Commits · a02071a12cb29d91ac7bb376e0a0744cad3cbb69 · change / sglang · GitLab

08 Sep, 2025 30 commits
- [Bench] feat: mooncake trace integration (#9839) · a02071a1
  Teng Ma authored Sep 09, 2025
```
Signed-off-by: Xuchun Shang <xuchun.shang@linux.alibaba.com>
Signed-off-by: Teng Ma <sima.mt@alibaba-inc.com>
Co-authored-by: Xuchun Shang <xuchun.shang@linux.alibaba.com>
```
  a02071a1
- Revert "[ModelOpt] Fix Weight Loading for DSR1-FP4 Quantization (#9712)" (#10176) · 45b3a6a2
  Yineng Zhang authored Sep 08, 2025
  
  45b3a6a2
- [fix] Relax white space rules in EBNFComposer (#9595) · 9a18aa54
  LukasBluebaum authored Sep 08, 2025
  
  9a18aa54
- pref: Add H20 fp8 fused MoE kernel configs for Qwen3 (#10166) · 91f0fd95
  Zhiy-Zhang authored Sep 09, 2025
```
Co-authored-by: qiufan.zzy <qiufan.zzy@antgroup.com>
```
  91f0fd95
- [Bug fix] Fix ascend mla in aclgraph (#9925) · 8085aca7
  alanhe151220037 authored Sep 09, 2025
  
  8085aca7
- [1/2] Speed up prefill mla attention (#10156) · 0096798e
  fzyzcjy authored Sep 09, 2025
  
  0096798e
- [CI] fix ambiguous argument in testing hybrid attentions. (#10161) · 2c2b19b1
  Liangsheng Yin authored Sep 08, 2025
  
  2c2b19b1
- Monkey patch uvicorn multi worker `is_alive` timeout (#10159) · 72f9fc5f
  Liangsheng Yin authored Sep 08, 2025
```
Co-authored-by: Huang Long <121648372+llll114@users.noreply.github.com>
```
  72f9fc5f
- [Hicache]: Add E2E CI For 3FS-KVStore (#10131) · ec99668a
  hzh0425 authored Sep 08, 2025
  
  ec99668a
- [1/N] DP-Refactor: move communicators into `tokenizer_communicator_mixin` (#10028) · 78f13981
  Liangsheng Yin authored Sep 08, 2025
  
  78f13981
- update xgrammar 0.1.24 and transformers 4.56.1 (#10155) · bfd7a18d
  Swipe4057 authored Sep 08, 2025
  
  bfd7a18d
- [Bug fix] Fix Gemma 2 and fix Gemma 3 multimodal with bs > 1 on NPU (#9871) · 5dd8c644
  ssshinigami authored Sep 08, 2025
```
Co-authored-by: Maksim <makcum888e@mail.ru>
```
  5dd8c644
- enable llama3.1-8B on xpu (#9434) · ee21817c
  Huaiyu, Zheng authored Sep 08, 2025
  
  ee21817c
- Revert "enable auto-round quantization model (#6226)" (#10148) · b7d1f17b
  Yineng Zhang authored Sep 07, 2025
  
  b7d1f17b
- enable auto-round quantization model (#6226) · c8295d23
  Weiwei authored Sep 08, 2025
```
Signed-off-by: Zhang, Weiwei1 <weiwei1.zhang@intel.com>
```
  c8295d23
- [Bugfix] Qwen3MoE aclrtMemcpy failed with NPUGraph (#10013) · b67c277f
  Even Zhou authored Sep 08, 2025
  
  b67c277f
- Fix: (glm4v) Add missing field (#10147) · 8116804e
  Xinyuan Tong authored Sep 08, 2025
  
  8116804e
- Add speculator attention backend switch (#9981) · 8c5930f0
  cicirori authored Sep 08, 2025
  
  8c5930f0
- [Bugfix] Retract not releasing enough memory when page size > 1 (#9989) · 3b99f23c
  Zhiqiang Xie authored Sep 07, 2025
  
