Commits · 04e3ff69753f5b18512a67094592a4d04a05859c · zhaoyu6 / sglang · GitLab

26 Mar, 2025 3 commits
- Support compressed tensors fp8w8a8 (#4743) · 04e3ff69
  Xiaoyu Zhang authored Mar 27, 2025
  
  04e3ff69
- Fix shared memory OOM on sm86 GPUs. (#4797) · 45fdf1f7
  Yi Pan authored Mar 27, 2025
  
  45fdf1f7
- fix(typo): fix `reply` to `replay` in `base_attn_backend.py` (#4784) · ced35a06
  Thysrael authored Mar 26, 2025
  
  ced35a06
25 Mar, 2025 3 commits
- Add endpoints to dump selected expert ids (#4435) · 199bb01d
  yuhsaun-t authored Mar 24, 2025
```
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
```
  199bb01d
- fix FlashMLA cudagraph config (#4691) · 57eec0bf
  lukec authored Mar 25, 2025
```
Co-authored-by: yinfan98 <1106310035@qq.com>
```
  57eec0bf
- Fix circular imports in gptq.py and unblock test explorer (#4736) · 4c584fc6
  Stefan He authored Mar 24, 2025
  
  4c584fc6
24 Mar, 2025 2 commits
- Support FA3 as Attention backend by using `--attention-backend fa3` (#4680) · 5d7edc8e
  Stefan He authored Mar 23, 2025
```
Co-authored-by: qsong <qsong@linkedin.com>
Co-authored-by: qingquansong <ustcsqq@gmail.com>
```
  5d7edc8e
- Fix RotaryEmbedding when using Triton backend for EXAONE-3.5-2.4B (#4064) · 2a206b22
  Kyungmin Lee authored Mar 24, 2025
  
  2a206b22
23 Mar, 2025 3 commits
- refactor: bug fixes and refactor for vlm (#4661) · 11577ced
  Mick authored Mar 23, 2025
  
  11577ced
- Support async in DeepEP (#4610) · ca75741e
  fzyzcjy authored Mar 23, 2025
```
Co-authored-by: Cheng Wan <cwan39@gatech.edu>
```
  ca75741e
- Multiple tiny code cleanups (#4608) · c6d549e7
  fzyzcjy authored Mar 23, 2025
  
  c6d549e7
22 Mar, 2025 3 commits
- Optimize Permute Kernel in DeepEP (#4643) · c2bd094d
  xutizhou authored Mar 23, 2025
```
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
```
  c2bd094d
- [Bug Fix] Add partial rotary factor support for Phi-4 and upgrade to transformers v4.50.0 (#3984) · f8f9244a
  Adarsh Shirawalmath authored Mar 23, 2025
```
Co-authored-by: Chayenne <zhaochen20@outlook.com>
```
  f8f9244a
- [quantization] fix channelwise conversion with scalar weight scale (#4596) · 8cd42504
  Yun Dai authored Mar 22, 2025
  
  8cd42504
21 Mar, 2025 2 commits
- Set deepgemm to the default value in the hopper architecture. (#4613) · 4c56e5db
  lukec authored Mar 21, 2025
  
  4c56e5db
- fix SUPPORT_CUTLASS_BLOCK_FP8 flag (#4640) · 7b5fc719
  Cheng Wan authored Mar 21, 2025
  
  7b5fc719
20 Mar, 2025 1 commit
- [fix] fix illegal mem access and clean up triton attention backend (#4571) · 9e93ef3f
  JieXin Liang authored Mar 20, 2025
  
