Skip to content

Alpha v0.11.5
Compare
Choose a tag to compare
@github-actions github-actions released this 28 Aug 02:51
· 1468 commits to main since this release
v0.11.5
17862ea

Another patch update is here! v0.12.0 is right around the corner, but for now, v0.11.5 brings a few fixes.

Dependency Updates

  • NCNN (#855)
    • I finally got arm64 (M1) macOS builds for ncnn-vulkan (my fork of NCNN) working. This means NCNN should now install properly on M1 macbooks. If it doesn't, or doesn't work at all, please let me know and I will try to resolve it.
    • I also got it to statically compile (I think so, at least) libomp. This means that (hopefully) no mac user will get the missing libomp library error anymore.
    • Remember to update if you haven't in a while. Starting with v0.12.0, changes made in the version that came with v0.11.0 will be required for NCNN upscaling to work.

Bug Fixes

  • Fixed iterator progress bars not working (#839)
  • Fixed an issue where the View Image creation shortcut button on image outputs wouldn't make new View Image nodes if one had already been created, regardless of if one was attached to the output that you clicked the button on (#841)
  • Restored missing node shadow (#840)
  • Fixed ONNX Save Model icon (#846)

Changes to Existing Features

  • Allow negative cropping in crop nodes (#844)
    • This crops from the center of the image rather than off the edges. I figured why not allow it as it might be useful.
  • Remove presets tab placeholder (#851)
    • The presets tab has been a placeholder since I started working on chaiNNer. Eventually presets will come, but for now we might as well get rid of the tab as it only serves to add confusion.
  • Add index outputs to Image File Iterator and Spritesheet Iterator (#852)
    • It was brought to my attention that these lacked an index output like the Video Frame Iterator has. This is useful for outputting files with the indices attached to the file name.

Known issues

  • "Failed to fetch" on macOS < 10.15. This is a known issue caused by opencv not being compiled for these older macOS versions. I have brought this up with the opencv team, and they were receptive to lowering the version target. So, there isn't anything I can do about this right now, but as soon as they release an update with a lower build target I will update the version chaiNNer uses.
  • Iterators are apparently being a bit finicky for some users. I also noticed that pausing and resuming doesn't quite work right anymore for them. We are investigating.
  • ONNX cannot be installed on M1 MacBooks. This is a known issue as ONNX does not have an arm64 version. There is nothing I can do about this.
  • Lots of RAM usage. This is partially due to the caching chaiNNer does (which I have partially fixed already, and the fix will be available in v0.12.0) as well as the fact that many users (particularly macOS ones) have been upscaling with CPU. There is no way around the RAM usage caused by CPU upscaling.
Assets 10
Loading