MMnode/nodes
Z User 46617b71b8 fix: snap the half-res first pass onto the even-latent grid
Multi-segment projects crashed on the first continuation segment with
RuntimeError: shape '[1, 24, 1, 1, 8, 2, 15, 2]' is invalid for input of
size 12240 (patchify_video). The hidden half-resolution first pass
declared its size with plain integer halving, so a 1920x1080 editor
sampled at 960x540 - and a latent encoded from 540px carries an odd
17-row count. Odd latents sample fine, but when the segment tail comes
back as motion-context keyframes on the next segment, the model's
patchify_video (2x2 latent patches) cannot reshape them.

The first pass now snaps onto the nearest 64px grid (h3_even_latent_size)
so every latent in the pipeline keeps even row/column counts:

- 1920x1080 editor samples at 960x512 (was 960x540 -> 17 rows)
- 1344x768 samples at 704x384 (672 had odd latent columns)
- 3840x2160 is unchanged at 1920x1088 - the size the VAE already rounds
  up to, so the 4K-editor trick behaves identically

Outputs and second-pass dimensions are unchanged; only continuation
segments that previously crashed now run. Also updates the stale
single-task expansion test, which still expected pre-half-res-trick
conditioning dimensions (editor size instead of the sampled size).
2026-10-15 02:05:48 +00:00
..
__init__.py init 2026-09-26 11:11:23 -07:00
audio.py init 2026-09-26 11:11:23 -07:00
basic.py feat: revert editor to pristine upstream, drop format label mapping and T2V==R2V, fix editor delivery 2026-09-27 21:24:56 +00:00
common.py init 2026-09-26 11:11:23 -07:00
image.py init 2026-09-26 11:11:23 -07:00
ltx.py init 2026-09-26 11:11:23 -07:00
ltxv.py init 2026-09-26 11:11:23 -07:00
minimax.py feat: translate all cache/task console logs to English 2026-09-26 23:27:09 +00:00
project.py fix: snap the half-res first pass onto the even-latent grid 2026-10-15 02:05:48 +00:00
subtitle.py init 2026-09-26 11:11:23 -07:00
video.py init 2026-09-26 11:11:23 -07:00
wan.py init 2026-09-26 11:11:23 -07:00