Commits · 4c54f4420217dcbed3aa66d94b5865665c4bac26 · change / sglang

04 Apr, 2025 2 commits
- Add DeepSeek V3/R1 shared experts fusion (#4918) · 924ca7c9
  Xiaoyu Zhang authored Apr 04, 2025
  
  924ca7c9
- Small refactor DeepEPDispatcher into subclasses (#4994) · febe21ce
  fzyzcjy authored Apr 04, 2025
  
  febe21ce
03 Apr, 2025 3 commits
- Revert "Replace enable_flashinfer_mla argument with attention_backend" (#5048) · 74885a84
  Lianmin Zheng authored Apr 03, 2025
  
  74885a84
- Small refactor DeepEPMode to clean up code a bit (#4992) · 8e10fec9
  fzyzcjy authored Apr 03, 2025
  
  8e10fec9
- Replace enable_flashinfer_mla argument with attention_backend (#5005) · e8999b13
  Baizhou Zhang authored Apr 03, 2025
  
  e8999b13
01 Apr, 2025 1 commit

[Feature] Support DeepEP Low Latency (#4767) · 23c764b1

Jinyan Chen authored Apr 02, 2025


Co-authored-by: sleepcoo <sleepcoo@gmail.com>
Co-authored-by: laixinn <xielx@shanghaitech.edu.cn>
Co-authored-by: ch-wan <cwan39@gatech.edu>

23c764b1

31 Mar, 2025 2 commits
- refactor: multimodal data (#4754) · 5cb552b1
  Mick authored Apr 01, 2025
  
  5cb552b1
- Fix DeepSeek bug causing 2.2% MMLU drop when TP!=DP (#4883) · a303325f
  fzyzcjy authored Mar 31, 2025
```
Co-authored-by: ch-wan <cwan39@gatech.edu>
```
  a303325f
30 Mar, 2025 1 commit
- Fix 2-gpu CI test and suppress some warnings (#4930) · 9adf178c
  Lianmin Zheng authored Mar 30, 2025
  
  9adf178c
29 Mar, 2025 1 commit
- [Feature] Support FA3 backend for MLA (#4831) · 20c90be2
  Baizhou Zhang authored Mar 28, 2025
  
  20c90be2
28 Mar, 2025 2 commits
- Clean up `import vllm` in quantization/__init__.py (#4834) · 74e0ac1d
  Lianmin Zheng authored Mar 28, 2025
  
  74e0ac1d
- Support (1 <= dp < tp) in the dp attention in DeepEP (#4770) · 7f19e083
  tarinkk authored Mar 27, 2025
```
Co-authored-by: Cheng Wan <cwan39@gatech.edu>
```
  7f19e083
26 Mar, 2025 1 commit
- Support compressed tensors fp8w8a8 (#4743) · 04e3ff69
  Xiaoyu Zhang authored Mar 27, 2025
  
  04e3ff69
25 Mar, 2025 1 commit
- 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
23 Mar, 2025 2 commits
- 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 1 commit
- 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
19 Mar, 2025 2 commits
- avoid cudaStreamSynchronize in DeepSeekV2AttentionMLA (#4577) · df7014a8
  strgrb authored Mar 20, 2025
```
Co-authored-by: Zhang Kaihong <zhangkaihong.zkh@alibaba-inc.com>
```
  df7014a8
- [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
18 Mar, 2025 2 commits
- Reduce computation and communication in DP attention (#4521) · 3196999f
  Cheng Wan authored Mar 18, 2025
  
  3196999f
- cleanup deps 3/n (#4541) · c16b33cc
  Yineng Zhang authored Mar 18, 2025
  
  c16b33cc
17 Mar, 2025 1 commit

[Feature] Support Deepseek-VL2 (#2798) · d6d21640

萝卜菜 authored Mar 17, 2025


Co-authored-by: Edenzzzz <wtan45@wisc.edu>
Co-authored-by: Chayenne <zhaochen20@outlook.com>
Co-authored-by: Yi Zhang <1109276519@qq.com>

d6d21640

16 Mar, 2025 1 commit
- cleanup deps 2/n (#4464) · 65b7c9b7
  Yineng Zhang authored Mar 15, 2025
  
  65b7c9b7
13 Mar, 2025 3 commits
- Improve DP attention (#4390) · 8e66fbec
  Lianmin Zheng authored Mar 13, 2025
```
Co-authored-by: dhou-xai <dhou@x.ai>
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
```
  8e66fbec
- Revert "[XPU][CPU] Enable the native path of DeepSeek" (#4367) · 45de8971
  Lianmin Zheng authored Mar 12, 2025
  
