MMnode/utils
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_gain.py init 2026-09-26 11:11:23 -07:00
audio.py feat: scrub EasyMedia/MiniMax branding from all user-facing output 2026-09-26 23:05:24 +00:00
bernini_s2v_model_patch.py init 2026-09-26 11:11:23 -07:00
bernini_s2v.py init 2026-09-26 11:11:23 -07:00
bernini.py init 2026-09-26 11:11:23 -07:00
h3_conditioning_cache.py feat: translate all cache/task console logs to English 2026-09-26 23:27:09 +00:00
h3_presets.py init 2026-09-26 11:11:23 -07:00
h3_project.py fix: snap the half-res first pass onto the even-latent grid 2026-10-15 02:05:48 +00:00
image.py init 2026-09-26 11:11:23 -07:00
llm_api.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
log.py feat: scrub EasyMedia/MiniMax branding from all user-facing output 2026-09-26 23:05:24 +00:00
media.py feat: scrub EasyMedia/MiniMax branding from all user-facing output 2026-09-26 23:05:24 +00:00
minimax.py init 2026-09-26 11:11:23 -07:00
model_memory.py feat: scrub EasyMedia/MiniMax branding from all user-facing output 2026-09-26 23:05:24 +00:00
models.py feat: scrub EasyMedia/MiniMax branding from all user-facing output 2026-09-26 23:05:24 +00:00
multitrack.py init 2026-09-26 11:11:23 -07:00
panorama.py init 2026-09-26 11:11:23 -07:00
project_memory.py feat: scrub EasyMedia/MiniMax branding from all user-facing output 2026-09-26 23:05:24 +00:00
prompt_builder.py init 2026-09-26 11:11:23 -07:00
prompt_override.py init 2026-09-26 11:11:23 -07:00
rtx_vsr.py feat: display Moxie Multimedia on the format combo, keep MiniMax internally; silence the upscale console line 2026-09-27 03:52:43 +00:00
sampling_preview.py feat: scrub EasyMedia/MiniMax branding from all user-facing output 2026-09-26 23:05:24 +00:00
speech.py init 2026-09-26 11:11:23 -07:00
subtitles.py init 2026-09-26 11:11:23 -07:00
video.py fix: also look for FFmpeg next to the running python interpreter 2026-09-29 06:47:43 +00:00
workflow_submission.py init 2026-09-26 11:11:23 -07:00