Commits · edefab0c6498c96a42228e718b3102220ce4b946 · change / sglang

08 Oct, 2025 2 commits
- [2/2] Support MHA prefill with FlashAttention 4. (#10937) · edefab0c
  Lifu Huang authored Oct 08, 2025
```
Co-authored-by: Hieu Pham <hyhieu@gmail.com>
```
  edefab0c
- [8/N] MoE Refactor: deprecate `EPMoE` (#11211) · 3c06b673
  Cheng Wan authored Oct 07, 2025
  
  3c06b673
07 Oct, 2025 1 commit
- [quantization] Properly ignore quantization for layers excluded in quant_config (#11205) · cd4b39a9
  Bowen Bao authored Oct 07, 2025
  
  cd4b39a9
06 Oct, 2025 3 commits

Enable native ModelOpt quantization support (1/3) (#7149) · 155cbb51
Zhiyu authored Oct 06, 2025
```
Signed-off-by: Zhiyu Cheng <zhiyuc@nvidia.com>
```
155cbb51

Support DeepSeek V3.2 Exp (#11061) · efbc687c

fzyzcjy authored Oct 06, 2025


Co-authored-by: Stefan He <11166516+hebiao064@users.noreply.github.com>
Co-authored-by: Liangsheng Yin <95566987+hnyls2002@users.noreply.github.com>
Co-authored-by: Baizhou Zhang <56809903+fridge003@users.noreply.github.com>
Co-authored-by: DarkSharpness <76582120+darksharpness@users.noreply.github.com>
Co-authored-by: ZhengdQin <46387172+zhengdqin@users.noreply.github.com>
Co-authored-by: DarkSharpness <2040703891@qq.com>
Co-authored-by: hnyls2002 <lsyincs@gmail.com>
Co-authored-by: Zhengda Qin <zhengdqin@gmail.com>
Co-authored-by: Liangsheng Yin <hnyls2002@gmail.com>
Co-authored-by: HAI <hixiao@gmail.com>
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>

efbc687c

[quantization] Enable aiter mxfp4 fused_moe for Quark (#10048) · baee0860
Bowen Bao authored Oct 05, 2025
```
Co-authored-by: HaiShaw <hixiao@gmail.com>
```
baee0860

05 Oct, 2025 1 commit
- [Feat] Support Torch Symm Mem AllReduce (#10571) · 590f2da0
  Yuan Luo authored Oct 06, 2025
```
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
```
  590f2da0
04 Oct, 2025 1 commit
- [fix]enable flashmla when using draft model P/D attention select (#11012) · 666da3d5
  Hank Han authored Oct 04, 2025
  
  666da3d5
03 Oct, 2025 1 commit
- [1/n] Enable DCA CUDA graph capture (#9537) · a2faf894
  b8zhong authored Oct 02, 2025
  
  a2faf894
02 Oct, 2025 7 commits
- Allow use of TRTLLM_MHA backend for hybrid attention on Blackwell (#11138) · e8100774
  Dom Brown authored Oct 03, 2025
  
  e8100774
- Tiny fix ep_gather behavior different in CI (#11130) · 12d68183
  fzyzcjy authored Oct 02, 2025
  
  12d68183
- Tiny cleanup deepseek_v2.py (#11163) · b65db028
  fzyzcjy authored Oct 02, 2025
  
  b65db028
- [model] Add mamba2 and Falcon-H1 support. (#10988) · 083629c2
  ilyasch2 authored Oct 02, 2025
```
Co-authored-by: Younes Belkada <younes.belkada@tii.ae>
Co-authored-by: Younes B <49240599+younesbelkada@users.noreply.github.com>
```
  083629c2
- Support single batch overlap (#10422) · 5e786cca
  fzyzcjy authored Oct 02, 2025
  
  5e786cca
- Support dispatch low latency (#10263) · 0b9dfba7
  fzyzcjy authored Oct 02, 2025
```
Co-authored-by: Kaixi Hou <4001424+kaixih@users.noreply.github.com>
```
  0b9dfba7
- Fuse quantize and rope in trtllm_mla MTP (#10779) · f35def86
  fzyzcjy authored Oct 02, 2025
  
  f35def86
01 Oct, 2025 3 commits
- feat: add fast_decode_plan from flashinfer, flashinfer to 0.4.0rc3 (#10760) · ac1f2928
  eigen authored Oct 01, 2025
```
Co-authored-by: Zihao Ye <yezihhhao@gmail.com>
Co-authored-by: Sleepcoo <Sleepcoo@gmail.com>
```
  ac1f2928
- Quick Fix: fix Qwen3-VL launch failure caused by MRotaryEmbedding arg (#10985) · 61305291
  yhyang201 authored Oct 01, 2025
  
  61305291
- Organize spec-related data structures (#10735) · 73d4a5f8
  Liangsheng Yin authored Oct 01, 2025
  