  45de8971
- [XPU][CPU] Enable the native path of DeepSeek (#4086) · 71046fcd
  Meng, Hengyu authored Mar 13, 2025
```
Co-authored-by: Zhang, Liangang <liangang.zhang@intel.com>
```
  71046fcd
12 Mar, 2025 1 commit
- unify is_cuda and is_hip (#4321) · d1da58e2
  Yineng Zhang authored Mar 11, 2025
  
  d1da58e2
10 Mar, 2025 1 commit
- [Bug fixed] fixed the crash when enable the dp-attention on the single card (#3958) · 4455b26e
  DavidChan authored Mar 10, 2025
  
  4455b26e
09 Mar, 2025 1 commit
- Support nextn for flashinfer mla attention backend (#4218) · 9fb48f95
  Baizhou Zhang authored Mar 09, 2025
  
  9fb48f95
07 Mar, 2025 1 commit

[Feature] DeepSeek V3/R1 INT8 Quantization (channel-wise) (#3888) · c7f25446

HandH1998 authored Mar 07, 2025


Co-authored-by: yych0745 <1398089567@qq.com>
Co-authored-by: sleepcoo <sleepcoo@gmail.com>
Co-authored-by: b0urnee <2769086541@qq.com>

c7f25446

05 Mar, 2025 2 commits
- [QUANT] Add GPTQModel Dynamic Quantization + `lm_head` Quantization (#3790) · 56a724eb
  Qubitium-ModelCloud authored Mar 05, 2025
```
Signed-off-by: ZX-ModelCloud <zx@modelcloud.ai>
Co-authored-by: ZX-ModelCloud <zx@modelcloud.ai>
```
  56a724eb
- [Minor] more code cleanup (#4077) · e074d84e
  Lianmin Zheng authored Mar 04, 2025
  
  e074d84e
03 Mar, 2025 1 commit
- Share target model embed and head weights for nextn (#4033) · 9fafa62d
  Ke Bao authored Mar 04, 2025
  
  9fafa62d
01 Mar, 2025 1 commit

[Feature]Support ragged prefill in flashinfer mla backend (#3967) · 90a4b7d9

Baizhou Zhang authored Feb 28, 2025


Co-authored-by: Yineng Zhang <me@zhyncs.com>
Co-authored-by: pankajroark <pankajroark@users.noreply.github.com>

90a4b7d9

25 Feb, 2025 2 commits
- Fix allgather ops inside cuda graphs (#3709) · 127998cc
  Nicolas Castet authored Feb 25, 2025
  
  127998cc
- [ROCm] Enable Fused MLA Triton kernel for DeepSeekV3 (#3237) · 6ce9dbe8
  Chaitanya Sri Krishna Lolla authored Feb 25, 2025
```
Co-authored-by: HAI <hixiao@gmail.com>
```
  6ce9dbe8
24 Feb, 2025 2 commits
- Feature DeepSeek V3/R1 INT8 Quantization (block-wise) (#3730) · 1a6e9757
  laixin authored Feb 24, 2025
```
Co-authored-by: HandH1998 <1335248067@qq.com>
```
  1a6e9757
- Refactor flashinfer logic for deepseek v3 and fix accuracy bug (#3785) · b1100846
  Baizhou Zhang authored Feb 24, 2025
  
  b1100846
17 Feb, 2025 1 commit
- feat: support flashinfer mla with prefix cache (#3643) · 714f3e63
  Yineng Zhang authored Feb 18, 2025
  
  714f3e63
14 Feb, 2025 1 commit
- Support NextN (MTP) speculative decoding for DeepSeek-V3/R1 (#3582) · 862dd76c
  Ke Bao authored Feb 15, 2025
  
  862dd76c