Rattus 023036ef9d vae_refiner: roll the convolution through temporal II ...

Roll the convolution through time using 2-latent-frame chunks and a
FIFO queue for the convolution seams.

Added support for encoder, lowered to 1 latent frame to save more
VRAM, made work for Hunyuan Image 3.0 (as code shared).

Fixed names, cleaned up code.

2025-11-20 22:12:56 -05:00

..

model.py

Fix TokenRefiner for fp16

2025-11-20 22:12:56 -05:00

upsampler.py

Support HunyuanVideo15 latent resampler

2025-11-20 22:12:56 -05:00

vae_refiner.py

vae_refiner: roll the convolution through temporal II

2025-11-20 22:12:56 -05:00

vae.py

Fix lowvram issue with hunyuan image vae. (#9794)

2025-09-10 02:15:34 -04:00