Commits · ec92b0cefe08ded451eaa8549b3d75941868f71c · change / sglang · GitLab

29 Oct, 2025 8 commits
- EPLB: prefer to use physical experts in the same gpu or node (#10874) · ec92b0ce
  Yingchun Lai authored Oct 29, 2025
  
  ec92b0ce
- doc: improve modelopt error description (#12269) · e03b6bee
  Liana Koleva authored Oct 28, 2025
  
  e03b6bee
- [metrics][EPLB]: Support selected count of physical experts on each GPU (#9825) · 5e36a0b4
  Yingchun Lai authored Oct 29, 2025
  
  5e36a0b4
- a tiny fix for support deepseek bf16 weights (#12313) · 0297773a
  Gao016 authored Oct 29, 2025
```
Co-authored-by: gaochang <gaochang@U-19PX2WQ1-0350.local>
```
  0297773a
- [hotfix] Fix pytest not found in CI (#12311) · 587deb15
  Baizhou Zhang authored Oct 28, 2025
  
  587deb15
- [hotfix] missing `w13_weight_fp8` and `w2_weight_fp8` in UE8M0 requantization (#12259) · 83087247
  Cheng Wan authored Oct 28, 2025
  
  83087247
- Add continuous_usage_stats support for streaming responses (#12241) · 334543ff
  Xiaoyu Zhang authored Oct 29, 2025
  
  334543ff
- fix: Llama 4 BF16 load on Blackwell (#12308) · c143f416
  b8zhong authored Oct 28, 2025
  
  c143f416
28 Oct, 2025 32 commits
- [router][grpc] Fix inconsistent behavior of conversation_id not found (#12299) · b48354c5
  Chang Su authored Oct 28, 2025
  
  b48354c5
- Super tiny fix expert distribution dump error (#12271) · 29195aaa
  fzyzcjy authored Oct 29, 2025
  
  29195aaa
- Update deepseek_v32.md (#12296) · 0ee831de
  hlu1 authored Oct 28, 2025
  
  0ee831de
- fix seqlen bug for trtllm_mla's draft_extend (#12295) · 8d6ab1cb
  bmac3 authored Oct 28, 2025
  
  8d6ab1cb
- [router] support arm, windows, mac, linux, reduce wheel size and number (#12285) · 84a9d0ea
  Simo Lin authored Oct 28, 2025
  
  84a9d0ea
- [rust][ci] Add end-to-end tests for Oracle history backend (#12233) · 737b58d6
  Keyang Ru authored Oct 28, 2025
  
  737b58d6
- Use Flashinfer TRT-LLM as Llama 4 compatible MoE backend (#11928) · 77225d60
  b8zhong authored Oct 28, 2025
  
  77225d60
- doc for logit_bias (#12188) · 9c6e25d2
  ybyang authored Oct 29, 2025
  
  9c6e25d2
- Tiny fix sgl-kernel related CI installing the wrong binary (#12283) · 2a3763c3
  fzyzcjy authored Oct 29, 2025
  
  2a3763c3
- Fix 'BypassedTopKOutput' object has no attribute 'topk_weights' for DeepEP (#12231) · fdd00295
  Trevor Morris authored Oct 28, 2025
  
  fdd00295
- [router] upgrade grpc dependency and py 3.13 3.14 support (#12284) · 25e73640
  Simo Lin authored Oct 28, 2025
  
  25e73640
- Modify rocm.Dockerfile (#12274) · 0da9845e
  sogalin authored Oct 28, 2025
  
  0da9845e
- [router] Fix type unmatch during validation (#12257) · 92885441
  Keyang Ru authored Oct 28, 2025
  
  92885441
- chore: cleanup quant deps (#12268) · 64cf868e
  Yineng Zhang authored Oct 28, 2025
  
  64cf868e
- Revert "[Feature] PD-Multiplexing Context and Scheduler." (#12267) · ea399527
  Yineng Zhang authored Oct 28, 2025
  
  ea399527
- Fix potential eos bug on decode instance when PD is enabled (#12206) · 41a11335
  Shangming Cai authored Oct 28, 2025
```
Signed-off-by: Shangming Cai <csmthu@gmail.com>
```
  41a11335
- [Bug fix] [PP] fix wrong dtype for quantified model (#12247) · a1f2dc90
  Xuchun Shang authored Oct 28, 2025
```
Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>
```
  a1f2dc90
- [Feature] Sglang Tracing: Fine-Grained Tracking for Request Latency - Part 2 (#10804) · ea961060
  Feng Su authored Oct 28, 2025
```
Signed-off-by: Feng Su <sufeng@linux.alibaba.com>
```
  ea961060
- [hotfix] Incorrect CombineOverlapArgs in SBO (#12230) · b1e13e7c
  Cheng Wan authored Oct 28, 2025
  
  b1e13e7c
- Feature/Add GET endpoint to query loaded LoRA adapters (#12229) · cc7b04a2
  Chenxi Li authored Oct 28, 2025
  
  cc7b04a2
- [router] configure workflow retries and timeout based on routerConfig (#12252) · d85d6dba
  Simo Lin authored Oct 28, 2025
  
  d85d6dba
- [router] use mcp struct from sdk and clean up code across codebase (#12249) · c5642a7a
  Simo Lin authored Oct 28, 2025
  
  c5642a7a
- Support releasing CUDA graph memory when paused (#7873) · 691c8534
  fzyzcjy authored Oct 28, 2025
```
Co-authored-by: ryang-max <y1cunhui.yang@gmail.com>
Co-authored-by: ryang <38470282+ryang-max@users.noreply.github.com>
```
  691c8534
- Opt fused triton moe: add tma for down proj kernel (#10567) · d2b8c412
  Yongfei Xu authored Oct 28, 2025
```
Co-authored-by: ybyang <10629930+whybeyoung@users.noreply.github.com>
```
  d2b8c412
- Add per-request retraction count (#11177) · bf8f7a94
  Scott Lee authored Oct 27, 2025
  
  bf8f7a94
- [DeepseekV32] Enable flashmla_prefill kernel with fp8 kvcache (#11655) · 81a632ac
  hlu1 authored Oct 27, 2025
```
Signed-off-by: Hao Lu <14827759+hlu1@users.noreply.github.com>
```
  81a632ac
- [router] remove code duplication (#12245) · 83b22400
  Simo Lin authored Oct 27, 2025
  
  83b22400
- docker: add CUDA13 support in dockerfile and update GDRCopy/NVSHMEM for blackwell support (#11517) · 285a8e69
  ishandhanani authored Oct 27, 2025
```
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
```
  285a8e69
- [2/2] Use moe_sum_reduce cuda kernel (#10654) · 813bd6f8
  Yuan Luo authored Oct 28, 2025
```
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
Co-authored-by: huangtingwei <141888744+huangtingwei9988@users.noreply.github.com>
```
  813bd6f8
- Update openai package version to 2.6.1 (#12222) · 729f612d
  Xinyuan Tong authored Oct 27, 2025
  
  729f612d
- Use explicit uint64 dtype for Tensor data_ptr() to avoid overflow (#11994) · 899453ac
  jianan-gu authored Oct 28, 2025
  
  899453ac
- Add env var to control custom Triton kernel cache and set CSGMV as default backend. (#12176) · ce832d70
  Lifu Huang authored Oct 27, 2025
  
  ce832d70