Skip to content

sd: sync with master-741-484baa4#2291

Open
wbruna wants to merge 6 commits into
LostRuins:concedo_experimentalfrom
wbruna:kcpp_sd_update_202606_7
Open

sd: sync with master-741-484baa4#2291
wbruna wants to merge 6 commits into
LostRuins:concedo_experimentalfrom
wbruna:kcpp_sd_update_202606_7

Conversation

@wbruna

@wbruna wbruna commented Jun 27, 2026

Copy link
Copy Markdown

The stable-diffusion.cpp CI is currently timing out, due to issues with the dynamic CPU backend changes, but I'm syncing anyway to test the fixes:

  • taehv fix for image models in image to image mode
  • Vulkan workarounds for Anima and ERNIE replaced by ggml fixes

I'll update this PR once a new release comes out.

@wbruna wbruna changed the base branch from concedo to concedo_experimental June 27, 2026 14:27
@wbruna wbruna force-pushed the kcpp_sd_update_202606_7 branch from 1f55455 to c98c400 Compare June 29, 2026 00:27
@wbruna wbruna changed the title sd: sync with master-721-8caa3f9+5 (9956436) sd: sync with master-731-9f855c9 Jun 29, 2026
@wbruna

wbruna commented Jun 29, 2026

Copy link
Copy Markdown
Author

master-731-9f855c9:

  • Qwen2D-VAE support
  • SeFi-Image support (only tested the 1b, and it's kind of sub-SD1.5 level, so no kobold image)

@wbruna wbruna force-pushed the kcpp_sd_update_202606_7 branch 2 times, most recently from a3ed9fe to c275d2e Compare June 29, 2026 21:51
@wbruna wbruna changed the title sd: sync with master-731-9f855c9 sd: sync with master-737-3b6c9ca Jun 29, 2026
@wbruna

wbruna commented Jun 29, 2026

Copy link
Copy Markdown
Author

master-737-3b6c9ca:

  • 'normal' alias for the discrete scheduler
  • Flux and Flux 2 specific schedulers
  • avoid crash when using Qwen 2D VAE with video

@wbruna wbruna force-pushed the kcpp_sd_update_202606_7 branch from c275d2e to 870388a Compare July 1, 2026 02:36
@wbruna wbruna changed the title sd: sync with master-737-3b6c9ca sd: sync with master-741-484baa4 Jul 1, 2026
@wbruna

wbruna commented Jul 1, 2026

Copy link
Copy Markdown
Author

master-741-484baa4:

  • beta scheduler
  • C API changes
  • byte order mark cleanup

Unfortunately, I'm not being able to build gpttype_adapter.cpp anymore: I'm hitting internal compiler errors all the time, no matter the amount of free RAM 😕 If we can't reduce its complexity, I'll need to update my build stack.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant