minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[diffusion] feat: support FastH3 (4-step VSA-distilled MiniMax-H3) with a VSA-H3 attention backend (#37480)

Browse Source 
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Kevin Mi
2026-09-02 21:39:54 +08:00
committed by GitHub
co-authored by Claude Fable 5
parent 99b9109553
commit f586654518
26 changed files with 1778 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/docs/sglang-diffusion/compatibility_matrix.mdx
+26
View File
@@ -137,6 +137,12 @@ Rows are grouped when a family shares the same runtime path or optimization supp
<td>T2VA / FL2VA / Ref2VA, 768p at 24 fps with synchronized audio</td>
<td><span className="sgd-chip">Cache-DiT</span><span className="sgd-chip">Sage</span><span className="sgd-chip">Online FP8</span><span className="sgd-chip">GGUF</span></td>
</tr>
<tr>
<td>FastH3</td>
<td><div className="sgd-id-list"><code>FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree</code></div></td>
<td>T2VA only, 4-step distilled, 768p at 24 fps with synchronized audio</td>
<td><span className="sgd-chip">VSA-H3</span></td>
</tr>
<tr>
<td>Wan2.1 Fun</td>
<td><div className="sgd-id-list"><code>weizhou03/Wan2.1-Fun-1.3B-InP-Diffusers</code></div></td>
@@ -621,6 +627,21 @@ Optimization columns are abbreviated to keep the matrix readable:
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>❌</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>❌</td>
</tr>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>FastH3 4-step (T2VA only)</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}><code>FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree</code></td>
<td style={{padding: "9px 8px", backgroundColor: "rgba(255,255,255,0.02)"}}>768p · 24 fps</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>❌</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>❌</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>✅</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>✅</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>❌</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>❌</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>❌</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>❌</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>❌</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>❌</td>
</tr>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>LTX-2.3 (one/two-stage/TI2V/HQ)</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}><code>Lightricks/LTX-2.3</code></td>
@@ -704,6 +725,11 @@ Optimization columns are abbreviated to keep the matrix readable:
difference is transformer depth and width, picked up from
`transformer/config.json` at load time. A single checkpoint serves T2V,
I2V (`--image-path`), and T2I (`--num-frames 1`).
6. FastH3's VSA column refers to the dedicated `video_sparse_attn_h3` (VSA-H3)
backend. Dense backends that run on base MiniMax-H3 also run on the FastH3
weights.
FastH3 serves `t2va` only and rejects `--model-variant` and
`quality: "high"`.
</Accordion>