  73d4a5f8
30 Sep, 2025 6 commits
- Fix DSR1 accuracy for flashinfer_trtllm MoE with FP8 quantization (#11081) · a6cc86df
  Trevor Morris authored Sep 30, 2025
  
  a6cc86df
- [Auto Sync] Update fused_moe_triton_config.py (20250930) (#11099) · 1237aa19
  Lianmin Zheng authored Sep 30, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
```
  1237aa19
- [Fix] Resolve performance drop in speculative decoding aiter backend (#11087) · 59911195
  jacky.cheng authored Sep 30, 2025
  
  59911195
- Enable optional FP32 compute for LM Head (#10729) · d17986f8
  narutolhy authored Sep 29, 2025
```
Thanks to MiniMax Team and Chenyang Zhao's support.
```
  d17986f8
- [model] added support for w8a8int8 used by neuralmagic/Qwen2-0.5B-Ins… (#9642) · 8831c55c
  DevashishLal-CB authored Sep 29, 2025
  
  8831c55c
- Remove hybrid_linear_attn attention backend and refactor attention registry (#10816) · 2bc61dd1
  li-kesen authored Sep 30, 2025
```
Co-authored-by: Yi Zhang <1109276519@qq.com>
```
  2bc61dd1
29 Sep, 2025 1 commit
- [speculative decoding] rename lookahead to ngram (#11010) · 24f7cb1e
  Zhihao Zhang authored Sep 29, 2025
```
Co-authored-by: a4zhangfei <a4zhangfei@qq.com>
```
  24f7cb1e
28 Sep, 2025 2 commits
- Fix FusedSetKVBufferArg in RotaryEmbedding (#11003) · 07440f5f
  Lianmin Zheng authored Sep 28, 2025
  
  07440f5f
- Fix CI failure of TypeError: RotaryEmbedding.forward_cpu() got an unexpected... · 66face35
  YanbingJiang authored Sep 28, 2025
```
Fix CI failure of TypeError: RotaryEmbedding.forward_cpu() got an unexpected keyword argument 'fused_set_kv_buffer_arg' (#11009)
```
  66face35
27 Sep, 2025 2 commits
- [Fix] RuntimeError: get_cfg Unsupported input_type:Float4_e2m1fn_x2 in using... · 8ebf72fe
  kk authored Sep 27, 2025
```
[Fix] RuntimeError: get_cfg Unsupported input_type:Float4_e2m1fn_x2 in using aiter-mxfp4-moe (#10981)
Co-authored-by: wunhuang <wunhuang@amd.com>
```
  8ebf72fe
- fix: fp8 quantization failure of qwen 2.5 VL 7B model (#10112) · 82605747
  Yueyang Pan authored Sep 27, 2025
```
Signed-off-by: PanJason <pyyjason@gmail.com>
```
  82605747
26 Sep, 2025 1 commit
- Revert "Refactor kv_cache_scheme handling for quantization (#10132)" (#10935) · 172bcf01
  Yineng Zhang authored Sep 25, 2025
  
  172bcf01
25 Sep, 2025 1 commit
- Refactor kv_cache_scheme handling for quantization (#10132) · cd4da1f1
  Mohammad Miadh Angkad authored Sep 26, 2025
  
  cd4da1f1
24 Sep, 2025 1 commit
- Restruct sgl-kernel benchmark (#10861) · c4e314f9
  Xiaoyu Zhang authored Sep 25, 2025
  
  c4e314f9
23 Sep, 2025 7 commits

[Auto Sync] Update load_config.py, model_config.py, configu... (20250923) (#10825) · f47a2c67
Lianmin Zheng authored Sep 23, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
```
f47a2c67

[Auto Sync] Update elementwise.py (20250923) (#10823) · e290303e

Lianmin Zheng authored Sep 23, 2025


Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>

e290303e

add tunning files for QWEN-3-NEXT (#10794) · 984730b7
Yiakwy authored Sep 24, 2025

984730b7

model: support qwen3-vl series (#10323) · 4f564b9e

Zheng Li authored Sep 24, 2025


Co-authored-by: ocss884 <ocss.lin@gmail.com>
Co-authored-by: cao1zhg <653506626@qq.com>
Co-authored-by: yhyang201 <yhyang201@gmail.com>
Co-authored-by: yhyang201 <47235274+yhyang201@users.noreply.github.com>
Co-authored-by: 瑀澈 <yuche.lz@alibaba-inc.com>
Co-authored-by: Mick <mickjagger19@icloud.com>
Co-authored-by: Yineng Zhang <me@zhyncs.com>

4f564b9e

Fix trtllm_mla slow concat kernel in MTP (#10777) · 063c3791
fzyzcjy authored Sep 23, 2025

063c3791
[Auto Sync] Update configurer.py (20250923) (#10765) · 113f8f65
Lianmin Zheng authored Sep 22, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
```
113f8f65
[Ascend]optimize Qwen3 on Ascend (#10574) · e22f3a5e
ronnie_zheng authored Sep 23, 2025
```
Co-authored-by: c30031083 <chenxu140@huawei.com>
```
e22f3a5e