Commits · e2ac7888b8cb1fd6c33a7ec58d27a5f5b5b24e0c · change / sglang · GitLab

22 Sep, 2025 4 commits
- [2/2] Support deterministic inference for temperature > 0 (#10678) · e2ac7888
  Qiaolin Yu authored Sep 21, 2025
```
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
Co-authored-by: hebiao064 <hebiaobuaa@gmail.com>
```
  e2ac7888
- [deterministic inference] Move batch invariant pkg to sglang (#10695) · 86527a47
  Stefan He authored Sep 21, 2025
  
  86527a47
- Support deterministic inference with triton backend (#10694) · 134b4f7e
  Ethan (Yusheng) Su authored Sep 21, 2025
  
  134b4f7e
- [Auto Sync] Update deepseek_v2.py (20250922) (#10717) · f67d1f45
  Yineng Zhang authored Sep 21, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Michael Granado <mgranado@together.ai>
```
  f67d1f45
21 Sep, 2025 5 commits
- Refactors radix cache for extra key support (#10317) · 12d6cf18
  Xinyuan Tong authored Sep 21, 2025
```
Signed-off-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
```
  12d6cf18
- [3/4] Speed up CSGMV backend perf by 10% through dynamic chunking + kernel optimization (#10592) · 08ecd0aa
  Lifu Huang authored Sep 20, 2025
  
  08ecd0aa
- Super tiny fix extra logs (#10697) · 720c1c8c
  fzyzcjy authored Sep 21, 2025
  
  720c1c8c
- feat: update server args (#10696) · d403c143
  Yineng Zhang authored Sep 20, 2025
```
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
```
  d403c143
- [Feature] Support deterministic inference with FA3 backend (#10651) · cba0d8c3
  Stefan He authored Sep 20, 2025
  
  cba0d8c3
20 Sep, 2025 8 commits
- [Auto Sync] Update modelopt_quant.py (20250920) (#10688) · f1d78923
  Yineng Zhang authored Sep 20, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
```
  f1d78923
- fix: remove awq_dequantize deps (#10686) · 7c876de7
  Yineng Zhang authored Sep 20, 2025
  
  7c876de7
- fix(disagg): fix sending KV cache in case of MLA for NIXL backend (#10673) · 2b7417bf
  dmitrygx authored Sep 20, 2025
  
  2b7417bf
- Replace os.environ in layernorm.py (#10684) · f1116495
  Baizhou Zhang authored Sep 20, 2025
  
  f1116495
- [Performance] Qwen3-Next: optimize causal_conv1d_fn triton kernel - up to 9% faster (#10680) · bd7eb020
  Binyao Jiang authored Sep 20, 2025
  
  bd7eb020
- [Auto Sync] Update deepseek_v2.py (20250920) (#10683) · b17e67df
  Yineng Zhang authored Sep 19, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
```
  b17e67df
- [1/2] Support deterministic inference with flashinfer attention backend (#10645) · 8ecef73f
  Baizhou Zhang authored Sep 19, 2025
```
Co-authored-by: hebiao064 <hebiaobuaa@gmail.com>
Co-authored-by: Qiaolin-Yu <liin1211@outlook.com>
```
  8ecef73f
- [Auto Sync] Update model_runner.py (20250920) (#10679) · 60e2a7ce
  Yineng Zhang authored Sep 19, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
```
  60e2a7ce
19 Sep, 2025 12 commits
- limit sgl-kernel causal conv1d to cuda only (#10648) · d88ef4a3
  Jinyan Chen authored Sep 20, 2025
```
Co-authored-by: Jinyan Chen <jinyanc@nvidia.com>
```
  d88ef4a3
- chore: cleanup docker image (#10671) · 6f993e8b
  Yineng Zhang authored Sep 19, 2025
  
