Commits · 0b14159fc4e07c57bed6f4603056bb3dd0303e8d · zhaoyu6 / sglang

"vscode:/vscode.git/clone" did not exist on "5fefaa5d4dc3445417d155b639a9657ef6378115"

15 Sep, 2025 14 commits
- Add reasoning examples for GPT-OSS in Markdown examples (#9626) · 0b14159f
  Vincent Zhong authored Sep 14, 2025
  
  0b14159f
- [docs / oneliner] update mmmu docs instruction (#9768) · 1489cd6c
  Vincent Zhong authored Sep 14, 2025
```
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
```
  1489cd6c
- metrics: support customer labels specified in request header (#10143) · fc2c3a3d
  Yingchun Lai authored Sep 15, 2025
  
  fc2c3a3d
- Fix label pr for ci (#10441) · 8f6a1758
  Lianmin Zheng authored Sep 14, 2025
  
  8f6a1758
- 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
- automatically label pr for ci (#10435) · b7d385e8
  Lianmin Zheng authored Sep 14, 2025
  
  b7d385e8
- [4/N]DP refactor: support watching mode `get_load` and shortest queue strategy (#10201) · 305c9e8c
  Liangsheng Yin authored Sep 15, 2025
  
  305c9e8c
- Revert "Fix FA4 import cause moe_fused_gate output be illegal memory" (#10432) · ca63f075
  fzyzcjy authored Sep 15, 2025
  
  ca63f075
- [router]: Add Embedding routing logic (#10129) · f9ee6ae1
  Jintao Zhang authored Sep 15, 2025
```
Signed-off-by: Jintao Zhang <zhangjintao9020@gmail.com>
Co-authored-by: Waël Boukhobza <wawa_wael@live.fr>
```
  f9ee6ae1
- feat: add dsv3 fp4 cutlass moe etp ut (#10433) · dcee42c2
  Yineng Zhang authored Sep 14, 2025
  
  dcee42c2
- 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 19 commits
- fix: enable cu124 and cu128 build on main push (#10431) · 7ce6c10e
  Yineng Zhang authored Sep 14, 2025
  
  7ce6c10e
- fix: use latest flashinfer (#10428) · 55025b92
  Yineng Zhang authored Sep 14, 2025
  
  55025b92
- [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
- Fix sgl-kernel + srt CI (#10419) · e3cf812f
  fzyzcjy authored Sep 14, 2025
  
  e3cf812f
- 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
- Let sgl-kernel changes be tested on srt (#10313) · a0f844ed
  fzyzcjy authored Sep 14, 2025
  
  a0f844ed
- Fix the global scale fix does not support EPLB and improve enabling condition (#10369) · 2df532ef
  fzyzcjy authored Sep 14, 2025
  
  2df532ef
- Auto determine sgl kernel version in blackwell CI (#10318) · abea9250
  fzyzcjy authored Sep 14, 2025
  
  abea9250
- 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 7 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
- Update ROCm docker image to add sgl-router support (#10406) · eca59f96
  kk authored Sep 14, 2025
```
Co-authored-by: Colin Wang <kangwang@amd.com>
```
  eca59f96
- [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