v0.10.0 release checklist

[ssube]

Issues:

• add support for SwinIR restoration #153
• add support for BSRGAN upscaling #154
• switch API container to waitress  #159
• tensor errors when model download is cut off #165
• add a linear blend stage for chain pipelines  #183
• add a download from HTTPS stage for chain pipelines  #184
• add a download from S3 stage for chain pipelines  #185
• add token-protected admin endpoints #207
• name context params consistently #278
• extract thread worker mains #279
• pytorch 2 support #292
• MIopen missing system database file #293
• SD upscaling fails with torch-fp16 #300
• add prompt construction and keywords to the user guide #301
• add support for ControlNet #302
• add support for the highres fix #303
• build an EXE bundle for windows #305
• clip skip prompt token doesn't work with fp16 models #309
• add support for LyCORIS networks: LoCON and LoHA #312
• missing directories for GFPGAN when using Windows bundle #315
• add a parameter for the number of highres iterations #316
• prompt in JSON file does not have network tokens #320
• run PyTorch GC for ROCm devices #323
• make sure bundle JS is served with correct content type #327
• update user guide for v0.10 #328
• add a notification when an image is ready #329
• add optimization for offloading models to CPU #330
• add parameter for img2img loopback #331
• fixes for diffusers 0.15 #336
• correctly handle empty prompts #340
• include an empty extras file in the Windows bundle #342
• inpainting is also outpainting for some reason, 512 inputs produce 1536 outputs #345
• add support for negative embeddings #348
• inpaint tab is resizing source images to 512 #351
• add python 3.8 to CI or drop support for it #354
• add optimization to enable attention slicing during conversion #355
• Add inpainting pipeline to diffuser conversion #356
• add support for tiled VAE #358
• Missing key(s) in state_dict error when converting v2 checkpoints #360
• add a tab for converting models and managing the extras file #365
• cannot convert models that were created using pytorch_lightning with the Windows bundle #366
• include the build version in the logs #367
• conversion can fail at ControlNet on machines with 16GB of system memory #369
• cannot use tabs in extras file #370
• error converting models from Huggingface website #371
• inpaint source image does not appear in canvas until you paint something on the mask #372
• inpaint mask does not line up correctly with non-square sources #373
• add parameters to image file EXIF #383
• split up input and output size in JSON file #385
• High VRAM usage after img2img when worker stays alive #387

Merges:

• TODO

Testing:

• txt2img
  • 256x256 with SD v1.5
    • should work
    • has automation
  • 512x512 with SD v1.5
    • DEIS Multi
      • should work
      • has automation
    • DPM Multi
      • should work
      • has automation
    • Euler A
      • should work
    • Heun
      • should work
      • has automation
    • UniPC Multi
      • should work
      • has automation
  • 512x512 with SD v2.1
    • should work
    • has automation
  • 768x768 with SD v2.1
    • should work, given sufficient memory
    • has automation
  • extra models
    • 512x512 with Knollingcase
      • should work
      • has automation
    • 512x512 with OpenJourney
      • should work
      • has automation
    • 256x256 with OpenJourney
      • should work
      • has automation
  • textual inversion
    • 512x512 with SD v1.5 + Autumn (embeddings)
      • should work
      • has automation
    • 512x512 with SD v1.5 + Minecraft (concept)
      • should work
      • has automation
  • LoRA weights
    • 512x512 with SD v1.5 + Taters
      • should work
      • has automation
• img2img
  • 256x256 input
    • should work: expanded into valid image
    • has automation
  • 512x512 input
    • should work
    • has automation
  • 1024x768 input
    • should work
    • has automation
• inpaint
  • regular inpaint
    • black mask
      • should keep all pixels, same image
      • has automation
    • white mask
      • should replace all pixels, different image
      • has automation
  • outpaint
    • 0 all sides
      • should work, run 1 tile
      • primarily a client-side test
    • 256 all sides
      • should work, run 8 tiles
      • has automation
    • 512 top and bottom, 0 left and right
      • should work, run 3 tiles
      • has automation
    • 512 left and right, 0 top and bottom
      • should work, run 3 tiles
      • has automation
• upscale
  • Real ESRGAN
    • x4 with CodeFormer
      • should work
      • has automation
    • x4 with GFPGAN
      • should work
      • has automation
    • x4 without face correction
      • should work
      • has automation
    • x2 without face correction
      • should work
      • has automation
    • x2 model and x4 scale
      • should sort of work: ignores scale and uses x2
    • x4 model and x2 scale
      • should fail: tiles worked, seemes to ignore scale
    • v3 model and x4 scale
      • should work
  • Stable Diffusion
    • using x2 scale
      • should fail: tiles
    • using x4 scale
      • should work
      • has automation
    • with CodeFormer
      • should work
      • has automation
    • with GFPGAN
      • should work
      • has automation
• blend
  • two 512x512 inputs
    • should work
    • has automation
  • two 1024x1024 inputs
    • should work
  • two different size inputs
    • 256x256 and 512x512
      • should work
    • 512x512 and 1024x1024
      • should work
• interactions
  • generate a new image
    • should request and then load an image from the server
  • delete a pending image
    • should remove a single image and leave the rest
  • delete a finished image
    • should remove a single image and leave the rest
  • copy an image to img2img
    • should switch to the img2img tab
    • should populate the image source
    • the generate button should be enabled
  • copy an image to inpaint
    • should switch to the inpaint tab
    • should populate the image source
    • the generate button should be enabled
  • copy an image to upscale
    • should switch to the upscale tab
    • should populate the image source
    • the generate button should be enabled
  • copy two images to blend
    • should switch to the blend tab
    • should populate the image sources
    • the generate button should be enabled once both sources have been populated
  • state should persist on refresh
    • loading images
    • switching tabs
    • images sources do not
• schedulers
  • DDIM
  • DDPM DDPM scheduler produces random colored chips #190
  • DEIS Multi
  • DPM Multi
  • DPM Single
  • Euler A
  • Euler
  • Heun
  • iPNDM ipndm scheduler produces random colored chips #297
  • KDPM2 A
  • KDPM2
  • Karras Ve karras ve scheduler throwing an error #189
  • LMS
  • PNDM
  • UniPC Multi
• providers
  • CUDA 11.7
  • CPU
  • DirectML
  • ROCm 5.4
• conversion
  • SD v1.5
    • fp32
    • onnx-fp16
    • torch-fp16
  • SD v2.1
    • fp32
    • onnx-fp16 broken: converting SD 2.1 models with onnx-fp16 optimization leads to black images #364
    • torch-fp16
  • SD upscaling
    • fp32
    • onnx-fp16
    • torch-fp16

Release:

• update readme features
• update user guide
• update release tests
• fix any late arrival bugs
• check and fix lint
• update readme screenshot
• update package versions and stage files
• run commit-and-tag-version --sign --git-tag-fallback --commit-all --release-as=minor to make VERSION
• make sure packages and images have been built
  • build ROCm container
  • build Windows bundle
• update GH pages bundle and default version
• post release on GH
• make follow up tickets
• close milestone and checklist