Commits · 698bc661a47e68ebf0239e224eb30b905eafd9af · change / sglang

10 Nov, 2025 5 commits
- Merge branch 'v0.5.4_dev_lzg' into 'v0.5.4_dev' · 698bc661
  lizhigong authored Nov 10, 2025
```
fix bug on flash attention use for chunkprefill and radix cache

See merge request OpenDAS/sglang!14
```
  698bc661
- Merge branch 'v0.5.4_rzc' into 'v0.5.4_dev' · 1eaad6d1
  lizhigong authored Nov 10, 2025
```
添加环境变量SGLANG_USE_LIGHTOP 控制 lightop的融合rotaty_emb和moe_gated算子，默认禁用；修复RMSNorm:forward_hip中的错误逻辑

See merge request OpenDAS/sglang!13
```
  1eaad6d1
- fix bug on flash attention use for chunkprefill and radix cache · 8d0b2f15
  lizhigong authored Nov 10, 2025
  
  8d0b2f15
- 添加环境变量SGLANG_USE_LIGHTOP 控制 lightop的融合rotaty_emb和moe_gated算子，默认禁用；修复RMSNorm:forward_hip中的错误逻辑 · f453578b
  renzhc authored Nov 10, 2025
  
  f453578b
- add merge_state_v2 in sgl_kernel · d297cda2
  lizhigong authored Nov 10, 2025
  
  d297cda2
07 Nov, 2025 4 commits
- Merge branch 'v0.5.4_dev_liucong' into 'v0.5.4_dev' · 63c8d8d0
  maxiao1 authored Nov 07, 2025
```
V0.5.4 dev liucong

See merge request OpenDAS/sglang!11
```
  63c8d8d0
- V0.5.4 dev liucong · 5073dd76
  liucong8560 authored Nov 07, 2025
  
  5073dd76
- Merge branch 'v0.5.4_dev_lzg' into 'v0.5.4_dev' · a5156371
  lizhigong authored Nov 07, 2025
```
修复extern报错、精度问题，支持radix cache和chunk prefill

See merge request OpenDAS/sglang!10
```
  a5156371
- 修复extern报错、精度问题，支持radix cache和chunk prefill · abe30c35
  lizhigong authored Nov 07, 2025
  
  abe30c35
04 Nov, 2025 4 commits
- Merge branch 'v0.5.4_dev_linhai' into 'v0.5.4_dev' · 46da9556
  maxiao1 authored Nov 04, 2025
```
V0.5.4 dev linhai

See merge request OpenDAS/sglang!9
```
  46da9556
- V0.5.4 dev linhai · ee775772
  linhai1 authored Nov 04, 2025
  
  ee775772
- Merge branch 'v0.5.4_dev_maxiao' into 'v0.5.4_dev' · a9e0e668
  maxiao1 authored Nov 04, 2025
```
enable custom_allreduce

See merge request OpenDAS/sglang!6
```
  a9e0e668
- enable custom_allreduce · 0c5532b0
  maxiao1 authored Nov 04, 2025
  
  0c5532b0
03 Nov, 2025 5 commits
- Merge branch 'v0.5.4_dev_maxiao' into 'v0.5.4_dev' · 785e5e90
  maxiao1 authored Nov 03, 2025
```
调用vllm里custom all reduce

See merge request OpenDAS/sglang!5
```
  785e5e90
- 调用vllm里custom all reduce · d2fdeac2
  maxiao1 authored Nov 03, 2025
  
  d2fdeac2
- Merge branch 'v0.5.4_dev_qwennext' into 'v0.5.4_dev' · 47e4d923
  maxiao1 authored Nov 03, 2025
```
适配qwen3-next

See merge request OpenDAS/sglang!4
```
  47e4d923
- Merge branch 'v0.5.4_dev_maxiao' into 'v0.5.4_dev' · 0fbecc43
  maxiao1 authored Nov 03, 2025
```
change sgl_kernel WARP_SIZE to 64

See merge request OpenDAS/sglang!3
```
  0fbecc43
- change sgl_kernel WARP_SIZE to 64 · 75cd34d1
  maxiao1 authored Nov 03, 2025
  
