Commits · 3dd6420a4dc0234d4582cdda7e0252a2ce6d4722 · change / sglang · GitLab

11 Sep, 2025 10 commits
- [CI] add pyproject.toml to deepseek w4a8 ci (#10314) · 3dd6420a
  Hank Han authored Sep 11, 2025
  
  3dd6420a
- chore: bump sgl-kernel 0.3.9.post2 (#10311) · 532f998b
  Yineng Zhang authored Sep 11, 2025
  
  532f998b
- Revert "Fix flashinfer version in sgl-kernel (#10135)" (#10310) · de15d140
  Yineng Zhang authored Sep 11, 2025
  
  de15d140
- [fix CI] Fix logical condition in fused MoE layer for compressed tensor quantization (#10299) · 37367da6
  Xiaoyu Zhang authored Sep 11, 2025
  
  37367da6
- [CPU] fix OOM when mem-fraction is not set (#9090) · ef959d7b
  Zaili Wang authored Sep 11, 2025
  
  ef959d7b
- [chore]Add sgl-router to npu images (#10229) · 4aa1e69b
  BourneSun0527 authored Sep 11, 2025
  
  4aa1e69b
- add flash linear attention triton kernel (#10239) · dc491b39
  Yi Zhang authored Sep 11, 2025
  
  dc491b39
- [Feature] Support DeepEP normal & Redundant Experts on NPU (#9881) · 5b64f006
  Even Zhou authored Sep 11, 2025
  
  5b64f006
- chore: bump sgl-kernel 0.3.9.post1 (#10294) · 5b7448de
  Yineng Zhang authored Sep 10, 2025
  
  5b7448de
- Revert "[1/2] Optimizations and refactors about quant kernel (#9534)" (#10292) · 6d55f60e
  Yineng Zhang authored Sep 10, 2025
  
  6d55f60e
10 Sep, 2025 21 commits
- [Auto Sync] Update serving_base.py, serving_chat.py, servin... (20250910) (#10282) · 033b75f5
  Lianmin Zheng authored Sep 10, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: cctry <shiyang@x.ai>
```
  033b75f5
- Feat: support disable tool parser (#10184) · f3b5db6e
  Xinyuan Tong authored Sep 10, 2025
  
  f3b5db6e
- pass a_scale from fp8 quant result instead of hard code to 1.0f (#10241) · 2286e85e
  Rain Jiang authored Sep 10, 2025
```
Co-authored-by: Yichen Wang <yichen.wang@bytedance.com>
Co-authored-by: Jinwu Guo <641876696@qq.com>
```
  2286e85e
- Add tests to AMD CI for MI35x (#9662) · 91b3555d
  Hubert Lu authored Sep 10, 2025
```
Co-authored-by: Sai Enduri <saimanas.enduri@amd.com>
```
  91b3555d
- add dual stream for qwen2_moe (#10252) · 9e2f7252
  Yi Zhang authored Sep 11, 2025
  
  9e2f7252
- [Bugfix] Fix Weightloading for the original nvidia/Deepseek-R1-FP4 checkpoint (#9940) · 21176b00
  Pavani Majety authored Sep 10, 2025
```
Signed-off-by: Pavani Majety <pmajety@nvidia.com>
Co-authored-by: Yineng Zhang <me@zhyncs.com>
Co-authored-by: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
```
  21176b00
- [1/2] Refactor LoRA to support backend-specific batch preprocessing. (#10251) · 94100294
  Lifu Huang authored Sep 10, 2025
  
  94100294
- [router] Add PD router mmlu test (#10256) · cda7e47c
  Keyang Ru authored Sep 10, 2025
  
  cda7e47c
- Fix potential flakiness in test_lora_qwen3 (#10250) · e903f695
  Lifu Huang authored Sep 10, 2025
  
  e903f695
- [Auto Sync] Update io_struct.py (20250910) (#10262) · 27760fc1
  Lianmin Zheng authored Sep 10, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Kan Wu <wukanustc@gmail.com>
```
  27760fc1
- Fix assertion typo in tp_worker.py (#9954) · 0ac809de
  Seunggeun Cho authored Sep 10, 2025
  
  0ac809de
- support vlm model spec bench (#10173) · 4efe2c57
  Lzhang-hub authored Sep 10, 2025
  
  4efe2c57
- Page first direct IO kernel (#10060) · 5be8c2f7
  huangtingwei authored Sep 10, 2025
```
Co-authored-by: Zhiqiang Xie <xiezhq@stanford.edu>
```
  5be8c2f7
- Fix: the default choice is wrong for flashinfer mxfp4 moe precision (#10253) · 737d73ed
  Yiyu Liu authored Sep 10, 2025
  
  737d73ed
- [doc] add walkthrough for implementing and hosting a simple llama wrapper m… (#10093) · ebd0e1c1
  Glen Liu authored Sep 10, 2025
  
  ebd0e1c1
- [Benchmark] Prefil-only benchmark scripts (#10240) · a1d03892
  Sundara Raman Ramachandran authored Sep 09, 2025
  
  a1d03892
- [UT for RL] Add UT to cover release/resume memory case for moe model (#8803) · dccf52f9
  ryang authored Sep 10, 2025
  
  dccf52f9
- make --speculative-draft-model an alias of --speculative-draft-model-path (#10246) · 676a7b51
  Lianmin Zheng authored Sep 09, 2025
  
  676a7b51
- refactor(InternVL): Use gpu to preprocess the input image (#9795) · 15f99347
  Kevin Tuan authored Sep 10, 2025
  
  15f99347
- Revert "chore: upgrade v0.3.9 sgl-kernel" (#10245) · bcf1955f
  Lianmin Zheng authored Sep 09, 2025
  
  bcf1955f
- [Auto Sync] Update collector.py, startup_func_log_and_timer... (20250910) (#10242) · a06bf664
  Lianmin Zheng authored Sep 09, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: cctry <shiyang@x.ai>
```
  a06bf664
09 Sep, 2025 9 commits
- [Auto Sync] Update io_struct.py (20250909) (#10236) · bf72b801
  Lianmin Zheng authored Sep 09, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: cctry <shiyang@x.ai>
```
  bf72b801
- Add mamba kernel (#10234) · 8cbe1538
  Yi Zhang authored Sep 10, 2025
  
  8cbe1538
- [HiCache] feat: add mooncake backend extra config (#10213) · 8471e5e6
  Teng Ma authored Sep 10, 2025
  
  8471e5e6
- Revert "Revert the changes on NCCL symmetric memory" (#10238) · 4582931a
  Lianmin Zheng authored Sep 09, 2025
  
  4582931a
- Revert the changes on NCCL symmetric memory (#10210) · d352c29a
  Lianmin Zheng authored Sep 09, 2025
```
Co-authored-by: Yineng Zhang <me@zhyncs.com>
```
  d352c29a
- chore: upgrade v0.3.9 sgl-kernel (#10220) · d3ee7098
  Yineng Zhang authored Sep 09, 2025
  
  d3ee7098
- [Fix] KV-cache eviction mismatch across PP ranks in DeepSeek V3/R1 (#10214) · 71fc7b7f
  Rain H authored Sep 09, 2025
  
  71fc7b7f
- add variable TP Decode > Prefill size support (#9960) · 9ab72f98
  shaharmor98 authored Sep 09, 2025
```
Signed-off-by: Shahar Mor <smor@nvidia.com>
```
  9ab72f98
- chore: bump v0.3.9 sgl-kernel (#10208) · f3817cb0
  Yineng Zhang authored Sep 09, 2025
  
  f3817cb0