AnalysisVisual AINovember 19, 2026

MOSS-Image SFT gets 4-step distillation LoRA via eDMD + LADD

Read original source →reddit.com

Community developer lvladikov released a 4-step distillation LoRA for MOSS-Image SFT, trained with a mix of eDMD and latent adversarial diffusion distillation (LADD). It supports text-to-image, text-to-image with layout boxes, and image-to-image edits.

1 source

More stories today

Open the live feed