Commits · 9a7ced4e4dea7647e4b5aee098b8b19b96cd2c8b · change / sglang

07 Sep, 2025 2 commits

[Feature] LMCache Connector Integration (#9741) · 9a7ced4e

Yuwei An authored Sep 06, 2025


Signed-off-by: Oasis-Git <ayw.sirius19@gmail.com>
Signed-off-by: YuhanLiu11 <yliu738@wisc.edu>
Co-authored-by: Zhiqiang Xie <xiezhq@stanford.edu>

9a7ced4e

Optimize moe_sum_reduce_kernel (#9477) · cb3918a0

Yuan Luo authored Sep 07, 2025


Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
Co-authored-by: Xiaoyu Zhang <35585791+BBuf@users.noreply.github.com>

cb3918a0

06 Sep, 2025 24 commits
- [Auto Sync] Update server_args.py (20250906) (#10117) · f3b67602
  Lianmin Zheng authored Sep 06, 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>
```
  f3b67602
- [router] Improve the router e2e tests (#10102) · 9eb50ecc
  Keyang Ru authored Sep 06, 2025
  
  9eb50ecc
- increase the rust e2e timeout (#10116) · b3e7a2ce
  Keyang Ru authored Sep 06, 2025
  
  b3e7a2ce
- [CI] Refactor disaggregation tests (#10068) · 00974e4f
  Shangming Cai authored Sep 06, 2025
```
Signed-off-by: Shangming Cai <csmthu@gmail.com>
```
  00974e4f
- [chore] Remove unused ep_moe cuda kernels (#9956) · 5f1eb204
  hlu1 authored Sep 06, 2025
  
  5f1eb204
- Remove non-accelerated targets(100 and up) from cmake (#10041) · 039cef76
  hlu1 authored Sep 06, 2025
```
Signed-off-by: Hao Lu <14827759+hlu1@users.noreply.github.com>
```
  039cef76
- Disable kernel cutlass_mla_decode on SM103 (#10058) · 4c22ebe2
  hlu1 authored Sep 06, 2025
```
Signed-off-by: Hao Lu <14827759+hlu1@users.noreply.github.com>
```
  4c22ebe2
- Fix circular import (#10107) · a5a03209
  Cheng Wan authored Sep 06, 2025
  
  a5a03209
- [Fix] Compatibility between DP attention and pipeline parallelism (#10100) · 21af5c04
  Cheng Wan authored Sep 06, 2025
```
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
```
  21af5c04
- perf: Avoid unnecessary data type conversions for DeepSeek-V3 on Blackwell (#9834) · 012584ec
  Jinyang Yuan authored Sep 06, 2025
```
Signed-off-by: Jinyang Yuan <154768711+jinyangyuan-nvidia@users.noreply.github.com>
```
  012584ec
- [NVIDIA] disable chunked prefix cache when dp and blackwell is used (#9861) · 90dfe3de
  Kaixi Hou authored Sep 05, 2025
  
  90dfe3de
- [NVIDIA] Remove unused `get_fused_moe_impl_class` function (#9764) · 9a719b7a
  Kaixi Hou authored Sep 05, 2025
  
  9a719b7a
- [7/N] MoE Refactor: the implementation of new framework (#9269) · 3fa62da7
  Cheng Wan authored Sep 05, 2025
  
  3fa62da7
- [Fix] illegal sync based on undefined behaviour (#9620) · dbb1235d
  DevashishLal-CB authored Sep 05, 2025
```
Signed-off-by: Devashish Lal <devashish@rivosinc.com>
Co-authored-by: Xiaoyu Zhang <35585791+BBuf@users.noreply.github.com>
```
  dbb1235d
- fix double sparsity initialization (#6905) · ad26f298
  Chi-Chih Chang authored Sep 06, 2025
  
  ad26f298
- Fix RMSNorm API CALL mismatch issue. (#10032) · 8d114f25