  9e93ef3f
19 Mar, 2025 5 commits
- Support FlashMLA backend cuda graph (#4514) · b6944f97
  lukec authored Mar 19, 2025
```
Co-authored-by: yinfan98 <1106310035@qq.com>
Co-authored-by: Hongbosherlock <hongbosherlock@gmail.com>
Co-authored-by: ispobock <ispobaoke@163.com>
```
  b6944f97
- [Feature] Integrate DeepEP into SGLang (#4232) · f44db16c
  Jinyan Chen authored Mar 19, 2025
```
Co-authored-by: Cheng Wan <cwan39@gatech.edu>
Co-authored-by: Xuting Zhou <xutingz@nvidia.com>
```
  f44db16c
- Create col-major and tma-aligned x_scale for deep_gemm.gemm_fp8_fp8_bf16_nt (#4515) · f9c53cbb
  strgrb authored Mar 19, 2025
```
Co-authored-by: Zhang Kaihong <zhangkaihong.zkh@alibaba-inc.com>
```
  f9c53cbb
- Optimize Triton decoding kernel for dynamic workload (#4553) · c0e9a36c
  JieXin Liang authored Mar 19, 2025
  
  c0e9a36c
- [Feature] Support Tensor Parallelism and Weight Slicing for Lora (#4274) · 588865f0
  aoshen524 authored Mar 18, 2025
```
Co-authored-by: ShenAo1111 <1377693092@qq.com>
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
```
  588865f0
18 Mar, 2025 6 commits
- Reduce computation and communication in DP attention (#4521) · 3196999f
  Cheng Wan authored Mar 18, 2025
  
  3196999f
- [Feature] Support EAGLE 3 (#4247) · 9e0186f3
  James Liu authored Mar 18, 2025
  
  9e0186f3
- cleanup deps 3/n (#4541) · c16b33cc
  Yineng Zhang authored Mar 18, 2025
  
  c16b33cc
- remove useless backend forward in rotary_embedding (#4500) · 804d250a
  Xiaoyu Zhang authored Mar 18, 2025
  
  804d250a
- [Hotfix] solve fp8 w8a8 ci test fail (#4531) · dd865bef
  Xiaoyu Zhang authored Mar 18, 2025
  
  dd865bef
- fix: second_per_grid_ts should be used to get mrope position (#3682) · d373a48c
  Mick authored Mar 18, 2025
  
  d373a48c
17 Mar, 2025 9 commits
- sglang quant module remove vllm dependency (#4507) · 9b81f9bd
  Xiaoyu Zhang authored Mar 18, 2025
  
  9b81f9bd
- Revert "feat: update grouped_topk to support softmax and sigmoid" (#4505) · 3ded4b21
  Ke Bao authored Mar 18, 2025
  
  3ded4b21
- Remove redundant type conversion (#4513) · 82dec1f7
  Lianmin Zheng authored Mar 17, 2025
  
  82dec1f7
- [ROCm] fix dtype (#4510) · 5f9b2c62
  yiakwy-xpu-ml-framework-team authored Mar 17, 2025
  
  5f9b2c62
- Fix data parallel + tensor parallel (#4499) · 5493c334
  Lianmin Zheng authored Mar 17, 2025
  
  5493c334
- Support Online Quantization for W8A8 (#4485) · ef3c2dd0
  Stefan He authored Mar 17, 2025
  
  ef3c2dd0
- Support serving DeepSeek-R1-Channel-INT8 with 32 L40S. (#4418) · 75b65648
  Wenbo Yang authored Mar 17, 2025
  
  75b65648
- config: Update fused moe config (#4493) · 0f52fb55
  Mick authored Mar 17, 2025
  
  0f52fb55
- Urgent model support: support gemma-3-it (#4424) · 9d02bb3e
  Mick authored Mar 17, 2025
  
  9d02bb3e
16 Mar, 2025 3 commits
- Support FlashMLA backend (#4472) · a53fe428
  lukec authored Mar 17, 2025
```
Co-authored-by: yinfan98 <1106310035@qq.com>
```
  a53fe428
- cleanup deps 2/n (#4464) · 65b7c9b7
  Yineng Zhang authored Mar 15, 2025
  
  65b7c9b7
- perf: update fused moe config (#4459) · 8ec2ce07
  Mick authored Mar 16, 2025
  
  8ec2ce07