num_patches = self.patch_embed.num_patches - patches_resolution = self.patch_embed.patches_resolution - self.patches_resolution = patches_resolution - - # merge non-overlapping patches into image - self.patch_unembed = PatchUnEmbed( - img_size=img_size, patch_size=patch_size, in_chans=embed_dim, embed_dim=embed_dim, - norm_layer=norm_layer if self.patch_norm else None) - - # absolute position embedding - if self.ape: - self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim)) - trunc_normal_(self.absolute_pos_embed, std=.02) - - self.pos_drop = nn.Dropout(p=drop_rate) - - # stochastic depth - dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))] # stochastic depth decay rule - - # build Residual Swin Transformer blocks (RSTB) - self.layers = nn.ModuleList() - for i_layer in range(self.num_layers): - layer = RSTB(dim=embed_dim, - input_resolution=(patches_resolution[0], - patches_resolution[1]), - depth=depths[i_layer], - num_heads=num_heads[i_layer], - window_size=window_size, - mlp_ratio=self.mlp_ratio, - qkv_bias=qkv_bias, - drop=drop_rate, attn_drop=attn_drop_rate, - drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])], # no impact on SR results - norm_layer=norm_layer, - downsample=None, - use_checkpoint=use_checkpoint, - img_size=img_size, - patch_size=patch_size, - resi_connection=resi_connection - - ) - self.layers.append(layer) - - if self.upsampler == 'pixelshuffle_hf': - self.layers_hf = nn.ModuleList() - for i_layer in range(self.num_layers): - layer = RSTB(dim=embed_dim, - input_resolution=(patches_resolution[0], - patches_resolution[1]), - depth=depths[i_layer], - num_heads=num_heads[i_layer], - window_size=window_size, - mlp_ratio=self.mlp_ratio, - qkv_bias=qkv_bias, - drop=drop_rate, attn_drop=attn_drop_rate, - drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])], # no impact on SR results - norm_layer=norm_layer, - downsample=None, - use_checkpoint=use_checkpoint, - img_size=img_size, - patch_size=patch_size, - resi_connection=resi_connection - - ) - self.layers_hf.append(layer) - - self.norm = norm_layer(self.num_features) - - # build the last conv layer in deep feature extraction - if resi_connection == '1conv': - self.conv_after_body = nn.Conv2d(embed_dim, embed_dim, 3, 1, 1) - elif resi_connection == '3conv': - # to save parameters and memory - self.conv_after_body = nn.Sequential(nn.Conv2d(embed_dim, embed_dim // 4, 3, 1, 1), - nn.LeakyReLU(negative_slope=0.2, inplace=True), - nn.Conv2d(embed_dim // 4, embed_dim // 4, 1, 1, 0), - nn.LeakyReLU(negative_slope=0.2, inplace=True), - nn.Conv2d(embed_dim // 4, embed_dim, 3, 1, 1)) - - ##################################################################################################### - ################################ 3, high quality image reconstruction ################################ - if self.upsampler == 'pixelshuffle': - # for classical SR - self.conv_before_upsample = nn.Sequential(nn.Conv2d(embed_dim, num_feat, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.upsample = Upsample(upscale, num_feat) - self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) - elif self.upsampler == 'pixelshuffle_aux': - self.conv_bicubic = nn.Conv2d(num_in_ch, num_feat, 3, 1, 1) - self.conv_before_upsample = nn.Sequential( - nn.Conv2d(embed_dim, num_feat, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.conv_aux = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) - self.conv_after_aux = nn.Sequential( - nn.Conv2d(3, num_feat, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.upsample = Upsample(upscale, num_feat) - self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) - - elif self.upsampler == 'pixelshuffle_hf': - self.conv_before_upsample = nn.Sequential(nn.Conv2d(embed_dim, num_feat, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.upsample = Upsample(upscale, num_feat) - self.upsample_hf = Upsample_hf(upscale, num_feat) - self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) - self.conv_first_hf = nn.Sequential(nn.Conv2d(num_feat, embed_dim, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.conv_after_body_hf = nn.Conv2d(embed_dim, embed_dim, 3, 1, 1) - self.conv_before_upsample_hf = nn.Sequential( - nn.Conv2d(embed_dim, num_feat, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.conv_last_hf = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) - - elif self.upsampler == 'pixelshuffledirect': - # for lightweight SR (to save parameters) - self.upsample = UpsampleOneStep(upscale, embed_dim, num_out_ch, - (patches_resolution[0], patches_resolution[1])) - elif self.upsampler == 'nearest+conv': - # for real-world SR (less artifacts) - assert self.upscale == 4, 'only support x4 now.' - self.conv_before_upsample = nn.Sequential(nn.Conv2d(embed_dim, num_feat, 3, 1, 1), - nn.LeakyReLU(inplace=True)) - self.conv_up1 = nn.Conv2d(num_feat, num_feat, 3, 1, 1) - self.conv_up2 = nn.Conv2d(num_feat, num_feat, 3, 1, 1) - self.conv_hr = nn.Conv2d(num_feat, num_feat, 3, 1, 1) - self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) - self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) - else: - # for image denoising and JPEG compression artifact reduction - self.conv_last = nn.Conv2d(embed_dim, num_out_ch, 3, 1, 1) - - self.apply(self._init_weights) - - def _init_weights(self, m): - if isinstance(m, nn.Linear): - trunc_normal_(m.weight, std=.02) - if isinstance(m, nn.Linear) and m.bias is not None: - nn.init.constant_(m.bias, 0) - elif isinstance(m, nn.LayerNorm): - nn.init.constant_(m.bias, 0) - nn.init.constant_(m.weight, 1.0) - - @torch.jit.ignore - def no_weight_decay(self): - return {'absolute_pos_embed'} - - @torch.jit.ignore - def no_weight_decay_keywords(self): - return {'relative_position_bias_table'} - - def check_image_size(self, x): - _, _, h, w = x.size() - mod_pad_h = (self.window_size - h % self.window_size) % self.window_size - mod_pad_w = (self.window_size - w % self.window_size) % self.window_size - x = F.pad(x, (0, mod_pad_w, 0, mod_pad_h), 'reflect') - return x - - def forward_features(self, x): - x_size = (x.shape[2], x.shape[3]) - x = self.patch_embed(x) - if self.ape: - x = x + self.absolute_pos_embed - x = self.pos_drop(x) - - for layer in self.layers: - x = layer(x, x_size) - - x = self.norm(x) # B L C - x = self.patch_unembed(x, x_size) - - return x - - def forward_features_hf(self, x): - x_size = (x.shape[2], x.shape[3]) - x = self.patch_embed(x) - if self.ape: - x = x + self.absolute_pos_embed - x = self.pos_drop(x) - - for layer in self.layers_hf: - x = layer(x, x_size) - - x = self.norm(x) # B L C - x = self.patch_unembed(x, x_size) - - return x - - def forward(self, x): - H, W = x.shape[2:] - x = self.check_image_size(x) - - self.mean = self.mean.type_as(x) - x = (x - self.mean) * self.img_range - - if self.upsampler == 'pixelshuffle': - # for classical SR - x = self.conv_first(x) - x = self.conv_after_body(self.forward_features(x)) + x - x = self.conv_before_upsample(x) - x = self.conv_last(self.upsample(x)) - elif self.upsampler == 'pixelshuffle_aux': - bicubic = F.interpolate(x, size=(H * self.upscale, W * self.upscale), mode='bicubic', align_corners=False) - bicubic = self.conv_bicubic(bicubic) - x = self.conv_first(x) - x = self.conv_after_body(self.forward_features(x)) + x - x = self.conv_before_upsample(x) - aux = self.conv_aux(x) # b, 3, LR_H, LR_W - x = self.conv_after_aux(aux) - x = self.upsample(x)[:, :, :H * self.upscale, :W * self.upscale] + bicubic[:, :, :H * self.upscale, :W * self.upscale] - x = self.conv_last(x) - aux = aux / self.img_range + self.mean - elif self.upsampler == 'pixelshuffle_hf': - # for classical SR with HF - x = self.conv_first(x) - x = self.conv_after_body(self.forward_features(x)) + x - x_before = self.conv_before_upsample(x) - x_out = self.conv_last(self.upsample(x_before)) - - x_hf = self.conv_first_hf(x_before) - x_hf = self.conv_after_body_hf(self.forward_features_hf(x_hf)) + x_hf - x_hf = self.conv_before_upsample_hf(x_hf) - x_hf = self.conv_last_hf(self.upsample_hf(x_hf)) - x = x_out + x_hf - x_hf = x_hf / self.img_range + self.mean - - elif self.upsampler == 'pixelshuffledirect': - # for lightweight SR - x = self.conv_first(x) - x = self.conv_after_body(self.forward_features(x)) + x - x = self.upsample(x) - elif self.upsampler == 'nearest+conv': - # for real-world SR - x = self.conv_first(x) - x = self.conv_after_body(self.forward_features(x)) + x - x = self.conv_before_upsample(x) - x = self.lrelu(self.conv_up1(torch.nn.functional.interpolate(x, scale_factor=2, mode='nearest'))) - x = self.lrelu(self.conv_up2(torch.nn.functional.interpolate(x, scale_factor=2, mode='nearest'))) - x = self.conv_last(self.lrelu(self.conv_hr(x))) - else: - # for image denoising and JPEG compression artifact reduction - x_first = self.conv_first(x) - res = self.conv_after_body(self.forward_features(x_first)) + x_first - x = x + self.conv_last(res) - - x = x / self.img_range + self.mean - if self.upsampler == "pixelshuffle_aux": - return x[:, :, :H*self.upscale, :W*self.upscale], aux - - elif self.upsampler == "pixelshuffle_hf": - x_out = x_out / self.img_range + self.mean - return x_out[:, :, :H*self.upscale, :W*self.upscale], x[:, :, :H*self.upscale, :W*self.upscale], x_hf[:, :, :H*self.upscale, :W*self.upscale] - - else: - return x[:, :, :H*self.upscale, :W*self.upscale] - - def flops(self): - flops = 0 - H, W = self.patches_resolution - flops += H * W * 3 * self.embed_dim * 9 - flops += self.patch_embed.flops() - for layer in self.layers: - flops += layer.flops() - flops += H * W * 3 * self.embed_dim * self.embed_dim - flops += self.upsample.flops() - return flops - - -if __name__ == '__main__': - upscale = 4 - window_size = 8 - height = (1024 // upscale // window_size + 1) * window_size - width = (720 // upscale // window_size + 1) * window_size - model = Swin2SR(upscale=2, img_size=(height, width), - window_size=window_size, img_range=1., depths=[6, 6, 6, 6], - embed_dim=60, num_heads=[6, 6, 6, 6], mlp_ratio=2, upsampler='pixelshuffledirect') - print(model) - print(height, width, model.flops() / 1e9) - - x = torch.randn((1, 3, height, width)) - x = model(x) - print(x.shape) diff --git a/extensions-builtin/canvas-zoom-and-pan/javascript/zoom.js b/extensions-builtin/canvas-zoom-and-pan/javascript/zoom.js index 45c7600ac5f..64e7a638a4c 100644 --- a/extensions-builtin/canvas-zoom-and-pan/javascript/zoom.js +++ b/extensions-builtin/canvas-zoom-and-pan/javascript/zoom.js @@ -218,6 +218,8 @@ onUiLoaded(async() => { canvas_hotkey_fullscreen: "KeyS", canvas_hotkey_move: "KeyF", canvas_hotkey_overlap: "KeyO", + canvas_hotkey_shrink_brush: "KeyQ", + canvas_hotkey_grow_brush: "KeyW", canvas_disabled_functions: [], canvas_show_tooltip: true, canvas_auto_expand: true, @@ -227,6 +229,8 @@ onUiLoaded(async() => { const functionMap = { "Zoom": "canvas_hotkey_zoom", "Adjust brush size": "canvas_hotkey_adjust", + "Hotkey shrink brush": "canvas_hotkey_shrink_brush", + "Hotkey enlarge brush": "canvas_hotkey_grow_brush", "Moving canvas": "canvas_hotkey_move", "Fullscreen": "canvas_hotkey_fullscreen", "Reset Zoom": "canvas_hotkey_reset", @@ -288,7 +292,7 @@ onUiLoaded(async() => { // Create tooltip function createTooltip() { - const toolTipElemnt = + const toolTipElement = targetElement.querySelector(".image-container"); const tooltip = document.createElement("div"); tooltip.className = "canvas-tooltip"; @@ -351,7 +355,7 @@ onUiLoaded(async() => { tooltip.appendChild(tooltipContent); // Add a hint element to the target element - toolTipElemnt.appendChild(tooltip); + toolTipElement.appendChild(tooltip); } //Show tool tip if setting enable @@ -686,7 +690,9 @@ onUiLoaded(async() => { const hotkeyActions = { [hotkeysConfig.canvas_hotkey_reset]: resetZoom, [hotkeysConfig.canvas_hotkey_overlap]: toggleOverlap, - [hotkeysConfig.canvas_hotkey_fullscreen]: fitToScreen + [hotkeysConfig.canvas_hotkey_fullscreen]: fitToScreen, + [hotkeysConfig.canvas_hotkey_shrink_brush]: () => adjustBrushSize(elemId, 10), + [hotkeysConfig.canvas_hotkey_grow_brush]: () => adjustBrushSize(elemId, -10) }; const action = hotkeyActions[event.code]; diff --git a/extensions-builtin/canvas-zoom-and-pan/scripts/hotkey_config.py b/extensions-builtin/canvas-zoom-and-pan/scripts/hotkey_config.py index 2d8d2d1c014..17b27b2741b 100644 --- a/extensions-builtin/canvas-zoom-and-pan/scripts/hotkey_config.py +++ b/extensions-builtin/canvas-zoom-and-pan/scripts/hotkey_config.py @@ -4,12 +4,14 @@ shared.options_templates.update(shared.options_section(('canvas_hotkey', "Canvas Hotkeys"), { "canvas_hotkey_zoom": shared.OptionInfo("Alt", "Zoom canvas", gr.Radio, {"choices": ["Shift","Ctrl", "Alt"]}).info("If you choose 'Shift' you cannot scroll horizontally, 'Alt' can cause a little trouble in firefox"), "canvas_hotkey_adjust": shared.OptionInfo("Ctrl", "Adjust brush size", gr.Radio, {"choices": ["Shift","Ctrl", "Alt"]}).info("If you choose 'Shift' you cannot scroll horizontally, 'Alt' can cause a little trouble in firefox"), + "canvas_hotkey_shrink_brush": shared.OptionInfo("Q", "Shrink the brush size"), + "canvas_hotkey_grow_brush": shared.OptionInfo("W", "Enlarge the brush size"), "canvas_hotkey_move": shared.OptionInfo("F", "Moving the canvas").info("To work correctly in firefox, turn off 'Automatically search the page text when typing' in the browser settings"), "canvas_hotkey_fullscreen": shared.OptionInfo("S", "Fullscreen Mode, maximizes the picture so that it fits into the screen and stretches it to its full width "), - "canvas_hotkey_reset": shared.OptionInfo("R", "Reset zoom and canvas positon"), - "canvas_hotkey_overlap": shared.OptionInfo("O", "Toggle overlap").info("Technical button, neededs for testing"), + "canvas_hotkey_reset": shared.OptionInfo("R", "Reset zoom and canvas position"), + "canvas_hotkey_overlap": shared.OptionInfo("O", "Toggle overlap").info("Technical button, needed for testing"), "canvas_show_tooltip": shared.OptionInfo(True, "Enable tooltip on the canvas"), "canvas_auto_expand": shared.OptionInfo(True, "Automatically expands an image that does not fit completely in the canvas area, similar to manually pressing the S and R buttons"), "canvas_blur_prompt": shared.OptionInfo(False, "Take the focus off the prompt when working with a canvas"), - "canvas_disabled_functions": shared.OptionInfo(["Overlap"], "Disable function that you don't use", gr.CheckboxGroup, {"choices": ["Zoom","Adjust brush size", "Moving canvas","Fullscreen","Reset Zoom","Overlap"]}), + "canvas_disabled_functions": shared.OptionInfo(["Overlap"], "Disable function that you don't use", gr.CheckboxGroup, {"choices": ["Zoom","Adjust brush size","Hotkey enlarge brush","Hotkey shrink brush","Moving canvas","Fullscreen","Reset Zoom","Overlap"]}), })) diff --git a/extensions-builtin/extra-options-section/scripts/extra_options_section.py b/extensions-builtin/extra-options-section/scripts/extra_options_section.py index ac2c3de4643..a91bea4fa9d 100644 --- a/extensions-builtin/extra-options-section/scripts/extra_options_section.py +++ b/extensions-builtin/extra-options-section/scripts/extra_options_section.py @@ -1,7 +1,7 @@ import math import gradio as gr -from modules import scripts, shared, ui_components, ui_settings, generation_parameters_copypaste +from modules import scripts, shared, ui_components, ui_settings, infotext_utils, errors from modules.ui_components import FormColumn @@ -25,7 +25,7 @@ def ui(self, is_img2img): extra_options = shared.opts.extra_options_img2img if is_img2img else shared.opts.extra_options_txt2img elem_id_tabname = "extra_options_" + ("img2img" if is_img2img else "txt2img") - mapping = {k: v for v, k in generation_parameters_copypaste.infotext_to_setting_name_mapping} + mapping = {k: v for v, k in infotext_utils.infotext_to_setting_name_mapping} with gr.Blocks() as interface: with gr.Accordion("Options", open=False, elem_id=elem_id_tabname) if shared.opts.extra_options_accordion and extra_options else gr.Group(elem_id=elem_id_tabname): @@ -42,7 +42,11 @@ def ui(self, is_img2img): setting_name = extra_options[index] with FormColumn(): - comp = ui_settings.create_setting_component(setting_name) + try: + comp = ui_settings.create_setting_component(setting_name) + except KeyError: + errors.report(f"Can't add extra options for {setting_name} in ui") + continue self.comps.append(comp) self.setting_names.append(setting_name) diff --git a/extensions-builtin/soft-inpainting/scripts/soft_inpainting.py b/extensions-builtin/soft-inpainting/scripts/soft_inpainting.py new file mode 100644 index 00000000000..f56e1e2266d --- /dev/null +++ b/extensions-builtin/soft-inpainting/scripts/soft_inpainting.py @@ -0,0 +1,761 @@ +import numpy as np +import gradio as gr +import math +from modules.ui_components import InputAccordion +import modules.scripts as scripts + + +class SoftInpaintingSettings: + def __init__(self, + mask_blend_power, + mask_blend_scale, + inpaint_detail_preservation, + composite_mask_influence, + composite_difference_threshold, + composite_difference_contrast): + self.mask_blend_power = mask_blend_power + self.mask_blend_scale = mask_blend_scale + self.inpaint_detail_preservation = inpaint_detail_preservation + self.composite_mask_influence = composite_mask_influence + self.composite_difference_threshold = composite_difference_threshold + self.composite_difference_contrast = composite_difference_contrast + + def add_generation_params(self, dest): + dest[enabled_gen_param_label] = True + dest[gen_param_labels.mask_blend_power] = self.mask_blend_power + dest[gen_param_labels.mask_blend_scale] = self.mask_blend_scale + dest[gen_param_labels.inpaint_detail_preservation] = self.inpaint_detail_preservation + dest[gen_param_labels.composite_mask_influence] = self.composite_mask_influence + dest[gen_param_labels.composite_difference_threshold] = self.composite_difference_threshold + dest[gen_param_labels.composite_difference_contrast] = self.composite_difference_contrast + + +# ------------------- Methods ------------------- + +def processing_uses_inpainting(p): + # TODO: Figure out a better way to determine if inpainting is being used by p + if getattr(p, "image_mask", None) is not None: + return True + + if getattr(p, "mask", None) is not None: + return True + + if getattr(p, "nmask", None) is not None: + return True + + return False + + +def latent_blend(settings, a, b, t): + """ + Interpolates two latent image representations according to the parameter t, + where the interpolated vectors' magnitudes are also interpolated separately. + The "detail_preservation" factor biases the magnitude interpolation towards + the larger of the two magnitudes. + """ + import torch + + # NOTE: We use inplace operations wherever possible. + + if len(t.shape) == 3: + # [4][w][h] to [1][4][w][h] + t2 = t.unsqueeze(0) + # [4][w][h] to [1][1][w][h] - the [4] seem redundant. + t3 = t[0].unsqueeze(0).unsqueeze(0) + else: + t2 = t + t3 = t[:, 0][:, None] + + one_minus_t2 = 1 - t2 + one_minus_t3 = 1 - t3 + + # Linearly interpolate the image vectors. + a_scaled = a * one_minus_t2 + b_scaled = b * t2 + image_interp = a_scaled + image_interp.add_(b_scaled) + result_type = image_interp.dtype + del a_scaled, b_scaled, t2, one_minus_t2 + + # Calculate the magnitude of the interpolated vectors. (We will remove this magnitude.) + # 64-bit operations are used here to allow large exponents. + current_magnitude = torch.norm(image_interp, p=2, dim=1, keepdim=True).to(torch.float64).add_(0.00001) + + # Interpolate the powered magnitudes, then un-power them (bring them back to a power of 1). + a_magnitude = torch.norm(a, p=2, dim=1, keepdim=True).to(torch.float64).pow_( + settings.inpaint_detail_preservation) * one_minus_t3 + b_magnitude = torch.norm(b, p=2, dim=1, keepdim=True).to(torch.float64).pow_( + settings.inpaint_detail_preservation) * t3 + desired_magnitude = a_magnitude + desired_magnitude.add_(b_magnitude).pow_(1 / settings.inpaint_detail_preservation) + del a_magnitude, b_magnitude, t3, one_minus_t3 + + # Change the linearly interpolated image vectors' magnitudes to the value we want. + # This is the last 64-bit operation. + image_interp_scaling_factor = desired_magnitude + image_interp_scaling_factor.div_(current_magnitude) + image_interp_scaling_factor = image_interp_scaling_factor.to(result_type) + image_interp_scaled = image_interp + image_interp_scaled.mul_(image_interp_scaling_factor) + del current_magnitude + del desired_magnitude + del image_interp + del image_interp_scaling_factor + del result_type + + return image_interp_scaled + + +def get_modified_nmask(settings, nmask, sigma): + """ + Converts a negative mask representing the transparency of the original latent vectors being overlaid + to a mask that is scaled according to the denoising strength for this step. + + Where: + 0 = fully opaque, infinite density, fully masked + 1 = fully transparent, zero density, fully unmasked + + We bring this transparency to a power, as this allows one to simulate N number of blending operations + where N can be any positive real value. Using this one can control the balance of influence between + the denoiser and the original latents according to the sigma value. + + NOTE: "mask" is not used + """ + import torch + return torch.pow(nmask, (sigma ** settings.mask_blend_power) * settings.mask_blend_scale) + + +def apply_adaptive_masks( + settings: SoftInpaintingSettings, + nmask, + latent_orig, + latent_processed, + overlay_images, + width, height, + paste_to): + import torch + import modules.processing as proc + import modules.images as images + from PIL import Image, ImageOps, ImageFilter + + # TODO: Bias the blending according to the latent mask, add adjustable parameter for bias control. + if len(nmask.shape) == 3: + latent_mask = nmask[0].float() + else: + latent_mask = nmask[:, 0].float() + # convert the original mask into a form we use to scale distances for thresholding + mask_scalar = 1 - (torch.clamp(latent_mask, min=0, max=1) ** (settings.mask_blend_scale / 2)) + mask_scalar = (0.5 * (1 - settings.composite_mask_influence) + + mask_scalar * settings.composite_mask_influence) + mask_scalar = mask_scalar / (1.00001 - mask_scalar) + mask_scalar = mask_scalar.cpu().numpy() + + latent_distance = torch.norm(latent_processed - latent_orig, p=2, dim=1) + + kernel, kernel_center = get_gaussian_kernel(stddev_radius=1.5, max_radius=2) + + masks_for_overlay = [] + + for i, (distance_map, overlay_image) in enumerate(zip(latent_distance, overlay_images)): + converted_mask = distance_map.float().cpu().numpy() + converted_mask = weighted_histogram_filter(converted_mask, kernel, kernel_center, + percentile_min=0.9, percentile_max=1, min_width=1) + converted_mask = weighted_histogram_filter(converted_mask, kernel, kernel_center, + percentile_min=0.25, percentile_max=0.75, min_width=1) + + # The distance at which opacity of original decreases to 50% + if len(mask_scalar.shape) == 3: + if mask_scalar.shape[0] > i: + half_weighted_distance = settings.composite_difference_threshold * mask_scalar[i] + else: + half_weighted_distance = settings.composite_difference_threshold * mask_scalar[0] + else: + half_weighted_distance = settings.composite_difference_threshold * mask_scalar + + converted_mask = converted_mask / half_weighted_distance + + converted_mask = 1 / (1 + converted_mask ** settings.composite_difference_contrast) + converted_mask = smootherstep(converted_mask) + converted_mask = 1 - converted_mask + converted_mask = 255. * converted_mask + converted_mask = converted_mask.astype(np.uint8) + converted_mask = Image.fromarray(converted_mask) + converted_mask = images.resize_image(2, converted_mask, width, height) + converted_mask = proc.create_binary_mask(converted_mask, round=False) + + # Remove aliasing artifacts using a gaussian blur. + converted_mask = converted_mask.filter(ImageFilter.GaussianBlur(radius=4)) + + # Expand the mask to fit the whole image if needed. + if paste_to is not None: + converted_mask = proc.uncrop(converted_mask, + (overlay_image.width, overlay_image.height), + paste_to) + + masks_for_overlay.append(converted_mask) + + image_masked = Image.new('RGBa', (overlay_image.width, overlay_image.height)) + image_masked.paste(overlay_image.convert("RGBA").convert("RGBa"), + mask=ImageOps.invert(converted_mask.convert('L'))) + + overlay_images[i] = image_masked.convert('RGBA') + + return masks_for_overlay + + +def apply_masks( + settings, + nmask, + overlay_images, + width, height, + paste_to): + import torch + import modules.processing as proc + import modules.images as images + from PIL import Image, ImageOps, ImageFilter + + converted_mask = nmask[0].float() + converted_mask = torch.clamp(converted_mask, min=0, max=1).pow_(settings.mask_blend_scale / 2) + converted_mask = 255. * converted_mask + converted_mask = converted_mask.cpu().numpy().astype(np.uint8) + converted_mask = Image.fromarray(converted_mask) + converted_mask = images.resize_image(2, converted_mask, width, height) + converted_mask = proc.create_binary_mask(converted_mask, round=False) + + # Remove aliasing artifacts using a gaussian blur. + converted_mask = converted_mask.filter(ImageFilter.GaussianBlur(radius=4)) + + # Expand the mask to fit the whole image if needed. + if paste_to is not None: + converted_mask = proc.uncrop(converted_mask, + (width, height), + paste_to) + + masks_for_overlay = [] + + for i, overlay_image in enumerate(overlay_images): + masks_for_overlay[i] = converted_mask + + image_masked = Image.new('RGBa', (overlay_image.width, overlay_image.height)) + image_masked.paste(overlay_image.convert("RGBA").convert("RGBa"), + mask=ImageOps.invert(converted_mask.convert('L'))) + + overlay_images[i] = image_masked.convert('RGBA') + + return masks_for_overlay + + +def weighted_histogram_filter(img, kernel, kernel_center, percentile_min=0.0, percentile_max=1.0, min_width=1.0): + """ + Generalization convolution filter capable of applying + weighted mean, median, maximum, and minimum filters + parametrically using an arbitrary kernel. + + Args: + img (nparray): + The image, a 2-D array of floats, to which the filter is being applied. + kernel (nparray): + The kernel, a 2-D array of floats. + kernel_center (nparray): + The kernel center coordinate, a 1-D array with two elements. + percentile_min (float): + The lower bound of the histogram window used by the filter, + from 0 to 1. + percentile_max (float): + The upper bound of the histogram window used by the filter, + from 0 to 1. + min_width (float): + The minimum size of the histogram window bounds, in weight units. + Must be greater than 0. + + Returns: + (nparray): A filtered copy of the input image "img", a 2-D array of floats. + """ + + # Converts an index tuple into a vector. + def vec(x): + return np.array(x) + + kernel_min = -kernel_center + kernel_max = vec(kernel.shape) - kernel_center + + def weighted_histogram_filter_single(idx): + idx = vec(idx) + min_index = np.maximum(0, idx + kernel_min) + max_index = np.minimum(vec(img.shape), idx + kernel_max) + window_shape = max_index - min_index + + class WeightedElement: + """ + An element of the histogram, its weight + and bounds. + """ + + def __init__(self, value, weight): + self.value: float = value + self.weight: float = weight + self.window_min: float = 0.0 + self.window_max: float = 1.0 + + # Collect the values in the image as WeightedElements, + # weighted by their corresponding kernel values. + values = [] + for window_tup in np.ndindex(tuple(window_shape)): + window_index = vec(window_tup) + image_index = window_index + min_index + centered_kernel_index = image_index - idx + kernel_index = centered_kernel_index + kernel_center + element = WeightedElement(img[tuple(image_index)], kernel[tuple(kernel_index)]) + values.append(element) + + def sort_key(x: WeightedElement): + return x.value + + values.sort(key=sort_key) + + # Calculate the height of the stack (sum) + # and each sample's range they occupy in the stack + sum = 0 + for i in range(len(values)): + values[i].window_min = sum + sum += values[i].weight + values[i].window_max = sum + + # Calculate what range of this stack ("window") + # we want to get the weighted average across. + window_min = sum * percentile_min + window_max = sum * percentile_max + window_width = window_max - window_min + + # Ensure the window is within the stack and at least a certain size. + if window_width < min_width: + window_center = (window_min + window_max) / 2 + window_min = window_center - min_width / 2 + window_max = window_center + min_width / 2 + + if window_max > sum: + window_max = sum + window_min = sum - min_width + + if window_min < 0: + window_min = 0 + window_max = min_width + + value = 0 + value_weight = 0 + + # Get the weighted average of all the samples + # that overlap with the window, weighted + # by the size of their overlap. + for i in range(len(values)): + if window_min >= values[i].window_max: + continue + if window_max <= values[i].window_min: + break + + s = max(window_min, values[i].window_min) + e = min(window_max, values[i].window_max) + w = e - s + + value += values[i].value * w + value_weight += w + + return value / value_weight if value_weight != 0 else 0 + + img_out = img.copy() + + # Apply the kernel operation over each pixel. + for index in np.ndindex(img.shape): + img_out[index] = weighted_histogram_filter_single(index) + + return img_out + + +def smoothstep(x): + """ + The smoothstep function, input should be clamped to 0-1 range. + Turns a diagonal line (f(x) = x) into a sigmoid-like curve. + """ + return x * x * (3 - 2 * x) + + +def smootherstep(x): + """ + The smootherstep function, input should be clamped to 0-1 range. + Turns a diagonal line (f(x) = x) into a sigmoid-like curve. + """ + return x * x * x * (x * (6 * x - 15) + 10) + + +def get_gaussian_kernel(stddev_radius=1.0, max_radius=2): + """ + Creates a Gaussian kernel with thresholded edges. + + Args: + stddev_radius (float): + Standard deviation of the gaussian kernel, in pixels. + max_radius (int): + The size of the filter kernel. The number of pixels is (max_radius*2+1) ** 2. + The kernel is thresholded so that any values one pixel beyond this radius + is weighted at 0. + + Returns: + (nparray, nparray): A kernel array (shape: (N, N)), its center coordinate (shape: (2)) + """ + + # Evaluates a 0-1 normalized gaussian function for a given square distance from the mean. + def gaussian(sqr_mag): + return math.exp(-sqr_mag / (stddev_radius * stddev_radius)) + + # Helper function for converting a tuple to an array. + def vec(x): + return np.array(x) + + """ + Since a gaussian is unbounded, we need to limit ourselves + to a finite range. + We taper the ends off at the end of that range so they equal zero + while preserving the maximum value of 1 at the mean. + """ + zero_radius = max_radius + 1.0 + gauss_zero = gaussian(zero_radius * zero_radius) + gauss_kernel_scale = 1 / (1 - gauss_zero) + + def gaussian_kernel_func(coordinate): + x = coordinate[0] ** 2.0 + coordinate[1] ** 2.0 + x = gaussian(x) + x -= gauss_zero + x *= gauss_kernel_scale + x = max(0.0, x) + return x + + size = max_radius * 2 + 1 + kernel_center = max_radius + kernel = np.zeros((size, size)) + + for index in np.ndindex(kernel.shape): + kernel[index] = gaussian_kernel_func(vec(index) - kernel_center) + + return kernel, kernel_center + + +# ------------------- Constants ------------------- + + +default = SoftInpaintingSettings(1, 0.5, 4, 0, 0.5, 2) + +enabled_ui_label = "Soft inpainting" +enabled_gen_param_label = "Soft inpainting enabled" +enabled_el_id = "soft_inpainting_enabled" + +ui_labels = SoftInpaintingSettings( + "Schedule bias", + "Preservation strength", + "Transition contrast boost", + "Mask influence", + "Difference threshold", + "Difference contrast") + +ui_info = SoftInpaintingSettings( + "Shifts when preservation of original content occurs during denoising.", + "How strongly partially masked content should be preserved.", + "Amplifies the contrast that may be lost in partially masked regions.", + "How strongly the original mask should bias the difference threshold.", + "How much an image region can change before the original pixels are not blended in anymore.", + "How sharp the transition should be between blended and not blended.") + +gen_param_labels = SoftInpaintingSettings( + "Soft inpainting schedule bias", + "Soft inpainting preservation strength", + "Soft inpainting transition contrast boost", + "Soft inpainting mask influence", + "Soft inpainting difference threshold", + "Soft inpainting difference contrast") + +el_ids = SoftInpaintingSettings( + "mask_blend_power", + "mask_blend_scale", + "inpaint_detail_preservation", + "composite_mask_influence", + "composite_difference_threshold", + "composite_difference_contrast") + + +# ------------------- Script ------------------- + + +class Script(scripts.Script): + def __init__(self): + self.section = "inpaint" + self.masks_for_overlay = None + self.overlay_images = None + + def title(self): + return "Soft Inpainting" + + def show(self, is_img2img): + return scripts.AlwaysVisible if is_img2img else False + + def ui(self, is_img2img): + if not is_img2img: + return + + with InputAccordion(False, label=enabled_ui_label, elem_id=enabled_el_id) as soft_inpainting_enabled: + with gr.Group(): + gr.Markdown( + """ + Soft inpainting allows you to **seamlessly blend original content with inpainted content** according to the mask opacity. + **High _Mask blur_** values are recommended! + """) + + power = \ + gr.Slider(label=ui_labels.mask_blend_power, + info=ui_info.mask_blend_power, + minimum=0, + maximum=8, + step=0.1, + value=default.mask_blend_power, + elem_id=el_ids.mask_blend_power) + scale = \ + gr.Slider(label=ui_labels.mask_blend_scale, + info=ui_info.mask_blend_scale, + minimum=0, + maximum=8, + step=0.05, + value=default.mask_blend_scale, + elem_id=el_ids.mask_blend_scale) + detail = \ + gr.Slider(label=ui_labels.inpaint_detail_preservation, + info=ui_info.inpaint_detail_preservation, + minimum=1, + maximum=32, + step=0.5, + value=default.inpaint_detail_preservation, + elem_id=el_ids.inpaint_detail_preservation) + + gr.Markdown( + """ + ### Pixel Composite Settings + """) + + mask_inf = \ + gr.Slider(label=ui_labels.composite_mask_influence, + info=ui_info.composite_mask_influence, + minimum=0, + maximum=1, + step=0.05, + value=default.composite_mask_influence, + elem_id=el_ids.composite_mask_influence) + + dif_thresh = \ + gr.Slider(label=ui_labels.composite_difference_threshold, + info=ui_info.composite_difference_threshold, + minimum=0, + maximum=8, + step=0.25, + value=default.composite_difference_threshold, + elem_id=el_ids.composite_difference_threshold) + + dif_contr = \ + gr.Slider(label=ui_labels.composite_difference_contrast, + info=ui_info.composite_difference_contrast, + minimum=0, + maximum=8, + step=0.25, + value=default.composite_difference_contrast, + elem_id=el_ids.composite_difference_contrast) + + with gr.Accordion("Help", open=False): + gr.Markdown( + f""" + ### {ui_labels.mask_blend_power} + + The blending strength of original content is scaled proportionally with the decreasing noise level values at each step (sigmas). + This ensures that the influence of the denoiser and original content preservation is roughly balanced at each step. + This balance can be shifted using this parameter, controlling whether earlier or later steps have stronger preservation. + + - **Below 1**: Stronger preservation near the end (with low sigma) + - **1**: Balanced (proportional to sigma) + - **Above 1**: Stronger preservation in the beginning (with high sigma) + """) + gr.Markdown( + f""" + ### {ui_labels.mask_blend_scale} + + Skews whether partially masked image regions should be more likely to preserve the original content or favor inpainted content. + This may need to be adjusted depending on the {ui_labels.mask_blend_power}, CFG Scale, prompt and Denoising strength. + + - **Low values**: Favors generated content. + - **High values**: Favors original content. + """) + gr.Markdown( + f""" + ### {ui_labels.inpaint_detail_preservation} + + This parameter controls how the original latent vectors and denoised latent vectors are interpolated. + With higher values, the magnitude of the resulting blended vector will be closer to the maximum of the two interpolated vectors. + This can prevent the loss of contrast that occurs with linear interpolation. + + - **Low values**: Softer blending, details may fade. + - **High values**: Stronger contrast, may over-saturate colors. + """) + + gr.Markdown( + """ + ## Pixel Composite Settings + + Masks are generated based on how much a part of the image changed after denoising. + These masks are used to blend the original and final images together. + If the difference is low, the original pixels are used instead of the pixels returned by the inpainting process. + """) + + gr.Markdown( + f""" + ### {ui_labels.composite_mask_influence} + + This parameter controls how much the mask should bias this sensitivity to difference. + + - **0**: Ignore the mask, only consider differences in image content. + - **1**: Follow the mask closely despite image content changes. + """) + + gr.Markdown( + f""" + ### {ui_labels.composite_difference_threshold} + + This value represents the difference at which the original pixels will have less than 50% opacity. + + - **Low values**: Two images patches must be almost the same in order to retain original pixels. + - **High values**: Two images patches can be very different and still retain original pixels. + """) + + gr.Markdown( + f""" + ### {ui_labels.composite_difference_contrast} + + This value represents the contrast between the opacity of the original and inpainted content. + + - **Low values**: The blend will be more gradual and have longer transitions, but may cause ghosting. + - **High values**: Ghosting will be less common, but transitions may be very sudden. + """) + + self.infotext_fields = [(soft_inpainting_enabled, enabled_gen_param_label), + (power, gen_param_labels.mask_blend_power), + (scale, gen_param_labels.mask_blend_scale), + (detail, gen_param_labels.inpaint_detail_preservation), + (mask_inf, gen_param_labels.composite_mask_influence), + (dif_thresh, gen_param_labels.composite_difference_threshold), + (dif_contr, gen_param_labels.composite_difference_contrast)] + + self.paste_field_names = [] + for _, field_name in self.infotext_fields: + self.paste_field_names.append(field_name) + + return [soft_inpainting_enabled, + power, + scale, + detail, + mask_inf, + dif_thresh, + dif_contr] + + def process(self, p, enabled, power, scale, detail_preservation, mask_inf, dif_thresh, dif_contr): + if not enabled: + return + + if not processing_uses_inpainting(p): + return + + # Shut off the rounding it normally does. + p.mask_round = False + + settings = SoftInpaintingSettings(power, scale, detail_preservation, mask_inf, dif_thresh, dif_contr) + + # p.extra_generation_params["Mask rounding"] = False + settings.add_generation_params(p.extra_generation_params) + + def on_mask_blend(self, p, mba: scripts.MaskBlendArgs, enabled, power, scale, detail_preservation, mask_inf, + dif_thresh, dif_contr): + if not enabled: + return + + if not processing_uses_inpainting(p): + return + + if mba.is_final_blend: + mba.blended_latent = mba.current_latent + return + + settings = SoftInpaintingSettings(power, scale, detail_preservation, mask_inf, dif_thresh, dif_contr) + + # todo: Why is sigma 2D? Both values are the same. + mba.blended_latent = latent_blend(settings, + mba.init_latent, + mba.current_latent, + get_modified_nmask(settings, mba.nmask, mba.sigma[0])) + + def post_sample(self, p, ps: scripts.PostSampleArgs, enabled, power, scale, detail_preservation, mask_inf, + dif_thresh, dif_contr): + if not enabled: + return + + if not processing_uses_inpainting(p): + return + + nmask = getattr(p, "nmask", None) + if nmask is None: + return + + from modules import images + from modules.shared import opts + + settings = SoftInpaintingSettings(power, scale, detail_preservation, mask_inf, dif_thresh, dif_contr) + + # since the original code puts holes in the existing overlay images, + # we have to rebuild them. + self.overlay_images = [] + for img in p.init_images: + + image = images.flatten(img, opts.img2img_background_color) + + if p.paste_to is None and p.resize_mode != 3: + image = images.resize_image(p.resize_mode, image, p.width, p.height) + + self.overlay_images.append(image.convert('RGBA')) + + if len(p.init_images) == 1: + self.overlay_images = self.overlay_images * p.batch_size + + if getattr(ps.samples, 'already_decoded', False): + self.masks_for_overlay = apply_masks(settings=settings, + nmask=nmask, + overlay_images=self.overlay_images, + width=p.width, + height=p.height, + paste_to=p.paste_to) + else: + self.masks_for_overlay = apply_adaptive_masks(settings=settings, + nmask=nmask, + latent_orig=p.init_latent, + latent_processed=ps.samples, + overlay_images=self.overlay_images, + width=p.width, + height=p.height, + paste_to=p.paste_to) + + def postprocess_maskoverlay(self, p, ppmo: scripts.PostProcessMaskOverlayArgs, enabled, power, scale, + detail_preservation, mask_inf, dif_thresh, dif_contr): + if not enabled: + return + + if not processing_uses_inpainting(p): + return + + if self.masks_for_overlay is None: + return + + if self.overlay_images is None: + return + + ppmo.mask_for_overlay = self.masks_for_overlay[ppmo.index] + ppmo.overlay_image = self.overlay_images[ppmo.index] diff --git a/html/extra-networks-card.html b/html/extra-networks-card.html index 39674666f1e..f1d959a6733 100644 --- a/html/extra-networks-card.html +++ b/html/extra-networks-card.html @@ -1,14 +1,9 @@ -
+
{background_image} -
- {metadata_button} - {edit_button} -
-
-
- -
- {name} - {description} +
{copy_path_button}{metadata_button}{edit_button}
+
+
{search_terms}
+ {name} + {description}
diff --git a/html/extra-networks-copy-path-button.html b/html/extra-networks-copy-path-button.html new file mode 100644 index 00000000000..8083bb03357 --- /dev/null +++ b/html/extra-networks-copy-path-button.html @@ -0,0 +1,5 @@ +
+
\ No newline at end of file diff --git a/html/extra-networks-edit-item-button.html b/html/extra-networks-edit-item-button.html new file mode 100644 index 00000000000..0fe43082ad1 --- /dev/null +++ b/html/extra-networks-edit-item-button.html @@ -0,0 +1,4 @@ +
+
\ No newline at end of file diff --git a/html/extra-networks-metadata-button.html b/html/extra-networks-metadata-button.html new file mode 100644 index 00000000000..285b5b3b658 --- /dev/null +++ b/html/extra-networks-metadata-button.html @@ -0,0 +1,4 @@ +
+
\ No newline at end of file diff --git a/html/extra-networks-pane.html b/html/extra-networks-pane.html new file mode 100644 index 00000000000..02a87108655 --- /dev/null +++ b/html/extra-networks-pane.html @@ -0,0 +1,55 @@ +
+ +
+
+ {tree_html} +
+
+ {items_html} +
+
+
\ No newline at end of file diff --git a/html/extra-networks-tree-button.html b/html/extra-networks-tree-button.html new file mode 100644 index 00000000000..9dc2e2a40c8 --- /dev/null +++ b/html/extra-networks-tree-button.html @@ -0,0 +1,23 @@ + +
+ + {action_list_item_action_leading} + + + {action_list_item_visual_leading} + + + {action_list_item_label} + + + {action_list_item_visual_trailing} + + + {action_list_item_action_trailing} + +
\ No newline at end of file diff --git a/html/licenses.html b/html/licenses.html index ef6f2c0a42b..9f5d1e9dc5c 100644 --- a/html/licenses.html +++ b/html/licenses.html @@ -4,107 +4,6 @@ #licenses pre { margin: 1em 0 2em 0;} -

CodeFormer

-Parts of CodeFormer code had to be copied to be compatible with GFPGAN. -
-S-Lab License 1.0
-
-Copyright 2022 S-Lab
-
-Redistribution and use for non-commercial purpose in source and
-binary forms, with or without modification, are permitted provided
-that the following conditions are met:
-
-1. Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-
-2. Redistributions in binary form must reproduce the above copyright
-   notice, this list of conditions and the following disclaimer in
-   the documentation and/or other materials provided with the
-   distribution.
-
-3. Neither the name of the copyright holder nor the names of its
-   contributors may be used to endorse or promote products derived
-   from this software without specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-In the event that redistribution and/or use for commercial purpose in
-source or binary forms, with or without modification is required,
-please contact the contributor(s) of the work.
-
- - -

ESRGAN

-Code for architecture and reading models copied. -
-MIT License
-
-Copyright (c) 2021 victorca25
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
-
- -

Real-ESRGAN

-Some code is copied to support ESRGAN models. -
-BSD 3-Clause License
-
-Copyright (c) 2021, Xintao Wang
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
-1. Redistributions of source code must retain the above copyright notice, this
-   list of conditions and the following disclaimer.
-
-2. Redistributions in binary form must reproduce the above copyright notice,
-   this list of conditions and the following disclaimer in the documentation
-   and/or other materials provided with the distribution.
-
-3. Neither the name of the copyright holder nor the names of its
-   contributors may be used to endorse or promote products derived from
-   this software without specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
-AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
-IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
-DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
-FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
-DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
-SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
-CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
-OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-

InvokeAI

Some code for compatibility with OSX is taken from lstein's repository. 
@@ -183,213 +82,6 @@ 
SwinIR

-Code added by contributors, most likely copied from this repository.
-
-
-                                 Apache License
-                           Version 2.0, January 2004
-                        http://www.apache.org/licenses/
-
-   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
-
-   1. Definitions.
-
-      "License" shall mean the terms and conditions for use, reproduction,
-      and distribution as defined by Sections 1 through 9 of this document.
-
-      "Licensor" shall mean the copyright owner or entity authorized by
-      the copyright owner that is granting the License.
-
-      "Legal Entity" shall mean the union of the acting entity and all
-      other entities that control, are controlled by, or are under common
-      control with that entity. For the purposes of this definition,
-      "control" means (i) the power, direct or indirect, to cause the
-      direction or management of such entity, whether by contract or
-      otherwise, or (ii) ownership of fifty percent (50%) or more of the
-      outstanding shares, or (iii) beneficial ownership of such entity.
-
-      "You" (or "Your") shall mean an individual or Legal Entity
-      exercising permissions granted by this License.
-
-      "Source" form shall mean the preferred form for making modifications,
-      including but not limited to software source code, documentation
-      source, and configuration files.
-
-      "Object" form shall mean any form resulting from mechanical
-      transformation or translation of a Source form, including but
-      not limited to compiled object code, generated documentation,
-      and conversions to other media types.
-
-      "Work" shall mean the work of authorship, whether in Source or
-      Object form, made available under the License, as indicated by a
-      copyright notice that is included in or attached to the work
-      (an example is provided in the Appendix below).
-
-      "Derivative Works" shall mean any work, whether in Source or Object
-      form, that is based on (or derived from) the Work and for which the
-      editorial revisions, annotations, elaborations, or other modifications
-      represent, as a whole, an original work of authorship. For the purposes
-      of this License, Derivative Works shall not include works that remain
-      separable from, or merely link (or bind by name) to the interfaces of,
-      the Work and Derivative Works thereof.
-
-      "Contribution" shall mean any work of authorship, including
-      the original version of the Work and any modifications or additions
-      to that Work or Derivative Works thereof, that is intentionally
-      submitted to Licensor for inclusion in the Work by the copyright owner
-      or by an individual or Legal Entity authorized to submit on behalf of
-      the copyright owner. For the purposes of this definition, "submitted"
-      means any form of electronic, verbal, or written communication sent
-      to the Licensor or its representatives, including but not limited to
-      communication on electronic mailing lists, source code control systems,
-      and issue tracking systems that are managed by, or on behalf of, the
-      Licensor for the purpose of discussing and improving the Work, but
-      excluding communication that is conspicuously marked or otherwise
-      designated in writing by the copyright owner as "Not a Contribution."
-
-      "Contributor" shall mean Licensor and any individual or Legal Entity
-      on behalf of whom a Contribution has been received by Licensor and
-      subsequently incorporated within the Work.
-
-   2. Grant of Copyright License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      copyright license to reproduce, prepare Derivative Works of,
-      publicly display, publicly perform, sublicense, and distribute the
-      Work and such Derivative Works in Source or Object form.
-
-   3. Grant of Patent License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      (except as stated in this section) patent license to make, have made,
-      use, offer to sell, sell, import, and otherwise transfer the Work,
-      where such license applies only to those patent claims licensable
-      by such Contributor that are necessarily infringed by their
-      Contribution(s) alone or by combination of their Contribution(s)
-      with the Work to which such Contribution(s) was submitted. If You
-      institute patent litigation against any entity (including a
-      cross-claim or counterclaim in a lawsuit) alleging that the Work
-      or a Contribution incorporated within the Work constitutes direct
-      or contributory patent infringement, then any patent licenses
-      granted to You under this License for that Work shall terminate
-      as of the date such litigation is filed.
-
-   4. Redistribution. You may reproduce and distribute copies of the
-      Work or Derivative Works thereof in any medium, with or without
-      modifications, and in Source or Object form, provided that You
-      meet the following conditions:
-
-      (a) You must give any other recipients of the Work or
-          Derivative Works a copy of this License; and
-
-      (b) You must cause any modified files to carry prominent notices
-          stating that You changed the files; and
-
-      (c) You must retain, in the Source form of any Derivative Works
-          that You distribute, all copyright, patent, trademark, and
-          attribution notices from the Source form of the Work,
-          excluding those notices that do not pertain to any part of
-          the Derivative Works; and
-
-      (d) If the Work includes a "NOTICE" text file as part of its
-          distribution, then any Derivative Works that You distribute must
-          include a readable copy of the attribution notices contained
-          within such NOTICE file, excluding those notices that do not
-          pertain to any part of the Derivative Works, in at least one
-          of the following places: within a NOTICE text file distributed
-          as part of the Derivative Works; within the Source form or
-          documentation, if provided along with the Derivative Works; or,
-          within a display generated by the Derivative Works, if and
-          wherever such third-party notices normally appear. The contents
-          of the NOTICE file are for informational purposes only and
-          do not modify the License. You may add Your own attribution
-          notices within Derivative Works that You distribute, alongside
-          or as an addendum to the NOTICE text from the Work, provided
-          that such additional attribution notices cannot be construed
-          as modifying the License.
-
-      You may add Your own copyright statement to Your modifications and
-      may provide additional or different license terms and conditions
-      for use, reproduction, or distribution of Your modifications, or
-      for any such Derivative Works as a whole, provided Your use,
-      reproduction, and distribution of the Work otherwise complies with
-      the conditions stated in this License.
-
-   5. Submission of Contributions. Unless You explicitly state otherwise,
-      any Contribution intentionally submitted for inclusion in the Work
-      by You to the Licensor shall be under the terms and conditions of
-      this License, without any additional terms or conditions.
-      Notwithstanding the above, nothing herein shall supersede or modify
-      the terms of any separate license agreement you may have executed
-      with Licensor regarding such Contributions.
-
-   6. Trademarks. This License does not grant permission to use the trade
-      names, trademarks, service marks, or product names of the Licensor,
-      except as required for reasonable and customary use in describing the
-      origin of the Work and reproducing the content of the NOTICE file.
-
-   7. Disclaimer of Warranty. Unless required by applicable law or
-      agreed to in writing, Licensor provides the Work (and each
-      Contributor provides its Contributions) on an "AS IS" BASIS,
-      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
-      implied, including, without limitation, any warranties or conditions
-      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
-      PARTICULAR PURPOSE. You are solely responsible for determining the
-      appropriateness of using or redistributing the Work and assume any
-      risks associated with Your exercise of permissions under this License.
-
-   8. Limitation of Liability. In no event and under no legal theory,
-      whether in tort (including negligence), contract, or otherwise,
-      unless required by applicable law (such as deliberate and grossly
-      negligent acts) or agreed to in writing, shall any Contributor be
-      liable to You for damages, including any direct, indirect, special,
-      incidental, or consequential damages of any character arising as a
-      result of this License or out of the use or inability to use the
-      Work (including but not limited to damages for loss of goodwill,
-      work stoppage, computer failure or malfunction, or any and all
-      other commercial damages or losses), even if such Contributor
-      has been advised of the possibility of such damages.
-
-   9. Accepting Warranty or Additional Liability. While redistributing
-      the Work or Derivative Works thereof, You may choose to offer,
-      and charge a fee for, acceptance of support, warranty, indemnity,
-      or other liability obligations and/or rights consistent with this
-      License. However, in accepting such obligations, You may act only
-      on Your own behalf and on Your sole responsibility, not on behalf
-      of any other Contributor, and only if You agree to indemnify,
-      defend, and hold each Contributor harmless for any liability
-      incurred by, or claims asserted against, such Contributor by reason
-      of your accepting any such warranty or additional liability.
-
-   END OF TERMS AND CONDITIONS
-
-   APPENDIX: How to apply the Apache License to your work.
-
-      To apply the Apache License to your work, attach the following
-      boilerplate notice, with the fields enclosed by brackets "[]"
-      replaced with your own identifying information. (Don't include
-      the brackets!)  The text should be enclosed in the appropriate
-      comment syntax for the file format. We also recommend that a
-      file or class name and description of purpose be included on the
-      same "printed page" as the copyright notice for easier
-      identification within third-party archives.
-
-   Copyright [2021] [SwinIR Authors]
-
-   Licensed under the Apache License, Version 2.0 (the "License");
-   you may not use this file except in compliance with the License.
-   You may obtain a copy of the License at
-
-       http://www.apache.org/licenses/LICENSE-2.0
-
-   Unless required by applicable law or agreed to in writing, software
-   distributed under the License is distributed on an "AS IS" BASIS,
-   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-   See the License for the specific language governing permissions and
-   limitations under the License.
-

-
 
Memory Efficient Attention

 The sub-quadratic cross attention optimization uses modified code from the Memory Efficient Attention package that Alex Birch optimized for 3D tensors. This license is updated to reflect that.
 
@@ -687,4 +379,4 @@ 
TAESD
\ No newline at end of file
+

diff --git a/javascript/aspectRatioOverlay.js b/javascript/aspectRatioOverlay.js
index 2cf2d571fc0..c8751fe494f 100644
--- a/javascript/aspectRatioOverlay.js
+++ b/javascript/aspectRatioOverlay.js
@@ -50,17 +50,17 @@ function dimensionChange(e, is_width, is_height) {
         var scaledx = targetElement.naturalWidth * viewportscale;
         var scaledy = targetElement.naturalHeight * viewportscale;
 
-        var cleintRectTop = (viewportOffset.top + window.scrollY);
-        var cleintRectLeft = (viewportOffset.left + window.scrollX);
-        var cleintRectCentreY = cleintRectTop + (targetElement.clientHeight / 2);
-        var cleintRectCentreX = cleintRectLeft + (targetElement.clientWidth / 2);
+        var clientRectTop = (viewportOffset.top + window.scrollY);
+        var clientRectLeft = (viewportOffset.left + window.scrollX);
+        var clientRectCentreY = clientRectTop + (targetElement.clientHeight / 2);
+        var clientRectCentreX = clientRectLeft + (targetElement.clientWidth / 2);
 
         var arscale = Math.min(scaledx / currentWidth, scaledy / currentHeight);
         var arscaledx = currentWidth * arscale;
         var arscaledy = currentHeight * arscale;
 
-        var arRectTop = cleintRectCentreY - (arscaledy / 2);
-        var arRectLeft = cleintRectCentreX - (arscaledx / 2);
+        var arRectTop = clientRectCentreY - (arscaledy / 2);
+        var arRectLeft = clientRectCentreX - (arscaledx / 2);
         var arRectWidth = arscaledx;
         var arRectHeight = arscaledy;
 
diff --git a/javascript/extensions.js b/javascript/extensions.js
index 312131b76eb..cc8ee220b17 100644
--- a/javascript/extensions.js
+++ b/javascript/extensions.js
@@ -2,8 +2,11 @@
 function extensions_apply(_disabled_list, _update_list, disable_all) {
     var disable = [];
     var update = [];
-
-    gradioApp().querySelectorAll('#extensions input[type="checkbox"]').forEach(function(x) {
+    const extensions_input = gradioApp().querySelectorAll('#extensions input[type="checkbox"]');
+    if (extensions_input.length == 0) {
+        throw Error("Extensions page not yet loaded.");
+    }
+    extensions_input.forEach(function(x) {
         if (x.name.startsWith("enable_") && !x.checked) {
             disable.push(x.name.substring(7));
         }
diff --git a/javascript/extraNetworks.js b/javascript/extraNetworks.js
index 98a7abb745c..584fd6c75db 100644
--- a/javascript/extraNetworks.js
+++ b/javascript/extraNetworks.js
@@ -16,99 +16,112 @@ function toggleCss(key, css, enable) {
 }
 
 function setupExtraNetworksForTab(tabname) {
-    gradioApp().querySelector('#' + tabname + '_extra_tabs').classList.add('extra-networks');
-
-    var tabs = gradioApp().querySelector('#' + tabname + '_extra_tabs > div');
-    var searchDiv = gradioApp().getElementById(tabname + '_extra_search');
-    var search = searchDiv.querySelector('textarea');
-    var sort = gradioApp().getElementById(tabname + '_extra_sort');
-    var sortOrder = gradioApp().getElementById(tabname + '_extra_sortorder');
-    var refresh = gradioApp().getElementById(tabname + '_extra_refresh');
-    var showDirsDiv = gradioApp().getElementById(tabname + '_extra_show_dirs');
-    var showDirs = gradioApp().querySelector('#' + tabname + '_extra_show_dirs input');
-    var promptContainer = gradioApp().querySelector('.prompt-container-compact#' + tabname + '_prompt_container');
-    var negativePrompt = gradioApp().querySelector('#' + tabname + '_neg_prompt');
-
-    tabs.appendChild(searchDiv);
-    tabs.appendChild(sort);
-    tabs.appendChild(sortOrder);
-    tabs.appendChild(refresh);
-    tabs.appendChild(showDirsDiv);
-
-    var applyFilter = function() {
-        var searchTerm = search.value.toLowerCase();
-
-        gradioApp().querySelectorAll('#' + tabname + '_extra_tabs div.card').forEach(function(elem) {
-            var searchOnly = elem.querySelector('.search_only');
-            var text = elem.querySelector('.name').textContent.toLowerCase() + " " + elem.querySelector('.search_term').textContent.toLowerCase();
-
-            var visible = text.indexOf(searchTerm) != -1;
-
-            if (searchOnly && searchTerm.length < 4) {
-                visible = false;
-            }
+    function registerPrompt(tabname, id) {
+        var textarea = gradioApp().querySelector("#" + id + " > label > textarea");
 
-            elem.style.display = visible ? "" : "none";
+        if (!activePromptTextarea[tabname]) {
+            activePromptTextarea[tabname] = textarea;
+        }
+
+        textarea.addEventListener("focus", function() {
+            activePromptTextarea[tabname] = textarea;
         });
+    }
 
-        applySort();
-    };
+    var tabnav = gradioApp().querySelector('#' + tabname + '_extra_tabs > div.tab-nav');
+    var controlsDiv = document.createElement('DIV');
+    controlsDiv.classList.add('extra-networks-controls-div');
+    tabnav.appendChild(controlsDiv);
+    tabnav.insertBefore(controlsDiv, null);
+
+    var this_tab = gradioApp().querySelector('#' + tabname + '_extra_tabs');
+    this_tab.querySelectorAll(":scope > [id^='" + tabname + "_']").forEach(function(elem) {
+        // tabname_full = {tabname}_{extra_networks_tabname}
+        var tabname_full = elem.id;
+        var search = gradioApp().querySelector("#" + tabname_full + "_extra_search");
+        var sort_mode = gradioApp().querySelector("#" + tabname_full + "_extra_sort");
+        var sort_dir = gradioApp().querySelector("#" + tabname_full + "_extra_sort_dir");
+        var refresh = gradioApp().querySelector("#" + tabname_full + "_extra_refresh");
+
+        // If any of the buttons above don't exist, we want to skip this iteration of the loop.
+        if (!search || !sort_mode || !sort_dir || !refresh) {
+            return; // `return` is equivalent of `continue` but for forEach loops.
+        }
 
-    var applySort = function() {
-        var cards = gradioApp().querySelectorAll('#' + tabname + '_extra_tabs div.card');
+        var applyFilter = function(force) {
+            var searchTerm = search.value.toLowerCase();
+            gradioApp().querySelectorAll('#' + tabname + '_extra_tabs div.card').forEach(function(elem) {
+                var searchOnly = elem.querySelector('.search_only');
+                var text = Array.prototype.map.call(elem.querySelectorAll('.search_terms'), function(t) {
+                    return t.textContent.toLowerCase();
+                }).join(" ");
+
+                var visible = text.indexOf(searchTerm) != -1;
+                if (searchOnly && searchTerm.length < 4) {
+                    visible = false;
+                }
+                if (visible) {
+                    elem.classList.remove("hidden");
+                } else {
+                    elem.classList.add("hidden");
+                }
+            });
 
-        var reverse = sortOrder.classList.contains("sortReverse");
-        var sortKey = sort.querySelector("input").value.toLowerCase().replace("sort", "").replaceAll(" ", "_").replace(/_+$/, "").trim() || "name";
-        sortKey = "sort" + sortKey.charAt(0).toUpperCase() + sortKey.slice(1);
-        var sortKeyStore = sortKey + "-" + (reverse ? "Descending" : "Ascending") + "-" + cards.length;
+            applySort(force);
+        };
 
-        if (sortKeyStore == sort.dataset.sortkey) {
-            return;
-        }
-        sort.dataset.sortkey = sortKeyStore;
+        var applySort = function(force) {
+            var cards = gradioApp().querySelectorAll('#' + tabname + '_extra_tabs div.card');
+            var reverse = sort_dir.dataset.sortdir == "Descending";
+            var sortKey = sort_mode.dataset.sortmode.toLowerCase().replace("sort", "").replaceAll(" ", "_").replace(/_+$/, "").trim() || "name";
+            sortKey = "sort" + sortKey.charAt(0).toUpperCase() + sortKey.slice(1);
+            var sortKeyStore = sortKey + "-" + (reverse ? "Descending" : "Ascending") + "-" + cards.length;
 
-        cards.forEach(function(card) {
-            card.originalParentElement = card.parentElement;
-        });
-        var sortedCards = Array.from(cards);
-        sortedCards.sort(function(cardA, cardB) {
-            var a = cardA.dataset[sortKey];
-            var b = cardB.dataset[sortKey];
-            if (!isNaN(a) && !isNaN(b)) {
-                return parseInt(a) - parseInt(b);
+            if (sortKeyStore == sort_mode.dataset.sortkey && !force) {
+                return;
             }
+            sort_mode.dataset.sortkey = sortKeyStore;
+
+            cards.forEach(function(card) {
+                card.originalParentElement = card.parentElement;
+            });
+            var sortedCards = Array.from(cards);
+            sortedCards.sort(function(cardA, cardB) {
+                var a = cardA.dataset[sortKey];
+                var b = cardB.dataset[sortKey];
+                if (!isNaN(a) && !isNaN(b)) {
+                    return parseInt(a) - parseInt(b);
+                }
 
-            return (a < b ? -1 : (a > b ? 1 : 0));
-        });
-        if (reverse) {
-            sortedCards.reverse();
-        }
-        cards.forEach(function(card) {
-            card.remove();
-        });
-        sortedCards.forEach(function(card) {
-            card.originalParentElement.appendChild(card);
-        });
-    };
-
-    search.addEventListener("input", applyFilter);
-    sortOrder.addEventListener("click", function() {
-        sortOrder.classList.toggle("sortReverse");
+                return (a < b ? -1 : (a > b ? 1 : 0));
+            });
+            if (reverse) {
+                sortedCards.reverse();
+            }
+            cards.forEach(function(card) {
+                card.remove();
+            });
+            sortedCards.forEach(function(card) {
+                card.originalParentElement.appendChild(card);
+            });
+        };
+
+        search.addEventListener("input", applyFilter);
         applySort();
-    });
-    applyFilter();
+        applyFilter();
+        extraNetworksApplySort[tabname_full] = applySort;
+        extraNetworksApplyFilter[tabname_full] = applyFilter;
 
-    extraNetworksApplySort[tabname] = applySort;
-    extraNetworksApplyFilter[tabname] = applyFilter;
+        var controls = gradioApp().querySelector("#" + tabname_full + "_controls");
+        controlsDiv.insertBefore(controls, null);
 
-    var showDirsUpdate = function() {
-        var css = '#' + tabname + '_extra_tabs .extra-network-subdirs { display: none; }';
-        toggleCss(tabname + '_extra_show_dirs_style', css, !showDirs.checked);
-        localSet('extra-networks-show-dirs', showDirs.checked ? 1 : 0);
-    };
-    showDirs.checked = localGet('extra-networks-show-dirs', 1) == 1;
-    showDirs.addEventListener("change", showDirsUpdate);
-    showDirsUpdate();
+        if (elem.style.display != "none") {
+            extraNetworksShowControlsForPage(tabname, tabname_full);
+        }
+    });
+
+    registerPrompt(tabname, tabname + "_prompt");
+    registerPrompt(tabname, tabname + "_neg_prompt");
 }
 
 function extraNetworksMovePromptToTab(tabname, id, showPrompt, showNegativePrompt) {
@@ -137,21 +150,42 @@ function extraNetworksMovePromptToTab(tabname, id, showPrompt, showNegativePromp
 }
 
 
-function extraNetworksUrelatedTabSelected(tabname) { // called from python when user selects an unrelated tab (generate)
+function extraNetworksShowControlsForPage(tabname, tabname_full) {
+    gradioApp().querySelectorAll('#' + tabname + '_extra_tabs .extra-networks-controls-div > div').forEach(function(elem) {
+        var targetId = tabname_full + "_controls";
+        elem.style.display = elem.id == targetId ? "" : "none";
+    });
+}
+
+
+function extraNetworksUnrelatedTabSelected(tabname) { // called from python when user selects an unrelated tab (generate)
     extraNetworksMovePromptToTab(tabname, '', false, false);
+
+    extraNetworksShowControlsForPage(tabname, null);
 }
 
-function extraNetworksTabSelected(tabname, id, showPrompt, showNegativePrompt) { // called from python when user selects an extra networks tab
+function extraNetworksTabSelected(tabname, id, showPrompt, showNegativePrompt, tabname_full) { // called from python when user selects an extra networks tab
     extraNetworksMovePromptToTab(tabname, id, showPrompt, showNegativePrompt);
 
+    extraNetworksShowControlsForPage(tabname, tabname_full);
 }
 
-function applyExtraNetworkFilter(tabname) {
-    setTimeout(extraNetworksApplyFilter[tabname], 1);
+function applyExtraNetworkFilter(tabname_full) {
+    var doFilter = function() {
+        var applyFunction = extraNetworksApplyFilter[tabname_full];
+
+        if (applyFunction) {
+            applyFunction(true);
+        }
+    };
+    setTimeout(doFilter, 1);
 }
 
-function applyExtraNetworkSort(tabname) {
-    setTimeout(extraNetworksApplySort[tabname], 1);
+function applyExtraNetworkSort(tabname_full) {
+    var doSort = function() {
+        extraNetworksApplySort[tabname_full](true);
+    };
+    setTimeout(doSort, 1);
 }
 
 var extraNetworksApplyFilter = {};
@@ -161,41 +195,24 @@ var activePromptTextarea = {};
 function setupExtraNetworks() {
     setupExtraNetworksForTab('txt2img');
     setupExtraNetworksForTab('img2img');
-
-    function registerPrompt(tabname, id) {
-        var textarea = gradioApp().querySelector("#" + id + " > label > textarea");
-
-        if (!activePromptTextarea[tabname]) {
-            activePromptTextarea[tabname] = textarea;
-        }
-
-        textarea.addEventListener("focus", function() {
-            activePromptTextarea[tabname] = textarea;
-        });
-    }
-
-    registerPrompt('txt2img', 'txt2img_prompt');
-    registerPrompt('txt2img', 'txt2img_neg_prompt');
-    registerPrompt('img2img', 'img2img_prompt');
-    registerPrompt('img2img', 'img2img_neg_prompt');
 }
 
-onUiLoaded(setupExtraNetworks);
-
 var re_extranet = /<([^:^>]+:[^:]+):[\d.]+>(.*)/;
 var re_extranet_g = /<([^:^>]+:[^:]+):[\d.]+>/g;
 
