Commits · bbcfbc1a024980e76926c114ca2daf8cec9098a2 · change / sglang · GitLab

20 Jul, 2025 12 commits
- feat: add h200 tp 16 kimi k2 moe config (#8183) · bbcfbc1a
  Qiaolin Yu authored Jul 20, 2025
  
  bbcfbc1a
- Feat: Support for Persimmon Model (#7983) · 83c104b1
  Praneth Paruchuri authored Jul 20, 2025
  
  83c104b1
- feat: update nccl 2.27.6 (#8182) · 2db6719c
  Yineng Zhang authored Jul 19, 2025
  
  2db6719c
- Revert "[Feature] Simple Improve Health Check Mechanism for Production-Grade Stability" (#8181) · 55381a46
  Lianmin Zheng authored Jul 19, 2025
  
  55381a46
- fix moe gate dtype, fix tbo, fix fake dispatch (#7825) · a589a071
  Atream authored Jul 20, 2025
  
  a589a071
- feat: add b200 tp 16 kimi k2 moe config (#8178) · f62d75b6
  Yineng Zhang authored Jul 19, 2025
  
  f62d75b6
- feat: add h200 tp 16 kimi k2 moe config (#8176) · 0f9b11e3
  Yineng Zhang authored Jul 19, 2025
  
  0f9b11e3
- Add get_hidden_dim to qwen3.py for correct lora (#7312) · 877e35d7
  Pavel Logachev authored Jul 20, 2025
  
  877e35d7
- Enable FlashInfer support encoder models and add head_dim padding workaround (#6230) · cbdfb771
  Clay authored Jul 20, 2025
  
  cbdfb771
- Add bf16 output option for dsv3_router_gemm kernel (#7999) · 282eb59f
  Baizhou Zhang authored Jul 19, 2025
  
  282eb59f
- [Feature] Simple Improve Health Check Mechanism for Production-Grade Stability (#8115) · 4540a466
  ybyang authored Jul 20, 2025
```
Signed-off-by: ybyang <ybyang7@iflytek.com>
```
  4540a466
- Fix tuning_fused_moe_triton.py (#8175) · abda2542
  Cheng Wan authored Jul 19, 2025
  
  abda2542
19 Jul, 2025 20 commits
- Clean warning logs for gate_proj loading in Lora (#8172) · 8cddfa56
  Baizhou Zhang authored Jul 19, 2025
  
  8cddfa56
- Support start up LoRA server without initial adapters (#8019) · 4e3defe5
  Lifu Huang authored Jul 19, 2025
  
  4e3defe5
- bugfix: fix sglang crash in NVIDIA MIG container (#8167) · 60468da4
  Garry Fang authored Jul 20, 2025
```
Signed-off-by: Garrybest <garrybest@foxmail.com>
```
  60468da4
- [router] add ut for worker and errors (#8170) · 41d33e47
  Simo Lin authored Jul 19, 2025
  
  41d33e47
- Fix Dockerfile.gb200 (#8169) · bfdd226f
  kyleliang-nv authored Jul 19, 2025
  
  bfdd226f
- Fix LoRA buffer contamination during adapter eviction (#8103) · 3de617a7
  Lifu Huang authored Jul 19, 2025
  
  3de617a7
- Clean up server args (#8161) · bb0e8a32
  Lianmin Zheng authored Jul 19, 2025
  
  bb0e8a32
- Update README.md (#8171) · 1b427dae
  Lianmin Zheng authored Jul 19, 2025
  
  1b427dae
- Fix suffix mismatch for the metrics. (#8168) · f3d97361
  Charles Chen authored Jul 20, 2025
```
Signed-off-by: Charles Chen <chenliqian@chenliqian.cn>
```
  f3d97361
- chore: upgrade sgl-kernel 0.2.6 (#8166) · 561dd7b2
  Yineng Zhang authored Jul 19, 2025
  
  561dd7b2
- chore: bump sgl-kernel v0.2.6 (#8165) · f98e88b9
  Yineng Zhang authored Jul 19, 2025
  
  f98e88b9
- [1/N] MoE Refactor: refactor `select_experts` (#7966) · 15ad6c90
  Cheng Wan authored Jul 19, 2025
  
  15ad6c90
- Add GB200 wide-EP docker (#8157) · cfab0ff6
  kyleliang-nv authored Jul 18, 2025
  
  cfab0ff6
- [router] allow router to have empty workers (#8160) · b763cf7e
  Simo Lin authored Jul 18, 2025
  
  b763cf7e
- [router] router metrics cleanup (#8158) · 8fcc55cf
  Simo Lin authored Jul 18, 2025
  
  8fcc55cf
- [health_generate] fix: fix the /health_generate always success bug (#8028) · 610381b7
  Yingchun Lai authored Jul 19, 2025
  
  610381b7
- [PD] Support non-MLA models PD different TP with DP attention (#7931) · 1403ea56
  Shangming Cai authored Jul 19, 2025
```
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
```
  1403ea56
- Feat: Support audio in Phi4-mm model (#8048) · b7e951a6
  Binyao Jiang authored Jul 18, 2025
  
  b7e951a6
- Support NVFP4 quantized dense models on AMD CDNA2/CDNA3 GPUs (#7302) · d918ab79
  Haohui Mai authored Jul 18, 2025
```
Co-authored-by: HAI <hixiao@gmail.com>
Co-authored-by: Sai Enduri <saimanas.enduri@amd.com>
```
  d918ab79
- chore: tune mem fraction static for vlm (#6881) · 3964b352
  Mick authored Jul 19, 2025
  
  3964b352
18 Jul, 2025 8 commits
- [Doc] Steps to add a new attention backend (#8155) · 9c7a4618
  Lianmin Zheng authored Jul 18, 2025
  
  9c7a4618
- [AMD] Add triton awq_dequantize kernel to support AWQ on ROCm (#7661) · 7750b91c
  Hubert Lu authored Jul 18, 2025
  
  7750b91c
- [router] Refactor router and policy traits with dependency injection (#7987) · c8f31042
  Simo Lin authored Jul 18, 2025
```
Co-authored-by: Jin Pan <jpan236@wisc.edu>
Co-authored-by: Keru Yang <rukeyang@gmail.com>
Co-authored-by: Yingyi Huang <yingyihuang2000@outlook.com>
Co-authored-by: Philip Zhu <phlipzhux@gmail.com>
```
  c8f31042
- [3/n] chore: decouple AWQ implementation from vLLM dependency (#8113) · 1f76fc87
  Hongbo Xu authored Jul 19, 2025
```
Co-authored-by: AniZpZ <zhuangsen.zp@antgroup.com>
```
  1f76fc87
- [Bugfix] Fix w8a8_int8 import error on NPU (#8147) · 6737671c
  Even Zhou authored Jul 19, 2025
  
  6737671c
- fix compressed tensors WNA16 imports (#8142) · fd63b62e
  Enrique Shockwave authored Jul 18, 2025
  
  fd63b62e
- feat: enchance green context stream creation robust with backward compatibility (#8136) · 719b29f2
  Peng Zhang authored Jul 18, 2025
  
  719b29f2
- Revert "Fix different device type adjustment in PP" (#8141) · d0510f08
  Sai Enduri authored Jul 18, 2025
  
  d0510f08