Commits · 1b427dae0269024ec7c7330bc7b5e181b557d342 · zhaoyu6 / sglang

"test/ut/sdk/test_model_speedup.py" did not exist on "80b6cb3b934d3c2965a7412128645e0eefdefaf2"

19 Jul, 2025 13 commits
- 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 15 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
- Hicache Storage Layer Prototype (#7704) · 9d33fcfb
  Zhiqiang Xie authored Jul 18, 2025
  
  9d33fcfb
- [Quantization][w8a8_int8] Fix weight loading issue for w8a8_int8 path with... · 7891bac1
  jianan-gu authored Jul 18, 2025
```
[Quantization][w8a8_int8] Fix weight loading issue for w8a8_int8 path with "ignore" layer list in quantization config (#7820)
```
  7891bac1
- [CPU][Llama4] Fix Llama4 MoE inputs with "apply_router_weight_on_input" (#7889) · 48c1fa7b
  jianan-gu authored Jul 18, 2025
  
  48c1fa7b
- load draft model fix (#7506) · 8aa5ae6b
  yilian49 authored Jul 18, 2025
  
  8aa5ae6b
- Feat: Support Granite 3.0 MoE in SGLang (#7959) · 8a323557
  Minglei Zhu authored Jul 17, 2025
  
  8a323557
- [Fix][Ready]Fix register spilling in cutlass nvfp4 gemm kernel on Blackwell (#8127) · 6e92da8f
  Qi Yuhang authored Jul 18, 2025
  
  6e92da8f
- refactor: simply MultimodalTokens logic (#7924) · e1020dc5
  Mick authored Jul 18, 2025
  
  e1020dc5
17 Jul, 2025 12 commits
- feat: add production metric for retracted requests due to insufficient kvcache (#7030) · 3586b4ce
  Zhao Chen authored Jul 18, 2025
```
Signed-off-by: Zhao Chen <zhaochen.zju@gmail.com>
```
  3586b4ce
- [Hunyuan]: Fix Dense Model Support (#8117) · 42960214
  Asher authored Jul 18, 2025
```
Signed-off-by: Asher Zhang <asherszhang@tencent.com>
```
  42960214
- fix: update HostKVCache init to report correct msg when available memory is not enough (#8102) · 01857fab
  Ziqi Fan authored Jul 17, 2025
  
  01857fab
- Super tiny fix typo (#8046) · 519ff5c8
  fzyzcjy authored Jul 17, 2025
  
  519ff5c8
- [kernel] opt moe align block kernel by block/warp scan algorithm (#7884) · af1cc8fe
  Yuan Luo authored Jul 17, 2025
  
  af1cc8fe
- Refactor: move all quantization-related code to `srt/layer/quantization` (#7989) · 49b87774
  Cheng Wan authored Jul 17, 2025
  
  49b87774
- [ci] recover 8-gpu deepep test (#8105) · 02404a1e
  Cheng Wan authored Jul 17, 2025
  
  02404a1e
- [Fix] ensure DeepGEMM is only enabled for FP8_W8A8 models (#8110) · 5c08a36c
  hzh0425 authored Jul 17, 2025
  
  5c08a36c
- [ci] disable memory imbalance check for draft worker (#8108) · 9069884b
  Cheng Wan authored Jul 16, 2025
  
  9069884b
- [ci] limit cmake build nproc (#8100) · 8a7a7770
  Simo Lin authored Jul 16, 2025
  
  8a7a7770
- feat: add tp_rank, pp_rank and dp_rank labels for scheduler metrics (#7597) · 795668dc
  Yingchun Lai authored Jul 17, 2025
```
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
```
  795668dc
- refactor: unify names of the feature field of MultimodalDataItem (#8075) · 4395c87a
  Mick authored Jul 17, 2025
  
  4395c87a