Commits · fcde67b016eb6bdc46afbd05e98d962d315b15d1 · zhaoyu6 / sglang

"projects/vscode:/vscode.git/clone" did not exist on "fe25f7a51d36e3702f961e198894580d83c4387b"

10 Jun, 2025 4 commits
- Fix fused_moe triton configs (#7029) · 81372f3b
  yudian0504 authored Jun 10, 2025
  
  81372f3b
- Add decode req pool (#6980) · c2b16795
  Byron Hsu authored Jun 09, 2025
  
  c2b16795
- Support both approximate and exact expert distribution collection (#6964) · f6ebba53
  fzyzcjy authored Jun 10, 2025
  
  f6ebba53
- Update default settings for blackwell (#7023) · 6716b417
  Baizhou Zhang authored Jun 09, 2025
  
  6716b417
09 Jun, 2025 12 commits
- Simplify prepare_extend_after_decode (#6987) · dc0705a5
  Lianmin Zheng authored Jun 09, 2025
  
  dc0705a5
- Fix torchvision version for Blackwell (#7015) · a968c888
  Wenxuan Tan authored Jun 09, 2025
  
  a968c888
- Fallback to lower triton version for unfound fused moe configs (#7013) · a979daac
  Baizhou Zhang authored Jun 09, 2025
  
  a979daac
- feat: add direct routing strategy to DP worker (#6884) · f1569876
  ishandhanani authored Jun 09, 2025
  
  f1569876
- Fix cutlass MLA gets almost zero accuracy (#6998) · e58423b2
  fzyzcjy authored Jun 10, 2025
  
  e58423b2
- chore: upgrade flashinfer v0.2.6.post1 jit (#6958) · 56ccd3c2
  Yineng Zhang authored Jun 09, 2025
```
Co-authored-by: alcanderian <alcanderian@gmail.com>
Co-authored-by: Qiaolin Yu <qy254@cornell.edu>
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
Co-authored-by: Mick <mickjagger19@icloud.com>
Co-authored-by: ispobock <ispobaoke@gmail.com>
```
  56ccd3c2
- Fix torch profiler bugs for bench_offline_throughput.py (#6557) · 98c00a2d
  Yueyang Pan authored Jun 09, 2025
  
  98c00a2d
- support qwen3 emebedding (#6990) · 451ffe74
  uylnap authored Jun 09, 2025
  
  451ffe74
- Eliminate stream sync to speed up LoRA batch init (#6960) · b1e5a33a
  Lifu Huang authored Jun 09, 2025
  
  b1e5a33a
- Use torch.compile to fuse flash attention decode metadata preparation (#6973) · 9d5fa68b
  Lianmin Zheng authored Jun 08, 2025
  
  9d5fa68b
- Minor remove one kernel for DeepSeek (#6977) · de1350ea
  fzyzcjy authored Jun 09, 2025
  
  de1350ea
- Fix expert distribution dumping causes OOM (#6967) · 86fe943b
  fzyzcjy authored Jun 09, 2025
  
  86fe943b
08 Jun, 2025 10 commits
- Sync cuda graph runners (#6976) · 0c1f03a2
  Lianmin Zheng authored Jun 08, 2025
  
  0c1f03a2
- Fuse routed scaling factor in deepseek (#6970) · 3712abfa
  Xiaoyu Zhang authored Jun 09, 2025
  
  3712abfa
- Extend cuda graph capture bs for B200 (#6937) · 971a0dfa
  Baizhou Zhang authored Jun 08, 2025
  
  971a0dfa
- Remove unnecessary kernels of num_token_non_padded (#6965) · 2fc12995
  fzyzcjy authored Jun 08, 2025
  
  2fc12995
- Fix CI and triton moe Configs (#6974) · 20d3ad3b
  Lianmin Zheng authored Jun 08, 2025
  
  20d3ad3b
- rebase h20 fused_moe config (#6966) · fa3592cf
  Xiaoyu Zhang authored Jun 08, 2025
  
  fa3592cf
- Slightly improve the sampler to skip unnecessary steps (#6956) · 608668e1
  Lianmin Zheng authored Jun 08, 2025
  
  608668e1
- Revert "Fuse routed scaling factor in topk_reduce kernel (#6220)" (#6968) · 1fb76ebb
  Yineng Zhang authored Jun 07, 2025
  
  1fb76ebb
- [DeepseekR1-FP4] Add Support for nvidia/DeepSeekR1-FP4 model (#6853) · c2c4f57f
  Pavani Majety authored Jun 07, 2025
```
Signed-off-by: Pavani Majety <pmajety@nvidia.com>
```
  c2c4f57f
- chore: upgrade sgl-kernel v0.1.6.post1 (#6957) · 23881fa6
  Yineng Zhang authored Jun 07, 2025
  
  23881fa6
07 Jun, 2025 14 commits
- Add a CUDA kernel for fusing mapping and weighted sum for MoE. (#6916) · 3e56f557
  Elfie Guo authored Jun 07, 2025
```
Co-authored-by: Elfie Guo <elfiegxf@gmail.com>
```
  3e56f557
- Add H20 fused MoE kernel tuning configs for DeepSeek-R1/V3 (#6885) · 62fec60d
  Xu Wenqing authored Jun 08, 2025
```
Signed-off-by: Xu Wenqing <xuwq1993@qq.com>
```
  62fec60d
- [misc] add is_cpu() (#6950) · e7759778
  JieXin Liang authored Jun 08, 2025
  
  e7759778
- Update server timeout time in AMD CI. (#6953) · 77e928d0
  Sai Enduri authored Jun 07, 2025
  
  77e928d0
- Fuse routed scaling factor in topk_reduce kernel (#6220) · 515ef4fa
  Xiaoyu Zhang authored Jun 08, 2025
  
  515ef4fa
- Refactor global_server_args_dict (#6866) · f5599ef1
  fzyzcjy authored Jun 07, 2025
  
  f5599ef1
- Add canary for EPLB rebalancing (#6895) · c499591a
  fzyzcjy authored Jun 07, 2025
  
  c499591a
- Disabling mixed chunked prefill when eagle is enabled (#6874) · e1ce44cd
  Swipe4057 authored Jun 07, 2025
  
  e1ce44cd
- chore: upgrade sgl-kernel v0.1.6 (#6945) · 6153f2ff
  JieXin Liang authored Jun 07, 2025
  
  6153f2ff
- reduce torch.zeros overhead in moe align block size kernel (#6369) · 8b5f83ed
  Xiaoyu Zhang authored Jun 07, 2025
  
  8b5f83ed
- Add triton version as a fused_moe_triton config search key to avoid performace... · 2a413829
  Xiaoyu Zhang authored Jun 07, 2025
```
Add triton version as a fused_moe_triton config search key to avoid performace decrease in different Triton version (#5955)
```
  2a413829
- Tiny re-introduce profile id logging (#6912) · d5c097a2
  fzyzcjy authored Jun 07, 2025
  
  d5c097a2
- [Bugfix] pipeline parallelism and Eagle Qwen2 (#6910) · 9736cd3b
  Swipe4057 authored Jun 07, 2025
  
  9736cd3b
- Minor compile fused topk (#6944) · 2f715f51
  fzyzcjy authored Jun 07, 2025
  
  2f715f51