CalamitousFelicitousness 726907dd5b fix(video): count minimax steps as transformer evaluations ...

MiniMaxH3Scheduler counts the terminal sigma in num_inference_steps, so
Steps N ran N-1 evaluations while every other model runs N. The shim
hands the scheduler p.steps + 1, the slider starts at 1, and the PDD
pin records the evaluation count while passing the scheduler its grid
argument. Metadata written before this change counted grid points.

2026-09-16 00:47:46 +01:00

..

__init__.py

linting

2026-07-03 22:05:33 +02:00

google_omni.py

update all google stuff

2026-09-05 21:53:29 +02:00

google_veo.py

update all google stuff

2026-09-05 21:53:29 +02:00

models_def.py

add minimax-nunchaku-lite

2026-09-13 12:49:57 +02:00

video_cache.py

cleanup logger

2026-02-19 11:09:13 +01:00

video_codecs.py

fix(video): correct codec option strings to the key=value form

2026-08-17 03:04:59 +01:00

video_load.py

fix: honor offline mode in model loaders

2026-09-06 10:10:36 -04:00

video_minimax.py

fix(video): count minimax steps as transformer evaluations

2026-09-16 00:47:46 +01:00

video_overrides.py

feat(video): absolute per-request minimax shift on every path

2026-09-15 23:01:41 +01:00

video_run.py

feat(video): absolute per-request minimax shift on every path

2026-09-15 23:01:41 +01:00

video_save.py

improve memory reporting, explicit interpolation and minimax tracking

2026-08-25 09:11:46 +02:00

video_ui.py

feat(ui): filter generated sampler and upscaler lists

2026-09-07 08:29:37 -07:00

video_upscale.py

video upscaling

2026-08-21 20:28:44 +02:00

video_utils.py

feat(ltx): add a keyword core to the ltx runner

2026-08-19 01:45:46 +01:00

video_vae.py

fix(video): route tiny decode through the vae hijack

2026-08-18 03:13:00 +01:00