Commits · 0dd6cf16bac0b02b69ff78a5595aea1a68c46647 · change / sglang · GitLab

16 Oct, 2025 10 commits
- [ci]use H20 to run disaggregation test (#11543) · 0dd6cf16
  Hank Han authored Oct 17, 2025
  
  0dd6cf16
- [Bugfix] Fix Qwen3/DSV3/DSV3.2 model support (#11510) · 3cceaa38
  Even Zhou authored Oct 16, 2025
  
  3cceaa38
- Set csgmv as default lora backend. (#11488) · b0d20cde
  Lifu Huang authored Oct 15, 2025
  
  b0d20cde
- Split test_intel_amx_attention_backend.py to pass CI of timeout (#11370) · cbac4997
  YanbingJiang authored Oct 16, 2025
```
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
```
  cbac4997
- Fix missing a2a backend init of GLM4.5 MoE Block (#11692) · 476c67d7
  Shangming Cai authored Oct 16, 2025
```
Signed-off-by: Shangming Cai <csmthu@gmail.com>
```
  476c67d7
- [PD] Add PD support for hybrid model (Qwen3-Next, DeepSeek V3.2 Exp) (#10912) · 868403f6
  Shangming Cai authored Oct 16, 2025
```
Signed-off-by: Shangming Cai <csmthu@gmail.com>
Co-authored-by: hzh0425 <hzh0425@apache.org>
Co-authored-by: ZeldaHuang <hzm414167@alibaba-inc.com>
```
  868403f6
- [Mamba] Increase default mamba_full_memory_ratio to 0.9 (#11679) · 97d857c0
  Hanming Lu authored Oct 15, 2025
  
  97d857c0
- Sync code and test CI; rename some env vars (#11686) · cd7e1bd5
  Lianmin Zheng authored Oct 15, 2025
  
  cd7e1bd5
- enable rmsnorm on XPU (#10248) · 729b7edf
  Huaiyu, Zheng authored Oct 16, 2025
  
  729b7edf
- [CI][XPU]enable sglang CI on Intel XPU (#9493) · 4c03dbaa
  DiweiSun authored Oct 16, 2025
```
Co-authored-by: huaiyuzh <huaiyu.zheng@intel.com>
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
```
  4c03dbaa
15 Oct, 2025 13 commits
- chore: bump SGLang version to 0.5.3.post2 (#11680) · baf277a9
  sglang-bot authored Oct 15, 2025
```
Co-authored-by: sglang-bot <sglang-bot@users.noreply.github.com>
```
  baf277a9
- Fix missing json imports in serving_responses.py (#11681) · f226d3da
  Chang Su authored Oct 15, 2025
  
  f226d3da
- [tool call] Fix prev_tool_call_arr management in base_format_detector.py (#11367) · 30ea4c46
  Chang Su authored Oct 15, 2025
```
Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
```
  30ea4c46
- Fix 1-step draft model forward (#11653) · 6d036468
  Shangming Cai authored Oct 15, 2025
```
Signed-off-by: Shangming Cai <csmthu@gmail.com>
Co-authored-by: Liangsheng Yin <lsyincs@gmail.com>
```
  6d036468
- Tiny cleanup some eagle unused codes (#11660) · 8221f9ae
  Liangsheng Yin authored Oct 15, 2025
  
  8221f9ae
- Clean up some Qwen3-Next and deterministic code (#11585) · 6b143d62
  Stefan He authored Oct 15, 2025
  
  6b143d62
- [BugFix][Qwen3-VL]: fix cu_seqlens in qwen3-vl (#11458) · b2c85669
  Zheng Wengang authored Oct 15, 2025
  
  b2c85669
- feat: add add_chunked_prefix_cache_attention_backend (#11636) · 91fc5bb5
  Yineng Zhang authored Oct 14, 2025
  
  91fc5bb5
- [Fix] Fix accuracy bug in CSGMV kernel caching key. (#11579) · 780fbf2f
  Lifu Huang authored Oct 14, 2025
  
