Commits · 94371dbbd6359139cb0a90912838a47a27f61bcc · change / sglang

17 Aug, 2025 1 commit

[feature] Ascend NPU graph support (#8027) · 94371dbb

VDV1985 authored Aug 17, 2025


Co-authored-by: ronnie_zheng <zl19940307@163.com>
Co-authored-by: yezhifeng (D) <y00897525@china.huawei.com>
Co-authored-by: anon189Ty <Stari_Falcon@outlook.com>
Co-authored-by: Maksim <makcum888e@mail.ru>
Co-authored-by: ssshinigami <44640852+ssshinigami@users.noreply.github.com>

94371dbb

16 Aug, 2025 3 commits
- [CI] add deepseek w4a8 test on h20 ci (#7758) · 81da16f6
  Hank Han authored Aug 16, 2025
  
  81da16f6
- Fix DP load for embedding (#9165) · bc938ea1
  Brayden Zhong authored Aug 16, 2025
  
  bc938ea1
- Add fp4 quantize before all-gather for Flashinfer cutlass MoE DP (max throughput) (#7667) · eff4eb3f
  Trevor Morris authored Aug 15, 2025
  
  eff4eb3f
15 Aug, 2025 11 commits
- Fix nan value generated after custom all reduce (#8663) · 983aa496
  kk authored Aug 16, 2025
```
Co-authored-by: wunhuang <wunhuang@amd.com>
```
  983aa496
- [AMD] Expand test coverage for AMD CI and enable... · 9c3e95d9
  Hubert Lu authored Aug 15, 2025
```
[AMD] Expand test coverage for AMD CI and enable apply_token_bitmask_inplace_cuda in sgl-kernel (#8268)
```
  9c3e95d9
- Cleanup MoE Refactor (#9223) · 84b006b2
  Cheng Wan authored Aug 15, 2025
  
  84b006b2
- [Eagle Warning fix] replace the deprecated 'and' with & (#9215) · 189af908
  Xuchun Shang authored Aug 15, 2025
```
Signed-off-by: Xuchun Shang <xuchun.shang@linux.alibaba.com>
```
  189af908
- Fix the deprecation warning for enable_flashinfer_mxfp4_moe (#9214) · e3e75a78
  Cheng Wan authored Aug 14, 2025
  
  e3e75a78
- fix: the store_dtype typo for ascend mla (#9208) · d4db9b02
  shilinlee authored Aug 15, 2025
```
Signed-off-by: shilinlee_com <836160610@qq.com>
```
  d4db9b02
- feat(hicache-3fs): 3FS-SGLang Hierarchical Cache Deployment Guide (#9213) · f7dd651d
  hzh0425 authored Aug 15, 2025
  
  f7dd651d
- [6/N] MoE Refactor: Cleanup MoE-related configs (#8849) · 29589512
  Cheng Wan authored Aug 14, 2025
  
  29589512
- fix: fix unsupported palette mode of images in bench_serving for mmmu (#9206) · 584e1ab2
  Mick authored Aug 15, 2025
  
  584e1ab2
- [typo fix] Fix a typo in communicator.py (#9183) · 004f7f19
  Philo authored Aug 15, 2025
```
Signed-off-by: Philo <lul16@foxmail.com>
```
  004f7f19
- feat: add fused moe config for Qwen3-235B-A22B-FP8 on B200 (#9204) · d2fbf2de
  zixuanzhang226 authored Aug 14, 2025
  
  d2fbf2de
14 Aug, 2025 21 commits
- chore: bump v0.5.0rc2 (#9203) · fab0f6e7
  Yineng Zhang authored Aug 14, 2025
  
  fab0f6e7
- feat: update model config (#9202) · 27985c27
  Yineng Zhang authored Aug 14, 2025
  
  27985c27
- chore: upgrade transformers 4.55.2 (#9197) · ac474869
  Yineng Zhang authored Aug 14, 2025
  
  ac474869
- [VLM] Improving multimodal tensor hash kernel (#9008) · 0b1e04f0
  Adarsh Shirawalmath authored Aug 15, 2025
  