-function tryToRemoveExtraNetworkFromPrompt(textarea, text) {
-    var m = text.match(re_extranet);
+var re_extranet_neg = /\(([^:^>]+:[\d.]+)\)/;
+var re_extranet_g_neg = /\(([^:^>]+:[\d.]+)\)/g;
+function tryToRemoveExtraNetworkFromPrompt(textarea, text, isNeg) {
+    var m = text.match(isNeg ? re_extranet_neg : re_extranet);
     var replaced = false;
     var newTextareaText;
+    var extraTextBeforeNet = opts.extra_networks_add_text_separator;
     if (m) {
-        var extraTextBeforeNet = opts.extra_networks_add_text_separator;
         var extraTextAfterNet = m[2];
         var partToSearch = m[1];
         var foundAtPosition = -1;
-        newTextareaText = textarea.value.replaceAll(re_extranet_g, function(found, net, pos) {
-            m = found.match(re_extranet);
+        newTextareaText = textarea.value.replaceAll(isNeg ? re_extranet_g_neg : re_extranet_g, function(found, net, pos) {
+            m = found.match(isNeg ? re_extranet_neg : re_extranet);
             if (m[1] == partToSearch) {
                 replaced = true;
                 foundAtPosition = pos;
@@ -203,9 +220,8 @@ function tryToRemoveExtraNetworkFromPrompt(textarea, text) {
             }
             return found;
         });
-
         if (foundAtPosition >= 0) {
-            if (newTextareaText.substr(foundAtPosition, extraTextAfterNet.length) == extraTextAfterNet) {
+            if (extraTextAfterNet && newTextareaText.substr(foundAtPosition, extraTextAfterNet.length) == extraTextAfterNet) {
                 newTextareaText = newTextareaText.substr(0, foundAtPosition) + newTextareaText.substr(foundAtPosition + extraTextAfterNet.length);
             }
             if (newTextareaText.substr(foundAtPosition - extraTextBeforeNet.length, extraTextBeforeNet.length) == extraTextBeforeNet) {
@@ -213,13 +229,8 @@ function tryToRemoveExtraNetworkFromPrompt(textarea, text) {
             }
         }
     } else {
-        newTextareaText = textarea.value.replaceAll(new RegExp(text, "g"), function(found) {
-            if (found == text) {
-                replaced = true;
-                return "";
-            }
-            return found;
-        });
+        newTextareaText = textarea.value.replaceAll(new RegExp(`((?:${extraTextBeforeNet})?${text})`, "g"), "");
+        replaced = (newTextareaText != textarea.value);
     }
 
     if (replaced) {
@@ -230,14 +241,22 @@ function tryToRemoveExtraNetworkFromPrompt(textarea, text) {
     return false;
 }
 
-function cardClicked(tabname, textToAdd, allowNegativePrompt) {
-    var textarea = allowNegativePrompt ? activePromptTextarea[tabname] : gradioApp().querySelector("#" + tabname + "_prompt > label > textarea");
-
-    if (!tryToRemoveExtraNetworkFromPrompt(textarea, textToAdd)) {
-        textarea.value = textarea.value + opts.extra_networks_add_text_separator + textToAdd;
+function updatePromptArea(text, textArea, isNeg) {
+    if (!tryToRemoveExtraNetworkFromPrompt(textArea, text, isNeg)) {
+        textArea.value = textArea.value + opts.extra_networks_add_text_separator + text;
     }
 
-    updateInput(textarea);
+    updateInput(textArea);
+}
+
+function cardClicked(tabname, textToAdd, textToAddNegative, allowNegativePrompt) {
+    if (textToAddNegative.length > 0) {
+        updatePromptArea(textToAdd, gradioApp().querySelector("#" + tabname + "_prompt > label > textarea"));
+        updatePromptArea(textToAddNegative, gradioApp().querySelector("#" + tabname + "_neg_prompt > label > textarea"), true);
+    } else {
+        var textarea = allowNegativePrompt ? activePromptTextarea[tabname] : gradioApp().querySelector("#" + tabname + "_prompt > label > textarea");
+        updatePromptArea(textToAdd, textarea);
+    }
 }
 
 function saveCardPreview(event, tabname, filename) {
@@ -253,13 +272,219 @@ function saveCardPreview(event, tabname, filename) {
     event.preventDefault();
 }
 
-function extraNetworksSearchButton(tabs_id, event) {
-    var searchTextarea = gradioApp().querySelector("#" + tabs_id + ' > label > textarea');
-    var button = event.target;
-    var text = button.classList.contains("search-all") ? "" : button.textContent.trim();
+function extraNetworksTreeProcessFileClick(event, btn, tabname, extra_networks_tabname) {
+    /**
+     * Processes `onclick` events when user clicks on files in tree.
+     *
+     * @param event                     The generated event.
+     * @param btn                       The clicked `tree-list-item` button.
+     * @param tabname                   The name of the active tab in the sd webui. Ex: txt2img, img2img, etc.
+     * @param extra_networks_tabname    The id of the active extraNetworks tab. Ex: lora, checkpoints, etc.
+     */
+    // NOTE: Currently unused.
+    return;
+}
+
+function extraNetworksTreeProcessDirectoryClick(event, btn, tabname, extra_networks_tabname) {
+    /**
+     * Processes `onclick` events when user clicks on directories in tree.
+     *
+     * Here is how the tree reacts to clicks for various states:
+     * unselected unopened directory: Directory is selected and expanded.
+     * unselected opened directory: Directory is selected.
+     * selected opened directory: Directory is collapsed and deselected.
+     * chevron is clicked: Directory is expanded or collapsed. Selected state unchanged.
+     *
+     * @param event                     The generated event.
+     * @param btn                       The clicked `tree-list-item` button.
+     * @param tabname                   The name of the active tab in the sd webui. Ex: txt2img, img2img, etc.
+     * @param extra_networks_tabname    The id of the active extraNetworks tab. Ex: lora, checkpoints, etc.
+     */
+    var ul = btn.nextElementSibling;
+    // This is the actual target that the user clicked on within the target button.
+    // We use this to detect if the chevron was clicked.
+    var true_targ = event.target;
+
+    function _expand_or_collapse(_ul, _btn) {
+        // Expands 
     if it is collapsed, collapses otherwise. Updates button attributes.
+        if (_ul.hasAttribute("hidden")) {
+            _ul.removeAttribute("hidden");
+            _btn.dataset.expanded = "";
+        } else {
+            _ul.setAttribute("hidden", "");
+            delete _btn.dataset.expanded;
+        }
+    }
+
+    function _remove_selected_from_all() {
+        // Removes the `selected` attribute from all buttons.
+        var sels = document.querySelectorAll("div.tree-list-content");
+        [...sels].forEach(el => {
+            delete el.dataset.selected;
+        });
+    }
+
+    function _select_button(_btn) {
+        // Removes `data-selected` attribute from all buttons then adds to passed button.
+        _remove_selected_from_all();
+        _btn.dataset.selected = "";
+    }
+
+    function _update_search(_tabname, _extra_networks_tabname, _search_text) {
+        // Update search input with select button's path.
+        var search_input_elem = gradioApp().querySelector("#" + tabname + "_" + extra_networks_tabname + "_extra_search");
+        search_input_elem.value = _search_text;
+        updateInput(search_input_elem);
+    }
+
+
+    // If user clicks on the chevron, then we do not select the folder.
+    if (true_targ.matches(".tree-list-item-action--leading, .tree-list-item-action-chevron")) {
+        _expand_or_collapse(ul, btn);
+    } else {
+        // User clicked anywhere else on the button.
+        if ("selected" in btn.dataset && !(ul.hasAttribute("hidden"))) {
+            // If folder is select and open, collapse and deselect button.
+            _expand_or_collapse(ul, btn);
+            delete btn.dataset.selected;
+            _update_search(tabname, extra_networks_tabname, "");
+        } else if (!(!("selected" in btn.dataset) && !(ul.hasAttribute("hidden")))) {
+            // If folder is open and not selected, then we don't collapse; just select.
+            // NOTE: Double inversion sucks but it is the clearest way to show the branching here.
+            _expand_or_collapse(ul, btn);
+            _select_button(btn, tabname, extra_networks_tabname);
+            _update_search(tabname, extra_networks_tabname, btn.dataset.path);
+        } else {
+            // All other cases, just select the button.
+            _select_button(btn, tabname, extra_networks_tabname);
+            _update_search(tabname, extra_networks_tabname, btn.dataset.path);
+        }
+    }
+}
+
+function extraNetworksTreeOnClick(event, tabname, extra_networks_tabname) {
+    /**
+     * Handles `onclick` events for buttons within an `extra-network-tree .tree-list--tree`.
+     *
+     * Determines whether the clicked button in the tree is for a file entry or a directory
+     * then calls the appropriate function.
+     *
+     * @param event                     The generated event.
+     * @param tabname                   The name of the active tab in the sd webui. Ex: txt2img, img2img, etc.
+     * @param extra_networks_tabname    The id of the active extraNetworks tab. Ex: lora, checkpoints, etc.
+     */
+    var btn = event.currentTarget;
+    var par = btn.parentElement;
+    if (par.dataset.treeEntryType === "file") {
+        extraNetworksTreeProcessFileClick(event, btn, tabname, extra_networks_tabname);
+    } else {
+        extraNetworksTreeProcessDirectoryClick(event, btn, tabname, extra_networks_tabname);
+    }
+}
 
-    searchTextarea.value = text;
-    updateInput(searchTextarea);
+function extraNetworksControlSortOnClick(event, tabname, extra_networks_tabname) {
+    /**
+     * Handles `onclick` events for the Sort Mode button.
+     *
+     * Modifies the data attributes of the Sort Mode button to cycle between
+     * various sorting modes.
+     *
+     * @param event                     The generated event.
+     * @param tabname                   The name of the active tab in the sd webui. Ex: txt2img, img2img, etc.
+     * @param extra_networks_tabname    The id of the active extraNetworks tab. Ex: lora, checkpoints, etc.
+     */
+    var curr_mode = event.currentTarget.dataset.sortmode;
+    var el_sort_dir = gradioApp().querySelector("#" + tabname + "_" + extra_networks_tabname + "_extra_sort_dir");
+    var sort_dir = el_sort_dir.dataset.sortdir;
+    if (curr_mode == "path") {
+        event.currentTarget.dataset.sortmode = "name";
+        event.currentTarget.dataset.sortkey = "sortName-" + sort_dir + "-640";
+        event.currentTarget.setAttribute("title", "Sort by filename");
+    } else if (curr_mode == "name") {
+        event.currentTarget.dataset.sortmode = "date_created";
+        event.currentTarget.dataset.sortkey = "sortDate_created-" + sort_dir + "-640";
+        event.currentTarget.setAttribute("title", "Sort by date created");
+    } else if (curr_mode == "date_created") {
+        event.currentTarget.dataset.sortmode = "date_modified";
+        event.currentTarget.dataset.sortkey = "sortDate_modified-" + sort_dir + "-640";
+        event.currentTarget.setAttribute("title", "Sort by date modified");
+    } else {
+        event.currentTarget.dataset.sortmode = "path";
+        event.currentTarget.dataset.sortkey = "sortPath-" + sort_dir + "-640";
+        event.currentTarget.setAttribute("title", "Sort by path");
+    }
+    applyExtraNetworkSort(tabname + "_" + extra_networks_tabname);
+}
+
+function extraNetworksControlSortDirOnClick(event, tabname, extra_networks_tabname) {
+    /**
+     * Handles `onclick` events for the Sort Direction button.
+     *
+     * Modifies the data attributes of the Sort Direction button to cycle between
+     * ascending and descending sort directions.
+     *
+     * @param event                     The generated event.
+     * @param tabname                   The name of the active tab in the sd webui. Ex: txt2img, img2img, etc.
+     * @param extra_networks_tabname    The id of the active extraNetworks tab. Ex: lora, checkpoints, etc.
+     */
+    if (event.currentTarget.dataset.sortdir == "Ascending") {
+        event.currentTarget.dataset.sortdir = "Descending";
+        event.currentTarget.setAttribute("title", "Sort descending");
+    } else {
+        event.currentTarget.dataset.sortdir = "Ascending";
+        event.currentTarget.setAttribute("title", "Sort ascending");
+    }
+    applyExtraNetworkSort(tabname + "_" + extra_networks_tabname);
+}
+
+function extraNetworksControlTreeViewOnClick(event, tabname, extra_networks_tabname) {
+    /**
+     * Handles `onclick` events for the Tree View button.
+     *
+     * Toggles the tree view in the extra networks pane.
+     *
+     * @param event                     The generated event.
+     * @param tabname                   The name of the active tab in the sd webui. Ex: txt2img, img2img, etc.
+     * @param extra_networks_tabname    The id of the active extraNetworks tab. Ex: lora, checkpoints, etc.
+     */
+    const tree = gradioApp().getElementById(tabname + "_" + extra_networks_tabname + "_tree");
+    const parent = tree.parentElement;
+    let resizeHandle = parent.querySelector('.resize-handle');
+    tree.classList.toggle("hidden");
+
+    if (tree.classList.contains("hidden")) {
+        tree.style.display = 'none';
+        parent.style.display = 'flex';
+        if (resizeHandle) {
+            resizeHandle.style.display = 'none';
+        }
+    } else {
+        tree.style.display = 'block';
+        parent.style.display = 'grid';
+        if (!resizeHandle) {
+            setupResizeHandle(parent);
+            resizeHandle = parent.querySelector('.resize-handle');
+        }
+        resizeHandle.style.display = 'block';
+    }
+    event.currentTarget.classList.toggle("extra-network-control--enabled");
+}
+
+function extraNetworksControlRefreshOnClick(event, tabname, extra_networks_tabname) {
+    /**
+     * Handles `onclick` events for the Refresh Page button.
+     *
+     * In order to actually call the python functions in `ui_extra_networks.py`
+     * to refresh the page, we created an empty gradio button in that file with an
+     * event handler that refreshes the page. So what this function here does
+     * is it manually raises a `click` event on that button.
+     *
+     * @param event                     The generated event.
+     * @param tabname                   The name of the active tab in the sd webui. Ex: txt2img, img2img, etc.
+     * @param extra_networks_tabname    The id of the active extraNetworks tab. Ex: lora, checkpoints, etc.
+     */
+    var btn_refresh_internal = gradioApp().getElementById(tabname + "_" + extra_networks_tabname + "_extra_refresh_internal");
+    btn_refresh_internal.dispatchEvent(new Event("click"));
 }
 
 var globalPopup = null;
@@ -303,12 +528,76 @@ function popupId(id) {
     popup(storedPopupIds[id]);
 }
 
+function extraNetworksFlattenMetadata(obj) {
+    const result = {};
+
+    // Convert any stringified JSON objects to actual objects
+    for (const key of Object.keys(obj)) {
+        if (typeof obj[key] === 'string') {
+            try {
+                const parsed = JSON.parse(obj[key]);
+                if (parsed && typeof parsed === 'object') {
+                    obj[key] = parsed;
+                }
+            } catch (error) {
+                continue;
+            }
+        }
+    }
+
+    // Flatten the object
+    for (const key of Object.keys(obj)) {
+        if (typeof obj[key] === 'object' && obj[key] !== null) {
+            const nested = extraNetworksFlattenMetadata(obj[key]);
+            for (const nestedKey of Object.keys(nested)) {
+                result[`${key}/${nestedKey}`] = nested[nestedKey];
+            }
+        } else {
+            result[key] = obj[key];
+        }
+    }
+
+    // Special case for handling modelspec keys
+    for (const key of Object.keys(result)) {
+        if (key.startsWith("modelspec.")) {
+            result[key.replaceAll(".", "/")] = result[key];
+            delete result[key];
+        }
+    }
+
+    // Add empty keys to designate hierarchy
+    for (const key of Object.keys(result)) {
+        const parts = key.split("/");
+        for (let i = 1; i < parts.length; i++) {
+            const parent = parts.slice(0, i).join("/");
+            if (!result[parent]) {
+                result[parent] = "";
+            }
+        }
+    }
+
+    return result;
+}
+
 function extraNetworksShowMetadata(text) {
+    try {
+        let parsed = JSON.parse(text);
+        if (parsed && typeof parsed === 'object') {
+            parsed = extraNetworksFlattenMetadata(parsed);
+            const table = createVisualizationTable(parsed, 0);
+            popup(table);
+            return;
+        }
+    } catch (error) {
+        console.eror(error);
+    }
+
     var elem = document.createElement('pre');
     elem.classList.add('popup-metadata');
     elem.textContent = text;
 
     popup(elem);
+    return;
 }
 
 function requestGet(url, data, handler, errorHandler) {
@@ -337,6 +626,11 @@ function requestGet(url, data, handler, errorHandler) {
     xhr.send(js);
 }
 
+function extraNetworksCopyCardPath(event, path) {
+    navigator.clipboard.writeText(path);
+    event.stopPropagation();
+}
+
 function extraNetworksRequestMetadata(event, extraPage, cardName) {
     var showError = function() {
         extraNetworksShowMetadata("there was an error getting metadata");
@@ -398,3 +692,39 @@ window.addEventListener("keydown", function(event) {
         closePopup();
     }
 });
+
+/**
+ * Setup custom loading for this script.
+ * We need to wait for all of our HTML to be generated in the extra networks tabs
+ * before we can actually run the `setupExtraNetworks` function.
+ * The `onUiLoaded` function actually runs before all of our extra network tabs are
+ * finished generating. Thus we needed this new method.
+ *
+ */
+
+var uiAfterScriptsCallbacks = [];
+var uiAfterScriptsTimeout = null;
+var executedAfterScripts = false;
+
+function scheduleAfterScriptsCallbacks() {
+    clearTimeout(uiAfterScriptsTimeout);
+    uiAfterScriptsTimeout = setTimeout(function() {
+        executeCallbacks(uiAfterScriptsCallbacks);
+    }, 200);
+}
+
+onUiLoaded(function() {
+    var mutationObserver = new MutationObserver(function(m) {
+        let existingSearchfields = gradioApp().querySelectorAll("[id$='_extra_search']").length;
+        let neededSearchfields = gradioApp().querySelectorAll("[id$='_extra_tabs'] > .tab-nav > button").length - 2;
+
+        if (!executedAfterScripts && existingSearchfields >= neededSearchfields) {
+            mutationObserver.disconnect();
+            executedAfterScripts = true;
+            scheduleAfterScriptsCallbacks();
+        }
+    });
+    mutationObserver.observe(gradioApp(), {childList: true, subtree: true});
+});
+
+uiAfterScriptsCallbacks.push(setupExtraNetworks);
diff --git a/javascript/profilerVisualization.js b/javascript/profilerVisualization.js
index 9d8e5f42f32..9822f4b2a2a 100644
--- a/javascript/profilerVisualization.js
+++ b/javascript/profilerVisualization.js
@@ -33,120 +33,141 @@ function createRow(table, cellName, items) {
     return res;
 }
 
-function showProfile(path, cutoff = 0.05) {
-    requestGet(path, {}, function(data) {
-        var table = document.createElement('table');
-        table.className = 'popup-table';
-
-        data.records['total'] = data.total;
-        var keys = Object.keys(data.records).sort(function(a, b) {
-            return data.records[b] - data.records[a];
+function createVisualizationTable(data, cutoff = 0, sort = "") {
+    var table = document.createElement('table');
+    table.className = 'popup-table';
+
+    var keys = Object.keys(data);
+    if (sort === "number") {
+        keys = keys.sort(function(a, b) {
+            return data[b] - data[a];
         });
-        var items = keys.map(function(x) {
-            return {key: x, parts: x.split('/'), time: data.records[x]};
+    } else {
+        keys = keys.sort();
+    }
+    var items = keys.map(function(x) {
+        return {key: x, parts: x.split('/'), value: data[x]};
+    });
+    var maxLength = items.reduce(function(a, b) {
+        return Math.max(a, b.parts.length);
+    }, 0);
+
+    var cols = createRow(
+        table,
+        'th',
+        [
+            cutoff === 0 ? 'key' : 'record',
+            cutoff === 0 ? 'value' : 'seconds'
+        ]
+    );
+    cols[0].colSpan = maxLength;
+
+    function arraysEqual(a, b) {
+        return !(a < b || b < a);
+    }
+
+    var addLevel = function(level, parent, hide) {
+        var matching = items.filter(function(x) {
+            return x.parts[level] && !x.parts[level + 1] && arraysEqual(x.parts.slice(0, level), parent);
         });
-        var maxLength = items.reduce(function(a, b) {
-            return Math.max(a, b.parts.length);
-        }, 0);
-
-        var cols = createRow(table, 'th', ['record', 'seconds']);
-        cols[0].colSpan = maxLength;
-
-        function arraysEqual(a, b) {
-            return !(a < b || b < a);
+        if (sort === "number") {
+            matching = matching.sort(function(a, b) {
+                return b.value - a.value;
+            });
+        } else {
+            matching = matching.sort();
         }
+        var othersTime = 0;
+        var othersList = [];
+        var othersRows = [];
+        var childrenRows = [];
+        matching.forEach(function(x) {
+            var visible = (cutoff === 0 && !hide) || (x.value >= cutoff && !hide);
+
+            var cells = [];
+            for (var i = 0; i < maxLength; i++) {
+                cells.push(x.parts[i]);
+            }
+            cells.push(cutoff === 0 ? x.value : x.value.toFixed(3));
+            var cols = createRow(table, 'td', cells);
+            for (i = 0; i < level; i++) {
+                cols[i].className = 'muted';
+            }
 
-        var addLevel = function(level, parent, hide) {
-            var matching = items.filter(function(x) {
-                return x.parts[level] && !x.parts[level + 1] && arraysEqual(x.parts.slice(0, level), parent);
-            });
-            var sorted = matching.sort(function(a, b) {
-                return b.time - a.time;
-            });
-            var othersTime = 0;
-            var othersList = [];
-            var othersRows = [];
-            var childrenRows = [];
-            sorted.forEach(function(x) {
-                var visible = x.time >= cutoff && !hide;
-
-                var cells = [];
-                for (var i = 0; i < maxLength; i++) {
-                    cells.push(x.parts[i]);
-                }
-                cells.push(x.time.toFixed(3));
-                var cols = createRow(table, 'td', cells);
-                for (i = 0; i < level; i++) {
-                    cols[i].className = 'muted';
-                }
-
-                var tr = cols[0].parentNode;
-                if (!visible) {
-                    tr.classList.add("hidden");
-                }
-
-                if (x.time >= cutoff) {
-                    childrenRows.push(tr);
-                } else {
-                    othersTime += x.time;
-                    othersList.push(x.parts[level]);
-                    othersRows.push(tr);
-                }
-
-                var children = addLevel(level + 1, parent.concat([x.parts[level]]), true);
-                if (children.length > 0) {
-                    var cell = cols[level];
-                    var onclick = function() {
-                        cell.classList.remove("link");
-                        cell.removeEventListener("click", onclick);
-                        children.forEach(function(x) {
-                            x.classList.remove("hidden");
-                        });
-                    };
-                    cell.classList.add("link");
-                    cell.addEventListener("click", onclick);
-                }
-            });
+            var tr = cols[0].parentNode;
+            if (!visible) {
+                tr.classList.add("hidden");
+            }
 
-            if (othersTime > 0) {
-                var cells = [];
-                for (var i = 0; i < maxLength; i++) {
-                    cells.push(parent[i]);
-                }
-                cells.push(othersTime.toFixed(3));
-                cells[level] = 'others';
-                var cols = createRow(table, 'td', cells);
-                for (i = 0; i < level; i++) {
-                    cols[i].className = 'muted';
-                }
+            if (cutoff === 0 || x.value >= cutoff) {
+                childrenRows.push(tr);
+            } else {
+                othersTime += x.value;
+                othersList.push(x.parts[level]);
+                othersRows.push(tr);
+            }
 
+            var children = addLevel(level + 1, parent.concat([x.parts[level]]), true);
+            if (children.length > 0) {
                 var cell = cols[level];
-                var tr = cell.parentNode;
                 var onclick = function() {
-                    tr.classList.add("hidden");
                     cell.classList.remove("link");
                     cell.removeEventListener("click", onclick);
-                    othersRows.forEach(function(x) {
+                    children.forEach(function(x) {
                         x.classList.remove("hidden");
                     });
                 };
-
-                cell.title = othersList.join(", ");
                 cell.classList.add("link");
                 cell.addEventListener("click", onclick);
+            }
+        });
 
-                if (hide) {
-                    tr.classList.add("hidden");
-                }
+        if (othersTime > 0) {
+            var cells = [];
+            for (var i = 0; i < maxLength; i++) {
+                cells.push(parent[i]);
+            }
+            cells.push(othersTime.toFixed(3));
+            cells[level] = 'others';
+            var cols = createRow(table, 'td', cells);
+            for (i = 0; i < level; i++) {
+                cols[i].className = 'muted';
+            }
 
-                childrenRows.push(tr);
+            var cell = cols[level];
+            var tr = cell.parentNode;
+            var onclick = function() {
+                tr.classList.add("hidden");
+                cell.classList.remove("link");
+                cell.removeEventListener("click", onclick);
+                othersRows.forEach(function(x) {
+                    x.classList.remove("hidden");
+                });
+            };
+
+            cell.title = othersList.join(", ");
+            cell.classList.add("link");
+            cell.addEventListener("click", onclick);
+
+            if (hide) {
+                tr.classList.add("hidden");
             }
 
-            return childrenRows;
-        };
+            childrenRows.push(tr);
+        }
+
+        return childrenRows;
+    };
 
-        addLevel(0, []);
+    addLevel(0, []);
+
+    return table;
+}
 
+function showProfile(path, cutoff = 0.05) {
+    requestGet(path, {}, function(data) {
+        data.records['total'] = data.total;
+        const table = createVisualizationTable(data.records, cutoff, "number");
         popup(table);
     });
 }
diff --git a/javascript/progressbar.js b/javascript/progressbar.js
index 777614954b2..f068bac6aba 100644
--- a/javascript/progressbar.js
+++ b/javascript/progressbar.js
@@ -45,8 +45,15 @@ function formatTime(secs) {
     }
 }
 
+
+var originalAppTitle = undefined;
+
+onUiLoaded(function() {
+    originalAppTitle = document.title;
+});
+
 function setTitle(progress) {
-    var title = 'Stable Diffusion';
+    var title = originalAppTitle;
 
     if (opts.show_progress_in_title && progress) {
         title = '[' + progress.trim() + '] ' + title;
diff --git a/javascript/resizeHandle.js b/javascript/resizeHandle.js
index 8c5c5169210..50251ffc1a5 100644
--- a/javascript/resizeHandle.js
+++ b/javascript/resizeHandle.js
@@ -1,8 +1,8 @@
 (function() {
     const GRADIO_MIN_WIDTH = 320;
-    const GRID_TEMPLATE_COLUMNS = '1fr 16px 1fr';
     const PAD = 16;
     const DEBOUNCE_TIME = 100;
+    const DOUBLE_TAP_DELAY = 200; //ms
 
     const R = {
         tracking: false,
@@ -11,6 +11,7 @@
         leftCol: null,
         leftColStartWidth: null,
         screenX: null,
+        lastTapTime: null,
     };
 
     let resizeTimer;
@@ -21,30 +22,29 @@
     }
 
     function displayResizeHandle(parent) {
+        if (!parent.needHideOnMoblie) {
+            return true;
+        }
         if (window.innerWidth < GRADIO_MIN_WIDTH * 2 + PAD * 4) {
             parent.style.display = 'flex';
-            if (R.handle != null) {
-                R.handle.style.opacity = '0';
-            }
+            parent.resizeHandle.style.display = "none";
             return false;
         } else {
             parent.style.display = 'grid';
-            if (R.handle != null) {
-                R.handle.style.opacity = '100';
-            }
+            parent.resizeHandle.style.display = "block";
             return true;
         }
     }
 
     function afterResize(parent) {
-        if (displayResizeHandle(parent) && parent.style.gridTemplateColumns != GRID_TEMPLATE_COLUMNS) {
+        if (displayResizeHandle(parent) && parent.style.gridTemplateColumns != parent.style.originalGridTemplateColumns) {
             const oldParentWidth = R.parentWidth;
             const newParentWidth = parent.offsetWidth;
             const widthL = parseInt(parent.style.gridTemplateColumns.split(' ')[0]);
 
             const ratio = newParentWidth / oldParentWidth;
 
-            const newWidthL = Math.max(Math.floor(ratio * widthL), GRADIO_MIN_WIDTH);
+            const newWidthL = Math.max(Math.floor(ratio * widthL), parent.minLeftColWidth);
             setLeftColGridTemplate(parent, newWidthL);
 
             R.parentWidth = newParentWidth;
@@ -52,6 +52,14 @@
     }
 
     function setup(parent) {
+
+        function onDoubleClick(evt) {
+            evt.preventDefault();
+            evt.stopPropagation();
+
+            parent.style.gridTemplateColumns = parent.style.originalGridTemplateColumns;
+        }
+
         const leftCol = parent.firstElementChild;
         const rightCol = parent.lastElementChild;
 
@@ -59,63 +67,109 @@
 
         parent.style.display = 'grid';
         parent.style.gap = '0';
-        parent.style.gridTemplateColumns = GRID_TEMPLATE_COLUMNS;
+        let leftColTemplate = "";
+        if (parent.children[0].style.flexGrow) {
+            leftColTemplate = `${parent.children[0].style.flexGrow}fr`;
+            parent.minLeftColWidth = GRADIO_MIN_WIDTH;
+            parent.minRightColWidth = GRADIO_MIN_WIDTH;
+            parent.needHideOnMoblie = true;
+        } else {
+            leftColTemplate = parent.children[0].style.flexBasis;
+            parent.minLeftColWidth = parent.children[0].style.flexBasis.slice(0, -2) / 2;
+            parent.minRightColWidth = 0;
+            parent.needHideOnMoblie = false;
+        }
+        const gridTemplateColumns = `${leftColTemplate} ${PAD}px ${parent.children[1].style.flexGrow}fr`;
+        parent.style.gridTemplateColumns = gridTemplateColumns;
+        parent.style.originalGridTemplateColumns = gridTemplateColumns;
 
         const resizeHandle = document.createElement('div');
         resizeHandle.classList.add('resize-handle');
         parent.insertBefore(resizeHandle, rightCol);
-
-        resizeHandle.addEventListener('mousedown', (evt) => {
-            if (evt.button !== 0) return;
-
-            evt.preventDefault();
-            evt.stopPropagation();
-
-            document.body.classList.add('resizing');
-
-            R.tracking = true;
-            R.parent = parent;
-            R.parentWidth = parent.offsetWidth;
-            R.handle = resizeHandle;
-            R.leftCol = leftCol;
-            R.leftColStartWidth = leftCol.offsetWidth;
-            R.screenX = evt.screenX;
+        parent.resizeHandle = resizeHandle;
+
+        ['mousedown', 'touchstart'].forEach((eventType) => {
+            resizeHandle.addEventListener(eventType, (evt) => {
+                if (eventType.startsWith('mouse')) {
+                    if (evt.button !== 0) return;
+                } else {
+                    if (evt.changedTouches.length !== 1) return;
+
+                    const currentTime = new Date().getTime();
+                    if (R.lastTapTime && currentTime - R.lastTapTime <= DOUBLE_TAP_DELAY) {
+                        onDoubleClick(evt);
+                        return;
+                    }
+
+                    R.lastTapTime = currentTime;
+                }
+
+                evt.preventDefault();
+                evt.stopPropagation();
+
+                document.body.classList.add('resizing');
+
+                R.tracking = true;
+                R.parent = parent;
+                R.parentWidth = parent.offsetWidth;
+                R.leftCol = leftCol;
+                R.leftColStartWidth = leftCol.offsetWidth;
+                if (eventType.startsWith('mouse')) {
+                    R.screenX = evt.screenX;
+                } else {
+                    R.screenX = evt.changedTouches[0].screenX;
+                }
+            });
         });
 
-        resizeHandle.addEventListener('dblclick', (evt) => {
-            evt.preventDefault();
-            evt.stopPropagation();
-
-            parent.style.gridTemplateColumns = GRID_TEMPLATE_COLUMNS;
-        });
+        resizeHandle.addEventListener('dblclick', onDoubleClick);
 
         afterResize(parent);
     }
 
-    window.addEventListener('mousemove', (evt) => {
-        if (evt.button !== 0) return;
-
-        if (R.tracking) {
-            evt.preventDefault();
-            evt.stopPropagation();
+    ['mousemove', 'touchmove'].forEach((eventType) => {
+        window.addEventListener(eventType, (evt) => {
+            if (eventType.startsWith('mouse')) {
+                if (evt.button !== 0) return;
+            } else {
+                if (evt.changedTouches.length !== 1) return;
+            }
 
-            const delta = R.screenX - evt.screenX;
-            const leftColWidth = Math.max(Math.min(R.leftColStartWidth - delta, R.parent.offsetWidth - GRADIO_MIN_WIDTH - PAD), GRADIO_MIN_WIDTH);
-            setLeftColGridTemplate(R.parent, leftColWidth);
-        }
+            if (R.tracking) {
+                if (eventType.startsWith('mouse')) {
+                    evt.preventDefault();
+                }
+                evt.stopPropagation();
+
+                let delta = 0;
+                if (eventType.startsWith('mouse')) {
+                    delta = R.screenX - evt.screenX;
+                } else {
+                    delta = R.screenX - evt.changedTouches[0].screenX;
+                }
+                const leftColWidth = Math.max(Math.min(R.leftColStartWidth - delta, R.parent.offsetWidth - R.parent.minRightColWidth - PAD), R.parent.minLeftColWidth);
+                setLeftColGridTemplate(R.parent, leftColWidth);
+            }
+        });
     });
 
-    window.addEventListener('mouseup', (evt) => {
-        if (evt.button !== 0) return;
+    ['mouseup', 'touchend'].forEach((eventType) => {
+        window.addEventListener(eventType, (evt) => {
+            if (eventType.startsWith('mouse')) {
+                if (evt.button !== 0) return;
+            } else {
+                if (evt.changedTouches.length !== 1) return;
+            }
 
-        if (R.tracking) {
-            evt.preventDefault();
-            evt.stopPropagation();
+            if (R.tracking) {
+                evt.preventDefault();
+                evt.stopPropagation();
 
-            R.tracking = false;
+                R.tracking = false;
 
-            document.body.classList.remove('resizing');
-        }
+                document.body.classList.remove('resizing');
+            }
+        });
     });
 
 
@@ -132,10 +186,15 @@
     setupResizeHandle = setup;
 })();
 
-onUiLoaded(function() {
+
+function setupAllResizeHandles() {
     for (var elem of gradioApp().querySelectorAll('.resize-handle-row')) {
-        if (!elem.querySelector('.resize-handle')) {
+        if (!elem.querySelector('.resize-handle') && !elem.children[0].classList.contains("hidden")) {
             setupResizeHandle(elem);
         }
     }
-});
+}
+
+
+onUiLoaded(setupAllResizeHandles);
+
diff --git a/javascript/settings.js b/javascript/settings.js
index e6009290ab3..b2d981c2144 100644
--- a/javascript/settings.js
+++ b/javascript/settings.js
@@ -55,8 +55,8 @@ onOptionsChanged(function() {
     });
 
     opts._categories.forEach(function(x) {
-        var section = x[0];
-        var category = x[1];
+        var section = localization[x[0]] ?? x[0];
+        var category = localization[x[1]] ?? x[1];
 
         var span = document.createElement('SPAN');
         span.textContent = category;
diff --git a/javascript/token-counters.js b/javascript/token-counters.js
index 2ecc7d91010..eeea7a5d26c 100644
--- a/javascript/token-counters.js
+++ b/javascript/token-counters.js
@@ -48,11 +48,6 @@ function setupTokenCounting(id, id_counter, id_button) {
     var counter = gradioApp().getElementById(id_counter);
     var textarea = gradioApp().querySelector(`#${id} > label > textarea`);
 
-    if (opts.disable_token_counters) {
-        counter.style.display = "none";
-        return;
-    }
-
     if (counter.parentElement == prompt.parentElement) {
         return;
     }
@@ -61,15 +56,32 @@ function setupTokenCounting(id, id_counter, id_button) {
     prompt.parentElement.style.position = "relative";
 
     var func = onEdit(id, textarea, 800, function() {
-        gradioApp().getElementById(id_button)?.click();
+        if (counter.classList.contains("token-counter-visible")) {
+            gradioApp().getElementById(id_button)?.click();
+        }
     });
     promptTokenCountUpdateFunctions[id] = func;
     promptTokenCountUpdateFunctions[id_button] = func;
 }
 
-function setupTokenCounters() {
-    setupTokenCounting('txt2img_prompt', 'txt2img_token_counter', 'txt2img_token_button');
-    setupTokenCounting('txt2img_neg_prompt', 'txt2img_negative_token_counter', 'txt2img_negative_token_button');
-    setupTokenCounting('img2img_prompt', 'img2img_token_counter', 'img2img_token_button');
-    setupTokenCounting('img2img_neg_prompt', 'img2img_negative_token_counter', 'img2img_negative_token_button');
+function toggleTokenCountingVisibility(id, id_counter, id_button) {
+    var counter = gradioApp().getElementById(id_counter);
+
+    counter.style.display = opts.disable_token_counters ? "none" : "block";
+    counter.classList.toggle("token-counter-visible", !opts.disable_token_counters);
 }
+
+function runCodeForTokenCounters(fun) {
+    fun('txt2img_prompt', 'txt2img_token_counter', 'txt2img_token_button');
+    fun('txt2img_neg_prompt', 'txt2img_negative_token_counter', 'txt2img_negative_token_button');
+    fun('img2img_prompt', 'img2img_token_counter', 'img2img_token_button');
+    fun('img2img_neg_prompt', 'img2img_negative_token_counter', 'img2img_negative_token_button');
+}
+
+onUiLoaded(function() {
+    runCodeForTokenCounters(setupTokenCounting);
+});
+
+onOptionsChanged(function() {
+    runCodeForTokenCounters(toggleTokenCountingVisibility);
+});
diff --git a/javascript/ui.js b/javascript/ui.js
index 18c9f891afc..1eef6d33799 100644
--- a/javascript/ui.js
+++ b/javascript/ui.js
@@ -119,9 +119,18 @@ function create_submit_args(args) {
     return res;
 }
 
+function setSubmitButtonsVisibility(tabname, showInterrupt, showSkip, showInterrupting) {
+    gradioApp().getElementById(tabname + '_interrupt').style.display = showInterrupt ? "block" : "none";
+    gradioApp().getElementById(tabname + '_skip').style.display = showSkip ? "block" : "none";
+    gradioApp().getElementById(tabname + '_interrupting').style.display = showInterrupting ? "block" : "none";
+}
+
 function showSubmitButtons(tabname, show) {
-    gradioApp().getElementById(tabname + '_interrupt').style.display = show ? "none" : "block";
-    gradioApp().getElementById(tabname + '_skip').style.display = show ? "none" : "block";
+    setSubmitButtonsVisibility(tabname, !show, !show, false);
+}
+
+function showSubmitInterruptingPlaceholder(tabname) {
+    setSubmitButtonsVisibility(tabname, false, true, true);
 }
 
 function showRestoreProgressButton(tabname, show) {
@@ -150,6 +159,14 @@ function submit() {
     return res;
 }
 
+function submit_txt2img_upscale() {
+    var res = submit(...arguments);
+
+    res[2] = selected_gallery_index();
+
+    return res;
+}
+
 function submit_img2img() {
     showSubmitButtons('img2img', false);
 
@@ -302,8 +319,6 @@ onAfterUiUpdate(function() {
     });
 
     json_elem.parentElement.style.display = "none";
-
-    setupTokenCounters();
 });
 
 onOptionsChanged(function() {
@@ -396,7 +411,7 @@ function switchWidthHeight(tabname) {
 
 var onEditTimers = {};
 
-// calls func after afterMs milliseconds has passed since the input elem has beed enited by user
+// calls func after afterMs milliseconds has passed since the input elem has been edited by user
 function onEdit(editId, elem, afterMs, func) {
     var edited = function() {
         var existingTimer = onEditTimers[editId];
diff --git a/modules/api/api.py b/modules/api/api.py
index b3d74e513a3..29fa0011a55 100644
--- a/modules/api/api.py
+++ b/modules/api/api.py
@@ -17,13 +17,13 @@
 from secrets import compare_digest
 
 import modules.shared as shared
-from modules import sd_samplers, deepbooru, sd_hijack, images, scripts, ui, postprocessing, errors, restart, shared_items, script_callbacks, generation_parameters_copypaste, sd_models
+from modules import sd_samplers, deepbooru, sd_hijack, images, scripts, ui, postprocessing, errors, restart, shared_items, script_callbacks, infotext_utils, sd_models
 from modules.api import models
 from modules.shared import opts
 from modules.processing import StableDiffusionProcessingTxt2Img, StableDiffusionProcessingImg2Img, process_images
 from modules.textual_inversion.textual_inversion import create_embedding, train_embedding
 from modules.hypernetworks.hypernetwork import create_hypernetwork, train_hypernetwork
-from PIL import PngImagePlugin, Image
+from PIL import PngImagePlugin
 from modules.sd_models_config import find_checkpoint_config_near_filename
 from modules.realesrgan_model import get_realesrgan_models
 from modules import devices
@@ -31,7 +31,7 @@
 import piexif
 import piexif.helper
 from contextlib import closing
-
+from modules.progress import create_task_id, add_task_to_queue, start_task, finish_task, current_task
 
 def script_name_to_index(name, scripts):
     try:
@@ -85,7 +85,7 @@ def decode_base64_to_image(encoding):
         headers = {'user-agent': opts.api_useragent} if opts.api_useragent else {}
         response = requests.get(encoding, timeout=30, headers=headers)
         try:
-            image = Image.open(BytesIO(response.content))
+            image = images.read(BytesIO(response.content))
             return image
         except Exception as e:
             raise HTTPException(status_code=500, detail="Invalid image url") from e
@@ -93,7 +93,7 @@ def decode_base64_to_image(encoding):
     if encoding.startswith("data:image/"):
         encoding = encoding.split(";")[1].split(",")[1]
     try:
-        image = Image.open(BytesIO(base64.b64decode(encoding)))
+        image = images.read(BytesIO(base64.b64decode(encoding)))
         return image
     except Exception as e:
         raise HTTPException(status_code=500, detail="Invalid encoded image") from e
@@ -230,6 +230,7 @@ def __init__(self, app: FastAPI, queue_lock: Lock):
         self.add_api_route("/sdapi/v1/realesrgan-models", self.get_realesrgan_models, methods=["GET"], response_model=list[models.RealesrganItem])
         self.add_api_route("/sdapi/v1/prompt-styles", self.get_prompt_styles, methods=["GET"], response_model=list[models.PromptStyleItem])
         self.add_api_route("/sdapi/v1/embeddings", self.get_embeddings, methods=["GET"], response_model=models.EmbeddingsResponse)
+        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/create/embedding", self.create_embedding, methods=["POST"], response_model=models.CreateResponse)
@@ -251,6 +252,24 @@ def __init__(self, app: FastAPI, queue_lock: Lock):
         self.default_script_arg_txt2img = []
         self.default_script_arg_img2img = []
 
+        txt2img_script_runner = scripts.scripts_txt2img
+        img2img_script_runner = scripts.scripts_img2img
+
+        if not txt2img_script_runner.scripts or not img2img_script_runner.scripts:
+            ui.create_ui()
+
+        if not txt2img_script_runner.scripts:
+            txt2img_script_runner.initialize_scripts(False)
+        if not self.default_script_arg_txt2img:
+            self.default_script_arg_txt2img = self.init_default_script_args(txt2img_script_runner)
+
+        if not img2img_script_runner.scripts:
+            img2img_script_runner.initialize_scripts(True)
+        if not self.default_script_arg_img2img:
+            self.default_script_arg_img2img = self.init_default_script_args(img2img_script_runner)
+
+
+
     def add_api_route(self, path: str, endpoint, **kwargs):
         if shared.cmd_opts.api_auth:
             return self.app.add_api_route(path, endpoint, dependencies=[Depends(self.auth)], **kwargs)
@@ -312,8 +331,13 @@ def init_default_script_args(self, script_runner):
                     script_args[script.args_from:script.args_to] = ui_default_values
         return script_args
 
-    def init_script_args(self, request, default_script_args, selectable_scripts, selectable_idx, script_runner):
+    def init_script_args(self, request, default_script_args, selectable_scripts, selectable_idx, script_runner, *, input_script_args=None):
         script_args = default_script_args.copy()
+
+        if input_script_args is not None:
+            for index, value in input_script_args.items():
+                script_args[index] = value
+
         # position 0 in script_arg is the idx+1 of the selectable script that is going to be run when using scripts.scripts_*2img.run()
         if selectable_scripts:
             script_args[selectable_scripts.args_from:selectable_scripts.args_to] = request.script_args
@@ -335,13 +359,83 @@ def init_script_args(self, request, default_script_args, selectable_scripts, sel
                         script_args[alwayson_script.args_from + idx] = request.alwayson_scripts[alwayson_script_name]["args"][idx]
         return script_args
 
+    def apply_infotext(self, request, tabname, *, script_runner=None, mentioned_script_args=None):
+        """Processes `infotext` field from the `request`, and sets other fields of the `request` according to what's in infotext.
+
+        If request already has a field set, and that field is encountered in infotext too, the value from infotext is ignored.
+
+        Additionally, fills `mentioned_script_args` dict with index: value pairs for script arguments read from infotext.
+        """
+
+        if not request.infotext:
+            return {}
+
+        possible_fields = infotext_utils.paste_fields[tabname]["fields"]
+        set_fields = request.model_dump(exclude_unset=True) if hasattr(request, "request") else request.dict(exclude_unset=True)  # pydantic v1/v2 have differenrt names for this
+        params = infotext_utils.parse_generation_parameters(request.infotext)
+
+        def get_field_value(field, params):
+            value = field.function(params) if field.function else params.get(field.label)
+            if value is None:
+                return None
+
+            if field.api in request.__fields__:
+                target_type = request.__fields__[field.api].type_
+            else:
+                target_type = type(field.component.value)
+
+            if target_type == type(None):
+                return None
+
+            if isinstance(value, dict) and value.get('__type__') == 'generic_update':  # this is a gradio.update rather than a value
+                value = value.get('value')
+
+            if value is not None and not isinstance(value, target_type):
+                value = target_type(value)
+
+            return value
+
+        for field in possible_fields:
+            if not field.api:
+                continue
+
+            if field.api in set_fields:
+                continue
+
+            value = get_field_value(field, params)
+            if value is not None:
+                setattr(request, field.api, value)
+
+        if request.override_settings is None:
+            request.override_settings = {}
+
+        overridden_settings = infotext_utils.get_override_settings(params)
+        for _, setting_name, value in overridden_settings:
+            if setting_name not in request.override_settings:
+                request.override_settings[setting_name] = value
+
+        if script_runner is not None and mentioned_script_args is not None:
+            indexes = {v: i for i, v in enumerate(script_runner.inputs)}
+            script_fields = ((field, indexes[field.component]) for field in possible_fields if field.component in indexes)
+
+            for field, index in script_fields:
+                value = get_field_value(field, params)
+
+                if value is None:
+                    continue
+
+                mentioned_script_args[index] = value
+
+        return params
+
     def text2imgapi(self, txt2imgreq: models.StableDiffusionTxt2ImgProcessingAPI):
+        task_id = txt2imgreq.force_task_id or create_task_id("txt2img")
+
         script_runner = scripts.scripts_txt2img
-        if not script_runner.scripts:
-            script_runner.initialize_scripts(False)
-            ui.create_ui()
-        if not self.default_script_arg_txt2img:
-            self.default_script_arg_txt2img = self.init_default_script_args(script_runner)
+
+        infotext_script_args = {}
+        self.apply_infotext(txt2imgreq, "txt2img", script_runner=script_runner, mentioned_script_args=infotext_script_args)
+
         selectable_scripts, selectable_script_idx = self.get_selectable_script(txt2imgreq.script_name, script_runner)
 
         populate = txt2imgreq.copy(update={  # Override __init__ params
@@ -356,12 +450,15 @@ def text2imgapi(self, txt2imgreq: models.StableDiffusionTxt2ImgProcessingAPI):
         args.pop('script_name', None)
         args.pop('script_args', None) # will refeed them to the pipeline directly after initializing them
         args.pop('alwayson_scripts', None)
+        args.pop('infotext', None)
 
-        script_args = self.init_script_args(txt2imgreq, self.default_script_arg_txt2img, selectable_scripts, selectable_script_idx, script_runner)
+        script_args = self.init_script_args(txt2imgreq, self.default_script_arg_txt2img, selectable_scripts, selectable_script_idx, script_runner, input_script_args=infotext_script_args)
 
         send_images = args.pop('send_images', True)
         args.pop('save_images', None)
 
+        add_task_to_queue(task_id)
+
         with self.queue_lock:
             with closing(StableDiffusionProcessingTxt2Img(sd_model=shared.sd_model, **args)) as p:
                 p.is_api = True
@@ -371,12 +468,14 @@ def text2imgapi(self, txt2imgreq: models.StableDiffusionTxt2ImgProcessingAPI):
 
                 try:
                     shared.state.begin(job="scripts_txt2img")
+                    start_task(task_id)
                     if selectable_scripts is not None:
                         p.script_args = script_args
                         processed = scripts.scripts_txt2img.run(p, *p.script_args) # Need to pass args as list here
                     else:
                         p.script_args = tuple(script_args) # Need to pass args as tuple here
                         processed = process_images(p)
+                    finish_task(task_id)
                 finally:
                     shared.state.end()
                     shared.total_tqdm.clear()
@@ -386,6 +485,8 @@ def text2imgapi(self, txt2imgreq: models.StableDiffusionTxt2ImgProcessingAPI):
         return models.TextToImageResponse(images=b64images, parameters=vars(txt2imgreq), info=processed.js())
 
     def img2imgapi(self, img2imgreq: models.StableDiffusionImg2ImgProcessingAPI):
+        task_id = img2imgreq.force_task_id or create_task_id("img2img")
+
         init_images = img2imgreq.init_images
         if init_images is None:
             raise HTTPException(status_code=404, detail="Init image not found")
@@ -395,11 +496,10 @@ def img2imgapi(self, img2imgreq: models.StableDiffusionImg2ImgProcessingAPI):
             mask = decode_base64_to_image(mask)
 
         script_runner = scripts.scripts_img2img
-        if not script_runner.scripts:
-            script_runner.initialize_scripts(True)
-            ui.create_ui()
-        if not self.default_script_arg_img2img:
-            self.default_script_arg_img2img = self.init_default_script_args(script_runner)
+
+        infotext_script_args = {}
+        self.apply_infotext(img2imgreq, "img2img", script_runner=script_runner, mentioned_script_args=infotext_script_args)
+
         selectable_scripts, selectable_script_idx = self.get_selectable_script(img2imgreq.script_name, script_runner)
 
         populate = img2imgreq.copy(update={  # Override __init__ params
@@ -416,12 +516,15 @@ def img2imgapi(self, img2imgreq: models.StableDiffusionImg2ImgProcessingAPI):
         args.pop('script_name', None)
         args.pop('script_args', None)  # will refeed them to the pipeline directly after initializing them
         args.pop('alwayson_scripts', None)
+        args.pop('infotext', None)
 
-        script_args = self.init_script_args(img2imgreq, self.default_script_arg_img2img, selectable_scripts, selectable_script_idx, script_runner)
+        script_args = self.init_script_args(img2imgreq, self.default_script_arg_img2img, selectable_scripts, selectable_script_idx, script_runner, input_script_args=infotext_script_args)
 
         send_images = args.pop('send_images', True)
         args.pop('save_images', None)
 
+        add_task_to_queue(task_id)
+
         with self.queue_lock:
             with closing(StableDiffusionProcessingImg2Img(sd_model=shared.sd_model, **args)) as p:
                 p.init_images = [decode_base64_to_image(x) for x in init_images]
@@ -432,12 +535,14 @@ def img2imgapi(self, img2imgreq: models.StableDiffusionImg2ImgProcessingAPI):
 
                 try:
                     shared.state.begin(job="scripts_img2img")
+                    start_task(task_id)
                     if selectable_scripts is not None:
                         p.script_args = script_args
                         processed = scripts.scripts_img2img.run(p, *p.script_args) # Need to pass args as list here
                     else:
                         p.script_args = tuple(script_args) # Need to pass args as tuple here
                         processed = process_images(p)
+                    finish_task(task_id)
                 finally:
                     shared.state.end()
                     shared.total_tqdm.clear()
@@ -480,7 +585,7 @@ def pnginfoapi(self, req: models.PNGInfoRequest):
         if geninfo is None:
             geninfo = ""
 
-        params = generation_parameters_copypaste.parse_generation_parameters(geninfo)
+        params = infotext_utils.parse_generation_parameters(geninfo)
         script_callbacks.infotext_pasted_callback(geninfo, params)
 
         return models.PNGInfoResponse(info=geninfo, items=items, parameters=params)
@@ -511,7 +616,7 @@ def progressapi(self, req: models.ProgressRequest = Depends()):
         if shared.state.current_image and not req.skip_current_image:
             current_image = encode_pil_to_base64(shared.state.current_image)
 
-        return models.ProgressResponse(progress=progress, eta_relative=eta_relative, state=shared.state.dict(), current_image=current_image, textinfo=shared.state.textinfo)
+        return models.ProgressResponse(progress=progress, eta_relative=eta_relative, state=shared.state.dict(), current_image=current_image, textinfo=shared.state.textinfo, current_task=current_task)
 
     def interrogateapi(self, interrogatereq: models.InterrogateRequest):
         image_b64 = interrogatereq.image
@@ -643,6 +748,10 @@ def convert_embeddings(embeddings):
             "skipped": convert_embeddings(db.skipped_embeddings),
         }
 
+    def refresh_embeddings(self):
+        with self.queue_lock:
+            sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings(force_reload=True)
+
     def refresh_checkpoints(self):
         with self.queue_lock:
             shared.refresh_checkpoints()
@@ -775,7 +884,15 @@ def get_extensions_list(self):
 
     def launch(self, server_name, port, root_path):
         self.app.include_router(self.router)
-        uvicorn.run(self.app, host=server_name, port=port, timeout_keep_alive=shared.cmd_opts.timeout_keep_alive, root_path=root_path)
+        uvicorn.run(
+            self.app,
+            host=server_name,
+            port=port,
+            timeout_keep_alive=shared.cmd_opts.timeout_keep_alive,
+            root_path=root_path,
+            ssl_keyfile=shared.cmd_opts.tls_keyfile,
+            ssl_certfile=shared.cmd_opts.tls_certfile
+        )
 
     def kill_webui(self):
         restart.stop_program()
diff --git a/modules/api/models.py b/modules/api/models.py
index 33894b3e694..16edf11cf83 100644
--- a/modules/api/models.py
+++ b/modules/api/models.py
@@ -107,6 +107,8 @@ def generate_model(self):
         {"key": "send_images", "type": bool, "default": True},
         {"key": "save_images", "type": bool, "default": False},
         {"key": "alwayson_scripts", "type": dict, "default": {}},
+        {"key": "force_task_id", "type": str, "default": None},
+        {"key": "infotext", "type": str, "default": None},
     ]
 ).generate_model()
 
@@ -124,6 +126,8 @@ def generate_model(self):
         {"key": "send_images", "type": bool, "default": True},
         {"key": "save_images", "type": bool, "default": False},
         {"key": "alwayson_scripts", "type": dict, "default": {}},
+        {"key": "force_task_id", "type": str, "default": None},
+        {"key": "infotext", "type": str, "default": None},
     ]
 ).generate_model()
 
diff --git a/modules/cache.py b/modules/cache.py
index 2d37e7b99d5..a9822a0eb21 100644
--- a/modules/cache.py
+++ b/modules/cache.py
@@ -62,16 +62,15 @@ def cache(subsection):
     if cache_data is None:
         with cache_lock:
             if cache_data is None:
-                if not os.path.isfile(cache_filename):
+                try:
+                    with open(cache_filename, "r", encoding="utf8") as file:
+                        cache_data = json.load(file)
+                except FileNotFoundError:
+                    cache_data = {}
+                except Exception:
+                    os.replace(cache_filename, os.path.join(script_path, "tmp", "cache.json"))
+                    print('[ERROR] issue occurred while trying to read cache.json, move current cache to tmp/cache.json and create new cache')
                     cache_data = {}
-                else:
-                    try:
-                        with open(cache_filename, "r", encoding="utf8") as file:
-                            cache_data = json.load(file)
-                    except Exception:
-                        os.replace(cache_filename, os.path.join(script_path, "tmp", "cache.json"))
-                        print('[ERROR] issue occurred while trying to read cache.json, move current cache to tmp/cache.json and create new cache')
-                        cache_data = {}
 
     s = cache_data.get(subsection, {})
     cache_data[subsection] = s
diff --git a/modules/call_queue.py b/modules/call_queue.py
index ddf0d57383c..b50931bcdb9 100644
--- a/modules/call_queue.py
+++ b/modules/call_queue.py
@@ -78,6 +78,7 @@ def f(*args, extra_outputs_array=extra_outputs, **kwargs):
 
         shared.state.skipped = False
         shared.state.interrupted = False
+        shared.state.stopping_generation = False
         shared.state.job_count = 0
 
         if not add_stats:
@@ -99,8 +100,8 @@ def f(*args, extra_outputs_array=extra_outputs, **kwargs):
             sys_pct = sys_peak/max(sys_total, 1) * 100
 
             toltip_a = "Active: peak amount of video memory used during generation (excluding cached data)"
-            toltip_r = "Reserved: total amout of video memory allocated by the Torch library "
-            toltip_sys = "System: peak amout of video memory allocated by all running programs, out of total capacity"
+            toltip_r = "Reserved: total amount of video memory allocated by the Torch library "
+            toltip_sys = "System: peak amount of video memory allocated by all running programs, out of total capacity"
 
             text_a = f"A: {active_peak/1024:.2f} GB"
             text_r = f"R: {reserved_peak/1024:.2f} GB"
diff --git a/modules/cmd_args.py b/modules/cmd_args.py
index da93eb2669f..bf355303151 100644
--- a/modules/cmd_args.py
+++ b/modules/cmd_args.py
@@ -1,7 +1,7 @@
 import argparse
 import json
 import os
-from modules.paths_internal import models_path, script_path, data_path, extensions_dir, extensions_builtin_dir, sd_default_config, sd_model_file  # noqa: F401
+from modules.paths_internal import normalized_filepath, models_path, script_path, data_path, extensions_dir, extensions_builtin_dir, sd_default_config, sd_model_file  # noqa: F401
 
 parser = argparse.ArgumentParser()
 