  6f993e8b
- [Auto Sync] Update .clang-format (20250919) (#10670) · dab4663b
  Yineng Zhang authored Sep 19, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
```
  dab4663b
- fix: resolve sync issue (#10668) · 610a6d6e
  Yineng Zhang authored Sep 19, 2025
  
  610a6d6e
- [HiCacheStorage]support page_first_direct layout for generic set&get (#10522) · 7f399e4b
  huangtingwei authored Sep 19, 2025
  
  7f399e4b
- Fix fast decode plan for flashinfer v0.4.0rc1 and upgrade sgl-kernel 0.3.11 (#10634) · 3fa3c22a
  Baizhou Zhang authored Sep 19, 2025
```
Co-authored-by: zhyncs <me@zhyncs.com>
```
  3fa3c22a
- [Feature]feat(get_ip): unify get_ip_xxx (#10081) · 56b991b1
  Jimmy authored Sep 19, 2025
  
  56b991b1
- Hicache L3 backend mooncake optimization configuration reading method (#10319) · 8b713c72
  FlyPanda authored Sep 19, 2025
```
Co-authored-by: Teng Ma <sima.mt@alibaba-inc.com>
Co-authored-by: huangtingwei <141888744+huangtingwei9988@users.noreply.github.com>
Co-authored-by: shicang <shicang@shicang>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
```
  8b713c72
- feat: add fused moe config for Qwen3-Next-80B-A3B-Instruct on B200 (#10631) · 8c52de6f
  zixuanzhang226 authored Sep 18, 2025
  
  8c52de6f
- model support: Sarashina2VisionForCausalLM (#10632) · c1815a99
  Chang Su authored Sep 18, 2025
  
  c1815a99
- [Performance] Qwen3-Next: speed up update_mamba_state_after_mtp_verify by 10x;... · 4e6c4923
  Binyao Jiang authored Sep 18, 2025
```
[Performance] Qwen3-Next: speed up update_mamba_state_after_mtp_verify by 10x; e2e up to 3.54% faster (#10586)
```
  4e6c4923
- [Performance] Qwen3-Next: replace arange to cached query_start_loc_li… (#10553) · b91cb67e
  Binyao Jiang authored Sep 18, 2025
  
  b91cb67e
18 Sep, 2025 11 commits
- [Feature] Speculative decoding support lookahead (#9873) · e7bc6003
  Zhihao Zhang authored Sep 19, 2025
```
Co-authored-by: a4zhangfei <a4zhangfei@qq.com>
Co-authored-by: Qiaolin-Yu <liin1211@outlook.com>
```
  e7bc6003
- refactor: use registry for _get_attention_backend_from_str (#10629) · 2a2ff9a8
  Yineng Zhang authored Sep 18, 2025
  
  2a2ff9a8
- Garbage collector regression in the online server (#10621) · 67073dde
  brayden-hai authored Sep 18, 2025
  
  67073dde
- feat: Add FlexAttention Backend for Efficient Sparse Attention (#9947) · 9a5c42f9
  yuk.igalaxy authored Sep 19, 2025
```
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
```
  9a5c42f9
- Fix bias handling in TritonMoeQuantInfo within quantization/mxfp4.py (#10579) · 388c05d5
  yhyang201 authored Sep 19, 2025
  
  388c05d5
- [Performance] qwen3-next improve causal conv1d in prefill phase (#10595) · fc809665
  Jinyan Chen authored Sep 19, 2025
```
Co-authored-by: Jinyan Chen <jinyanc@nvidia.com>
```
  fc809665
- support qwen3-next-fp8 deepep (#10622) · 1344ebc8
  Yi Zhang authored Sep 19, 2025
  
  1344ebc8
- Feat/add heartbeat mechanism for nixl conn (#10222) · 52f248cd
  shaharmor98 authored Sep 18, 2025
```
Signed-off-by: Shahar Mor <smor@nvidia.com>
```
  52f248cd
- [RL] Add destroy process group api (#9979) · 93f75778
  penguin_wwy authored Sep 19, 2025
  
  93f75778
- fix deepep assert when PD disaggregation == null (#8274) · 4039c626
  AlphaBaby authored Sep 19, 2025
```
Co-authored-by: fujianhao.fjh <fujianhao.fjh@antgroup.com>
```
  4039c626
- Scale kkt after reduction (#10604) · db71c38f
  Yi Zhang authored Sep 18, 2025
  
  db71c38f