Commits · 60bcbf2a35e290a351386f4dbe85f5b0534acb14 · change / sglang

"vscode:/vscode.git/clone" did not exist on "0e0eef00ce465256d3839018371bd98b9a82df3e"

11 Apr, 2025 18 commits
- remove moe_align_block_size torch.zeros in small batch/expert mode (#5298) · 60bcbf2a
  Xiaoyu Zhang authored Apr 12, 2025
  
  60bcbf2a
- [Docs] Remove the older supported docs section (#5301) · a0a9f6d6
  Adarsh Shirawalmath authored Apr 12, 2025
  
  a0a9f6d6
- fix: update update_wheel_index for cu128 (#5300) · 80aa8ca8
  Yineng Zhang authored Apr 11, 2025
  
  80aa8ca8
- [Docs] Supported Model Docs - Major restructuring (#5290) · 4aa6bab0
  Adarsh Shirawalmath authored Apr 11, 2025
```
Co-authored-by: zhaochenyang20 <zhaochen20@outlook.com>
```
  4aa6bab0
- [PD] fix: skip warmup request in disaggregation mode to prevent crash on timeout (#5292) · c35dcfdb
  Yusong Gao authored Apr 11, 2025
  
  c35dcfdb
- chore: bump sgl-kernel v0.0.8.post1 (#5289) · c163bf4f
  Yineng Zhang authored Apr 11, 2025
  
  c163bf4f
- chore: relax the torch version restriction for sgl-kernel compilation (#5288) · 55986343
  Yineng Zhang authored Apr 11, 2025
  
  55986343
- feat: add cu128 identifier for sgl-kernel (#5287) · b75275b6
  Yineng Zhang authored Apr 11, 2025
  
  b75275b6
- fix: enable fp4 compilation on cu128 (#5286) · 7074e9ca
  Yineng Zhang authored Apr 11, 2025
  
  7074e9ca
- Fix a 404 link in send_request.ipynb (#5280) · fc14cca0
  Michael Yao authored Apr 11, 2025
```
Signed-off-by: windsonsea <haifeng.yao@daocloud.io>
```
  fc14cca0
- fix: examples for token_in_token_out_vlm (#5193) · e7beff8a
  XinyuanTong authored Apr 11, 2025
  
  e7beff8a
- doc: nested loop code for offline engine (#5244) · 4d2e3051
  mlmz authored Apr 11, 2025
  
  4d2e3051
- [fix] fix mrope positions not picked up (#5265) · e53a0b3d
  Mick authored Apr 11, 2025
  
  e53a0b3d
- Support `--enable-llama4-multimodal` (#5254) · 038bc5d5
  Cheng Wan authored Apr 11, 2025
  
  038bc5d5
- Optimize GPU memory usage in FlashAttentionBackend's strided indexing (#5262) · aee62d74
  Chang Su authored Apr 11, 2025
```
Co-authored-by: ch-wan <cwan39@gatech.edu>
```
  aee62d74
- Optimize attention in llama4 (#5127) · cd7e32e2
  fzyzcjy authored Apr 11, 2025
  
  cd7e32e2
- ROCm/AITER CK_MoE: update 2-stage kernels & support both Activations (#5228) · 88799448
  HAI authored Apr 10, 2025
  
  88799448
- Fix torch.compile cacheing (#5259) · a879811c
  Richard Zou authored Apr 10, 2025
```
Co-authored-by: zhyncs <me@zhyncs.com>
```
  a879811c
10 Apr, 2025 10 commits
- Update Makefile / build script to avoid installing incompatible torch dependency (#5245) · a222945d
  Elfie Guo authored Apr 10, 2025
  
  a222945d
- [Misc] Clean sgl-kernel test (#5216) · ed01b451
  PGFLMG authored Apr 11, 2025
  
  ed01b451
- ROCm sgl-kernel: compatible to later torch (#5167) · d050df36
  HAI authored Apr 10, 2025
  
  d050df36
- Fix deepseek-v3 with torch.compile in PyTorch 2.6. (#5213) · 76f44c2a
  Richard Zou authored Apr 10, 2025
  
  76f44c2a
- Update deps for mllama4 (#5215) · 1078396f
  Ke Bao authored Apr 11, 2025
  
  1078396f
- [PD] Add get_contiguous_buf_infos interface for MLATokenToKVPool (#5204) · 7e4f72dd
  Teng Ma authored Apr 10, 2025
  
  7e4f72dd
- [PD] Support KV transfer with mooncake (#4880) · 4c31ae9f
  Teng Ma authored Apr 10, 2025
```
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
Co-authored-by: Shangming Cai <caishangming@linux.alibaba.com>
Co-authored-by: Xuchun Shang <xuchun.shang@linux.alibaba.com>
Co-authored-by: shangmingc <csmthu@gmail.com>
```
  4c31ae9f
- reduce moe_align_block_size_kernel small batch mode overhead (#5086) · f730362e
  Xiaoyu Zhang authored Apr 10, 2025
  
  f730362e
- Fix DeepSeek error when using DeepEP mode (#5190) · e3c4bd31
  fzyzcjy authored Apr 10, 2025
  
  e3c4bd31
- [metrics] Add in queue metrics (#4444) · 5db37c86
  Stefan He authored Apr 09, 2025
  
  5db37c86
09 Apr, 2025 12 commits
- fix: log warning when disable cuda graph (#5209) · 4cb53ecd
  Yineng Zhang authored Apr 09, 2025
  
  4cb53ecd
- Add H20 dtype fp8_w8a8 fused MoE kernel tuning configs for DeepSeek V3/R1 (#5196) · 456b008b
  Zhaoyang Hao authored Apr 10, 2025
  
  456b008b
- sgl-kernel use cutlass latest version for fp8 blockwise gemm (#5207) · ebf495f0
  Yi Zhang authored Apr 10, 2025
  
  ebf495f0
- update grok test (#5171) · 7f875f12
  saienduri authored Apr 09, 2025
  
  7f875f12
- model: support mllama4 (#5144) · fbebcb7a
  Mick authored Apr 10, 2025
  
  fbebcb7a
- [ci] fix ci test fused_moe op (#5102) · 87eddedf
  Xiaoyu Zhang authored Apr 09, 2025
  
  87eddedf
- Support Llama4 fp8 inference (#5194) · 40652482
  HandH1998 authored Apr 09, 2025
```
Co-authored-by: laixinn <xielx@shanghaitech.edu.cn>
Co-authored-by: sleepcoo <sleepcoo@gmail.com>
Co-authored-by: zhyncs <me@zhyncs.com>
```
  40652482
- Optimize topk operation in llama4 (#5128) · 86a876d8
  fzyzcjy authored Apr 09, 2025
  
  86a876d8
- Fix ci test "test_eval_fp8_accuracy" failed (#5185) · 92823069
  kk authored Apr 09, 2025
```
Co-authored-by: wunhuang <wunhuang@amd.com>
```
  92823069
- [Misc] clean up vllm in sgl-kernel test (#5189) · d2e507df
  yinfan98 authored Apr 09, 2025
  
  d2e507df
- Let `bench_one_batch` support `enable_dp_attention` (#4058) · 61970b08
  fzyzcjy authored Apr 09, 2025
  
  61970b08
- [DeepEP] fix: import buffer error (#5179) · 76c48a09
  Cheng Wan authored Apr 09, 2025
  
  76c48a09