H3 image/video-conditioned generation: wire the encoder vision tower to real weights (row/H3-CONDITIONED-E2E) - #86

Closed

localai-bot wants to merge 4 commits into

mainfrom

row/H3-CONDITIONED-E2E

Commits on Aug 7, 2026

record(h3): reconcile encoder vision-tower record — math CPU-gated synthetic, never real-wired; encoder GGUF carries visual.* (row/H3-CONDITIONED-E2E)
mudler
committed
fa8697b
feat(h3): GGUF visual.* loader wires the encoder vision tower to real weights + driver probe
mudler
committed
e82b7ee
fix(h3): --prompt-image vision probe bypasses the --dit/--vae/--out usage check
mudler
committed
56ad628
record(h3): GB10-verify vision tower on real weights; fl2va COHERENT, ref2va still grids
mudler
committed
beac172