v0.6.0 release checklist

[ssube]

Issues:

• implement upload output images to an S3-compatible endpoint #7
• add add devcontainer definitions for codespaces #25
• implement add prompt emphasis and length check #27
• add add cancel button to loading card #60
• add add support for Stable Diffusion x4 Upscaler model #66
• add add support for CodeFormer correction model #68
• add add a way to include custom models in the convert script without code changes #70
• implement run outpainting at full resolution #71
• test and document images with one dimension < 256 are neon noise #79
• implement switch to the appropriate tab when using the copy to img2img and inpaint buttons #80
• implement store image parameters along with output images #81
• fix set a reasonable default platform in each container #82
• add add an option for the server to disable certain platforms #83
• add add a way to chain multiple pipelines #84, chain pipelines
• implement implement spiral tile processing #85
• implement ONNX pipeline for Stable Diffusion upscaler #86, upstream as add OnnxStableDiffusionUpscalePipeline pipeline huggingface/diffusers#2158
• implement add a way to send images to chain pipelines #88
• implement cache community pipelines #96
• fix fix Euler scheduler with LPW pipeline #99
• fix using CodeFormer correction returns incorrect image size #100
• fix onnx parameter error while inpainting #101
• fix upscaling still runs when the checkbox is cleared if the scale factor is > 1 #106
• fix using Real ESRGAN with GFPGAN is still producing scattered tiles #108
• fix upscaling with Real ESRGAN v3 produces a neon image #109
• fix upscaling with Stable Diffusion while using CodeFormer and outscale > 1 produces scattered tiles #110
• add add a parameter for correction outscale #111
• fix unnecessary warning when convert script is run with --extras="" #112

Merges:

• support for multiple devices #91
• support multiple cuda devices #92
• enable LPW #93
• add codeformer support #94
• make LPW a request parameter #102

Testing:

• txt2img
  • 256x256 with SD v1.5
    • should fail: neon blobs
    • seems to work (sometimes?) with DPM multi
    • attempt 2
  • 512x512 with SD v1.5
    • DPM Multi
      • should work
      • attempt 2
    • Euler A
      • should work
      • attempt 2
  • 512x512 with SD v2.1
    • should work
    • attempt 2
  • 768x768 with SD v2.1
    • should work, given sufficient memory
    • attempt 2
  • extra models
    • 512x512 with Knollingcase
      • should work
      • attempt 2
    • 512x512 with OpenJourney
      • should work
      • 256x256 seems to work as well
      • attempt 2
• img2img
  • 256x256 input
    • should fail: neon blobs
    • attempt 2
  • 512x512 input
    • should work
    • attempt 2
  • 1024x768 input
    • should work
    • attempt 2
• inpaint
  • outpaint
    • 0 all sides
      • should work, run 1 tile
      • attempt 2
    • 256 all sides
      • should work, run 8 tiles: image came out ok, but card reported Size: 1536x1536
      • attempt 2
    • 512 top and bottom, 0 left and right
      • should work, run 3 tiles
      • attempt 2
    • 512 left and right, 0 top and bottom
      • should work, run 3 tiles
      • attempt 2
• upscale
  • Real ESRGAN
    • with CodeFormer
      • should work
      • attempt 2
    • with GFPGAN
      • should work failed: scattered tiles, using Real ESRGAN with GFPGAN is still producing scattered tiles #108
      • attempt 2
    • without face correction
      • should work
      • attempt 2
    • using x2 model and x4 scale
      • should fail: tiles worked: seems to have ignored scale and run at x2, reported size is correct
      • attempt 2
    • using x4 model and x2 scale
      • should fail: tiles worked: seems to have ignored scale and run at x4
      • attempt 2
    • using v3 model and x2 scale
      • should work failed: neon overlay, upscaling with Real ESRGAN v3 produces a neon image #109
      • attempt 2
    • using v3 model and x4 scale
      • should work failed: neon overlay, upscaling with Real ESRGAN v3 produces a neon image #109
      • attempt 2
  • Stable Diffusion
    • using x2 scale
      • should fail: tiles
      • attempt 2
    • using x4 scale
      • should work
      • attempt 2
    • with CodeFormer
      • should work failed: scattered tiles, upscaling with Stable Diffusion while using CodeFormer and outscale > 1 produces scattered tiles #110
      • attempt 2
    • with GFPGAN
      • should fail: using GFPGAN correction with SD upscaling causes a model error in Real ESRGAN #87
      • attempt 2
    • without face correction
      • should work
      • attempt 2
• 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
  • persist on refresh
    • loading images
    • switching tabs

