minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[dLLM] Make FDFO a framework capability for all dLLM algorithms (#27551)

Browse Source 
Co-authored-by: Xiaoyu Zhang <1182563586@qq.com>
This commit is contained in:
Chenchen Hong
2026-07-11 11:05:05 +08:00
committed by GitHub
co-authored by Xiaoyu Zhang
parent fc2ef35308
commit e3ceccf781
14 changed files with 468 additions and 204 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs_new/docs/advanced_features/server_arguments.mdx
+6
View File
@@ -2217,6 +2217,12 @@ Please consult the documentation below and [server_args.py](https://github.com/s
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>`None`</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>Type: str</td>
</tr>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>`--dllm-fdfo`</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>First-Done-First-Out (FDFO) scheduling lets completed requests leave the batch immediately instead of waiting for slower requests, eliminating head-of-line blocking. Enabled by default; pass `--no-dllm-fdfo` to fall back to synchronous lockstep scheduling. Works with any dLLM algorithm.</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>`True`</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>Type: bool</td>
</tr>
</tbody>
</table>
docs_new/docs/supported-models/diffusion_language_models.mdx
+13
View File
@@ -16,6 +16,19 @@ python3 -m sglang.launch_server \
--port 30000
```
## First-Done-First-Out (FDFO) Scheduling
FDFO scheduling is **enabled by default**: each request leaves the batch as soon as its block is resolved, instead of advancing in lockstep where fast-converging requests must wait for slow long-tail requests before leaving the batch (head-of-line blocking). This improves throughput and is orthogonal to `--dllm-algorithm`, so it works with any dLLM algorithm. Pass `--no-dllm-fdfo` to fall back to synchronous lockstep scheduling:
```bash Command
python3 -m sglang.launch_server \
--model-path inclusionAI/LLaDA2.0-mini \
--dllm-algorithm LowConfidence \
--no-dllm-fdfo \
--host 0.0.0.0 \
--port 30000
```
## Example Configuration File
Depending on the algorithm selected, the configuration parameters vary.