  3b99f23c
- [1/N][Bug] Fix w4afp8 MoE NaN issue (sgl-kernel, fixed) (#10108) · ee0b3c5b
  Yuhao Yao authored Sep 08, 2025
  
  ee0b3c5b
- move compile threads to an option to avoid OOM on low memory host (#10123) · 6049ca20
  Rain Jiang authored Sep 07, 2025
  
  6049ca20
- Add graph runner support with torch compile on CPU (#7843) · 7577f0e4
  Cao E authored Sep 08, 2025
  
  7577f0e4
- Standalone speculative decoding (#10090) · 8cda5a62
  Qiaolin Yu authored Sep 07, 2025
  
  8cda5a62
- Fix run time error in dsv3-fp8 model on mi35x (#10104) · 400d3b97
  kk authored Sep 08, 2025
```
Co-authored-by: wunhuang <wunhuang@amd.com>
Co-authored-by: HaiShaw <hixiao@gmail.com>
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>
```
  400d3b97
- Qwen2.5-VL eagle3 infer (#8801) · 37d83c6e
  Lzhang-hub authored Sep 08, 2025
  
  37d83c6e
- fix the fp8 topk_config.correction_bias is none bug (#10040) · 7802586c
  Rain Jiang authored Sep 07, 2025
  
  7802586c
- Fix slow fused add RMSNorm (#10141) · bc5fc332
  fzyzcjy authored Sep 08, 2025
  
  bc5fc332
- vlm: enable GLM4.1V server testing & fix video processing (#10095) · f3440adc
  Xinyuan Tong authored Sep 08, 2025
```
Signed-off-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
Co-authored-by: Binyao Jiang <byjiang1996@gmail.com>
```
  f3440adc
- [MoE] fix: incorrect weight initialization for cutlass_fused_experts_fp8 (#10144) · 5a7e10fe
  Cheng Wan authored Sep 07, 2025
  
  5a7e10fe
- [BUG FIX] add fail check when get fail in case wait complete block (#9971) · 33467c05
  Shisong Ma authored Sep 08, 2025
```
Co-authored-by: mashisong <mashisong@bytedance.com>
Co-authored-by: Zhiqiang Xie <xiezhq@stanford.edu>
```
  33467c05
07 Sep, 2025 10 commits
- [DOC]: some minor updates (#10134) · b0fcbb74
  eigen authored Sep 07, 2025
  
  b0fcbb74
- Fix flashinfer version in sgl-kernel (#10135) · 76a2c86b
  Lianmin Zheng authored Sep 07, 2025
  
  76a2c86b
- [1/2] Refactor multi-tokenizer manager (#10074) · e719bb0e
  Liangsheng Yin authored Sep 07, 2025
  
  e719bb0e
- [Minor] fix lint in main (#10128) · 06724683
  DarkSharpness authored Sep 07, 2025
  
  06724683
- [Auto Sync] Update parallel_state.py (20250907) (#10126) · 617aa2b2
  Lianmin Zheng authored Sep 07, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: jzhou-xai <jzhou@x.ai>
```
  617aa2b2
- add dataset_path for bench_one_batch_server.py (#10113) · 111b1379
  miter authored Sep 07, 2025
```
Signed-off-by: linhuang <linhuang@ruijie.com.cn>
Co-authored-by: linhuang <linhuang@ruijie.com.cn>
```
  111b1379
- [HiCache] fix: check clear() method for storage backend (#10096) · 41628dc1
  Teng Ma authored Sep 07, 2025
```
Co-authored-by: hzh0425 <hzh0425@apache.org>
```
  41628dc1
- Fix cuda graph mode in flashinfer attn backend (#10056) · a12061df
  Ben Barsdell authored Sep 07, 2025
  
  a12061df
- Optimize nvfp4 block scaled gemm kernel when M is small. (#10101) · 85ed8e0a
  Qi Yuhang authored Sep 07, 2025
  
  85ed8e0a
- CUTLASS fp8 blockwise gemm support of sm120 (#9969) · dd1e2689
  Jianying authored Sep 07, 2025
  
  dd1e2689