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

[diffusion] chore: use native qwen2.5-vl generation (#34896)

Browse Source
This commit is contained in:
Mick
2026-08-16 09:57:39 +08:00
committed by GitHub
parent 4a6dc267e1
commit eb6b773149
16 changed files with 1010 additions and 116 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/docs/sglang-diffusion/compatibility_matrix.mdx
+4
View File
@@ -37,6 +37,10 @@ Rows are grouped when a family shares the same runtime path or optimization supp
<td>Qwen-Image</td>
<td><div className="sgd-id-list"><code>Qwen/Qwen-Image</code><code>Qwen/Qwen-Image-2512</code><code>Qwen/Qwen-Image-Edit</code><code>Qwen/Qwen-Image-Edit-2509</code><code>Qwen/Qwen-Image-Edit-2511</code><code>Qwen/Qwen-Image-Layered</code></div></td>
</tr>
<tr>
<td>LongCat-Image</td>
<td><div className="sgd-id-list"><code>meituan-longcat/LongCat-Image</code></div></td>
</tr>
<tr>
<td>SD3 / SD3.5</td>
<td><div className="sgd-id-list"><code>stabilityai/stable-diffusion-3-medium</code><code>stabilityai/stable-diffusion-3-medium-diffusers</code><code>stabilityai/stable-diffusion-3.5-medium</code><code>stabilityai/stable-diffusion-3.5-medium-diffusers</code><code>stabilityai/stable-diffusion-3.5-large</code><code>stabilityai/stable-diffusion-3.5-large-diffusers</code></div></td>
docs/docs/sglang-diffusion/models_with_ar.mdx
+20 -5
View File
@@ -1,11 +1,16 @@
---
title: "Diffusion models with AR stage like GLM-Image"
description: "Run diffusion pipelines that delegate an AR stage to a separate SGLang server, such as GLM-Image."
title: "Diffusion models with autoregressive stages"
description: "Run diffusion pipelines with in-process or separately deployed autoregressive encoders."
---
## Quick Start
SGLang Diffusion supports two AR execution paths. Qwen Image Layered and
LongCat-Image use the native Qwen2.5-VL component in process. GLM-Image can use
its bundled Transformers implementation or delegate AR inference to a separate
SGLang server.
Run model with transformers implementation for AR stage (default)
## GLM-Image Quick Start
Run GLM-Image with its bundled Transformers implementation (default):
```bash
# Terminal 1 : launch server
sglang serve --model-path zai-org/GLM-Image --port ${PORT}
@@ -20,7 +25,7 @@ curl http://${HOST}:${PORT}/v1/images/generations \
"size": "widthxheight"
}'
```
Run model with SGLang srt implementation for AR stage (high performance)
Run GLM-Image with a separate SGLang server for its AR stage:
```bash
# Terminal 1 : launch server with AR model
sglang serve --model-path /path/to/zai-org/GLM-Image/vision_language_encoder/ \
@@ -64,6 +69,16 @@ curl http://${HOST}:${PORT}/v1/images/generations \
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)", whiteSpace: "nowrap"}}>T2I, I2I, V2I</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>T2I</td>
</tr>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>Qwen Image Layered</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>Not used</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>I2I, in-process native Qwen2.5-VL</td>
</tr>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>LongCat-Image</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>Not used</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>T2I, in-process native Qwen2.5-VL</td>
</tr>
</tbody>
</table>