@@ -19,21 +19,21 @@
 parser.add_argument("--dump-sysinfo", action='store_true', help="launch.py argument: dump limited sysinfo file (without information about extensions, options) to disk and quit")
 parser.add_argument("--loglevel", type=str, help="log level; one of: CRITICAL, ERROR, WARNING, INFO, DEBUG", default=None)
 parser.add_argument("--do-not-download-clip", action='store_true', help="do not download CLIP model even if it's not included in the checkpoint")
-parser.add_argument("--data-dir", type=str, default=os.path.dirname(os.path.dirname(os.path.realpath(__file__))), help="base path where all user data is stored")
-parser.add_argument("--config", type=str, default=sd_default_config, help="path to config which constructs model",)
-parser.add_argument("--ckpt", type=str, default=sd_model_file, help="path to checkpoint of stable diffusion model; if specified, this checkpoint will be added to the list of checkpoints and loaded",)
-parser.add_argument("--ckpt-dir", type=str, default=None, help="Path to directory with stable diffusion checkpoints")
-parser.add_argument("--vae-dir", type=str, default=None, help="Path to directory with VAE files")
-parser.add_argument("--gfpgan-dir", type=str, help="GFPGAN directory", default=('./src/gfpgan' if os.path.exists('./src/gfpgan') else './GFPGAN'))
-parser.add_argument("--gfpgan-model", type=str, help="GFPGAN model file name", default=None)
+parser.add_argument("--data-dir", type=normalized_filepath, default=os.path.dirname(os.path.dirname(os.path.realpath(__file__))), help="base path where all user data is stored")
+parser.add_argument("--config", type=normalized_filepath, default=sd_default_config, help="path to config which constructs model",)
+parser.add_argument("--ckpt", type=normalized_filepath, default=sd_model_file, help="path to checkpoint of stable diffusion model; if specified, this checkpoint will be added to the list of checkpoints and loaded",)
+parser.add_argument("--ckpt-dir", type=normalized_filepath, default=None, help="Path to directory with stable diffusion checkpoints")
+parser.add_argument("--vae-dir", type=normalized_filepath, default=None, help="Path to directory with VAE files")
+parser.add_argument("--gfpgan-dir", type=normalized_filepath, help="GFPGAN directory", default=('./src/gfpgan' if os.path.exists('./src/gfpgan') else './GFPGAN'))
+parser.add_argument("--gfpgan-model", type=normalized_filepath, help="GFPGAN model file name", default=None)
 parser.add_argument("--no-half", action='store_true', help="do not switch the model to 16-bit floats")
 parser.add_argument("--no-half-vae", action='store_true', help="do not switch the VAE model to 16-bit floats")
 parser.add_argument("--no-progressbar-hiding", action='store_true', help="do not hide progressbar in gradio UI (we hide it because it slows down ML if you have hardware acceleration in browser)")
 parser.add_argument("--max-batch-count", type=int, default=16, help="maximum batch count value for the UI")
-parser.add_argument("--embeddings-dir", type=str, default=os.path.join(data_path, 'embeddings'), help="embeddings directory for textual inversion (default: embeddings)")
-parser.add_argument("--textual-inversion-templates-dir", type=str, default=os.path.join(script_path, 'textual_inversion_templates'), help="directory with textual inversion templates")
-parser.add_argument("--hypernetwork-dir", type=str, default=os.path.join(models_path, 'hypernetworks'), help="hypernetwork directory")
-parser.add_argument("--localizations-dir", type=str, default=os.path.join(script_path, 'localizations'), help="localizations directory")
+parser.add_argument("--embeddings-dir", type=normalized_filepath, default=os.path.join(data_path, 'embeddings'), help="embeddings directory for textual inversion (default: embeddings)")
+parser.add_argument("--textual-inversion-templates-dir", type=normalized_filepath, default=os.path.join(script_path, 'textual_inversion_templates'), help="directory with textual inversion templates")
+parser.add_argument("--hypernetwork-dir", type=normalized_filepath, default=os.path.join(models_path, 'hypernetworks'), help="hypernetwork directory")
+parser.add_argument("--localizations-dir", type=normalized_filepath, default=os.path.join(script_path, 'localizations'), help="localizations directory")
 parser.add_argument("--allow-code", action='store_true', help="allow custom script execution from webui")
 parser.add_argument("--medvram", action='store_true', help="enable stable diffusion model optimizations for sacrificing a little speed for low VRM usage")
 parser.add_argument("--medvram-sdxl", action='store_true', help="enable --medvram optimization just for SDXL models")
@@ -48,12 +48,13 @@
 parser.add_argument("--ngrok-region", type=str, help="does not do anything.", default="")
 parser.add_argument("--ngrok-options", type=json.loads, help='The options to pass to ngrok in JSON format, e.g.: \'{"authtoken_from_env":true, "basic_auth":"user:password", "oauth_provider":"google", "oauth_allow_emails":"user@asdf.com"}\'', default=dict())
 parser.add_argument("--enable-insecure-extension-access", action='store_true', help="enable extensions tab regardless of other options")
-parser.add_argument("--codeformer-models-path", type=str, help="Path to directory with codeformer model file(s).", default=os.path.join(models_path, 'Codeformer'))
-parser.add_argument("--gfpgan-models-path", type=str, help="Path to directory with GFPGAN model file(s).", default=os.path.join(models_path, 'GFPGAN'))
-parser.add_argument("--esrgan-models-path", type=str, help="Path to directory with ESRGAN model file(s).", default=os.path.join(models_path, 'ESRGAN'))
-parser.add_argument("--bsrgan-models-path", type=str, help="Path to directory with BSRGAN model file(s).", default=os.path.join(models_path, 'BSRGAN'))
-parser.add_argument("--realesrgan-models-path", type=str, help="Path to directory with RealESRGAN model file(s).", default=os.path.join(models_path, 'RealESRGAN'))
-parser.add_argument("--clip-models-path", type=str, help="Path to directory with CLIP model file(s).", default=None)
+parser.add_argument("--codeformer-models-path", type=normalized_filepath, help="Path to directory with codeformer model file(s).", default=os.path.join(models_path, 'Codeformer'))
+parser.add_argument("--gfpgan-models-path", type=normalized_filepath, help="Path to directory with GFPGAN model file(s).", default=os.path.join(models_path, 'GFPGAN'))
+parser.add_argument("--esrgan-models-path", type=normalized_filepath, help="Path to directory with ESRGAN model file(s).", default=os.path.join(models_path, 'ESRGAN'))
+parser.add_argument("--bsrgan-models-path", type=normalized_filepath, help="Path to directory with BSRGAN model file(s).", default=os.path.join(models_path, 'BSRGAN'))
+parser.add_argument("--realesrgan-models-path", type=normalized_filepath, help="Path to directory with RealESRGAN model file(s).", default=os.path.join(models_path, 'RealESRGAN'))
+parser.add_argument("--dat-models-path", type=normalized_filepath, help="Path to directory with DAT model file(s).", default=os.path.join(models_path, 'DAT'))
+parser.add_argument("--clip-models-path", type=normalized_filepath, help="Path to directory with CLIP model file(s).", default=None)
 parser.add_argument("--xformers", action='store_true', help="enable xformers for cross attention layers")
 parser.add_argument("--force-enable-xformers", action='store_true', help="enable xformers for cross attention layers regardless of whether the checking code thinks you can run it; do not make bug reports if this fails to work")
 parser.add_argument("--xformers-flash-attention", action='store_true', help="enable xformers with Flash Attention to improve reproducibility (supported for SD2.x or variant only)")
@@ -77,22 +78,24 @@
 parser.add_argument("--show-negative-prompt", action='store_true', help="does not do anything", default=False)
 parser.add_argument("--ui-config-file", type=str, help="filename to use for ui configuration", default=os.path.join(data_path, 'ui-config.json'))
 parser.add_argument("--hide-ui-dir-config", action='store_true', help="hide directory configuration from webui", default=False)
-parser.add_argument("--freeze-settings", action='store_true', help="disable editing settings", default=False)
+parser.add_argument("--freeze-settings", action='store_true', help="disable editing of all settings globally", default=False)
+parser.add_argument("--freeze-settings-in-sections", type=str, help='disable editing settings in specific sections of the settings page by specifying a comma-delimited list such like "saving-images,upscaling". The list of setting names can be found in the modules/shared_options.py file', default=None)
+parser.add_argument("--freeze-specific-settings", type=str, help='disable editing of individual settings by specifying a comma-delimited list like "samples_save,samples_format". The list of setting names can be found in the config.json file', default=None)
 parser.add_argument("--ui-settings-file", type=str, help="filename to use for ui settings", default=os.path.join(data_path, 'config.json'))
 parser.add_argument("--gradio-debug",  action='store_true', help="launch gradio with --debug option")
 parser.add_argument("--gradio-auth", type=str, help='set gradio authentication like "username:password"; or comma-delimit multiple like "u1:p1,u2:p2,u3:p3"', default=None)
-parser.add_argument("--gradio-auth-path", type=str, help='set gradio authentication file path ex. "/path/to/auth/file" same auth format as --gradio-auth', default=None)
+parser.add_argument("--gradio-auth-path", type=normalized_filepath, help='set gradio authentication file path ex. "/path/to/auth/file" same auth format as --gradio-auth', default=None)
 parser.add_argument("--gradio-img2img-tool", type=str, help='does not do anything')
 parser.add_argument("--gradio-inpaint-tool", type=str, help="does not do anything")
 parser.add_argument("--gradio-allowed-path", action='append', help="add path to gradio's allowed_paths, make it possible to serve files from it", default=[data_path])
 parser.add_argument("--opt-channelslast", action='store_true', help="change memory type for stable diffusion to channels last")
-parser.add_argument("--styles-file", type=str, help="filename to use for styles", default=os.path.join(data_path, 'styles.csv'))
+parser.add_argument("--styles-file", type=str, action='append', help="path or wildcard path of styles files, allow multiple entries.", default=[])
 parser.add_argument("--autolaunch", action='store_true', help="open the webui URL in the system's default browser upon launch", default=False)
 parser.add_argument("--theme", type=str, help="launches the UI with light or dark theme", default=None)
 parser.add_argument("--use-textbox-seed", action='store_true', help="use textbox for seeds in UI (no up/down, but possible to input long seeds)", default=False)
 parser.add_argument("--disable-console-progressbars", action='store_true', help="do not output progressbars to console", default=False)
 parser.add_argument("--enable-console-prompts", action='store_true', help="does not do anything", default=False)  # Legacy compatibility, use as default value shared.opts.enable_console_prompts
-parser.add_argument('--vae-path', type=str, help='Checkpoint to use as VAE; setting this argument disables all settings related to VAE', default=None)
+parser.add_argument('--vae-path', type=normalized_filepath, help='Checkpoint to use as VAE; setting this argument disables all settings related to VAE', default=None)
 parser.add_argument("--disable-safe-unpickle", action='store_true', help="disable checking pytorch models for malicious code", default=False)
 parser.add_argument("--api", action='store_true', help="use api=True to launch the API together with the webui (use --nowebui instead for only the API)")
 parser.add_argument("--api-auth", type=str, help='Set authentication for API like "username:password"; or comma-delimit multiple like "u1:p1,u2:p2,u3:p3"', default=None)
@@ -118,4 +121,6 @@
 parser.add_argument('--timeout-keep-alive', type=int, default=30, help='set timeout_keep_alive for uvicorn')
 parser.add_argument("--disable-all-extensions", action='store_true', help="prevent all extensions from running regardless of any other settings", default=False)
 parser.add_argument("--disable-extra-extensions", action='store_true', help="prevent all extensions except built-in from running regardless of any other settings", default=False)
-parser.add_argument("--skip-load-model-at-start", action='store_true', help="if load a model at web start, only take effect when --nowebui", )
+parser.add_argument("--skip-load-model-at-start", action='store_true', help="if load a model at web start, only take effect when --nowebui")
+parser.add_argument("--unix-filenames-sanitization", action='store_true', help="allow any symbols except '/' in filenames. May conflict with your browser and file system")
+parser.add_argument("--filenames-max-length", type=int, default=128, help='maximal length of filenames of saved images. If you override it, it can conflict with your file system')
diff --git a/modules/codeformer/codeformer_arch.py b/modules/codeformer/codeformer_arch.py
deleted file mode 100644
index 12db6814268..00000000000
--- a/modules/codeformer/codeformer_arch.py
+++ /dev/null
@@ -1,276 +0,0 @@
-# this file is copied from CodeFormer repository. Please see comment in modules/codeformer_model.py
-
-import math
-import torch
-from torch import nn, Tensor
-import torch.nn.functional as F
-from typing import Optional
-
-from modules.codeformer.vqgan_arch import VQAutoEncoder, ResBlock
-from basicsr.utils.registry import ARCH_REGISTRY
-
-def calc_mean_std(feat, eps=1e-5):
-    """Calculate mean and std for adaptive_instance_normalization.
-
-    Args:
-        feat (Tensor): 4D tensor.
-        eps (float): A small value added to the variance to avoid
-            divide-by-zero. Default: 1e-5.
-    """
-    size = feat.size()
-    assert len(size) == 4, 'The input feature should be 4D tensor.'
-    b, c = size[:2]
-    feat_var = feat.view(b, c, -1).var(dim=2) + eps
-    feat_std = feat_var.sqrt().view(b, c, 1, 1)
-    feat_mean = feat.view(b, c, -1).mean(dim=2).view(b, c, 1, 1)
-    return feat_mean, feat_std
-
-
-def adaptive_instance_normalization(content_feat, style_feat):
-    """Adaptive instance normalization.
-
-    Adjust the reference features to have the similar color and illuminations
-    as those in the degradate features.
-
-    Args:
-        content_feat (Tensor): The reference feature.
-        style_feat (Tensor): The degradate features.
-    """
-    size = content_feat.size()
-    style_mean, style_std = calc_mean_std(style_feat)
-    content_mean, content_std = calc_mean_std(content_feat)
-    normalized_feat = (content_feat - content_mean.expand(size)) / content_std.expand(size)
-    return normalized_feat * style_std.expand(size) + style_mean.expand(size)
-
-
-class PositionEmbeddingSine(nn.Module):
-    """
-    This is a more standard version of the position embedding, very similar to the one
-    used by the Attention is all you need paper, generalized to work on images.
-    """
-
-    def __init__(self, num_pos_feats=64, temperature=10000, normalize=False, scale=None):
-        super().__init__()
-        self.num_pos_feats = num_pos_feats
-        self.temperature = temperature
-        self.normalize = normalize
-        if scale is not None and normalize is False:
-            raise ValueError("normalize should be True if scale is passed")
-        if scale is None:
-            scale = 2 * math.pi
-        self.scale = scale
-
-    def forward(self, x, mask=None):
-        if mask is None:
-            mask = torch.zeros((x.size(0), x.size(2), x.size(3)), device=x.device, dtype=torch.bool)
-        not_mask = ~mask
-        y_embed = not_mask.cumsum(1, dtype=torch.float32)
-        x_embed = not_mask.cumsum(2, dtype=torch.float32)
-        if self.normalize:
-            eps = 1e-6
-            y_embed = y_embed / (y_embed[:, -1:, :] + eps) * self.scale
-            x_embed = x_embed / (x_embed[:, :, -1:] + eps) * self.scale
-
-        dim_t = torch.arange(self.num_pos_feats, dtype=torch.float32, device=x.device)
-        dim_t = self.temperature ** (2 * (dim_t // 2) / self.num_pos_feats)
-
-        pos_x = x_embed[:, :, :, None] / dim_t
-        pos_y = y_embed[:, :, :, None] / dim_t
-        pos_x = torch.stack(
-            (pos_x[:, :, :, 0::2].sin(), pos_x[:, :, :, 1::2].cos()), dim=4
-        ).flatten(3)
-        pos_y = torch.stack(
-            (pos_y[:, :, :, 0::2].sin(), pos_y[:, :, :, 1::2].cos()), dim=4
-        ).flatten(3)
-        pos = torch.cat((pos_y, pos_x), dim=3).permute(0, 3, 1, 2)
-        return pos
-
-def _get_activation_fn(activation):
-    """Return an activation function given a string"""
-    if activation == "relu":
-        return F.relu
-    if activation == "gelu":
-        return F.gelu
-    if activation == "glu":
-        return F.glu
-    raise RuntimeError(F"activation should be relu/gelu, not {activation}.")
-
-
-class TransformerSALayer(nn.Module):
-    def __init__(self, embed_dim, nhead=8, dim_mlp=2048, dropout=0.0, activation="gelu"):
-        super().__init__()
-        self.self_attn = nn.MultiheadAttention(embed_dim, nhead, dropout=dropout)
-        # Implementation of Feedforward model - MLP
-        self.linear1 = nn.Linear(embed_dim, dim_mlp)
-        self.dropout = nn.Dropout(dropout)
-        self.linear2 = nn.Linear(dim_mlp, embed_dim)
-
-        self.norm1 = nn.LayerNorm(embed_dim)
-        self.norm2 = nn.LayerNorm(embed_dim)
-        self.dropout1 = nn.Dropout(dropout)
-        self.dropout2 = nn.Dropout(dropout)
-
-        self.activation = _get_activation_fn(activation)
-
-    def with_pos_embed(self, tensor, pos: Optional[Tensor]):
-        return tensor if pos is None else tensor + pos
-
-    def forward(self, tgt,
-                tgt_mask: Optional[Tensor] = None,
-                tgt_key_padding_mask: Optional[Tensor] = None,
-                query_pos: Optional[Tensor] = None):
-
-        # self attention
-        tgt2 = self.norm1(tgt)
-        q = k = self.with_pos_embed(tgt2, query_pos)
-        tgt2 = self.self_attn(q, k, value=tgt2, attn_mask=tgt_mask,
-                              key_padding_mask=tgt_key_padding_mask)[0]
-        tgt = tgt + self.dropout1(tgt2)
-
-        # ffn
-        tgt2 = self.norm2(tgt)
-        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt2))))
-        tgt = tgt + self.dropout2(tgt2)
-        return tgt
-
-class Fuse_sft_block(nn.Module):
-    def __init__(self, in_ch, out_ch):
-        super().__init__()
-        self.encode_enc = ResBlock(2*in_ch, out_ch)
-
-        self.scale = nn.Sequential(
-                    nn.Conv2d(in_ch, out_ch, kernel_size=3, padding=1),
-                    nn.LeakyReLU(0.2, True),
-                    nn.Conv2d(out_ch, out_ch, kernel_size=3, padding=1))
-
-        self.shift = nn.Sequential(
-                    nn.Conv2d(in_ch, out_ch, kernel_size=3, padding=1),
-                    nn.LeakyReLU(0.2, True),
-                    nn.Conv2d(out_ch, out_ch, kernel_size=3, padding=1))
-
-    def forward(self, enc_feat, dec_feat, w=1):
-        enc_feat = self.encode_enc(torch.cat([enc_feat, dec_feat], dim=1))
-        scale = self.scale(enc_feat)
-        shift = self.shift(enc_feat)
-        residual = w * (dec_feat * scale + shift)
-        out = dec_feat + residual
-        return out
-
-
-@ARCH_REGISTRY.register()
-class CodeFormer(VQAutoEncoder):
-    def __init__(self, dim_embd=512, n_head=8, n_layers=9,
-                codebook_size=1024, latent_size=256,
-                connect_list=('32', '64', '128', '256'),
-                fix_modules=('quantize', 'generator')):
-        super(CodeFormer, self).__init__(512, 64, [1, 2, 2, 4, 4, 8], 'nearest',2, [16], codebook_size)
-
-        if fix_modules is not None:
-            for module in fix_modules:
-                for param in getattr(self, module).parameters():
-                    param.requires_grad = False
-
-        self.connect_list = connect_list
-        self.n_layers = n_layers
-        self.dim_embd = dim_embd
-        self.dim_mlp = dim_embd*2
-
-        self.position_emb = nn.Parameter(torch.zeros(latent_size, self.dim_embd))
-        self.feat_emb = nn.Linear(256, self.dim_embd)
-
-        # transformer
-        self.ft_layers = nn.Sequential(*[TransformerSALayer(embed_dim=dim_embd, nhead=n_head, dim_mlp=self.dim_mlp, dropout=0.0)
-                                    for _ in range(self.n_layers)])
-
-        # logits_predict head
-        self.idx_pred_layer = nn.Sequential(
-            nn.LayerNorm(dim_embd),
-            nn.Linear(dim_embd, codebook_size, bias=False))
-
-        self.channels = {
-            '16': 512,
-            '32': 256,
-            '64': 256,
-            '128': 128,
-            '256': 128,
-            '512': 64,
-        }
-
-        # after second residual block for > 16, before attn layer for ==16
-        self.fuse_encoder_block = {'512':2, '256':5, '128':8, '64':11, '32':14, '16':18}
-        # after first residual block for > 16, before attn layer for ==16
-        self.fuse_generator_block = {'16':6, '32': 9, '64':12, '128':15, '256':18, '512':21}
-
-        # fuse_convs_dict
-        self.fuse_convs_dict = nn.ModuleDict()
-        for f_size in self.connect_list:
-            in_ch = self.channels[f_size]
-            self.fuse_convs_dict[f_size] = Fuse_sft_block(in_ch, in_ch)
-
-    def _init_weights(self, module):
-        if isinstance(module, (nn.Linear, nn.Embedding)):
-            module.weight.data.normal_(mean=0.0, std=0.02)
-            if isinstance(module, nn.Linear) and module.bias is not None:
-                module.bias.data.zero_()
-        elif isinstance(module, nn.LayerNorm):
-            module.bias.data.zero_()
-            module.weight.data.fill_(1.0)
-
-    def forward(self, x, w=0, detach_16=True, code_only=False, adain=False):
-        # ################### Encoder #####################
-        enc_feat_dict = {}
-        out_list = [self.fuse_encoder_block[f_size] for f_size in self.connect_list]
-        for i, block in enumerate(self.encoder.blocks):
-            x = block(x)
-            if i in out_list:
-                enc_feat_dict[str(x.shape[-1])] = x.clone()
-
-        lq_feat = x
-        # ################# Transformer ###################
-        # quant_feat, codebook_loss, quant_stats = self.quantize(lq_feat)
-        pos_emb = self.position_emb.unsqueeze(1).repeat(1,x.shape[0],1)
-        # BCHW -> BC(HW) -> (HW)BC
-        feat_emb = self.feat_emb(lq_feat.flatten(2).permute(2,0,1))
-        query_emb = feat_emb
-        # Transformer encoder
-        for layer in self.ft_layers:
-            query_emb = layer(query_emb, query_pos=pos_emb)
-
-        # output logits
-        logits = self.idx_pred_layer(query_emb) # (hw)bn
-        logits = logits.permute(1,0,2) # (hw)bn -> b(hw)n
-
-        if code_only: # for training stage II
-          # logits doesn't need softmax before cross_entropy loss
-            return logits, lq_feat
-
-        # ################# Quantization ###################
-        # if self.training:
-        #     quant_feat = torch.einsum('btn,nc->btc', [soft_one_hot, self.quantize.embedding.weight])
-        #     # b(hw)c -> bc(hw) -> bchw
-        #     quant_feat = quant_feat.permute(0,2,1).view(lq_feat.shape)
-        # ------------
-        soft_one_hot = F.softmax(logits, dim=2)
-        _, top_idx = torch.topk(soft_one_hot, 1, dim=2)
-        quant_feat = self.quantize.get_codebook_feat(top_idx, shape=[x.shape[0],16,16,256])
-        # preserve gradients
-        # quant_feat = lq_feat + (quant_feat - lq_feat).detach()
-
-        if detach_16:
-            quant_feat = quant_feat.detach() # for training stage III
-        if adain:
-            quant_feat = adaptive_instance_normalization(quant_feat, lq_feat)
-
-        # ################## Generator ####################
-        x = quant_feat
-        fuse_list = [self.fuse_generator_block[f_size] for f_size in self.connect_list]
-
-        for i, block in enumerate(self.generator.blocks):
-            x = block(x)
-            if i in fuse_list: # fuse after i-th block
-                f_size = str(x.shape[-1])
-                if w>0:
-                    x = self.fuse_convs_dict[f_size](enc_feat_dict[f_size].detach(), x, w)
-        out = x
-        # logits doesn't need softmax before cross_entropy loss
-        return out, logits, lq_feat
diff --git a/modules/codeformer/vqgan_arch.py b/modules/codeformer/vqgan_arch.py
deleted file mode 100644
index 09ee6660dc5..00000000000
--- a/modules/codeformer/vqgan_arch.py
+++ /dev/null
@@ -1,435 +0,0 @@
-# this file is copied from CodeFormer repository. Please see comment in modules/codeformer_model.py
-
-'''
-VQGAN code, adapted from the original created by the Unleashing Transformers authors:
-https://github.com/samb-t/unleashing-transformers/blob/master/models/vqgan.py
-
-'''
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-from basicsr.utils import get_root_logger
-from basicsr.utils.registry import ARCH_REGISTRY
-
-def normalize(in_channels):
-    return torch.nn.GroupNorm(num_groups=32, num_channels=in_channels, eps=1e-6, affine=True)
-
-
-@torch.jit.script
-def swish(x):
-    return x*torch.sigmoid(x)
-
-
-#  Define VQVAE classes
-class VectorQuantizer(nn.Module):
-    def __init__(self, codebook_size, emb_dim, beta):
-        super(VectorQuantizer, self).__init__()
-        self.codebook_size = codebook_size  # number of embeddings
-        self.emb_dim = emb_dim  # dimension of embedding
-        self.beta = beta  # commitment cost used in loss term, beta * ||z_e(x)-sg[e]||^2
-        self.embedding = nn.Embedding(self.codebook_size, self.emb_dim)
-        self.embedding.weight.data.uniform_(-1.0 / self.codebook_size, 1.0 / self.codebook_size)
-
-    def forward(self, z):
-        # reshape z -> (batch, height, width, channel) and flatten
-        z = z.permute(0, 2, 3, 1).contiguous()
-        z_flattened = z.view(-1, self.emb_dim)
-
-        # distances from z to embeddings e_j (z - e)^2 = z^2 + e^2 - 2 e * z
-        d = (z_flattened ** 2).sum(dim=1, keepdim=True) + (self.embedding.weight**2).sum(1) - \
-            2 * torch.matmul(z_flattened, self.embedding.weight.t())
-
-        mean_distance = torch.mean(d)
-        # find closest encodings
-        # min_encoding_indices = torch.argmin(d, dim=1).unsqueeze(1)
-        min_encoding_scores, min_encoding_indices = torch.topk(d, 1, dim=1, largest=False)
-        # [0-1], higher score, higher confidence
-        min_encoding_scores = torch.exp(-min_encoding_scores/10)
-
-        min_encodings = torch.zeros(min_encoding_indices.shape[0], self.codebook_size).to(z)
-        min_encodings.scatter_(1, min_encoding_indices, 1)
-
-        # get quantized latent vectors
-        z_q = torch.matmul(min_encodings, self.embedding.weight).view(z.shape)
-        # compute loss for embedding
-        loss = torch.mean((z_q.detach()-z)**2) + self.beta * torch.mean((z_q - z.detach()) ** 2)
-        # preserve gradients
-        z_q = z + (z_q - z).detach()
-
-        # perplexity
-        e_mean = torch.mean(min_encodings, dim=0)
-        perplexity = torch.exp(-torch.sum(e_mean * torch.log(e_mean + 1e-10)))
-        # reshape back to match original input shape
-        z_q = z_q.permute(0, 3, 1, 2).contiguous()
-
-        return z_q, loss, {
-            "perplexity": perplexity,
-            "min_encodings": min_encodings,
-            "min_encoding_indices": min_encoding_indices,
-            "min_encoding_scores": min_encoding_scores,
-            "mean_distance": mean_distance
-            }
-
-    def get_codebook_feat(self, indices, shape):
-        # input indices: batch*token_num -> (batch*token_num)*1
-        # shape: batch, height, width, channel
-        indices = indices.view(-1,1)
-        min_encodings = torch.zeros(indices.shape[0], self.codebook_size).to(indices)
-        min_encodings.scatter_(1, indices, 1)
-        # get quantized latent vectors
-        z_q = torch.matmul(min_encodings.float(), self.embedding.weight)
-
-        if shape is not None:  # reshape back to match original input shape
-            z_q = z_q.view(shape).permute(0, 3, 1, 2).contiguous()
-
-        return z_q
-
-
-class GumbelQuantizer(nn.Module):
-    def __init__(self, codebook_size, emb_dim, num_hiddens, straight_through=False, kl_weight=5e-4, temp_init=1.0):
-        super().__init__()
-        self.codebook_size = codebook_size  # number of embeddings
-        self.emb_dim = emb_dim  # dimension of embedding
-        self.straight_through = straight_through
-        self.temperature = temp_init
-        self.kl_weight = kl_weight
-        self.proj = nn.Conv2d(num_hiddens, codebook_size, 1)  # projects last encoder layer to quantized logits
-        self.embed = nn.Embedding(codebook_size, emb_dim)
-
-    def forward(self, z):
-        hard = self.straight_through if self.training else True
-
-        logits = self.proj(z)
-
-        soft_one_hot = F.gumbel_softmax(logits, tau=self.temperature, dim=1, hard=hard)
-
-        z_q = torch.einsum("b n h w, n d -> b d h w", soft_one_hot, self.embed.weight)
-
-        # + kl divergence to the prior loss
-        qy = F.softmax(logits, dim=1)
-        diff = self.kl_weight * torch.sum(qy * torch.log(qy * self.codebook_size + 1e-10), dim=1).mean()
-        min_encoding_indices = soft_one_hot.argmax(dim=1)
-
-        return z_q, diff, {
-            "min_encoding_indices": min_encoding_indices
-        }
-
-
-class Downsample(nn.Module):
-    def __init__(self, in_channels):
-        super().__init__()
-        self.conv = torch.nn.Conv2d(in_channels, in_channels, kernel_size=3, stride=2, padding=0)
-
-    def forward(self, x):
-        pad = (0, 1, 0, 1)
-        x = torch.nn.functional.pad(x, pad, mode="constant", value=0)
-        x = self.conv(x)
-        return x
-
-
-class Upsample(nn.Module):
-    def __init__(self, in_channels):
-        super().__init__()
-        self.conv = nn.Conv2d(in_channels, in_channels, kernel_size=3, stride=1, padding=1)
-
-    def forward(self, x):
-        x = F.interpolate(x, scale_factor=2.0, mode="nearest")
-        x = self.conv(x)
-
-        return x
-
-
-class ResBlock(nn.Module):
-    def __init__(self, in_channels, out_channels=None):
-        super(ResBlock, self).__init__()
-        self.in_channels = in_channels
-        self.out_channels = in_channels if out_channels is None else out_channels
-        self.norm1 = normalize(in_channels)
-        self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=1, padding=1)
-        self.norm2 = normalize(out_channels)
-        self.conv2 = nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1)
-        if self.in_channels != self.out_channels:
-            self.conv_out = nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=1, padding=0)
-
-    def forward(self, x_in):
-        x = x_in
-        x = self.norm1(x)
-        x = swish(x)
-        x = self.conv1(x)
-        x = self.norm2(x)
-        x = swish(x)
-        x = self.conv2(x)
-        if self.in_channels != self.out_channels:
-            x_in = self.conv_out(x_in)
-
-        return x + x_in
-
-
-class AttnBlock(nn.Module):
-    def __init__(self, in_channels):
-        super().__init__()
-        self.in_channels = in_channels
-
-        self.norm = normalize(in_channels)
-        self.q = torch.nn.Conv2d(
-            in_channels,
-            in_channels,
-            kernel_size=1,
-            stride=1,
-            padding=0
-        )
-        self.k = torch.nn.Conv2d(
-            in_channels,
-            in_channels,
-            kernel_size=1,
-            stride=1,
-            padding=0
-        )
-        self.v = torch.nn.Conv2d(
-            in_channels,
-            in_channels,
-            kernel_size=1,
-            stride=1,
-            padding=0
-        )
-        self.proj_out = torch.nn.Conv2d(
-            in_channels,
-            in_channels,
-            kernel_size=1,
-            stride=1,
-            padding=0
-        )
-
-    def forward(self, x):
-        h_ = x
-        h_ = self.norm(h_)
-        q = self.q(h_)
-        k = self.k(h_)
-        v = self.v(h_)
-
-        # compute attention
-        b, c, h, w = q.shape
-        q = q.reshape(b, c, h*w)
-        q = q.permute(0, 2, 1)
-        k = k.reshape(b, c, h*w)
-        w_ = torch.bmm(q, k)
-        w_ = w_ * (int(c)**(-0.5))
-        w_ = F.softmax(w_, dim=2)
-
-        # attend to values
-        v = v.reshape(b, c, h*w)
-        w_ = w_.permute(0, 2, 1)
-        h_ = torch.bmm(v, w_)
-        h_ = h_.reshape(b, c, h, w)
-
-        h_ = self.proj_out(h_)
-
-        return x+h_
-
-
-class Encoder(nn.Module):
-    def __init__(self, in_channels, nf, emb_dim, ch_mult, num_res_blocks, resolution, attn_resolutions):
-        super().__init__()
-        self.nf = nf
-        self.num_resolutions = len(ch_mult)
-        self.num_res_blocks = num_res_blocks
-        self.resolution = resolution
-        self.attn_resolutions = attn_resolutions
-
-        curr_res = self.resolution
-        in_ch_mult = (1,)+tuple(ch_mult)
-
-        blocks = []
-        # initial convultion
-        blocks.append(nn.Conv2d(in_channels, nf, kernel_size=3, stride=1, padding=1))
-
-        # residual and downsampling blocks, with attention on smaller res (16x16)
-        for i in range(self.num_resolutions):
-            block_in_ch = nf * in_ch_mult[i]
-            block_out_ch = nf * ch_mult[i]
-            for _ in range(self.num_res_blocks):
-                blocks.append(ResBlock(block_in_ch, block_out_ch))
-                block_in_ch = block_out_ch
-                if curr_res in attn_resolutions:
-                    blocks.append(AttnBlock(block_in_ch))
-
-            if i != self.num_resolutions - 1:
-                blocks.append(Downsample(block_in_ch))
-                curr_res = curr_res // 2
-
-        # non-local attention block
-        blocks.append(ResBlock(block_in_ch, block_in_ch))
-        blocks.append(AttnBlock(block_in_ch))
-        blocks.append(ResBlock(block_in_ch, block_in_ch))
-
-        # normalise and convert to latent size
-        blocks.append(normalize(block_in_ch))
-        blocks.append(nn.Conv2d(block_in_ch, emb_dim, kernel_size=3, stride=1, padding=1))
-        self.blocks = nn.ModuleList(blocks)
-
-    def forward(self, x):
-        for block in self.blocks:
-            x = block(x)
-
-        return x
-
-
-class Generator(nn.Module):
-    def __init__(self, nf, emb_dim, ch_mult, res_blocks, img_size, attn_resolutions):
-        super().__init__()
-        self.nf = nf
-        self.ch_mult = ch_mult
-        self.num_resolutions = len(self.ch_mult)
-        self.num_res_blocks = res_blocks
-        self.resolution = img_size
-        self.attn_resolutions = attn_resolutions
-        self.in_channels = emb_dim
-        self.out_channels = 3
-        block_in_ch = self.nf * self.ch_mult[-1]
-        curr_res = self.resolution // 2 ** (self.num_resolutions-1)
-
-        blocks = []
-        # initial conv
-        blocks.append(nn.Conv2d(self.in_channels, block_in_ch, kernel_size=3, stride=1, padding=1))
-
-        # non-local attention block
-        blocks.append(ResBlock(block_in_ch, block_in_ch))
-        blocks.append(AttnBlock(block_in_ch))
-        blocks.append(ResBlock(block_in_ch, block_in_ch))
-
-        for i in reversed(range(self.num_resolutions)):
-            block_out_ch = self.nf * self.ch_mult[i]
-
-            for _ in range(self.num_res_blocks):
-                blocks.append(ResBlock(block_in_ch, block_out_ch))
-                block_in_ch = block_out_ch
-
-                if curr_res in self.attn_resolutions:
-                    blocks.append(AttnBlock(block_in_ch))
-
-            if i != 0:
-                blocks.append(Upsample(block_in_ch))
-                curr_res = curr_res * 2
-
-        blocks.append(normalize(block_in_ch))
-        blocks.append(nn.Conv2d(block_in_ch, self.out_channels, kernel_size=3, stride=1, padding=1))
-
-        self.blocks = nn.ModuleList(blocks)
-
-
-    def forward(self, x):
-        for block in self.blocks:
-            x = block(x)
-
-        return x
-
-
-@ARCH_REGISTRY.register()
-class VQAutoEncoder(nn.Module):
-    def __init__(self, img_size, nf, ch_mult, quantizer="nearest", res_blocks=2, attn_resolutions=None, codebook_size=1024, emb_dim=256,
-                beta=0.25, gumbel_straight_through=False, gumbel_kl_weight=1e-8, model_path=None):
-        super().__init__()
-        logger = get_root_logger()
-        self.in_channels = 3
-        self.nf = nf
-        self.n_blocks = res_blocks
-        self.codebook_size = codebook_size
-        self.embed_dim = emb_dim
-        self.ch_mult = ch_mult
-        self.resolution = img_size
-        self.attn_resolutions = attn_resolutions or [16]
-        self.quantizer_type = quantizer
-        self.encoder = Encoder(
-            self.in_channels,
-            self.nf,
-            self.embed_dim,
-            self.ch_mult,
-            self.n_blocks,
-            self.resolution,
-            self.attn_resolutions
-        )
-        if self.quantizer_type == "nearest":
-            self.beta = beta #0.25
-            self.quantize = VectorQuantizer(self.codebook_size, self.embed_dim, self.beta)
-        elif self.quantizer_type == "gumbel":
-            self.gumbel_num_hiddens = emb_dim
-            self.straight_through = gumbel_straight_through
-            self.kl_weight = gumbel_kl_weight
-            self.quantize = GumbelQuantizer(
-                self.codebook_size,
-                self.embed_dim,
-                self.gumbel_num_hiddens,
-                self.straight_through,
-                self.kl_weight
-            )
-        self.generator = Generator(
-            self.nf,
-            self.embed_dim,
-            self.ch_mult,
-            self.n_blocks,
-            self.resolution,
-            self.attn_resolutions
-        )
-
-        if model_path is not None:
-            chkpt = torch.load(model_path, map_location='cpu')
-            if 'params_ema' in chkpt:
-                self.load_state_dict(torch.load(model_path, map_location='cpu')['params_ema'])
-                logger.info(f'vqgan is loaded from: {model_path} [params_ema]')
-            elif 'params' in chkpt:
-                self.load_state_dict(torch.load(model_path, map_location='cpu')['params'])
-                logger.info(f'vqgan is loaded from: {model_path} [params]')
-            else:
-                raise ValueError('Wrong params!')
-
-
-    def forward(self, x):
-        x = self.encoder(x)
-        quant, codebook_loss, quant_stats = self.quantize(x)
-        x = self.generator(quant)
-        return x, codebook_loss, quant_stats
-
-
-
-# patch based discriminator
-@ARCH_REGISTRY.register()
-class VQGANDiscriminator(nn.Module):
-    def __init__(self, nc=3, ndf=64, n_layers=4, model_path=None):
-        super().__init__()
-
-        layers = [nn.Conv2d(nc, ndf, kernel_size=4, stride=2, padding=1), nn.LeakyReLU(0.2, True)]
-        ndf_mult = 1
-        ndf_mult_prev = 1
-        for n in range(1, n_layers):  # gradually increase the number of filters
-            ndf_mult_prev = ndf_mult
-            ndf_mult = min(2 ** n, 8)
-            layers += [
-                nn.Conv2d(ndf * ndf_mult_prev, ndf * ndf_mult, kernel_size=4, stride=2, padding=1, bias=False),
-                nn.BatchNorm2d(ndf * ndf_mult),
-                nn.LeakyReLU(0.2, True)
-            ]
-
-        ndf_mult_prev = ndf_mult
-        ndf_mult = min(2 ** n_layers, 8)
-
-        layers += [
-            nn.Conv2d(ndf * ndf_mult_prev, ndf * ndf_mult, kernel_size=4, stride=1, padding=1, bias=False),
-            nn.BatchNorm2d(ndf * ndf_mult),
-            nn.LeakyReLU(0.2, True)
-        ]
-
-        layers += [
-            nn.Conv2d(ndf * ndf_mult, 1, kernel_size=4, stride=1, padding=1)]  # output 1 channel prediction map
-        self.main = nn.Sequential(*layers)
-
-        if model_path is not None:
-            chkpt = torch.load(model_path, map_location='cpu')
-            if 'params_d' in chkpt:
-                self.load_state_dict(torch.load(model_path, map_location='cpu')['params_d'])
-            elif 'params' in chkpt:
-                self.load_state_dict(torch.load(model_path, map_location='cpu')['params'])
-            else:
-                raise ValueError('Wrong params!')
-
-    def forward(self, x):
-        return self.main(x)
diff --git a/modules/codeformer_model.py b/modules/codeformer_model.py
index da42b5e9932..44b84618ec0 100644
--- a/modules/codeformer_model.py
+++ b/modules/codeformer_model.py
@@ -1,132 +1,64 @@
-import os
+from __future__ import annotations
 
-import cv2
-import torch
-
-import modules.face_restoration
-import modules.shared
-from modules import shared, devices, modelloader, errors
-from modules.paths import models_path
-
-# codeformer people made a choice to include modified basicsr library to their project which makes
-# it utterly impossible to use it alongside with other libraries that also use basicsr, like GFPGAN.
-# I am making a choice to include some files from codeformer to work around this issue.
-model_dir = "Codeformer"
-model_path = os.path.join(models_path, model_dir)
-model_url = 'https://github.com/sczhou/CodeFormer/releases/download/v0.1.0/codeformer.pth'
-
-codeformer = None
-
-
-def setup_model(dirname):
-    os.makedirs(model_path, exist_ok=True)
-
-    path = modules.paths.paths.get("CodeFormer", None)
-    if path is None:
-        return
-
-    try:
-        from torchvision.transforms.functional import normalize
-        from modules.codeformer.codeformer_arch import CodeFormer
-        from basicsr.utils import img2tensor, tensor2img
-        from facelib.utils.face_restoration_helper import FaceRestoreHelper
-        from facelib.detection.retinaface import retinaface
-
-        net_class = CodeFormer
-
-        class FaceRestorerCodeFormer(modules.face_restoration.FaceRestoration):
-            def name(self):
-                return "CodeFormer"
-
-            def __init__(self, dirname):
-                self.net = None
-                self.face_helper = None
-                self.cmd_dir = dirname
+import logging
 
-            def create_models(self):
-
-                if self.net is not None and self.face_helper is not None:
-                    self.net.to(devices.device_codeformer)
-                    return self.net, self.face_helper
-                model_paths = modelloader.load_models(model_path, model_url, self.cmd_dir, download_name='codeformer-v0.1.0.pth', ext_filter=['.pth'])
-                if len(model_paths) != 0:
-                    ckpt_path = model_paths[0]
-                else:
-                    print("Unable to load codeformer model.")
-                    return None, None
-                net = net_class(dim_embd=512, codebook_size=1024, n_head=8, n_layers=9, connect_list=['32', '64', '128', '256']).to(devices.device_codeformer)
-                checkpoint = torch.load(ckpt_path)['params_ema']
-                net.load_state_dict(checkpoint)
-                net.eval()
-
-                if hasattr(retinaface, 'device'):
-                    retinaface.device = devices.device_codeformer
-                face_helper = FaceRestoreHelper(1, face_size=512, crop_ratio=(1, 1), det_model='retinaface_resnet50', save_ext='png', use_parse=True, device=devices.device_codeformer)
-
-                self.net = net
-                self.face_helper = face_helper
-
-                return net, face_helper
-
-            def send_model_to(self, device):
-                self.net.to(device)
-                self.face_helper.face_det.to(device)
-                self.face_helper.face_parse.to(device)
-
-            def restore(self, np_image, w=None):
-                np_image = np_image[:, :, ::-1]
-
-                original_resolution = np_image.shape[0:2]
+import torch
 
-                self.create_models()
-                if self.net is None or self.face_helper is None:
-                    return np_image
+from modules import (
+    devices,
+    errors,
+    face_restoration,
+    face_restoration_utils,
+    modelloader,
+    shared,
+)
 
-                self.send_model_to(devices.device_codeformer)
+logger = logging.getLogger(__name__)
 
-                self.face_helper.clean_all()
-                self.face_helper.read_image(np_image)
-                self.face_helper.get_face_landmarks_5(only_center_face=False, resize=640, eye_dist_threshold=5)
-                self.face_helper.align_warp_face()
+model_url = 'https://github.com/sczhou/CodeFormer/releases/download/v0.1.0/codeformer.pth'
+model_download_name = 'codeformer-v0.1.0.pth'
 
-                for cropped_face in self.face_helper.cropped_faces:
-                    cropped_face_t = img2tensor(cropped_face / 255., bgr2rgb=True, float32=True)
-                    normalize(cropped_face_t, (0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True)
-                    cropped_face_t = cropped_face_t.unsqueeze(0).to(devices.device_codeformer)
+# used by e.g. postprocessing_codeformer.py
+codeformer: face_restoration.FaceRestoration | None = None
 
-                    try:
-                        with torch.no_grad():
-                            output = self.net(cropped_face_t, w=w if w is not None else shared.opts.code_former_weight, adain=True)[0]
-                            restored_face = tensor2img(output, rgb2bgr=True, min_max=(-1, 1))
-                        del output
-                        devices.torch_gc()
-                    except Exception:
-                        errors.report('Failed inference for CodeFormer', exc_info=True)
-                        restored_face = tensor2img(cropped_face_t, rgb2bgr=True, min_max=(-1, 1))
 
-                    restored_face = restored_face.astype('uint8')
-                    self.face_helper.add_restored_face(restored_face)
+class FaceRestorerCodeFormer(face_restoration_utils.CommonFaceRestoration):
+    def name(self):
+        return "CodeFormer"
 
-                self.face_helper.get_inverse_affine(None)
+    def load_net(self) -> torch.Module:
+        for model_path in modelloader.load_models(
+            model_path=self.model_path,
+            model_url=model_url,
+            command_path=self.model_path,
+            download_name=model_download_name,
+            ext_filter=['.pth'],
+        ):
+            return modelloader.load_spandrel_model(
+                model_path,
+                device=devices.device_codeformer,
+                expected_architecture='CodeFormer',
+            ).model
+        raise ValueError("No codeformer model found")
 
-                restored_img = self.face_helper.paste_faces_to_input_image()
-                restored_img = restored_img[:, :, ::-1]
+    def get_device(self):
+        return devices.device_codeformer
 
-                if original_resolution != restored_img.shape[0:2]:
-                    restored_img = cv2.resize(restored_img, (0, 0), fx=original_resolution[1]/restored_img.shape[1], fy=original_resolution[0]/restored_img.shape[0], interpolation=cv2.INTER_LINEAR)
+    def restore(self, np_image, w: float | None = None):
+        if w is None:
+            w = getattr(shared.opts, "code_former_weight", 0.5)
 
-                self.face_helper.clean_all()
+        def restore_face(cropped_face_t):
+            assert self.net is not None
+            return self.net(cropped_face_t, w=w, adain=True)[0]
 
-                if shared.opts.face_restoration_unload:
-                    self.send_model_to(devices.cpu)
+        return self.restore_with_helper(np_image, restore_face)
 
-                return restored_img
 
-        global codeformer
+def setup_model(dirname: str) -> None:
+    global codeformer
+    try:
         codeformer = FaceRestorerCodeFormer(dirname)
         shared.face_restorers.append(codeformer)
-
     except Exception:
         errors.report("Error setting up CodeFormer", exc_info=True)
-
-   # sys.path = stored_sys_path
diff --git a/modules/dat_model.py b/modules/dat_model.py
new file mode 100644
index 00000000000..495d5f4937d
--- /dev/null
+++ b/modules/dat_model.py
@@ -0,0 +1,79 @@
+import os
+
+from modules import modelloader, errors
+from modules.shared import cmd_opts, opts
+from modules.upscaler import Upscaler, UpscalerData
+from modules.upscaler_utils import upscale_with_model
+
+
+class UpscalerDAT(Upscaler):
+    def __init__(self, user_path):
+        self.name = "DAT"
+        self.user_path = user_path
+        self.scalers = []
+        super().__init__()
+
+        for file in self.find_models(ext_filter=[".pt", ".pth"]):
+            name = modelloader.friendly_name(file)
+            scaler_data = UpscalerData(name, file, upscaler=self, scale=None)
+            self.scalers.append(scaler_data)
+
+        for model in get_dat_models(self):
+            if model.name in opts.dat_enabled_models:
+                self.scalers.append(model)
+
+    def do_upscale(self, img, path):
+        try:
+            info = self.load_model(path)
+        except Exception:
+            errors.report(f"Unable to load DAT model {path}", exc_info=True)
+            return img
+
+        model_descriptor = modelloader.load_spandrel_model(
+            info.local_data_path,
+            device=self.device,
+            prefer_half=(not cmd_opts.no_half and not cmd_opts.upcast_sampling),
+            expected_architecture="DAT",
+        )
+        return upscale_with_model(
+            model_descriptor,
+            img,
+            tile_size=opts.DAT_tile,
+            tile_overlap=opts.DAT_tile_overlap,
+        )
+
+    def load_model(self, path):
+        for scaler in self.scalers:
+            if scaler.data_path == path:
+                if scaler.local_data_path.startswith("http"):
+                    scaler.local_data_path = modelloader.load_file_from_url(
+                        scaler.data_path,
+                        model_dir=self.model_download_path,
+                    )
+                if not os.path.exists(scaler.local_data_path):
+                    raise FileNotFoundError(f"DAT data missing: {scaler.local_data_path}")
+                return scaler
+        raise ValueError(f"Unable to find model info: {path}")
+
+
+def get_dat_models(scaler):
+    return [
+        UpscalerData(
+            name="DAT x2",
+            path="https://github.com/n0kovo/dat_upscaler_models/raw/main/DAT/DAT_x2.pth",
+            scale=2,
+            upscaler=scaler,
+        ),
+        UpscalerData(
+            name="DAT x3",
+            path="https://github.com/n0kovo/dat_upscaler_models/raw/main/DAT/DAT_x3.pth",
+            scale=3,
+            upscaler=scaler,
+        ),
+        UpscalerData(
+            name="DAT x4",
+            path="https://github.com/n0kovo/dat_upscaler_models/raw/main/DAT/DAT_x4.pth",
+            scale=4,
+            upscaler=scaler,
+        ),
+    ]
diff --git a/modules/devices.py b/modules/devices.py
index ea1f712f950..e4f671ac659 100644
--- a/modules/devices.py
+++ b/modules/devices.py
@@ -3,7 +3,7 @@
 from functools import lru_cache
 
 import torch
-from modules import errors, shared
+from modules import errors, shared, npu_specific
 
 if sys.platform == "darwin":
     from modules import mac_specific
@@ -23,6 +23,23 @@ def has_mps() -> bool:
         return mac_specific.has_mps
 
 
+def cuda_no_autocast(device_id=None) -> bool:
+    if device_id is None:
+        device_id = get_cuda_device_id()
+    return (
+        torch.cuda.get_device_capability(device_id) == (7, 5)
+        and torch.cuda.get_device_name(device_id).startswith("NVIDIA GeForce GTX 16")
+    )
+
+
+def get_cuda_device_id():
+    return (
+        int(shared.cmd_opts.device_id)
+        if shared.cmd_opts.device_id is not None and shared.cmd_opts.device_id.isdigit()
+        else 0
+    ) or torch.cuda.current_device()
+
+
 def get_cuda_device_string():
     if shared.cmd_opts.device_id is not None:
         return f"cuda:{shared.cmd_opts.device_id}"
@@ -40,6 +57,9 @@ def get_optimal_device_name():
     if has_xpu():
         return xpu_specific.get_xpu_device_string()
 
+    if npu_specific.has_npu:
+        return npu_specific.get_npu_device_string()
+
     return "cpu"
 
 
@@ -67,14 +87,23 @@ def torch_gc():
     if has_xpu():
         xpu_specific.torch_xpu_gc()
 
+    if npu_specific.has_npu:
+        torch_npu_set_device()
+        npu_specific.torch_npu_gc()
+
+
+def torch_npu_set_device():
+    # Work around due to bug in torch_npu, revert me after fixed, @see https://gitee.com/ascend/pytorch/issues/I8KECW?from=project-issue
+    if npu_specific.has_npu:
+        torch.npu.set_device(0)
+
 
 def enable_tf32():
     if torch.cuda.is_available():
 
         # enabling benchmark option seems to enable a range of cards to do fp16 when they otherwise can't
         # see https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/4407
-        device_id = (int(shared.cmd_opts.device_id) if shared.cmd_opts.device_id is not None and shared.cmd_opts.device_id.isdigit() else 0) or torch.cuda.current_device()
-        if torch.cuda.get_device_capability(device_id) == (7, 5) and torch.cuda.get_device_name(device_id).startswith("NVIDIA GeForce GTX 16"):
+        if cuda_no_autocast():
             torch.backends.cudnn.benchmark = True
 
         torch.backends.cuda.matmul.allow_tf32 = True
@@ -84,6 +113,7 @@ def enable_tf32():
 errors.run(enable_tf32, "Enabling TF32")
 
 cpu: torch.device = torch.device("cpu")
+fp8: bool = False
 device: torch.device = None
 device_interrogate: torch.device = None
 device_gfpgan: torch.device = None
@@ -92,6 +122,7 @@ def enable_tf32():
 dtype: torch.dtype = torch.float16
 dtype_vae: torch.dtype = torch.float16
 dtype_unet: torch.dtype = torch.float16
+dtype_inference: torch.dtype = torch.float16
 unet_needs_upcast = False
 
 
@@ -104,15 +135,89 @@ def cond_cast_float(input):
 
 
 nv_rng = None
+patch_module_list = [
+    torch.nn.Linear,
+    torch.nn.Conv2d,
+    torch.nn.MultiheadAttention,
+    torch.nn.GroupNorm,
+    torch.nn.LayerNorm,
+]
+
+
+def manual_cast_forward(target_dtype):
+    def forward_wrapper(self, *args, **kwargs):
+        if any(
+            isinstance(arg, torch.Tensor) and arg.dtype != target_dtype
+            for arg in args
+        ):
+            args = [arg.to(target_dtype) if isinstance(arg, torch.Tensor) else arg for arg in args]
+            kwargs = {k: v.to(target_dtype) if isinstance(v, torch.Tensor) else v for k, v in kwargs.items()}
+
+        org_dtype = target_dtype
+        for param in self.parameters():
+            if param.dtype != target_dtype:
+                org_dtype = param.dtype
+                break
+
+        if org_dtype != target_dtype:
+            self.to(target_dtype)
+        result = self.org_forward(*args, **kwargs)
+        if org_dtype != target_dtype:
+            self.to(org_dtype)
+
+        if target_dtype != dtype_inference:
+            if isinstance(result, tuple):
+                result = tuple(
+                    i.to(dtype_inference)
+                    if isinstance(i, torch.Tensor)
+                    else i
+                    for i in result
+                )
+            elif isinstance(result, torch.Tensor):
+                result = result.to(dtype_inference)
+        return result
+    return forward_wrapper
+
+
+@contextlib.contextmanager
+def manual_cast(target_dtype):
+    applied = False
+    for module_type in patch_module_list:
+        if hasattr(module_type, "org_forward"):
+            continue
+        applied = True
+        org_forward = module_type.forward
+        if module_type == torch.nn.MultiheadAttention:
+            module_type.forward = manual_cast_forward(torch.float32)
+        else:
+            module_type.forward = manual_cast_forward(target_dtype)
+        module_type.org_forward = org_forward
+    try:
+        yield None
+    finally:
+        if applied:
+            for module_type in patch_module_list:
+                if hasattr(module_type, "org_forward"):
+                    module_type.forward = module_type.org_forward
+                    delattr(module_type, "org_forward")
 
 
 def autocast(disable=False):
     if disable:
         return contextlib.nullcontext()
 
-    if dtype == torch.float32 or shared.cmd_opts.precision == "full":
+    if fp8 and device==cpu:
+        return torch.autocast("cpu", dtype=torch.bfloat16, enabled=True)
+
+    if fp8 and dtype_inference == torch.float32:
+        return manual_cast(dtype)
+
+    if dtype == torch.float32 or dtype_inference == torch.float32:
         return contextlib.nullcontext()
 
+    if has_xpu() or has_mps() or cuda_no_autocast():
+        return manual_cast(dtype)
+
     return torch.autocast("cuda")
 
 
@@ -154,7 +259,7 @@ def test_for_nans(x, where):
 def first_time_calculation():
     """
     just do any calculation with pytorch layers - the first time this is done it allocaltes about 700MB of memory and
-    spends about 2.7 seconds doing that, at least wih NVidia.
+    spends about 2.7 seconds doing that, at least with NVidia.
     """
 
     x = torch.zeros((1, 1)).to(device, dtype)
@@ -164,4 +269,3 @@ def first_time_calculation():
     x = torch.zeros((1, 1, 3, 3)).to(device, dtype)
     conv2d = torch.nn.Conv2d(1, 1, (3, 3)).to(device, dtype)
     conv2d(x)
-
diff --git a/modules/errors.py b/modules/errors.py
index eb234a83811..48aa13a1728 100644
--- a/modules/errors.py
+++ b/modules/errors.py
@@ -107,8 +107,8 @@ def check_versions():
     import torch
     import gradio
 
-    expected_torch_version = "2.0.0"
-    expected_xformers_version = "0.0.20"
+    expected_torch_version = "2.1.2"
+    expected_xformers_version = "0.0.23.post1"
     expected_gradio_version = "3.41.2"
 
     if version.parse(torch.__version__) < version.parse(expected_torch_version):
diff --git a/modules/esrgan_model.py b/modules/esrgan_model.py
index 02a1727d280..70041ab0234 100644
--- a/modules/esrgan_model.py
+++ b/modules/esrgan_model.py
@@ -1,121 +1,7 @@
-import sys
-
-import numpy as np
-import torch
-from PIL import Image
-
-import modules.esrgan_model_arch as arch
-from modules import modelloader, images, devices
+from modules import modelloader, devices, errors
 from modules.shared import opts
 from modules.upscaler import Upscaler, UpscalerData
-
-
-def mod2normal(state_dict):
-    # this code is copied from https://github.com/victorca25/iNNfer
-    if 'conv_first.weight' in state_dict:
-        crt_net = {}
-        items = list(state_dict)
-
-        crt_net['model.0.weight'] = state_dict['conv_first.weight']
-        crt_net['model.0.bias'] = state_dict['conv_first.bias']
-
-        for k in items.copy():
-            if 'RDB' in k:
-                ori_k = k.replace('RRDB_trunk.', 'model.1.sub.')
-                if '.weight' in k:
-                    ori_k = ori_k.replace('.weight', '.0.weight')
-                elif '.bias' in k:
-                    ori_k = ori_k.replace('.bias', '.0.bias')
-                crt_net[ori_k] = state_dict[k]
-                items.remove(k)
-
-        crt_net['model.1.sub.23.weight'] = state_dict['trunk_conv.weight']
-        crt_net['model.1.sub.23.bias'] = state_dict['trunk_conv.bias']
-        crt_net['model.3.weight'] = state_dict['upconv1.weight']
-        crt_net['model.3.bias'] = state_dict['upconv1.bias']
-        crt_net['model.6.weight'] = state_dict['upconv2.weight']
-        crt_net['model.6.bias'] = state_dict['upconv2.bias']
-        crt_net['model.8.weight'] = state_dict['HRconv.weight']
-        crt_net['model.8.bias'] = state_dict['HRconv.bias']
-        crt_net['model.10.weight'] = state_dict['conv_last.weight']
-        crt_net['model.10.bias'] = state_dict['conv_last.bias']
-        state_dict = crt_net
-    return state_dict
-
-
-def resrgan2normal(state_dict, nb=23):
-    # this code is copied from https://github.com/victorca25/iNNfer
-    if "conv_first.weight" in state_dict and "body.0.rdb1.conv1.weight" in state_dict:
-        re8x = 0
-        crt_net = {}
-        items = list(state_dict)
-
-        crt_net['model.0.weight'] = state_dict['conv_first.weight']
-        crt_net['model.0.bias'] = state_dict['conv_first.bias']
-
-        for k in items.copy():
-            if "rdb" in k:
-                ori_k = k.replace('body.', 'model.1.sub.')
-                ori_k = ori_k.replace('.rdb', '.RDB')
-                if '.weight' in k:
-                    ori_k = ori_k.replace('.weight', '.0.weight')
-                elif '.bias' in k:
-                    ori_k = ori_k.replace('.bias', '.0.bias')
-                crt_net[ori_k] = state_dict[k]
-                items.remove(k)
-
-        crt_net[f'model.1.sub.{nb}.weight'] = state_dict['conv_body.weight']
-        crt_net[f'model.1.sub.{nb}.bias'] = state_dict['conv_body.bias']
-        crt_net['model.3.weight'] = state_dict['conv_up1.weight']
-        crt_net['model.3.bias'] = state_dict['conv_up1.bias']
-        crt_net['model.6.weight'] = state_dict['conv_up2.weight']
-        crt_net['model.6.bias'] = state_dict['conv_up2.bias']
-
-        if 'conv_up3.weight' in state_dict:
-            # modification supporting: https://github.com/ai-forever/Real-ESRGAN/blob/main/RealESRGAN/rrdbnet_arch.py
-            re8x = 3
-            crt_net['model.9.weight'] = state_dict['conv_up3.weight']
-            crt_net['model.9.bias'] = state_dict['conv_up3.bias']
-
-        crt_net[f'model.{8+re8x}.weight'] = state_dict['conv_hr.weight']
-        crt_net[f'model.{8+re8x}.bias'] = state_dict['conv_hr.bias']
-        crt_net[f'model.{10+re8x}.weight'] = state_dict['conv_last.weight']
-        crt_net[f'model.{10+re8x}.bias'] = state_dict['conv_last.bias']
-
-        state_dict = crt_net
-    return state_dict
-
-
-def infer_params(state_dict):
-    # this code is copied from https://github.com/victorca25/iNNfer
-    scale2x = 0
-    scalemin = 6
-    n_uplayer = 0
-    plus = False
-
-    for block in list(state_dict):
-        parts = block.split(".")
-        n_parts = len(parts)
-        if n_parts == 5 and parts[2] == "sub":
-            nb = int(parts[3])
-        elif n_parts == 3:
-            part_num = int(parts[1])
-            if (part_num > scalemin
-                and parts[0] == "model"
-                and parts[2] == "weight"):
-                scale2x += 1
-            if part_num > n_uplayer:
-                n_uplayer = part_num
-                out_nc = state_dict[block].shape[0]
-        if not plus and "conv1x1" in block:
-            plus = True
-
-    nf = state_dict["model.0.weight"].shape[0]
-    in_nc = state_dict["model.0.weight"].shape[1]
-    out_nc = out_nc
-    scale = 2 ** scale2x
-
-    return in_nc, out_nc, nf, nb, plus, scale
+from modules.upscaler_utils import upscale_with_model
 
 
 class UpscalerESRGAN(Upscaler):
@@ -143,12 +29,11 @@ def __init__(self, dirname):
     def do_upscale(self, img, selected_model):
         try:
             model = self.load_model(selected_model)
-        except Exception as e:
-            print(f"Unable to load ESRGAN model {selected_model}: {e}", file=sys.stderr)
+        except Exception:
+            errors.report(f"Unable to load ESRGAN model {selected_model}", exc_info=True)
             return img
         model.to(devices.device_esrgan)
-        img = esrgan_upscale(model, img)
-        return img
+        return esrgan_upscale(model, img)
 
     def load_model(self, path: str):
         if path.startswith("http"):
@@ -161,69 +46,17 @@ def load_model(self, path: str):
         else:
             filename = path
 
-        state_dict = torch.load(filename, map_location='cpu' if devices.device_esrgan.type == 'mps' else None)
-
-        if "params_ema" in state_dict:
-            state_dict = state_dict["params_ema"]
-        elif "params" in state_dict:
-            state_dict = state_dict["params"]
-            num_conv = 16 if "realesr-animevideov3" in filename else 32
-            model = arch.SRVGGNetCompact(num_in_ch=3, num_out_ch=3, num_feat=64, num_conv=num_conv, upscale=4, act_type='prelu')
-            model.load_state_dict(state_dict)
-            model.eval()
-            return model
-
-        if "body.0.rdb1.conv1.weight" in state_dict and "conv_first.weight" in state_dict:
-            nb = 6 if "RealESRGAN_x4plus_anime_6B" in filename else 23
-            state_dict = resrgan2normal(state_dict, nb)
-        elif "conv_first.weight" in state_dict:
-            state_dict = mod2normal(state_dict)
-        elif "model.0.weight" not in state_dict:
-            raise Exception("The file is not a recognized ESRGAN model.")
-
-        in_nc, out_nc, nf, nb, plus, mscale = infer_params(state_dict)
-
-        model = arch.RRDBNet(in_nc=in_nc, out_nc=out_nc, nf=nf, nb=nb, upscale=mscale, plus=plus)
-        model.load_state_dict(state_dict)
-        model.eval()
-
-        return model
-
-
-def upscale_without_tiling(model, img):
-    img = np.array(img)
-    img = img[:, :, ::-1]
-    img = np.ascontiguousarray(np.transpose(img, (2, 0, 1))) / 255
-    img = torch.from_numpy(img).float()
-    img = img.unsqueeze(0).to(devices.device_esrgan)
-    with torch.no_grad():
-        output = model(img)
-    output = output.squeeze().float().cpu().clamp_(0, 1).numpy()
-    output = 255. * np.moveaxis(output, 0, 2)
-    output = output.astype(np.uint8)
-    output = output[:, :, ::-1]
-    return Image.fromarray(output, 'RGB')
+        return modelloader.load_spandrel_model(
+            filename,
+            device=('cpu' if devices.device_esrgan.type == 'mps' else None),
+            expected_architecture='ESRGAN',
+        )
 
 
 def esrgan_upscale(model, img):
-    if opts.ESRGAN_tile == 0:
-        return upscale_without_tiling(model, img)
-
-    grid = images.split_grid(img, opts.ESRGAN_tile, opts.ESRGAN_tile, opts.ESRGAN_tile_overlap)
-    newtiles = []
-    scale_factor = 1
-
-    for y, h, row in grid.tiles:
-        newrow = []
-        for tiledata in row:
-            x, w, tile = tiledata
-
-            output = upscale_without_tiling(model, tile)
-            scale_factor = output.width // tile.width
-
-            newrow.append([x * scale_factor, w * scale_factor, output])
-        newtiles.append([y * scale_factor, h * scale_factor, newrow])
-
-    newgrid = images.Grid(newtiles, grid.tile_w * scale_factor, grid.tile_h * scale_factor, grid.image_w * scale_factor, grid.image_h * scale_factor, grid.overlap * scale_factor)
-    output = images.combine_grid(newgrid)
-    return output
+    return upscale_with_model(
+        model,
+        img,
+        tile_size=opts.ESRGAN_tile,
+        tile_overlap=opts.ESRGAN_tile_overlap,
+    )
diff --git a/modules/esrgan_model_arch.py b/modules/esrgan_model_arch.py
deleted file mode 100644
index 2b9888bafbc..00000000000
--- a/modules/esrgan_model_arch.py
+++ /dev/null
@@ -1,465 +0,0 @@
-# this file is adapted from https://github.com/victorca25/iNNfer
-
-from collections import OrderedDict
-import math
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-
-####################
-# RRDBNet Generator
-####################
-
-class RRDBNet(nn.Module):
-    def __init__(self, in_nc, out_nc, nf, nb, nr=3, gc=32, upscale=4, norm_type=None,
-            act_type='leakyrelu', mode='CNA', upsample_mode='upconv', convtype='Conv2D',
-            finalact=None, gaussian_noise=False, plus=False):
-        super(RRDBNet, self).__init__()
-        n_upscale = int(math.log(upscale, 2))
-        if upscale == 3:
-            n_upscale = 1
-
-        self.resrgan_scale = 0
-        if in_nc % 16 == 0:
-            self.resrgan_scale = 1
-        elif in_nc != 4 and in_nc % 4 == 0:
-            self.resrgan_scale = 2
-
-        fea_conv = conv_block(in_nc, nf, kernel_size=3, norm_type=None, act_type=None, convtype=convtype)
-        rb_blocks = [RRDB(nf, nr, kernel_size=3, gc=32, stride=1, bias=1, pad_type='zero',
-            norm_type=norm_type, act_type=act_type, mode='CNA', convtype=convtype,
-            gaussian_noise=gaussian_noise, plus=plus) for _ in range(nb)]
-        LR_conv = conv_block(nf, nf, kernel_size=3, norm_type=norm_type, act_type=None, mode=mode, convtype=convtype)
-
-        if upsample_mode == 'upconv':
-            upsample_block = upconv_block
-        elif upsample_mode == 'pixelshuffle':
-            upsample_block = pixelshuffle_block
-        else:
-            raise NotImplementedError(f'upsample mode [{upsample_mode}] is not found')
-        if upscale == 3:
-            upsampler = upsample_block(nf, nf, 3, act_type=act_type, convtype=convtype)
-        else:
-            upsampler = [upsample_block(nf, nf, act_type=act_type, convtype=convtype) for _ in range(n_upscale)]
-        HR_conv0 = conv_block(nf, nf, kernel_size=3, norm_type=None, act_type=act_type, convtype=convtype)
-        HR_conv1 = conv_block(nf, out_nc, kernel_size=3, norm_type=None, act_type=None, convtype=convtype)
-
-        outact = act(finalact) if finalact else None
-
-        self.model = sequential(fea_conv, ShortcutBlock(sequential(*rb_blocks, LR_conv)),
-            *upsampler, HR_conv0, HR_conv1, outact)
-
-    def forward(self, x, outm=None):
-        if self.resrgan_scale == 1:
-            feat = pixel_unshuffle(x, scale=4)
-        elif self.resrgan_scale == 2:
-            feat = pixel_unshuffle(x, scale=2)
-        else:
-            feat = x
-
-        return self.model(feat)
-
-
-class RRDB(nn.Module):
-    """
-    Residual in Residual Dense Block
-    (ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks)
-    """
-
-    def __init__(self, nf, nr=3, kernel_size=3, gc=32, stride=1, bias=1, pad_type='zero',
-            norm_type=None, act_type='leakyrelu', mode='CNA', convtype='Conv2D',
-            spectral_norm=False, gaussian_noise=False, plus=False):
-        super(RRDB, self).__init__()
-        # This is for backwards compatibility with existing models
-        if nr == 3:
-            self.RDB1 = ResidualDenseBlock_5C(nf, kernel_size, gc, stride, bias, pad_type,
-                    norm_type, act_type, mode, convtype, spectral_norm=spectral_norm,
-                    gaussian_noise=gaussian_noise, plus=plus)
-            self.RDB2 = ResidualDenseBlock_5C(nf, kernel_size, gc, stride, bias, pad_type,
-                    norm_type, act_type, mode, convtype, spectral_norm=spectral_norm,
-                    gaussian_noise=gaussian_noise, plus=plus)
-            self.RDB3 = ResidualDenseBlock_5C(nf, kernel_size, gc, stride, bias, pad_type,
-                    norm_type, act_type, mode, convtype, spectral_norm=spectral_norm,
-                    gaussian_noise=gaussian_noise, plus=plus)
-        else:
-            RDB_list = [ResidualDenseBlock_5C(nf, kernel_size, gc, stride, bias, pad_type,
-                                              norm_type, act_type, mode, convtype, spectral_norm=spectral_norm,
-                                              gaussian_noise=gaussian_noise, plus=plus) for _ in range(nr)]
-            self.RDBs = nn.Sequential(*RDB_list)
-
-    def forward(self, x):
-        if hasattr(self, 'RDB1'):
-            out = self.RDB1(x)
-            out = self.RDB2(out)
-            out = self.RDB3(out)
-        else:
-            out = self.RDBs(x)
-        return out * 0.2 + x
-
-
-class ResidualDenseBlock_5C(nn.Module):
-    """
-    Residual Dense Block
-    The core module of paper: (Residual Dense Network for Image Super-Resolution, CVPR 18)
-    Modified options that can be used:
-        - "Partial Convolution based Padding" arXiv:1811.11718
-        - "Spectral normalization" arXiv:1802.05957
-        - "ICASSP 2020 - ESRGAN+ : Further Improving ESRGAN" N. C.
-            {Rakotonirina} and A. {Rasoanaivo}
-    """
-
-    def __init__(self, nf=64, kernel_size=3, gc=32, stride=1, bias=1, pad_type='zero',
-            norm_type=None, act_type='leakyrelu', mode='CNA', convtype='Conv2D',
-            spectral_norm=False, gaussian_noise=False, plus=False):
-        super(ResidualDenseBlock_5C, self).__init__()
-
-        self.noise = GaussianNoise() if gaussian_noise else None
-        self.conv1x1 = conv1x1(nf, gc) if plus else None
-
-        self.conv1 = conv_block(nf, gc, kernel_size, stride, bias=bias, pad_type=pad_type,
-            norm_type=norm_type, act_type=act_type, mode=mode, convtype=convtype,
-            spectral_norm=spectral_norm)
-        self.conv2 = conv_block(nf+gc, gc, kernel_size, stride, bias=bias, pad_type=pad_type,
-            norm_type=norm_type, act_type=act_type, mode=mode, convtype=convtype,
-            spectral_norm=spectral_norm)
-        self.conv3 = conv_block(nf+2*gc, gc, kernel_size, stride, bias=bias, pad_type=pad_type,
-            norm_type=norm_type, act_type=act_type, mode=mode, convtype=convtype,
-            spectral_norm=spectral_norm)
-        self.conv4 = conv_block(nf+3*gc, gc, kernel_size, stride, bias=bias, pad_type=pad_type,
-            norm_type=norm_type, act_type=act_type, mode=mode, convtype=convtype,
-            spectral_norm=spectral_norm)
-        if mode == 'CNA':
-            last_act = None
-        else:
-            last_act = act_type
-        self.conv5 = conv_block(nf+4*gc, nf, 3, stride, bias=bias, pad_type=pad_type,
-            norm_type=norm_type, act_type=last_act, mode=mode, convtype=convtype,
-            spectral_norm=spectral_norm)
-
-    def forward(self, x):
-        x1 = self.conv1(x)
-        x2 = self.conv2(torch.cat((x, x1), 1))
-        if self.conv1x1:
-            x2 = x2 + self.conv1x1(x)
-        x3 = self.conv3(torch.cat((x, x1, x2), 1))
-        x4 = self.conv4(torch.cat((x, x1, x2, x3), 1))
-        if self.conv1x1:
-            x4 = x4 + x2
-        x5 = self.conv5(torch.cat((x, x1, x2, x3, x4), 1))
-        if self.noise:
-            return self.noise(x5.mul(0.2) + x)
-        else:
-            return x5 * 0.2 + x
-
-
-####################
-# ESRGANplus
-####################
-
-class GaussianNoise(nn.Module):
-    def __init__(self, sigma=0.1, is_relative_detach=False):
-        super().__init__()
-        self.sigma = sigma
-        self.is_relative_detach = is_relative_detach
-        self.noise = torch.tensor(0, dtype=torch.float)
-
-    def forward(self, x):
-        if self.training and self.sigma != 0:
-            self.noise = self.noise.to(x.device)
-            scale = self.sigma * x.detach() if self.is_relative_detach else self.sigma * x
-            sampled_noise = self.noise.repeat(*x.size()).normal_() * scale
-            x = x + sampled_noise
-        return x
-
-def conv1x1(in_planes, out_planes, stride=1):
-    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)
-
-
-####################
-# SRVGGNetCompact
-####################
-
-class SRVGGNetCompact(nn.Module):
-    """A compact VGG-style network structure for super-resolution.
-    This class is copied from https://github.com/xinntao/Real-ESRGAN
-    """
-
-    def __init__(self, num_in_ch=3, num_out_ch=3, num_feat=64, num_conv=16, upscale=4, act_type='prelu'):
-        super(SRVGGNetCompact, self).__init__()
-        self.num_in_ch = num_in_ch
-        self.num_out_ch = num_out_ch
-        self.num_feat = num_feat
-        self.num_conv = num_conv
-        self.upscale = upscale
-        self.act_type = act_type
-
-        self.body = nn.ModuleList()
-        # the first conv
-        self.body.append(nn.Conv2d(num_in_ch, num_feat, 3, 1, 1))
-        # the first activation
-        if act_type == 'relu':
-            activation = nn.ReLU(inplace=True)
-        elif act_type == 'prelu':
-            activation = nn.PReLU(num_parameters=num_feat)
-        elif act_type == 'leakyrelu':
-            activation = nn.LeakyReLU(negative_slope=0.1, inplace=True)
-        self.body.append(activation)
-
-        # the body structure
-        for _ in range(num_conv):
-            self.body.append(nn.Conv2d(num_feat, num_feat, 3, 1, 1))
-            # activation
-            if act_type == 'relu':
-                activation = nn.ReLU(inplace=True)
-            elif act_type == 'prelu':
-                activation = nn.PReLU(num_parameters=num_feat)
-            elif act_type == 'leakyrelu':
-                activation = nn.LeakyReLU(negative_slope=0.1, inplace=True)
-            self.body.append(activation)
-
-        # the last conv
-        self.body.append(nn.Conv2d(num_feat, num_out_ch * upscale * upscale, 3, 1, 1))
-        # upsample
-        self.upsampler = nn.PixelShuffle(upscale)
-
-    def forward(self, x):
-        out = x
-        for i in range(0, len(self.body)):
-            out = self.body[i](out)
-
-        out = self.upsampler(out)
-        # add the nearest upsampled image, so that the network learns the residual
-        base = F.interpolate(x, scale_factor=self.upscale, mode='nearest')
-        out += base
-        return out
-
-
-####################
-# Upsampler
-####################
-
-class Upsample(nn.Module):
-    r"""Upsamples a given multi-channel 1D (temporal), 2D (spatial) or 3D (volumetric) data.
-    The input data is assumed to be of the form
-    `minibatch x channels x [optional depth] x [optional height] x width`.
-    """
-
-    def __init__(self, size=None, scale_factor=None, mode="nearest", align_corners=None):
-        super(Upsample, self).__init__()
-        if isinstance(scale_factor, tuple):
-            self.scale_factor = tuple(float(factor) for factor in scale_factor)
-        else:
-            self.scale_factor = float(scale_factor) if scale_factor else None
-        self.mode = mode
-        self.size = size
-        self.align_corners = align_corners
-
-    def forward(self, x):
-        return nn.functional.interpolate(x, size=self.size, scale_factor=self.scale_factor, mode=self.mode, align_corners=self.align_corners)
-
-    def extra_repr(self):
-        if self.scale_factor is not None:
-            info = f'scale_factor={self.scale_factor}'
-        else:
-            info = f'size={self.size}'
-        info += f', mode={self.mode}'
-        return info
-
-
-def pixel_unshuffle(x, scale):
-    """ Pixel unshuffle.
-    Args:
-        x (Tensor): Input feature with shape (b, c, hh, hw).
-        scale (int): Downsample ratio.
-    Returns:
-        Tensor: the pixel unshuffled feature.
-    """
-    b, c, hh, hw = x.size()
-    out_channel = c * (scale**2)
-    assert hh % scale == 0 and hw % scale == 0
-    h = hh // scale
-    w = hw // scale
-    x_view = x.view(b, c, h, scale, w, scale)
-    return x_view.permute(0, 1, 3, 5, 2, 4).reshape(b, out_channel, h, w)
-
-
-def pixelshuffle_block(in_nc, out_nc, upscale_factor=2, kernel_size=3, stride=1, bias=True,
-                        pad_type='zero', norm_type=None, act_type='relu', convtype='Conv2D'):
-    """
-    Pixel shuffle layer
-    (Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional
-    Neural Network, CVPR17)
-    """
-    conv = conv_block(in_nc, out_nc * (upscale_factor ** 2), kernel_size, stride, bias=bias,
-                        pad_type=pad_type, norm_type=None, act_type=None, convtype=convtype)
-    pixel_shuffle = nn.PixelShuffle(upscale_factor)
-
-    n = norm(norm_type, out_nc) if norm_type else None
-    a = act(act_type) if act_type else None
-    return sequential(conv, pixel_shuffle, n, a)
-
-
-def upconv_block(in_nc, out_nc, upscale_factor=2, kernel_size=3, stride=1, bias=True,
-                pad_type='zero', norm_type=None, act_type='relu', mode='nearest', convtype='Conv2D'):
-    """ Upconv layer """
-    upscale_factor = (1, upscale_factor, upscale_factor) if convtype == 'Conv3D' else upscale_factor
-    upsample = Upsample(scale_factor=upscale_factor, mode=mode)
-    conv = conv_block(in_nc, out_nc, kernel_size, stride, bias=bias,
-                        pad_type=pad_type, norm_type=norm_type, act_type=act_type, convtype=convtype)
-    return sequential(upsample, conv)
-
-
-
-
-
-
-
-
-####################
-# Basic blocks
-####################
-
-
-def make_layer(basic_block, num_basic_block, **kwarg):
-    """Make layers by stacking the same blocks.
-    Args:
-        basic_block (nn.module): nn.module class for basic block. (block)
-        num_basic_block (int): number of blocks. (n_layers)
-    Returns:
-        nn.Sequential: Stacked blocks in nn.Sequential.
-    """
-    layers = []
-    for _ in range(num_basic_block):
-        layers.append(basic_block(**kwarg))
-    return nn.Sequential(*layers)
-
-
-def act(act_type, inplace=True, neg_slope=0.2, n_prelu=1, beta=1.0):
-    """ activation helper """
-    act_type = act_type.lower()
-    if act_type == 'relu':
-        layer = nn.ReLU(inplace)
-    elif act_type in ('leakyrelu', 'lrelu'):
-        layer = nn.LeakyReLU(neg_slope, inplace)
-    elif act_type == 'prelu':
-        layer = nn.PReLU(num_parameters=n_prelu, init=neg_slope)
-    elif act_type == 'tanh':  # [-1, 1] range output
-        layer = nn.Tanh()
-    elif act_type == 'sigmoid':  # [0, 1] range output
-        layer = nn.Sigmoid()
-    else:
-        raise NotImplementedError(f'activation layer [{act_type}] is not found')
-    return layer
-
-
-class Identity(nn.Module):
-    def __init__(self, *kwargs):
-        super(Identity, self).__init__()
-
-    def forward(self, x, *kwargs):
-        return x
-
-
-def norm(norm_type, nc):
-    """ Return a normalization layer """
-    norm_type = norm_type.lower()
-    if norm_type == 'batch':
-        layer = nn.BatchNorm2d(nc, affine=True)
-    elif norm_type == 'instance':
-        layer = nn.InstanceNorm2d(nc, affine=False)
-    elif norm_type == 'none':
-        def norm_layer(x): return Identity()
-    else:
-        raise NotImplementedError(f'normalization layer [{norm_type}] is not found')
-    return layer
-
-
-def pad(pad_type, padding):
-    """ padding layer helper """
-    pad_type = pad_type.lower()
-    if padding == 0:
-        return None
-    if pad_type == 'reflect':
-        layer = nn.ReflectionPad2d(padding)
-    elif pad_type == 'replicate':
-        layer = nn.ReplicationPad2d(padding)
-    elif pad_type == 'zero':
-        layer = nn.ZeroPad2d(padding)
-    else:
-        raise NotImplementedError(f'padding layer [{pad_type}] is not implemented')
-    return layer
-
-
-def get_valid_padding(kernel_size, dilation):
-    kernel_size = kernel_size + (kernel_size - 1) * (dilation - 1)
-    padding = (kernel_size - 1) // 2
-    return padding
-
-
-class ShortcutBlock(nn.Module):
-    """ Elementwise sum the output of a submodule to its input """
-    def __init__(self, submodule):
-        super(ShortcutBlock, self).__init__()
-        self.sub = submodule
-
-    def forward(self, x):
-        output = x + self.sub(x)
-        return output
-
-    def __repr__(self):
-        return 'Identity + \n|' + self.sub.__repr__().replace('\n', '\n|')
-
-
-def sequential(*args):
-    """ Flatten Sequential. It unwraps nn.Sequential. """
-    if len(args) == 1:
-        if isinstance(args[0], OrderedDict):
-            raise NotImplementedError('sequential does not support OrderedDict input.')
-        return args[0]  # No sequential is needed.
-    modules = []
-    for module in args:
-        if isinstance(module, nn.Sequential):
-            for submodule in module.children():
-                modules.append(submodule)
-        elif isinstance(module, nn.Module):
-            modules.append(module)
-    return nn.Sequential(*modules)
-
-
-def conv_block(in_nc, out_nc, kernel_size, stride=1, dilation=1, groups=1, bias=True,
-               pad_type='zero', norm_type=None, act_type='relu', mode='CNA', convtype='Conv2D',
-               spectral_norm=False):
-    """ Conv layer with padding, normalization, activation """
-    assert mode in ['CNA', 'NAC', 'CNAC'], f'Wrong conv mode [{mode}]'
-    padding = get_valid_padding(kernel_size, dilation)
-    p = pad(pad_type, padding) if pad_type and pad_type != 'zero' else None
-    padding = padding if pad_type == 'zero' else 0
-
-    if convtype=='PartialConv2D':
-        from torchvision.ops import PartialConv2d  # this is definitely not going to work, but PartialConv2d doesn't work anyway and this shuts up static analyzer
-        c = PartialConv2d(in_nc, out_nc, kernel_size=kernel_size, stride=stride, padding=padding,
-               dilation=dilation, bias=bias, groups=groups)
-    elif convtype=='DeformConv2D':
-        from torchvision.ops import DeformConv2d  # not tested
-        c = DeformConv2d(in_nc, out_nc, kernel_size=kernel_size, stride=stride, padding=padding,
-               dilation=dilation, bias=bias, groups=groups)
-    elif convtype=='Conv3D':
-        c = nn.Conv3d(in_nc, out_nc, kernel_size=kernel_size, stride=stride, padding=padding,
-                dilation=dilation, bias=bias, groups=groups)
-    else:
-        c = nn.Conv2d(in_nc, out_nc, kernel_size=kernel_size, stride=stride, padding=padding,
-                dilation=dilation, bias=bias, groups=groups)
-
-    if spectral_norm:
-        c = nn.utils.spectral_norm(c)
-
-    a = act(act_type) if act_type else None
-    if 'CNA' in mode:
-        n = norm(norm_type, out_nc) if norm_type else None
-        return sequential(p, c, n, a)
-    elif mode == 'NAC':
-        if norm_type is None and act_type is not None:
-            a = act(act_type, inplace=False)
-        n = norm(norm_type, in_nc) if norm_type else None
-        return sequential(n, a, p, c)
diff --git a/modules/extensions.py b/modules/extensions.py
index 1899cd52975..04bda297e79 100644
--- a/modules/extensions.py
+++ b/modules/extensions.py
@@ -32,11 +32,12 @@ def __init__(self, path, canonical_name):
         self.config = configparser.ConfigParser()
 
         filepath = os.path.join(path, self.filename)
-        if os.path.isfile(filepath):
-            try:
-                self.config.read(filepath)
-            except Exception:
-                errors.report(f"Error reading {self.filename} for extension {canonical_name}.", exc_info=True)
+        # `self.config.read()` will quietly swallow OSErrors (which FileNotFoundError is),
+        # so no need to check whether the file exists beforehand.
+        try:
+            self.config.read(filepath)
+        except Exception:
+            errors.report(f"Error reading {self.filename} for extension {canonical_name}.", exc_info=True)
 
         self.canonical_name = self.config.get("Extension", "Name", fallback=canonical_name)
         self.canonical_name = canonical_name.lower().strip()
@@ -223,13 +224,16 @@ def list_extensions():
 
     # check for requirements
     for extension in extensions:
+        if not extension.enabled:
+            continue
+
         for req in extension.metadata.requires:
             required_extension = loaded_extensions.get(req)
             if required_extension is None:
                 errors.report(f'Extension "{extension.name}" requires "{req}" which is not installed.', exc_info=False)
                 continue
 
-            if not extension.enabled:
+            if not required_extension.enabled:
                 errors.report(f'Extension "{extension.name}" requires "{required_extension.name}" which is disabled.', exc_info=False)
                 continue
 
diff --git a/modules/extra_networks.py b/modules/extra_networks.py
index b9533677887..ae8d42d9b38 100644
--- a/modules/extra_networks.py
+++ b/modules/extra_networks.py
@@ -60,7 +60,7 @@ def activate(self, p, params_list):
         Where name matches the name of this ExtraNetwork object, and arg1:arg2:arg3 are any natural number of text arguments
         separated by colon.
 