  0b1e04f0
- feat: Add model version tracking with API endpoints and response metadata (#8795) · c1c7dc45
  Chengxing Xie authored Aug 15, 2025
  
  c1c7dc45
- [4/n]decouple quantization implementation from vLLM dependency (#9191) · 2cc9eeab
  Hongbo Xu authored Aug 15, 2025
```
Co-authored-by: AniZpZ <aniz1905@gmail.com>
Co-authored-by: Yineng Zhang <me@zhyncs.com>
```
  2cc9eeab
- refine mxfp4 shuffling log (#9194) · 63d82a77
  Xiaoyu Zhang authored Aug 15, 2025
  
  63d82a77
- [2/n]decouple quantization implementation from vLLM dependency (#8112) · 5aa1ebd2
  Peng Zhang authored Aug 14, 2025
```
Co-authored-by: walker-ai <yiyun.wyt@antgroup.com>
Co-authored-by: leoneo <1320612015@qq.com>
```
  5aa1ebd2
- fix: zero_init buffer (#9065) · 4dbf4360
  eigen authored Aug 14, 2025
```
Co-authored-by: Yineng Zhang <me@zhyncs.com>
```
  4dbf4360
- add w8a8-fp8-block-wise H20-3e triton config (#8018) · 3d6be1fb
  lukec authored Aug 14, 2025
  
  3d6be1fb
- Add H200 fused MoE kernel configs for DeepSeek-V3 in triton 3.3.1 (#7687) · 4063234c
  Jun Liu authored Aug 14, 2025
  
  4063234c
- Add H20 fused MoE kernel configs for Dpsk & Qwen3 (#7631) · 83feef5b
  Tommy Yang authored Aug 14, 2025
  
  83feef5b
- Add Triton Fused MoE kernel config for E=16 on B200 (#7004) · 2871eacc
  Brayden Zhong authored Aug 14, 2025
  
  2871eacc
- Add H200 fused MoE kernel tuning configs for Qwen3-Coder-480B-A35B-Instruct (#8852) · ac15bdc1
  forestlee95 authored Aug 14, 2025
  
  ac15bdc1
- Add A800 fused MoE kernel tuning configs for GLM4.5 and GLM4.5-Air (#8808) · d6451c3f
  Li Hui authored Aug 14, 2025
  
  d6451c3f
- fix io group (#9154) · 733446dd
  pansicheng authored Aug 14, 2025
```
Co-authored-by: Zhiqiang Xie <xiezhq@stanford.edu>
```
  733446dd
- Feature: support qwen and llama4 reducescatter for dp attention padding (#9101) · 4c22897a
  wxzhoucs authored Aug 14, 2025
  
  4c22897a
- Faster weight processing (trtllm-gen moe nvfp4) (#9162) · 1bc183c6
  Alex Yang authored Aug 13, 2025
  
  1bc183c6
- [DP Attention] Refactor: adding some utility functions (#9136) · b87aacb5
  Cheng Wan authored Aug 13, 2025
  
  b87aacb5
- [Bugfix] Avoid unnecessary reduce-scatter call in prepare_mlp (#9169) · 98457c04
  Huaixin Chang authored Aug 14, 2025
  
  98457c04
- Replace `sglang.srt.layers.quantization.scalar_types` with `sgl_kernel.scalar_type` (#8951) · a669bc2f
  Hongbo Xu authored Aug 14, 2025
  
  a669bc2f
13 Aug, 2025 4 commits
- Fix broken trtllm_mha attn backend with gpt-oss (#9161) · 6b7c2471
  Nicolas Castet authored Aug 13, 2025
  
  6b7c2471
- [Generative Score API] Optimization to Remove Decode. (#8840) · a027a9b4
  Sundara Raman Ramachandran authored Aug 13, 2025
  
  a027a9b4
- Clean up allocators (#9134) · 9e426466
  Lianmin Zheng authored Aug 13, 2025
```
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
```
  9e426466
- [Model] Support Qwen3ForSequenceClassification for Qwen3-Embed Model (#7957) · 65736dc5
  Zhihao Liu authored Aug 14, 2025
  
  65736dc5