Commits · b43263307f40a206f1371e4064d410a136d4e004 · change / sglang · GitLab

23 Jul, 2025 1 commit
- Hicache IO kernel refactoring (#8264) · b4326330
  Zhiqiang Xie authored Jul 23, 2025
  
  b4326330
20 Jul, 2025 1 commit
- Add bf16 output option for dsv3_router_gemm kernel (#7999) · 282eb59f
  Baizhou Zhang authored Jul 19, 2025
  
  282eb59f
18 Jul, 2025 2 commits
- feat: enchance green context stream creation robust with backward compatibility (#8136) · 719b29f2
  Peng Zhang authored Jul 18, 2025
  
  719b29f2
- [Fix][Ready]Fix register spilling in cutlass nvfp4 gemm kernel on Blackwell (#8127) · 6e92da8f
  Qi Yuhang authored Jul 18, 2025
  
  6e92da8f
17 Jul, 2025 1 commit
- [kernel] opt moe align block kernel by block/warp scan algorithm (#7884) · af1cc8fe
  Yuan Luo authored Jul 17, 2025
  
  af1cc8fe
16 Jul, 2025 1 commit
- fix greenctx stream compability (#8090) · 6dc4af49
  Peng Zhang authored Jul 16, 2025
  
  6dc4af49
14 Jul, 2025 1 commit
- [Feature] CUDA Green Context Support (#7649) · 1ebec1a8
  ykcombat authored Jul 15, 2025
  
  1ebec1a8
10 Jul, 2025 1 commit
- Fix bug of deepseek-v3 under DP+EP mode with large batchsize/seqlen (#6449) · 4a0d1919
  likesen-alibaba authored Jul 10, 2025
  
  4a0d1919
09 Jul, 2025 2 commits
- [CPU] [FP8] set SGLANG_CPU_FP8_CVT_FTZ in CMakeLists.txt (#7885) · ac80f4da
  Chunyuan WU authored Jul 09, 2025
  
  ac80f4da
- [CPU]convert topk_weights to fp32 for INT8 and FP8 paths (for llama4) and fix... · 128f16a8
  Chunyuan WU authored Jul 09, 2025
```
[CPU]convert topk_weights to fp32 for INT8 and FP8 paths (for llama4) and fix LmHead weight pack (#7818)
```
  128f16a8
07 Jul, 2025 1 commit
- Optimize moe align block size kernel (#7794) · a3398d84
  Ke Bao authored Jul 07, 2025
  
  a3398d84
05 Jul, 2025 4 commits
- Add treemask mode to build_eagle_tree & release sgl-kernel 0.2.3 (#7756) · 5589b750
  Lianmin Zheng authored Jul 05, 2025
```
Co-authored-by: Pranjal Shankhdhar <pranjal.ssh@gmail.com>
```
  5589b750
- fix: fix apply_shuffle_mul_sum (#7444) · c7973222
  Mick authored Jul 05, 2025
  
  c7973222
- Optimize Hopper CUTLASS FP8 Blockwise Grouped GEMM Kernel in Small K Scenario (#7782) · 8e9fb43d
  Qi Yuhang authored Jul 05, 2025
  
  8e9fb43d
- [1/n]: add cutlass W4A8 moe kernel for hopper architecture (#7772) · da3890e8
  SijiaYang authored Jul 05, 2025
```
Signed-off-by: yangsijia.614 <yangsijia.614@bytedance.com>
Co-authored-by: yicwang <yichen.wang@bytedance.com>
```
  da3890e8
03 Jul, 2025 4 commits
- [optimize] fuse renormalize into moe_topk_softmax (#7744) · 2998c4bd
  Yi Zhang authored Jul 04, 2025
```
Co-authored-by: ispobock <ispobaoke@gmail.com>
```
  2998c4bd
- Add CUTLASS FP8 Blockscale MoE kernel for Hopper architecture (#7278) · 2c4feaf3
  ayrnb authored Jul 03, 2025
```
Co-authored-by: HydraQYH <QYH820@Outlook.com>
Co-authored-by: TianQiLin666666 <1834987979@qq.com>
```
  2c4feaf3
- [CPU] [sgl-kernel] set dispatch key of initialize to CatchAll (#7734) · 36cc3ffd
  Chunyuan WU authored Jul 03, 2025
  
  36cc3ffd
- Support non-contiguous query input for extend/decode attention (#7462) · b044400d
  YanbingJiang authored Jul 03, 2025
  
  b044400d
02 Jul, 2025 1 commit

[1/n] apply wna16marlin kernel in moe weight only quantization (#7683) · 8e03b641