  75cd34d1
30 Oct, 2025 1 commit
- 适配qwen3-next · 477fddf2
  maxiao1 authored Oct 30, 2025
  
  477fddf2
29 Oct, 2025 3 commits
- Merge branch 'v0.5.4_dev_maxiao' into 'v0.5.4_dev' · 8fc55263
  maxiao1 authored Oct 29, 2025
```
适配w8a8模型

See merge request OpenDAS/sglang!1
```
  8fc55263
- update UNBALANCED_MODEL_LOADING_TIMEOUT_S=3600 · eb4ba1c2
  maxiao1 authored Oct 29, 2025
  
  eb4ba1c2
- 适配w8a8模型 · 4b9b337b
  maxiao1 authored Oct 29, 2025
  
  4b9b337b
28 Oct, 2025 3 commits
- 增加hipprof支持、修复异步调度中的同步问题 · f6528b74
  lizhigong authored Oct 28, 2025
  
  f6528b74
- 关闭custom_allreduce保持正确性 · a5718531
  maxiao1 authored Oct 28, 2025
  
  a5718531
- 补充 bench_serving.py里tpot等指标 · c333f125
  guobj authored Oct 28, 2025
  
  c333f125
27 Oct, 2025 1 commit
- fix fused_add_rms_norm bug · f9a026ad
  maxiao authored Oct 27, 2025
  
  f9a026ad
25 Oct, 2025 6 commits
- adaptation w4a8 tp · b80ae5e9
  maxiao1 authored Oct 25, 2025
  
  b80ae5e9
- adapt w4a8 marlin deepep dp ep · b091a7a5
  lizhigong authored Oct 22, 2025
```
(cherry picked from commit a0fb70e9)
```
  b091a7a5
- adaptation w4A8 quantization · 143ec5f3
  lizhigong authored Oct 21, 2025
```
(cherry picked from commit 848c5b82)
```
  143ec5f3
- adaptation part w4A8 quantization · 67510e01
  lizhigong authored Oct 21, 2025
```
(cherry picked from commit 68277eac)
```
  67510e01
- 修改sgl-kernel下的setup_hip.py · 32b1ccaf
  maxiao1 authored Oct 25, 2025
  
  32b1ccaf
- 适配v0.5.4 · 251235c2
  maxiao1 authored Oct 25, 2025
  
  251235c2
24 Oct, 2025 2 commits
- chore: bump SGLang version to 0.5.4 (#12027) · 1053e1be
  sglang-bot authored Oct 24, 2025
```
Co-authored-by: sglang-bot <sglang-bot@users.noreply.github.com>
```
  1053e1be
- Fixed aarch64 flash-mla (#12009) · 9a71500c
  nvjullin authored Oct 24, 2025
  
  9a71500c
23 Oct, 2025 6 commits
- [router] Add builder pattern for RouterConfig with zero duplication (#12030) · 6d6e24bc
  Simo Lin authored Oct 23, 2025
  
  6d6e24bc
- Update Github action title for kernel build (#12029) · 2c057fbf
  Kangyan-Zhou authored Oct 23, 2025
  
  2c057fbf
- Fix mamba radix cache eviction logic in `alloc_req_slots` (#11616) · dbd9435d
  Roger Young authored Oct 24, 2025
```
Signed-off-by: rogeryoungh <rogeryoungh@foxmail.com>
```
  dbd9435d
- Revert "[ROCm] Remove vLLM rope dependency & use AITER impl" (#12028) · 8ae9d4bb
  b8zhong authored Oct 23, 2025
  
  8ae9d4bb
- Log iteration # for prefill and decode (#9366) · 1c304aa9
  Nicolas Castet authored Oct 23, 2025
  
  1c304aa9
- model: support deepseek-ocr (#11891) · 770529a7
  Mick authored Oct 24, 2025
```
Co-authored-by: yhyang201 <47235274+yhyang201@users.noreply.github.com>
Co-authored-by: yhyang201 <yhyang201@gmail.com>
Co-authored-by: Shi Shuai <126407087+shuaills@users.noreply.github.com>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
```
  770529a7