kevin-mi/FastH3-4step-Preview-overlay

FastH3 4-step Preview Overlay

Overlay metadata repo for using FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree with native SGLang Diffusion MiniMax-H3 support.

This repo contains only overlay metadata and materialization logic. Source weights remain in FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree; SGLang's registry pins a revision of this repo so re-materializing the model does not drift.

The materializer reshapes the flat native-Diffusers release into the base MiniMax-H3 partition layout SGLang's loaders consume:

• model_index.json declares the t2va-only release contract (tasks: ["t2va"], video/audio sigma shifts 12/3).
• _overlay/materialize.py re-serializes the video VAE into the fused source form (bit-identical tensor values; names and fused-QKV row order differ), derives the rope.inv_freq buffer the Diffusers export drops, and patches the VAE config class names. Everything else is symlinked from the source snapshot.

Usage (no manual steps; SGLang resolves this overlay automatically):

sglang serve \
  --model-path FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree \
  --num-gpus 4 \
  --attention-backend video_sparse_attn_h3 \
  --component-attention-backends text_encoder=fa

The model inherits the MiniMax-H3 Community License from its base checkpoint.