-        Even if the user does not mention this ExtraNetwork in his prompt, the call will stil be made, with empty params_list -
+        Even if the user does not mention this ExtraNetwork in his prompt, the call will still be made, with empty params_list -
         in this case, all effects of this extra networks should be disabled.
 
         Can be called multiple times before deactivate() - each new call should override the previous call completely.
@@ -206,7 +206,7 @@ def parse_prompts(prompts):
     return res, extra_data
 
 
-def get_user_metadata(filename):
+def get_user_metadata(filename, lister=None):
     if filename is None:
         return {}
 
@@ -215,7 +215,8 @@ def get_user_metadata(filename):
 
     metadata = {}
     try:
-        if os.path.isfile(metadata_filename):
+        exists = lister.exists(metadata_filename) if lister else os.path.exists(metadata_filename)
+        if exists:
             with open(metadata_filename, "r", encoding="utf8") as file:
                 metadata = json.load(file)
     except Exception as e:
diff --git a/modules/face_restoration_utils.py b/modules/face_restoration_utils.py
new file mode 100644
index 00000000000..1cbac236480
--- /dev/null
+++ b/modules/face_restoration_utils.py
@@ -0,0 +1,180 @@
+from __future__ import annotations
+
+import logging
+import os
+from functools import cached_property
+from typing import TYPE_CHECKING, Callable
+
+import cv2
+import numpy as np
+import torch
+
+from modules import devices, errors, face_restoration, shared
+
+if TYPE_CHECKING:
+    from facexlib.utils.face_restoration_helper import FaceRestoreHelper
+
+logger = logging.getLogger(__name__)
+
+
+def bgr_image_to_rgb_tensor(img: np.ndarray) -> torch.Tensor:
+    """Convert a BGR NumPy image in [0..1] range to a PyTorch RGB float32 tensor."""
+    assert img.shape[2] == 3, "image must be RGB"
+    if img.dtype == "float64":
+        img = img.astype("float32")
+    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    return torch.from_numpy(img.transpose(2, 0, 1)).float()
+
+
+def rgb_tensor_to_bgr_image(tensor: torch.Tensor, *, min_max=(0.0, 1.0)) -> np.ndarray:
+    """
+    Convert a PyTorch RGB tensor in range `min_max` to a BGR NumPy image in [0..1] range.
+    """
+    tensor = tensor.squeeze(0).float().detach().cpu().clamp_(*min_max)
+    tensor = (tensor - min_max[0]) / (min_max[1] - min_max[0])
+    assert tensor.dim() == 3, "tensor must be RGB"
+    img_np = tensor.numpy().transpose(1, 2, 0)
+    if img_np.shape[2] == 1:  # gray image, no RGB/BGR required
+        return np.squeeze(img_np, axis=2)
+    return cv2.cvtColor(img_np, cv2.COLOR_BGR2RGB)
+
+
+def create_face_helper(device) -> FaceRestoreHelper:
+    from facexlib.detection import retinaface
+    from facexlib.utils.face_restoration_helper import FaceRestoreHelper
+    if hasattr(retinaface, 'device'):
+        retinaface.device = device
+    return FaceRestoreHelper(
+        upscale_factor=1,
+        face_size=512,
+        crop_ratio=(1, 1),
+        det_model='retinaface_resnet50',
+        save_ext='png',
+        use_parse=True,
+        device=device,
+    )
+
+
+def restore_with_face_helper(
+    np_image: np.ndarray,
+    face_helper: FaceRestoreHelper,
+    restore_face: Callable[[torch.Tensor], torch.Tensor],
+) -> np.ndarray:
+    """
+    Find faces in the image using face_helper, restore them using restore_face, and paste them back into the image.
+
+    `restore_face` should take a cropped face image and return a restored face image.
+    """
+    from torchvision.transforms.functional import normalize
+    np_image = np_image[:, :, ::-1]
+    original_resolution = np_image.shape[0:2]
+
+    try:
+        logger.debug("Detecting faces...")
+        face_helper.clean_all()
+        face_helper.read_image(np_image)
+        face_helper.get_face_landmarks_5(only_center_face=False, resize=640, eye_dist_threshold=5)
+        face_helper.align_warp_face()
+        logger.debug("Found %d faces, restoring", len(face_helper.cropped_faces))
+        for cropped_face in face_helper.cropped_faces:
+            cropped_face_t = bgr_image_to_rgb_tensor(cropped_face / 255.0)
+            normalize(cropped_face_t, (0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True)
+            cropped_face_t = cropped_face_t.unsqueeze(0).to(devices.device_codeformer)
+
+            try:
+                with torch.no_grad():
+                    cropped_face_t = restore_face(cropped_face_t)
+                devices.torch_gc()
+            except Exception:
+                errors.report('Failed face-restoration inference', exc_info=True)
+
+            restored_face = rgb_tensor_to_bgr_image(cropped_face_t, min_max=(-1, 1))
+            restored_face = (restored_face * 255.0).astype('uint8')
+            face_helper.add_restored_face(restored_face)
+
+        logger.debug("Merging restored faces into image")
+        face_helper.get_inverse_affine(None)
+        img = face_helper.paste_faces_to_input_image()
+        img = img[:, :, ::-1]
+        if original_resolution != img.shape[0:2]:
+            img = cv2.resize(
+                img,
+                (0, 0),
+                fx=original_resolution[1] / img.shape[1],
+                fy=original_resolution[0] / img.shape[0],
+                interpolation=cv2.INTER_LINEAR,
+            )
+        logger.debug("Face restoration complete")
+    finally:
+        face_helper.clean_all()
+    return img
+
+
+class CommonFaceRestoration(face_restoration.FaceRestoration):
+    net: torch.Module | None
+    model_url: str
+    model_download_name: str
+
+    def __init__(self, model_path: str):
+        super().__init__()
+        self.net = None
+        self.model_path = model_path
+        os.makedirs(model_path, exist_ok=True)
+
+    @cached_property
+    def face_helper(self) -> FaceRestoreHelper:
+        return create_face_helper(self.get_device())
+
+    def send_model_to(self, device):
+        if self.net:
+            logger.debug("Sending %s to %s", self.net, device)
+            self.net.to(device)
+        if self.face_helper:
+            logger.debug("Sending face helper to %s", device)
+            self.face_helper.face_det.to(device)
+            self.face_helper.face_parse.to(device)
+
+    def get_device(self):
+        raise NotImplementedError("get_device must be implemented by subclasses")
+
+    def load_net(self) -> torch.Module:
+        raise NotImplementedError("load_net must be implemented by subclasses")
+
+    def restore_with_helper(
+        self,
+        np_image: np.ndarray,
+        restore_face: Callable[[torch.Tensor], torch.Tensor],
+    ) -> np.ndarray:
+        try:
+            if self.net is None:
+                self.net = self.load_net()
+        except Exception:
+            logger.warning("Unable to load face-restoration model", exc_info=True)
+            return np_image
+
+        try:
+            self.send_model_to(self.get_device())
+            return restore_with_face_helper(np_image, self.face_helper, restore_face)
+        finally:
+            if shared.opts.face_restoration_unload:
+                self.send_model_to(devices.cpu)
+
+
+def patch_facexlib(dirname: str) -> None:
+    import facexlib.detection
+    import facexlib.parsing
+
+    det_facex_load_file_from_url = facexlib.detection.load_file_from_url
+    par_facex_load_file_from_url = facexlib.parsing.load_file_from_url
+
+    def update_kwargs(kwargs):
+        return dict(kwargs, save_dir=dirname, model_dir=None)
+
+    def facex_load_file_from_url(**kwargs):
+        return det_facex_load_file_from_url(**update_kwargs(kwargs))
+
+    def facex_load_file_from_url2(**kwargs):
+        return par_facex_load_file_from_url(**update_kwargs(kwargs))
+
+    facexlib.detection.load_file_from_url = facex_load_file_from_url
+    facexlib.parsing.load_file_from_url = facex_load_file_from_url2
diff --git a/modules/gfpgan_model.py b/modules/gfpgan_model.py
index 01d668ecdaf..445b040925e 100644
--- a/modules/gfpgan_model.py
+++ b/modules/gfpgan_model.py
@@ -1,125 +1,71 @@
+from __future__ import annotations
+
+import logging
 import os
 
-import facexlib
-import gfpgan
+import torch
 
-import modules.face_restoration
-from modules import paths, shared, devices, modelloader, errors
+from modules import (
+    devices,
+    errors,
+    face_restoration,
+    face_restoration_utils,
+    modelloader,
+    shared,
+)
 
-model_dir = "GFPGAN"
-user_path = None
-model_path = os.path.join(paths.models_path, model_dir)
-model_file_path = None
+logger = logging.getLogger(__name__)
 model_url = "https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.4.pth"
-have_gfpgan = False
-loaded_gfpgan_model = None
-
-
-def gfpgann():
-    global loaded_gfpgan_model
-    global model_path
-    global model_file_path
-    if loaded_gfpgan_model is not None:
-        loaded_gfpgan_model.gfpgan.to(devices.device_gfpgan)
-        return loaded_gfpgan_model
-
-    if gfpgan_constructor is None:
-        return None
-
-    models = modelloader.load_models(model_path, model_url, user_path, ext_filter=['.pth'])
-
-    if len(models) == 1 and models[0].startswith("http"):
-        model_file = models[0]
-    elif len(models) != 0:
-        gfp_models = []
-        for item in models:
-            if 'GFPGAN' in os.path.basename(item):
-                gfp_models.append(item)
-        latest_file = max(gfp_models, key=os.path.getctime)
-        model_file = latest_file
-    else:
-        print("Unable to load gfpgan model!")
-        return None
-
-    if hasattr(facexlib.detection.retinaface, 'device'):
-        facexlib.detection.retinaface.device = devices.device_gfpgan
-    model_file_path = model_file
-    model = gfpgan_constructor(model_path=model_file, upscale=1, arch='clean', channel_multiplier=2, bg_upsampler=None, device=devices.device_gfpgan)
-    loaded_gfpgan_model = model
-
-    return model
-
-
-def send_model_to(model, device):
-    model.gfpgan.to(device)
-    model.face_helper.face_det.to(device)
-    model.face_helper.face_parse.to(device)
+model_download_name = "GFPGANv1.4.pth"
+gfpgan_face_restorer: face_restoration.FaceRestoration | None = None
+
+
+class FaceRestorerGFPGAN(face_restoration_utils.CommonFaceRestoration):
+    def name(self):
+        return "GFPGAN"
+
+    def get_device(self):
+        return devices.device_gfpgan
+
+    def load_net(self) -> torch.Module:
+        for model_path in modelloader.load_models(
+            model_path=self.model_path,
+            model_url=model_url,
+            command_path=self.model_path,
+            download_name=model_download_name,
+            ext_filter=['.pth'],
+        ):
+            if 'GFPGAN' in os.path.basename(model_path):
+                model = modelloader.load_spandrel_model(
+                    model_path,
+                    device=self.get_device(),
+                    expected_architecture='GFPGAN',
+                ).model
+                model.different_w = True  # see https://github.com/chaiNNer-org/spandrel/pull/81
+                return model
+        raise ValueError("No GFPGAN model found")
+
+    def restore(self, np_image):
+        def restore_face(cropped_face_t):
+            assert self.net is not None
+            return self.net(cropped_face_t, return_rgb=False)[0]
+
+        return self.restore_with_helper(np_image, restore_face)
 
 
 def gfpgan_fix_faces(np_image):
-    model = gfpgann()
-    if model is None:
-        return np_image
-
-    send_model_to(model, devices.device_gfpgan)
-
-    np_image_bgr = np_image[:, :, ::-1]
-    cropped_faces, restored_faces, gfpgan_output_bgr = model.enhance(np_image_bgr, has_aligned=False, only_center_face=False, paste_back=True)
-    np_image = gfpgan_output_bgr[:, :, ::-1]
-
-    model.face_helper.clean_all()
-
-    if shared.opts.face_restoration_unload:
-        send_model_to(model, devices.cpu)
-
+    if gfpgan_face_restorer:
+        return gfpgan_face_restorer.restore(np_image)
+    logger.warning("GFPGAN face restorer not set up")
     return np_image
 
 
-gfpgan_constructor = None
+def setup_model(dirname: str) -> None:
+    global gfpgan_face_restorer
 
-
-def setup_model(dirname):
     try:
-        os.makedirs(model_path, exist_ok=True)
-        from gfpgan import GFPGANer
-        from facexlib import detection, parsing  # noqa: F401
-        global user_path
-        global have_gfpgan
-        global gfpgan_constructor
-        global model_file_path
-
-        facexlib_path = model_path
-
-        if dirname is not None:
-            facexlib_path = dirname
-
-        load_file_from_url_orig = gfpgan.utils.load_file_from_url
-        facex_load_file_from_url_orig = facexlib.detection.load_file_from_url
-        facex_load_file_from_url_orig2 = facexlib.parsing.load_file_from_url
-
-        def my_load_file_from_url(**kwargs):
-            return load_file_from_url_orig(**dict(kwargs, model_dir=model_file_path))
-
-        def facex_load_file_from_url(**kwargs):
-            return facex_load_file_from_url_orig(**dict(kwargs, save_dir=facexlib_path, model_dir=None))
-
-        def facex_load_file_from_url2(**kwargs):
-            return facex_load_file_from_url_orig2(**dict(kwargs, save_dir=facexlib_path, model_dir=None))
-
-        gfpgan.utils.load_file_from_url = my_load_file_from_url
-        facexlib.detection.load_file_from_url = facex_load_file_from_url
-        facexlib.parsing.load_file_from_url = facex_load_file_from_url2
-        user_path = dirname
-        have_gfpgan = True
-        gfpgan_constructor = GFPGANer
-
-        class FaceRestorerGFPGAN(modules.face_restoration.FaceRestoration):
-            def name(self):
-                return "GFPGAN"
-
-            def restore(self, np_image):
-                return gfpgan_fix_faces(np_image)
-
-        shared.face_restorers.append(FaceRestorerGFPGAN())
+        face_restoration_utils.patch_facexlib(dirname)
+        gfpgan_face_restorer = FaceRestorerGFPGAN(model_path=dirname)
+        shared.face_restorers.append(gfpgan_face_restorer)
     except Exception:
         errors.report("Error setting up GFPGAN", exc_info=True)
diff --git a/modules/hashes.py b/modules/hashes.py
index b7a33b427c5..d22e5fadc47 100644
--- a/modules/hashes.py
+++ b/modules/hashes.py
@@ -21,7 +21,10 @@ def calculate_sha256(filename):
 
 def sha256_from_cache(filename, title, use_addnet_hash=False):
     hashes = cache("hashes-addnet") if use_addnet_hash else cache("hashes")
-    ondisk_mtime = os.path.getmtime(filename)
+    try:
+        ondisk_mtime = os.path.getmtime(filename)
+    except FileNotFoundError:
+        return None
 
     if title not in hashes:
         return None
diff --git a/modules/hat_model.py b/modules/hat_model.py
new file mode 100644
index 00000000000..7f2abb41660
--- /dev/null
+++ b/modules/hat_model.py
@@ -0,0 +1,43 @@
+import os
+import sys
+
+from modules import modelloader, devices
+from modules.shared import opts
+from modules.upscaler import Upscaler, UpscalerData
+from modules.upscaler_utils import upscale_with_model
+
+
+class UpscalerHAT(Upscaler):
+    def __init__(self, dirname):
+        self.name = "HAT"
+        self.scalers = []
+        self.user_path = dirname
+        super().__init__()
+        for file in self.find_models(ext_filter=[".pt", ".pth"]):
+            name = modelloader.friendly_name(file)
+            scale = 4  # TODO: scale might not be 4, but we can't know without loading the model
+            scaler_data = UpscalerData(name, file, upscaler=self, scale=scale)
+            self.scalers.append(scaler_data)
+
+    def do_upscale(self, img, selected_model):
+        try:
+            model = self.load_model(selected_model)
+        except Exception as e:
+            print(f"Unable to load HAT model {selected_model}: {e}", file=sys.stderr)
+            return img
+        model.to(devices.device_esrgan)  # TODO: should probably be device_hat
+        return upscale_with_model(
+            model,
+            img,
+            tile_size=opts.ESRGAN_tile,  # TODO: should probably be HAT_tile
+            tile_overlap=opts.ESRGAN_tile_overlap,  # TODO: should probably be HAT_tile_overlap
+        )
+
+    def load_model(self, path: str):
+        if not os.path.isfile(path):
+            raise FileNotFoundError(f"Model file {path} not found")
+        return modelloader.load_spandrel_model(
+            path,
+            device=devices.device_esrgan,  # TODO: should probably be device_hat
+            expected_architecture='HAT',
+        )
diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index be3e4648486..6082d9cb3e0 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -95,6 +95,7 @@ def __init__(self, dim, state_dict=None, layer_structure=None, activation_func=N
                         zeros_(b)
                     else:
                         raise KeyError(f"Key {weight_init} is not defined as initialization!")
+        devices.torch_npu_set_device()
         self.to(devices.device)
 
     def fix_old_state_dict(self, state_dict):
diff --git a/modules/images.py b/modules/images.py
index daf4eebe4b1..c50b2455dee 100644
--- a/modules/images.py
+++ b/modules/images.py
@@ -12,7 +12,7 @@
 import numpy as np
 import piexif
 import piexif.helper
-from PIL import Image, ImageFont, ImageDraw, ImageColor, PngImagePlugin
+from PIL import Image, ImageFont, ImageDraw, ImageColor, PngImagePlugin, ImageOps
 import string
 import json
 import hashlib
@@ -61,12 +61,17 @@ def image_grid(imgs, batch_size=1, rows=None):
     return grid
 
 
-Grid = namedtuple("Grid", ["tiles", "tile_w", "tile_h", "image_w", "image_h", "overlap"])
+class Grid(namedtuple("_Grid", ["tiles", "tile_w", "tile_h", "image_w", "image_h", "overlap"])):
+    @property
+    def tile_count(self) -> int:
+        """
+        The total number of tiles in the grid.
+        """
+        return sum(len(row[2]) for row in self.tiles)
 
 
-def split_grid(image, tile_w=512, tile_h=512, overlap=64):
-    w = image.width
-    h = image.height
+def split_grid(image: Image.Image, tile_w: int = 512, tile_h: int = 512, overlap: int = 64) -> Grid:
+    w, h = image.size
 
     non_overlap_width = tile_w - overlap
     non_overlap_height = tile_h - overlap
@@ -316,13 +321,16 @@ def resize(im, w, h):
     return res
 
 
-invalid_filename_chars = '<>:"/\\|?*\n\r\t'
+if not shared.cmd_opts.unix_filenames_sanitization:
+    invalid_filename_chars = '#<>:"/\\|?*\n\r\t'
+else:
+    invalid_filename_chars = '/'
 invalid_filename_prefix = ' '
 invalid_filename_postfix = ' .'
 re_nonletters = re.compile(r'[\s' + string.punctuation + ']+')
 re_pattern = re.compile(r"(.*?)(?:\[([^\[\]]+)\]|$)")
 re_pattern_arg = re.compile(r"(.*)<([^>]*)>$")
-max_filename_part_length = 128
+max_filename_part_length = shared.cmd_opts.filenames_max_length
 NOTHING_AND_SKIP_PREVIOUS_TEXT = object()
 
 
@@ -765,7 +773,7 @@ def image_data(data):
     import gradio as gr
 
     try:
-        image = Image.open(io.BytesIO(data))
+        image = read(io.BytesIO(data))
         textinfo, _ = read_info_from_image(image)
         return textinfo, None
     except Exception:
@@ -791,3 +799,31 @@ def flatten(img, bgcolor):
         img = background
 
     return img.convert('RGB')
+
+
+def read(fp, **kwargs):
+    image = Image.open(fp, **kwargs)
+    image = fix_image(image)
+
+    return image
+
+
+def fix_image(image: Image.Image):
+    if image is None:
+        return None
+
+    try:
+        image = ImageOps.exif_transpose(image)
+        image = fix_png_transparency(image)
+    except Exception:
+        pass
+
+    return image
+
+
+def fix_png_transparency(image: Image.Image):
+    if image.mode not in ("RGB", "P") or not isinstance(image.info.get("transparency"), bytes):
+        return image
+
+    image = image.convert("RGBA")
+    return image
diff --git a/modules/img2img.py b/modules/img2img.py
index c583290a0ea..e7fb3ea3c33 100644
--- a/modules/img2img.py
+++ b/modules/img2img.py
@@ -6,8 +6,8 @@
 from PIL import Image, ImageOps, ImageFilter, ImageEnhance, UnidentifiedImageError
 import gradio as gr
 
-from modules import images as imgutil
-from modules.generation_parameters_copypaste import create_override_settings_dict, parse_generation_parameters
+from modules import images
+from modules.infotext_utils import create_override_settings_dict, parse_generation_parameters
 from modules.processing import Processed, StableDiffusionProcessingImg2Img, process_images
 from modules.shared import opts, state
 from modules.sd_models import get_closet_checkpoint_match
@@ -21,7 +21,7 @@ def process_batch(p, input_dir, output_dir, inpaint_mask_dir, args, to_scale=Fal
     output_dir = output_dir.strip()
     processing.fix_seed(p)
 
-    images = list(shared.walk_files(input_dir, allowed_extensions=(".png", ".jpg", ".jpeg", ".webp", ".tif", ".tiff")))
+    batch_images = list(shared.walk_files(input_dir, allowed_extensions=(".png", ".jpg", ".jpeg", ".webp", ".tif", ".tiff")))
 
     is_inpaint_batch = False
     if inpaint_mask_dir:
@@ -31,9 +31,9 @@ def process_batch(p, input_dir, output_dir, inpaint_mask_dir, args, to_scale=Fal
         if is_inpaint_batch:
             print(f"\nInpaint batch is enabled. {len(inpaint_masks)} masks found.")
 
-    print(f"Will process {len(images)} images, creating {p.n_iter * p.batch_size} new images for each.")
+    print(f"Will process {len(batch_images)} images, creating {p.n_iter * p.batch_size} new images for each.")
 
-    state.job_count = len(images) * p.n_iter
+    state.job_count = len(batch_images) * p.n_iter
 
     # extract "default" params to use in case getting png info fails
     prompt = p.prompt
@@ -46,16 +46,16 @@ def process_batch(p, input_dir, output_dir, inpaint_mask_dir, args, to_scale=Fal
     sd_model_checkpoint_override = get_closet_checkpoint_match(override_settings.get("sd_model_checkpoint", None))
     batch_results = None
     discard_further_results = False
-    for i, image in enumerate(images):
-        state.job = f"{i+1} out of {len(images)}"
+    for i, image in enumerate(batch_images):
+        state.job = f"{i+1} out of {len(batch_images)}"
         if state.skipped:
             state.skipped = False
 
-        if state.interrupted:
+        if state.interrupted or state.stopping_generation:
             break
 
         try:
-            img = Image.open(image)
+            img = images.read(image)
         except UnidentifiedImageError as e:
             print(e)
             continue
@@ -86,7 +86,7 @@ def process_batch(p, input_dir, output_dir, inpaint_mask_dir, args, to_scale=Fal
                 # otherwise user has many masks with the same name but different extensions
                 mask_image_path = masks_found[0]
 
-            mask_image = Image.open(mask_image_path)
+            mask_image = images.read(mask_image_path)
             p.image_mask = mask_image
 
         if use_png_info:
@@ -94,8 +94,8 @@ def process_batch(p, input_dir, output_dir, inpaint_mask_dir, args, to_scale=Fal
                 info_img = img
                 if png_info_dir:
                     info_img_path = os.path.join(png_info_dir, os.path.basename(image))
-                    info_img = Image.open(info_img_path)
-                geninfo, _ = imgutil.read_info_from_image(info_img)
+                    info_img = images.read(info_img_path)
+                geninfo, _ = images.read_info_from_image(info_img)
                 parsed_parameters = parse_generation_parameters(geninfo)
                 parsed_parameters = {k: v for k, v in parsed_parameters.items() if k in (png_info_props or {})}
             except Exception:
@@ -175,9 +175,8 @@ def img2img(id_task: str, mode: int, prompt: str, negative_prompt: str, prompt_s
         image = None
         mask = None
 
-    # Use the EXIF orientation of photos taken by smartphones.
-    if image is not None:
-        image = ImageOps.exif_transpose(image)
+    image = images.fix_image(image)
+    mask = images.fix_image(mask)
 
     if selected_scale_tab == 1 and not is_batch:
         assert image, "Can't scale by because no image is selected"
@@ -222,9 +221,6 @@ def img2img(id_task: str, mode: int, prompt: str, negative_prompt: str, prompt_s
     if shared.opts.enable_console_prompts:
         print(f"\nimg2img: {prompt}", file=shared.progress_print_out)
 
-    if mask:
-        p.extra_generation_params["Mask blur"] = mask_blur
-
     with closing(p):
         if is_batch:
             assert not shared.cmd_opts.hide_ui_dir_config, "Launched with --hide-ui-dir-config, batch img2img disabled"
diff --git a/modules/generation_parameters_copypaste.py b/modules/infotext_utils.py
similarity index 76%
rename from modules/generation_parameters_copypaste.py
rename to modules/infotext_utils.py
index 4efe53e0c48..db1866449ad 100644
--- a/modules/generation_parameters_copypaste.py
+++ b/modules/infotext_utils.py
@@ -4,12 +4,15 @@
 import json
 import os
 import re
+import sys
 
 import gradio as gr
 from modules.paths import data_path
-from modules import shared, ui_tempdir, script_callbacks, processing
+from modules import shared, ui_tempdir, script_callbacks, processing, infotext_versions, images, prompt_parser
 from PIL import Image
 
+sys.modules['modules.generation_parameters_copypaste'] = sys.modules[__name__]  # alias for old name
+
 re_param_code = r'\s*(\w[\w \-/]+):\s*("(?:\\.|[^\\"])+"|[^,]*)(?:,|$)'
 re_param = re.compile(re_param_code)
 re_imagesize = re.compile(r"^(\d+)x(\d+)$")
@@ -28,6 +31,19 @@ def __init__(self, paste_button, tabname, source_text_component=None, source_ima
         self.paste_field_names = paste_field_names or []
 
 
+class PasteField(tuple):
+    def __new__(cls, component, target, *, api=None):
+        return super().__new__(cls, (component, target))
+
+    def __init__(self, component, target, *, api=None):
+        super().__init__()
+
+        self.api = api
+        self.component = component
+        self.label = target if isinstance(target, str) else None
+        self.function = target if callable(target) else None
+
+
 paste_fields: dict[str, dict] = {}
 registered_param_bindings: list[ParamBinding] = []
 
@@ -67,7 +83,7 @@ def image_from_url_text(filedata):
         assert is_in_right_dir, 'trying to open image file outside of allowed directories'
 
         filename = filename.rsplit('?', 1)[0]
-        return Image.open(filename)
+        return images.read(filename)
 
     if type(filedata) == list:
         if len(filedata) == 0:
@@ -79,11 +95,17 @@ def image_from_url_text(filedata):
         filedata = filedata[len("data:image/png;base64,"):]
 
     filedata = base64.decodebytes(filedata.encode('utf-8'))
-    image = Image.open(io.BytesIO(filedata))
+    image = images.read(io.BytesIO(filedata))
     return image
 
 
 def add_paste_fields(tabname, init_img, fields, override_settings_component=None):
+
+    if fields:
+        for i in range(len(fields)):
+            if not isinstance(fields[i], PasteField):
+                fields[i] = PasteField(*fields[i])
+
     paste_fields[tabname] = {"init_img": init_img, "fields": fields, "override_settings_component": override_settings_component}
 
     # backwards compatibility for existing extensions
@@ -208,7 +230,7 @@ def restore_old_hires_fix_params(res):
     res['Hires resize-2'] = height
 