Release:

• update package versions and stage files
• run commit-and-tag-version --sign --git-tag-fallback --commit-all --release-as=minor to make v0.6.0
• make sure packages and images have been built
• fix lint error (line length) from recent patches
• run commit-and-tag-version --sign --git-tag-fallback --commit-all --release-as=minor to make v0.6.1
• make sure packages and images have been built
• update GH pages bundle and default version
• post release on GH
• make follow up tickets
• close milestone and checklist

[ssube]

Real ESRGAN + GFPGAN failed with:

[2023-02-05 23:33:39,374] WARNING: onnx_web.device_pool: job txt2img_902278279_b2f2069fbb92453a47865671c194efa56bd10c955f3d264dc3a72de692bfcd5d_1675661585.png failed with an error: ['Traceback (most
 recent call last):\n', '  File "C:\\Users\\ssube\\stabdiff\\onnx-try-2\\onnx-web\\api\\onnx_web\\device_pool.py", line 192, in job_done\n    f.result()\n', '  File "C:\\Users\\ssube\\AppData\\Local
\\Programs\\Python\\Python310\\lib\\concurrent\\futures\\_base.py", line 451, in result\n    return self.__get_result()\n', '  File "C:\\Users\\ssube\\AppData\\Local\\Programs\\Python\\Python310\\li
b\\concurrent\\futures\\_base.py", line 403, in __get_result\n    raise self._exception\n', '  File "C:\\Users\\ssube\\AppData\\Local\\Programs\\Python\\Python310\\lib\\concurrent\\futures\\thread.p
y", line 58, in run\n    result = self.fn(*self.args, **self.kwargs)\n', '  File "C:\\Users\\ssube\\stabdiff\\onnx-try-2\\onnx-web\\api\\onnx_web\\diffusion\\run.py", line 67, in run_txt2img_pipelin
e\n    image = run_upscale_correction(\n', '  File "C:\\Users\\ssube\\stabdiff\\onnx-try-2\\onnx-web\\api\\onnx_web\\upscale.py", line 56, in run_upscale_correction\n    return chain(job, server, pa
rams, image, prompt=params.prompt, upscale=upscale)\n', '  File "C:\\Users\\ssube\\stabdiff\\onnx-try-2\\onnx-web\\api\\onnx_web\\chain\\base.py", line 103, in __call__\n    image = process_tile_gri
d(\n', '  File "C:\\Users\\ssube\\stabdiff\\onnx-try-2\\onnx-web\\api\\onnx_web\\chain\\utils.py", line 36, in process_tile_grid\n    tile_image = filter(tile_image, (left, top, tile))\n', '  File "
C:\\Users\\ssube\\stabdiff\\onnx-try-2\\onnx-web\\api\\onnx_web\\chain\\base.py", line 96, in stage_tile\n    tile = stage_pipe(job, server, stage_params, params, tile, **kwargs)\n', '  File "C:\\Us
ers\\ssube\\stabdiff\\onnx-try-2\\onnx-web\\api\\onnx_web\\chain\\correct_gfpgan.py", line 72, in correct_gfpgan\n    gfpgan = load_gfpgan(server, upscale, upsampler=upsampler)\n', '  File "C:\\User
s\\ssube\\stabdiff\\onnx-try-2\\onnx-web\\api\\onnx_web\\chain\\correct_gfpgan.py", line 30, in load_gfpgan\n    upsampler = load_resrgan(ctx, bg_upscale)\n', "TypeError: load_resrgan() missing 1 re
quired positional argument: 'device'\n"]

[ssube]

Fixed the GFPGAN device error, but using Real ESRGAN with GFPGAN is still producing scattered tiles: #108

[ssube]

Most of the app seems to be working, but there are a few issues around combined upscaling + correction and the outscale parameter that I want to fix before cutting the release.