Commits · 4dbf43601d49d1c62018b62c377aa4560d492394 · change / sglang · GitLab

14 Aug, 2025 13 commits
- 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 12 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
- chore: bump v0.5.0rc1 (#9069) · 7b56e494
  Yineng Zhang authored Aug 13, 2025
  
  7b56e494
- Support FA3 backend for gpt-oss (#9028) · 0ff6d1fc
  Ke Bao authored Aug 14, 2025
  
  0ff6d1fc
- [PD] feat: mooncake use batch reg/dereg (#8910) · 4a16a71c
  Teng Ma authored Aug 14, 2025
```
Co-authored-by: Shangming Cai <csmthu@gmail.com>
```
  4a16a71c
- [PD] optimize kv cache transfer directly using batch transfer (#9149) · a16923ef
  Francis authored Aug 14, 2025
```
Co-authored-by: Shangming Cai <csmthu@gmail.com>
```
  a16923ef
- Double vision prefill throughput by defaulting to optimal vision attention backend (#8484) · 3b3b3baf
  Kevin Xiang Li authored Aug 13, 2025
```
Co-authored-by: Xiang (Kevin) Li <lik@nvidia.com>
```
  3b3b3baf
- Fix gpt-oss ~2x memory consumption issue (#9146) · 9394ed63
  fzyzcjy authored Aug 13, 2025
  
  9394ed63
- Support Triton FP8 Gemm can handle hidden_dim not divisible by 16 (#9093) · 930fe467
  Stefan He authored Aug 12, 2025
```
Co-authored-by: Xiaoyu Zhang <35585791+BBuf@users.noreply.github.com>
```
  930fe467
- Use FlashInfer's TRTLLM FP8 Blockscale GEMM (#8588) · 8723b4f1
  Elfie Guo authored Aug 12, 2025
  
  8723b4f1
12 Aug, 2025 15 commits
- Support page first layout zero copy for mooncake store (#8651) · 0edda320
  huangtingwei authored Aug 13, 2025
```
Co-authored-by: Zhiqiang Xie <xiezhq@stanford.edu>
```
  0edda320
- [AMD] Support Wave attention backend with AMD GPU optimizations (#8660) · 25caa7a8
  jacky.cheng authored Aug 13, 2025
```
Signed-off-by: Stanley Winata <stanley.winata@amd.com>
Signed-off-by: Harsh Menon <harsh@nod-labs.com>
Signed-off-by: nithinsubbiah <nithinsubbiah@gmail.com>
Signed-off-by: Ivan Butygin <ivan.butygin@gmail.com>
Signed-off-by: xintin <gaurav.verma@amd.com>
Co-authored-by: Harsh Menon <harsh@nod-labs.com>
Co-authored-by: Stanley Winata <stanley.winata@amd.com>
Co-authored-by: Stanley Winata <68087699+raikonenfnu@users.noreply.github.com>
Co-authored-by: Stanley Winata <stanley@nod-labs.com>
Co-authored-by: Ivan Butygin <ivan.butygin@gmail.com>
Co-authored-by: nithinsubbiah <nithinsubbiah@gmail.com>
Co-authored-by: Nithin Meganathan <18070964+nithinsubbiah@users.noreply.github.com>
Co-authored-by: Ivan Butygin <ibutygin@amd.com>
```
  25caa7a8
- [Quantization] Supported w8a8 int8 quantized Gemma3 and Qwen-VL models (#8619) · 83123f48
  ichernob authored Aug 12, 2025
```
Co-authored-by: ronnie_zheng <zl19940307@163.com>
```
  83123f48
- [feat] Enable Ascend profiling on SGLang (#8610) · 48afa8f1
  ronnie_zheng authored Aug 13, 2025
```
Co-authored-by: liyou_b <2953090824@qq.com>
```
  48afa8f1
- Fuse writing KV buffer into rope kernel (part 2: srt) (#9014) · c9ee7385
  Jiaqi Gu authored Aug 12, 2025
```
Co-authored-by: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
```
  c9ee7385
- [5/n] DP Enhancement: Correct `num_token_non_padded` (#9107) · 5f5b3b24
  Cheng Wan authored Aug 12, 2025
  
  5f5b3b24
- [Bugfix] Fix accuracy-test-1-gpu failure caused by `builtin_tools` (#9114) · f2a5de28
  Chang Su authored Aug 12, 2025
  
  f2a5de28
- Fuse writing KV buffer into rope kernel (part 1: sgl-kernel) (#9077) · 9aea2555
  fzyzcjy authored Aug 12, 2025
  
  9aea2555
- Fuse two kernels of hidden states padding into quantization kernel (#9005) · 5190ba7f
  fzyzcjy authored Aug 12, 2025
```
Co-authored-by: Qiaolin-Yu <liin1211@outlook.com>
```
  5190ba7f
- bugfix: Fix the commentary msg extraction in GptOssDetector (#9097) · 9c83d74d
  Chang Su authored Aug 11, 2025
  
  9c83d74d
- [Fix] Fix dual chunk model default behavior (#9032) · b4ac2b9c
  DarkSharpness authored Aug 11, 2025
  
  b4ac2b9c
- Fix mismatch between padded_scales shape and reshape dimensions in modelopt quantization (#8766) · 83262dcb
  Jianwei Dong authored Aug 12, 2025
  
  83262dcb
- feat: add fused moe config for Qwen3-30B-A3B on B200 (#9087) · c46c75f8
  zixuanzhang226 authored Aug 11, 2025
  
  c46c75f8
- Optimization for AscendPagedTokenToKVPoolAllocator (#8293) · 2aaf22c4
  Makcum888e authored Aug 12, 2025
```
Co-authored-by: ronnie_zheng <zl19940307@163.com>
Co-authored-by: VDV1985 <vladdv85@mail.ru>
```
  2aaf22c4
- Fix race condition in async lora unload (#9084) · 5ded39ca
  Lifu Huang authored Aug 11, 2025
  
  5ded39ca