  sogalin authored Sep 06, 2025
```
Co-authored-by: Hubert Lu <Hubert.Lu@amd.com>
```
  8d114f25
- Revert "[1/N][Bug] Fix w4afp8 MoE NaN issue (sgl-kernel) (#9953)" (#10097) · 0e78c63c
  Yineng Zhang authored Sep 05, 2025
  
  0e78c63c
- Modify ci workflow for auto-partitioning in 2-GPU backend tests (#10029) · 1a3d6f31
  hzh0425 authored Sep 06, 2025
  
  1a3d6f31
- [HiStorage] Remove delete and clear as necessary methods (#10039) · 0b8c5721
  Zhiqiang Xie authored Sep 05, 2025
  
  0b8c5721
- Add lora_path argument to bench_multiturn.py (#10092) · beac202b
  Baizhou Zhang authored Sep 05, 2025
  
  beac202b
- [router] Introduce router integration tests (#10086) · 21b9a4b4
  Keyang Ru authored Sep 05, 2025
  
  21b9a4b4
- [router] move to mcp sdk instead (#10057) · db37422c
  Simo Lin authored Sep 05, 2025
  
  db37422c
- support Llama4 with non uniformed intermediate size across layers for… (#10047) · ab62b135
  gongwei-130 authored Sep 05, 2025
  
  ab62b135
- [Minor] Refactors KV memory pool (#9842) · 273b2834
  Xinyuan Tong authored Sep 06, 2025
  
  273b2834
05 Sep, 2025 14 commits
- Add storage read/write bandwidth logs to monitor kvcache performance (#9965) · f84db115
  pansicheng authored Sep 06, 2025
```
Co-authored-by: Zhiqiang Xie <xiezhq@stanford.edu>
```
  f84db115
- Update wave-lang to 3.7.0 and unify Wave kernel buffer options (#10069) · efb0de2c
  jacky.cheng authored Sep 06, 2025
  
  efb0de2c
- [Bug Fix] Fix Glm4vVisionBlock norm (#9884) · 0f6ac5e2
  Adam Yanxiao Zhao authored Sep 06, 2025
  
  0f6ac5e2
- Update flashinfer to 0.3.1 for B300 support (#10087) · 29850900
  hlu1 authored Sep 05, 2025
```
Signed-off-by: Hao Lu <14827759+hlu1@users.noreply.github.com>
```
  29850900
- [bugfix]: use correct cache location for cross attention in torch native backend (#8622) · e678cc71
  Mahmoud Ashraf authored Sep 05, 2025
  
  e678cc71
- enable aiter gemm_a8w8_bpreshuffle for ptpc gemm (#8555) · 4efe844a
  Morpheus Guo authored Sep 06, 2025
  
  4efe844a
- [router] add rust cache in benchmark ci (#10080) · bde73ee4
  Simo Lin authored Sep 05, 2025
  
  bde73ee4
- [router] add rust cache for rust unit test (#10079) · 4f0e28d7
  Keyang Ru authored Sep 05, 2025
  
  4f0e28d7
- [router] add py binding unit tests to coverage 80% (#10043) · 045ab92d
  Keyang Ru authored Sep 05, 2025
  
  045ab92d
- Support copying tensor from cpu to gpu without using copy engines (#10007) · bd7f8821
  fzyzcjy authored Sep 05, 2025
  
  bd7f8821
- Tiny let DeepGEMM scale checks cover more cases (#7182) · 5e5c30d9
  fzyzcjy authored Sep 05, 2025
```
Co-authored-by: Yineng Zhang <me@zhyncs.com>
```
  5e5c30d9
- Fix and enhance dumper (#8725) · 9f00ec44
  fzyzcjy authored Sep 05, 2025
  
  9f00ec44
- Support simple evals in text comparator (#8867) · 8e85ee88
  fzyzcjy authored Sep 05, 2025
  
  8e85ee88
- Forbid DeepEP racing condition when too many tokens (#9567) · adf73175
  fzyzcjy authored Sep 05, 2025
  
  adf73175