Commits · 86a32bb5cd50ae1f06fddea9db2633f2f469482e · change / sglang · GitLab

15 Sep, 2025 15 commits
- chore: bump v0.5.3rc0 (#10468) · 86a32bb5
  Yineng Zhang authored Sep 15, 2025
  
  86a32bb5
- feat: support pip install sglang (#10465) · 5afd0365
  Yineng Zhang authored Sep 15, 2025
  
  5afd0365
- Fix trtllm_moe wrong correction bias (#10440) · 059c13de
  fzyzcjy authored Sep 15, 2025
  
  059c13de
- Run tests based on labels (#10456) · 50dc0c1e
  Lianmin Zheng authored Sep 15, 2025
  
  50dc0c1e
- Add rtx5880 moe triton (#10439) · 76becc1d
  Jimmy_L authored Sep 15, 2025
  
  76becc1d
- fix(server_args): Skip chunked_prefill_size validation when disaggregation mode is decode (#10358) · 3795b6a4
  Jimmy authored Sep 15, 2025
  
  3795b6a4
- fix: fix max_new_tokens uninitialized error (#9343) · 0549f21c
  Mick authored Sep 15, 2025
  
  0549f21c
- Add reasoning examples for GPT-OSS in Markdown examples (#9626) · 0b14159f
  Vincent Zhong authored Sep 14, 2025
  
  0b14159f
- metrics: support customer labels specified in request header (#10143) · fc2c3a3d
  Yingchun Lai authored Sep 15, 2025
  
  fc2c3a3d
- Tiny fix wrong naming (#10437) · 01018138
  fzyzcjy authored Sep 15, 2025
  
  01018138
- Refactor TopK to ensure readability and extensibility (#9338) · 4844fac9
  Cheng Wan authored Sep 14, 2025
  
  4844fac9
- [4/N]DP refactor: support watching mode `get_load` and shortest queue strategy (#10201) · 305c9e8c
  Liangsheng Yin authored Sep 15, 2025
  
  305c9e8c
- Fix correction bias undefined behavior for nvfp4 models (#10426) · 258d02c8
  fzyzcjy authored Sep 15, 2025
  
  258d02c8
- Add split tile size for Triton attention (#10425) · 60d7beda
  Ke Bao authored Sep 15, 2025
  
  60d7beda
- [Fix] MoE: fix w8a8_fp8 MoE and add tests to cover this code path (#10429) · 2f8ba6fe
  Cheng Wan authored Sep 14, 2025
  
  2f8ba6fe
14 Sep, 2025 14 commits
- [Feature] Sglang Tracing: Fine-Grained Tracking for Request Latency - Part 1 (#9962) · 4c21b090
  Feng Su authored Sep 15, 2025
```
Signed-off-by: Feng Su <sufeng@linux.alibaba.com>
Signed-off-by: Huaixin Chang <changhuaixin@linux.alibaba.com>
Signed-off-by: Peng Wang <rocking@linux.alibaba.com>
```
  4c21b090
- Typo: in `--enable-custom-logit-processor`: agree with cli arg (#10076) · 165abeeb
  艾力可 authored Sep 14, 2025
  
  165abeeb
- [PD metrics] Fix some uncompleted PD related metrics (#8627) · 21ca4c3a
  Yingchun Lai authored Sep 14, 2025
  
  21ca4c3a
- Support profile args in Engine API (#6539) · 4da55336
  fzyzcjy authored Sep 14, 2025
  
  4da55336
- Support global scale in addition to per expert scale for cutedsl moe (#10270) · ac964d2e
  fzyzcjy authored Sep 14, 2025
  
  ac964d2e
- Support offloading in fp8 (#9948) · fa46e2bd
  fzyzcjy authored Sep 14, 2025
  
  fa46e2bd
- [2/2] Speed up prefill mla attention concat (#10157) · b047b553
  fzyzcjy authored Sep 14, 2025
  
  b047b553
- Fix the global scale fix does not support EPLB and improve enabling condition (#10369) · 2df532ef
  fzyzcjy authored Sep 14, 2025
  
  2df532ef
- Add h200 fused moe config for Qwen3-Next (#10404) · b3c97762
  Ximingwang-09 authored Sep 14, 2025
```
Co-authored-by: 纬杭 <ximing.wxm@antgroup.com>
```
  b3c97762
- Add self.capture_aux_hidden_states For GLM-4.5V (#10228) · b8347b40
  Yuxuan Zhang authored Sep 14, 2025
  
  b8347b40
- Fix cutlass moe accuracy drop caused by attention UB from DP padding mode (#10414) · 72dfa96a
  fzyzcjy authored Sep 14, 2025
  
  72dfa96a
- fix duplicated logger in eager_utils (#10410) · 05b01ef4
  lijin23 authored Sep 14, 2025
  
  05b01ef4
- [Hack] Add pd-disaggregation decode polling interval (#10411) · 55a6e644
  Liangsheng Yin authored Sep 14, 2025
  
  55a6e644
- Remove repeatedly lists adding in `init_incremental_detokenization` (#10412) · 6897e06b
  Liangsheng Yin authored Sep 14, 2025
  
  6897e06b
13 Sep, 2025 8 commits
- [Generative Score API] Scoring(Prefill-only) optimizations. (#9748) · a360511d
  Sundara Raman Ramachandran authored Sep 13, 2025
  
  a360511d
- [Performance] Dynamic Batch Tokenizer (#9382) · 94d0f656
  Sundara Raman Ramachandran authored Sep 13, 2025
  
  94d0f656
- [Fix] Support qwen3-next MTP+DP (#10392) · 97528610
  Binyao Jiang authored Sep 13, 2025
  
  97528610
- support qwen3_next blackwell (#10403) · 297d3745
  Yi Zhang authored Sep 13, 2025
  
  297d3745
- [Fix] Init mamba related memory pools with torch.zeros (#10400) · 31e9d3a5
  Binyao Jiang authored Sep 12, 2025
  
  31e9d3a5
- fix: tool parse in large streaming chunk beginning with normal content (#10397) · 6f4676ef
  Xinyuan Tong authored Sep 12, 2025
```
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
```
  6f4676ef
- fix probs name which without temp scaling name (#9984) · 99757cc3
  narutolhy authored Sep 12, 2025
  
  99757cc3
- [Auto Sync] Update xgrammar_backend.py (20250913) (#10395) · cdddab05
  Lianmin Zheng authored Sep 12, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
```
  cdddab05
12 Sep, 2025 3 commits
- [HiCache] doc: update deployment in readme (#10332) · 49f169d5
  Teng Ma authored Sep 13, 2025
```
Signed-off-by: Teng Ma <sima.mt@alibaba-inc.com>
```
  49f169d5
- [HiCache] fix mooncake config in different tp size (#10377) · 7fce2fd9
  Teng Ma authored Sep 13, 2025
  
  7fce2fd9
- Support Qwen3-Next on Ascend NPU (#10379) · 16cd550c
  Even Zhou authored Sep 13, 2025
  
  16cd550c