Commits · c0c6f543e472981022c5f25f8b9ffa8d68c49141 · change / sglang · GitLab

16 Sep, 2025 10 commits
- chore: upgrade sgl-kernel 0.3.10 (#10500) · c0c6f543
  Yineng Zhang authored Sep 16, 2025
  
  c0c6f543
- Minor fix lint introduced by #10466 (#10507) · edd6a07b
  Shangming Cai authored Sep 16, 2025
```
Signed-off-by: Shangming Cai <csmthu@gmail.com>
```
  edd6a07b
- feat: update support for qwen3next model (#10466) · b6dd4bcb
  cao1zhg authored Sep 16, 2025
  
  b6dd4bcb
- Cache the result of `is_blackwell` platform check (#10498) · b2435be6
  b8zhong authored Sep 15, 2025
  
  b2435be6
- Remove wrong imports `from sglang.python` (#10493) · fa5d0bf6
  Liangsheng Yin authored Sep 16, 2025
  
  fa5d0bf6
- [ModelOpt] Respect `kv_cache_quant_algo` in ModelOpt checkpoints (#10336) · 80572c83
  brayden-hai authored Sep 15, 2025
```
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
```
  80572c83
- [Hicache] Evaluate Per-Round Metrics in Multiturn Bench (#10203) · 4bb08f6e
  ykwd authored Sep 16, 2025
```
Co-authored-by: Teng Ma <sima.mt@alibaba-inc.com>
```
  4bb08f6e
- Temporay work-around for rocm 7.0.0 alpha with enabling data-parallel issue (#10434) · ec272dda
  kk authored Sep 16, 2025
```
Co-authored-by: wunhuang <wunhuang@amd.com>
Co-authored-by: Sai Enduri <saimanas.enduri@amd.com>
```
  ec272dda
- fix crash of DeepSeek-V3 update_weights_from_disk (#8863) · a220c14f
  scut-cbq authored Sep 16, 2025
```
Co-authored-by: parkeychen <parkeychen@tencent.com>
```
  a220c14f
- [Auto Sync] Update registry.py (20250915) (#10484) · 31fb19a0
  Lianmin Zheng authored Sep 15, 2025
```
Co-authored-by: cctry <shiyang@x.ai>
```
  31fb19a0
15 Sep, 2025 22 commits
- [2/2] Introduce Chunked-SGMV kernels and corresponding LoRA backend for... · 3f41b48c
  Lifu Huang authored Sep 15, 2025
```
[2/2] Introduce Chunked-SGMV kernels and corresponding LoRA backend for improved performance (#10286)
```
  3f41b48c
- [Env] minimal version for organizing envs (#10479) · c3c26f76
  Liangsheng Yin authored Sep 16, 2025
  
  c3c26f76
- Fix `--dataset-path` in `bench_one_batch_server` (#10475) · 2cf811a9
  Liangsheng Yin authored Sep 16, 2025
  
  2cf811a9
- model: support solar (#8189) · a45d9a4e
  Praneth Paruchuri authored Sep 15, 2025
  
  a45d9a4e
- fix: gpt-oss streaming dropping normal content when tools are provided but not used (#9657) · 28c79dc8
  Jonas authored Sep 15, 2025
  
  28c79dc8
- fix(internvl): fix accuracy issue of normalization (#10375) · 1fcccda4
  Kevin Tuan authored Sep 16, 2025
  
  1fcccda4
- [PD metrics] Add latency Histogram metrics of each stage for generate requests (#8710) · b1721edb
  Yingchun Lai authored Sep 16, 2025
  
  b1721edb
- 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 8 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