AniZpZ authored Jul 02, 2025


Co-authored-by: 晟海 <huangtingwei.htw@antgroup.com>
Co-authored-by: yych0745 <1398089567@qq.com>
Co-authored-by: HandH1998 <1335248067@qq.com>
Co-authored-by: 弋云 <yiyun.wyt@antgroup.com>
Co-authored-by: walker-ai <2398833647@qq.com>

8e03b641

01 Jul, 2025 1 commit
- [CPU] remove process_group from inputs of shm_allreduce and shm_allgather (#7486) · 6005ecee
  Chunyuan WU authored Jul 01, 2025
  
  6005ecee
30 Jun, 2025 2 commits
- Add dsv3 router gemm kernel (#7627) · 7248272c
  Baizhou Zhang authored Jun 29, 2025
  
  7248272c
- [CPU] add c++ kernel to bind CPU cores and memory node (#7524) · c5131f7a
  Chunyuan WU authored Jun 30, 2025
  
  c5131f7a
29 Jun, 2025 1 commit
- Add dsv3 fused a gemm to sgl-kernel (#7630) · 04b35190
  Ke Bao authored Jun 29, 2025
  
  04b35190
25 Jun, 2025 2 commits
- [CPU] [BF16] Call fused_experts_cpu, weight_packed_linear and bmm_cpu kernel... · 7eb47b0f
  Chunyuan WU authored Jun 25, 2025
```
[CPU] [BF16] Call fused_experts_cpu, weight_packed_linear and bmm_cpu kernel in DeepSeek model (#6641)
Co-authored-by: Thien Tran <gau.nernst@yahoo.com.sg>
```
  7eb47b0f
- Fuse sorted_token_ids padding to moe_align_block_size kernel (#7437) · 57ab7769
  Ke Bao authored Jun 25, 2025
  
  57ab7769
23 Jun, 2025 1 commit
- kvcache io kernels and test case (#7382) · 34c3f9b2
  Zhiqiang Xie authored Jun 23, 2025
  
  34c3f9b2
17 Jun, 2025 1 commit
- Fix AWQ Dequant and Weight Loading of deepseek v2 (#6842) · 3eb4a800
  AniZpZ authored Jun 18, 2025
  
  3eb4a800
16 Jun, 2025 1 commit
- Fix sampling for speculative decoding & simplify kernels (#7207) · cfceb83d
  Lianmin Zheng authored Jun 16, 2025
  
  cfceb83d
14 Jun, 2025 1 commit
- [fix] fix cutlass_mla_backend with cuda_graph and add sm_scale for sgl-kernel cutlass_mla (#7184) · ab1a4fa5
  JieXin Liang authored Jun 15, 2025
  
  ab1a4fa5
13 Jun, 2025 2 commits
- Support new DeepGEMM format in per token group quant (#7146) · 5c66c442
  fzyzcjy authored Jun 13, 2025
  
  5c66c442
- Remove 200us slow concat kernel (part 1: kernel) (#7145) · aa46ed34
  fzyzcjy authored Jun 13, 2025
  
  aa46ed34
12 Jun, 2025 1 commit
- [sgl-kernel] Add cuda kernel for moe_ep_silu_and_mul (#6919) · 84727a51
  Yuan Luo authored Jun 12, 2025
```
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
```
  84727a51
10 Jun, 2025 2 commits
- Tiny fix cutlass_mla_get_workspace_size stub incorrect signature (#7057) · 19995dd7
  fzyzcjy authored Jun 11, 2025
  
  19995dd7
- CPU: map changes from developing branch in sgl-kernel (#6833) · fcde67b0
  YanbingJiang authored Jun 10, 2025
```
Co-authored-by: mingfeima <mingfei.ma@intel.com>
```
  fcde67b0
09 Jun, 2025 1 commit
- [perf][sgl-kernel] extend cutlass_mla_decode to support num_head < 128 (#6929) · 18efb5e8
  JieXin Liang authored Jun 09, 2025
  
  18efb5e8
07 Jun, 2025 2 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
- reduce torch.zeros overhead in moe align block size kernel (#6369) · 8b5f83ed
  Xiaoyu Zhang authored Jun 07, 2025
  
  8b5f83ed
05 Jun, 2025 2 commits
- [EP] Add cuda kernel for moe_ep_post_reorder (#6837) · 43baba64
  Yuan Luo authored Jun 05, 2025
```
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
```
  43baba64
- support 1 shot allreduce in 1-node and 2-node using mscclpp (#6277) · 8e3797be
  zyksir authored Jun 05, 2025
  
  8e3797be