 
-def parse_generation_parameters(x: str):
+def parse_generation_parameters(x: str, skip_fields: list[str] | None = None):
     """parses generation parameters string, the one you see in text field under the picture in UI:
 ```
 girl with an artist's beret, determined, blue eyes, desert scene, computer monitors, heavy makeup, by Alphonse Mucha and Charlie Bowater, ((eyeshadow)), (coquettish), detailed, intricate
@@ -218,6 +240,8 @@ def parse_generation_parameters(x: str):
 
     returns a dict with field values
     """
+    if skip_fields is None:
+        skip_fields = shared.opts.infotext_skip_pasting
 
     res = {}
 
@@ -290,6 +314,18 @@ def parse_generation_parameters(x: str):
     if "Hires negative prompt" not in res:
         res["Hires negative prompt"] = ""
 
+    if "Mask mode" not in res:
+        res["Mask mode"] = "Inpaint masked"
+
+    if "Masked content" not in res:
+        res["Masked content"] = 'original'
+
+    if "Inpaint area" not in res:
+        res["Inpaint area"] = "Whole picture"
+
+    if "Masked area padding" not in res:
+        res["Masked area padding"] = 32
+
     restore_old_hires_fix_params(res)
 
     # Missing RNG means the default was set, which is GPU RNG
@@ -314,8 +350,25 @@ def parse_generation_parameters(x: str):
     if "VAE Decoder" not in res:
         res["VAE Decoder"] = "Full"
 
-    skip = set(shared.opts.infotext_skip_pasting)
-    res = {k: v for k, v in res.items() if k not in skip}
+    if "FP8 weight" not in res:
+        res["FP8 weight"] = "Disable"
+
+    if "Cache FP16 weight for LoRA" not in res and res["FP8 weight"] != "Disable":
+        res["Cache FP16 weight for LoRA"] = False
+
+    prompt_attention = prompt_parser.parse_prompt_attention(prompt)
+    prompt_attention += prompt_parser.parse_prompt_attention(negative_prompt)
+    prompt_uses_emphasis = len(prompt_attention) != len([p for p in prompt_attention if p[1] == 1.0 or p[0] == 'BREAK'])
+    if "Emphasis" not in res and prompt_uses_emphasis:
+        res["Emphasis"] = "Original"
+
+    if "Refiner switch by sampling steps" not in res:
+        res["Refiner switch by sampling steps"] = False
+
+    infotext_versions.backcompat(res)
+
+    for key in skip_fields:
+        res.pop(key, None)
 
     return res
 
@@ -365,13 +418,57 @@ def create_override_settings_dict(text_pairs):
     return res
 
 
+def get_override_settings(params, *, skip_fields=None):
+    """Returns a list of settings overrides from the infotext parameters dictionary.
+
+    This function checks the `params` dictionary for any keys that correspond to settings in `shared.opts` and returns
+    a list of tuples containing the parameter name, setting name, and new value cast to correct type.
+
+    It checks for conditions before adding an override:
+    - ignores settings that match the current value
+    - ignores parameter keys present in skip_fields argument.
+
+    Example input:
+        {"Clip skip": "2"}
+
+    Example output:
+        [("Clip skip", "CLIP_stop_at_last_layers", 2)]
+    """
+
+    res = []
+
+    mapping = [(info.infotext, k) for k, info in shared.opts.data_labels.items() if info.infotext]
+    for param_name, setting_name in mapping + infotext_to_setting_name_mapping:
+        if param_name in (skip_fields or {}):
+            continue
+
+        v = params.get(param_name, None)
+        if v is None:
+            continue
+
+        if setting_name == "sd_model_checkpoint" and shared.opts.disable_weights_auto_swap:
+            continue
+
+        v = shared.opts.cast_value(setting_name, v)
+        current_value = getattr(shared.opts, setting_name, None)
+
+        if v == current_value:
+            continue
+
+        res.append((param_name, setting_name, v))
+
+    return res
+
+
 def connect_paste(button, paste_fields, input_comp, override_settings_component, tabname):
     def paste_func(prompt):
         if not prompt and not shared.cmd_opts.hide_ui_dir_config:
             filename = os.path.join(data_path, "params.txt")
-            if os.path.exists(filename):
+            try:
                 with open(filename, "r", encoding="utf8") as file:
                     prompt = file.read()
+            except OSError:
+                pass
 
         params = parse_generation_parameters(prompt)
         script_callbacks.infotext_pasted_callback(prompt, params)
@@ -393,6 +490,8 @@ def paste_func(prompt):
 
                     if valtype == bool and v == "False":
                         val = False
+                    elif valtype == int:
+                        val = float(v)
                     else:
                         val = valtype(v)
 
@@ -406,29 +505,9 @@ def paste_func(prompt):
         already_handled_fields = {key: 1 for _, key in paste_fields}
 
         def paste_settings(params):
-            vals = {}
-
-            mapping = [(info.infotext, k) for k, info in shared.opts.data_labels.items() if info.infotext]
-            for param_name, setting_name in mapping + infotext_to_setting_name_mapping:
-                if param_name in already_handled_fields:
-                    continue
-
-                v = params.get(param_name, None)
-                if v is None:
-                    continue
-
-                if setting_name == "sd_model_checkpoint" and shared.opts.disable_weights_auto_swap:
-                    continue
-
-                v = shared.opts.cast_value(setting_name, v)
-                current_value = getattr(shared.opts, setting_name, None)
-
-                if v == current_value:
-                    continue
-
-                vals[param_name] = v
+            vals = get_override_settings(params, skip_fields=already_handled_fields)
 
-            vals_pairs = [f"{k}: {v}" for k, v in vals.items()]
+            vals_pairs = [f"{infotext_text}: {value}" for infotext_text, setting_name, value in vals]
 
             return gr.Dropdown.update(value=vals_pairs, choices=vals_pairs, visible=bool(vals_pairs))
 
diff --git a/modules/infotext_versions.py b/modules/infotext_versions.py
new file mode 100644
index 00000000000..b5552a31286
--- /dev/null
+++ b/modules/infotext_versions.py
@@ -0,0 +1,45 @@
+from modules import shared
+from packaging import version
+import re
+
+
+v160 = version.parse("1.6.0")
+v170_tsnr = version.parse("v1.7.0-225")
+v180 = version.parse("1.8.0")
+
+
+def parse_version(text):
+    if text is None:
+        return None
+
+    m = re.match(r'([^-]+-[^-]+)-.*', text)
+    if m:
+        text = m.group(1)
+
+    try:
+        return version.parse(text)
+    except Exception:
+        return None
+
+
+def backcompat(d):
+    """Checks infotext Version field, and enables backwards compatibility options according to it."""
+
+    if not shared.opts.auto_backcompat:
+        return
+
+    ver = parse_version(d.get("Version"))
+    if ver is None:
+        return
+
+    if ver < v160 and '[' in d.get('Prompt', ''):
+        d["Old prompt editing timelines"] = True
+
+    if ver < v160 and d.get('Sampler', '') in ('DDIM', 'PLMS'):
+        d["Pad conds v0"] = True
+
+    if ver < v170_tsnr:
+        d["Downcast alphas_cumprod"] = True
+
+    if ver < v180 and d.get('Refiner'):
+        d["Refiner switch by sampling steps"] = True
diff --git a/modules/initialize.py b/modules/initialize.py
index ac95fc6f00c..08ad4c0b0e9 100644
--- a/modules/initialize.py
+++ b/modules/initialize.py
@@ -1,5 +1,6 @@
 import importlib
 import logging
+import os
 import sys
 import warnings
 from threading import Thread
@@ -18,6 +19,7 @@ def imports():
     warnings.filterwarnings(action="ignore", category=DeprecationWarning, module="pytorch_lightning")
     warnings.filterwarnings(action="ignore", category=UserWarning, module="torchvision")
 
+    os.environ.setdefault('GRADIO_ANALYTICS_ENABLED', 'False')
     import gradio  # noqa: F401
     startup_timer.record("import gradio")
 
@@ -54,9 +56,6 @@ def initialize():
     initialize_util.configure_sigint_handler()
     initialize_util.configure_opts_onchange()
 
-    from modules import modelloader
-    modelloader.cleanup_models()
-
     from modules import sd_models
     sd_models.setup_model()
     startup_timer.record("setup SD model")
@@ -140,16 +139,17 @@ def load_model():
         """
         Accesses shared.sd_model property to load model.
         After it's available, if it has been loaded before this access by some extension,
-        its optimization may be None because the list of optimizaers has neet been filled
+        its optimization may be None because the list of optimizers has not been filled
         by that time, so we apply optimization again.
         """
+        from modules import devices
+        devices.torch_npu_set_device()
 
         shared.sd_model  # noqa: B018
 
         if sd_hijack.current_optimizer is None:
             sd_hijack.apply_optimizations()
 
-        from modules import devices
         devices.first_time_calculation()
     if not shared.cmd_opts.skip_load_model_at_start:
         Thread(target=load_model).start()
diff --git a/modules/initialize_util.py b/modules/initialize_util.py
index 2e9b6d895f4..b6767138dd5 100644
--- a/modules/initialize_util.py
+++ b/modules/initialize_util.py
@@ -177,6 +177,8 @@ def configure_opts_onchange():
     shared.opts.onchange("temp_dir", ui_tempdir.on_tmpdir_changed)
     shared.opts.onchange("gradio_theme", shared.reload_gradio_theme)
     shared.opts.onchange("cross_attention_optimization", wrap_queued_call(lambda: sd_hijack.model_hijack.redo_hijack(shared.sd_model)), call=False)
+    shared.opts.onchange("fp8_storage", wrap_queued_call(lambda: sd_models.reload_model_weights()), call=False)
+    shared.opts.onchange("cache_fp16_weight", wrap_queued_call(lambda: sd_models.reload_model_weights(forced_reload=True)), call=False)
     startup_timer.record("opts onchange")
 
 
diff --git a/modules/interrogate.py b/modules/interrogate.py
index 3045560d0ae..c93e7aa861d 100644
--- a/modules/interrogate.py
+++ b/modules/interrogate.py
@@ -10,14 +10,14 @@
 from torchvision import transforms
 from torchvision.transforms.functional import InterpolationMode
 
-from modules import devices, paths, shared, lowvram, modelloader, errors
+from modules import devices, paths, shared, lowvram, modelloader, errors, torch_utils
 
 blip_image_eval_size = 384
 clip_model_name = 'ViT-L/14'
 
 Category = namedtuple("Category", ["name", "topn", "items"])
 
-re_topn = re.compile(r"\.top(\d+)\.")
+re_topn = re.compile(r"\.top(\d+)$")
 
 def category_types():
     return [f.stem for f in Path(shared.interrogator.content_dir).glob('*.txt')]
@@ -131,7 +131,7 @@ def load(self):
 
         self.clip_model = self.clip_model.to(devices.device_interrogate)
 
-        self.dtype = next(self.clip_model.parameters()).dtype
+        self.dtype = torch_utils.get_param(self.clip_model).dtype
 
     def send_clip_to_ram(self):
         if not shared.opts.interrogate_keep_models_in_memory:
diff --git a/modules/launch_utils.py b/modules/launch_utils.py
index 29506f24965..5812b0e5855 100644
--- a/modules/launch_utils.py
+++ b/modules/launch_utils.py
@@ -27,8 +27,7 @@
 # Whether to default to printing command output
 default_command_live = (os.environ.get('WEBUI_LAUNCH_LIVE_OUTPUT') == "1")
 
-if 'GRADIO_ANALYTICS_ENABLED' not in os.environ:
-    os.environ['GRADIO_ANALYTICS_ENABLED'] = 'False'
+os.environ.setdefault('GRADIO_ANALYTICS_ENABLED', 'False')
 
 
 def check_python_version():
@@ -56,7 +55,7 @@ def check_python_version():
 
 You can download 3.10 Python from here: https://www.python.org/downloads/release/python-3106/
 
-{"Alternatively, use a binary release of WebUI: https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases" if is_windows else ""}
+{"Alternatively, use a binary release of WebUI: https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre" if is_windows else ""}
 
 Use --skip-python-version-check to suppress this warning.
 """)
@@ -189,7 +188,7 @@ def git_clone(url, dir, name, commithash=None):
         return
 
     try:
-        run(f'"{git}" clone "{url}" "{dir}"', f"Cloning {name} into {dir}...", f"Couldn't clone {name}", live=True)
+        run(f'"{git}" clone --config core.filemode=false "{url}" "{dir}"', f"Cloning {name} into {dir}...", f"Couldn't clone {name}", live=True)
     except RuntimeError:
         shutil.rmtree(dir, ignore_errors=True)
         raise
@@ -245,11 +244,13 @@ def list_extensions(settings_file):
     settings = {}
 
     try:
-        if os.path.isfile(settings_file):
-            with open(settings_file, "r", encoding="utf8") as file:
-                settings = json.load(file)
+        with open(settings_file, "r", encoding="utf8") as file:
+            settings = json.load(file)
+    except FileNotFoundError:
+        pass
     except Exception:
-        errors.report("Could not load settings", exc_info=True)
+        errors.report(f'\nCould not load settings\nThe config file "{settings_file}" is likely corrupted\nIt has been moved to the "tmp/config.json"\nReverting config to default\n\n''', exc_info=True)
+        os.replace(settings_file, os.path.join(script_path, "tmp", "config.json"))
 
     disabled_extensions = set(settings.get('disabled_extensions', []))
     disable_all_extensions = settings.get('disable_all_extensions', 'none')
@@ -314,8 +315,8 @@ def requirements_met(requirements_file):
 
 
 def prepare_environment():
-    torch_index_url = os.environ.get('TORCH_INDEX_URL', "https://download.pytorch.org/whl/cu118")
-    torch_command = os.environ.get('TORCH_COMMAND', f"pip install torch==2.0.1 torchvision==0.15.2 --extra-index-url {torch_index_url}")
+    torch_index_url = os.environ.get('TORCH_INDEX_URL', "https://download.pytorch.org/whl/cu121")
+    torch_command = os.environ.get('TORCH_COMMAND', f"pip install torch==2.1.2 torchvision==0.16.2 --extra-index-url {torch_index_url}")
     if args.use_ipex:
         if platform.system() == "Windows":
             # The "Nuullll/intel-extension-for-pytorch" wheels were built from IPEX source for Intel Arc GPU: https://github.com/intel/intel-extension-for-pytorch/tree/xpu-main
@@ -337,21 +338,22 @@ def prepare_environment():
             torch_index_url = os.environ.get('TORCH_INDEX_URL', "https://pytorch-extension.intel.com/release-whl/stable/xpu/us/")
             torch_command = os.environ.get('TORCH_COMMAND', f"pip install torch==2.0.0a0 intel-extension-for-pytorch==2.0.110+gitba7f6c1 --extra-index-url {torch_index_url}")
     requirements_file = os.environ.get('REQS_FILE', "requirements_versions.txt")
+    requirements_file_for_npu = os.environ.get('REQS_FILE_FOR_NPU', "requirements_npu.txt")
 
-    xformers_package = os.environ.get('XFORMERS_PACKAGE', 'xformers==0.0.20')
+    xformers_package = os.environ.get('XFORMERS_PACKAGE', 'xformers==0.0.23.post1')
     clip_package = os.environ.get('CLIP_PACKAGE', "https://github.com/openai/CLIP/archive/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1.zip")
     openclip_package = os.environ.get('OPENCLIP_PACKAGE', "https://github.com/mlfoundations/open_clip/archive/bb6e834e9c70d9c27d0dc3ecedeebeaeb1ffad6b.zip")
 
+    assets_repo = os.environ.get('ASSETS_REPO', "https://github.com/AUTOMATIC1111/stable-diffusion-webui-assets.git")
     stable_diffusion_repo = os.environ.get('STABLE_DIFFUSION_REPO', "https://github.com/Stability-AI/stablediffusion.git")
     stable_diffusion_xl_repo = os.environ.get('STABLE_DIFFUSION_XL_REPO', "https://github.com/Stability-AI/generative-models.git")
     k_diffusion_repo = os.environ.get('K_DIFFUSION_REPO', 'https://github.com/crowsonkb/k-diffusion.git')
-    codeformer_repo = os.environ.get('CODEFORMER_REPO', 'https://github.com/sczhou/CodeFormer.git')
     blip_repo = os.environ.get('BLIP_REPO', 'https://github.com/salesforce/BLIP.git')
 
+    assets_commit_hash = os.environ.get('ASSETS_COMMIT_HASH', "6f7db241d2f8ba7457bac5ca9753331f0c266917")
     stable_diffusion_commit_hash = os.environ.get('STABLE_DIFFUSION_COMMIT_HASH', "cf1d67a6fd5ea1aa600c4df58e5b47da45f6bdbf")
     stable_diffusion_xl_commit_hash = os.environ.get('STABLE_DIFFUSION_XL_COMMIT_HASH', "45c443b316737a4ab6e40413d7794a7f5657c19f")
     k_diffusion_commit_hash = os.environ.get('K_DIFFUSION_COMMIT_HASH', "ab527a9a6d347f364e3d185ba6d714e22d80cb3c")
-    codeformer_commit_hash = os.environ.get('CODEFORMER_COMMIT_HASH', "c5b4593074ba6214284d6acd5f1719b6c5d739af")
     blip_commit_hash = os.environ.get('BLIP_COMMIT_HASH', "48211a1594f1321b00f14c9f7a5b4813144b2fb9")
 
     try:
@@ -405,18 +407,14 @@ def prepare_environment():
 
     os.makedirs(os.path.join(script_path, dir_repos), exist_ok=True)
 
+    git_clone(assets_repo, repo_dir('stable-diffusion-webui-assets'), "assets", assets_commit_hash)
     git_clone(stable_diffusion_repo, repo_dir('stable-diffusion-stability-ai'), "Stable Diffusion", stable_diffusion_commit_hash)
     git_clone(stable_diffusion_xl_repo, repo_dir('generative-models'), "Stable Diffusion XL", stable_diffusion_xl_commit_hash)
     git_clone(k_diffusion_repo, repo_dir('k-diffusion'), "K-diffusion", k_diffusion_commit_hash)
-    git_clone(codeformer_repo, repo_dir('CodeFormer'), "CodeFormer", codeformer_commit_hash)
     git_clone(blip_repo, repo_dir('BLIP'), "BLIP", blip_commit_hash)
 
     startup_timer.record("clone repositores")
 
-    if not is_installed("lpips"):
-        run_pip(f"install -r \"{os.path.join(repo_dir('CodeFormer'), 'requirements.txt')}\"", "requirements for CodeFormer")
-        startup_timer.record("install CodeFormer requirements")
-
     if not os.path.isfile(requirements_file):
         requirements_file = os.path.join(script_path, requirements_file)
 
@@ -424,6 +422,13 @@ def prepare_environment():
         run_pip(f"install -r \"{requirements_file}\"", "requirements")
         startup_timer.record("install requirements")
 
+    if not os.path.isfile(requirements_file_for_npu):
+        requirements_file_for_npu = os.path.join(script_path, requirements_file_for_npu)
+
+    if "torch_npu" in torch_command and not requirements_met(requirements_file_for_npu):
+        run_pip(f"install -r \"{requirements_file_for_npu}\"", "requirements_for_npu")
+        startup_timer.record("install requirements_for_npu")
+
     if not args.skip_install:
         run_extensions_installers(settings_file=args.ui_settings_file)
 
diff --git a/modules/logging_config.py b/modules/logging_config.py
index 79269875608..8e31d8c9fd1 100644
--- a/modules/logging_config.py
+++ b/modules/logging_config.py
@@ -1,41 +1,58 @@
-import os
 import logging
+import os
 
 try:
-    from tqdm.auto import tqdm
+    from tqdm import tqdm
+
 
     class TqdmLoggingHandler(logging.Handler):
-        def __init__(self, level=logging.INFO):
-            super().__init__(level)
+        def __init__(self, fallback_handler: logging.Handler):
+            super().__init__()
+            self.fallback_handler = fallback_handler
 
         def emit(self, record):
             try:
-                msg = self.format(record)
-                tqdm.write(msg)
-                self.flush()
+                # If there are active tqdm progress bars,
+                # attempt to not interfere with them.
+                if tqdm._instances:
+                    tqdm.write(self.format(record))
+                else:
+                    self.fallback_handler.emit(record)
             except Exception:
-                self.handleError(record)
+                self.fallback_handler.emit(record)
 
-    TQDM_IMPORTED = True
 except ImportError:
-    # tqdm does not exist before first launch
-    # I will import once the UI finishes seting up the enviroment and reloads.
-    TQDM_IMPORTED = False
+    TqdmLoggingHandler = None
+
 
 def setup_logging(loglevel):
     if loglevel is None:
         loglevel = os.environ.get("SD_WEBUI_LOG_LEVEL")
 
-    loghandlers = []
+    if not loglevel:
+        return
+
+    if logging.root.handlers:
+        # Already configured, do not interfere
+        return
+
+    formatter = logging.Formatter(
+        '%(asctime)s %(levelname)s [%(name)s] %(message)s',
+        '%Y-%m-%d %H:%M:%S',
+    )
+
+    if os.environ.get("SD_WEBUI_RICH_LOG"):
+        from rich.logging import RichHandler
+        handler = RichHandler()
+    else:
+        handler = logging.StreamHandler()
+        handler.setFormatter(formatter)
+
+    if TqdmLoggingHandler:
+        handler = TqdmLoggingHandler(handler)
 
-    if TQDM_IMPORTED:
-        loghandlers.append(TqdmLoggingHandler())
+    handler.setFormatter(formatter)
 
-    if loglevel:
-        log_level = getattr(logging, loglevel.upper(), None) or logging.INFO
-        logging.basicConfig(
-            level=log_level,
-            format='%(asctime)s %(levelname)s [%(name)s] %(message)s',
-            datefmt='%Y-%m-%d %H:%M:%S',
-            handlers=loghandlers
-        )
+    log_level = getattr(logging, loglevel.upper(), None) or logging.INFO
+    logging.root.setLevel(log_level)
+    logging.root.addHandler(handler)
diff --git a/modules/mac_specific.py b/modules/mac_specific.py
index d96d86d792c..039689f32e1 100644
--- a/modules/mac_specific.py
+++ b/modules/mac_specific.py
@@ -12,7 +12,7 @@
 
 # before torch version 1.13, has_mps is only available in nightly pytorch and macOS 12.3+,
 # use check `getattr` and try it for compatibility.
-# in torch version 1.13, backends.mps.is_available() and backends.mps.is_built() are introduced in to check mps availabilty,
+# in torch version 1.13, backends.mps.is_available() and backends.mps.is_built() are introduced in to check mps availability,
 # since torch 2.0.1+ nightly build, getattr(torch, 'has_mps', False) was deprecated, see https://github.com/pytorch/pytorch/pull/103279
 def check_for_mps() -> bool:
     if version.parse(torch.__version__) <= version.parse("2.0.1"):
diff --git a/modules/masking.py b/modules/masking.py
index be9f84c76f6..29a3945278e 100644
--- a/modules/masking.py
+++ b/modules/masking.py
@@ -3,40 +3,15 @@
 
 def get_crop_region(mask, pad=0):
     """finds a rectangular region that contains all masked ares in an image. Returns (x1, y1, x2, y2) coordinates of the rectangle.
-    For example, if a user has painted the top-right part of a 512x512 image", the result may be (256, 0, 512, 256)"""
-
-    h, w = mask.shape
-
-    crop_left = 0
-    for i in range(w):
-        if not (mask[:, i] == 0).all():
-            break
-        crop_left += 1
-
-    crop_right = 0
-    for i in reversed(range(w)):
-        if not (mask[:, i] == 0).all():
-            break
-        crop_right += 1
-
-    crop_top = 0
-    for i in range(h):
-        if not (mask[i] == 0).all():
-            break
-        crop_top += 1
-
-    crop_bottom = 0
-    for i in reversed(range(h)):
-        if not (mask[i] == 0).all():
-            break
-        crop_bottom += 1
-
-    return (
-        int(max(crop_left-pad, 0)),
-        int(max(crop_top-pad, 0)),
-        int(min(w - crop_right + pad, w)),
-        int(min(h - crop_bottom + pad, h))
-    )
+    For example, if a user has painted the top-right part of a 512x512 image, the result may be (256, 0, 512, 256)"""
+    mask_img = mask if isinstance(mask, Image.Image) else Image.fromarray(mask)
+    box = mask_img.getbbox()
+    if box:
+        x1, y1, x2, y2 = box
+    else:  # when no box is found
+        x1, y1 = mask_img.size
+        x2 = y2 = 0
+    return max(x1 - pad, 0), max(y1 - pad, 0), min(x2 + pad, mask_img.size[0]), min(y2 + pad, mask_img.size[1])
 
 
 def expand_crop_region(crop_region, processing_width, processing_height, image_width, image_height):
diff --git a/modules/modelloader.py b/modules/modelloader.py
index 098bcb79336..115415c8e65 100644
--- a/modules/modelloader.py
+++ b/modules/modelloader.py
@@ -1,13 +1,20 @@
 from __future__ import annotations
 
-import os
-import shutil
 import importlib
+import logging
+import os
+from typing import TYPE_CHECKING
 from urllib.parse import urlparse
 
+import torch
+
 from modules import shared
 from modules.upscaler import Upscaler, UpscalerLanczos, UpscalerNearest, UpscalerNone
-from modules.paths import script_path, models_path
+
+if TYPE_CHECKING:
+    import spandrel
+
+logger = logging.getLogger(__name__)
 
 
 def load_file_from_url(
@@ -90,54 +97,6 @@ def friendly_name(file: str):
     return model_name
 
 
-def cleanup_models():
-    # This code could probably be more efficient if we used a tuple list or something to store the src/destinations
-    # and then enumerate that, but this works for now. In the future, it'd be nice to just have every "model" scaler
-    # somehow auto-register and just do these things...
-    root_path = script_path
-    src_path = models_path
-    dest_path = os.path.join(models_path, "Stable-diffusion")
-    move_files(src_path, dest_path, ".ckpt")
-    move_files(src_path, dest_path, ".safetensors")
-    src_path = os.path.join(root_path, "ESRGAN")
-    dest_path = os.path.join(models_path, "ESRGAN")
-    move_files(src_path, dest_path)
-    src_path = os.path.join(models_path, "BSRGAN")
-    dest_path = os.path.join(models_path, "ESRGAN")
-    move_files(src_path, dest_path, ".pth")
-    src_path = os.path.join(root_path, "gfpgan")
-    dest_path = os.path.join(models_path, "GFPGAN")
-    move_files(src_path, dest_path)
-    src_path = os.path.join(root_path, "SwinIR")
-    dest_path = os.path.join(models_path, "SwinIR")
-    move_files(src_path, dest_path)
-    src_path = os.path.join(root_path, "repositories/latent-diffusion/experiments/pretrained_models/")
-    dest_path = os.path.join(models_path, "LDSR")
-    move_files(src_path, dest_path)
-
-
-def move_files(src_path: str, dest_path: str, ext_filter: str = None):
-    try:
-        os.makedirs(dest_path, exist_ok=True)
-        if os.path.exists(src_path):
-            for file in os.listdir(src_path):
-                fullpath = os.path.join(src_path, file)
-                if os.path.isfile(fullpath):
-                    if ext_filter is not None:
-                        if ext_filter not in file:
-                            continue
-                    print(f"Moving {file} from {src_path} to {dest_path}.")
-                    try:
-                        shutil.move(fullpath, dest_path)
-                    except Exception:
-                        pass
-            if len(os.listdir(src_path)) == 0:
-                print(f"Removing empty folder: {src_path}")
-                shutil.rmtree(src_path, True)
-    except Exception:
-        pass
-
-
 def load_upscalers():
     # We can only do this 'magic' method to dynamically load upscalers if they are referenced,
     # so we'll try to import any _model.py files before looking in __subclasses__
@@ -151,7 +110,7 @@ def load_upscalers():
             except Exception:
                 pass
 
-    datas = []
+    data = []
     commandline_options = vars(shared.cmd_opts)
 
     # some of upscaler classes will not go away after reloading their modules, and we'll end
@@ -170,10 +129,41 @@ def load_upscalers():
         scaler = cls(commandline_model_path)
         scaler.user_path = commandline_model_path
         scaler.model_download_path = commandline_model_path or scaler.model_path
-        datas += scaler.scalers
+        data += scaler.scalers
 
     shared.sd_upscalers = sorted(
-        datas,
+        data,
         # Special case for UpscalerNone keeps it at the beginning of the list.
         key=lambda x: x.name.lower() if not isinstance(x.scaler, (UpscalerNone, UpscalerLanczos, UpscalerNearest)) else ""
     )
+
+
+def load_spandrel_model(
+    path: str | os.PathLike,
+    *,
+    device: str | torch.device | None,
+    prefer_half: bool = False,
+    dtype: str | torch.dtype | None = None,
+    expected_architecture: str | None = None,
+) -> spandrel.ModelDescriptor:
+    import spandrel
+    model_descriptor = spandrel.ModelLoader(device=device).load_from_file(str(path))
+    if expected_architecture and model_descriptor.architecture != expected_architecture:
+        logger.warning(
+            f"Model {path!r} is not a {expected_architecture!r} model (got {model_descriptor.architecture!r})",
+        )
+    half = False
+    if prefer_half:
+        if model_descriptor.supports_half:
+            model_descriptor.model.half()
+            half = True
+        else:
+            logger.info("Model %s does not support half precision, ignoring --half", path)
+    if dtype:
+        model_descriptor.model.to(dtype=dtype)
+    model_descriptor.model.eval()
+    logger.debug(
+        "Loaded %s from %s (device=%s, half=%s, dtype=%s)",
+        model_descriptor, path, device, half, dtype,
+    )
+    return model_descriptor
diff --git a/modules/models/diffusion/ddpm_edit.py b/modules/models/diffusion/ddpm_edit.py
index 6db340da40b..7b51c83c5d9 100644
--- a/modules/models/diffusion/ddpm_edit.py
+++ b/modules/models/diffusion/ddpm_edit.py
@@ -341,7 +341,7 @@ def p_losses(self, x_start, t, noise=None):
         elif self.parameterization == "x0":
             target = x_start
         else:
-            raise NotImplementedError(f"Paramterization {self.parameterization} not yet supported")
+            raise NotImplementedError(f"Parameterization {self.parameterization} not yet supported")
 
         loss = self.get_loss(model_out, target, mean=False).mean(dim=[1, 2, 3])
 
@@ -901,7 +901,7 @@ def forward(self, x, c, *args, **kwargs):
     def apply_model(self, x_noisy, t, cond, return_ids=False):
 
         if isinstance(cond, dict):
-            # hybrid case, cond is exptected to be a dict
+            # hybrid case, cond is expected to be a dict
             pass
         else:
             if not isinstance(cond, list):
@@ -937,7 +937,7 @@ def apply_model(self, x_noisy, t, cond, return_ids=False):
                 cond_list = [{c_key: [c[:, :, :, :, i]]} for i in range(c.shape[-1])]
 
             elif self.cond_stage_key == 'coordinates_bbox':
-                assert 'original_image_size' in self.split_input_params, 'BoudingBoxRescaling is missing original_image_size'
+                assert 'original_image_size' in self.split_input_params, 'BoundingBoxRescaling is missing original_image_size'
 
                 # assuming padding of unfold is always 0 and its dilation is always 1
                 n_patches_per_row = int((w - ks[0]) / stride[0] + 1)
@@ -947,7 +947,7 @@ def apply_model(self, x_noisy, t, cond, return_ids=False):
                 num_downs = self.first_stage_model.encoder.num_resolutions - 1
                 rescale_latent = 2 ** (num_downs)
 
-                # get top left postions of patches as conforming for the bbbox tokenizer, therefore we
+                # get top left positions of patches as conforming for the bbbox tokenizer, therefore we
                 # need to rescale the tl patch coordinates to be in between (0,1)
                 tl_patch_coordinates = [(rescale_latent * stride[0] * (patch_nr % n_patches_per_row) / full_img_w,
                                          rescale_latent * stride[1] * (patch_nr // n_patches_per_row) / full_img_h)
diff --git a/modules/npu_specific.py b/modules/npu_specific.py
new file mode 100644
index 00000000000..9410069110f
--- /dev/null
+++ b/modules/npu_specific.py
@@ -0,0 +1,31 @@
+import importlib
+import torch
+
+from modules import shared
+
+
+def check_for_npu():
+    if importlib.util.find_spec("torch_npu") is None:
+        return False
+    import torch_npu
+
+    try:
+        # Will raise a RuntimeError if no NPU is found
+        _ = torch_npu.npu.device_count()
+        return torch.npu.is_available()
+    except RuntimeError:
+        return False
+
+
+def get_npu_device_string():
+    if shared.cmd_opts.device_id is not None:
+        return f"npu:{shared.cmd_opts.device_id}"
+    return "npu:0"
+
+
+def torch_npu_gc():
+    with torch.npu.device(get_npu_device_string()):
+        torch.npu.empty_cache()
+
+
+has_npu = check_for_npu()
diff --git a/modules/options.py b/modules/options.py
index 4fead690cea..35ccade25be 100644
--- a/modules/options.py
+++ b/modules/options.py
@@ -1,3 +1,4 @@
+import os
 import json
 import sys
 from dataclasses import dataclass
@@ -6,6 +7,7 @@
 
 from modules import errors
 from modules.shared_cmd_options import cmd_opts
+from modules.paths_internal import script_path
 
 
 class OptionInfo:
@@ -91,18 +93,35 @@ def __setattr__(self, key, value):
 
         if self.data is not None:
             if key in self.data or key in self.data_labels:
+
+                # Check that settings aren't globally frozen
                 assert not cmd_opts.freeze_settings, "changing settings is disabled"
 
+                # Get the info related to the setting being changed
                 info = self.data_labels.get(key, None)
                 if info.do_not_save:
                     return
 
+                # Restrict component arguments
                 comp_args = info.component_args if info else None
                 if isinstance(comp_args, dict) and comp_args.get('visible', True) is False:
-                    raise RuntimeError(f"not possible to set {key} because it is restricted")
+                    raise RuntimeError(f"not possible to set '{key}' because it is restricted")
+
+                # Check that this section isn't frozen
+                if cmd_opts.freeze_settings_in_sections is not None:
+                    frozen_sections = list(map(str.strip, cmd_opts.freeze_settings_in_sections.split(','))) # Trim whitespace from section names
+                    section_key = info.section[0]
+                    section_name = info.section[1]
+                    assert section_key not in frozen_sections, f"not possible to set '{key}' because settings in section '{section_name}' ({section_key}) are frozen with --freeze-settings-in-sections"
+
+                # Check that this section of the settings isn't frozen
+                if cmd_opts.freeze_specific_settings is not None:
+                    frozen_keys = list(map(str.strip, cmd_opts.freeze_specific_settings.split(','))) # Trim whitespace from setting keys
+                    assert key not in frozen_keys, f"not possible to set '{key}' because this setting is frozen with --freeze-specific-settings"
 
+                # Check shorthand option which disables editing options in "saving-paths"
                 if cmd_opts.hide_ui_dir_config and key in self.restricted_opts:
-                    raise RuntimeError(f"not possible to set {key} because it is restricted")
+                    raise RuntimeError(f"not possible to set '{key}' because it is restricted with --hide_ui_dir_config")
 
                 self.data[key] = value
                 return
@@ -176,9 +195,15 @@ def same_type(self, x, y):
         return type_x == type_y
 
     def load(self, filename):
-        with open(filename, "r", encoding="utf8") as file:
-            self.data = json.load(file)
-
+        try:
+            with open(filename, "r", encoding="utf8") as file:
+                self.data = json.load(file)
+        except FileNotFoundError:
+            self.data = {}
+        except Exception:
+            errors.report(f'\nCould not load settings\nThe config file "{filename}" is likely corrupted\nIt has been moved to the "tmp/config.json"\nReverting config to default\n\n''', exc_info=True)
+            os.replace(filename, os.path.join(script_path, "tmp", "config.json"))
+            self.data = {}
         # 1.6.0 VAE defaults
         if self.data.get('sd_vae_as_default') is not None and self.data.get('sd_vae_overrides_per_model_preferences') is None:
             self.data['sd_vae_overrides_per_model_preferences'] = not self.data.get('sd_vae_as_default')
diff --git a/modules/paths.py b/modules/paths.py
index 187b949612a..030646519c3 100644
--- a/modules/paths.py
+++ b/modules/paths.py
@@ -38,7 +38,6 @@ class Dummy:
 path_dirs = [
     (sd_path, 'ldm', 'Stable Diffusion', []),
     (os.path.join(sd_path, '../generative-models'), 'sgm', 'Stable Diffusion XL', ["sgm"]),
-    (os.path.join(sd_path, '../CodeFormer'), 'inference_codeformer.py', 'CodeFormer', []),
     (os.path.join(sd_path, '../BLIP'), 'models/blip.py', 'BLIP', []),
     (os.path.join(sd_path, '../k-diffusion'), 'k_diffusion/sampling.py', 'k_diffusion', ["atstart"]),
 ]
diff --git a/modules/paths_internal.py b/modules/paths_internal.py
index 89131a54fa1..6058b0cdef6 100644
--- a/modules/paths_internal.py
+++ b/modules/paths_internal.py
@@ -4,6 +4,10 @@
 import os
 import sys
 import shlex
+from pathlib import Path
+
+
+normalized_filepath = lambda filepath: str(Path(filepath).absolute())
 
 commandline_args = os.environ.get('COMMANDLINE_ARGS', "")
 sys.argv += shlex.split(commandline_args)
@@ -28,5 +32,6 @@
 extensions_dir = os.path.join(data_path, "extensions")
 extensions_builtin_dir = os.path.join(script_path, "extensions-builtin")
 config_states_dir = os.path.join(script_path, "config_states")
+default_output_dir = os.path.join(data_path, "output")
 
 roboto_ttf_file = os.path.join(modules_path, 'Roboto-Regular.ttf')
diff --git a/modules/postprocessing.py b/modules/postprocessing.py
index 0c59fad480b..754cc9e3ab2 100644
--- a/modules/postprocessing.py
+++ b/modules/postprocessing.py
@@ -2,7 +2,7 @@
 
 from PIL import Image
 
-from modules import shared, images, devices, scripts, scripts_postprocessing, ui_common, generation_parameters_copypaste
+from modules import shared, images, devices, scripts, scripts_postprocessing, ui_common, infotext_utils
 from modules.shared import opts
 
 
@@ -17,10 +17,10 @@ def get_images(extras_mode, image, image_folder, input_dir):
         if extras_mode == 1:
             for img in image_folder:
                 if isinstance(img, Image.Image):
-                    image = img
+                    image = images.fix_image(img)
                     fn = ''
                 else:
-                    image = Image.open(os.path.abspath(img.name))
+                    image = images.read(os.path.abspath(img.name))
                     fn = os.path.splitext(img.orig_name)[0]
                 yield image, fn
         elif extras_mode == 2:
@@ -56,14 +56,12 @@ def get_images(extras_mode, image, image_folder, input_dir):
 
         if isinstance(image_placeholder, str):
             try:
-                image_data = Image.open(image_placeholder)
+                image_data = images.read(image_placeholder)
             except Exception:
                 continue
         else:
             image_data = image_placeholder
 
-        shared.state.assign_current_image(image_data)
-
         parameters, existing_pnginfo = images.read_info_from_image(image_data)
         if parameters:
             existing_pnginfo["parameters"] = parameters
@@ -86,22 +84,25 @@ def get_images(extras_mode, image, image_folder, input_dir):
                 basename = ''
                 forced_filename = None
 
-            infotext = ", ".join([k if k == v else f'{k}: {generation_parameters_copypaste.quote(v)}' for k, v in pp.info.items() if v is not None])
+            infotext = ", ".join([k if k == v else f'{k}: {infotext_utils.quote(v)}' for k, v in pp.info.items() if v is not None])
 
             if opts.enable_pnginfo:
                 pp.image.info = existing_pnginfo
                 pp.image.info["postprocessing"] = infotext
 
+            shared.state.assign_current_image(pp.image)
+
             if save_output:
                 fullfn, _ = images.save_image(pp.image, path=outpath, basename=basename, extension=opts.samples_format, info=infotext, short_filename=True, no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=forced_filename, suffix=suffix)
 
                 if pp.caption:
                     caption_filename = os.path.splitext(fullfn)[0] + ".txt"
-                    if os.path.isfile(caption_filename):
+                    existing_caption = ""
+                    try:
                         with open(caption_filename, encoding="utf8") as file:
                             existing_caption = file.read().strip()
-                    else:
-                        existing_caption = ""
+                    except FileNotFoundError:
+                        pass
 
                     action = shared.opts.postprocessing_existing_caption_action
                     if action == 'Prepend' and existing_caption:
diff --git a/modules/processing.py b/modules/processing.py
index 6f01c95f5b6..93493f80e66 100644
--- a/modules/processing.py
+++ b/modules/processing.py
@@ -16,7 +16,7 @@
 from typing import Any
 
 import modules.sd_hijack
-from modules import devices, prompt_parser, masking, sd_samplers, lowvram, generation_parameters_copypaste, extra_networks, sd_vae_approx, scripts, sd_samplers_common, sd_unet, errors, rng
+from modules import devices, prompt_parser, masking, sd_samplers, lowvram, infotext_utils, extra_networks, sd_vae_approx, scripts, sd_samplers_common, sd_unet, errors, rng
 from modules.rng import slerp # noqa: F401
 from modules.sd_hijack import model_hijack
 from modules.sd_samplers_common import images_tensor_to_samples, decode_first_stage, approximation_indexes
@@ -62,28 +62,37 @@ def apply_color_correction(correction, original_image):
     return image.convert('RGB')
 
 
-def apply_overlay(image, paste_loc, index, overlays):
-    if overlays is None or index >= len(overlays):
-        return image
+def uncrop(image, dest_size, paste_loc):
+    x, y, w, h = paste_loc
+    base_image = Image.new('RGBA', dest_size)
+    image = images.resize_image(1, image, w, h)
+    base_image.paste(image, (x, y))
+    image = base_image
 
-    overlay = overlays[index]
+    return image
+
+
+def apply_overlay(image, paste_loc, overlay):
+    if overlay is None:
+        return image, image.copy()
 
     if paste_loc is not None:
-        x, y, w, h = paste_loc
-        base_image = Image.new('RGBA', (overlay.width, overlay.height))
-        image = images.resize_image(1, image, w, h)
-        base_image.paste(image, (x, y))
-        image = base_image
+        image = uncrop(image, (overlay.width, overlay.height), paste_loc)
+
+    original_denoised_image = image.copy()
 
     image = image.convert('RGBA')
     image.alpha_composite(overlay)
     image = image.convert('RGB')
 
-    return image
+    return image, original_denoised_image
 
-def create_binary_mask(image):
+def create_binary_mask(image, round=True):
     if image.mode == 'RGBA' and image.getextrema()[-1] != (255, 255):
-        image = image.split()[-1].convert("L").point(lambda x: 255 if x > 128 else 0)
+        if round:
+            image = image.split()[-1].convert("L").point(lambda x: 255 if x > 128 else 0)
+        else:
+            image = image.split()[-1].convert("L")
     else:
         image = image.convert('L')
     return image
@@ -106,6 +115,21 @@ def txt2img_image_conditioning(sd_model, x, width, height):
         return x.new_zeros(x.shape[0], 2*sd_model.noise_augmentor.time_embed.dim, dtype=x.dtype, device=x.device)
 
     else:
+        sd = sd_model.model.state_dict()
+        diffusion_model_input = sd.get('diffusion_model.input_blocks.0.0.weight', None)
+        if diffusion_model_input is not None:
+            if diffusion_model_input.shape[1] == 9:
+                # The "masked-image" in this case will just be all 0.5 since the entire image is masked.
+                image_conditioning = torch.ones(x.shape[0], 3, height, width, device=x.device) * 0.5
+                image_conditioning = images_tensor_to_samples(image_conditioning,
+                                                              approximation_indexes.get(opts.sd_vae_encode_method))
+
+                # Add the fake full 1s mask to the first dimension.
+                image_conditioning = torch.nn.functional.pad(image_conditioning, (0, 0, 0, 0, 1, 0), value=1.0)
+                image_conditioning = image_conditioning.to(x.dtype)
+
+                return image_conditioning
+
         # Dummy zero conditioning if we're not using inpainting or unclip models.
         # Still takes up a bit of memory, but no encoder call.
         # Pretty sure we can just make this a 1x1 image since its not going to be used besides its batch size.
@@ -157,6 +181,7 @@ class StableDiffusionProcessing:
     token_merging_ratio = 0
     token_merging_ratio_hr = 0
     disable_extra_networks: bool = False
+    firstpass_image: Image = None
 
     scripts_value: scripts.ScriptRunner = field(default=None, init=False)
     script_args_value: list = field(default=None, init=False)
@@ -308,7 +333,7 @@ def unclip_image_conditioning(self, source_image):
             c_adm = torch.cat((c_adm, noise_level_emb), 1)
         return c_adm
 
-    def inpainting_image_conditioning(self, source_image, latent_image, image_mask=None):
+    def inpainting_image_conditioning(self, source_image, latent_image, image_mask=None, round_image_mask=True):
         self.is_using_inpainting_conditioning = True
 
         # Handle the different mask inputs
@@ -320,8 +345,10 @@ def inpainting_image_conditioning(self, source_image, latent_image, image_mask=N
                 conditioning_mask = conditioning_mask.astype(np.float32) / 255.0
                 conditioning_mask = torch.from_numpy(conditioning_mask[None, None])
 
-                # Inpainting model uses a discretized mask as input, so we round to either 1.0 or 0.0
-                conditioning_mask = torch.round(conditioning_mask)
+                if round_image_mask:
+                    # Caller is requesting a discretized mask as input, so we round to either 1.0 or 0.0
+                    conditioning_mask = torch.round(conditioning_mask)
+
         else:
             conditioning_mask = source_image.new_ones(1, 1, *source_image.shape[-2:])
 
@@ -345,7 +372,7 @@ def inpainting_image_conditioning(self, source_image, latent_image, image_mask=N
 
         return image_conditioning
 
-    def img2img_image_conditioning(self, source_image, latent_image, image_mask=None):
+    def img2img_image_conditioning(self, source_image, latent_image, image_mask=None, round_image_mask=True):
         source_image = devices.cond_cast_float(source_image)
 
         # HACK: Using introspection as the Depth2Image model doesn't appear to uniquely
@@ -357,11 +384,17 @@ def img2img_image_conditioning(self, source_image, latent_image, image_mask=None
             return self.edit_image_conditioning(source_image)
 
         if self.sampler.conditioning_key in {'hybrid', 'concat'}:
-            return self.inpainting_image_conditioning(source_image, latent_image, image_mask=image_mask)
+            return self.inpainting_image_conditioning(source_image, latent_image, image_mask=image_mask, round_image_mask=round_image_mask)
 
         if self.sampler.conditioning_key == "crossattn-adm":
             return self.unclip_image_conditioning(source_image)
 
+        sd = self.sampler.model_wrap.inner_model.model.state_dict()
+        diffusion_model_input = sd.get('diffusion_model.input_blocks.0.0.weight', None)
+        if diffusion_model_input is not None:
+            if diffusion_model_input.shape[1] == 9:
+                return self.inpainting_image_conditioning(source_image, latent_image, image_mask=image_mask)
+
         # Dummy zero conditioning if we're not using inpainting or depth model.
         return latent_image.new_zeros(latent_image.shape[0], 5, 1, 1)
 
@@ -422,6 +455,9 @@ def cached_params(self, required_prompts, steps, extra_network_data, hires_steps
             opts.sdxl_crop_top,
             self.width,
             self.height,
+            opts.fp8_storage,
+            opts.cache_fp16_weight,
+            opts.emphasis,
         )
 
     def get_conds_with_caching(self, function, required_prompts, steps, caches, extra_network_data, hires_steps=None):
@@ -596,20 +632,33 @@ def decode_latent_batch(model, batch, target_device=None, check_for_nans=False):
         sample = decode_first_stage(model, batch[i:i + 1])[0]
 
         if check_for_nans:
+
             try:
                 devices.test_for_nans(sample, "vae")
             except devices.NansException as e:
-                if devices.dtype_vae == torch.float32 or not shared.opts.auto_vae_precision:
+                if shared.opts.auto_vae_precision_bfloat16:
+                    autofix_dtype = torch.bfloat16
+                    autofix_dtype_text = "bfloat16"
+                    autofix_dtype_setting = "Automatically convert VAE to bfloat16"
+                    autofix_dtype_comment = ""
+                elif shared.opts.auto_vae_precision:
+                    autofix_dtype = torch.float32
+                    autofix_dtype_text = "32-bit float"
+                    autofix_dtype_setting = "Automatically revert VAE to 32-bit floats"
+                    autofix_dtype_comment = "\nTo always start with 32-bit VAE, use --no-half-vae commandline flag."
+                else:
+                    raise e
+
+                if devices.dtype_vae == autofix_dtype:
                     raise e
 
                 errors.print_error_explanation(
                     "A tensor with all NaNs was produced in VAE.\n"
-                    "Web UI will now convert VAE into 32-bit float and retry.\n"
-                    "To disable this behavior, disable the 'Automatically revert VAE to 32-bit floats' setting.\n"
-                    "To always start with 32-bit VAE, use --no-half-vae commandline flag."
+                    f"Web UI will now convert VAE into {autofix_dtype_text} and retry.\n"
+                    f"To disable this behavior, disable the '{autofix_dtype_setting}' setting.{autofix_dtype_comment}"
                 )
 
-                devices.dtype_vae = torch.float32
+                devices.dtype_vae = autofix_dtype
                 model.first_stage_model.to(devices.dtype_vae)
                 batch = batch.to(devices.dtype_vae)
 
@@ -679,12 +728,14 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments=None, iter
         "Size": f"{p.width}x{p.height}",
         "Model hash": p.sd_model_hash if opts.add_model_hash_to_info else None,
         "Model": p.sd_model_name if opts.add_model_name_to_info else None,
+        "FP8 weight": opts.fp8_storage if devices.fp8 else None,
+        "Cache FP16 weight for LoRA": opts.cache_fp16_weight if devices.fp8 else None,
         "VAE hash": p.sd_vae_hash if opts.add_vae_hash_to_info else None,
         "VAE": p.sd_vae_name if opts.add_vae_name_to_info else None,
         "Variation seed": (None if p.subseed_strength == 0 else (p.all_subseeds[0] if use_main_prompt else all_subseeds[index])),
         "Variation seed strength": (None if p.subseed_strength == 0 else p.subseed_strength),
         "Seed resize from": (None if p.seed_resize_from_w <= 0 or p.seed_resize_from_h <= 0 else f"{p.seed_resize_from_w}x{p.seed_resize_from_h}"),
-        "Denoising strength": getattr(p, 'denoising_strength', None),
+        "Denoising strength": p.extra_generation_params.get("Denoising strength"),
         "Conditional mask weight": getattr(p, "inpainting_mask_weight", shared.opts.inpainting_mask_weight) if p.is_using_inpainting_conditioning else None,
         "Clip skip": None if clip_skip <= 1 else clip_skip,
         "ENSD": opts.eta_noise_seed_delta if uses_ensd else None,
@@ -699,7 +750,7 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments=None, iter
         "User": p.user if opts.add_user_name_to_info else None,
     }
 
-    generation_params_text = ", ".join([k if k == v else f'{k}: {generation_parameters_copypaste.quote(v)}' for k, v in generation_params.items() if v is not None])
+    generation_params_text = ", ".join([k if k == v else f'{k}: {infotext_utils.quote(v)}' for k, v in generation_params.items() if v is not None])
 
     prompt_text = p.main_prompt if use_main_prompt else all_prompts[index]
     negative_prompt_text = f"\nNegative prompt: {p.main_negative_prompt if use_main_prompt else all_negative_prompts[index]}" if all_negative_prompts[index] else ""
@@ -818,7 +869,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
             if state.skipped:
                 state.skipped = False
 
-            if state.interrupted:
+            if state.interrupted or state.stopping_generation:
                 break
 
             sd_models.reload_model_weights()  # model can be changed for example by refiner
@@ -845,6 +896,10 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
             if p.scripts is not None:
                 p.scripts.process_batch(p, batch_number=n, prompts=p.prompts, seeds=p.seeds, subseeds=p.subseeds)
 
+            p.setup_conds()
+
+            p.extra_generation_params.update(model_hijack.extra_generation_params)
+
             # params.txt should be saved after scripts.process_batch, since the
             # infotext could be modified by that callback
             # Example: a wildcard processed by process_batch sets an extra model
@@ -854,19 +909,22 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
                     processed = Processed(p, [])
                     file.write(processed.infotext(p, 0))
 
-            p.setup_conds()
-
             for comment in model_hijack.comments:
                 p.comment(comment)
 
-            p.extra_generation_params.update(model_hijack.extra_generation_params)
-
             if p.n_iter > 1:
                 shared.state.job = f"Batch {n+1} out of {p.n_iter}"
 
+            sd_models.apply_alpha_schedule_override(p.sd_model, p)
+
             with devices.without_autocast() if devices.unet_needs_upcast else devices.autocast():
                 samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts)
 
+            if p.scripts is not None:
+                ps = scripts.PostSampleArgs(samples_ddim)
+                p.scripts.post_sample(p, ps)
+                samples_ddim = ps.samples
+
             if getattr(samples_ddim, 'already_decoded', False):
                 x_samples_ddim = samples_ddim
             else:
@@ -922,13 +980,37 @@ def infotext(index=0, use_main_prompt=False):
                     pp = scripts.PostprocessImageArgs(image)
                     p.scripts.postprocess_image(p, pp)
                     image = pp.image
+
+                mask_for_overlay = getattr(p, "mask_for_overlay", None)
+
+                if not shared.opts.overlay_inpaint:
+                    overlay_image = None
+                elif getattr(p, "overlay_images", None) is not None and i < len(p.overlay_images):
+                    overlay_image = p.overlay_images[i]
+                else:
+                    overlay_image = None
+
+                if p.scripts is not None:
+                    ppmo = scripts.PostProcessMaskOverlayArgs(i, mask_for_overlay, overlay_image)
+                    p.scripts.postprocess_maskoverlay(p, ppmo)
+                    mask_for_overlay, overlay_image = ppmo.mask_for_overlay, ppmo.overlay_image
+
                 if p.color_corrections is not None and i < len(p.color_corrections):
                     if save_samples and opts.save_images_before_color_correction:
-                        image_without_cc = apply_overlay(image, p.paste_to, i, p.overlay_images)
+                        image_without_cc, _ = apply_overlay(image, p.paste_to, overlay_image)
                         images.save_image(image_without_cc, p.outpath_samples, "", p.seeds[i], p.prompts[i], opts.samples_format, info=infotext(i), p=p, suffix="-before-color-correction")
                     image = apply_color_correction(p.color_corrections[i], image)
 
-                image = apply_overlay(image, p.paste_to, i, p.overlay_images)
+                # If the intention is to show the output from the model
+                # that is being composited over the original image,
+                # we need to keep the original image around
+                # and use it in the composite step.
+                image, original_denoised_image = apply_overlay(image, p.paste_to, overlay_image)
+
+                if p.scripts is not None:
+                    pp = scripts.PostprocessImageArgs(image)
+                    p.scripts.postprocess_image_after_composite(p, pp)
+                    image = pp.image
 
                 if save_samples:
                     images.save_image(image, p.outpath_samples, "", p.seeds[i], p.prompts[i], opts.samples_format, info=infotext(i), p=p)
@@ -938,16 +1020,17 @@ def infotext(index=0, use_main_prompt=False):
                 if opts.enable_pnginfo:
                     image.info["parameters"] = text
                 output_images.append(image)
-                if hasattr(p, 'mask_for_overlay') and p.mask_for_overlay:
+
+                if mask_for_overlay is not None:
                     if opts.return_mask or opts.save_mask:
-                        image_mask = p.mask_for_overlay.convert('RGB')
+                        image_mask = mask_for_overlay.convert('RGB')
                         if save_samples and opts.save_mask:
                             images.save_image(image_mask, p.outpath_samples, "", p.seeds[i], p.prompts[i], opts.samples_format, info=infotext(i), p=p, suffix="-mask")
                         if opts.return_mask:
                             output_images.append(image_mask)
 
                     if opts.return_mask_composite or opts.save_mask_composite:
-                        image_mask_composite = Image.composite(image.convert('RGBA').convert('RGBa'), Image.new('RGBa', image.size), images.resize_image(2, p.mask_for_overlay, image.width, image.height).convert('L')).convert('RGBA')
+                        image_mask_composite = Image.composite(original_denoised_image.convert('RGBA').convert('RGBa'), Image.new('RGBa', image.size), images.resize_image(2, mask_for_overlay, image.width, image.height).convert('L')).convert('RGBA')
                         if save_samples and opts.save_mask_composite:
                             images.save_image(image_mask_composite, p.outpath_samples, "", p.seeds[i], p.prompts[i], opts.samples_format, info=infotext(i), p=p, suffix="-mask-composite")
                         if opts.return_mask_composite:
@@ -1025,6 +1108,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
     hr_sampler_name: str = None
     hr_prompt: str = ''
     hr_negative_prompt: str = ''
+    force_task_id: str = None
 
     cached_hr_uc = [None, None]
     cached_hr_c = [None, None]
@@ -1097,7 +1181,9 @@ def calculate_target_resolution(self):
 
     def init(self, all_prompts, all_seeds, all_subseeds):
         if self.enable_hr:
-            if self.hr_checkpoint_name:
+            self.extra_generation_params["Denoising strength"] = self.denoising_strength
+
+            if self.hr_checkpoint_name and self.hr_checkpoint_name != 'Use same checkpoint':
                 self.hr_checkpoint_info = sd_models.get_closet_checkpoint_match(self.hr_checkpoint_name)
 
                 if self.hr_checkpoint_info is None:
@@ -1124,8 +1210,11 @@ def init(self, all_prompts, all_seeds, all_subseeds):
             if not state.processing_has_refined_job_count:
                 if state.job_count == -1:
                     state.job_count = self.n_iter
-
-                shared.total_tqdm.updateTotal((self.steps + (self.hr_second_pass_steps or self.steps)) * state.job_count)
+                if getattr(self, 'txt2img_upscale', False):
+                    total_steps = (self.hr_second_pass_steps or self.steps) * state.job_count
+                else:
+                    total_steps = (self.steps + (self.hr_second_pass_steps or self.steps)) * state.job_count
+                shared.total_tqdm.updateTotal(total_steps)
                 state.job_count = state.job_count * 2
                 state.processing_has_refined_job_count = True
 
@@ -1138,18 +1227,45 @@ def init(self, all_prompts, all_seeds, all_subseeds):
     def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength, prompts):
         self.sampler = sd_samplers.create_sampler(self.sampler_name, self.sd_model)
 
-        x = self.rng.next()
-        samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.txt2img_image_conditioning(x))
-        del x
+        if self.firstpass_image is not None and self.enable_hr:
+            # here we don't need to generate image, we just take self.firstpass_image and prepare it for hires fix
 
-        if not self.enable_hr:
-            return samples
-        devices.torch_gc()
+            if self.latent_scale_mode is None:
+                image = np.array(self.firstpass_image).astype(np.float32) / 255.0 * 2.0 - 1.0
+                image = np.moveaxis(image, 2, 0)
+
+                samples = None
+                decoded_samples = torch.asarray(np.expand_dims(image, 0))
+
+            else:
+                image = np.array(self.firstpass_image).astype(np.float32) / 255.0
+                image = np.moveaxis(image, 2, 0)
+                image = torch.from_numpy(np.expand_dims(image, axis=0))
+                image = image.to(shared.device, dtype=devices.dtype_vae)
+
+                if opts.sd_vae_encode_method != 'Full':
+                    self.extra_generation_params['VAE Encoder'] = opts.sd_vae_encode_method
+
+                samples = images_tensor_to_samples(image, approximation_indexes.get(opts.sd_vae_encode_method), self.sd_model)
+                decoded_samples = None
+                devices.torch_gc()
 
-        if self.latent_scale_mode is None:
-            decoded_samples = torch.stack(decode_latent_batch(self.sd_model, samples, target_device=devices.cpu, check_for_nans=True)).to(dtype=torch.float32)
         else:
-            decoded_samples = None
+            # here we generate an image normally
+
+            x = self.rng.next()
+            samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.txt2img_image_conditioning(x))
+            del x
+
+            if not self.enable_hr:
+                return samples
+
+            devices.torch_gc()
+
+            if self.latent_scale_mode is None:
+                decoded_samples = torch.stack(decode_latent_batch(self.sd_model, samples, target_device=devices.cpu, check_for_nans=True)).to(dtype=torch.float32)
+            else:
+                decoded_samples = None
 
         with sd_models.SkipWritingToConfig():
             sd_models.reload_model_weights(info=self.hr_checkpoint_info)
@@ -1351,12 +1467,14 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing):
     mask_blur_x: int = 4
     mask_blur_y: int = 4
     mask_blur: int = None
+    mask_round: bool = True
     inpainting_fill: int = 0
     inpaint_full_res: bool = True
     inpaint_full_res_padding: int = 0
     inpainting_mask_invert: int = 0
     initial_noise_multiplier: float = None
     latent_mask: Image = None
+    force_task_id: str = None
 
     image_mask: Any = field(default=None, init=False)
 
@@ -1386,6 +1504,8 @@ def mask_blur(self, value):
             self.mask_blur_y = value
 
     def init(self, all_prompts, all_seeds, all_subseeds):
+        self.extra_generation_params["Denoising strength"] = self.denoising_strength
+
         self.image_cfg_scale: float = self.image_cfg_scale if shared.sd_model.cond_stage_key == "edit" else None
 
         self.sampler = sd_samplers.create_sampler(self.sampler_name, self.sd_model)
@@ -1396,10 +1516,11 @@ def init(self, all_prompts, all_seeds, all_subseeds):
         if image_mask is not None:
             # image_mask is passed in as RGBA by Gradio to support alpha masks,
             # but we still want to support binary masks.
-            image_mask = create_binary_mask(image_mask)
+            image_mask = create_binary_mask(image_mask, round=self.mask_round)
 
             if self.inpainting_mask_invert:
                 image_mask = ImageOps.invert(image_mask)
+                self.extra_generation_params["Mask mode"] = "Inpaint not masked"
 
             if self.mask_blur_x > 0:
                 np_mask = np.array(image_mask)
@@ -1413,16 +1534,22 @@ def init(self, all_prompts, all_seeds, all_subseeds):
                 np_mask = cv2.GaussianBlur(np_mask, (1, kernel_size), self.mask_blur_y)
                 image_mask = Image.fromarray(np_mask)
 
+            if self.mask_blur_x > 0 or self.mask_blur_y > 0:
+                self.extra_generation_params["Mask blur"] = self.mask_blur
+
             if self.inpaint_full_res:
                 self.mask_for_overlay = image_mask
                 mask = image_mask.convert('L')
-                crop_region = masking.get_crop_region(np.array(mask), self.inpaint_full_res_padding)
+                crop_region = masking.get_crop_region(mask, self.inpaint_full_res_padding)
                 crop_region = masking.expand_crop_region(crop_region, self.width, self.height, mask.width, mask.height)
                 x1, y1, x2, y2 = crop_region
 
                 mask = mask.crop(crop_region)
                 image_mask = images.resize_image(2, mask, self.width, self.height)
                 self.paste_to = (x1, y1, x2-x1, y2-y1)
+
+                self.extra_generation_params["Inpaint area"] = "Only masked"
+                self.extra_generation_params["Masked area padding"] = self.inpaint_full_res_padding
             else:
                 image_mask = images.resize_image(self.resize_mode, image_mask, self.width, self.height)
                 np_mask = np.array(image_mask)
@@ -1442,7 +1569,7 @@ def init(self, all_prompts, all_seeds, all_subseeds):
             # Save init image
             if opts.save_init_img:
                 self.init_img_hash = hashlib.md5(img.tobytes()).hexdigest()
-                images.save_image(img, path=opts.outdir_init_images, basename=None, forced_filename=self.init_img_hash, save_to_dirs=False)
+                images.save_image(img, path=opts.outdir_init_images, basename=None, forced_filename=self.init_img_hash, save_to_dirs=False, existing_info=img.info)
 
             image = images.flatten(img, opts.img2img_background_color)
 
@@ -1464,6 +1591,9 @@ def init(self, all_prompts, all_seeds, all_subseeds):
                 if self.inpainting_fill != 1:
                     image = masking.fill(image, latent_mask)
 
+                    if self.inpainting_fill == 0:
+                        self.extra_generation_params["Masked content"] = 'fill'
+
             if add_color_corrections:
                 self.color_corrections.append(setup_color_correction(image))
 
@@ -1503,7 +1633,8 @@ def init(self, all_prompts, all_seeds, all_subseeds):
             latmask = init_mask.convert('RGB').resize((self.init_latent.shape[3], self.init_latent.shape[2]))
             latmask = np.moveaxis(np.array(latmask, dtype=np.float32), 2, 0) / 255
             latmask = latmask[0]
-            latmask = np.around(latmask)
+            if self.mask_round:
+                latmask = np.around(latmask)
             latmask = np.tile(latmask[None], (4, 1, 1))
 
             self.mask = torch.asarray(1.0 - latmask).to(shared.device).type(self.sd_model.dtype)
@@ -1512,10 +1643,13 @@ def init(self, all_prompts, all_seeds, all_subseeds):
             # this needs to be fixed to be done in sample() using actual seeds for batches
             if self.inpainting_fill == 2:
                 self.init_latent = self.init_latent * self.mask + create_random_tensors(self.init_latent.shape[1:], all_seeds[0:self.init_latent.shape[0]]) * self.nmask
+                self.extra_generation_params["Masked content"] = 'latent noise'
+
             elif self.inpainting_fill == 3:
                 self.init_latent = self.init_latent * self.mask
+                self.extra_generation_params["Masked content"] = 'latent nothing'
 
-        self.image_conditioning = self.img2img_image_conditioning(image * 2 - 1, self.init_latent, image_mask)
+        self.image_conditioning = self.img2img_image_conditioning(image * 2 - 1, self.init_latent, image_mask, self.mask_round)
 
     def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength, prompts):
         x = self.rng.next()
@@ -1527,7 +1661,14 @@ def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subs
         samples = self.sampler.sample_img2img(self, self.init_latent, x, conditioning, unconditional_conditioning, image_conditioning=self.image_conditioning)
 
         if self.mask is not None:
-            samples = samples * self.nmask + self.init_latent * self.mask
+            blended_samples = samples * self.nmask + self.init_latent * self.mask
+
+            if self.scripts is not None:
+                mba = scripts.MaskBlendArgs(samples, self.nmask, self.init_latent, self.mask, blended_samples)
+                self.scripts.on_mask_blend(self, mba)
+                blended_samples = mba.blended_latent
+
+            samples = blended_samples
 
         del x
         devices.torch_gc()
diff --git a/modules/processing_scripts/comments.py b/modules/processing_scripts/comments.py
new file mode 100644
index 00000000000..638e39f2989
--- /dev/null
+++ b/modules/processing_scripts/comments.py
@@ -0,0 +1,42 @@
+from modules import scripts, shared, script_callbacks
+import re
+
+
+def strip_comments(text):
+    text = re.sub('(^|\n)#[^\n]*(\n|$)', '\n', text)  # while line comment
+    text = re.sub('#[^\n]*(\n|$)', '\n', text)  # in the middle of the line comment
+
+    return text
+
+
+class ScriptStripComments(scripts.Script):
+    def title(self):
+        return "Comments"
+
+    def show(self, is_img2img):
+        return scripts.AlwaysVisible
+
+    def process(self, p, *args):
+        if not shared.opts.enable_prompt_comments:
+            return
+
+        p.all_prompts = [strip_comments(x) for x in p.all_prompts]
+        p.all_negative_prompts = [strip_comments(x) for x in p.all_negative_prompts]
+
+        p.main_prompt = strip_comments(p.main_prompt)
+        p.main_negative_prompt = strip_comments(p.main_negative_prompt)
+
+
+def before_token_counter(params: script_callbacks.BeforeTokenCounterParams):
+    if not shared.opts.enable_prompt_comments:
+        return
+
+    params.prompt = strip_comments(params.prompt)
+
+
+script_callbacks.on_before_token_counter(before_token_counter)
+
+
+shared.options_templates.update(shared.options_section(('sd', "Stable Diffusion", "sd"), {
+    "enable_prompt_comments": shared.OptionInfo(True, "Enable comments").info("Use # anywhere in the prompt to hide the text between # and the end of the line from the generation."),
+}))
diff --git a/modules/processing_scripts/refiner.py b/modules/processing_scripts/refiner.py
index 29ccb78f903..ba33d8a4b80 100644
--- a/modules/processing_scripts/refiner.py
+++ b/modules/processing_scripts/refiner.py
@@ -1,6 +1,7 @@
 import gradio as gr
 
 from modules import scripts, sd_models
+from modules.infotext_utils import PasteField
 from modules.ui_common import create_refresh_button
 from modules.ui_components import InputAccordion
 
@@ -31,9 +32,9 @@ def lookup_checkpoint(title):
             return None if info is None else info.title
 
         self.infotext_fields = [
-            (enable_refiner, lambda d: 'Refiner' in d),
-            (refiner_checkpoint, lambda d: lookup_checkpoint(d.get('Refiner'))),
-            (refiner_switch_at, 'Refiner switch at'),
+            PasteField(enable_refiner, lambda d: 'Refiner' in d),
+            PasteField(refiner_checkpoint, lambda d: lookup_checkpoint(d.get('Refiner')), api="refiner_checkpoint"),
+            PasteField(refiner_switch_at, 'Refiner switch at', api="refiner_switch_at"),
         ]
 
         return enable_refiner, refiner_checkpoint, refiner_switch_at
diff --git a/modules/processing_scripts/seed.py b/modules/processing_scripts/seed.py
index dc9c2da5000..7a4c0159831 100644
--- a/modules/processing_scripts/seed.py
+++ b/modules/processing_scripts/seed.py
@@ -3,8 +3,10 @@
 import gradio as gr
 
 from modules import scripts, ui, errors
+from modules.infotext_utils import PasteField
 from modules.shared import cmd_opts
 from modules.ui_components import ToolButton
+from modules import infotext_utils
 
 
 class ScriptSeed(scripts.ScriptBuiltinUI):
@@ -51,12 +53,12 @@ def ui(self, is_img2img):
         seed_checkbox.change(lambda x: gr.update(visible=x), show_progress=False, inputs=[seed_checkbox], outputs=[seed_extras])
 
         self.infotext_fields = [
-            (self.seed, "Seed"),
-            (seed_checkbox, lambda d: "Variation seed" in d or "Seed resize from-1" in d),
-            (subseed, "Variation seed"),
-            (subseed_strength, "Variation seed strength"),
-            (seed_resize_from_w, "Seed resize from-1"),
-            (seed_resize_from_h, "Seed resize from-2"),
+            PasteField(self.seed, "Seed", api="seed"),
+            PasteField(seed_checkbox, lambda d: "Variation seed" in d or "Seed resize from-1" in d),
+            PasteField(subseed, "Variation seed", api="subseed"),
+            PasteField(subseed_strength, "Variation seed strength", api="subseed_strength"),
+            PasteField(seed_resize_from_w, "Seed resize from-1", api="seed_resize_from_h"),
+            PasteField(seed_resize_from_h, "Seed resize from-2", api="seed_resize_from_w"),
         ]
 
         self.on_after_component(lambda x: connect_reuse_seed(self.seed, reuse_seed, x.component, False), elem_id=f'generation_info_{self.tabname}')
@@ -76,7 +78,6 @@ def setup(self, p, seed, seed_checkbox, subseed, subseed_strength, seed_resize_f
             p.seed_resize_from_h = seed_resize_from_h
 
 
-
 def connect_reuse_seed(seed: gr.Number, reuse_seed: gr.Button, generation_info: gr.Textbox, is_subseed):
     """ Connects a 'reuse (sub)seed' button's click event so that it copies last used
         (sub)seed value from generation info the to the seed field. If copying subseed and subseed strength
@@ -84,21 +85,14 @@ def connect_reuse_seed(seed: gr.Number, reuse_seed: gr.Button, generation_info:
 
     def copy_seed(gen_info_string: str, index):
         res = -1
-
         try:
             gen_info = json.loads(gen_info_string)
-            index -= gen_info.get('index_of_first_image', 0)
-
-            if is_subseed and gen_info.get('subseed_strength', 0) > 0:
-                all_subseeds = gen_info.get('all_subseeds', [-1])
-                res = all_subseeds[index if 0 <= index < len(all_subseeds) else 0]
-            else:
-                all_seeds = gen_info.get('all_seeds', [-1])
-                res = all_seeds[index if 0 <= index < len(all_seeds) else 0]
-
-        except json.decoder.JSONDecodeError:
+            infotext = gen_info.get('infotexts')[index]
+            gen_parameters = infotext_utils.parse_generation_parameters(infotext, [])
+            res = int(gen_parameters.get('Variation seed' if is_subseed else 'Seed', -1))
+        except Exception:
             if gen_info_string:
-                errors.report(f"Error parsing JSON generation info: {gen_info_string}")
+                errors.report(f"Error retrieving seed from generation info: {gen_info_string}", exc_info=True)
 
         return [res, gr.update()]
 
diff --git a/modules/progress.py b/modules/progress.py
index 69921de7281..85255e821f5 100644
--- a/modules/progress.py
+++ b/modules/progress.py
@@ -8,10 +8,13 @@
 from modules.shared import opts
 
 import modules.shared as shared
-
+from collections import OrderedDict
+import string
+import random
+from typing import List
 
 current_task = None
-pending_tasks = {}
+pending_tasks = OrderedDict()
 finished_tasks = []
 recorded_results = []
 recorded_results_limit = 2
@@ -34,6 +37,11 @@ def finish_task(id_task):
     if len(finished_tasks) > 16:
         finished_tasks.pop(0)
 
+def create_task_id(task_type):
+    N = 7
+    res = ''.join(random.choices(string.ascii_uppercase +
+    string.digits, k=N))
+    return f"task({task_type}-{res})"
 
 def record_results(id_task, res):
     recorded_results.append((id_task, res))
@@ -44,6 +52,9 @@ def record_results(id_task, res):
 def add_task_to_queue(id_job):
     pending_tasks[id_job] = time.time()
 
+class PendingTasksResponse(BaseModel):
+    size: int = Field(title="Pending task size")
+    tasks: List[str] = Field(title="Pending task ids")
 
 class ProgressRequest(BaseModel):
     id_task: str = Field(default=None, title="Task ID", description="id of the task to get progress for")
@@ -63,9 +74,16 @@ class ProgressResponse(BaseModel):
 
 
 def setup_progress_api(app):
+    app.add_api_route("/internal/pending-tasks", get_pending_tasks, methods=["GET"])
     return app.add_api_route("/internal/progress", progressapi, methods=["POST"], response_model=ProgressResponse)
 
 
+def get_pending_tasks():
+    pending_tasks_ids = list(pending_tasks)
+    pending_len = len(pending_tasks_ids)
+    return PendingTasksResponse(size=pending_len, tasks=pending_tasks_ids)
+
+
 def progressapi(req: ProgressRequest):
     active = req.id_task == current_task
     queued = req.id_task in pending_tasks
diff --git a/modules/realesrgan_model.py b/modules/realesrgan_model.py
index 02841c30289..ff9d8ac0d69 100644
--- a/modules/realesrgan_model.py
+++ b/modules/realesrgan_model.py
@@ -1,12 +1,9 @@
 import os
 
-import numpy as np
-from PIL import Image
-from realesrgan import RealESRGANer
-
-from modules.upscaler import Upscaler, UpscalerData
-from modules.shared import cmd_opts, opts
 from modules import modelloader, errors
+from modules.shared import cmd_opts, opts
+from modules.upscaler import Upscaler, UpscalerData
+from modules.upscaler_utils import upscale_with_model
 
 
 class UpscalerRealESRGAN(Upscaler):
@@ -14,29 +11,20 @@ def __init__(self, path):
         self.name = "RealESRGAN"
         self.user_path = path
         super().__init__()
-        try:
-            from basicsr.archs.rrdbnet_arch import RRDBNet  # noqa: F401
-            from realesrgan import RealESRGANer  # noqa: F401
-            from realesrgan.archs.srvgg_arch import SRVGGNetCompact  # noqa: F401
-            self.enable = True
-            self.scalers = []
-            scalers = self.load_models(path)
+        self.enable = True
+        self.scalers = []
+        scalers = get_realesrgan_models(self)
 
-            local_model_paths = self.find_models(ext_filter=[".pth"])
-            for scaler in scalers:
-                if scaler.local_data_path.startswith("http"):
-                    filename = modelloader.friendly_name(scaler.local_data_path)
-                    local_model_candidates = [local_model for local_model in local_model_paths if local_model.endswith(f"{filename}.pth")]
-                    if local_model_candidates:
-                        scaler.local_data_path = local_model_candidates[0]
+        local_model_paths = self.find_models(ext_filter=[".pth"])
+        for scaler in scalers:
+            if scaler.local_data_path.startswith("http"):
+                filename = modelloader.friendly_name(scaler.local_data_path)
+                local_model_candidates = [local_model for local_model in local_model_paths if local_model.endswith(f"{filename}.pth")]
+                if local_model_candidates:
+                    scaler.local_data_path = local_model_candidates[0]
 
-                if scaler.name in opts.realesrgan_enabled_models:
-                    self.scalers.append(scaler)
-
-        except Exception:
-            errors.report("Error importing Real-ESRGAN", exc_info=True)
-            self.enable = False
-            self.scalers = []
+            if scaler.name in opts.realesrgan_enabled_models:
+                self.scalers.append(scaler)
 
     def do_upscale(self, img, path):
         if not self.enable:
@@ -48,20 +36,19 @@ def do_upscale(self, img, path):
             errors.report(f"Unable to load RealESRGAN model {path}", exc_info=True)
             return img
 
-        upsampler = RealESRGANer(
-            scale=info.scale,
-            model_path=info.local_data_path,
-            model=info.model(),
-            half=not cmd_opts.no_half and not cmd_opts.upcast_sampling,
-            tile=opts.ESRGAN_tile,
-            tile_pad=opts.ESRGAN_tile_overlap,
+        model_descriptor = modelloader.load_spandrel_model(
+            info.local_data_path,
             device=self.device,
+            prefer_half=(not cmd_opts.no_half and not cmd_opts.upcast_sampling),
+            expected_architecture="ESRGAN",  # "RealESRGAN" isn't a specific thing for Spandrel
+        )
+        return upscale_with_model(
+            model_descriptor,
+            img,
+            tile_size=opts.ESRGAN_tile,
+            tile_overlap=opts.ESRGAN_tile_overlap,
+            # TODO: `outscale`?
         )
-
-        upsampled = upsampler.enhance(np.array(img), outscale=info.scale)[0]
-
-        image = Image.fromarray(upsampled)
-        return image
 
     def load_model(self, path):
         for scaler in self.scalers:
@@ -76,58 +63,43 @@ def load_model(self, path):
                 return scaler
         raise ValueError(f"Unable to find model info: {path}")
 
-    def load_models(self, _):
-        return get_realesrgan_models(self)
-
 
-def get_realesrgan_models(scaler):
-    try:
-        from basicsr.archs.rrdbnet_arch import RRDBNet
-        from realesrgan.archs.srvgg_arch import SRVGGNetCompact
-        models = [
-            UpscalerData(
-                name="R-ESRGAN General 4xV3",
-                path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-x4v3.pth",
-                scale=4,
-                upscaler=scaler,
-                model=lambda: SRVGGNetCompact(num_in_ch=3, num_out_ch=3, num_feat=64, num_conv=32, upscale=4, act_type='prelu')
-            ),
-            UpscalerData(
-                name="R-ESRGAN General WDN 4xV3",
-                path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-wdn-x4v3.pth",
-                scale=4,
-                upscaler=scaler,
-                model=lambda: SRVGGNetCompact(num_in_ch=3, num_out_ch=3, num_feat=64, num_conv=32, upscale=4, act_type='prelu')
-            ),
-            UpscalerData(
-                name="R-ESRGAN AnimeVideo",
-                path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-animevideov3.pth",
-                scale=4,
-                upscaler=scaler,
-                model=lambda: SRVGGNetCompact(num_in_ch=3, num_out_ch=3, num_feat=64, num_conv=16, upscale=4, act_type='prelu')
-            ),
-            UpscalerData(
-                name="R-ESRGAN 4x+",
-                path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth",
-                scale=4,
-                upscaler=scaler,
-                model=lambda: RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=4)
-            ),
-            UpscalerData(
-                name="R-ESRGAN 4x+ Anime6B",
-                path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth",
-                scale=4,
-                upscaler=scaler,
-                model=lambda: RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=6, num_grow_ch=32, scale=4)
-            ),
-            UpscalerData(
-                name="R-ESRGAN 2x+",
-                path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.1/RealESRGAN_x2plus.pth",
-                scale=2,
-                upscaler=scaler,
-                model=lambda: RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=2)
-            ),
-        ]
-        return models
-    except Exception:
-        errors.report("Error making Real-ESRGAN models list", exc_info=True)
+def get_realesrgan_models(scaler: UpscalerRealESRGAN):
+    return [
+        UpscalerData(
+            name="R-ESRGAN General 4xV3",
+            path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-x4v3.pth",
+            scale=4,
+            upscaler=scaler,
+        ),
+        UpscalerData(
+            name="R-ESRGAN General WDN 4xV3",
+            path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-wdn-x4v3.pth",
+            scale=4,
+            upscaler=scaler,
+        ),
+        UpscalerData(
+            name="R-ESRGAN AnimeVideo",
+            path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-animevideov3.pth",
+            scale=4,
+            upscaler=scaler,
+        ),
+        UpscalerData(
+            name="R-ESRGAN 4x+",
+            path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth",
+            scale=4,
+            upscaler=scaler,
+        ),
+        UpscalerData(
+            name="R-ESRGAN 4x+ Anime6B",
+            path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth",
+            scale=4,
+            upscaler=scaler,
+        ),
+        UpscalerData(
+            name="R-ESRGAN 2x+",
+            path="https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.1/RealESRGAN_x2plus.pth",
+            scale=2,
+            upscaler=scaler,
+        ),
+    ]
diff --git a/modules/rng.py b/modules/rng.py
index 8934d39bf9a..5390d1bb73e 100644
--- a/modules/rng.py
+++ b/modules/rng.py
@@ -34,7 +34,7 @@ def randn_local(seed, shape):
 
 
 def randn_like(x):
-    """Generate a tensor with random numbers from a normal distribution using the previously initialized genrator.
+    """Generate a tensor with random numbers from a normal distribution using the previously initialized generator.
 
     Use either randn() or manual_seed() to initialize the generator."""
 
@@ -48,7 +48,7 @@ def randn_like(x):
 
 
 def randn_without_seed(shape, generator=None):
-    """Generate a tensor with random numbers from a normal distribution using the previously initialized genrator.
+    """Generate a tensor with random numbers from a normal distribution using the previously initialized generator.
 
     Use either randn() or manual_seed() to initialize the generator."""
 
diff --git a/modules/script_callbacks.py b/modules/script_callbacks.py
index 9ed7ad21d1b..08bc525641d 100644
--- a/modules/script_callbacks.py
+++ b/modules/script_callbacks.py
@@ -1,3 +1,4 @@
+import dataclasses
 import inspect
 import os
 from collections import namedtuple
@@ -41,7 +42,7 @@ def __init__(self, noise, x, xi):
 
 
 class CFGDenoiserParams:
-    def __init__(self, x, image_cond, sigma, sampling_step, total_sampling_steps, text_cond, text_uncond):
+    def __init__(self, x, image_cond, sigma, sampling_step, total_sampling_steps, text_cond, text_uncond, denoiser=None):
         self.x = x
         """Latent image representation in the process of being denoised"""
 
@@ -63,6 +64,9 @@ def __init__(self, x, image_cond, sigma, sampling_step, total_sampling_steps, te
         self.text_uncond = text_uncond
         """ Encoder hidden states of text conditioning from negative prompt"""
 
+        self.denoiser = denoiser
+        """Current CFGDenoiser object with processing parameters"""
+
 
 class CFGDenoisedParams:
     def __init__(self, x, sampling_step, total_sampling_steps, inner_model):
@@ -103,6 +107,15 @@ def __init__(self, imgs, cols, rows):
         self.rows = rows
 
 
+@dataclasses.dataclass
+class BeforeTokenCounterParams:
+    prompt: str
+    steps: int
+    styles: list
+
+    is_positive: bool = True
+
+
 ScriptCallback = namedtuple("ScriptCallback", ["script", "callback"])
 callback_map = dict(
     callbacks_app_started=[],
@@ -125,6 +138,7 @@ def __init__(self, imgs, cols, rows):
     callbacks_on_reload=[],
     callbacks_list_optimizers=[],
     callbacks_list_unets=[],
+    callbacks_before_token_counter=[],
 )
 
 
@@ -306,6 +320,14 @@ def list_unets_callback():
     return res
 
 
+def before_token_counter_callback(params: BeforeTokenCounterParams):
+    for c in callback_map['callbacks_before_token_counter']:
+        try:
+            c.callback(params)
+        except Exception:
+            report_exception(c, 'before_token_counter')
+
+
 def add_callback(callbacks, fun):
     stack = [x for x in inspect.stack() if x.filename != __file__]
     filename = stack[0].filename if stack else 'unknown file'
@@ -480,3 +502,10 @@ def on_list_unets(callback):
     The function will be called with one argument, a list, and shall add objects of type modules.sd_unet.SdUnetOption to it."""
 
     add_callback(callback_map['callbacks_list_unets'], callback)
+
+
+def on_before_token_counter(callback):
+    """register a function to be called when UI is counting tokens for a prompt.
+    The function will be called with one argument of type BeforeTokenCounterParams, and should modify its fields if necessary."""
+
+    add_callback(callback_map['callbacks_before_token_counter'], callback)
diff --git a/modules/scripts.py b/modules/scripts.py
index 7f9454eb578..77f5e4f3e86 100644
--- a/modules/scripts.py
+++ b/modules/scripts.py
@@ -11,11 +11,31 @@
 
 AlwaysVisible = object()
 
+class MaskBlendArgs:
+    def __init__(self, current_latent, nmask, init_latent, mask, blended_latent, denoiser=None, sigma=None):
+        self.current_latent = current_latent
+        self.nmask = nmask
+        self.init_latent = init_latent
+        self.mask = mask
+        self.blended_latent = blended_latent
+
+        self.denoiser = denoiser
+        self.is_final_blend = denoiser is None
+        self.sigma = sigma
+
+class PostSampleArgs:
+    def __init__(self, samples):
+        self.samples = samples
 
 class PostprocessImageArgs:
     def __init__(self, image):
         self.image = image
 
+class PostProcessMaskOverlayArgs:
+    def __init__(self, index, mask_for_overlay, overlay_image):
+        self.index = index
+        self.mask_for_overlay = mask_for_overlay
+        self.overlay_image = overlay_image
 
 class PostprocessBatchListArgs:
     def __init__(self, images):
@@ -71,6 +91,9 @@ class Script:
     setup_for_ui_only = False
     """If true, the script setup will only be run in Gradio UI, not in API"""
 
+    controls = None
+    """A list of controls returned by the ui()."""
+
     def title(self):
         """this function should return the title of the script. This is what will be displayed in the dropdown menu."""
 
@@ -86,7 +109,7 @@ def ui(self, is_img2img):
 
     def show(self, is_img2img):
         """
-        is_img2img is True if this function is called for the img2img interface, and Fasle otherwise
+        is_img2img is True if this function is called for the img2img interface, and False otherwise
 
         This function should return:
          - False if the script should not be shown in UI at all
@@ -206,6 +229,25 @@ def postprocess_batch_list(self, p, pp: PostprocessBatchListArgs, *args, **kwarg
 
         pass
 
+    def on_mask_blend(self, p, mba: MaskBlendArgs, *args):
+        """
+        Called in inpainting mode when the original content is blended with the inpainted content.
+        This is called at every step in the denoising process and once at the end.
+        If is_final_blend is true, this is called for the final blending stage.
+        Otherwise, denoiser and sigma are defined and may be used to inform the procedure.
+        """
+
+        pass
+
+    def post_sample(self, p, ps: PostSampleArgs, *args):
+        """
+        Called after the samples have been generated,
+        but before they have been decoded by the VAE, if applicable.
+        Check getattr(samples, 'already_decoded', False) to test if the images are decoded.
+        """
+
+        pass
+
     def postprocess_image(self, p, pp: PostprocessImageArgs, *args):
         """
         Called for every image after it has been generated.
@@ -213,6 +255,22 @@ def postprocess_image(self, p, pp: PostprocessImageArgs, *args):
 
         pass
 
+    def postprocess_maskoverlay(self, p, ppmo: PostProcessMaskOverlayArgs, *args):
+        """
+        Called for every image after it has been generated.
+        """
+
+        pass
+
+    def postprocess_image_after_composite(self, p, pp: PostprocessImageArgs, *args):
+        """
+        Called for every image after it has been generated.
+        Same as postprocess_image but after inpaint_full_res composite
+        So that it operates on the full image instead of the inpaint_full_res crop region.
+        """
+
+        pass
+
     def postprocess(self, p, processed, *args):
         """
         This function is called after processing ends for AlwaysVisible scripts.
@@ -520,7 +578,12 @@ def initialize_scripts(self, is_img2img):
         auto_processing_scripts = scripts_auto_postprocessing.create_auto_preprocessing_script_data()
 
         for script_data in auto_processing_scripts + scripts_data:
-            script = script_data.script_class()
+            try:
+                script = script_data.script_class()
+            except Exception:
+                errors.report(f"Error # failed to initialize Script {script_data.module}: ", exc_info=True)
+                continue
+
             script.filename = script_data.path
             script.is_txt2img = not is_img2img
             script.is_img2img = is_img2img
@@ -573,6 +636,7 @@ def create_script_ui_inner(self, script):
         import modules.api.models as api_models
 
         controls = wrap_call(script.ui, script.filename, "ui", script.is_img2img)
+        script.controls = controls
 
         if controls is None:
             return
@@ -645,6 +709,8 @@ def setup_ui(self):
         self.setup_ui_for_section(None, self.selectable_scripts)
 
         def select_script(script_index):
+            if script_index is None:
+                script_index = 0
             selected_script = self.selectable_scripts[script_index - 1] if script_index>0 else None
 
             return [gr.update(visible=selected_script == s) for s in self.selectable_scripts]
@@ -688,7 +754,7 @@ def onload_script_visibility(params):
     def run(self, p, *args):
         script_index = args[0]
 
-        if script_index == 0:
+        if script_index == 0 or script_index is None:
             return None
 
         script = self.selectable_scripts[script_index-1]
@@ -767,6 +833,22 @@ def postprocess_batch_list(self, p, pp: PostprocessBatchListArgs, **kwargs):
             except Exception:
                 errors.report(f"Error running postprocess_batch_list: {script.filename}", exc_info=True)
 
+    def post_sample(self, p, ps: PostSampleArgs):
+        for script in self.alwayson_scripts:
+            try:
+                script_args = p.script_args[script.args_from:script.args_to]
+                script.post_sample(p, ps, *script_args)
+            except Exception:
+                errors.report(f"Error running post_sample: {script.filename}", exc_info=True)
+
+    def on_mask_blend(self, p, mba: MaskBlendArgs):
+        for script in self.alwayson_scripts:
+            try:
+                script_args = p.script_args[script.args_from:script.args_to]
+                script.on_mask_blend(p, mba, *script_args)
+            except Exception:
+                errors.report(f"Error running post_sample: {script.filename}", exc_info=True)
+
     def postprocess_image(self, p, pp: PostprocessImageArgs):
         for script in self.alwayson_scripts:
             try:
@@ -775,6 +857,22 @@ def postprocess_image(self, p, pp: PostprocessImageArgs):
             except Exception:
                 errors.report(f"Error running postprocess_image: {script.filename}", exc_info=True)
 
+    def postprocess_maskoverlay(self, p, ppmo: PostProcessMaskOverlayArgs):
+        for script in self.alwayson_scripts:
+            try:
+                script_args = p.script_args[script.args_from:script.args_to]
+                script.postprocess_maskoverlay(p, ppmo, *script_args)
+            except Exception:
+                errors.report(f"Error running postprocess_image: {script.filename}", exc_info=True)
+
+    def postprocess_image_after_composite(self, p, pp: PostprocessImageArgs):
+        for script in self.alwayson_scripts:
+            try:
+                script_args = p.script_args[script.args_from:script.args_to]
+                script.postprocess_image_after_composite(p, pp, *script_args)
+            except Exception:
+                errors.report(f"Error running postprocess_image_after_composite: {script.filename}", exc_info=True)
+
     def before_component(self, component, **kwargs):
         for callback, script in self.on_before_component_elem_id.get(kwargs.get("elem_id"), []):
             try:
@@ -841,6 +939,35 @@ def setup_scrips(self, p, *, is_ui=True):
             except Exception:
                 errors.report(f"Error running setup: {script.filename}", exc_info=True)
 
+    def set_named_arg(self, args, script_name, arg_elem_id, value, fuzzy=False):
+        """Locate an arg of a specific script in script_args and set its value
+        Args:
+            args: all script args of process p, p.script_args
+            script_name: the name target script name to
+            arg_elem_id: the elem_id of the target arg
+            value: the value to set
+            fuzzy: if True, arg_elem_id can be a substring of the control.elem_id else exact match
+        Returns:
+            Updated script args
+        when script_name in not found or arg_elem_id is not found in script controls, raise RuntimeError
+        """
+        script = next((x for x in self.scripts if x.name == script_name), None)
+        if script is None:
+            raise RuntimeError(f"script {script_name} not found")
+
+        for i, control in enumerate(script.controls):
+            if arg_elem_id in control.elem_id if fuzzy else arg_elem_id == control.elem_id:
+                index = script.args_from + i
+
+                if isinstance(args, tuple):
+                    return args[:index] + (value,) + args[index + 1:]
+                elif isinstance(args, list):
+                    args[index] = value
+                    return args
+                else:
+                    raise RuntimeError(f"args is not a list or tuple, but {type(args)}")
+        raise RuntimeError(f"arg_elem_id {arg_elem_id} not found in script {script_name}")
+
 
 scripts_txt2img: ScriptRunner = None
 scripts_img2img: ScriptRunner = None
diff --git a/modules/sd_emphasis.py b/modules/sd_emphasis.py
new file mode 100644
index 00000000000..49ef1a6acb3
--- /dev/null
+++ b/modules/sd_emphasis.py
@@ -0,0 +1,70 @@
+from __future__ import annotations
+import torch
+
+
+class Emphasis:
+    """Emphasis class decides how to death with (emphasized:1.1) text in prompts"""
+
+    name: str = "Base"
+    description: str = ""
+
+    tokens: list[list[int]]
+    """tokens from the chunk of the prompt"""
+
+    multipliers: torch.Tensor
+    """tensor with multipliers, once for each token"""
+
+    z: torch.Tensor
+    """output of cond transformers network (CLIP)"""
+
+    def after_transformers(self):
+        """Called after cond transformers network has processed the chunk of the prompt; this function should modify self.z to apply the emphasis"""
+
+        pass
+
+
+class EmphasisNone(Emphasis):
+    name = "None"
+    description = "disable the mechanism entirely and treat (:.1.1) as literal characters"
+
+
+class EmphasisIgnore(Emphasis):
+    name = "Ignore"
+    description = "treat all empasised words as if they have no emphasis"
+
+
+class EmphasisOriginal(Emphasis):
+    name = "Original"
+    description = "the original emphasis implementation"
+
+    def after_transformers(self):
+        original_mean = self.z.mean()
+        self.z = self.z * self.multipliers.reshape(self.multipliers.shape + (1,)).expand(self.z.shape)
+
+        # restoring original mean is likely not correct, but it seems to work well to prevent artifacts that happen otherwise
+        new_mean = self.z.mean()
+        self.z = self.z * (original_mean / new_mean)
+
+
+class EmphasisOriginalNoNorm(EmphasisOriginal):
+    name = "No norm"
+    description = "same as original, but without normalization (seems to work better for SDXL)"
+
+    def after_transformers(self):
+        self.z = self.z * self.multipliers.reshape(self.multipliers.shape + (1,)).expand(self.z.shape)
+
+
+def get_current_option(emphasis_option_name):
+    return next(iter([x for x in options if x.name == emphasis_option_name]), EmphasisOriginal)
+
+
+def get_options_descriptions():
+    return ", ".join(f"{x.name}: {x.description}" for x in options)
+
+
+options = [
+    EmphasisNone,
+    EmphasisIgnore,
+    EmphasisOriginal,
+    EmphasisOriginalNoNorm,
+]
diff --git a/modules/sd_hijack_clip.py b/modules/sd_hijack_clip.py
index 8f29057a9cf..6ef10ac7cd8 100644
--- a/modules/sd_hijack_clip.py
+++ b/modules/sd_hijack_clip.py
@@ -3,7 +3,7 @@
 
 import torch
 
-from modules import prompt_parser, devices, sd_hijack
+from modules import prompt_parser, devices, sd_hijack, sd_emphasis
 from modules.shared import opts
 
 
@@ -23,7 +23,7 @@ def __init__(self):
 
 PromptChunkFix = namedtuple('PromptChunkFix', ['offset', 'embedding'])
 """An object of this type is a marker showing that textual inversion embedding's vectors have to placed at offset in the prompt
-chunk. Thos objects are found in PromptChunk.fixes and, are placed into FrozenCLIPEmbedderWithCustomWordsBase.hijack.fixes, and finally
+chunk. Those objects are found in PromptChunk.fixes and, are placed into FrozenCLIPEmbedderWithCustomWordsBase.hijack.fixes, and finally
 are applied by sd_hijack.EmbeddingsWithFixes's forward function."""
 
 
@@ -66,7 +66,7 @@ def tokenize(self, texts):
 
     def encode_with_transformers(self, tokens):
         """
-        converts a batch of token ids (in python lists) into a single tensor with numeric respresentation of those tokens;
+        converts a batch of token ids (in python lists) into a single tensor with numeric representation of those tokens;
         All python lists with tokens are assumed to have same length, usually 77.
         if input is a list with B elements and each element has T tokens, expected output shape is (B, T, C), where C depends on
         model - can be 768 and 1024.
@@ -88,7 +88,7 @@ def tokenize_line(self, line):
         Returns the list and the total number of tokens in the prompt.
         """
 
-        if opts.enable_emphasis:
+        if opts.emphasis != "None":
             parsed = prompt_parser.parse_prompt_attention(line)
         else:
             parsed = [[line, 1.0]]
@@ -136,7 +136,7 @@ def next_chunk(is_last=False):
                 if token == self.comma_token:
                     last_comma = len(chunk.tokens)
 
-                # this is when we are at the end of alloted 75 tokens for the current chunk, and the current token is not a comma. opts.comma_padding_backtrack
+                # this is when we are at the end of allotted 75 tokens for the current chunk, and the current token is not a comma. opts.comma_padding_backtrack
                 # is a setting that specifies that if there is a comma nearby, the text after the comma should be moved out of this chunk and into the next.
                 elif opts.comma_padding_backtrack != 0 and len(chunk.tokens) == self.chunk_length and last_comma != -1 and len(chunk.tokens) - last_comma <= opts.comma_padding_backtrack:
                     break_location = last_comma + 1
@@ -206,7 +206,7 @@ def forward(self, texts):
         be a multiple of 77; and C is dimensionality of each token - for SD1 it's 768, for SD2 it's 1024, and for SDXL it's 1280.
         An example shape returned by this function can be: (2, 77, 768).
         For SDXL, instead of returning one tensor avobe, it returns a tuple with two: the other one with shape (B, 1280) with pooled values.
-        Webui usually sends just one text at a time through this function - the only time when texts is an array with more than one elemenet
+        Webui usually sends just one text at a time through this function - the only time when texts is an array with more than one element
         is when you do prompt editing: "a picture of a [cat:dog:0.4] eating ice cream"
         """
 
@@ -230,7 +230,7 @@ def forward(self, texts):
             for fixes in self.hijack.fixes:
                 for _position, embedding in fixes:
                     used_embeddings[embedding.name] = embedding
-
+            devices.torch_npu_set_device()
             z = self.process_tokens(tokens, multipliers)
             zs.append(z)
 
@@ -249,6 +249,9 @@ def forward(self, texts):
                     hashes.append(self.hijack.extra_generation_params.get("TI hashes"))
                 self.hijack.extra_generation_params["TI hashes"] = ", ".join(hashes)
 
+        if any(x for x in texts if "(" in x or "[" in x) and opts.emphasis != "Original":
+            self.hijack.extra_generation_params["Emphasis"] = opts.emphasis
+
         if getattr(self.wrapped, 'return_pooled', False):
             return torch.hstack(zs), zs[0].pooled
         else:
@@ -274,12 +277,14 @@ def process_tokens(self, remade_batch_tokens, batch_multipliers):
 
         pooled = getattr(z, 'pooled', None)
 
-        # restoring original mean is likely not correct, but it seems to work well to prevent artifacts that happen otherwise
-        batch_multipliers = torch.asarray(batch_multipliers).to(devices.device)
-        original_mean = z.mean()
-        z = z * batch_multipliers.reshape(batch_multipliers.shape + (1,)).expand(z.shape)
-        new_mean = z.mean()
-        z = z * (original_mean / new_mean)
+        emphasis = sd_emphasis.get_current_option(opts.emphasis)()
+        emphasis.tokens = remade_batch_tokens
+        emphasis.multipliers = torch.asarray(batch_multipliers).to(devices.device)
+        emphasis.z = z
+
+        emphasis.after_transformers()
+
+        z = emphasis.z
 
         if pooled is not None:
             z.pooled = pooled
diff --git a/modules/sd_hijack_clip_old.py b/modules/sd_hijack_clip_old.py
index c5c6270b9c4..43e9b9529e8 100644
--- a/modules/sd_hijack_clip_old.py
+++ b/modules/sd_hijack_clip_old.py
@@ -32,7 +32,7 @@ def process_text_old(self: sd_hijack_clip.FrozenCLIPEmbedderWithCustomWordsBase,
 
                 embedding, embedding_length_in_tokens = self.hijack.embedding_db.find_embedding_at_position(tokens, i)
 
-                mult_change = self.token_mults.get(token) if shared.opts.enable_emphasis else None
+                mult_change = self.token_mults.get(token) if shared.opts.emphasis != "None" else None
                 if mult_change is not None:
                     mult *= mult_change
                     i += 1
diff --git a/modules/sd_hijack_utils.py b/modules/sd_hijack_utils.py
index f8684475ec5..79bf6e46862 100644
--- a/modules/sd_hijack_utils.py
+++ b/modules/sd_hijack_utils.py
@@ -11,10 +11,14 @@ def __new__(cls, orig_func, sub_func, cond_func):
                     break
                 except ImportError:
                     pass
-            for attr_name in func_path[i:-1]:
-                resolved_obj = getattr(resolved_obj, attr_name)
-            orig_func = getattr(resolved_obj, func_path[-1])
-            setattr(resolved_obj, func_path[-1], lambda *args, **kwargs: self(*args, **kwargs))
+            try:
+                for attr_name in func_path[i:-1]:
+                    resolved_obj = getattr(resolved_obj, attr_name)
+                orig_func = getattr(resolved_obj, func_path[-1])
+                setattr(resolved_obj, func_path[-1], lambda *args, **kwargs: self(*args, **kwargs))
+            except AttributeError:
+                print(f"Warning: Failed to resolve {orig_func} for CondFunc hijack")
+                pass
         self.__init__(orig_func, sub_func, cond_func)
         return lambda *args, **kwargs: self(*args, **kwargs)
     def __init__(self, orig_func, sub_func, cond_func):
diff --git a/modules/sd_models.py b/modules/sd_models.py
index 9355f1e16b7..b35aecbcaae 100644
--- a/modules/sd_models.py
+++ b/modules/sd_models.py
@@ -15,6 +15,7 @@
 
 from modules import paths, shared, modelloader, devices, script_callbacks, sd_vae, sd_disable_initialization, errors, hashes, sd_models_config, sd_unet, sd_models_xl, cache, extra_networks, processing, lowvram, sd_hijack, patches
 from modules.timer import Timer
+from modules.shared import opts
 import tomesd
 import numpy as np
 
@@ -348,10 +349,28 @@ def __exit__(self, exc_type, exc_value, exc_traceback):
         SkipWritingToConfig.skip = self.previous
 
 
+def check_fp8(model):
+    if model is None:
+        return None
+    if devices.get_optimal_device_name() == "mps":
+        enable_fp8 = False
+    elif shared.opts.fp8_storage == "Enable":
+        enable_fp8 = True
+    elif getattr(model, "is_sdxl", False) and shared.opts.fp8_storage == "Enable for SDXL":
+        enable_fp8 = True
+    else:
+        enable_fp8 = False
+    return enable_fp8
+
+
 def load_model_weights(model, checkpoint_info: CheckpointInfo, state_dict, timer):
     sd_model_hash = checkpoint_info.calculate_shorthash()
     timer.record("calculate hash")
 
+    if devices.fp8:
+        # prevent model to load state dict in fp8
+        model.half()
+
     if not SkipWritingToConfig.skip:
         shared.opts.data["sd_model_checkpoint"] = checkpoint_info.title
 
@@ -383,6 +402,7 @@ def load_model_weights(model, checkpoint_info: CheckpointInfo, state_dict, timer
 
     if shared.cmd_opts.no_half:
         model.float()
+        model.alphas_cumprod_original = model.alphas_cumprod
         devices.dtype_unet = torch.float32
         timer.record("apply float()")
     else:
@@ -396,7 +416,11 @@ def load_model_weights(model, checkpoint_info: CheckpointInfo, state_dict, timer
         if shared.cmd_opts.upcast_sampling and depth_model:
             model.depth_model = None
 
+        alphas_cumprod = model.alphas_cumprod
+        model.alphas_cumprod = None
         model.half()
+        model.alphas_cumprod = alphas_cumprod
+        model.alphas_cumprod_original = alphas_cumprod
         model.first_stage_model = vae
         if depth_model:
             model.depth_model = depth_model
@@ -404,6 +428,30 @@ def load_model_weights(model, checkpoint_info: CheckpointInfo, state_dict, timer
         devices.dtype_unet = torch.float16
         timer.record("apply half()")
 
+    apply_alpha_schedule_override(model)
+
+    for module in model.modules():
+        if hasattr(module, 'fp16_weight'):
+            del module.fp16_weight
+        if hasattr(module, 'fp16_bias'):
+            del module.fp16_bias
+
+    if check_fp8(model):
+        devices.fp8 = True
+        first_stage = model.first_stage_model
+        model.first_stage_model = None
+        for module in model.modules():
+            if isinstance(module, (torch.nn.Conv2d, torch.nn.Linear)):
+                if shared.opts.cache_fp16_weight:
+                    module.fp16_weight = module.weight.data.clone().cpu().half()
+                    if module.bias is not None:
+                        module.fp16_bias = module.bias.data.clone().cpu().half()
+                module.to(torch.float8_e4m3fn)
+        model.first_stage_model = first_stage
+        timer.record("apply fp8")
+    else:
+        devices.fp8 = False
+
     devices.unet_needs_upcast = shared.cmd_opts.upcast_sampling and devices.dtype == torch.float16 and devices.dtype_unet == torch.float16
 
     model.first_stage_model.to(devices.dtype_vae)
@@ -505,6 +553,48 @@ def repair_config(sd_config):
         sd_config.model.params.noise_aug_config.params.clip_stats_path = sd_config.model.params.noise_aug_config.params.clip_stats_path.replace("checkpoints/karlo_models", karlo_path)
 
 
+def rescale_zero_terminal_snr_abar(alphas_cumprod):
+    alphas_bar_sqrt = alphas_cumprod.sqrt()
+
+    # Store old values.
+    alphas_bar_sqrt_0 = alphas_bar_sqrt[0].clone()
+    alphas_bar_sqrt_T = alphas_bar_sqrt[-1].clone()
+
+    # Shift so the last timestep is zero.
+    alphas_bar_sqrt -= (alphas_bar_sqrt_T)
+
+    # Scale so the first timestep is back to the old value.
+    alphas_bar_sqrt *= alphas_bar_sqrt_0 / (alphas_bar_sqrt_0 - alphas_bar_sqrt_T)
+
+    # Convert alphas_bar_sqrt to betas
+    alphas_bar = alphas_bar_sqrt ** 2  # Revert sqrt
+    alphas_bar[-1] = 4.8973451890853435e-08
+    return alphas_bar
+
+
+def apply_alpha_schedule_override(sd_model, p=None):
+    """
+    Applies an override to the alpha schedule of the model according to settings.
+    - downcasts the alpha schedule to half precision
+    - rescales the alpha schedule to have zero terminal SNR
+    """
+
+    if not hasattr(sd_model, 'alphas_cumprod') or not hasattr(sd_model, 'alphas_cumprod_original'):
+        return
+
+    sd_model.alphas_cumprod = sd_model.alphas_cumprod_original.to(shared.device)
+
+    if opts.use_downcasted_alpha_bar:
+        if p is not None:
+            p.extra_generation_params['Downcast alphas_cumprod'] = opts.use_downcasted_alpha_bar
+        sd_model.alphas_cumprod = sd_model.alphas_cumprod.half().to(shared.device)
+
+    if opts.sd_noise_schedule == "Zero Terminal SNR":
+        if p is not None:
+            p.extra_generation_params['Noise Schedule'] = opts.sd_noise_schedule
+        sd_model.alphas_cumprod = rescale_zero_terminal_snr_abar(sd_model.alphas_cumprod).to(shared.device)
+
+
 sd1_clip_weight = 'cond_stage_model.transformer.text_model.embeddings.token_embedding.weight'
 sd2_clip_weight = 'cond_stage_model.model.transformer.resblocks.0.attn.in_proj_weight'
 sdxl_clip_weight = 'conditioner.embedders.1.model.ln_final.weight'
@@ -651,6 +741,7 @@ def load_model(checkpoint_info=None, already_loaded_state_dict=None):
     else:
         weight_dtype_conversion = {
             'first_stage_model': None,
+            'alphas_cumprod': None,
             '': torch.float16,
         }
 
@@ -693,7 +784,7 @@ def reuse_model_from_already_loaded(sd_model, checkpoint_info, timer):
     If it is loaded, returns that (moving it to GPU if necessary, and moving the currently loadded model to CPU if necessary).
     If not, returns the model that can be used to load weights from checkpoint_info's file.
     If no such model exists, returns None.
-    Additionaly deletes loaded models that are over the limit set in settings (sd_checkpoints_limit).
+    Additionally deletes loaded models that are over the limit set in settings (sd_checkpoints_limit).
     """
 
     already_loaded = None
@@ -746,7 +837,7 @@ def reuse_model_from_already_loaded(sd_model, checkpoint_info, timer):
         return None
 
 
-def reload_model_weights(sd_model=None, info=None):
+def reload_model_weights(sd_model=None, info=None, forced_reload=False):
     checkpoint_info = info or select_checkpoint()
 
     timer = Timer()
@@ -758,11 +849,14 @@ def reload_model_weights(sd_model=None, info=None):
         current_checkpoint_info = None
     else:
         current_checkpoint_info = sd_model.sd_checkpoint_info
-        if sd_model.sd_model_checkpoint == checkpoint_info.filename:
+        if check_fp8(sd_model) != devices.fp8:
+            # load from state dict again to prevent extra numerical errors
+            forced_reload = True
+        elif sd_model.sd_model_checkpoint == checkpoint_info.filename and not forced_reload:
             return sd_model
 
     sd_model = reuse_model_from_already_loaded(sd_model, checkpoint_info, timer)
-    if sd_model is not None and sd_model.sd_checkpoint_info.filename == checkpoint_info.filename:
+    if not forced_reload and sd_model is not None and sd_model.sd_checkpoint_info.filename == checkpoint_info.filename:
         return sd_model
 
     if sd_model is not None:
@@ -793,13 +887,13 @@ def reload_model_weights(sd_model=None, info=None):
         sd_hijack.model_hijack.hijack(sd_model)
         timer.record("hijack")
 
-        script_callbacks.model_loaded_callback(sd_model)
-        timer.record("script callbacks")
-
         if not sd_model.lowvram:
             sd_model.to(devices.device)
             timer.record("move model to device")
 
+        script_callbacks.model_loaded_callback(sd_model)
+        timer.record("script callbacks")
+
     print(f"Weights loaded in {timer.summary()}.")
 
     model_data.set_sd_model(sd_model)
diff --git a/modules/sd_models_config.py b/modules/sd_models_config.py
index deab2f6e237..b38137eb5a9 100644
--- a/modules/sd_models_config.py
+++ b/modules/sd_models_config.py
@@ -15,6 +15,7 @@
 config_sd2_inpainting = os.path.join(sd_repo_configs_path, "v2-inpainting-inference.yaml")
 config_sdxl = os.path.join(sd_xl_repo_configs_path, "sd_xl_base.yaml")
 config_sdxl_refiner = os.path.join(sd_xl_repo_configs_path, "sd_xl_refiner.yaml")
+config_sdxl_inpainting = os.path.join(sd_configs_path, "sd_xl_inpaint.yaml")
 config_depth_model = os.path.join(sd_repo_configs_path, "v2-midas-inference.yaml")
 config_unclip = os.path.join(sd_repo_configs_path, "v2-1-stable-unclip-l-inference.yaml")
 config_unopenclip = os.path.join(sd_repo_configs_path, "v2-1-stable-unclip-h-inference.yaml")
@@ -71,7 +72,10 @@ def guess_model_config_from_state_dict(sd, filename):
     sd2_variations_weight = sd.get('embedder.model.ln_final.weight', None)
 
     if sd.get('conditioner.embedders.1.model.ln_final.weight', None) is not None:
-        return config_sdxl
+        if diffusion_model_input.shape[1] == 9:
+            return config_sdxl_inpainting
+        else:
+            return config_sdxl
     if sd.get('conditioner.embedders.0.model.ln_final.weight', None) is not None:
         return config_sdxl_refiner
     elif sd.get('depth_model.model.pretrained.act_postprocess3.0.project.0.bias', None) is not None:
diff --git a/modules/sd_models_xl.py b/modules/sd_models_xl.py
index 0112332161f..0de17af3db2 100644
--- a/modules/sd_models_xl.py
+++ b/modules/sd_models_xl.py
@@ -6,6 +6,7 @@
 import sgm.modules.diffusionmodules.denoiser_scaling
 import sgm.modules.diffusionmodules.discretizer
 from modules import devices, shared, prompt_parser
+from modules import torch_utils
 
 
 def get_learned_conditioning(self: sgm.models.diffusion.DiffusionEngine, batch: prompt_parser.SdConditioning | list[str]):
@@ -34,6 +35,12 @@ def get_learned_conditioning(self: sgm.models.diffusion.DiffusionEngine, batch:
 
 
 def apply_model(self: sgm.models.diffusion.DiffusionEngine, x, t, cond):
+    sd = self.model.state_dict()
+    diffusion_model_input = sd.get('diffusion_model.input_blocks.0.0.weight', None)
+    if diffusion_model_input is not None:
+        if diffusion_model_input.shape[1] == 9:
+            x = torch.cat([x] + cond['c_concat'], dim=1)
+
     return self.model(x, t, cond)
 
 
@@ -84,7 +91,7 @@ def get_target_prompt_token_count(self, token_count):
 def extend_sdxl(model):
     """this adds a bunch of parameters to make SDXL model look a bit more like SD1.5 to the rest of the codebase."""
 
-    dtype = next(model.model.diffusion_model.parameters()).dtype
+    dtype = torch_utils.get_param(model.model.diffusion_model).dtype
     model.model.diffusion_model.dtype = dtype
     model.model.conditioning_key = 'crossattn'
     model.cond_stage_key = 'txt'
@@ -93,7 +100,7 @@ def extend_sdxl(model):
     model.parameterization = "v" if isinstance(model.denoiser.scaling, sgm.modules.diffusionmodules.denoiser_scaling.VScaling) else "eps"
 
     discretization = sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization()
-    model.alphas_cumprod = torch.asarray(discretization.alphas_cumprod, device=devices.device, dtype=dtype)
+    model.alphas_cumprod = torch.asarray(discretization.alphas_cumprod, device=devices.device, dtype=torch.float32)
 
     model.conditioner.wrapped = torch.nn.Module()
 
diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py
index 45faae62821..a58528a0b52 100644
--- a/modules/sd_samplers.py
+++ b/modules/sd_samplers.py
@@ -1,4 +1,4 @@
-from modules import sd_samplers_kdiffusion, sd_samplers_timesteps, shared
+from modules import sd_samplers_kdiffusion, sd_samplers_timesteps, sd_samplers_lcm, shared
 
 # imports for functions that previously were here and are used by other modules
 from modules.sd_samplers_common import samples_to_image_grid, sample_to_image  # noqa: F401
@@ -6,6 +6,7 @@
 all_samplers = [
     *sd_samplers_kdiffusion.samplers_data_k_diffusion,
     *sd_samplers_timesteps.samplers_data_timesteps,
+    *sd_samplers_lcm.samplers_data_lcm,
 ]
 all_samplers_map = {x.name: x for x in all_samplers}
 
