Skip to content

sd: sync with master-812-ea7f0c8 - #2371

Merged
LostRuins merged 6 commits into
LostRuins:concedo_experimentalfrom
wbruna:kcpp_sd_update_202607_4
Aug 6, 2026
Merged

sd: sync with master-812-ea7f0c8#2371
LostRuins merged 6 commits into
LostRuins:concedo_experimentalfrom
wbruna:kcpp_sd_update_202607_4

Conversation

@wbruna

@wbruna wbruna commented Jul 30, 2026

Copy link
Copy Markdown

master-801-9cfe2af:

  • fix missing sampler names
  • fix null pointer with malformed LoHa
  • backend-aware parameter loading
  • show number of prompt tokens for SD models

Didn't merge master-802-e92e86f yet because it has a few non-trivial conflicts with the int8 convrot changes.

@wbruna

wbruna commented Jul 30, 2026

Copy link
Copy Markdown
Author

master-802-e92e86f:

  • fix: prevent torch checkpoint offset overflow

@wbruna wbruna changed the title sd: sync with master-801-9cfe2af sd: sync with master-802-e92e86f Jul 30, 2026
@wbruna wbruna changed the title sd: sync with master-802-e92e86f sd: sync with master-805-e31a86c Jul 31, 2026
@wbruna

wbruna commented Jul 31, 2026

Copy link
Copy Markdown
Author

master-805-e31a86c:

  • support for changing alphaand beta for the Beta scheduler
  • fix Photomaker crash when 'img' isn't specified
  • CLIP prefix conversion refactor

@wbruna
wbruna force-pushed the kcpp_sd_update_202607_4 branch from 934d28a to 579b76b Compare July 31, 2026 10:46
@wbruna
wbruna force-pushed the kcpp_sd_update_202607_4 branch from 579b76b to a4dd6c1 Compare August 2, 2026 18:14
@wbruna wbruna changed the title sd: sync with master-805-e31a86c sd: sync with master-810-db99efd Aug 2, 2026
@wbruna

wbruna commented Aug 2, 2026

Copy link
Copy Markdown
Author

master-810-db99efd:

  • model loading refactor (needed to solve a few conflicts)
  • LMS sampler
  • IP-Adapter Plus
  • LoRA compatibility changes (Kroma-v0.1)

@wbruna wbruna changed the title sd: sync with master-810-db99efd sd: sync with master-812-ea7f0c8 Aug 5, 2026
@wbruna

wbruna commented Aug 5, 2026

Copy link
Copy Markdown
Author

master-812-ea7f0c8:

  • minimax-h3 (only sync for now)

Had to deal with conflicts on the LTX audio changes (untested).

@wbruna

wbruna commented Aug 5, 2026

Copy link
Copy Markdown
Author
output-2585.mp4

(audio still not working) edit: fixed.

@wbruna
wbruna force-pushed the kcpp_sd_update_202607_4 branch from f2f919d to e56c638 Compare August 5, 2026 22:31
@wbruna
wbruna force-pushed the kcpp_sd_update_202607_4 branch from e56c638 to af05c62 Compare August 6, 2026 03:13
@LostRuins

LostRuins commented Aug 6, 2026

Copy link
Copy Markdown
Owner

looks coool lol. running some regression tests now

edit: noticing some changes with preset lora multipliers, output is wrong but if the lora is set dynamically it is correct.
also noticing a regression where the model weights are being lazy loaded again
investigating now

edit2: setting lora to runtime seems to fix it, guess i'll add that later

edit3: havent tested out H3 yet but regression testing looks good. merging first.

@LostRuins
LostRuins merged commit 0ddb919 into LostRuins:concedo_experimental Aug 6, 2026
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.

2 participants