  780fbf2f
- [1/N]Support DeepSeek-R1 w4a8 normal deepep (#8247) · 825432fc
  Jinwu authored Oct 15, 2025
```
Co-authored-by: Hank Han <hanhan7630@outlook.com>
```
  825432fc
- [1/N] Introduce Mooncake Backend and Mooncake EP to Support Elastic EP (#10423) · a40229f6
  Xun Sun authored Oct 15, 2025
```
Co-authored-by: Hank Han <hanhan7630@outlook.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
```
  a40229f6
- fix: upgrade transformers to 4.57.1 (#11628) · e9e120ac
  Sahithi Chigurupati authored Oct 14, 2025
```
Signed-off-by: Sahithi Chigurupati <chigurupati.sahithi@gmail.com>
Co-authored-by: zhyncs <me@zhyncs.com>
```
  e9e120ac
- Refactor kv cache free (#11351) · 1d7f7835
  cctry authored Oct 14, 2025
  
  1d7f7835
14 Oct, 2025 17 commits
- [router][grpc] add warm up to grpc server (#11627) · 32595146
  Simo Lin authored Oct 14, 2025
```
Co-authored-by: Chang Su <chang.s.su@oracle.com>
```
  32595146
- [Minor] Update xgrammar dependency (#11622) · e28c9e52
  DarkSharpness authored Oct 15, 2025
  
  e28c9e52
- [Auto Sync] Update collector.py (20251014) (#11625) · b98cf398
  Lianmin Zheng authored Oct 14, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Byron Hsu <byronhsu1230@gmail.com>
```
  b98cf398
- [Auto Sync] Update scheduler.py, server_args.py (20251014) (#11623) · 27d71045
  Lianmin Zheng authored Oct 14, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
```
  27d71045
- Fix log for chunked prefix cache (#11624) · c224a4c6
  Baizhou Zhang authored Oct 14, 2025
  
  c224a4c6
- use non_blocking h2d in ForwardBatch.prepare_mlp_sync_batch. (#11605) · 94d26d85
  strgrb authored Oct 15, 2025
  
  94d26d85
- Reduce one step decode for draft model. (#11561) · 5ea96ac7
  Liangsheng Yin authored Oct 14, 2025
  
  5ea96ac7
- move eagle draft post process to cuda graph (#11434) · 56222658
  yinghui authored Oct 14, 2025
```
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>
```
  56222658
- make radix cache deterministic (#10721) · dc965db0
  Alex Chi Z authored Oct 14, 2025
```
Signed-off-by: Alex Chi Z <iskyzh@gmail.com>
```
  dc965db0
- Refactor spec decoding metrics calculation into separate `TokenizerManager`... · 817e46f4
  Scott Lee authored Oct 14, 2025
```
Refactor spec decoding metrics calculation into separate `TokenizerManager` utility function (#11586)
```
  817e46f4
- Optimize Triton Draft Backend (#11556) · 5a33c3aa
  Liangsheng Yin authored Oct 14, 2025
  
  5a33c3aa
- Add fused_moe_triton config: triton_3_4_0/E=256,N=256,device_name=NVIDIA_B200.json (#11587) · e4358a45
  Qiaolin Yu authored Oct 13, 2025
  
  e4358a45
- [Auto Sync] Update model_config.py (20251014) (#11580) · ba2ce28f
  Lianmin Zheng authored Oct 13, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Hanming Lu <69857889+hanming-lu@users.noreply.github.com>
```
  ba2ce28f
- Implement LRU eviction policy for LoRA adapters (#11041) · 28f80b12
  Chenxi Li authored Oct 13, 2025
  
  28f80b12
- bench_serving support PD Disaggregation (#11542) · 88a6f9da
  Xiaoyu Zhang authored Oct 14, 2025
  
  88a6f9da
- Make DeepEP combine recv do not overlap (#11535) · cb8ed2c0
  fzyzcjy authored Oct 14, 2025
  
  cb8ed2c0
- [DSv32] Use torch.compile for _get_logits_head_gate (#11565) · 38473363
  Trevor Morris authored Oct 13, 2025
  
  38473363