diff --git a/modules/sd_samplers_cfg_denoiser.py b/modules/sd_samplers_cfg_denoiser.py
index b8101d38dc3..93581c9acc6 100644
--- a/modules/sd_samplers_cfg_denoiser.py
+++ b/modules/sd_samplers_cfg_denoiser.py
@@ -53,9 +53,13 @@ def __init__(self, sampler):
         self.step = 0
         self.image_cfg_scale = None
         self.padded_cond_uncond = False
+        self.padded_cond_uncond_v0 = False
         self.sampler = sampler
         self.model_wrap = None
         self.p = None
+
+        # NOTE: masking before denoising can cause the original latents to be oversmoothed
+        # as the original latents do not have noise
         self.mask_before_denoising = False
 
     @property
@@ -88,11 +92,67 @@ def update_inner_model(self):
         self.sampler.sampler_extra_args['cond'] = c
         self.sampler.sampler_extra_args['uncond'] = uc
 
+    def pad_cond_uncond(self, cond, uncond):
+        empty = shared.sd_model.cond_stage_model_empty_prompt
+        num_repeats = (cond.shape[1] - uncond.shape[1]) // empty.shape[1]
+
+        if num_repeats < 0:
+            cond = pad_cond(cond, -num_repeats, empty)
+            self.padded_cond_uncond = True
+        elif num_repeats > 0:
+            uncond = pad_cond(uncond, num_repeats, empty)
+            self.padded_cond_uncond = True
+
+        return cond, uncond
+
+    def pad_cond_uncond_v0(self, cond, uncond):
+        """
+        Pads the 'uncond' tensor to match the shape of the 'cond' tensor.
+
+        If 'uncond' is a dictionary, it is assumed that the 'crossattn' key holds the tensor to be padded.
+        If 'uncond' is a tensor, it is padded directly.
+
+        If the number of columns in 'uncond' is less than the number of columns in 'cond', the last column of 'uncond'
+        is repeated to match the number of columns in 'cond'.
+
+        If the number of columns in 'uncond' is greater than the number of columns in 'cond', 'uncond' is truncated
+        to match the number of columns in 'cond'.
+
+        Args:
+            cond (torch.Tensor or DictWithShape): The condition tensor to match the shape of 'uncond'.
+            uncond (torch.Tensor or DictWithShape): The tensor to be padded, or a dictionary containing the tensor to be padded.
+
+        Returns:
+            tuple: A tuple containing the 'cond' tensor and the padded 'uncond' tensor.
+
+        Note:
+            This is the padding that was always used in DDIM before version 1.6.0
+        """
+
+        is_dict_cond = isinstance(uncond, dict)
+        uncond_vec = uncond['crossattn'] if is_dict_cond else uncond
+
+        if uncond_vec.shape[1] < cond.shape[1]:
+            last_vector = uncond_vec[:, -1:]
+            last_vector_repeated = last_vector.repeat([1, cond.shape[1] - uncond_vec.shape[1], 1])
+            uncond_vec = torch.hstack([uncond_vec, last_vector_repeated])
+            self.padded_cond_uncond_v0 = True
+        elif uncond_vec.shape[1] > cond.shape[1]:
+            uncond_vec = uncond_vec[:, :cond.shape[1]]
+            self.padded_cond_uncond_v0 = True
+
+        if is_dict_cond:
+            uncond['crossattn'] = uncond_vec
+        else:
+            uncond = uncond_vec
+
+        return cond, uncond
+
     def forward(self, x, sigma, uncond, cond, cond_scale, s_min_uncond, image_cond):
         if state.interrupted or state.skipped:
             raise sd_samplers_common.InterruptedException
 
-        if sd_samplers_common.apply_refiner(self):
+        if sd_samplers_common.apply_refiner(self, sigma):
             cond = self.sampler.sampler_extra_args['cond']
             uncond = self.sampler.sampler_extra_args['uncond']
 
@@ -105,8 +165,21 @@ def forward(self, x, sigma, uncond, cond, cond_scale, s_min_uncond, image_cond):
 
         assert not is_edit_model or all(len(conds) == 1 for conds in conds_list), "AND is not supported for InstructPix2Pix checkpoint (unless using Image CFG scale = 1.0)"
 
+        # If we use masks, blending between the denoised and original latent images occurs here.
+        def apply_blend(current_latent):
+            blended_latent = current_latent * self.nmask + self.init_latent * self.mask
+
+            if self.p.scripts is not None:
+                from modules import scripts
+                mba = scripts.MaskBlendArgs(current_latent, self.nmask, self.init_latent, self.mask, blended_latent, denoiser=self, sigma=sigma)
+                self.p.scripts.on_mask_blend(self.p, mba)
+                blended_latent = mba.blended_latent
+
+            return blended_latent
+
+        # Blend in the original latents (before)
         if self.mask_before_denoising and self.mask is not None:
-            x = self.init_latent * self.mask + self.nmask * x
+            x = apply_blend(x)
 
         batch_size = len(conds_list)
         repeats = [len(conds_list[i]) for i in range(batch_size)]
@@ -130,7 +203,7 @@ def forward(self, x, sigma, uncond, cond, cond_scale, s_min_uncond, image_cond):
             sigma_in = torch.cat([torch.stack([sigma[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [sigma] + [sigma])
             image_cond_in = torch.cat([torch.stack([image_cond[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [image_uncond] + [torch.zeros_like(self.init_latent)])
 
-        denoiser_params = CFGDenoiserParams(x_in, image_cond_in, sigma_in, state.sampling_step, state.sampling_steps, tensor, uncond)
+        denoiser_params = CFGDenoiserParams(x_in, image_cond_in, sigma_in, state.sampling_step, state.sampling_steps, tensor, uncond, self)
         cfg_denoiser_callback(denoiser_params)
         x_in = denoiser_params.x
         image_cond_in = denoiser_params.image_cond
@@ -146,16 +219,11 @@ def forward(self, x, sigma, uncond, cond, cond_scale, s_min_uncond, image_cond):
             sigma_in = sigma_in[:-batch_size]
 
         self.padded_cond_uncond = False
-        if shared.opts.pad_cond_uncond and tensor.shape[1] != uncond.shape[1]:
-            empty = shared.sd_model.cond_stage_model_empty_prompt
-            num_repeats = (tensor.shape[1] - uncond.shape[1]) // empty.shape[1]
-
-            if num_repeats < 0:
-                tensor = pad_cond(tensor, -num_repeats, empty)
-                self.padded_cond_uncond = True
-            elif num_repeats > 0:
-                uncond = pad_cond(uncond, num_repeats, empty)
-                self.padded_cond_uncond = True
+        self.padded_cond_uncond_v0 = False
+        if shared.opts.pad_cond_uncond_v0 and tensor.shape[1] != uncond.shape[1]:
+            tensor, uncond = self.pad_cond_uncond_v0(tensor, uncond)
+        elif shared.opts.pad_cond_uncond and tensor.shape[1] != uncond.shape[1]:
+            tensor, uncond = self.pad_cond_uncond(tensor, uncond)
 
         if tensor.shape[1] == uncond.shape[1] or skip_uncond:
             if is_edit_model:
@@ -207,8 +275,9 @@ def forward(self, x, sigma, uncond, cond, cond_scale, s_min_uncond, image_cond):
         else:
             denoised = self.combine_denoised(x_out, conds_list, uncond, cond_scale)
 
+        # Blend in the original latents (after)
         if not self.mask_before_denoising and self.mask is not None:
-            denoised = self.init_latent * self.mask + self.nmask * denoised
+            denoised = apply_blend(denoised)
 
         self.sampler.last_latent = self.get_pred_x0(torch.cat([x_in[i:i + 1] for i in denoised_image_indexes]), torch.cat([x_out[i:i + 1] for i in denoised_image_indexes]), sigma)
 
diff --git a/modules/sd_samplers_common.py b/modules/sd_samplers_common.py
index 58efcad2374..bda578cc5b8 100644
--- a/modules/sd_samplers_common.py
+++ b/modules/sd_samplers_common.py
@@ -155,8 +155,19 @@ def torchsde_randn(size, dtype, device, seed):
 replace_torchsde_browinan()
 
 
-def apply_refiner(cfg_denoiser):
-    completed_ratio = cfg_denoiser.step / cfg_denoiser.total_steps
+def apply_refiner(cfg_denoiser, sigma=None):
+    if opts.refiner_switch_by_sample_steps or sigma is None:
+        completed_ratio = cfg_denoiser.step / cfg_denoiser.total_steps
+        cfg_denoiser.p.extra_generation_params["Refiner switch by sampling steps"] = True
+
+    else:
+        # torch.max(sigma) only to handle rare case where we might have different sigmas in the same batch
+        try:
+            timestep = torch.argmin(torch.abs(cfg_denoiser.inner_model.sigmas - torch.max(sigma)))
+        except AttributeError:  # for samplers that don't use sigmas (DDIM) sigma is actually the timestep
+            timestep = torch.max(sigma).to(dtype=int)
+        completed_ratio = (999 - timestep) / 1000
+
     refiner_switch_at = cfg_denoiser.p.refiner_switch_at
     refiner_checkpoint_info = cfg_denoiser.p.refiner_checkpoint_info
 
@@ -335,3 +346,10 @@ def sample(self, p, x, conditioning, unconditional_conditioning, steps=None, ima
 
     def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning, steps=None, image_conditioning=None):
         raise NotImplementedError()
+
+    def add_infotext(self, p):
+        if self.model_wrap_cfg.padded_cond_uncond:
+            p.extra_generation_params["Pad conds"] = True
+
+        if self.model_wrap_cfg.padded_cond_uncond_v0:
+            p.extra_generation_params["Pad conds v0"] = True
diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py
index 8a8c87e0d01..337106c0224 100644
--- a/modules/sd_samplers_kdiffusion.py
+++ b/modules/sd_samplers_kdiffusion.py
@@ -187,8 +187,7 @@ def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning,
 
         samples = self.launch_sampling(t_enc + 1, lambda: self.func(self.model_wrap_cfg, xi, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
 
-        if self.model_wrap_cfg.padded_cond_uncond:
-            p.extra_generation_params["Pad conds"] = True
+        self.add_infotext(p)
 
         return samples
 
@@ -234,8 +233,7 @@ def sample(self, p, x, conditioning, unconditional_conditioning, steps=None, ima
 
         samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
 
-        if self.model_wrap_cfg.padded_cond_uncond:
-            p.extra_generation_params["Pad conds"] = True
+        self.add_infotext(p)
 
         return samples
 
diff --git a/modules/sd_samplers_lcm.py b/modules/sd_samplers_lcm.py
new file mode 100644
index 00000000000..59839b720dd
--- /dev/null
+++ b/modules/sd_samplers_lcm.py
@@ -0,0 +1,104 @@
+import torch
+
+from k_diffusion import utils, sampling
+from k_diffusion.external import DiscreteEpsDDPMDenoiser
+from k_diffusion.sampling import default_noise_sampler, trange
+
+from modules import shared, sd_samplers_cfg_denoiser, sd_samplers_kdiffusion, sd_samplers_common
+
+
+class LCMCompVisDenoiser(DiscreteEpsDDPMDenoiser):
+    def __init__(self, model):
+        timesteps = 1000
+        original_timesteps = 50     # LCM Original Timesteps (default=50, for current version of LCM)
+        self.skip_steps = timesteps // original_timesteps
+
+        alphas_cumprod_valid = torch.zeros((original_timesteps), dtype=torch.float32, device=model.device)
+        for x in range(original_timesteps):
+            alphas_cumprod_valid[original_timesteps - 1 - x] = model.alphas_cumprod[timesteps - 1 - x * self.skip_steps]
+
+        super().__init__(model, alphas_cumprod_valid, quantize=None)
+
+
+    def get_sigmas(self, n=None,):
+        if n is None:
+            return sampling.append_zero(self.sigmas.flip(0))
+
+        start = self.sigma_to_t(self.sigma_max)
+        end = self.sigma_to_t(self.sigma_min)
+
+        t = torch.linspace(start, end, n, device=shared.sd_model.device)
+
+        return sampling.append_zero(self.t_to_sigma(t))
+
+
+    def sigma_to_t(self, sigma, quantize=None):
+        log_sigma = sigma.log()
+        dists = log_sigma - self.log_sigmas[:, None]
+        return dists.abs().argmin(dim=0).view(sigma.shape) * self.skip_steps + (self.skip_steps - 1)
+
+
+    def t_to_sigma(self, timestep):
+        t = torch.clamp(((timestep - (self.skip_steps - 1)) / self.skip_steps).float(), min=0, max=(len(self.sigmas) - 1))
+        return super().t_to_sigma(t)
+
+
+    def get_eps(self, *args, **kwargs):
+        return self.inner_model.apply_model(*args, **kwargs)
+
+
+    def get_scaled_out(self, sigma, output, input):
+        sigma_data = 0.5
+        scaled_timestep = utils.append_dims(self.sigma_to_t(sigma), output.ndim) * 10.0
+
+        c_skip = sigma_data**2 / (scaled_timestep**2 + sigma_data**2)
+        c_out = scaled_timestep / (scaled_timestep**2 + sigma_data**2) ** 0.5
+
+        return c_out * output + c_skip * input
+
+
+    def forward(self, input, sigma, **kwargs):
+        c_out, c_in = [utils.append_dims(x, input.ndim) for x in self.get_scalings(sigma)]
+        eps = self.get_eps(input * c_in, self.sigma_to_t(sigma), **kwargs)
+        return self.get_scaled_out(sigma, input + eps * c_out, input)
+
+
+def sample_lcm(model, x, sigmas, extra_args=None, callback=None, disable=None, noise_sampler=None):
+    extra_args = {} if extra_args is None else extra_args
+    noise_sampler = default_noise_sampler(x) if noise_sampler is None else noise_sampler
+    s_in = x.new_ones([x.shape[0]])
+
+    for i in trange(len(sigmas) - 1, disable=disable):
+        denoised = model(x, sigmas[i] * s_in, **extra_args)
+
+        if callback is not None:
+            callback({'x': x, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i], 'denoised': denoised})
+
+        x = denoised
+        if sigmas[i + 1] > 0:
+            x += sigmas[i + 1] * noise_sampler(sigmas[i], sigmas[i + 1])
+    return x
+
+
+class CFGDenoiserLCM(sd_samplers_cfg_denoiser.CFGDenoiser):
+    @property
+    def inner_model(self):
+        if self.model_wrap is None:
+            denoiser = LCMCompVisDenoiser
+            self.model_wrap = denoiser(shared.sd_model)
+
+        return self.model_wrap
+
+
+class LCMSampler(sd_samplers_kdiffusion.KDiffusionSampler):
+    def __init__(self, funcname, sd_model, options=None):
+        super().__init__(funcname, sd_model, options)
+        self.model_wrap_cfg = CFGDenoiserLCM(self)
+        self.model_wrap = self.model_wrap_cfg.inner_model
+
+
+samplers_lcm = [('LCM', sample_lcm, ['k_lcm'], {})]
+samplers_data_lcm = [
+    sd_samplers_common.SamplerData(label, lambda model, funcname=funcname: LCMSampler(funcname, model), aliases, options)
+    for label, funcname, aliases, options in samplers_lcm
+]
diff --git a/modules/sd_samplers_timesteps.py b/modules/sd_samplers_timesteps.py
index b17a8f93c2b..8cc7d3848aa 100644
--- a/modules/sd_samplers_timesteps.py
+++ b/modules/sd_samplers_timesteps.py
@@ -36,7 +36,7 @@ def __init__(self, model, *args, **kwargs):
         self.inner_model = model
 
     def predict_eps_from_z_and_v(self, x_t, t, v):
-        return self.inner_model.sqrt_alphas_cumprod[t.to(torch.int), None, None, None] * v + self.inner_model.sqrt_one_minus_alphas_cumprod[t.to(torch.int), None, None, None] * x_t
+        return torch.sqrt(self.inner_model.alphas_cumprod)[t.to(torch.int), None, None, None] * v + torch.sqrt(1 - self.inner_model.alphas_cumprod)[t.to(torch.int), None, None, None] * x_t
 
     def forward(self, input, timesteps, **kwargs):
         model_output = self.inner_model.apply_model(input, timesteps, **kwargs)
@@ -80,6 +80,7 @@ def __init__(self, funcname, sd_model):
         self.eta_default = 0.0
 
         self.model_wrap_cfg = CFGDenoiserTimesteps(self)
+        self.model_wrap = self.model_wrap_cfg.inner_model
 
     def get_timesteps(self, p, steps):
         discard_next_to_last_sigma = self.config is not None and self.config.options.get('discard_next_to_last_sigma', False)
@@ -132,8 +133,7 @@ def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning,
 
         samples = self.launch_sampling(t_enc + 1, lambda: self.func(self.model_wrap_cfg, xi, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
 
-        if self.model_wrap_cfg.padded_cond_uncond:
-            p.extra_generation_params["Pad conds"] = True
+        self.add_infotext(p)
 
         return samples
 
@@ -157,8 +157,7 @@ def sample(self, p, x, conditioning, unconditional_conditioning, steps=None, ima
         }
         samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
 
-        if self.model_wrap_cfg.padded_cond_uncond:
-            p.extra_generation_params["Pad conds"] = True
+        self.add_infotext(p)
 
         return samples
 
diff --git a/modules/sd_vae.py b/modules/sd_vae.py
index 31306d8ba4b..43687e48dcf 100644
--- a/modules/sd_vae.py
+++ b/modules/sd_vae.py
@@ -273,10 +273,11 @@ def reload_vae_weights(sd_model=None, vae_file=unspecified):
     load_vae(sd_model, vae_file, vae_source)
 
     sd_hijack.model_hijack.hijack(sd_model)
-    script_callbacks.model_loaded_callback(sd_model)
 
     if not sd_model.lowvram:
         sd_model.to(devices.device)
 
+    script_callbacks.model_loaded_callback(sd_model)
+
     print("VAE weights loaded.")
     return sd_model
diff --git a/modules/shared.py b/modules/shared.py
index 636619391fc..b4ba14ad7c6 100644
--- a/modules/shared.py
+++ b/modules/shared.py
@@ -1,3 +1,4 @@
+import os
 import sys
 
 import gradio as gr
@@ -11,7 +12,7 @@
 
 batch_cond_uncond = True  # old field, unused now in favor of shared.opts.batch_cond_uncond
 parallel_processing_allowed = True
-styles_filename = cmd_opts.styles_file
+styles_filename = cmd_opts.styles_file = cmd_opts.styles_file if len(cmd_opts.styles_file) > 0 else [os.path.join(data_path, 'styles.csv')]
 config_filename = cmd_opts.ui_settings_file
 hide_dirs = {"visible": not cmd_opts.hide_ui_dir_config}
 
@@ -42,7 +43,7 @@
 sd_model: sd_models_types.WebuiSdModel = None
 
 settings_components = None
-"""assinged from ui.py, a mapping on setting names to gradio components repsponsible for those settings"""
+"""assigned from ui.py, a mapping on setting names to gradio components repsponsible for those settings"""
 
 tab_names = []
 
diff --git a/modules/shared_gradio_themes.py b/modules/shared_gradio_themes.py
index 822db0a951d..b6dc31450bc 100644
--- a/modules/shared_gradio_themes.py
+++ b/modules/shared_gradio_themes.py
@@ -65,3 +65,7 @@ def reload_gradio_theme(theme_name=None):
         except Exception as e:
             errors.display(e, "changing gradio theme")
             shared.gradio_theme = gr.themes.Default(**default_theme_args)
+
+    # append additional values gradio_theme
+    shared.gradio_theme.sd_webui_modal_lightbox_toolbar_opacity = shared.opts.sd_webui_modal_lightbox_toolbar_opacity
+    shared.gradio_theme.sd_webui_modal_lightbox_icon_opacity = shared.opts.sd_webui_modal_lightbox_icon_opacity
diff --git a/modules/shared_init.py b/modules/shared_init.py
index d3fb687e0cd..935e3a21cf2 100644
--- a/modules/shared_init.py
+++ b/modules/shared_init.py
@@ -18,8 +18,10 @@ def initialize():
     shared.options_templates = shared_options.options_templates
     shared.opts = options.Options(shared_options.options_templates, shared_options.restricted_opts)
     shared.restricted_opts = shared_options.restricted_opts
-    if os.path.exists(shared.config_filename):
+    try:
         shared.opts.load(shared.config_filename)
+    except FileNotFoundError:
+        pass
 
     from modules import devices
     devices.device, devices.device_interrogate, devices.device_gfpgan, devices.device_esrgan, devices.device_codeformer = \
@@ -27,6 +29,7 @@ def initialize():
 
     devices.dtype = torch.float32 if cmd_opts.no_half else torch.float16
     devices.dtype_vae = torch.float32 if cmd_opts.no_half or cmd_opts.no_half_vae else torch.float16
+    devices.dtype_inference = torch.float32 if cmd_opts.precision == 'full' else devices.dtype
 
     shared.device = devices.device
     shared.weight_load_location = None if cmd_opts.lowram else "cpu"
diff --git a/modules/shared_items.py b/modules/shared_items.py
index 991971ad0fb..88f636452c7 100644
--- a/modules/shared_items.py
+++ b/modules/shared_items.py
@@ -8,6 +8,11 @@ def realesrgan_models_names():
     return [x.name for x in modules.realesrgan_model.get_realesrgan_models(None)]
 
 
+def dat_models_names():
+    import modules.dat_model
+    return [x.name for x in modules.dat_model.get_dat_models(None)]
+
+
 def postprocessing_scripts():
     import modules.scripts
 
@@ -67,14 +72,14 @@ def reload_hypernetworks():
 
 
 def get_infotext_names():
-    from modules import generation_parameters_copypaste, shared
+    from modules import infotext_utils, shared
     res = {}
 
     for info in shared.opts.data_labels.values():
         if info.infotext:
             res[info.infotext] = 1
 
-    for tab_data in generation_parameters_copypaste.paste_fields.values():
+    for tab_data in infotext_utils.paste_fields.values():
         for _, name in tab_data.get("fields") or []:
             if isinstance(name, str):
                 res[name] = 1
diff --git a/modules/shared_options.py b/modules/shared_options.py
index d2e86ff10b3..536766dbe62 100644
--- a/modules/shared_options.py
+++ b/modules/shared_options.py
@@ -1,7 +1,8 @@
+import os
 import gradio as gr
 
-from modules import localization, ui_components, shared_items, shared, interrogate, shared_gradio_themes
-from modules.paths_internal import models_path, script_path, data_path, sd_configs_path, sd_default_config, sd_model_file, default_sd_model_file, extensions_dir, extensions_builtin_dir  # noqa: F401
+from modules import localization, ui_components, shared_items, shared, interrogate, shared_gradio_themes, util, sd_emphasis
+from modules.paths_internal import models_path, script_path, data_path, sd_configs_path, sd_default_config, sd_model_file, default_sd_model_file, extensions_dir, extensions_builtin_dir, default_output_dir  # noqa: F401
 from modules.shared_cmd_options import cmd_opts
 from modules.options import options_section, OptionInfo, OptionHTML, categories
 
@@ -74,14 +75,14 @@
 
 options_templates.update(options_section(('saving-paths', "Paths for saving", "saving"), {
     "outdir_samples": OptionInfo("", "Output directory for images; if empty, defaults to three directories below", component_args=hide_dirs),
-    "outdir_txt2img_samples": OptionInfo("outputs/txt2img-images", 'Output directory for txt2img images', component_args=hide_dirs),
-    "outdir_img2img_samples": OptionInfo("outputs/img2img-images", 'Output directory for img2img images', component_args=hide_dirs),
-    "outdir_extras_samples": OptionInfo("outputs/extras-images", 'Output directory for images from extras tab', component_args=hide_dirs),
+    "outdir_txt2img_samples": OptionInfo(util.truncate_path(os.path.join(default_output_dir, 'txt2img-images')), 'Output directory for txt2img images', component_args=hide_dirs),
+    "outdir_img2img_samples": OptionInfo(util.truncate_path(os.path.join(default_output_dir, 'img2img-images')), 'Output directory for img2img images', component_args=hide_dirs),
+    "outdir_extras_samples": OptionInfo(util.truncate_path(os.path.join(default_output_dir, 'extras-images')), 'Output directory for images from extras tab', component_args=hide_dirs),
     "outdir_grids": OptionInfo("", "Output directory for grids; if empty, defaults to two directories below", component_args=hide_dirs),
-    "outdir_txt2img_grids": OptionInfo("outputs/txt2img-grids", 'Output directory for txt2img grids', component_args=hide_dirs),
-    "outdir_img2img_grids": OptionInfo("outputs/img2img-grids", 'Output directory for img2img grids', component_args=hide_dirs),
-    "outdir_save": OptionInfo("log/images", "Directory for saving images using the Save button", component_args=hide_dirs),
-    "outdir_init_images": OptionInfo("outputs/init-images", "Directory for saving init images when using img2img", component_args=hide_dirs),
+    "outdir_txt2img_grids": OptionInfo(util.truncate_path(os.path.join(default_output_dir, 'txt2img-grids')), 'Output directory for txt2img grids', component_args=hide_dirs),
+    "outdir_img2img_grids": OptionInfo(util.truncate_path(os.path.join(default_output_dir, 'img2img-grids')), 'Output directory for img2img grids', component_args=hide_dirs),
+    "outdir_save": OptionInfo(util.truncate_path(os.path.join(data_path, 'log', 'images')), "Directory for saving images using the Save button", component_args=hide_dirs),
+    "outdir_init_images": OptionInfo(util.truncate_path(os.path.join(default_output_dir, 'init-images')), "Directory for saving init images when using img2img", component_args=hide_dirs),
 }))
 
 options_templates.update(options_section(('saving-to-dirs', "Saving to a directory", "saving"), {
@@ -96,6 +97,9 @@
     "ESRGAN_tile": OptionInfo(192, "Tile size for ESRGAN upscalers.", gr.Slider, {"minimum": 0, "maximum": 512, "step": 16}).info("0 = no tiling"),
     "ESRGAN_tile_overlap": OptionInfo(8, "Tile overlap for ESRGAN upscalers.", gr.Slider, {"minimum": 0, "maximum": 48, "step": 1}).info("Low values = visible seam"),
     "realesrgan_enabled_models": OptionInfo(["R-ESRGAN 4x+", "R-ESRGAN 4x+ Anime6B"], "Select which Real-ESRGAN models to show in the web UI.", gr.CheckboxGroup, lambda: {"choices": shared_items.realesrgan_models_names()}),
+    "dat_enabled_models": OptionInfo(["DAT x2", "DAT x3", "DAT x4"], "Select which DAT models to show in the web UI.", gr.CheckboxGroup, lambda: {"choices": shared_items.dat_models_names()}),
+    "DAT_tile": OptionInfo(192, "Tile size for DAT upscalers.", gr.Slider, {"minimum": 0, "maximum": 512, "step": 16}).info("0 = no tiling"),
+    "DAT_tile_overlap": OptionInfo(8, "Tile overlap for DAT upscalers.", gr.Slider, {"minimum": 0, "maximum": 48, "step": 1}).info("Low values = visible seam"),
     "upscaler_for_img2img": OptionInfo(None, "Upscaler for img2img", gr.Dropdown, lambda: {"choices": [x.name for x in shared.sd_upscalers]}),
 }))
 
@@ -114,6 +118,7 @@
     "memmon_poll_rate": OptionInfo(8, "VRAM usage polls per second during generation.", gr.Slider, {"minimum": 0, "maximum": 40, "step": 1}).info("0 = disable"),
     "samples_log_stdout": OptionInfo(False, "Always print all generation info to standard output"),
     "multiple_tqdm": OptionInfo(True, "Add a second progress bar to the console that shows progress for an entire job."),
+    "enable_upscale_progressbar": OptionInfo(True, "Show a progress bar in the console for tiled upscaling."),
     "print_hypernet_extra": OptionInfo(False, "Print extra hypernetwork information to console."),
     "list_hidden_files": OptionInfo(True, "Load models/files in hidden directories").info("directory is hidden if its name starts with \".\""),
     "disable_mmap_load_safetensors": OptionInfo(False, "Disable memmapping for loading .safetensors files.").info("fixes very slow loading speed in some cases"),
@@ -149,7 +154,7 @@
     "sd_checkpoint_cache": OptionInfo(0, "Checkpoints to cache in RAM", gr.Slider, {"minimum": 0, "maximum": 10, "step": 1}).info("obsolete; set to 0 and use the two settings above instead"),
     "sd_unet": OptionInfo("Automatic", "SD Unet", gr.Dropdown, lambda: {"choices": shared_items.sd_unet_items()}, refresh=shared_items.refresh_unet_list).info("choose Unet model: Automatic = use one with same filename as checkpoint; None = use Unet from checkpoint"),
     "enable_quantization": OptionInfo(False, "Enable quantization in K samplers for sharper and cleaner results. This may change existing seeds").needs_reload_ui(),
-    "enable_emphasis": OptionInfo(True, "Enable emphasis").info("use (text) to make model pay more attention to text and [text] to make it pay less attention"),
+    "emphasis": OptionInfo("Original", "Emphasis mode", gr.Radio, lambda: {"choices": [x.name for x in sd_emphasis.options]}, infotext="Emphasis").info("makes it possible to make model to pay (more:1.1) or (less:0.9) attention to text when you use the syntax in prompt; " + sd_emphasis.get_options_descriptions()),
     "enable_batch_seeds": OptionInfo(True, "Make K-diffusion samplers produce same images in a batch as when making a single image"),
     "comma_padding_backtrack": OptionInfo(20, "Prompt word wrap length limit", gr.Slider, {"minimum": 0, "maximum": 74, "step": 1}).info("in tokens - for texts shorter than specified, if they don't fit into 75 token limit, move them to the next 75 token chunk"),
     "CLIP_stop_at_last_layers": OptionInfo(1, "Clip skip", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}, infotext="Clip skip").link("wiki", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#clip-skip").info("ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer"),
@@ -176,6 +181,7 @@
     "sd_vae_checkpoint_cache": OptionInfo(0, "VAE Checkpoints to cache in RAM", gr.Slider, {"minimum": 0, "maximum": 10, "step": 1}),
     "sd_vae": OptionInfo("Automatic", "SD VAE", gr.Dropdown, lambda: {"choices": shared_items.sd_vae_items()}, refresh=shared_items.refresh_vae_list, infotext='VAE').info("choose VAE model: Automatic = use one with same filename as checkpoint; None = use VAE from checkpoint"),
     "sd_vae_overrides_per_model_preferences": OptionInfo(True, "Selected VAE overrides per-model preferences").info("you can set per-model VAE either by editing user metadata for checkpoints, or by making the VAE have same name as checkpoint"),
+    "auto_vae_precision_bfloat16": OptionInfo(False, "Automatically convert VAE to bfloat16").info("triggers when a tensor with NaNs is produced in VAE; disabling the option in this case will result in a black square image; if enabled, overrides the option below"),
     "auto_vae_precision": OptionInfo(True, "Automatically revert VAE to 32-bit floats").info("triggers when a tensor with NaNs is produced in VAE; disabling the option in this case will result in a black square image"),
     "sd_vae_encode_method": OptionInfo("Full", "VAE type for encode", gr.Radio, {"choices": ["Full", "TAESD"]}, infotext='VAE Encoder').info("method to encode image to latent (use in img2img, hires-fix or inpaint mask)"),
     "sd_vae_decode_method": OptionInfo("Full", "VAE type for decode", gr.Radio, {"choices": ["Full", "TAESD"]}, infotext='VAE Decoder').info("method to decode latent to image"),
@@ -195,6 +201,7 @@
     "return_mask": OptionInfo(False, "For inpainting, include the greyscale mask in results for web"),
     "return_mask_composite": OptionInfo(False, "For inpainting, include masked composite in results for web"),
     "img2img_batch_show_results_limit": OptionInfo(32, "Show the first N batch img2img results in UI", gr.Slider, {"minimum": -1, "maximum": 1000, "step": 1}).info('0: disable, -1: show all images. Too many images can cause lag'),
+    "overlay_inpaint": OptionInfo(True, "Overlay original for inpaint").info("when inpainting, overlay the original image over the areas that weren't inpainted."),
 }))
 
 options_templates.update(options_section(('optimizations', "Optimizations", "sd"), {
@@ -203,12 +210,16 @@
     "token_merging_ratio": OptionInfo(0.0, "Token merging ratio", gr.Slider, {"minimum": 0.0, "maximum": 0.9, "step": 0.1}, infotext='Token merging ratio').link("PR", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/9256").info("0=disable, higher=faster"),
     "token_merging_ratio_img2img": OptionInfo(0.0, "Token merging ratio for img2img", gr.Slider, {"minimum": 0.0, "maximum": 0.9, "step": 0.1}).info("only applies if non-zero and overrides above"),
     "token_merging_ratio_hr": OptionInfo(0.0, "Token merging ratio for high-res pass", gr.Slider, {"minimum": 0.0, "maximum": 0.9, "step": 0.1}, infotext='Token merging ratio hr').info("only applies if non-zero and overrides above"),
-    "pad_cond_uncond": OptionInfo(False, "Pad prompt/negative prompt to be same length", infotext='Pad conds').info("improves performance when prompt and negative prompt have different lengths; changes seeds"),
+    "pad_cond_uncond": OptionInfo(False, "Pad prompt/negative prompt", infotext='Pad conds').info("improves performance when prompt and negative prompt have different lengths; changes seeds"),
+    "pad_cond_uncond_v0": OptionInfo(False, "Pad prompt/negative prompt (v0)", infotext='Pad conds v0').info("alternative implementation for the above; used prior to 1.6.0 for DDIM sampler; overrides the above if set; WARNING: truncates negative prompt if it's too long; changes seeds"),
     "persistent_cond_cache": OptionInfo(True, "Persistent cond cache").info("do not recalculate conds from prompts if prompts have not changed since previous calculation"),
-    "batch_cond_uncond": OptionInfo(True, "Batch cond/uncond").info("do both conditional and unconditional denoising in one batch; uses a bit more VRAM during sampling, but improves speed; previously this was controlled by --always-batch-cond-uncond comandline argument"),
+    "batch_cond_uncond": OptionInfo(True, "Batch cond/uncond").info("do both conditional and unconditional denoising in one batch; uses a bit more VRAM during sampling, but improves speed; previously this was controlled by --always-batch-cond-uncond commandline argument"),
+    "fp8_storage": OptionInfo("Disable", "FP8 weight", gr.Radio, {"choices": ["Disable", "Enable for SDXL", "Enable"]}).info("Use FP8 to store Linear/Conv layers' weight. Require pytorch>=2.1.0."),
+    "cache_fp16_weight": OptionInfo(False, "Cache FP16 weight for LoRA").info("Cache fp16 weight when enabling FP8, will increase the quality of LoRA. Use more system ram."),
 }))
 
 options_templates.update(options_section(('compatibility', "Compatibility", "sd"), {
+    "auto_backcompat": OptionInfo(True, "Automatic backward compatibility").info("automatically enable options for backwards compatibility when importing generation parameters from infotext that has program version."),
     "use_old_emphasis_implementation": OptionInfo(False, "Use old emphasis implementation. Can be useful to reproduce old seeds."),
     "use_old_karras_scheduler_sigmas": OptionInfo(False, "Use old karras scheduler sigmas (0.1 to 10)."),
     "no_dpmpp_sde_batch_determinism": OptionInfo(False, "Do not make DPM++ SDE deterministic across different batch sizes."),
@@ -216,6 +227,8 @@
     "dont_fix_second_order_samplers_schedule": OptionInfo(False, "Do not fix prompt schedule for second order samplers."),
     "hires_fix_use_firstpass_conds": OptionInfo(False, "For hires fix, calculate conds of second pass using extra networks of first pass."),
     "use_old_scheduling": OptionInfo(False, "Use old prompt editing timelines.", infotext="Old prompt editing timelines").info("For [red:green:N]; old: If N < 1, it's a fraction of steps (and hires fix uses range from 0 to 1), if N >= 1, it's an absolute number of steps; new: If N has a decimal point in it, it's a fraction of steps (and hires fix uses range from 1 to 2), othewrwise it's an absolute number of steps"),
+    "use_downcasted_alpha_bar": OptionInfo(False, "Downcast model alphas_cumprod to fp16 before sampling. For reproducing old seeds.", infotext="Downcast alphas_cumprod"),
+    "refiner_switch_by_sample_steps": OptionInfo(False, "Switch to refiner by sampling steps instead of model timesteps. Old behavior for refiner.", infotext="Refiner switch by sampling steps")
 }))
 
 options_templates.update(options_section(('interrogate', "Interrogate"), {
@@ -242,8 +255,11 @@
     "extra_networks_card_height": OptionInfo(0, "Card height for Extra Networks").info("in pixels"),
     "extra_networks_card_text_scale": OptionInfo(1.0, "Card text scale", gr.Slider, {"minimum": 0.0, "maximum": 2.0, "step": 0.01}).info("1 = original size"),
     "extra_networks_card_show_desc": OptionInfo(True, "Show description on card"),
+    "extra_networks_card_description_is_html": OptionInfo(False, "Treat card description as HTML"),
     "extra_networks_card_order_field": OptionInfo("Path", "Default order field for Extra Networks cards", gr.Dropdown, {"choices": ['Path', 'Name', 'Date Created', 'Date Modified']}).needs_reload_ui(),
     "extra_networks_card_order": OptionInfo("Ascending", "Default order for Extra Networks cards", gr.Dropdown, {"choices": ['Ascending', 'Descending']}).needs_reload_ui(),
+    "extra_networks_tree_view_default_enabled": OptionInfo(False, "Enables the Extra Networks directory tree view by default").needs_reload_ui(),
+    "extra_networks_tree_view_default_width": OptionInfo(180, "Default width for the Extra Networks directory tree view", gr.Number).needs_reload_ui(),
     "extra_networks_add_text_separator": OptionInfo(" ", "Extra networks separator").info("extra text to add before <...> when adding extra network to prompt"),
     "ui_extra_networks_tab_reorder": OptionInfo("", "Extra networks tab order").needs_reload_ui(),
     "textual_inversion_print_at_load": OptionInfo(False, "Print a list of Textual Inversion embeddings when loading model"),
@@ -257,7 +273,8 @@
     "keyedit_delimiters": OptionInfo(r".,\/!?%^*;:{}=`~() ", "Word delimiters when editing the prompt with Ctrl+up/down"),
     "keyedit_delimiters_whitespace": OptionInfo(["Tab", "Carriage Return", "Line Feed"], "Ctrl+up/down whitespace delimiters", gr.CheckboxGroup, lambda: {"choices": ["Tab", "Carriage Return", "Line Feed"]}),
     "keyedit_move": OptionInfo(True, "Alt+left/right moves prompt elements"),
-    "disable_token_counters": OptionInfo(False, "Disable prompt token counters").needs_reload_ui(),
+    "disable_token_counters": OptionInfo(False, "Disable prompt token counters"),
+    "include_styles_into_token_counters": OptionInfo(True, "Count tokens of enabled styles").info("When calculating how many tokens the prompt has, also consider tokens added by enabled styles."),
 }))
 
 options_templates.update(options_section(('ui_gallery', "Gallery", "ui"), {
@@ -267,7 +284,10 @@
     "js_modal_lightbox_initially_zoomed": OptionInfo(True, "Full page image viewer: show images zoomed in by default"),
     "js_modal_lightbox_gamepad": OptionInfo(False, "Full page image viewer: navigate with gamepad"),
     "js_modal_lightbox_gamepad_repeat": OptionInfo(250, "Full page image viewer: gamepad repeat period").info("in milliseconds"),
+    "sd_webui_modal_lightbox_icon_opacity": OptionInfo(1, "Full page image viewer: control icon unfocused opacity", gr.Slider, {"minimum": 0.0, "maximum": 1, "step": 0.01}, onchange=shared.reload_gradio_theme).info('for mouse only').needs_reload_ui(),
+    "sd_webui_modal_lightbox_toolbar_opacity": OptionInfo(0.9, "Full page image viewer: tool bar opacity", gr.Slider, {"minimum": 0.0, "maximum": 1, "step": 0.01}, onchange=shared.reload_gradio_theme).info('for mouse only').needs_reload_ui(),
     "gallery_height": OptionInfo("", "Gallery height", gr.Textbox).info("can be any valid CSS value, for example 768px or 20em").needs_reload_ui(),
+    "open_dir_button_choice": OptionInfo("Subdirectory", "What directory the [📂] button opens", gr.Radio, {"choices": ["Output Root", "Subdirectory", "Subdirectory (even temp dir)"]}),
 }))
 
 options_templates.update(options_section(('ui_alternatives', "UI alternatives", "ui"), {
@@ -279,6 +299,7 @@
     "hires_fix_show_prompts": OptionInfo(False, "Hires fix: show hires prompt and negative prompt").needs_reload_ui(),
     "txt2img_settings_accordion": OptionInfo(False, "Settings in txt2img hidden under Accordion").needs_reload_ui(),
     "img2img_settings_accordion": OptionInfo(False, "Settings in img2img hidden under Accordion").needs_reload_ui(),
+    "interrupt_after_current": OptionInfo(True, "Don't Interrupt in the middle").info("when using Interrupt button, if generating more than one image, stop after the generation of an image has finished, instead of immediately"),
 }))
 
 options_templates.update(options_section(('ui', "User interface", "ui"), {
@@ -349,11 +370,12 @@
     'rho':  OptionInfo(0.0, "rho", gr.Number, infotext='Schedule rho').info("0 = default (7 for karras, 1 for polyexponential); higher values result in a steeper noise schedule (decreases faster)"),
     'eta_noise_seed_delta': OptionInfo(0, "Eta noise seed delta", gr.Number, {"precision": 0}, infotext='ENSD').info("ENSD; does not improve anything, just produces different results for ancestral samplers - only useful for reproducing images"),
     'always_discard_next_to_last_sigma': OptionInfo(False, "Always discard next-to-last sigma", infotext='Discard penultimate sigma').link("PR", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/6044"),
-    'sgm_noise_multiplier': OptionInfo(False, "SGM noise multiplier", infotext='SGM noise multplier').link("PR", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/12818").info("Match initial noise to official SDXL implementation - only useful for reproducing images"),
+    'sgm_noise_multiplier': OptionInfo(False, "SGM noise multiplier", infotext='SGM noise multiplier').link("PR", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/12818").info("Match initial noise to official SDXL implementation - only useful for reproducing images"),
     'uni_pc_variant': OptionInfo("bh1", "UniPC variant", gr.Radio, {"choices": ["bh1", "bh2", "vary_coeff"]}, infotext='UniPC variant'),
     'uni_pc_skip_type': OptionInfo("time_uniform", "UniPC skip type", gr.Radio, {"choices": ["time_uniform", "time_quadratic", "logSNR"]}, infotext='UniPC skip type'),
     'uni_pc_order': OptionInfo(3, "UniPC order", gr.Slider, {"minimum": 1, "maximum": 50, "step": 1}, infotext='UniPC order').info("must be < sampling steps"),
     'uni_pc_lower_order_final': OptionInfo(True, "UniPC lower order final", infotext='UniPC lower order final'),
+    'sd_noise_schedule': OptionInfo("Default", "Noise schedule for sampling", gr.Radio, {"choices": ["Default", "Zero Terminal SNR"]}, infotext="Noise Schedule").info("for use with zero terminal SNR trained models")
 }))
 
 options_templates.update(options_section(('postprocessing', "Postprocessing", "postprocessing"), {
diff --git a/modules/shared_state.py b/modules/shared_state.py
index a68789cc815..f74eafc5895 100644
--- a/modules/shared_state.py
+++ b/modules/shared_state.py
@@ -12,6 +12,7 @@
 class State:
     skipped = False
     interrupted = False
+    stopping_generation = False
     job = ""
     job_no = 0
     job_count = 0
@@ -79,6 +80,10 @@ def interrupt(self):
         self.interrupted = True
         log.info("Received interrupt request")
 
+    def stop_generating(self):
+        self.stopping_generation = True
+        log.info("Received stop generating request")
+
     def nextjob(self):
         if shared.opts.live_previews_enable and shared.opts.show_progress_every_n_steps == -1:
             self.do_set_current_image()
@@ -91,6 +96,7 @@ def dict(self):
         obj = {
             "skipped": self.skipped,
             "interrupted": self.interrupted,
+            "stopping_generation": self.stopping_generation,
             "job": self.job,
             "job_count": self.job_count,
             "job_timestamp": self.job_timestamp,
@@ -114,6 +120,7 @@ def begin(self, job: str = "(unknown)"):
         self.id_live_preview = 0
         self.skipped = False
         self.interrupted = False
+        self.stopping_generation = False
         self.textinfo = None
         self.job = job
         devices.torch_gc()
@@ -150,10 +157,12 @@ def do_set_current_image(self):
             self.current_image_sampling_step = self.sampling_step
 
         except Exception:
-            # when switching models during genration, VAE would be on CPU, so creating an image will fail.
+            # when switching models during generation, VAE would be on CPU, so creating an image will fail.
             # we silently ignore this error
             errors.record_exception()
 
     def assign_current_image(self, image):
+        if shared.opts.live_previews_image_format == 'jpeg' and image.mode == 'RGBA':
+            image = image.convert('RGB')
         self.current_image = image
         self.id_live_preview += 1
diff --git a/modules/styles.py b/modules/styles.py
index 81d9800d184..a9d8636a98d 100644
--- a/modules/styles.py
+++ b/modules/styles.py
@@ -1,16 +1,16 @@
+from pathlib import Path
+from modules import errors
 import csv
-import fnmatch
 import os
-import os.path
 import typing
 import shutil
 
 
 class PromptStyle(typing.NamedTuple):
     name: str
-    prompt: str
-    negative_prompt: str
-    path: str = None
+    prompt: str | None
+    negative_prompt: str | None
+    path: str | None = None
 
 
 def merge_prompts(style_prompt: str, prompt: str) -> str:
@@ -30,38 +30,29 @@ def apply_styles_to_prompt(prompt, styles):
     return prompt
 
 
-def unwrap_style_text_from_prompt(style_text, prompt):
-    """
-    Checks the prompt to see if the style text is wrapped around it. If so,
-    returns True plus the prompt text without the style text. Otherwise, returns
-    False with the original prompt.
+def extract_style_text_from_prompt(style_text, prompt):
+    """This function extracts the text from a given prompt based on a provided style text. It checks if the style text contains the placeholder {prompt} or if it appears at the end of the prompt. If a match is found, it returns True along with the extracted text. Otherwise, it returns False and the original prompt.
 
-    Note that the "cleaned" version of the style text is only used for matching
-    purposes here. It isn't returned; the original style text is not modified.
+    extract_style_text_from_prompt("masterpiece", "1girl, art by greg, masterpiece") outputs (True, "1girl, art by greg")
+    extract_style_text_from_prompt("masterpiece, {prompt}", "masterpiece, 1girl, art by greg") outputs (True, "1girl, art by greg")
+    extract_style_text_from_prompt("masterpiece, {prompt}", "exquisite, 1girl, art by greg") outputs (False, "exquisite, 1girl, art by greg")
     """
-    stripped_prompt = prompt
-    stripped_style_text = style_text
+
+    stripped_prompt = prompt.strip()
+    stripped_style_text = style_text.strip()
+
     if "{prompt}" in stripped_style_text:
-        # Work out whether the prompt is wrapped in the style text. If so, we
-        # return True and the "inner" prompt text that isn't part of the style.
-        try:
-            left, right = stripped_style_text.split("{prompt}", 2)
-        except ValueError as e:
-            # If the style text has multple "{prompt}"s, we can't split it into
-            # two parts. This is an error, but we can't do anything about it.
-            print(f"Unable to compare style text to prompt:\n{style_text}")
-            print(f"Error: {e}")
-            return False, prompt
+        left, _, right = stripped_style_text.partition("{prompt}")
         if stripped_prompt.startswith(left) and stripped_prompt.endswith(right):
-            prompt = stripped_prompt[len(left) : len(stripped_prompt) - len(right)]
+            prompt = stripped_prompt[len(left):len(stripped_prompt)-len(right)]
             return True, prompt
     else:
-        # Work out whether the given prompt ends with the style text. If so, we
-        # return True and the prompt text up to where the style text starts.
         if stripped_prompt.endswith(stripped_style_text):
-            prompt = stripped_prompt[: len(stripped_prompt) - len(stripped_style_text)]
-            if prompt.endswith(", "):
+            prompt = stripped_prompt[:len(stripped_prompt)-len(stripped_style_text)]
+
+            if prompt.endswith(', '):
                 prompt = prompt[:-2]
+
             return True, prompt
 
     return False, prompt
@@ -76,15 +67,11 @@ def extract_original_prompts(style: PromptStyle, prompt, negative_prompt):
     if not style.prompt and not style.negative_prompt:
         return False, prompt, negative_prompt
 
-    match_positive, extracted_positive = unwrap_style_text_from_prompt(
-        style.prompt, prompt
-    )
+    match_positive, extracted_positive = extract_style_text_from_prompt(style.prompt, prompt)
     if not match_positive:
         return False, prompt, negative_prompt
 
-    match_negative, extracted_negative = unwrap_style_text_from_prompt(
-        style.negative_prompt, negative_prompt
-    )
+    match_negative, extracted_negative = extract_style_text_from_prompt(style.negative_prompt, negative_prompt)
     if not match_negative:
         return False, prompt, negative_prompt
 
@@ -92,14 +79,19 @@ def extract_original_prompts(style: PromptStyle, prompt, negative_prompt):
 
 
 class StyleDatabase:
-    def __init__(self, path: str):
+    def __init__(self, paths: list[str | Path]):
         self.no_style = PromptStyle("None", "", "", None)
         self.styles = {}
-        self.path = path
-
-        folder, file = os.path.split(self.path)
-        filename, _, ext = file.partition('*')
-        self.default_path = os.path.join(folder, filename + ext)
+        self.paths = paths
+        self.all_styles_files: list[Path] = []
+
+        folder, file = os.path.split(self.paths[0])
+        if '*' in file or '?' in file:
+            # if the first path is a wildcard pattern, find the first match else use "folder/styles.csv" as the default path
+            self.default_path = next(Path(folder).glob(file), Path(os.path.join(folder, 'styles.csv')))
+            self.paths.insert(0, self.default_path)
+        else:
+            self.default_path = Path(self.paths[0])
 
         self.prompt_fields = [field for field in PromptStyle._fields if field != "path"]
 
@@ -112,57 +104,58 @@ def reload(self):
         """
         self.styles.clear()
 
-        path, filename = os.path.split(self.path)
-
-        if "*" in filename:
-            fileglob = filename.split("*")[0] + "*.csv"
-            filelist = []
-            for file in os.listdir(path):
-                if fnmatch.fnmatch(file, fileglob):
-                    filelist.append(file)
-                    # Add a visible divider to the style list
-                    half_len = round(len(file) / 2)
-                    divider = f"{'-' * (20 - half_len)} {file.upper()}"
-                    divider = f"{divider} {'-' * (40 - len(divider))}"
-                    self.styles[divider] = PromptStyle(
-                        f"{divider}", None, None, "do_not_save"
+        # scans for all styles files
+        all_styles_files = []
+        for pattern in self.paths:
+            folder, file = os.path.split(pattern)
+            if '*' in file or '?' in file:
+                found_files = Path(folder).glob(file)
+                [all_styles_files.append(file) for file in found_files]
+            else:
+                # if os.path.exists(pattern):
+                all_styles_files.append(Path(pattern))
+
+        # Remove any duplicate entries
+        seen = set()
+        self.all_styles_files = [s for s in all_styles_files if not (s in seen or seen.add(s))]
+
+        for styles_file in self.all_styles_files:
+            if len(all_styles_files) > 1:
+                # add divider when more than styles file
+                # '---------------- STYLES ----------------'
+                divider = f' {styles_file.stem.upper()} '.center(40, '-')
+                self.styles[divider] = PromptStyle(f"{divider}", None, None, "do_not_save")
+            if styles_file.is_file():
+                self.load_from_csv(styles_file)
+
+    def load_from_csv(self, path: str | Path):
+        try:
+            with open(path, "r", encoding="utf-8-sig", newline="") as file:
+                reader = csv.DictReader(file, skipinitialspace=True)
+                for row in reader:
+                    # Ignore empty rows or rows starting with a comment
+                    if not row or row["name"].startswith("#"):
+                        continue
+                    # Support loading old CSV format with "name, text"-columns
+                    prompt = row["prompt"] if "prompt" in row else row["text"]
+                    negative_prompt = row.get("negative_prompt", "")
+                    # Add style to database
+                    self.styles[row["name"]] = PromptStyle(
+                        row["name"], prompt, negative_prompt, str(path)
                     )
-                    # Add styles from this CSV file
-                    self.load_from_csv(os.path.join(path, file))
-            if len(filelist) == 0:
-                print(f"No styles found in {path} matching {fileglob}")
-                return
-        elif not os.path.exists(self.path):
-            print(f"Style database not found: {self.path}")
-            return
-        else:
-            self.load_from_csv(self.path)
-
-    def load_from_csv(self, path: str):
-        with open(path, "r", encoding="utf-8-sig", newline="") as file:
-            reader = csv.DictReader(file, skipinitialspace=True)
-            for row in reader:
-                # Ignore empty rows or rows starting with a comment
-                if not row or row["name"].startswith("#"):
-                    continue
-                # Support loading old CSV format with "name, text"-columns
-                prompt = row["prompt"] if "prompt" in row else row["text"]
-                negative_prompt = row.get("negative_prompt", "")
-                # Add style to database
-                self.styles[row["name"]] = PromptStyle(
-                    row["name"], prompt, negative_prompt, path
-                )
+        except Exception:
+            errors.report(f'Error loading styles from {path}: ', exc_info=True)
 
     def get_style_paths(self) -> set:
         """Returns a set of all distinct paths of files that styles are loaded from."""
         # Update any styles without a path to the default path
         for style in list(self.styles.values()):
             if not style.path:
-                self.styles[style.name] = style._replace(path=self.default_path)
+                self.styles[style.name] = style._replace(path=str(self.default_path))
 
         # Create a list of all distinct paths, including the default path
         style_paths = set()
-        style_paths.add(self.default_path)
+        style_paths.add(str(self.default_path))
         for _, style in self.styles.items():
             if style.path:
                 style_paths.add(style.path)
@@ -190,7 +183,6 @@ def apply_negative_styles_to_prompt(self, prompt, styles):
 
     def save_styles(self, path: str = None) -> None:
         # The path argument is deprecated, but kept for backwards compatibility
-        _ = path
 
         style_paths = self.get_style_paths()
 
diff --git a/modules/sysinfo.py b/modules/sysinfo.py
index b669edd0cfd..f336251e445 100644
--- a/modules/sysinfo.py
+++ b/modules/sysinfo.py
@@ -24,13 +24,13 @@
     "XFORMERS_PACKAGE",
     "CLIP_PACKAGE",
     "OPENCLIP_PACKAGE",
+    "ASSETS_REPO",
     "STABLE_DIFFUSION_REPO",
     "K_DIFFUSION_REPO",
-    "CODEFORMER_REPO",
     "BLIP_REPO",
+    "ASSETS_COMMIT_HASH",
     "STABLE_DIFFUSION_COMMIT_HASH",
     "K_DIFFUSION_COMMIT_HASH",
-    "CODEFORMER_COMMIT_HASH",
     "BLIP_COMMIT_HASH",
     "COMMANDLINE_ARGS",
     "IGNORE_CMD_ARGS_ERRORS",
diff --git a/modules/textual_inversion/autocrop.py b/modules/textual_inversion/autocrop.py
index e223a2e0cc9..ca858ef4c4a 100644
--- a/modules/textual_inversion/autocrop.py
+++ b/modules/textual_inversion/autocrop.py
@@ -65,7 +65,7 @@ def crop_image(im, settings):
         rect[3] -= 1
         d.rectangle(rect, outline=GREEN)
         results.append(im_debug)
-        if settings.destop_view_image:
+        if settings.desktop_view_image:
             im_debug.show()
 
     return results
@@ -341,5 +341,5 @@ def __init__(self, crop_width=512, crop_height=512, corner_points_weight=0.5, en
         self.entropy_points_weight = entropy_points_weight
         self.face_points_weight = face_points_weight
         self.annotate_image = annotate_image
-        self.destop_view_image = False
+        self.desktop_view_image = False
         self.dnn_model_path = dnn_model_path
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 7ee05061545..71c032df76d 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -2,7 +2,6 @@
 import numpy as np
 import PIL
 import torch
-from PIL import Image
 from torch.utils.data import Dataset, DataLoader, Sampler
 from torchvision import transforms
 from collections import defaultdict
@@ -10,7 +9,7 @@
 
 import random
 import tqdm
-from modules import devices, shared
+from modules import devices, shared, images
 import re
 
 from ldm.modules.distributions.distributions import DiagonalGaussianDistribution
@@ -61,7 +60,7 @@ def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_to
             if shared.state.interrupted:
                 raise Exception("interrupted")
             try:
-                image = Image.open(path)
+                image = images.read(path)
                 #Currently does not work for single color transparency
                 #We would need to read image.info['transparency'] for that
                 if use_weight and 'A' in image.getbands():
diff --git a/modules/textual_inversion/image_embedding.py b/modules/textual_inversion/image_embedding.py
index 81cff7bf17c..ea4b88333ac 100644
--- a/modules/textual_inversion/image_embedding.py
+++ b/modules/textual_inversion/image_embedding.py
@@ -193,11 +193,11 @@ def caption_image_overlay(srcimage, title, footerLeft, footerMid, footerRight, t
 
     embedded_image = insert_image_data_embed(cap_image, test_embed)
 
-    retrived_embed = extract_image_data_embed(embedded_image)
+    retrieved_embed = extract_image_data_embed(embedded_image)
 
-    assert str(retrived_embed) == str(test_embed)
+    assert str(retrieved_embed) == str(test_embed)
 
-    embedded_image2 = insert_image_data_embed(cap_image, retrived_embed)
+    embedded_image2 = insert_image_data_embed(cap_image, retrieved_embed)
 
     assert embedded_image == embedded_image2
 
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index 04dda585cc9..c206ef5fd01 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -11,7 +11,6 @@
 
 import numpy as np
 from PIL import Image, PngImagePlugin
-from torch.utils.tensorboard import SummaryWriter
 
 from modules import shared, devices, sd_hijack, sd_models, images, sd_samplers, sd_hijack_checkpoint, errors, hashes
 import modules.textual_inversion.dataset
@@ -151,6 +150,7 @@ def register_embedding_by_name(self, embedding, model, name):
         return embedding
 
     def get_expected_shape(self):
+        devices.torch_npu_set_device()
         vec = shared.sd_model.cond_stage_model.encode_embedding_init_text(",", 1)
         return vec.shape[1]
 
@@ -172,7 +172,7 @@ def load_from_file(self, path, filename):
                 if data:
                     name = data.get('name', name)
                 else:
-                    # if data is None, means this is not an embeding, just a preview image
+                    # if data is None, means this is not an embedding, just a preview image
                     return
         elif ext in ['.BIN', '.PT']:
             data = torch.load(path, map_location="cpu")
@@ -344,6 +344,7 @@ def write_loss(log_directory, filename, step, epoch_len, values):
         })
 
 def tensorboard_setup(log_directory):
+    from torch.utils.tensorboard import SummaryWriter
     os.makedirs(os.path.join(log_directory, "tensorboard"), exist_ok=True)
     return SummaryWriter(
             log_dir=os.path.join(log_directory, "tensorboard"),
@@ -448,8 +449,12 @@ def train_embedding(id_task, embedding_name, learn_rate, batch_size, gradient_st
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     old_parallel_processing_allowed = shared.parallel_processing_allowed
 
+    tensorboard_writer = None
     if shared.opts.training_enable_tensorboard:
-        tensorboard_writer = tensorboard_setup(log_directory)
+        try:
+            tensorboard_writer = tensorboard_setup(log_directory)
+        except ImportError:
+            errors.report("Error initializing tensorboard", exc_info=True)
 
     pin_memory = shared.opts.pin_memory
 
@@ -622,7 +627,7 @@ def train_embedding(id_task, embedding_name, learn_rate, batch_size, gradient_st
                         last_saved_image, last_text_info = images.save_image(image, images_dir, "", p.seed, p.prompt, shared.opts.samples_format, processed.infotexts[0], p=p, forced_filename=forced_filename, save_to_dirs=False)
                         last_saved_image += f", prompt: {preview_text}"
 
-                        if shared.opts.training_enable_tensorboard and shared.opts.training_tensorboard_save_images:
+                        if tensorboard_writer and shared.opts.training_tensorboard_save_images:
                             tensorboard_add_image(tensorboard_writer, f"Validation at epoch {epoch_num}", image, embedding.step)
 
                     if save_image_with_stored_embedding and os.path.exists(last_saved_file) and embedding_yet_to_be_embedded:
diff --git a/modules/torch_utils.py b/modules/torch_utils.py
new file mode 100644
index 00000000000..e5b52393ec8
--- /dev/null
+++ b/modules/torch_utils.py
@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import torch.nn
+
+
+def get_param(model) -> torch.nn.Parameter:
+    """
+    Find the first parameter in a model or module.
+    """
+    if hasattr(model, "model") and hasattr(model.model, "parameters"):
+        # Unpeel a model descriptor to get at the actual Torch module.
+        model = model.model
+
+    for param in model.parameters():
+        return param
+
+    raise ValueError(f"No parameters found in model {model!r}")
diff --git a/modules/txt2img.py b/modules/txt2img.py
index e4e18ceb6dd..fc56b8a86c4 100644
--- a/modules/txt2img.py
+++ b/modules/txt2img.py
@@ -1,17 +1,22 @@
+import json
 from contextlib import closing
 
 import modules.scripts
-from modules import processing
-from modules.generation_parameters_copypaste import create_override_settings_dict
+from modules import processing, infotext_utils
+from modules.infotext_utils import create_override_settings_dict, parse_generation_parameters
 from modules.shared import opts
 import modules.shared as shared
 from modules.ui import plaintext_to_html
+from PIL import Image
 import gradio as gr
 
 
-def txt2img(id_task: str, prompt: str, negative_prompt: str, prompt_styles, steps: int, sampler_name: str, n_iter: int, batch_size: int, cfg_scale: float, height: int, width: int, enable_hr: bool, denoising_strength: float, hr_scale: float, hr_upscaler: str, hr_second_pass_steps: int, hr_resize_x: int, hr_resize_y: int, hr_checkpoint_name: str, hr_sampler_name: str, hr_prompt: str, hr_negative_prompt, override_settings_texts, request: gr.Request, *args):
+def txt2img_create_processing(id_task: str, request: gr.Request, prompt: str, negative_prompt: str, prompt_styles, steps: int, sampler_name: str, n_iter: int, batch_size: int, cfg_scale: float, height: int, width: int, enable_hr: bool, denoising_strength: float, hr_scale: float, hr_upscaler: str, hr_second_pass_steps: int, hr_resize_x: int, hr_resize_y: int, hr_checkpoint_name: str, hr_sampler_name: str, hr_prompt: str, hr_negative_prompt, override_settings_texts, *args, force_enable_hr=False):
     override_settings = create_override_settings_dict(override_settings_texts)
 
+    if force_enable_hr:
+        enable_hr = True
+
     p = processing.StableDiffusionProcessingTxt2Img(
         sd_model=shared.sd_model,
         outpath_samples=opts.outdir_samples or opts.outdir_txt2img_samples,
@@ -27,7 +32,7 @@ def txt2img(id_task: str, prompt: str, negative_prompt: str, prompt_styles, step
         width=width,
         height=height,
         enable_hr=enable_hr,
-        denoising_strength=denoising_strength if enable_hr else None,
+        denoising_strength=denoising_strength,
         hr_scale=hr_scale,
         hr_upscaler=hr_upscaler,
         hr_second_pass_steps=hr_second_pass_steps,
@@ -48,8 +53,58 @@ def txt2img(id_task: str, prompt: str, negative_prompt: str, prompt_styles, step
     if shared.opts.enable_console_prompts:
         print(f"\ntxt2img: {prompt}", file=shared.progress_print_out)
 
+    return p
+
+
+def txt2img_upscale(id_task: str, request: gr.Request, gallery, gallery_index, generation_info, *args):
+    assert len(gallery) > 0, 'No image to upscale'
+    assert 0 <= gallery_index < len(gallery), f'Bad image index: {gallery_index}'
+
+    p = txt2img_create_processing(id_task, request, *args, force_enable_hr=True)
+    p.batch_size = 1
+    p.n_iter = 1
+    # txt2img_upscale attribute that signifies this is called by txt2img_upscale
+    p.txt2img_upscale = True
+
+    geninfo = json.loads(generation_info)
+
+    image_info = gallery[gallery_index] if 0 <= gallery_index < len(gallery) else gallery[0]
+    p.firstpass_image = infotext_utils.image_from_url_text(image_info)
+
+    parameters = parse_generation_parameters(geninfo.get('infotexts')[gallery_index], [])
+    p.seed = parameters.get('Seed', -1)
+    p.subseed = parameters.get('Variation seed', -1)
+
+    p.override_settings['save_images_before_highres_fix'] = False
+
+    with closing(p):
+        processed = modules.scripts.scripts_txt2img.run(p, *p.script_args)
+
+        if processed is None:
+            processed = processing.process_images(p)
+
+    shared.total_tqdm.clear()
+
+    new_gallery = []
+    for i, image in enumerate(gallery):
+        if i == gallery_index:
+            geninfo["infotexts"][gallery_index: gallery_index+1] = processed.infotexts
+            new_gallery.extend(processed.images)
+        else:
+            fake_image = Image.new(mode="RGB", size=(1, 1))
+            fake_image.already_saved_as = image["name"].rsplit('?', 1)[0]
+            new_gallery.append(fake_image)
+
+    geninfo["infotexts"][gallery_index] = processed.info
+
+    return new_gallery, json.dumps(geninfo), plaintext_to_html(processed.info), plaintext_to_html(processed.comments, classname="comments")
+
+
+def txt2img(id_task: str, request: gr.Request, *args):
+    p = txt2img_create_processing(id_task, request, *args)
+
     with closing(p):
-        processed = modules.scripts.scripts_txt2img.run(p, *args)
+        processed = modules.scripts.scripts_txt2img.run(p, *p.script_args)
 
         if processed is None:
             processed = processing.process_images(p)
diff --git a/modules/ui.py b/modules/ui.py
index d80486dd4ac..7b4341627b8 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -13,7 +13,7 @@
 from modules.call_queue import wrap_gradio_gpu_call, wrap_queued_call, wrap_gradio_call
 
 from modules import gradio_extensons  # noqa: F401
-from modules import sd_hijack, sd_models, script_callbacks, ui_extensions, deepbooru, extra_networks, ui_common, ui_postprocessing, progress, ui_loadsave, shared_items, ui_settings, timer, sysinfo, ui_checkpoint_merger, scripts, sd_samplers, processing, ui_extra_networks, ui_toprow
+from modules import sd_hijack, sd_models, script_callbacks, ui_extensions, deepbooru, extra_networks, ui_common, ui_postprocessing, progress, ui_loadsave, shared_items, ui_settings, timer, sysinfo, ui_checkpoint_merger, scripts, sd_samplers, processing, ui_extra_networks, ui_toprow, launch_utils
 from modules.ui_components import FormRow, FormGroup, ToolButton, FormHTML, InputAccordion, ResizeHandleRow
 from modules.paths import script_path
 from modules.ui_common import create_refresh_button
@@ -21,14 +21,14 @@
 
 from modules.shared import opts, cmd_opts
 
-import modules.generation_parameters_copypaste as parameters_copypaste
+import modules.infotext_utils as parameters_copypaste
 import modules.hypernetworks.ui as hypernetworks_ui
 import modules.textual_inversion.ui as textual_inversion_ui
 import modules.textual_inversion.textual_inversion as textual_inversion
 import modules.shared as shared
 from modules import prompt_parser
 from modules.sd_hijack import model_hijack
-from modules.generation_parameters_copypaste import image_from_url_text
+from modules.infotext_utils import image_from_url_text, PasteField
 
 create_setting_component = ui_settings.create_setting_component
 
@@ -151,7 +151,18 @@ def connect_clear_prompt(button):
     )
 
 
-def update_token_counter(text, steps, *, is_positive=True):
+def update_token_counter(text, steps, styles, *, is_positive=True):
+    params = script_callbacks.BeforeTokenCounterParams(text, steps, styles, is_positive=is_positive)
+    script_callbacks.before_token_counter_callback(params)
+    text = params.prompt
+    steps = params.steps
+    styles = params.styles
+    is_positive = params.is_positive
+
+    if shared.opts.include_styles_into_token_counters:
+        apply_styles = shared.prompt_styles.apply_styles_to_prompt if is_positive else shared.prompt_styles.apply_negative_styles_to_prompt
+        text = apply_styles(text, styles)
+
     try:
         text, _ = extra_networks.parse_prompt(text)
 
@@ -173,9 +184,8 @@ def update_token_counter(text, steps, *, is_positive=True):
     return f"{token_count}/{max_length}"
 
 
-def update_negative_prompt_token_counter(text, steps):
-    return update_token_counter(text, steps, is_positive=False)
-
+def update_negative_prompt_token_counter(*args):
+    return update_token_counter(*args, is_positive=False)
 
 
 def setup_progressbar(*args, **kwargs):
@@ -259,6 +269,9 @@ def create_ui():
 
     parameters_copypaste.reset()
 
+    settings = ui_settings.UiSettings()
+    settings.register_settings()
+
     scripts.scripts_current = scripts.scripts_txt2img
     scripts.scripts_txt2img.initialize_scripts(is_img2img=False)
 
@@ -267,7 +280,7 @@ def create_ui():
 
         dummy_component = gr.Label(visible=False)
 
-        extra_tabs = gr.Tabs(elem_id="txt2img_extra_tabs")
+        extra_tabs = gr.Tabs(elem_id="txt2img_extra_tabs", elem_classes=["extra-networks"])
         extra_tabs.__enter__()
 
         with gr.Tab("Generation", id="txt2img_generation") as txt2img_generation_tab, ResizeHandleRow(equal_height=False):
@@ -376,50 +389,60 @@ def create_ui():
                     show_progress=False,
                 )
 
-            txt2img_gallery, generation_info, html_info, html_log = create_output_panel("txt2img", opts.outdir_txt2img_samples, toprow)
+            output_panel = create_output_panel("txt2img", opts.outdir_txt2img_samples, toprow)
+
+            txt2img_inputs = [
+                dummy_component,
+                toprow.prompt,
+                toprow.negative_prompt,
+                toprow.ui_styles.dropdown,
+                steps,
+                sampler_name,
+                batch_count,
+                batch_size,
+                cfg_scale,
+                height,
+                width,
+                enable_hr,
+                denoising_strength,
+                hr_scale,
+                hr_upscaler,
+                hr_second_pass_steps,
+                hr_resize_x,
+                hr_resize_y,
+                hr_checkpoint_name,
+                hr_sampler_name,
+                hr_prompt,
+                hr_negative_prompt,
+                override_settings,
+            ] + custom_inputs
+
+            txt2img_outputs = [
+                output_panel.gallery,
+                output_panel.generation_info,
+                output_panel.infotext,
+                output_panel.html_log,
+            ]
 
             txt2img_args = dict(
                 fn=wrap_gradio_gpu_call(modules.txt2img.txt2img, extra_outputs=[None, '', '']),
                 _js="submit",
-                inputs=[
-                    dummy_component,
-                    toprow.prompt,
-                    toprow.negative_prompt,
-                    toprow.ui_styles.dropdown,
-                    steps,
-                    sampler_name,
-                    batch_count,
-                    batch_size,
-                    cfg_scale,
-                    height,
-                    width,
-                    enable_hr,
-                    denoising_strength,
-                    hr_scale,
-                    hr_upscaler,
-                    hr_second_pass_steps,
-                    hr_resize_x,
-                    hr_resize_y,
-                    hr_checkpoint_name,
-                    hr_sampler_name,
-                    hr_prompt,
-                    hr_negative_prompt,
-                    override_settings,
-
-                ] + custom_inputs,
-
-                outputs=[
-                    txt2img_gallery,
-                    generation_info,
-                    html_info,
-                    html_log,
-                ],
+                inputs=txt2img_inputs,
+                outputs=txt2img_outputs,
                 show_progress=False,
             )
 
             toprow.prompt.submit(**txt2img_args)
             toprow.submit.click(**txt2img_args)
 
+            output_panel.button_upscale.click(
+                fn=wrap_gradio_gpu_call(modules.txt2img.txt2img_upscale, extra_outputs=[None, '', '']),
+                _js="submit_txt2img_upscale",
+                inputs=txt2img_inputs[0:1] + [output_panel.gallery, dummy_component, output_panel.generation_info] + txt2img_inputs[1:],
+                outputs=txt2img_outputs,
+                show_progress=False,
+            )
+
             res_switch_btn.click(fn=None, _js="function(){switchWidthHeight('txt2img')}", inputs=None, outputs=None, show_progress=False)
 
             toprow.restore_progress_button.click(
@@ -427,37 +450,37 @@ def create_ui():
                 _js="restoreProgressTxt2img",
                 inputs=[dummy_component],
                 outputs=[
-                    txt2img_gallery,
-                    generation_info,
-                    html_info,
-                    html_log,
+                    output_panel.gallery,
+                    output_panel.generation_info,
+                    output_panel.infotext,
+                    output_panel.html_log,
                 ],
                 show_progress=False,
             )
 
             txt2img_paste_fields = [
-                (toprow.prompt, "Prompt"),
-                (toprow.negative_prompt, "Negative prompt"),
-                (steps, "Steps"),
-                (sampler_name, "Sampler"),
-                (cfg_scale, "CFG scale"),
-                (width, "Size-1"),
-                (height, "Size-2"),
-                (batch_size, "Batch size"),
-                (toprow.ui_styles.dropdown, lambda d: d["Styles array"] if isinstance(d.get("Styles array"), list) else gr.update()),
-                (denoising_strength, "Denoising strength"),
-                (enable_hr, lambda d: "Denoising strength" in d and ("Hires upscale" in d or "Hires upscaler" in d or "Hires resize-1" in d)),
-                (hr_scale, "Hires upscale"),
-                (hr_upscaler, "Hires upscaler"),
-                (hr_second_pass_steps, "Hires steps"),
-                (hr_resize_x, "Hires resize-1"),
-                (hr_resize_y, "Hires resize-2"),
-                (hr_checkpoint_name, "Hires checkpoint"),
-                (hr_sampler_name, "Hires sampler"),
-                (hr_sampler_container, lambda d: gr.update(visible=True) if d.get("Hires sampler", "Use same sampler") != "Use same sampler" or d.get("Hires checkpoint", "Use same checkpoint") != "Use same checkpoint" else gr.update()),
-                (hr_prompt, "Hires prompt"),
-                (hr_negative_prompt, "Hires negative prompt"),
-                (hr_prompts_container, lambda d: gr.update(visible=True) if d.get("Hires prompt", "") != "" or d.get("Hires negative prompt", "") != "" else gr.update()),
+                PasteField(toprow.prompt, "Prompt", api="prompt"),
+                PasteField(toprow.negative_prompt, "Negative prompt", api="negative_prompt"),
+                PasteField(steps, "Steps", api="steps"),
+                PasteField(sampler_name, "Sampler", api="sampler_name"),
+                PasteField(cfg_scale, "CFG scale", api="cfg_scale"),
+                PasteField(width, "Size-1", api="width"),
+                PasteField(height, "Size-2", api="height"),
+                PasteField(batch_size, "Batch size", api="batch_size"),
+                PasteField(toprow.ui_styles.dropdown, lambda d: d["Styles array"] if isinstance(d.get("Styles array"), list) else gr.update(), api="styles"),
+                PasteField(denoising_strength, "Denoising strength", api="denoising_strength"),
+                PasteField(enable_hr, lambda d: "Denoising strength" in d and ("Hires upscale" in d or "Hires upscaler" in d or "Hires resize-1" in d), api="enable_hr"),
+                PasteField(hr_scale, "Hires upscale", api="hr_scale"),
+                PasteField(hr_upscaler, "Hires upscaler", api="hr_upscaler"),
+                PasteField(hr_second_pass_steps, "Hires steps", api="hr_second_pass_steps"),
+                PasteField(hr_resize_x, "Hires resize-1", api="hr_resize_x"),
+                PasteField(hr_resize_y, "Hires resize-2", api="hr_resize_y"),
+                PasteField(hr_checkpoint_name, "Hires checkpoint", api="hr_checkpoint_name"),
+                PasteField(hr_sampler_name, "Hires sampler", api="hr_sampler_name"),
+                PasteField(hr_sampler_container, lambda d: gr.update(visible=True) if d.get("Hires sampler", "Use same sampler") != "Use same sampler" or d.get("Hires checkpoint", "Use same checkpoint") != "Use same checkpoint" else gr.update()),
+                PasteField(hr_prompt, "Hires prompt", api="hr_prompt"),
+                PasteField(hr_negative_prompt, "Hires negative prompt", api="hr_negative_prompt"),
+                PasteField(hr_prompts_container, lambda d: gr.update(visible=True) if d.get("Hires prompt", "") != "" or d.get("Hires negative prompt", "") != "" else gr.update()),
                 *scripts.scripts_txt2img.infotext_fields
             ]
             parameters_copypaste.add_paste_fields("txt2img", None, txt2img_paste_fields, override_settings)
@@ -476,11 +499,13 @@ def create_ui():
                 height,
             ]
 
-            toprow.token_button.click(fn=wrap_queued_call(update_token_counter), inputs=[toprow.prompt, steps], outputs=[toprow.token_counter])
-            toprow.negative_token_button.click(fn=wrap_queued_call(update_negative_prompt_token_counter), inputs=[toprow.negative_prompt, steps], outputs=[toprow.negative_token_counter])
+            toprow.ui_styles.dropdown.change(fn=wrap_queued_call(update_token_counter), inputs=[toprow.prompt, steps, toprow.ui_styles.dropdown], outputs=[toprow.token_counter])
+            toprow.ui_styles.dropdown.change(fn=wrap_queued_call(update_negative_prompt_token_counter), inputs=[toprow.negative_prompt, steps, toprow.ui_styles.dropdown], outputs=[toprow.negative_token_counter])
+            toprow.token_button.click(fn=wrap_queued_call(update_token_counter), inputs=[toprow.prompt, steps, toprow.ui_styles.dropdown], outputs=[toprow.token_counter])
+            toprow.negative_token_button.click(fn=wrap_queued_call(update_negative_prompt_token_counter), inputs=[toprow.negative_prompt, steps, toprow.ui_styles.dropdown], outputs=[toprow.negative_token_counter])
 
         extra_networks_ui = ui_extra_networks.create_ui(txt2img_interface, [txt2img_generation_tab], 'txt2img')
-        ui_extra_networks.setup_ui(extra_networks_ui, txt2img_gallery)
+        ui_extra_networks.setup_ui(extra_networks_ui, output_panel.gallery)
 
         extra_tabs.__exit__()
 
@@ -490,7 +515,7 @@ def create_ui():
     with gr.Blocks(analytics_enabled=False) as img2img_interface:
         toprow = ui_toprow.Toprow(is_img2img=True, is_compact=shared.opts.compact_prompt_box)
 
-        extra_tabs = gr.Tabs(elem_id="img2img_extra_tabs")
+        extra_tabs = gr.Tabs(elem_id="img2img_extra_tabs", elem_classes=["extra-networks"])
         extra_tabs.__enter__()
 
         with gr.Tab("Generation", id="img2img_generation") as img2img_generation_tab, ResizeHandleRow(equal_height=False):
@@ -523,7 +548,7 @@ def add_copy_image_controls(tab_name, elem):
 
                     if category == "image":
                         with gr.Tabs(elem_id="mode_img2img"):
-                            img2img_selected_tab = gr.State(0)
+                            img2img_selected_tab = gr.Number(value=0, visible=False)
 
                             with gr.TabItem('img2img', id='img2img', elem_id="img2img_img2img_tab") as tab_img2img:
                                 init_img = gr.Image(label="Image for img2img", elem_id="img2img_image", show_label=False, source="upload", interactive=True, type="pil", tool="editor", image_mode="RGBA", height=opts.img2img_editor_height)
@@ -604,7 +629,7 @@ def copy_image(img):
                     elif category == "dimensions":
                         with FormRow():
                             with gr.Column(elem_id="img2img_column_size", scale=4):
-                                selected_scale_tab = gr.State(value=0)
+                                selected_scale_tab = gr.Number(value=0, visible=False)
 
                                 with gr.Tabs():
                                     with gr.Tab(label="Resize to", elem_id="img2img_tab_resize_to") as tab_scale_to:
@@ -711,7 +736,7 @@ def select_img2img_tab(tab):
                     outputs=[inpaint_controls, mask_alpha],
                 )
 
-            img2img_gallery, generation_info, html_info, html_log = create_output_panel("img2img", opts.outdir_img2img_samples, toprow)
+            output_panel = create_output_panel("img2img", opts.outdir_img2img_samples, toprow)
 
             img2img_args = dict(
                 fn=wrap_gradio_gpu_call(modules.img2img.img2img, extra_outputs=[None, '', '']),
@@ -756,10 +781,10 @@ def select_img2img_tab(tab):
                     img2img_batch_png_info_dir,
                 ] + custom_inputs,
                 outputs=[
-                    img2img_gallery,
-                    generation_info,
-                    html_info,
-                    html_log,
+                    output_panel.gallery,
+                    output_panel.generation_info,
+                    output_panel.infotext,
+                    output_panel.html_log,
                 ],
                 show_progress=False,
             )
@@ -797,10 +822,10 @@ def select_img2img_tab(tab):
                 _js="restoreProgressImg2img",
                 inputs=[dummy_component],
                 outputs=[
-                    img2img_gallery,
-                    generation_info,
-                    html_info,
-                    html_log,
+                    output_panel.gallery,
+                    output_panel.generation_info,
+                    output_panel.infotext,
+                    output_panel.html_log,
                 ],
                 show_progress=False,
             )
@@ -815,8 +840,10 @@ def select_img2img_tab(tab):
                 **interrogate_args,
             )
 
-            toprow.token_button.click(fn=update_token_counter, inputs=[toprow.prompt, steps], outputs=[toprow.token_counter])
-            toprow.negative_token_button.click(fn=wrap_queued_call(update_token_counter), inputs=[toprow.negative_prompt, steps], outputs=[toprow.negative_token_counter])
+            toprow.ui_styles.dropdown.change(fn=wrap_queued_call(update_token_counter), inputs=[toprow.prompt, steps, toprow.ui_styles.dropdown], outputs=[toprow.token_counter])
+            toprow.ui_styles.dropdown.change(fn=wrap_queued_call(update_negative_prompt_token_counter), inputs=[toprow.negative_prompt, steps, toprow.ui_styles.dropdown], outputs=[toprow.negative_token_counter])
+            toprow.token_button.click(fn=update_token_counter, inputs=[toprow.prompt, steps, toprow.ui_styles.dropdown], outputs=[toprow.token_counter])
+            toprow.negative_token_button.click(fn=wrap_queued_call(update_negative_prompt_token_counter), inputs=[toprow.negative_prompt, steps, toprow.ui_styles.dropdown], outputs=[toprow.negative_token_counter])
 
             img2img_paste_fields = [
                 (toprow.prompt, "Prompt"),
@@ -831,6 +858,10 @@ def select_img2img_tab(tab):
                 (toprow.ui_styles.dropdown, lambda d: d["Styles array"] if isinstance(d.get("Styles array"), list) else gr.update()),
                 (denoising_strength, "Denoising strength"),
                 (mask_blur, "Mask blur"),
+                (inpainting_mask_invert, 'Mask mode'),
+                (inpainting_fill, 'Masked content'),
+                (inpaint_full_res, 'Inpaint area'),
+                (inpaint_full_res_padding, 'Masked area padding'),
                 *scripts.scripts_img2img.infotext_fields
             ]
             parameters_copypaste.add_paste_fields("img2img", init_img, img2img_paste_fields, override_settings)
@@ -840,7 +871,7 @@ def select_img2img_tab(tab):
             ))
 
         extra_networks_ui_img2img = ui_extra_networks.create_ui(img2img_interface, [img2img_generation_tab], 'img2img')
-        ui_extra_networks.setup_ui(extra_networks_ui_img2img, img2img_gallery)
+        ui_extra_networks.setup_ui(extra_networks_ui_img2img, output_panel.gallery)
 
         extra_tabs.__exit__()
 
@@ -850,7 +881,7 @@ def select_img2img_tab(tab):
         ui_postprocessing.create_ui()
 
     with gr.Blocks(analytics_enabled=False) as pnginfo_interface:
-        with gr.Row(equal_height=False):
+        with ResizeHandleRow(equal_height=False):
             with gr.Column(variant='panel'):
                 image = gr.Image(elem_id="pnginfo_image", label="Source", source="upload", interactive=True, type="pil")
 
@@ -878,7 +909,7 @@ def select_img2img_tab(tab):
         with gr.Row(equal_height=False):
             gr.HTML(value="
    See wiki for detailed explanation.
    ")
 
-        with gr.Row(variant="compact", equal_height=False):
+        with ResizeHandleRow(variant="compact", equal_height=False):
             with gr.Tabs(elem_id="train_tabs"):
 
                 with gr.Tab(label="Create embedding", id="create_embedding"):
@@ -1086,8 +1117,8 @@ def get_textual_inversion_template_names():
         )
 
     loadsave = ui_loadsave.UiLoadsave(cmd_opts.ui_config_file)
+    ui_settings_from_file = loadsave.ui_settings.copy()
 
-    settings = ui_settings.UiSettings()
     settings.create_ui(loadsave, dummy_component)
 
     interfaces = [
@@ -1146,7 +1177,8 @@ def get_textual_inversion_template_names():
 
         modelmerger_ui.setup_ui(dummy_component=dummy_component, sd_model_checkpoint_component=settings.component_dict['sd_model_checkpoint'])
 
-    loadsave.dump_defaults()
+    if ui_settings_from_file != loadsave.ui_settings:
+        loadsave.dump_defaults()
     demo.ui_loadsave = loadsave
 
     return demo
@@ -1208,3 +1240,5 @@ def download_sysinfo(attachment=False):
     app.add_api_route("/internal/sysinfo", download_sysinfo, methods=["GET"])
     app.add_api_route("/internal/sysinfo-download", lambda: download_sysinfo(attachment=True), methods=["GET"])
 
+    import fastapi.staticfiles
+    app.mount("/webui-assets", fastapi.staticfiles.StaticFiles(directory=launch_utils.repo_dir('stable-diffusion-webui-assets')), name="webui-assets")
diff --git a/modules/ui_common.py b/modules/ui_common.py
index 032ec4af762..31b5492eacd 100644
--- a/modules/ui_common.py
+++ b/modules/ui_common.py
@@ -1,3 +1,5 @@
+import csv
+import dataclasses
 import json
 import html
 import os
@@ -7,11 +9,11 @@
 import gradio as gr
 import subprocess as sp
 
-from modules import call_queue, shared
-from modules.generation_parameters_copypaste import image_from_url_text
+from modules import call_queue, shared, ui_tempdir
+from modules.infotext_utils import image_from_url_text
 import modules.images
 from modules.ui_components import ToolButton
-import modules.generation_parameters_copypaste as parameters_copypaste
+import modules.infotext_utils as parameters_copypaste
 
 folder_symbol = '\U0001f4c2'  # 📂
 refresh_symbol = '\U0001f504'  # 🔄
@@ -35,12 +37,38 @@ def plaintext_to_html(text, classname=None):
     return f"
    {content}
    " if classname else f"
    {content}
    "
 
 
+def update_logfile(logfile_path, fields):
+    """Update a logfile from old format to new format to maintain CSV integrity."""
+    with open(logfile_path, "r", encoding="utf8", newline="") as file:
+        reader = csv.reader(file)
+        rows = list(reader)
+
+    # blank file: leave it as is
+    if not rows:
+        return
+
+    # file is already synced, do nothing
+    if len(rows[0]) == len(fields):
+        return
+
+    rows[0] = fields
+
+    # append new fields to each row as empty values
+    for row in rows[1:]:
+        while len(row) < len(fields):
+            row.append("")
+
+    with open(logfile_path, "w", encoding="utf8", newline="") as file:
+        writer = csv.writer(file)
+        writer.writerows(rows)
+
+
 def save_files(js_data, images, do_make_zip, index):
-    import csv
     filenames = []
     fullfns = []
+    parsed_infotexts = []
 
-    #quick dictionary to class object conversion. Its necessary due apply_filename_pattern requiring it
+    # quick dictionary to class object conversion. Its necessary due apply_filename_pattern requiring it
     class MyObject:
         def __init__(self, d=None):
             if d is not None:
@@ -48,35 +76,55 @@ def __init__(self, d=None):
                     setattr(self, key, value)
 
     data = json.loads(js_data)
-
     p = MyObject(data)
+
     path = shared.opts.outdir_save
     save_to_dirs = shared.opts.use_save_to_dirs_for_ui
     extension: str = shared.opts.samples_format
     start_index = 0
-    only_one = False
 
     if index > -1 and shared.opts.save_selected_only and (index >= data["index_of_first_image"]):  # ensures we are looking at a specific non-grid picture, and we have save_selected_only
-        only_one = True
         images = [images[index]]
         start_index = index
 
     os.makedirs(shared.opts.outdir_save, exist_ok=True)
 
-    with open(os.path.join(shared.opts.outdir_save, "log.csv"), "a", encoding="utf8", newline='') as file:
+    fields = [
+        "prompt",
+        "seed",
+        "width",
+        "height",
+        "sampler",
+        "cfgs",
+        "steps",
+        "filename",
+        "negative_prompt",
+        "sd_model_name",
+        "sd_model_hash",
+    ]
+    logfile_path = os.path.join(shared.opts.outdir_save, "log.csv")
+
+    # NOTE: ensure csv integrity when fields are added by
+    # updating headers and padding with delimiters where needed
+    if os.path.exists(logfile_path):
+        update_logfile(logfile_path, fields)
+
+    with open(logfile_path, "a", encoding="utf8", newline='') as file:
         at_start = file.tell() == 0
         writer = csv.writer(file)
         if at_start:
-            writer.writerow(["prompt", "seed", "width", "height", "sampler", "cfgs", "steps", "filename", "negative_prompt"])
+            writer.writerow(fields)
 
         for image_index, filedata in enumerate(images, start_index):
             image = image_from_url_text(filedata)
 
             is_grid = image_index < p.index_of_first_image
-            i = 0 if is_grid else (image_index - p.index_of_first_image)
 
             p.batch_index = image_index-1
-            fullfn, txt_fullfn = modules.images.save_image(image, path, "", seed=p.all_seeds[i], prompt=p.all_prompts[i], extension=extension, info=p.infotexts[image_index], grid=is_grid, p=p, save_to_dirs=save_to_dirs)
+
+            parameters = parameters_copypaste.parse_generation_parameters(data["infotexts"][image_index], [])
+            parsed_infotexts.append(parameters)
+            fullfn, txt_fullfn = modules.images.save_image(image, path, "", seed=parameters['Seed'], prompt=parameters['Prompt'], extension=extension, info=p.infotexts[image_index], grid=is_grid, p=p, save_to_dirs=save_to_dirs)
 
             filename = os.path.relpath(fullfn, path)
             filenames.append(filename)
@@ -85,12 +133,12 @@ def __init__(self, d=None):
                 filenames.append(os.path.basename(txt_fullfn))
                 fullfns.append(txt_fullfn)
 
-        writer.writerow([data["prompt"], data["seed"], data["width"], data["height"], data["sampler_name"], data["cfg_scale"], data["steps"], filenames[0], data["negative_prompt"]])
+        writer.writerow([parsed_infotexts[0]['Prompt'], parsed_infotexts[0]['Seed'], data["width"], data["height"], data["sampler_name"], data["cfg_scale"], data["steps"], filenames[0], parsed_infotexts[0]['Negative prompt'], data["sd_model_name"], data["sd_model_hash"]])
 
     # Make Zip
     if do_make_zip:
-        zip_fileseed = p.all_seeds[index-1] if only_one else p.all_seeds[0]
-        namegen = modules.images.FilenameGenerator(p, zip_fileseed, p.all_prompts[0], image, True)
+        p.all_seeds = [parameters['Seed'] for parameters in parsed_infotexts]
+        namegen = modules.images.FilenameGenerator(p, parsed_infotexts[0]['Seed'], parsed_infotexts[0]['Prompt'], image, True)
         zip_filename = namegen.apply(shared.opts.grid_zip_filename_pattern or "[datetime]_[[model_name]]_[seed]-[seed_last]")
         zip_filepath = os.path.join(path, f"{zip_filename}.zip")
 
@@ -104,31 +152,55 @@ def __init__(self, d=None):
     return gr.File.update(value=fullfns, visible=True), plaintext_to_html(f"Saved: {filenames[0]}")
 
 
+@dataclasses.dataclass
+class OutputPanel:
+    gallery = None
+    generation_info = None
+    infotext = None
+    html_log = None
+    button_upscale = None
+
+
 def create_output_panel(tabname, outdir, toprow=None):
+    res = OutputPanel()
+
+    def open_folder(f, images=None, index=None):
+        if shared.cmd_opts.hide_ui_dir_config:
+            return
+
+        try:
+            if 'Sub' in shared.opts.open_dir_button_choice:
+                image_dir = os.path.split(images[index]["name"].rsplit('?', 1)[0])[0]
+                if 'temp' in shared.opts.open_dir_button_choice or not ui_tempdir.is_gradio_temp_path(image_dir):
+                    f = image_dir
+        except Exception:
+            pass
 
-    def open_folder(f):
         if not os.path.exists(f):
-            print(f'Folder "{f}" does not exist. After you create an image, the folder will be created.')
+            msg = f'Folder "{f}" does not exist. After you create an image, the folder will be created.'
+            print(msg)
+            gr.Info(msg)
             return
         elif not os.path.isdir(f):
-            print(f"""
+            msg = f"""
 WARNING
 An open_folder request was made with an argument that is not a folder.
 This could be an error or a malicious attempt to run code on your computer.
 Requested path was: {f}
-""", file=sys.stderr)
+"""
+            print(msg, file=sys.stderr)
+            gr.Warning(msg)
             return
 
-        if not shared.cmd_opts.hide_ui_dir_config:
-            path = os.path.normpath(f)
-            if platform.system() == "Windows":
-                os.startfile(path)
-            elif platform.system() == "Darwin":
-                sp.Popen(["open", path])
-            elif "microsoft-standard-WSL2" in platform.uname().release:
-                sp.Popen(["wsl-open", path])
-            else:
-                sp.Popen(["xdg-open", path])
+        path = os.path.normpath(f)
+        if platform.system() == "Windows":
+            os.startfile(path)
+        elif platform.system() == "Darwin":
+            sp.Popen(["open", path])
+        elif "microsoft-standard-WSL2" in platform.uname().release:
+            sp.Popen(["wsl-open", path])
+        else:
+            sp.Popen(["xdg-open", path])
 
     with gr.Column(elem_id=f"{tabname}_results"):
         if toprow:
@@ -136,9 +208,8 @@ def open_folder(f):
 
         with gr.Column(variant='panel', elem_id=f"{tabname}_results_panel"):
             with gr.Group(elem_id=f"{tabname}_gallery_container"):
-                result_gallery = gr.Gallery(label='Output', show_label=False, elem_id=f"{tabname}_gallery", columns=4, preview=True, height=shared.opts.gallery_height or None)
+                res.gallery = gr.Gallery(label='Output', show_label=False, elem_id=f"{tabname}_gallery", columns=4, preview=True, height=shared.opts.gallery_height or None)
 
-            generation_info = None
             with gr.Row(elem_id=f"image_buttons_{tabname}", elem_classes="image-buttons"):
                 open_folder_button = ToolButton(folder_symbol, elem_id=f'{tabname}_open_folder', visible=not shared.cmd_opts.hide_ui_dir_config, tooltip="Open images output directory.")
 
@@ -152,9 +223,16 @@ def open_folder(f):
                     'extras': ToolButton('📐', elem_id=f'{tabname}_send_to_extras', tooltip="Send image and generation parameters to extras tab.")
                 }
 
+                if tabname == 'txt2img':
+                    res.button_upscale = ToolButton('✨', elem_id=f'{tabname}_upscale', tooltip="Create an upscaled version of the current image using hires fix settings.")
+
             open_folder_button.click(
-                fn=lambda: open_folder(shared.opts.outdir_samples or outdir),
-                inputs=[],
+                fn=lambda images, index: open_folder(shared.opts.outdir_samples or outdir, images, index),
+                _js="(y, w) => [y, selected_gallery_index()]",
+                inputs=[
+                    res.gallery,
+                    open_folder_button,  # placeholder for index
+                ],
                 outputs=[],
             )
 
@@ -162,17 +240,17 @@ def open_folder(f):
                 download_files = gr.File(None, file_count="multiple", interactive=False, show_label=False, visible=False, elem_id=f'download_files_{tabname}')
 
                 with gr.Group():
-                    html_info = gr.HTML(elem_id=f'html_info_{tabname}', elem_classes="infotext")
-                    html_log = gr.HTML(elem_id=f'html_log_{tabname}', elem_classes="html-log")
+                    res.infotext = gr.HTML(elem_id=f'html_info_{tabname}', elem_classes="infotext")
+                    res.html_log = gr.HTML(elem_id=f'html_log_{tabname}', elem_classes="html-log")
 
-                    generation_info = gr.Textbox(visible=False, elem_id=f'generation_info_{tabname}')
+                    res.generation_info = gr.Textbox(visible=False, elem_id=f'generation_info_{tabname}')
                     if tabname == 'txt2img' or tabname == 'img2img':
                         generation_info_button = gr.Button(visible=False, elem_id=f"{tabname}_generation_info_button")
                         generation_info_button.click(
                             fn=update_generation_info,
                             _js="function(x, y, z){ return [x, y, selected_gallery_index()] }",
-                            inputs=[generation_info, html_info, html_info],
-                            outputs=[html_info, html_info],
+                            inputs=[res.generation_info, res.infotext, res.infotext],
+                            outputs=[res.infotext, res.infotext],
                             show_progress=False,
                         )
 
@@ -180,14 +258,14 @@ def open_folder(f):
                         fn=call_queue.wrap_gradio_call(save_files),
                         _js="(x, y, z, w) => [x, y, false, selected_gallery_index()]",
                         inputs=[
-                            generation_info,
-                            result_gallery,
-                            html_info,
-                            html_info,
+                            res.generation_info,
+                            res.gallery,
+                            res.infotext,
+                            res.infotext,
                         ],
                         outputs=[
                             download_files,
-                            html_log,
+                            res.html_log,
                         ],
                         show_progress=False,
                     )
@@ -196,21 +274,21 @@ def open_folder(f):
                         fn=call_queue.wrap_gradio_call(save_files),
                         _js="(x, y, z, w) => [x, y, true, selected_gallery_index()]",
                         inputs=[
-                            generation_info,
-                            result_gallery,
-                            html_info,
-                            html_info,
+                            res.generation_info,
+                            res.gallery,
+                            res.infotext,
+                            res.infotext,
                         ],
                         outputs=[
                             download_files,
-                            html_log,
+                            res.html_log,
                         ]
                     )
 
             else:
-                html_info_x = gr.HTML(elem_id=f'html_info_x_{tabname}')
-                html_info = gr.HTML(elem_id=f'html_info_{tabname}', elem_classes="infotext")
-                html_log = gr.HTML(elem_id=f'html_log_{tabname}')
+                res.generation_info = gr.HTML(elem_id=f'html_info_x_{tabname}')
+                res.infotext = gr.HTML(elem_id=f'html_info_{tabname}', elem_classes="infotext")
+                res.html_log = gr.HTML(elem_id=f'html_log_{tabname}')
 
             paste_field_names = []
             if tabname == "txt2img":
@@ -220,11 +298,11 @@ def open_folder(f):
 
             for paste_tabname, paste_button in buttons.items():
                 parameters_copypaste.register_paste_params_button(parameters_copypaste.ParamBinding(
-                    paste_button=paste_button, tabname=paste_tabname, source_tabname="txt2img" if tabname == "txt2img" else None, source_image_component=result_gallery,
+                    paste_button=paste_button, tabname=paste_tabname, source_tabname="txt2img" if tabname == "txt2img" else None, source_image_component=res.gallery,
                     paste_field_names=paste_field_names
                 ))
 
-            return result_gallery, generation_info if tabname != "extras" else html_info_x, html_info, html_log
+    return res
 
 
 def create_refresh_button(refresh_component, refresh_method, refreshed_args, elem_id):
diff --git a/modules/ui_components.py b/modules/ui_components.py
index 55979f62629..9cf67722a3d 100644
--- a/modules/ui_components.py
+++ b/modules/ui_components.py
@@ -88,7 +88,7 @@ def get_block_name(self):
 class InputAccordion(gr.Checkbox):
     """A gr.Accordion that can be used as an input - returns True if open, False if closed.
 
-    Actaully just a hidden checkbox, but creates an accordion that follows and is followed by the state of the checkbox.
+    Actually just a hidden checkbox, but creates an accordion that follows and is followed by the state of the checkbox.
     """
 
     global_index = 0
diff --git a/modules/ui_extensions.py b/modules/ui_extensions.py
index dc1e34c8af8..913e1444e95 100644
--- a/modules/ui_extensions.py
+++ b/modules/ui_extensions.py
@@ -380,7 +380,7 @@ def install_extension_from_url(dirname, url, branch_name=None):
         except OSError as err:
             if err.errno == errno.EXDEV:
                 # Cross device link, typical in docker or when tmp/ and extensions/ are on different file systems
-                # Since we can't use a rename, do the slower but more versitile shutil.move()
+                # Since we can't use a rename, do the slower but more versatile shutil.move()
                 shutil.move(tmpdir, target_dir)
             else:
                 # Something else, not enough free space, permissions, etc.  rethrow it so that it gets handled.
@@ -548,6 +548,7 @@ def create_ui():
                     extensions_disable_all = gr.Radio(label="Disable all extensions", choices=["none", "extra", "all"], value=shared.opts.disable_all_extensions, elem_id="extensions_disable_all")
                     extensions_disabled_list = gr.Text(elem_id="extensions_disabled_list", visible=False, container=False)
                     extensions_update_list = gr.Text(elem_id="extensions_update_list", visible=False, container=False)
+                    refresh = gr.Button(value='Refresh', variant="compact")
 
                 html = ""
 
@@ -566,7 +567,8 @@ def create_ui():
                 with gr.Row(elem_classes="progress-container"):
                     extensions_table = gr.HTML('Loading...', elem_id="extensions_installed_html")
 
-                ui.load(fn=extension_table, inputs=[], outputs=[extensions_table])
+                ui.load(fn=extension_table, inputs=[], outputs=[extensions_table], show_progress=False)
+                refresh.click(fn=extension_table, inputs=[], outputs=[extensions_table], show_progress=False)
 
                 apply.click(
                     fn=apply_and_restart,
diff --git a/modules/ui_extra_networks.py b/modules/ui_extra_networks.py
index fe5d3ba3338..ad2c23054d9 100644
--- a/modules/ui_extra_networks.py
+++ b/modules/ui_extra_networks.py
@@ -2,23 +2,22 @@
 import os.path
 import urllib.parse
 from pathlib import Path
+from typing import Optional, Union
+from dataclasses import dataclass
 
-from modules import shared, ui_extra_networks_user_metadata, errors, extra_networks
+from modules import shared, ui_extra_networks_user_metadata, errors, extra_networks, util
 from modules.images import read_info_from_image, save_image_with_geninfo
 import gradio as gr
 import json
 import html
 from fastapi.exceptions import HTTPException
 
-from modules.generation_parameters_copypaste import image_from_url_text
-from modules.ui_components import ToolButton
+from modules.infotext_utils import image_from_url_text
 
 extra_pages = []
 allowed_dirs = set()
-
 default_allowed_preview_extensions = ["png", "jpg", "jpeg", "webp", "gif"]
 
-
 @functools.cache
 def allowed_preview_extensions_with_extra(extra_extensions=None):
     return set(default_allowed_preview_extensions) | set(extra_extensions or [])
@@ -28,6 +27,62 @@ def allowed_preview_extensions():
     return allowed_preview_extensions_with_extra((shared.opts.samples_format, ))
 
 
+@dataclass
+class ExtraNetworksItem:
+    """Wrapper for dictionaries representing ExtraNetworks items."""
+    item: dict
+
+
+def get_tree(paths: Union[str, list[str]], items: dict[str, ExtraNetworksItem]) -> dict:
+    """Recursively builds a directory tree.
+
+    Args:
+        paths: Path or list of paths to directories. These paths are treated as roots from which
+            the tree will be built.
+        items: A dictionary associating filepaths to an ExtraNetworksItem instance.
+
+    Returns:
+        The result directory tree.
+    """
+    if isinstance(paths, (str,)):
+        paths = [paths]
+
+    def _get_tree(_paths: list[str], _root: str):
+        _res = {}
+        for path in _paths:
+            relpath = os.path.relpath(path, _root)
+            if os.path.isdir(path):
+                dir_items = os.listdir(path)
+                # Ignore empty directories.
+                if not dir_items:
+                    continue
+                dir_tree = _get_tree([os.path.join(path, x) for x in dir_items], _root)
+                # We only want to store non-empty folders in the tree.
+                if dir_tree:
+                    _res[relpath] = dir_tree
+            else:
+                if path not in items:
+                    continue
+                # Add the ExtraNetworksItem to the result.
+                _res[relpath] = items[path]
+        return _res
+
+    res = {}
+    # Handle each root directory separately.
+    # Each root WILL have a key/value at the root of the result dict though
+    # the value can be an empty dict if the directory is empty. We want these
+    # placeholders for empty dirs so we can inform the user later.
+    for path in paths:
+        root = os.path.dirname(path)
+        relpath = os.path.relpath(path, root)
+        # Wrap the path in a list since that is what the `_get_tree` expects.
+        res[relpath] = _get_tree([path], root)
+        if res[relpath]:
+            # We need to pull the inner path out one for these root dirs.
+            res[relpath] = res[relpath][relpath]
+
+    return res
+
 def register_page(page):
     """registers extra networks page for the UI; recommend doing it in on_before_ui() callback for extensions"""
 
@@ -79,8 +134,8 @@ def get_single_card(page: str = "", tabname: str = "", name: str = ""):
         errors.display(e, "creating item for extra network")
         item = page.items.get(name)
 
-    page.read_user_metadata(item)
-    item_html = page.create_html_for_item(item, tabname)
+    page.read_user_metadata(item, use_cache=False)
+    item_html = page.create_item_html(tabname, item, shared.html("extra-networks-card.html"))
 
     return JSONResponse({"html": item_html})
 
@@ -96,24 +151,31 @@ def quote_js(s):
     s = s.replace('"', '\\"')
     return f'"{s}"'
 
-
 class ExtraNetworksPage:
     def __init__(self, title):
         self.title = title
         self.name = title.lower()
-        self.id_page = self.name.replace(" ", "_")
-        self.card_page = shared.html("extra-networks-card.html")
+        # This is the actual name of the extra networks tab (not txt2img/img2img).
+        self.extra_networks_tabname = self.name.replace(" ", "_")
         self.allow_prompt = True
         self.allow_negative_prompt = False
         self.metadata = {}
         self.items = {}
+        self.lister = util.MassFileLister()
+        # HTML Templates
+        self.pane_tpl = shared.html("extra-networks-pane.html")
+        self.card_tpl = shared.html("extra-networks-card.html")
+        self.btn_tree_tpl = shared.html("extra-networks-tree-button.html")
+        self.btn_copy_path_tpl = shared.html("extra-networks-copy-path-button.html")
+        self.btn_metadata_tpl = shared.html("extra-networks-metadata-button.html")
+        self.btn_edit_item_tpl = shared.html("extra-networks-edit-item-button.html")
 
     def refresh(self):
         pass
 
-    def read_user_metadata(self, item):
+    def read_user_metadata(self, item, use_cache=True):
         filename = item.get("filename", None)
-        metadata = extra_networks.get_user_metadata(filename)
+        metadata = extra_networks.get_user_metadata(filename, lister=self.lister if use_cache else None)
 
         desc = metadata.get("description", None)
         if desc is not None:
@@ -123,117 +185,74 @@ def read_user_metadata(self, item):
 
     def link_preview(self, filename):
         quoted_filename = urllib.parse.quote(filename.replace('\\', '/'))
-        mtime = os.path.getmtime(filename)
+        mtime, _ = self.lister.mctime(filename)
         return f"./sd_extra_networks/thumb?filename={quoted_filename}&mtime={mtime}"
 
     def search_terms_from_path(self, filename, possible_directories=None):
         abspath = os.path.abspath(filename)
-
         for parentdir in (possible_directories if possible_directories is not None else self.allowed_directories_for_previews()):
-            parentdir = os.path.abspath(parentdir)
+            parentdir = os.path.dirname(os.path.abspath(parentdir))
             if abspath.startswith(parentdir):
-                return abspath[len(parentdir):].replace('\\', '/')
+                return os.path.relpath(abspath, parentdir)
 
         return ""
 
-    def create_html(self, tabname):
-        items_html = ''
-
-        self.metadata = {}
-
-        subdirs = {}
-        for parentdir in [os.path.abspath(x) for x in self.allowed_directories_for_previews()]:
-            for root, dirs, _ in sorted(os.walk(parentdir, followlinks=True), key=lambda x: shared.natural_sort_key(x[0])):
-                for dirname in sorted(dirs, key=shared.natural_sort_key):
-                    x = os.path.join(root, dirname)
-
-                    if not os.path.isdir(x):
-                        continue
-
-                    subdir = os.path.abspath(x)[len(parentdir):].replace("\\", "/")
-
-                    if shared.opts.extra_networks_dir_button_function:
-                        if not subdir.startswith("/"):
-                            subdir = "/" + subdir
-                    else:
-                        while subdir.startswith("/"):
-                            subdir = subdir[1:]
-
-                    is_empty = len(os.listdir(x)) == 0
-                    if not is_empty and not subdir.endswith("/"):
-                        subdir = subdir + "/"
-
-                    if ("/." in subdir or subdir.startswith(".")) and not shared.opts.extra_networks_show_hidden_directories:
-                        continue
-
-                    subdirs[subdir] = 1
-
-        if subdirs:
-            subdirs = {"": 1, **subdirs}
-
-        subdirs_html = "".join([f"""
-
-""" for subdir in subdirs])
-
-        self.items = {x["name"]: x for x in self.list_items()}
-        for item in self.items.values():
-            metadata = item.get("metadata")
-            if metadata:
-                self.metadata[item["name"]] = metadata
-
-            if "user_metadata" not in item:
-                self.read_user_metadata(item)
-
-            items_html += self.create_html_for_item(item, tabname)
-
-        if items_html == '':
-            dirs = "".join([f"
  • {x}
    • " for x in self.allowed_directories_for_previews()])
-            items_html = shared.html("extra-networks-no-cards.html").format(dirs=dirs)
-
-        self_name_id = self.name.replace(" ", "_")
-
-        res = f"""
-
    
-{subdirs_html}
-
    
-
    
-{items_html}
-
    
-"""
-
-        return res
-
-    def create_item(self, name, index=None):
-        raise NotImplementedError()
-
-    def list_items(self):
-        raise NotImplementedError()
-
-    def allowed_directories_for_previews(self):
-        return []
-
-    def create_html_for_item(self, item, tabname):
-        """
-        Create HTML for card item in tab tabname; can return empty string if the item is not meant to be shown.
+    def create_item_html(
+        self,
+        tabname: str,
+        item: dict,
+        template: Optional[str] = None,
+    ) -> Union[str, dict]:
+        """Generates HTML for a single ExtraNetworks Item.
+
+        Args:
+            tabname: The name of the active tab.
+            item: Dictionary containing item information.
+            template: Optional template string to use.
+
+        Returns:
+            If a template is passed: HTML string generated for this item.
+                Can be empty if the item is not meant to be shown.
+            If no template is passed: A dictionary containing the generated item's attributes.
         """
-
         preview = item.get("preview", None)
+        style_height = f"height: {shared.opts.extra_networks_card_height}px;" if shared.opts.extra_networks_card_height else ''
+        style_width = f"width: {shared.opts.extra_networks_card_width}px;" if shared.opts.extra_networks_card_width else ''
+        style_font_size = f"font-size: {shared.opts.extra_networks_card_text_scale*100}%;"
+        card_style = style_height + style_width + style_font_size
+        background_image = f'' if preview else ''
 
         onclick = item.get("onclick", None)
         if onclick is None:
-            onclick = '"' + html.escape(f"""return cardClicked({quote_js(tabname)}, {item["prompt"]}, {"true" if self.allow_negative_prompt else "false"})""") + '"'
-
-        height = f"height: {shared.opts.extra_networks_card_height}px;" if shared.opts.extra_networks_card_height else ''
-        width = f"width: {shared.opts.extra_networks_card_width}px;" if shared.opts.extra_networks_card_width else ''
-        background_image = f'' if preview else ''
-        metadata_button = ""
+            # Don't quote prompt/neg_prompt since they are stored as js strings already.
+            onclick_js_tpl = "cardClicked('{tabname}', {prompt}, {neg_prompt}, {allow_neg});"
+            onclick = onclick_js_tpl.format(
+                **{
+                    "tabname": tabname,
+                    "prompt": item["prompt"],
+                    "neg_prompt": item.get("negative_prompt", "''"),
+                    "allow_neg": str(self.allow_negative_prompt).lower(),
+                }
+            )
+            onclick = html.escape(onclick)
+
+        btn_copy_path = self.btn_copy_path_tpl.format(**{"filename": item["filename"]})
+        btn_metadata = ""
         metadata = item.get("metadata")
         if metadata:
-            metadata_button = f"
    "
-
-        edit_button = f"
    "
+            btn_metadata = self.btn_metadata_tpl.format(
+                **{
+                    "extra_networks_tabname": self.extra_networks_tabname,
+                    "name": html.escape(item["name"]),
+                }
+            )
+        btn_edit_item = self.btn_edit_item_tpl.format(
+            **{
+                "tabname": tabname,
+                "extra_networks_tabname": self.extra_networks_tabname,
+                "name": html.escape(item["name"]),
+            }
+        )
 
         local_path = ""
         filename = item.get("filename", "")
@@ -253,38 +272,310 @@ def create_html_for_item(self, item, tabname):
         if search_only and shared.opts.extra_networks_hidden_models == "Never":
             return ""
 
-        sort_keys = " ".join([f'data-sort-{k}="{html.escape(str(v))}"' for k, v in item.get("sort_keys", {}).items()]).strip()
-
+        sort_keys = " ".join(
+            [
+                f'data-sort-{k}="{html.escape(str(v))}"'
+                for k, v in item.get("sort_keys", {}).items()
+            ]
+        ).strip()
+
+        search_terms_html = ""
+        search_term_template = ""
+        for search_term in item.get("search_terms", []):
+            search_terms_html += search_term_template.format(
+                **{
+                    "class": f"search_terms{' search_only' if search_only else ''}",
+                    "search_term": search_term,
+                }
+            )
+
+        description = (item.get("description", "") or "" if shared.opts.extra_networks_card_show_desc else "")
+        if not shared.opts.extra_networks_card_description_is_html:
+            description = html.escape(description)
+
+        # Some items here might not be used depending on HTML template used.
         args = {
             "background_image": background_image,
-            "style": f"'display: none; {height}{width}; font-size: {shared.opts.extra_networks_card_text_scale*100}%'",
-            "prompt": item.get("prompt", None),
-            "tabname": quote_js(tabname),
+            "card_clicked": onclick,
+            "copy_path_button": btn_copy_path,
+            "description": description,
+            "edit_button": btn_edit_item,
             "local_preview": quote_js(item["local_preview"]),
+            "metadata_button": btn_metadata,
             "name": html.escape(item["name"]),
-            "description": (item.get("description") or "" if shared.opts.extra_networks_card_show_desc else ""),
-            "card_clicked": onclick,
-            "save_card_preview": '"' + html.escape(f"""return saveCardPreview(event, {quote_js(tabname)}, {quote_js(item["local_preview"])})""") + '"',
-            "search_term": item.get("search_term", ""),
-            "metadata_button": metadata_button,
-            "edit_button": edit_button,
+            "prompt": item.get("prompt", None),
+            "save_card_preview": html.escape(f"return saveCardPreview(event, '{tabname}', '{item['local_preview']}');"),
             "search_only": " search_only" if search_only else "",
+            "search_terms": search_terms_html,
             "sort_keys": sort_keys,
+            "style": card_style,
+            "tabname": tabname,
+            "extra_networks_tabname": self.extra_networks_tabname,
         }
 
-        return self.card_page.format(**args)
+        if template:
+            return template.format(**args)
+        else:
+            return args
+
+    def create_tree_dir_item_html(
+        self,
+        tabname: str,
+        dir_path: str,
+        content: Optional[str] = None,
+    ) -> Optional[str]:
+        """Generates HTML for a directory item in the tree.
+
+        The generated HTML is of the format:
+        ```html
+        
  • 
+            
    
