EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[PD]decode: add CLIP_MAX_NEW_TOKEN for pop_preallocated (#8866)

Browse Source
This commit is contained in:
Jimmy
2025-08-12 04:08:11 +08:00
committed by GitHub
parent 3d64fda376
commit 0d9e89ec69
2 changed files with 11 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
docs/advanced_features/pd_disaggregation.md
View File

@@ -67,6 +67,7 @@ Please be aware that this setting will cause prefill instances to take a longer
| **`SGLANG_DISAGGREGATION_HEARTBEAT_INTERVAL`** | Interval (seconds) between health checks to prefill bootstrap servers | `5.0` |
| **`SGLANG_DISAGGREGATION_HEARTBEAT_MAX_FAILURE`** | Consecutive heartbeat failures before marking prefill server offline | `2` |
| **`SGLANG_DISAGGREGATION_WAITING_TIMEOUT`** | Timeout (seconds) for receiving KV Cache after request initialization | `300` |
| **`SGLANG_CLIP_MAX_NEW_TOKENS_ESTIMATION`** | Clip request param "max_tokens" to pre_allocate | `4096` |
If a greater mean TTFT is acceptable, you can `export SGLANG_DISAGGREGATION_WAITING_TIMEOUT=600` (10 minutes) to relax the timeout condition.