+            
      
+                {content}
+            
    
+        
    • 
+        ```
+
+        Args:
+            tabname: The name of the active tab.
+            dir_path: Path to the directory for this item.
+            content: Optional HTML string that will be wrapped by this 
      .
+
+        Returns:
+            HTML formatted string.
+        """
+        if not content:
+            return None
+
+        btn = self.btn_tree_tpl.format(
+            **{
+                "search_terms": "",
+                "subclass": "tree-list-content-dir",
+                "tabname": tabname,
+                "extra_networks_tabname": self.extra_networks_tabname,
+                "onclick_extra": "",
+                "data_path": dir_path,
+                "data_hash": "",
+                "action_list_item_action_leading": "",
+                "action_list_item_visual_leading": "🗀",
+                "action_list_item_label": os.path.basename(dir_path),
+                "action_list_item_visual_trailing": "",
+                "action_list_item_action_trailing": "",
+            }
+        )
+        ul = f""
+        return (
+            "
    • "
+            f"{btn}{ul}"
+            "
      • "
+        )
+
+    def create_tree_file_item_html(self, tabname: str, file_path: str, item: dict) -> str:
+        """Generates HTML for a file item in the tree.
+
+        The generated HTML is of the format:
+        ```html
+        
    • 
+            
+            
      
+        
      • 
+        ```
+
+        Args:
+            tabname: The name of the active tab.
+            file_path: The path to the file for this item.
+            item: Dictionary containing the item information.
+
+        Returns:
+            HTML formatted string.
+        """
+        item_html_args = self.create_item_html(tabname, item)
+        action_buttons = "".join(
+            [
+                item_html_args["copy_path_button"],
+                item_html_args["metadata_button"],
+                item_html_args["edit_button"],
+            ]
+        )
+        action_buttons = f"
      {action_buttons}
      "
+        btn = self.btn_tree_tpl.format(
+            **{
+                "search_terms": "",
+                "subclass": "tree-list-content-file",
+                "tabname": tabname,
+                "extra_networks_tabname": self.extra_networks_tabname,
+                "onclick_extra": item_html_args["card_clicked"],
+                "data_path": file_path,
+                "data_hash": item["shorthash"],
+                "action_list_item_action_leading": "",
+                "action_list_item_visual_leading": "🗎",
+                "action_list_item_label": item["name"],
+                "action_list_item_visual_trailing": "",
+                "action_list_item_action_trailing": action_buttons,
+            }
+        )
+        return (
+            "
    • "
+            f"{btn}"
+            "
      • "
+        )
+
+    def create_tree_view_html(self, tabname: str) -> str:
+        """Generates HTML for displaying folders in a tree view.
+
+        Args:
+            tabname: The name of the active tab.
+
+        Returns:
+            HTML string generated for this tree view.
+        """
+        res = ""
+
+        # Setup the tree dictionary.
+        roots = self.allowed_directories_for_previews()
+        tree_items = {v["filename"]: ExtraNetworksItem(v) for v in self.items.values()}
+        tree = get_tree([os.path.abspath(x) for x in roots], items=tree_items)
+
+        if not tree:
+            return res
+
+        def _build_tree(data: Optional[dict[str, ExtraNetworksItem]] = None) -> Optional[str]:
+            """Recursively builds HTML for a tree.
+
+            Args:
+                data: Dictionary representing a directory tree. Can be NoneType.
+                    Data keys should be absolute paths from the root and values
+                    should be subdirectory trees or an ExtraNetworksItem.
+
+            Returns:
+                If data is not None: HTML string
+                Else: None
+            """
+            if not data:
+                return None
+
+            # Lists for storing 
    •  items html for directories and files separately.
+            _dir_li = []
+            _file_li = []
+
+            for k, v in sorted(data.items(), key=lambda x: shared.natural_sort_key(x[0])):
+                if isinstance(v, (ExtraNetworksItem,)):
+                    _file_li.append(self.create_tree_file_item_html(tabname, k, v.item))
+                else:
+                    _dir_li.append(self.create_tree_dir_item_html(tabname, k, _build_tree(v)))
+
+            # Directories should always be displayed before files so we order them here.
+            return "".join(_dir_li) + "".join(_file_li)
+
+        # Add each root directory to the tree.
+        for k, v in sorted(tree.items(), key=lambda x: shared.natural_sort_key(x[0])):
+            item_html = self.create_tree_dir_item_html(tabname, k, _build_tree(v))
+            # Only add non-empty entries to the tree.
+            if item_html is not None:
+                res += item_html
+
+        return f"
        {res}
      "
+
+    def create_card_view_html(self, tabname: str, *, none_message) -> str:
+        """Generates HTML for the network Card View section for a tab.
+
+        This HTML goes into the `extra-networks-pane.html` 
       with
+        `id='{tabname}_{extra_networks_tabname}_cards`.
+
+        Args:
+            tabname: The name of the active tab.
+            none_message: HTML text to show when there are no cards.
+
+        Returns:
+            HTML formatted string.
+        """
+        res = ""
+        for item in self.items.values():
+            res += self.create_item_html(tabname, item, self.card_tpl)
+
+        if res == "":
+            dirs = "".join([f"
    • {x}
      • " for x in self.allowed_directories_for_previews()])
+            res = none_message or shared.html("extra-networks-no-cards.html").format(dirs=dirs)
+
+        return res
+
+    def create_html(self, tabname, *, empty=False):
+        """Generates an HTML string for the current pane.
+
+        The generated HTML uses `extra-networks-pane.html` as a template.
+
+        Args:
+            tabname: The name of the active tab.
+            empty: create an empty HTML page with no items
+
+        Returns:
+            HTML formatted string.
+        """
+        self.lister.reset()
+        self.metadata = {}
+
+        items_list = [] if empty else self.list_items()
+        self.items = {x["name"]: x for x in items_list}
+
+        # Populate the instance metadata for each item.
+        for item in self.items.values():
+            metadata = item.get("metadata")
+            if metadata:
+                self.metadata[item["name"]] = metadata
+
+            if "user_metadata" not in item:
+                self.read_user_metadata(item)
+
+        data_sortdir = shared.opts.extra_networks_card_order
+        data_sortmode = shared.opts.extra_networks_card_order_field.lower().replace("sort", "").replace(" ", "_").rstrip("_").strip()
+        data_sortkey = f"{data_sortmode}-{data_sortdir}-{len(self.items)}"
+        tree_view_btn_extra_class = ""
+        tree_view_div_extra_class = "hidden"
+        tree_view_div_default_display = "none"
+        extra_network_pane_content_default_display = "flex"
+        if shared.opts.extra_networks_tree_view_default_enabled:
+            tree_view_btn_extra_class = "extra-network-control--enabled"
+            tree_view_div_extra_class = ""
+            tree_view_div_default_display = "block"
+            extra_network_pane_content_default_display = "grid"
+
+        return self.pane_tpl.format(
+            **{
+                "tabname": tabname,
+                "extra_networks_tabname": self.extra_networks_tabname,
+                "data_sortmode": data_sortmode,
+                "data_sortkey": data_sortkey,
+                "data_sortdir": data_sortdir,
+                "tree_view_btn_extra_class": tree_view_btn_extra_class,
+                "tree_view_div_extra_class": tree_view_div_extra_class,
+                "tree_html": self.create_tree_view_html(tabname),
+                "items_html": self.create_card_view_html(tabname, none_message="Loading..." if empty else None),
+                "extra_networks_tree_view_default_width": shared.opts.extra_networks_tree_view_default_width,
+                "tree_view_div_default_display": tree_view_div_default_display,
+                "extra_network_pane_content_default_display": extra_network_pane_content_default_display,
+            }
+        )
+
+    def create_item(self, name, index=None):
+        raise NotImplementedError()
+
+    def list_items(self):
+        raise NotImplementedError()
+
+    def allowed_directories_for_previews(self):
+        return []
 
     def get_sort_keys(self, path):
         """
         List of default keys used for sorting in the UI.
         """
         pth = Path(path)
-        stat = pth.stat()
+        mtime, ctime = self.lister.mctime(path)
         return {
-            "date_created": int(stat.st_ctime or 0),
-            "date_modified": int(stat.st_mtime or 0),
+            "date_created": int(mtime),
+            "date_modified": int(ctime),
             "name": pth.name.lower(),
-            "path": str(pth.parent).lower(),
+            "path": str(pth).lower(),
         }
 
     def find_preview(self, path):
@@ -292,10 +583,10 @@ def find_preview(self, path):
         Find a preview PNG for a given path (without extension) and call link_preview on it.
         """
 
-        potential_files = sum([[path + "." + ext, path + ".preview." + ext] for ext in allowed_preview_extensions()], [])
+        potential_files = sum([[f"{path}.{ext}", f"{path}.preview.{ext}"] for ext in allowed_preview_extensions()], [])
 
         for file in potential_files:
-            if os.path.isfile(file):
+            if self.lister.exists(file):
                 return self.link_preview(file)
 
         return None
@@ -305,6 +596,9 @@ def find_description(self, path):
         Find and read a description file for a given path (without extension).
         """
         for file in [f"{path}.txt", f"{path}.description.txt"]:
+            if not self.lister.exists(file):
+                continue
+
             try:
                 with open(file, "r", encoding="utf-8", errors="replace") as f:
                     return f.read()
@@ -362,8 +656,6 @@ def tab_name_score(name):
 
 
 def create_ui(interface: gr.Blocks, unrelated_tabs, tabname):
-    from modules.ui import switch_values_symbol
-
     ui = ExtraNetworksUi()
     ui.pages = []
     ui.pages_contents = []
@@ -374,62 +666,51 @@ def create_ui(interface: gr.Blocks, unrelated_tabs, tabname):
     related_tabs = []
 
     for page in ui.stored_extra_pages:
-        with gr.Tab(page.title, elem_id=f"{tabname}_{page.id_page}", elem_classes=["extra-page"]) as tab:
-            with gr.Column(elem_id=f"{tabname}_{page.id_page}_prompts", elem_classes=["extra-page-prompts"]):
+        with gr.Tab(page.title, elem_id=f"{tabname}_{page.extra_networks_tabname}", elem_classes=["extra-page"]) as tab:
+            with gr.Column(elem_id=f"{tabname}_{page.extra_networks_tabname}_prompts", elem_classes=["extra-page-prompts"]):
                 pass
 
-            elem_id = f"{tabname}_{page.id_page}_cards_html"
-            page_elem = gr.HTML('Loading...', elem_id=elem_id)
+            elem_id = f"{tabname}_{page.extra_networks_tabname}_cards_html"
+            page_elem = gr.HTML(page.create_html(tabname, empty=True), elem_id=elem_id)
             ui.pages.append(page_elem)
-
-            page_elem.change(fn=lambda: None, _js='function(){applyExtraNetworkFilter(' + quote_js(tabname) + '); return []}', inputs=[], outputs=[])
-
             editor = page.create_user_metadata_editor(ui, tabname)
             editor.create_ui()
             ui.user_metadata_editors.append(editor)
-
             related_tabs.append(tab)
 
-    edit_search = gr.Textbox('', show_label=False, elem_id=tabname+"_extra_search", elem_classes="search", placeholder="Search...", visible=False, interactive=True)
-    dropdown_sort = gr.Dropdown(choices=['Path', 'Name', 'Date Created', 'Date Modified', ], value=shared.opts.extra_networks_card_order_field, elem_id=tabname+"_extra_sort", elem_classes="sort", multiselect=False, visible=False, show_label=False, interactive=True, label=tabname+"_extra_sort_order")
-    button_sortorder = ToolButton(switch_values_symbol, elem_id=tabname+"_extra_sortorder", elem_classes=["sortorder"] + ([] if shared.opts.extra_networks_card_order == "Ascending" else ["sortReverse"]), visible=False, tooltip="Invert sort order")
-    button_refresh = gr.Button('Refresh', elem_id=tabname+"_extra_refresh", visible=False)
-    checkbox_show_dirs = gr.Checkbox(True, label='Show dirs', elem_id=tabname+"_extra_show_dirs", elem_classes="show-dirs", visible=False)
-
-    ui.button_save_preview = gr.Button('Save preview', elem_id=tabname+"_save_preview", visible=False)
-    ui.preview_target_filename = gr.Textbox('Preview save filename', elem_id=tabname+"_preview_filename", visible=False)
-
-    tab_controls = [edit_search, dropdown_sort, button_sortorder, button_refresh, checkbox_show_dirs]
+    ui.button_save_preview = gr.Button('Save preview', elem_id=f"{tabname}_save_preview", visible=False)
+    ui.preview_target_filename = gr.Textbox('Preview save filename', elem_id=f"{tabname}_preview_filename", visible=False)
 
     for tab in unrelated_tabs:
-        tab.select(fn=lambda: [gr.update(visible=False) for _ in tab_controls], _js='function(){ extraNetworksUrelatedTabSelected("' + tabname + '"); }', inputs=[], outputs=tab_controls, show_progress=False)
+        tab.select(fn=None, _js=f"function(){{extraNetworksUnrelatedTabSelected('{tabname}');}}", inputs=[], outputs=[], show_progress=False)
 
     for page, tab in zip(ui.stored_extra_pages, related_tabs):
-        allow_prompt = "true" if page.allow_prompt else "false"
-        allow_negative_prompt = "true" if page.allow_negative_prompt else "false"
-
-        jscode = 'extraNetworksTabSelected("' + tabname + '", "' + f"{tabname}_{page.id_page}_prompts" + '", ' + allow_prompt + ', ' + allow_negative_prompt + ');'
-
-        tab.select(fn=lambda: [gr.update(visible=True) for _ in tab_controls],  _js='function(){ ' + jscode + ' }', inputs=[], outputs=tab_controls, show_progress=False)
-
-    dropdown_sort.change(fn=lambda: None, _js="function(){ applyExtraNetworkSort('" + tabname + "'); }")
+        jscode = (
+            "function(){{"
+            f"extraNetworksTabSelected('{tabname}', '{tabname}_{page.extra_networks_tabname}_prompts', {str(page.allow_prompt).lower()}, {str(page.allow_negative_prompt).lower()}, '{tabname}_{page.extra_networks_tabname}');"
+            f"applyExtraNetworkFilter('{tabname}_{page.extra_networks_tabname}');"
+            "}}"
+        )
+        tab.select(fn=None, _js=jscode, inputs=[], outputs=[], show_progress=False)
+
+        def refresh():
+            for pg in ui.stored_extra_pages:
+                pg.refresh()
+            create_html()
+            return ui.pages_contents
+
+        button_refresh = gr.Button("Refresh", elem_id=f"{tabname}_{page.extra_networks_tabname}_extra_refresh_internal", visible=False)
+        button_refresh.click(fn=refresh, inputs=[], outputs=ui.pages).then(fn=lambda: None, _js="function(){ " + f"applyExtraNetworkFilter('{tabname}_{page.extra_networks_tabname}');" + " }").then(fn=lambda: None, _js='setupAllResizeHandles')
+
+    def create_html():
+        ui.pages_contents = [pg.create_html(ui.tabname) for pg in ui.stored_extra_pages]
 
     def pages_html():
         if not ui.pages_contents:
-            return refresh()
-
+            create_html()
         return ui.pages_contents
 
-    def refresh():
-        for pg in ui.stored_extra_pages:
-            pg.refresh()
-
-        ui.pages_contents = [pg.create_html(ui.tabname) for pg in ui.stored_extra_pages]
-
-        return ui.pages_contents
-
-    interface.load(fn=pages_html, inputs=[], outputs=[*ui.pages])
-    button_refresh.click(fn=refresh, inputs=[], outputs=ui.pages)
+    interface.load(fn=pages_html, inputs=[], outputs=ui.pages).then(fn=lambda: None, _js='setupAllResizeHandles')
 
     return ui
 
@@ -478,5 +759,3 @@ def save_preview(index, images, filename):
 
     for editor in ui.user_metadata_editors:
         editor.setup_ui(gallery)
-
-
diff --git a/modules/ui_extra_networks_checkpoints.py b/modules/ui_extra_networks_checkpoints.py
index 1693e71f16f..d69d144dba4 100644
--- a/modules/ui_extra_networks_checkpoints.py
+++ b/modules/ui_extra_networks_checkpoints.py
@@ -2,7 +2,6 @@
 import os
 
 from modules import shared, ui_extra_networks, sd_models
-from modules.ui_extra_networks import quote_js
 from modules.ui_extra_networks_checkpoints_user_metadata import CheckpointUserMetadataEditor
 
 
@@ -21,14 +20,17 @@ def create_item(self, name, index=None, enable_filter=True):
             return
 
         path, ext = os.path.splitext(checkpoint.filename)
+        search_terms = [self.search_terms_from_path(checkpoint.filename)]
+        if checkpoint.sha256:
+            search_terms.append(checkpoint.sha256)
         return {
             "name": checkpoint.name_for_extra,
             "filename": checkpoint.filename,
             "shorthash": checkpoint.shorthash,
             "preview": self.find_preview(path),
             "description": self.find_description(path),
-            "search_term": self.search_terms_from_path(checkpoint.filename) + " " + (checkpoint.sha256 or ""),
-            "onclick": '"' + html.escape(f"""return selectCheckpoint({quote_js(name)})""") + '"',
+            "search_terms": search_terms,
+            "onclick": html.escape(f"return selectCheckpoint({ui_extra_networks.quote_js(name)})"),
             "local_preview": f"{path}.{shared.opts.samples_format}",
             "metadata": checkpoint.metadata,
             "sort_keys": {'default': index, **self.get_sort_keys(checkpoint.filename)},
diff --git a/modules/ui_extra_networks_hypernets.py b/modules/ui_extra_networks_hypernets.py
index c96c4fa3b12..2fb4bd190a1 100644
--- a/modules/ui_extra_networks_hypernets.py
+++ b/modules/ui_extra_networks_hypernets.py
@@ -20,14 +20,16 @@ def create_item(self, name, index=None, enable_filter=True):
         path, ext = os.path.splitext(full_path)
         sha256 = sha256_from_cache(full_path, f'hypernet/{name}')
         shorthash = sha256[0:10] if sha256 else None
-
+        search_terms = [self.search_terms_from_path(path)]
+        if sha256:
+            search_terms.append(sha256)
         return {
             "name": name,
             "filename": full_path,
             "shorthash": shorthash,
             "preview": self.find_preview(path),
             "description": self.find_description(path),
-            "search_term": self.search_terms_from_path(path) + " " + (sha256 or ""),
+            "search_terms": search_terms,
             "prompt": quote_js(f""),
             "local_preview": f"{path}.preview.{shared.opts.samples_format}",
             "sort_keys": {'default': index, **self.get_sort_keys(path + ext)},
diff --git a/modules/ui_extra_networks_textual_inversion.py b/modules/ui_extra_networks_textual_inversion.py
index 1b334fda174..deb7cb8733b 100644
--- a/modules/ui_extra_networks_textual_inversion.py
+++ b/modules/ui_extra_networks_textual_inversion.py
@@ -18,13 +18,16 @@ def create_item(self, name, index=None, enable_filter=True):
             return
 
         path, ext = os.path.splitext(embedding.filename)
+        search_terms = [self.search_terms_from_path(embedding.filename)]
+        if embedding.hash:
+            search_terms.append(embedding.hash)
         return {
             "name": name,
             "filename": embedding.filename,
             "shorthash": embedding.shorthash,
             "preview": self.find_preview(path),
             "description": self.find_description(path),
-            "search_term": self.search_terms_from_path(embedding.filename) + " " + (embedding.hash or ""),
+            "search_terms": search_terms,
             "prompt": quote_js(embedding.name),
             "local_preview": f"{path}.preview.{shared.opts.samples_format}",
             "sort_keys": {'default': index, **self.get_sort_keys(embedding.filename)},
diff --git a/modules/ui_extra_networks_user_metadata.py b/modules/ui_extra_networks_user_metadata.py
index 36a807fcdf9..2ca937fd117 100644
--- a/modules/ui_extra_networks_user_metadata.py
+++ b/modules/ui_extra_networks_user_metadata.py
@@ -5,7 +5,7 @@
 
 import gradio as gr
 
-from modules import generation_parameters_copypaste, images, sysinfo, errors, ui_extra_networks
+from modules import infotext_utils, images, sysinfo, errors, ui_extra_networks
 
 
 class UserMetadataEditor:
@@ -14,7 +14,7 @@ def __init__(self, ui, tabname, page):
         self.ui = ui
         self.tabname = tabname
         self.page = page
-        self.id_part = f"{self.tabname}_{self.page.id_page}_edit_user_metadata"
+        self.id_part = f"{self.tabname}_{self.page.extra_networks_tabname}_edit_user_metadata"
 
         self.box = None
 
@@ -181,7 +181,7 @@ def save_preview(self, index, gallery, name):
         index = len(gallery) - 1 if index >= len(gallery) else index
 
         img_info = gallery[index if index >= 0 else 0]
-        image = generation_parameters_copypaste.image_from_url_text(img_info)
+        image = infotext_utils.image_from_url_text(img_info)
         geninfo, items = images.read_info_from_image(image)
 
         images.save_image_with_geninfo(image, geninfo, item["local_preview"])
diff --git a/modules/ui_gradio_extensions.py b/modules/ui_gradio_extensions.py
index 0d368f8b2c4..f5278d22f02 100644
--- a/modules/ui_gradio_extensions.py
+++ b/modules/ui_gradio_extensions.py
@@ -1,17 +1,12 @@
 import os
 import gradio as gr
 
-from modules import localization, shared, scripts
-from modules.paths import script_path, data_path, cwd
+from modules import localization, shared, scripts, util
+from modules.paths import script_path, data_path
 
 
 def webpath(fn):
-    if fn.startswith(cwd):
-        web_path = os.path.relpath(fn, cwd)
-    else:
-        web_path = os.path.abspath(fn)
-
-    return f'file={web_path}?{os.path.getmtime(fn)}'
+    return f'file={util.truncate_path(fn)}?{os.path.getmtime(fn)}'
 
 
 def javascript_html():
@@ -40,13 +35,11 @@ def stylesheet(fn):
         return f''
 
     for cssfile in scripts.list_files_with_name("style.css"):
-        if not os.path.isfile(cssfile):
-            continue
-
         head += stylesheet(cssfile)
 
-    if os.path.exists(os.path.join(data_path, "user.css")):
-        head += stylesheet(os.path.join(data_path, "user.css"))
+    user_css = os.path.join(data_path, "user.css")
+    if os.path.exists(user_css):
+        head += stylesheet(user_css)
 
     return head
 
diff --git a/modules/ui_loadsave.py b/modules/ui_loadsave.py
index 7826786ccde..2555cdb6c60 100644
--- a/modules/ui_loadsave.py
+++ b/modules/ui_loadsave.py
@@ -26,8 +26,9 @@ def __init__(self, filename):
         self.ui_defaults_review = None
 
         try:
-            if os.path.exists(self.filename):
-                self.ui_settings = self.read_from_file()
+            self.ui_settings = self.read_from_file()
+        except FileNotFoundError:
+            pass
         except Exception as e:
             self.error_loading = True
             errors.display(e, "loading settings")
@@ -144,7 +145,7 @@ def write_to_file(self, current_ui_settings):
             json.dump(current_ui_settings, file, indent=4, ensure_ascii=False)
 
     def dump_defaults(self):
-        """saves default values to a file unless tjhe file is present and there was an error loading default values at start"""
+        """saves default values to a file unless the file is present and there was an error loading default values at start"""
 
         if self.error_loading and os.path.exists(self.filename):
             return
diff --git a/modules/ui_postprocessing.py b/modules/ui_postprocessing.py
index 13d888e48d1..7261c2df8a0 100644
--- a/modules/ui_postprocessing.py
+++ b/modules/ui_postprocessing.py
@@ -1,13 +1,14 @@
 import gradio as gr
 from modules import scripts, shared, ui_common, postprocessing, call_queue, ui_toprow
-import modules.generation_parameters_copypaste as parameters_copypaste
+import modules.infotext_utils as parameters_copypaste
+from modules.ui_components import ResizeHandleRow
 
 
 def create_ui():
     dummy_component = gr.Label(visible=False)
-    tab_index = gr.State(value=0)
+    tab_index = gr.Number(value=0, visible=False)
 
-    with gr.Row(equal_height=False, variant='compact'):
+    with ResizeHandleRow(equal_height=False, variant='compact'):
         with gr.Column(variant='compact'):
             with gr.Tabs(elem_id="mode_extras"):
                 with gr.TabItem('Single Image', id="single_image", elem_id="extras_single_tab") as tab_single:
@@ -28,7 +29,7 @@ def create_ui():
             toprow.create_inline_toprow_image()
             submit = toprow.submit
 
-            result_images, html_info_x, html_info, html_log = ui_common.create_output_panel("extras", shared.opts.outdir_extras_samples)
+            output_panel = ui_common.create_output_panel("extras", shared.opts.outdir_extras_samples)
 
     tab_single.select(fn=lambda: 0, inputs=[], outputs=[tab_index])
     tab_batch.select(fn=lambda: 1, inputs=[], outputs=[tab_index])
@@ -48,9 +49,9 @@ def create_ui():
             *script_inputs
         ],
         outputs=[
-            result_images,
-            html_info_x,
-            html_log,
+            output_panel.gallery,
+            output_panel.generation_info,
+            output_panel.html_log,
         ],
         show_progress=False,
     )
diff --git a/modules/ui_prompt_styles.py b/modules/ui_prompt_styles.py
index 0d74c23fa19..f71b40c419b 100644
--- a/modules/ui_prompt_styles.py
+++ b/modules/ui_prompt_styles.py
@@ -22,9 +22,12 @@ def save_style(name, prompt, negative_prompt):
     if not name:
         return gr.update(visible=False)
 
-    style = styles.PromptStyle(name, prompt, negative_prompt)
+    existing_style = shared.prompt_styles.styles.get(name)
+    path = existing_style.path if existing_style is not None else None
+
+    style = styles.PromptStyle(name, prompt, negative_prompt, path)
     shared.prompt_styles.styles[style.name] = style
-    shared.prompt_styles.save_styles(shared.styles_filename)
+    shared.prompt_styles.save_styles()
 
     return gr.update(visible=True)
 
@@ -34,7 +37,7 @@ def delete_style(name):
         return
 
     shared.prompt_styles.styles.pop(name, None)
-    shared.prompt_styles.save_styles(shared.styles_filename)
+    shared.prompt_styles.save_styles()
 
     return '', '', ''
 
@@ -64,7 +67,7 @@ def __init__(self, tabname, main_ui_prompt, main_ui_negative_prompt):
             with gr.Row():
                 self.selection = gr.Dropdown(label="Styles", elem_id=f"{tabname}_styles_edit_select", choices=list(shared.prompt_styles.styles), value=[], allow_custom_value=True, info="Styles allow you to add custom text to prompt. Use the {prompt} token in style text, and it will be replaced with user's prompt when applying style. Otherwise, style's text will be added to the end of the prompt.")
                 ui_common.create_refresh_button([self.dropdown, self.selection], shared.prompt_styles.reload, lambda: {"choices": list(shared.prompt_styles.styles)}, f"refresh_{tabname}_styles")
-                self.materialize = ui_components.ToolButton(value=styles_materialize_symbol, elem_id=f"{tabname}_style_apply_dialog", tooltip="Apply all selected styles from the style selction dropdown in main UI to the prompt.")
+                self.materialize = ui_components.ToolButton(value=styles_materialize_symbol, elem_id=f"{tabname}_style_apply_dialog", tooltip="Apply all selected styles from the style selection dropdown in main UI to the prompt.")
                 self.copy = ui_components.ToolButton(value=styles_copy_symbol, elem_id=f"{tabname}_style_copy", tooltip="Copy main UI prompt to style.")
 
             with gr.Row():
diff --git a/modules/ui_settings.py b/modules/ui_settings.py
index e054d00ab04..d17ef1d9558 100644
--- a/modules/ui_settings.py
+++ b/modules/ui_settings.py
@@ -98,6 +98,9 @@ def run_settings_single(self, value, key):
 
         return get_value_for_setting(key), opts.dumpjson()
 
+    def register_settings(self):
+        script_callbacks.ui_settings_callback()
+
     def create_ui(self, loadsave, dummy_component):
         self.components = []
         self.component_dict = {}
@@ -105,7 +108,6 @@ def create_ui(self, loadsave, dummy_component):
 
         shared.settings_components = self.component_dict
 
-        script_callbacks.ui_settings_callback()
         opts.reorder()
 
         with gr.Blocks(analytics_enabled=False) as settings_interface:
diff --git a/modules/ui_tempdir.py b/modules/ui_tempdir.py
index 85015db56b5..ecd6bdec355 100644
--- a/modules/ui_tempdir.py
+++ b/modules/ui_tempdir.py
@@ -35,12 +35,9 @@ def save_pil_to_file(self, pil_image, dir=None, format="png"):
     already_saved_as = getattr(pil_image, 'already_saved_as', None)
     if already_saved_as and os.path.isfile(already_saved_as):
         register_tmp_file(shared.demo, already_saved_as)
-        filename = already_saved_as
-
-        if not shared.opts.save_images_add_number:
-            filename += f'?{os.path.getmtime(already_saved_as)}'
-
-        return filename
+        filename_with_mtime = f'{already_saved_as}?{os.path.getmtime(already_saved_as)}'
+        register_tmp_file(shared.demo, filename_with_mtime)
+        return filename_with_mtime
 
     if shared.opts.temp_dir != "":
         dir = shared.opts.temp_dir
@@ -86,3 +83,18 @@ def cleanup_tmpdr():
 
             filename = os.path.join(root, name)
             os.remove(filename)
+
+
+def is_gradio_temp_path(path):
+    """
+    Check if the path is a temp dir used by gradio
+    """
+    path = Path(path)
+    if shared.opts.temp_dir and path.is_relative_to(shared.opts.temp_dir):
+        return True
+    if gradio_temp_dir := os.environ.get("GRADIO_TEMP_DIR"):
+        if path.is_relative_to(gradio_temp_dir):
+            return True
+    if path.is_relative_to(Path(tempfile.gettempdir()) / "gradio"):
+        return True
+    return False
diff --git a/modules/ui_toprow.py b/modules/ui_toprow.py
index 88838f97749..dc3c3aa3837 100644
--- a/modules/ui_toprow.py
+++ b/modules/ui_toprow.py
@@ -17,6 +17,7 @@ class Toprow:
     button_deepbooru = None
 
     interrupt = None
+    interrupting = None
     skip = None
     submit = None
 
@@ -79,11 +80,11 @@ def create_inline_toprow_image(self):
     def create_prompts(self):
         with gr.Column(elem_id=f"{self.id_part}_prompt_container", elem_classes=["prompt-container-compact"] if self.is_compact else [], scale=6):
             with gr.Row(elem_id=f"{self.id_part}_prompt_row", elem_classes=["prompt-row"]):
-                self.prompt = gr.Textbox(label="Prompt", elem_id=f"{self.id_part}_prompt", show_label=False, lines=3, placeholder="Prompt (press Ctrl+Enter or Alt+Enter to generate)", elem_classes=["prompt"])
+                self.prompt = gr.Textbox(label="Prompt", elem_id=f"{self.id_part}_prompt", show_label=False, lines=3, placeholder="Prompt\n(Press Ctrl+Enter to generate, Alt+Enter to skip, Esc to interrupt)", elem_classes=["prompt"])
                 self.prompt_img = gr.File(label="", elem_id=f"{self.id_part}_prompt_image", file_count="single", type="binary", visible=False)
 
             with gr.Row(elem_id=f"{self.id_part}_neg_prompt_row", elem_classes=["prompt-row"]):
-                self.negative_prompt = gr.Textbox(label="Negative prompt", elem_id=f"{self.id_part}_neg_prompt", show_label=False, lines=3, placeholder="Negative prompt (press Ctrl+Enter or Alt+Enter to generate)", elem_classes=["prompt"])
+                self.negative_prompt = gr.Textbox(label="Negative prompt", elem_id=f"{self.id_part}_neg_prompt", show_label=False, lines=3, placeholder="Negative prompt\n(Press Ctrl+Enter to generate, Alt+Enter to skip, Esc to interrupt)", elem_classes=["prompt"])
 
         self.prompt_img.change(
             fn=modules.images.image_data,
@@ -96,21 +97,21 @@ def create_submit_box(self):
         with gr.Row(elem_id=f"{self.id_part}_generate_box", elem_classes=["generate-box"] + (["generate-box-compact"] if self.is_compact else []), render=not self.is_compact) as submit_box:
             self.submit_box = submit_box
 
-            self.interrupt = gr.Button('Interrupt', elem_id=f"{self.id_part}_interrupt", elem_classes="generate-box-interrupt")
-            self.skip = gr.Button('Skip', elem_id=f"{self.id_part}_skip", elem_classes="generate-box-skip")
-            self.submit = gr.Button('Generate', elem_id=f"{self.id_part}_generate", variant='primary')
+            self.interrupt = gr.Button('Interrupt', elem_id=f"{self.id_part}_interrupt", elem_classes="generate-box-interrupt", tooltip="End generation immediately or after completing current batch")
+            self.skip = gr.Button('Skip', elem_id=f"{self.id_part}_skip", elem_classes="generate-box-skip", tooltip="Stop generation of current batch and continues onto next batch")
+            self.interrupting = gr.Button('Interrupting...', elem_id=f"{self.id_part}_interrupting", elem_classes="generate-box-interrupting", tooltip="Interrupting generation...")
+            self.submit = gr.Button('Generate', elem_id=f"{self.id_part}_generate", variant='primary', tooltip="Right click generate forever menu")
 
-            self.skip.click(
-                fn=lambda: shared.state.skip(),
-                inputs=[],
-                outputs=[],
-            )
+            def interrupt_function():
+                if not shared.state.stopping_generation and shared.state.job_count > 1 and shared.opts.interrupt_after_current:
+                    shared.state.stop_generating()
+                    gr.Info("Generation will stop after finishing this image, click again to stop immediately.")
+                else:
+                    shared.state.interrupt()
 
-            self.interrupt.click(
-                fn=lambda: shared.state.interrupt(),
-                inputs=[],
-                outputs=[],
-            )
+            self.skip.click(fn=shared.state.skip)
+            self.interrupt.click(fn=interrupt_function, _js='function(){ showSubmitInterruptingPlaceholder("' + self.id_part + '"); }')
+            self.interrupting.click(fn=interrupt_function)
 
     def create_tools_row(self):
         with gr.Row(elem_id=f"{self.id_part}_tools"):
@@ -126,9 +127,9 @@ def create_tools_row(self):
 
             self.restore_progress_button = ToolButton(value=restore_progress_symbol, elem_id=f"{self.id_part}_restore_progress", visible=False, tooltip="Restore progress")
 
-            self.token_counter = gr.HTML(value="0/75", elem_id=f"{self.id_part}_token_counter", elem_classes=["token-counter"])
+            self.token_counter = gr.HTML(value="0/75", elem_id=f"{self.id_part}_token_counter", elem_classes=["token-counter"], visible=False)
             self.token_button = gr.Button(visible=False, elem_id=f"{self.id_part}_token_button")
-            self.negative_token_counter = gr.HTML(value="0/75", elem_id=f"{self.id_part}_negative_token_counter", elem_classes=["token-counter"])
+            self.negative_token_counter = gr.HTML(value="0/75", elem_id=f"{self.id_part}_negative_token_counter", elem_classes=["token-counter"], visible=False)
             self.negative_token_button = gr.Button(visible=False, elem_id=f"{self.id_part}_negative_token_button")
 
             self.clear_prompt_button.click(
diff --git a/modules/upscaler.py b/modules/upscaler.py
index b256e085b6d..3aee69db8d2 100644
--- a/modules/upscaler.py
+++ b/modules/upscaler.py
@@ -98,6 +98,9 @@ def __init__(self, name: str, path: str, upscaler: Upscaler = None, scale: int =
         self.scale = scale
         self.model = model
 
+    def __repr__(self):
+        return f""
+
 
 class UpscalerNone(Upscaler):
     name = "None"
diff --git a/modules/upscaler_utils.py b/modules/upscaler_utils.py
new file mode 100644
index 00000000000..17223ca0da1
--- /dev/null
+++ b/modules/upscaler_utils.py
@@ -0,0 +1,188 @@
+import logging
+from typing import Callable
+
+import numpy as np
+import torch
+import tqdm
+from PIL import Image
+
+from modules import devices, images, shared, torch_utils
+
+logger = logging.getLogger(__name__)
+
+
+def pil_image_to_torch_bgr(img: Image.Image) -> torch.Tensor:
+    img = np.array(img.convert("RGB"))
+    img = img[:, :, ::-1]  # flip RGB to BGR
+    img = np.transpose(img, (2, 0, 1))  # HWC to CHW
+    img = np.ascontiguousarray(img) / 255  # Rescale to [0, 1]
+    return torch.from_numpy(img)
+
+
+def torch_bgr_to_pil_image(tensor: torch.Tensor) -> Image.Image:
+    if tensor.ndim == 4:
+        # If we're given a tensor with a batch dimension, squeeze it out
+        # (but only if it's a batch of size 1).
+        if tensor.shape[0] != 1:
+            raise ValueError(f"{tensor.shape} does not describe a BCHW tensor")
+        tensor = tensor.squeeze(0)
+    assert tensor.ndim == 3, f"{tensor.shape} does not describe a CHW tensor"
+    # TODO: is `tensor.float().cpu()...numpy()` the most efficient idiom?
+    arr = tensor.float().cpu().clamp_(0, 1).numpy()  # clamp
+    arr = 255.0 * np.moveaxis(arr, 0, 2)  # CHW to HWC, rescale
+    arr = arr.round().astype(np.uint8)
+    arr = arr[:, :, ::-1]  # flip BGR to RGB
+    return Image.fromarray(arr, "RGB")
+
+
+def upscale_pil_patch(model, img: Image.Image) -> Image.Image:
+    """
+    Upscale a given PIL image using the given model.
+    """
+    param = torch_utils.get_param(model)
+
+    with torch.no_grad():
+        tensor = pil_image_to_torch_bgr(img).unsqueeze(0)  # add batch dimension
+        tensor = tensor.to(device=param.device, dtype=param.dtype)
+        with devices.without_autocast():
+            return torch_bgr_to_pil_image(model(tensor))
+
+
+def upscale_with_model(
+    model: Callable[[torch.Tensor], torch.Tensor],
+    img: Image.Image,
+    *,
+    tile_size: int,
+    tile_overlap: int = 0,
+    desc="tiled upscale",
+) -> Image.Image:
+    if tile_size <= 0:
+        logger.debug("Upscaling %s without tiling", img)
+        output = upscale_pil_patch(model, img)
+        logger.debug("=> %s", output)
+        return output
+
+    grid = images.split_grid(img, tile_size, tile_size, tile_overlap)
+    newtiles = []
+
+    with tqdm.tqdm(total=grid.tile_count, desc=desc, disable=not shared.opts.enable_upscale_progressbar) as p:
+        for y, h, row in grid.tiles:
+            newrow = []
+            for x, w, tile in row:
+                output = upscale_pil_patch(model, tile)
+                scale_factor = output.width // tile.width
+                newrow.append([x * scale_factor, w * scale_factor, output])
+                p.update(1)
+            newtiles.append([y * scale_factor, h * scale_factor, newrow])
+
+    newgrid = images.Grid(
+        newtiles,
+        tile_w=grid.tile_w * scale_factor,
+        tile_h=grid.tile_h * scale_factor,
+        image_w=grid.image_w * scale_factor,
+        image_h=grid.image_h * scale_factor,
+        overlap=grid.overlap * scale_factor,
+    )
+    return images.combine_grid(newgrid)
+
+
+def tiled_upscale_2(
+    img: torch.Tensor,
+    model,
+    *,
+    tile_size: int,
+    tile_overlap: int,
+    scale: int,
+    device: torch.device,
+    desc="Tiled upscale",
+):
+    # Alternative implementation of `upscale_with_model` originally used by
+    # SwinIR and ScuNET.  It differs from `upscale_with_model` in that tiling and
+    # weighting is done in PyTorch space, as opposed to `images.Grid` doing it in
+    # Pillow space without weighting.
+
+    b, c, h, w = img.size()
+    tile_size = min(tile_size, h, w)
+
+    if tile_size <= 0:
+        logger.debug("Upscaling %s without tiling", img.shape)
+        return model(img)
+
+    stride = tile_size - tile_overlap
+    h_idx_list = list(range(0, h - tile_size, stride)) + [h - tile_size]
+    w_idx_list = list(range(0, w - tile_size, stride)) + [w - tile_size]
+    result = torch.zeros(
+        b,
+        c,
+        h * scale,
+        w * scale,
+        device=device,
+        dtype=img.dtype,
+    )
+    weights = torch.zeros_like(result)
+    logger.debug("Upscaling %s to %s with tiles", img.shape, result.shape)
+    with tqdm.tqdm(total=len(h_idx_list) * len(w_idx_list), desc=desc, disable=not shared.opts.enable_upscale_progressbar) as pbar:
+        for h_idx in h_idx_list:
+            if shared.state.interrupted or shared.state.skipped:
+                break
+
+            for w_idx in w_idx_list:
+                if shared.state.interrupted or shared.state.skipped:
+                    break
+
+                # Only move this patch to the device if it's not already there.
+                in_patch = img[
+                    ...,
+                    h_idx : h_idx + tile_size,
+                    w_idx : w_idx + tile_size,
+                ].to(device=device)
+
+                out_patch = model(in_patch)
+
+                result[
+                    ...,
+                    h_idx * scale : (h_idx + tile_size) * scale,
+                    w_idx * scale : (w_idx + tile_size) * scale,
+                ].add_(out_patch)
+
+                out_patch_mask = torch.ones_like(out_patch)
+
+                weights[
+                    ...,
+                    h_idx * scale : (h_idx + tile_size) * scale,
+                    w_idx * scale : (w_idx + tile_size) * scale,
+                ].add_(out_patch_mask)
+
+                pbar.update(1)
+
+    output = result.div_(weights)
+
+    return output
+
+
+def upscale_2(
+    img: Image.Image,
+    model,
+    *,
+    tile_size: int,
+    tile_overlap: int,
+    scale: int,
+    desc: str,
+):
+    """
+    Convenience wrapper around `tiled_upscale_2` that handles PIL images.
+    """
+    param = torch_utils.get_param(model)
+    tensor = pil_image_to_torch_bgr(img).to(dtype=param.dtype).unsqueeze(0)  # add batch dimension
+
+    with torch.no_grad():
+        output = tiled_upscale_2(
+            tensor,
+            model,
+            tile_size=tile_size,
+            tile_overlap=tile_overlap,
+            scale=scale,
+            desc=desc,
+            device=param.device,
+        )
+    return torch_bgr_to_pil_image(output)
diff --git a/modules/util.py b/modules/util.py
index 60afc0670c7..8d1aea44f5c 100644
--- a/modules/util.py
+++ b/modules/util.py
@@ -2,7 +2,7 @@
 import re
 
 from modules import shared
-from modules.paths_internal import script_path
+from modules.paths_internal import script_path, cwd
 
 
 def natural_sort_key(s, regex=re.compile('([0-9]+)')):
@@ -21,11 +21,11 @@ def html_path(filename):
 def html(filename):
     path = html_path(filename)
 
-    if os.path.exists(path):
+    try:
         with open(path, encoding="utf8") as file:
             return file.read()
-
-    return ""
+    except OSError:
+        return ""
 
 
 def walk_files(path, allowed_extensions=None):
@@ -42,7 +42,7 @@ def walk_files(path, allowed_extensions=None):
         for filename in sorted(files, key=natural_sort_key):
             if allowed_extensions is not None:
                 _, ext = os.path.splitext(filename)
-                if ext not in allowed_extensions:
+                if ext.lower() not in allowed_extensions:
                     continue
 
             if not shared.opts.list_hidden_files and ("/." in root or "\\." in root):
@@ -56,3 +56,83 @@ def ldm_print(*args, **kwargs):
         return
 
     print(*args, **kwargs)
+
+
+def truncate_path(target_path, base_path=cwd):
+    abs_target, abs_base = os.path.abspath(target_path), os.path.abspath(base_path)
+    try:
+        if os.path.commonpath([abs_target, abs_base]) == abs_base:
+            return os.path.relpath(abs_target, abs_base)
+    except ValueError:
+        pass
+    return abs_target
+
+
+class MassFileListerCachedDir:
+    """A class that caches file metadata for a specific directory."""
+
+    def __init__(self, dirname):
+        self.files = None
+        self.files_cased = None
+        self.dirname = dirname
+
+        stats = ((x.name, x.stat(follow_symlinks=False)) for x in os.scandir(self.dirname))
+        files = [(n, s.st_mtime, s.st_ctime) for n, s in stats]
+        self.files = {x[0].lower(): x for x in files}
+        self.files_cased = {x[0]: x for x in files}
+
+
+class MassFileLister:
+    """A class that provides a way to check for the existence and mtime/ctile of files without doing more than one stat call per file."""
+
+    def __init__(self):
+        self.cached_dirs = {}
+
+    def find(self, path):
+        """
+        Find the metadata for a file at the given path.
+
+        Returns:
+            tuple or None: A tuple of (name, mtime, ctime) if the file exists, or None if it does not.
+        """
+
+        dirname, filename = os.path.split(path)
+
+        cached_dir = self.cached_dirs.get(dirname)
+        if cached_dir is None:
+            cached_dir = MassFileListerCachedDir(dirname)
+            self.cached_dirs[dirname] = cached_dir
+
+        stats = cached_dir.files_cased.get(filename)
+        if stats is not None:
+            return stats
+
+        stats = cached_dir.files.get(filename.lower())
+        if stats is None:
+            return None
+
+        try:
+            os_stats = os.stat(path, follow_symlinks=False)
+            return filename, os_stats.st_mtime, os_stats.st_ctime
+        except Exception:
+            return None
+
+    def exists(self, path):
+        """Check if a file exists at the given path."""
+
+        return self.find(path) is not None
+
+    def mctime(self, path):
+        """
+        Get the modification and creation times for a file at the given path.
+
+        Returns:
+            tuple: A tuple of (mtime, ctime) if the file exists, or (0, 0) if it does not.
+        """
+
+        stats = self.find(path)
+        return (0, 0) if stats is None else stats[1:3]
+
+    def reset(self):
+        """Clear the cache of all directories."""
+        self.cached_dirs.clear()
diff --git a/modules/xlmr.py b/modules/xlmr.py
index a407a3cade8..319771b7bf0 100644
--- a/modules/xlmr.py
+++ b/modules/xlmr.py
@@ -5,6 +5,9 @@
 from transformers import XLMRobertaModel,XLMRobertaTokenizer
 from typing import Optional
 
+from modules import torch_utils
+
+
 class BertSeriesConfig(BertConfig):
     def __init__(self, vocab_size=30522, hidden_size=768, num_hidden_layers=12, num_attention_heads=12, intermediate_size=3072, hidden_act="gelu", hidden_dropout_prob=0.1, attention_probs_dropout_prob=0.1, max_position_embeddings=512, type_vocab_size=2, initializer_range=0.02, layer_norm_eps=1e-12, pad_token_id=0, position_embedding_type="absolute", use_cache=True, classifier_dropout=None,project_dim=512, pooler_fn="average",learn_encoder=False,model_type='bert',**kwargs):
 
@@ -62,7 +65,7 @@ def __init__(self, config=None, **kargs):
         self.post_init()
 
     def encode(self,c):
-        device = next(self.parameters()).device
+        device = torch_utils.get_param(self).device
         text = self.tokenizer(c,
                         truncation=True,
                         max_length=77,
diff --git a/modules/xlmr_m18.py b/modules/xlmr_m18.py
index a727e865529..f60555049f5 100644
--- a/modules/xlmr_m18.py
+++ b/modules/xlmr_m18.py
@@ -4,6 +4,8 @@
 from transformers.models.xlm_roberta.configuration_xlm_roberta import XLMRobertaConfig
 from transformers import XLMRobertaModel,XLMRobertaTokenizer
 from typing import Optional
+from modules import torch_utils
+
 
 class BertSeriesConfig(BertConfig):
     def __init__(self, vocab_size=30522, hidden_size=768, num_hidden_layers=12, num_attention_heads=12, intermediate_size=3072, hidden_act="gelu", hidden_dropout_prob=0.1, attention_probs_dropout_prob=0.1, max_position_embeddings=512, type_vocab_size=2, initializer_range=0.02, layer_norm_eps=1e-12, pad_token_id=0, position_embedding_type="absolute", use_cache=True, classifier_dropout=None,project_dim=512, pooler_fn="average",learn_encoder=False,model_type='bert',**kwargs):
@@ -68,7 +70,7 @@ def __init__(self, config=None, **kargs):
         self.post_init()
 
     def encode(self,c):
-        device = next(self.parameters()).device
+        device = torch_utils.get_param(self).device
         text = self.tokenizer(c,
                         truncation=True,
                         max_length=77,
diff --git a/modules/xpu_specific.py b/modules/xpu_specific.py
index d8da94a0efd..2971dbc3cf5 100644
--- a/modules/xpu_specific.py
+++ b/modules/xpu_specific.py
@@ -27,11 +27,90 @@ def torch_xpu_gc():
 
 has_xpu = check_for_xpu()
 
+
+# Arc GPU cannot allocate a single block larger than 4GB: https://github.com/intel/compute-runtime/issues/627
+# Here we implement a slicing algorithm to split large batch size into smaller chunks,
+# so that SDPA of each chunk wouldn't require any allocation larger than ARC_SINGLE_ALLOCATION_LIMIT.
+# The heuristic limit (TOTAL_VRAM // 8) is tuned for Intel Arc A770 16G and Arc A750 8G,
+# which is the best trade-off between VRAM usage and performance.
+ARC_SINGLE_ALLOCATION_LIMIT = {}
+orig_sdp_attn_func = torch.nn.functional.scaled_dot_product_attention
+def torch_xpu_scaled_dot_product_attention(
+    query, key, value, attn_mask=None, dropout_p=0.0, is_causal=False, *args, **kwargs
+):
+    # cast to same dtype first
+    key = key.to(query.dtype)
+    value = value.to(query.dtype)
+    if attn_mask is not None and attn_mask.dtype != torch.bool:
+        attn_mask = attn_mask.to(query.dtype)
+
+    N = query.shape[:-2]  # Batch size
+    L = query.size(-2)  # Target sequence length
+    E = query.size(-1)  # Embedding dimension of the query and key
+    S = key.size(-2)  # Source sequence length
+    Ev = value.size(-1)  # Embedding dimension of the value
+
+    total_batch_size = torch.numel(torch.empty(N))
+    device_id = query.device.index
+    if device_id not in ARC_SINGLE_ALLOCATION_LIMIT:
+        ARC_SINGLE_ALLOCATION_LIMIT[device_id] = min(torch.xpu.get_device_properties(device_id).total_memory // 8, 4 * 1024 * 1024 * 1024)
+    batch_size_limit = max(1, ARC_SINGLE_ALLOCATION_LIMIT[device_id] // (L * S * query.element_size()))
+
+    if total_batch_size <= batch_size_limit:
+        return orig_sdp_attn_func(
+            query,
+            key,
+            value,
+            attn_mask,
+            dropout_p,
+            is_causal,
+            *args, **kwargs
+        )
+
+    query = torch.reshape(query, (-1, L, E))
+    key = torch.reshape(key, (-1, S, E))
+    value = torch.reshape(value, (-1, S, Ev))
+    if attn_mask is not None:
+        attn_mask = attn_mask.view(-1, L, S)
+    chunk_count = (total_batch_size + batch_size_limit - 1) // batch_size_limit
+    outputs = []
+    for i in range(chunk_count):
+        attn_mask_chunk = (
+            None
+            if attn_mask is None
+            else attn_mask[i * batch_size_limit : (i + 1) * batch_size_limit, :, :]
+        )
+        chunk_output = orig_sdp_attn_func(
+            query[i * batch_size_limit : (i + 1) * batch_size_limit, :, :],
+            key[i * batch_size_limit : (i + 1) * batch_size_limit, :, :],
+            value[i * batch_size_limit : (i + 1) * batch_size_limit, :, :],
+            attn_mask_chunk,
+            dropout_p,
+            is_causal,
+            *args, **kwargs
+        )
+        outputs.append(chunk_output)
+    result = torch.cat(outputs, dim=0)
+    return torch.reshape(result, (*N, L, Ev))
+
+
+def is_xpu_device(device: str | torch.device = None):
+    if device is None:
+        return False
+    if isinstance(device, str):
+        return device.startswith("xpu")
+    return device.type == "xpu"
+
+
 if has_xpu:
-    # W/A for https://github.com/intel/intel-extension-for-pytorch/issues/452: torch.Generator API doesn't support XPU device
-    CondFunc('torch.Generator',
-        lambda orig_func, device=None: torch.xpu.Generator(device),
-        lambda orig_func, device=None: device is not None and device.type == "xpu")
+    try:
+        # torch.Generator supports "xpu" device since 2.1
+        torch.Generator("xpu")
+    except RuntimeError:
+        # W/A for https://github.com/intel/intel-extension-for-pytorch/issues/452: torch.Generator API doesn't support XPU device (for torch < 2.1)
+        CondFunc('torch.Generator',
+            lambda orig_func, device=None: torch.xpu.Generator(device),
+            lambda orig_func, device=None: is_xpu_device(device))
 
     # W/A for some OPs that could not handle different input dtypes
     CondFunc('torch.nn.functional.layer_norm',
@@ -55,5 +134,5 @@ def torch_xpu_gc():
         lambda orig_func, tensors, dim=0, out=None: orig_func([t.to(tensors[0].dtype) for t in tensors], dim=dim, out=out),
         lambda orig_func, tensors, dim=0, out=None: not all(t.dtype == tensors[0].dtype for t in tensors))
     CondFunc('torch.nn.functional.scaled_dot_product_attention',
-        lambda orig_func, query, key, value, attn_mask=None, dropout_p=0.0, is_causal=False: orig_func(query, key.to(query.dtype), value.to(query.dtype), attn_mask, dropout_p, is_causal),
-        lambda orig_func, query, key, value, attn_mask=None, dropout_p=0.0, is_causal=False: query.dtype != key.dtype or query.dtype != value.dtype)
+        lambda orig_func, *args, **kwargs: torch_xpu_scaled_dot_product_attention(*args, **kwargs),
+        lambda orig_func, query, *args, **kwargs: query.is_xpu)
diff --git a/requirements.txt b/requirements.txt
index 80b438455ce..731a1be7d9b 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -2,12 +2,11 @@ GitPython
 Pillow
 accelerate
 
-basicsr
 blendmodes
 clean-fid
 einops
+facexlib
 fastapi>=0.90.1
-gfpgan
 gradio==3.41.2
 inflection
 jsonmerge
@@ -20,13 +19,11 @@ open-clip-torch
 piexif
 psutil
 pytorch_lightning
-realesrgan
 requests
 resize-right
 
 safetensors
 scikit-image>=0.19
-timm
 tomesd
 torch
 torchdiffeq
diff --git a/requirements_npu.txt b/requirements_npu.txt
new file mode 100644
index 00000000000..5e6a43646a0
--- /dev/null
+++ b/requirements_npu.txt
@@ -0,0 +1,4 @@
+cloudpickle
+decorator
+synr==0.5.0
+tornado
diff --git a/requirements_versions.txt b/requirements_versions.txt
index cb7403a9d4b..5e30b5ea18b 100644
--- a/requirements_versions.txt
+++ b/requirements_versions.txt
@@ -1,29 +1,27 @@
 GitPython==3.1.32
 Pillow==9.5.0
 accelerate==0.21.0
-basicsr==1.4.2
 blendmodes==2022
 clean-fid==0.1.35
 einops==0.4.1
+facexlib==0.3.0
 fastapi==0.94.0
-gfpgan==1.3.8
 gradio==3.41.2
 httpcore==0.15
 inflection==0.5.1
 jsonmerge==1.8.0
 kornia==0.6.7
 lark==1.1.2
-numpy==1.23.5
+numpy==1.26.2
 omegaconf==2.2.3
 open-clip-torch==2.20.0
 piexif==1.1.3
 psutil==5.9.5
 pytorch_lightning==1.9.4
-realesrgan==0.3.0
 resize-right==0.0.2
-safetensors==0.3.1
+safetensors==0.4.2
 scikit-image==0.21.0
-timm==0.9.2
+spandrel==0.1.6
 tomesd==0.1.3
 torch
 torchdiffeq==0.2.3
diff --git a/script.js b/script.js
index c0e678ea702..f069b1ef002 100644
--- a/script.js
+++ b/script.js
@@ -121,16 +121,22 @@ document.addEventListener("DOMContentLoaded", function() {
 });
 
 /**
- * Add a ctrl+enter as a shortcut to start a generation
+ * Add keyboard shortcuts:
+ * Ctrl+Enter to start/restart a generation
+ * Alt/Option+Enter to skip a generation
+ * Esc to interrupt a generation
  */
 document.addEventListener('keydown', function(e) {
     const isEnter = e.key === 'Enter' || e.keyCode === 13;
-    const isModifierKey = e.metaKey || e.ctrlKey || e.altKey;
+    const isCtrlKey = e.metaKey || e.ctrlKey;
+    const isAltKey = e.altKey;
+    const isEsc = e.key === 'Escape';
 
-    const interruptButton = get_uiCurrentTabContent().querySelector('button[id$=_interrupt]');
     const generateButton = get_uiCurrentTabContent().querySelector('button[id$=_generate]');
+    const interruptButton = get_uiCurrentTabContent().querySelector('button[id$=_interrupt]');
+    const skipButton = get_uiCurrentTabContent().querySelector('button[id$=_skip]');
 
-    if (isEnter && isModifierKey) {
+    if (isCtrlKey && isEnter) {
         if (interruptButton.style.display === 'block') {
             interruptButton.click();
             const callback = (mutationList) => {
@@ -150,6 +156,23 @@ document.addEventListener('keydown', function(e) {
         }
         e.preventDefault();
     }
+
+    if (isAltKey && isEnter) {
+        skipButton.click();
+        e.preventDefault();
+    }
+
+    if (isEsc) {
+        const globalPopup = document.querySelector('.global-popup');
+        const lightboxModal = document.querySelector('#lightboxModal');
+        if (!globalPopup || globalPopup.style.display === 'none') {
+            if (document.activeElement === lightboxModal) return;
+            if (interruptButton.style.display === 'block') {
+                interruptButton.click();
+                e.preventDefault();
+            }
+        }
+    }
 });
 
 /**
diff --git a/scripts/loopback.py b/scripts/loopback.py
index 2d5feaf9b26..800ee882a16 100644
--- a/scripts/loopback.py
+++ b/scripts/loopback.py
@@ -95,7 +95,7 @@ def calculate_denoising_strength(loop):
                 processed = processing.process_images(p)
 
                 # Generation cancelled.
-                if state.interrupted:
+                if state.interrupted or state.stopping_generation:
                     break
 
                 if initial_seed is None:
@@ -122,8 +122,8 @@ def calculate_denoising_strength(loop):
 
             p.inpainting_fill = original_inpainting_fill
 
-            if state.interrupted:
-                    break
+            if state.interrupted or state.stopping_generation:
+                break
 
         if len(history) > 1:
             grid = images.image_grid(history, rows=1)
diff --git a/scripts/outpainting_mk_2.py b/scripts/outpainting_mk_2.py
index c98ab48098e..5df9dff9c48 100644
--- a/scripts/outpainting_mk_2.py
+++ b/scripts/outpainting_mk_2.py
@@ -102,7 +102,7 @@ def _get_masked_window_rgb(np_mask_grey, hardness=1.):
     shaped_noise_fft = _fft2(noise_rgb)
     shaped_noise_fft[:, :, :] = np.absolute(shaped_noise_fft[:, :, :]) ** 2 * (src_dist ** noise_q) * src_phase  # perform the actual shaping
 
-    brightness_variation = 0.  # color_variation # todo: temporarily tieing brightness variation to color variation for now
+    brightness_variation = 0.  # color_variation # todo: temporarily tying brightness variation to color variation for now
     contrast_adjusted_np_src = _np_src_image[:] * (brightness_variation + 1.) - brightness_variation * 2.
 
     # scikit-image is used for histogram matching, very convenient!
diff --git a/scripts/postprocessing_caption.py b/scripts/postprocessing_caption.py
index 243e3ad9c62..5592a89870e 100644
--- a/scripts/postprocessing_caption.py
+++ b/scripts/postprocessing_caption.py
@@ -4,7 +4,7 @@
 
 class ScriptPostprocessingCeption(scripts_postprocessing.ScriptPostprocessing):
     name = "Caption"
-    order = 4000
+    order = 4040
 
     def ui(self):
         with ui_components.InputAccordion(False, label="Caption") as enable:
@@ -25,6 +25,6 @@ def process(self, pp: scripts_postprocessing.PostprocessedImage, enable, option)
             captions.append(deepbooru.model.tag(pp.image))
 
         if "BLIP" in option:
-            captions.append(shared.interrogator.generate_caption(pp.image))
+            captions.append(shared.interrogator.interrogate(pp.image.convert("RGB")))
 
         pp.caption = ", ".join([x for x in captions if x])
diff --git a/scripts/postprocessing_create_flipped_copies.py b/scripts/postprocessing_create_flipped_copies.py
index 3425571dc3b..b673003b6ea 100644
--- a/scripts/postprocessing_create_flipped_copies.py
+++ b/scripts/postprocessing_create_flipped_copies.py
@@ -6,7 +6,7 @@
 
 class ScriptPostprocessingCreateFlippedCopies(scripts_postprocessing.ScriptPostprocessing):
     name = "Create flipped copies"
-    order = 4000
+    order = 4030
 
     def ui(self):
         with ui_components.InputAccordion(False, label="Create flipped copies") as enable:
diff --git a/scripts/postprocessing_focal_crop.py b/scripts/postprocessing_focal_crop.py
index d3baf29878a..cff1dbc5470 100644
--- a/scripts/postprocessing_focal_crop.py
+++ b/scripts/postprocessing_focal_crop.py
@@ -7,7 +7,7 @@
 
 class ScriptPostprocessingFocalCrop(scripts_postprocessing.ScriptPostprocessing):
     name = "Auto focal point crop"
-    order = 4000
+    order = 4010
 
     def ui(self):
         with ui_components.InputAccordion(False, label="Auto focal point crop") as enable:
diff --git a/scripts/postprocessing_split_oversized.py b/scripts/postprocessing_split_oversized.py
index c4a03160fc6..133e199b838 100644
--- a/scripts/postprocessing_split_oversized.py
+++ b/scripts/postprocessing_split_oversized.py
@@ -61,7 +61,7 @@ def process(self, pp: scripts_postprocessing.PostprocessedImage, enable, split_t
             ratio = (pp.image.height * width) / (pp.image.width * height)
             inverse_xy = True
 
-        if ratio >= 1.0 and ratio > split_threshold:
+        if ratio >= 1.0 or ratio > split_threshold:
             return
 
         result, *others = split_pic(pp.image, inverse_xy, width, height, overlap_ratio)
diff --git a/scripts/postprocessing_upscale.py b/scripts/postprocessing_upscale.py
index ed709688de4..e269682d0f3 100644
--- a/scripts/postprocessing_upscale.py
+++ b/scripts/postprocessing_upscale.py
@@ -15,7 +15,7 @@ class ScriptPostprocessingUpscale(scripts_postprocessing.ScriptPostprocessing):
     order = 1000
 
     def ui(self):
-        selected_tab = gr.State(value=0)
+        selected_tab = gr.Number(value=0, visible=False)
 
         with gr.Column():
             with FormRow():
@@ -26,8 +26,8 @@ def ui(self):
                     with gr.TabItem('Scale to', elem_id="extras_scale_to_tab") as tab_scale_to:
                         with FormRow():
                             with gr.Column(elem_id="upscaling_column_size", scale=4):
-                                upscaling_resize_w = gr.Slider(minimum=64, maximum=2048, step=8, label="Width", value=512, elem_id="extras_upscaling_resize_w")
-                                upscaling_resize_h = gr.Slider(minimum=64, maximum=2048, step=8, label="Height", value=512, elem_id="extras_upscaling_resize_h")
+                                upscaling_resize_w = gr.Slider(minimum=64, maximum=8192, step=8, label="Width", value=512, elem_id="extras_upscaling_resize_w")
+                                upscaling_resize_h = gr.Slider(minimum=64, maximum=8192, step=8, label="Height", value=512, elem_id="extras_upscaling_resize_h")
                             with gr.Column(elem_id="upscaling_dimensions_row", scale=1, elem_classes="dimensions-tools"):
                                 upscaling_res_switch_btn = ToolButton(value=switch_values_symbol, elem_id="upscaling_res_switch_btn", tooltip="Switch width/height")
                                 upscaling_crop = gr.Checkbox(label='Crop to fit', value=True, elem_id="extras_upscaling_crop")
diff --git a/scripts/processing_autosized_crop.py b/scripts/processing_autosized_crop.py
index c098022645d..7e674989814 100644
--- a/scripts/processing_autosized_crop.py
+++ b/scripts/processing_autosized_crop.py
@@ -28,7 +28,7 @@ def multicrop_pic(image: Image, mindim, maxdim, minarea, maxarea, objective, thr
 
 class ScriptPostprocessingAutosizedCrop(scripts_postprocessing.ScriptPostprocessing):
     name = "Auto-sized crop"
-    order = 4000
+    order = 4020
 
     def ui(self):
         with ui_components.InputAccordion(False, label="Auto-sized crop") as enable:
diff --git a/scripts/xyz_grid.py b/scripts/xyz_grid.py
index 0dc255bc43d..57ee47088cf 100644
--- a/scripts/xyz_grid.py
+++ b/scripts/xyz_grid.py
@@ -45,7 +45,7 @@ def apply_prompt(p, x, xs):
 def apply_order(p, x, xs):
     token_order = []
 
-    # Initally grab the tokens from the prompt, so they can be replaced in order of earliest seen
+    # Initially grab the tokens from the prompt, so they can be replaced in order of earliest seen
     for token in x:
         token_order.append((p.prompt.find(token), token))
 
@@ -270,6 +270,7 @@ def __init__(self, *args, **kwargs):
     AxisOption("Refiner checkpoint", str, apply_field('refiner_checkpoint'), format_value=format_remove_path, confirm=confirm_checkpoints_or_none, cost=1.0, choices=lambda: ['None'] + sorted(sd_models.checkpoints_list, key=str.casefold)),
     AxisOption("Refiner switch at", float, apply_field('refiner_switch_at')),
     AxisOption("RNG source", str, apply_override("randn_source"), choices=lambda: ["GPU", "CPU", "NV"]),
+    AxisOption("FP8 mode", str, apply_override("fp8_storage"), cost=0.9, choices=lambda: ["Disable", "Enable for SDXL", "Enable"]),
 ]
 
 
@@ -437,13 +438,16 @@ def ui(self, is_img2img):
             with gr.Column():
                 draw_legend = gr.Checkbox(label='Draw legend', value=True, elem_id=self.elem_id("draw_legend"))
                 no_fixed_seeds = gr.Checkbox(label='Keep -1 for seeds', value=False, elem_id=self.elem_id("no_fixed_seeds"))
+                with gr.Row():
+                    vary_seeds_x = gr.Checkbox(label='Vary seeds for X', value=False, min_width=80, elem_id=self.elem_id("vary_seeds_x"), tooltip="Use different seeds for images along X axis.")
+                    vary_seeds_y = gr.Checkbox(label='Vary seeds for Y', value=False, min_width=80, elem_id=self.elem_id("vary_seeds_y"), tooltip="Use different seeds for images along Y axis.")
+                    vary_seeds_z = gr.Checkbox(label='Vary seeds for Z', value=False, min_width=80, elem_id=self.elem_id("vary_seeds_z"), tooltip="Use different seeds for images along Z axis.")
             with gr.Column():
                 include_lone_images = gr.Checkbox(label='Include Sub Images', value=False, elem_id=self.elem_id("include_lone_images"))
                 include_sub_grids = gr.Checkbox(label='Include Sub Grids', value=False, elem_id=self.elem_id("include_sub_grids"))
+                csv_mode = gr.Checkbox(label='Use text inputs instead of dropdowns', value=False, elem_id=self.elem_id("csv_mode"))
             with gr.Column():
                 margin_size = gr.Slider(label="Grid margins (px)", minimum=0, maximum=500, value=0, step=2, elem_id=self.elem_id("margin_size"))
-            with gr.Column():
-                csv_mode = gr.Checkbox(label='Use text inputs instead of dropdowns', value=False, elem_id=self.elem_id("csv_mode"))
 
         with gr.Row(variant="compact", elem_id="swap_axes"):
             swap_xy_axes_button = gr.Button(value="Swap X/Y axes", elem_id="xy_grid_swap_axes_button")
@@ -475,6 +479,8 @@ def fill(axis_type, csv_mode):
         fill_z_button.click(fn=fill, inputs=[z_type, csv_mode], outputs=[z_values, z_values_dropdown])
 
         def select_axis(axis_type, axis_values, axis_values_dropdown, csv_mode):
+            axis_type = axis_type or 0  # if axle type is None set to 0
+
             choices = self.current_axis_options[axis_type].choices
             has_choices = choices is not None
 
@@ -522,9 +528,11 @@ def get_dropdown_update_from_params(axis, params):
             (z_values_dropdown, lambda params: get_dropdown_update_from_params("Z", params)),
         )
 
-        return [x_type, x_values, x_values_dropdown, y_type, y_values, y_values_dropdown, z_type, z_values, z_values_dropdown, draw_legend, include_lone_images, include_sub_grids, no_fixed_seeds, margin_size, csv_mode]
+        return [x_type, x_values, x_values_dropdown, y_type, y_values, y_values_dropdown, z_type, z_values, z_values_dropdown, draw_legend, include_lone_images, include_sub_grids, no_fixed_seeds, vary_seeds_x, vary_seeds_y, vary_seeds_z, margin_size, csv_mode]
+
+    def run(self, p, x_type, x_values, x_values_dropdown, y_type, y_values, y_values_dropdown, z_type, z_values, z_values_dropdown, draw_legend, include_lone_images, include_sub_grids, no_fixed_seeds, vary_seeds_x, vary_seeds_y, vary_seeds_z, margin_size, csv_mode):
+        x_type, y_type, z_type = x_type or 0, y_type or 0, z_type or 0  # if axle type is None set to 0
 
-    def run(self, p, x_type, x_values, x_values_dropdown, y_type, y_values, y_values_dropdown, z_type, z_values, z_values_dropdown, draw_legend, include_lone_images, include_sub_grids, no_fixed_seeds, margin_size, csv_mode):
         if not no_fixed_seeds:
             modules.processing.fix_seed(p)
 
@@ -546,6 +554,8 @@ def process_axis(opt, vals, vals_dropdown):
                 valslist_ext = []
 
                 for val in valslist:
+                    if val.strip() == '':
+                        continue
                     m = re_range.fullmatch(val)
                     mc = re_range_count.fullmatch(val)
                     if m is not None:
@@ -568,6 +578,8 @@ def process_axis(opt, vals, vals_dropdown):
                 valslist_ext = []
 
                 for val in valslist:
+                    if val.strip() == '':
+                        continue
                     m = re_range_float.fullmatch(val)
                     mc = re_range_count_float.fullmatch(val)
                     if m is not None:
@@ -688,7 +700,7 @@ def fix_axis_seeds(axis_opt, axis_list):
         grid_infotext = [None] * (1 + len(zs))
 
         def cell(x, y, z, ix, iy, iz):
-            if shared.state.interrupted:
+            if shared.state.interrupted or state.stopping_generation:
                 return Processed(p, [], p.seed, "")
 
             pc = copy(p)
@@ -697,6 +709,16 @@ def cell(x, y, z, ix, iy, iz):
             y_opt.apply(pc, y, ys)
             z_opt.apply(pc, z, zs)
 
+            xdim = len(xs) if vary_seeds_x else 1
+            ydim = len(ys) if vary_seeds_y else 1
+
+            if vary_seeds_x:
+               pc.seed += ix
+            if vary_seeds_y:
+               pc.seed += iy * xdim
+            if vary_seeds_z:
+               pc.seed += iz * xdim * ydim
+
             try:
                 res = process_images(pc)
             except Exception as e:
diff --git a/style.css b/style.css
index ee39a57b73e..004038f899a 100644
--- a/style.css
+++ b/style.css
@@ -1,6 +1,6 @@
 /* temporary fix to load default gradio font in frontend instead of backend */
 
-@import url('https://fonts.googleapis.com/css2?family=Source+Sans+Pro:wght@400;600&display=swap');
+@import url('webui-assets/css/sourcesanspro.css');
 
 
 /* temporary fix to hide gradio crop tool until it's fixed https://github.com/gradio-app/gradio/issues/3810 */
@@ -28,7 +28,7 @@ div.gradio-container{
 }
 
 .hidden{
-    display: none;
+    display: none !important;
 }
 
 .compact{
@@ -222,6 +222,10 @@ input[type="checkbox"].input-accordion-checkbox{
     top: -0.75em;
 }
 
+.block.token-counter-visible{
+    display: block !important;
+}
+
 .block.token-counter span{
     background: var(--input-background-fill) !important;
     box-shadow: 0 0 0.0 0.3em rgba(192,192,192,0.15), inset 0 0 0.6em rgba(192,192,192,0.075);
@@ -331,17 +335,17 @@ input[type="checkbox"].input-accordion-checkbox{
 .generate-box{
     position: relative;
 }
-.gradio-button.generate-box-skip, .gradio-button.generate-box-interrupt{
+.gradio-button.generate-box-skip, .gradio-button.generate-box-interrupt, .gradio-button.generate-box-interrupting{
     position: absolute;
     width: 50%;
     height: 100%;
     display: none;
     background: #b4c0cc;
 }
-.gradio-button.generate-box-skip:hover, .gradio-button.generate-box-interrupt:hover{
+.gradio-button.generate-box-skip:hover, .gradio-button.generate-box-interrupt:hover, .gradio-button.generate-box-interrupting:hover{
     background: #c2cfdb;
 }
-.gradio-button.generate-box-interrupt{
+.gradio-button.generate-box-interrupt, .gradio-button.generate-box-interrupting{
     left: 0;
     border-radius: 0.5rem 0 0 0.5rem;
 }
@@ -679,7 +683,7 @@ table.popup-table .link{
     transition: 0.2s ease background-color;
 }
 .modalControls:hover {
-    background-color:rgba(0,0,0,0.9);
+    background-color:rgba(0,0,0, var(--sd-webui-modal-lightbox-toolbar-opacity));
 }
 .modalClose {
     margin-left: auto;
@@ -749,6 +753,22 @@ table.popup-table .link{
     display: none;
 }
 
+@media (pointer: fine) {
+    .modalPrev:hover,
+    .modalNext:hover,
+    .modalControls:hover ~ .modalPrev,
+    .modalControls:hover ~ .modalNext,
+    .modalControls:hover .cursor {
+        opacity: 1;
+    }
+
+    .modalPrev,
+    .modalNext,
+    .modalControls .cursor {
+        opacity: var(--sd-webui-modal-lightbox-icon-opacity);
+    }
+}
+
 /* context menu (ie for the generate button) */
 
 #context-menu{
@@ -830,6 +850,20 @@ table.popup-table .link{
     display: inline-block;
 }
 
+/* extensions tab table row hover highlight */
+
+#extensions tr:hover td,
+#config_state_extensions tr:hover td,
+#available_extensions tr:hover td {
+    background: rgba(0, 0, 0, 0.15);
+}
+
+.dark #extensions tr:hover td ,
+.dark #config_state_extensions tr:hover td ,
+.dark #available_extensions tr:hover td {
+    background: rgba(255, 255, 255, 0.15);
+}
+
 /* replace original footer with ours */
 
 footer {
@@ -863,31 +897,21 @@ footer {
     margin-bottom: 1em;
 }
 
-.extra-network-cards{
-    height: calc(100vh - 24rem);
-    overflow: clip scroll;
-    resize: vertical;
-    min-height: 52rem;
+.extra-networks > div.tab-nav{
+    min-height: 2.7rem;
 }
 
-.extra-networks > div.tab-nav{
-    min-height: 3.4rem;
+.extra-networks-controls-div{
+    align-self: center;
+    margin-left: auto;
 }
 
 .extra-networks > div > [id *= '_extra_']{
     margin: 0.3em;
 }
 
-.extra-network-subdirs{
-    padding: 0.2em 0.35em;
-}
-
-.extra-network-subdirs button{
-    margin: 0 0.15em;
-}
 .extra-networks .tab-nav .search,
-.extra-networks .tab-nav .sort,
-.extra-networks .tab-nav .show-dirs
+.extra-networks .tab-nav .sort
 {
     margin: 0.3em;
     align-self: center;
@@ -908,53 +932,69 @@ footer {
     width: auto;
 }
 
-.extra-network-cards .nocards{
+.extra-network-pane .nocards{
     margin: 1.25em 0.5em 0.5em 0.5em;
 }
 
-.extra-network-cards .nocards h1{
+.extra-network-pane .nocards h1{
     font-size: 1.5em;
     margin-bottom: 1em;
 }
 
-.extra-network-cards .nocards li{
+.extra-network-pane .nocards li{
     margin-left: 0.5em;
 }
 
+.extra-network-pane .card .button-row{
+    display: inline-flex;
+    visibility: hidden;
+    color: white;
+}
 
-.extra-network-cards .card .button-row{
-    display: none;
+.extra-network-pane .card .button-row {
     position: absolute;
-    color: white;
     right: 0;
-    z-index: 1
+    z-index: 1;
 }
-.extra-network-cards .card:hover .button-row{
-    display: flex;
+
+.extra-network-pane .card:hover .button-row{
+    visibility: visible;
 }
 
-.extra-network-cards .card .card-button{
+.extra-network-pane .card-button{
     color: white;
 }
 
-.extra-network-cards .card .metadata-button:before{
+.extra-network-pane .copy-path-button::before {
+    content: "⎘";
+}
+
+.extra-network-pane .metadata-button::before{
     content: "🛈";
 }
 
-.extra-network-cards .card .edit-button:before{
+.extra-network-pane .edit-button::before{
     content: "🛠";
 }
 
-.extra-network-cards .card .card-button {
+.extra-network-pane .card-button {
+    width: 1.5em;
     text-shadow: 2px 2px 3px black;
+    color: white;
     padding: 0.25em 0.1em;
-    font-size: 200%;
-    width: 1.5em;
 }
-.extra-network-cards .card .card-button:hover{
+
+.extra-network-pane .card-button:hover{
     color: red;
 }
 
+.extra-network-pane .card .card-button {
+    font-size: 2rem;
+}
+
+.extra-network-pane .card-minimal .card-button {
+    font-size: 1rem;
+}
 
 .standalone-card-preview.card .preview{
     position: absolute;
@@ -963,7 +1003,7 @@ footer {
     height:100%;
 }
 
-.extra-network-cards .card, .standalone-card-preview.card{
+.extra-network-pane .card, .standalone-card-preview.card{
     display: inline-block;
     margin: 0.5rem;
     width: 16rem;
@@ -980,15 +1020,15 @@ footer {
     background-image: url('./file=html/card-no-preview.png')
 }
 
-.extra-network-cards .card:hover{
+.extra-network-pane .card:hover{
     box-shadow: 0 0 2px 0.3em rgba(0, 128, 255, 0.35);
 }
 
-.extra-network-cards .card .actions .additional{
+.extra-network-pane .card .actions .additional{
     display: none;
 }
 
-.extra-network-cards .card .actions{
+.extra-network-pane .card .actions{
     position: absolute;
     bottom: 0;
     left: 0;
@@ -999,45 +1039,45 @@ footer {
     text-shadow: 0 0 0.2em black;
 }
 
-.extra-network-cards .card .actions *{
+.extra-network-pane .card .actions *{
     color: white;
 }
 
-.extra-network-cards .card .actions .name{
+.extra-network-pane .card .actions .name{
     font-size: 1.7em;
     font-weight: bold;
     line-break: anywhere;
 }
 
-.extra-network-cards .card .actions .description {
+.extra-network-pane .card .actions .description {
     display: block;
     max-height: 3em;
     white-space: pre-wrap;
     line-height: 1.1;
 }
 
-.extra-network-cards .card .actions .description:hover {
+.extra-network-pane .card .actions .description:hover {
     max-height: none;
 }
 
-.extra-network-cards .card .actions:hover .additional{
+.extra-network-pane .card .actions:hover .additional{
     display: block;
 }
 
-.extra-network-cards .card ul{
+.extra-network-pane .card ul{
     margin: 0.25em 0 0.75em 0.25em;
     cursor: unset;
 }
 
-.extra-network-cards .card ul a{
+.extra-network-pane .card ul a{
     cursor: pointer;
 }
 
-.extra-network-cards .card ul a:hover{
+.extra-network-pane .card ul a:hover{
     color: red;
 }
 
-.extra-network-cards .card .preview{
+.extra-network-pane .card .preview{
     position: absolute;
     object-fit: cover;
     width: 100%;
@@ -1080,9 +1120,6 @@ div.block.gradio-box.edit-user-metadata {
     margin-top: 1.5em;
 }
 
-
-
-
 div.block.gradio-box.popup-dialog, .popup-dialog {
     width: 56em;
     background: var(--body-background-fill);
@@ -1157,3 +1194,431 @@ body.resizing .resize-handle {
     left: 7.5px;
     border-left: 1px dashed var(--border-color-primary);
 }
+
+/* ========================= */
+.extra-network-pane {
+    display: flex;
+    height: calc(100vh - 24rem);
+    resize: vertical;
+    min-height: 52rem;
+    flex-direction: column;
+    overflow: hidden;
+}
+
+.extra-network-pane .extra-network-pane-content {
+    display: flex;
+    flex: 1;
+    overflow: hidden;
+}
+
+.extra-network-pane .extra-network-tree {
+    flex: 1;
+    font-size: 1rem;
+    border: 1px solid var(--block-border-color);
+    overflow: clip auto !important;
+}
+
+.extra-network-pane .extra-network-cards {
+    flex: 3;
+    overflow: clip auto !important;
+    border: 1px solid var(--block-border-color);
+}
+
+.extra-network-pane .extra-network-tree .tree-list {
+    flex: 1;
+    display: flex;
+    flex-direction: column;
+    padding: 0;
+    width: 100%;
+    overflow: hidden;
+}
+
+
+.extra-network-pane .extra-network-cards::-webkit-scrollbar,
+.extra-network-pane .extra-network-tree::-webkit-scrollbar {
+    background-color: transparent;
+    width: 16px;
+}
+
+.extra-network-pane .extra-network-cards::-webkit-scrollbar-track,
+.extra-network-pane .extra-network-tree::-webkit-scrollbar-track {
+    background-color: transparent;
+    background-clip: content-box;
+}
+
+.extra-network-pane .extra-network-cards::-webkit-scrollbar-thumb,
+.extra-network-pane .extra-network-tree::-webkit-scrollbar-thumb {
+    background-color: var(--border-color-primary);
+    border-radius: 16px;
+    border: 4px solid var(--background-fill-primary);
+}
+
+.extra-network-pane .extra-network-cards::-webkit-scrollbar-button,
+.extra-network-pane .extra-network-tree::-webkit-scrollbar-button {
+    display: none;
+}
+
+.extra-network-control {
+    position: relative;
+    display: grid;
+    width: 100%;
+    padding: 0 !important;
+    margin-top: 0 !important;
+    margin-bottom: 0 !important;
+    font-size: 1rem;
+    text-align: left;
+    user-select: none;
+    background-color: transparent;
+    border: none;
+    transition: background 33.333ms linear;
+    grid-template-rows: min-content;
+    grid-template-columns: minmax(0, auto) repeat(4, min-content);
+    grid-gap: 0.1rem;
+    align-items: start;
+}
+
+.extra-network-tree .tree-list--tree {}
+
+/* Remove auto indentation from tree. Will be overridden later. */
+.extra-network-tree .tree-list--subgroup {
+    margin: 0 !important;
+    padding: 0 !important;
+    box-shadow: 0.5rem 0 0 var(--body-background-fill) inset,
+                0.7rem 0 0 var(--neutral-800) inset;
+}
+
+/* Set indentation for each depth of tree. */
+.extra-network-tree .tree-list--subgroup > .tree-list-item {
+    margin-left: 0.4rem !important;
+    padding-left: 0.4rem !important;
+}
+
+/* Styles for tree 
    •  elements. */
+.extra-network-tree .tree-list-item {
+    list-style: none;
+    position: relative;
+    background-color: transparent;
+}
+
+/* Directory 
         visibility based on data-expanded attribute. */
+.extra-network-tree .tree-list-content+.tree-list--subgroup {
+    height: 0;
+    visibility: hidden;
+    opacity: 0;
+}
+
+.extra-network-tree .tree-list-content[data-expanded]+.tree-list--subgroup {
+    height: auto;
+    visibility: visible;
+    opacity: 1;
+}
+
+/* File 
      •  */
+.extra-network-tree .tree-list-item--subitem {
+    padding-top: 0 !important;
+    padding-bottom: 0 !important;
+    margin-top: 0 !important;
+    margin-bottom: 0 !important;
+}
+
+/* 
      •  containing 
           */
+.extra-network-tree .tree-list-item--has-subitem {}
+
+/* BUTTON ELEMENTS */
+/*