Commits · c0f0b209bb177386ca7c4fc8d7d08ad00bb7e7cd · OpenDAS / vllm_cscc

20 Aug, 2025 3 commits
- [BugFix] fix CUTLASS MLA full cudagraph (#23200) · 3a8708f6
  Lucas Wilkinson authored Aug 19, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: simon-mo <simon.mo@hey.com>
```
  3a8708f6
- update utils.py · ad038b4e
  zhuwenwen authored Aug 20, 2025
  
  ad038b4e
- fix ds-v2 run error · 78d833ae
  zhuwenwen authored Aug 20, 2025
  
  78d833ae
18 Aug, 2025 2 commits
- 增加marlin对cache13的支持，以及新增flash mla的kvcache fp8的支持 · 22a46529
  zhuwenwen authored Aug 18, 2025
  
  22a46529
- add v1 engine + deepseek r1 mtp + zero-overhead scheduler · 22a95571
  zhuwenwen authored Aug 18, 2025
  
  22a95571
16 Aug, 2025 2 commits
- [Bugfix gpt-oss] Fix float32 convert for flashinfer sink support (#23016) · 000cceca
  Michael Goin authored Aug 16, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  000cceca
- [Kernel] Simplify `get_kv_cache_layout` and cache `use_trtllm_attention` env-dependent bit (#22735) · 070da660
  Nicolò Lucchesi authored Aug 16, 2025
```
Signed-off-by: NickLucche <nlucches@redhat.com>
```
  070da660
15 Aug, 2025 5 commits

minor: zero workspace buffer init for flashinfer trtllm-gen attn (#22603) · 1723ef1a
eigen authored Aug 15, 2025

1723ef1a

[Core] Allow full cudagraph with separate attention routines and orthogonal to... · 74f441f4

fhl2000 authored Aug 15, 2025


[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)
Signed-off-by: fhl <2410591650@qq.com>
Signed-off-by: fhl2000 <63384265+fhl2000@users.noreply.github.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>

74f441f4

[V1] - Split Prefill and Decode for Mamba1 models (#22653) · fe91ce95

amirai21 authored Aug 15, 2025


Signed-off-by: amirk <amirk@ai21.com>
Signed-off-by: asafg <asafg@ai21.com>
Co-authored-by: asafg <asafg@ai21.com>
Co-authored-by: Asaf Joseph Gardin <39553475+Josephasafg@users.noreply.github.com>

fe91ce95

[Mamba] - refactor: Renamed mamba_attn to mamba2_attn (#22818) · 3d232dbd
Asaf Joseph Gardin authored Aug 15, 2025
```
Signed-off-by: asafg <asafg@ai21.com>
Co-authored-by: asafg <asafg@ai21.com>
```
3d232dbd
[Feature] Full Cuda Graph Support for Cutlass MLA and 6% E2E Throughput Improvement (#22763) · 5c3fbfe4
Wentao Ye authored Aug 15, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
5c3fbfe4

13 Aug, 2025 5 commits
- [Bugfix][mamba] Fix type annotation of Mamba2Metadata (#22787) · fceafaf5
  Chen Zhang authored Aug 13, 2025
```
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
```
  fceafaf5
- [V1] Add tree drafting tests for eagle spec decoding (#22705) · d94e3026
  Giancarlo Delfin authored Aug 13, 2025
```
Signed-off-by: Giancarlo Delfin <gdelfin@meta.com>
```
  d94e3026
- add VLLM_USE_PA_PRINT_PARAM to print v1 fa size · 43e67933
  zhuwenwen authored Aug 13, 2025
  
  43e67933
- [feat]优化mtp/eagle的计算逻辑，减少第1层并行解码的计算重复（num_accepted_tokens_tensor修改暂未合入） · 4874e3e0
  zhuwenwen authored Aug 13, 2025
  
  4874e3e0
- Force TRTLLM attention for gpt-oss on SM100 (#22678) · c6b92879
  Michael Goin authored Aug 13, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  c6b92879
12 Aug, 2025 2 commits

[Kernel][AMD] Avoid D2H copy and cumsum kernel (#22683) · 6bd8ebf0

Xiaozhu Meng authored Aug 12, 2025


Signed-off-by: Xiaozhu <mxz297@gmail.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

6bd8ebf0

[Bugfix] Fix ModernBert load & Enable sliding window attention for... · 6d729c43

wang.yuqi authored Aug 12, 2025


[Bugfix] Fix ModernBert load & Enable sliding window attention for bidirectional attention. (#22637)
Signed-off-by: wang.yuqi <noooop@126.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Co-authored-by: Max de Bayser <mbayser@br.ibm.com>

6d729c43

11 Aug, 2025 1 commit
- [fix]修复v1 mtp接受率低的问题 · db554c6c
  zhuwenwen authored Aug 11, 2025
```
更改默认的full _cuda_graph启动方式为false
```
  db554c6c
10 Aug, 2025 1 commit
- [V1] [Hybrid] Enable Full CUDA Graph (decode-only) for Mamba layers (#21401) · 61f67d8a
  Thomas Parnell authored Aug 10, 2025
```
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
```
  61f67d8a
09 Aug, 2025 1 commit
- [V1] [Hybrid] Support Minimax-Text-01 in V1 (#22151) · 6ade99ea
  Thomas Parnell authored Aug 09, 2025
```
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
```
  6ade99ea
08 Aug, 2025 3 commits
- [BugFix] Fix IMA FlashMLA full cuda-graph and DP + Update FlashMLA (#21691) · cd9b9de1
  Lucas Wilkinson authored Aug 08, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
```
  cd9b9de1
- update fa full_cuda_graph support · 71b02b7a
  zhuwenwen authored Aug 08, 2025
  
  71b02b7a
- [bugfix] Fix Llama3/4 issues caused by FlashInfer 0.2.10 (#22426) · af473f0a
  Po-Han Huang (NVIDIA) authored Aug 08, 2025
```
Signed-off-by: Po-Han Huang <pohanh@nvidia.com>
```
  af473f0a
07 Aug, 2025 5 commits

[feat]支持mtp模型full_cuda_graph · ffe9e7db
zhuwenwen authored Aug 07, 2025

ffe9e7db

[Attention] Support multiple attention metadata builders per kv_cache_spec +... · 1dc8a70b

Lucas Wilkinson authored Aug 06, 2025


[Attention] Support multiple attention metadata builders per kv_cache_spec  + proper local attention no hybrid kv cache fix (#21588)
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>

1dc8a70b

Support encoder_only attention for FlexAttention (#22273) · f825c6bd
Maximilien de Bayser authored Aug 06, 2025
```
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
```
f825c6bd

Fix trtllm-gen attention env and add attention sink (#22378) · 9a3835aa

Lain authored Aug 06, 2025


Signed-off-by: Siyuan Fu <siyuanf@nvidia.com>
Signed-off-by: Lain <fusiyuan2000@hotmail.com>
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Yongye Zhu <zyy1102000@gmail.com>

9a3835aa

[v1] - Mamba1 Attention Metadata (#21249) · 46a13949
Asaf Joseph Gardin authored Aug 07, 2025
```
Signed-off-by: asafg <asafg@ai21.com>
Co-authored-by: asafg <asafg@ai21.com>
```
46a13949

06 Aug, 2025 3 commits

[gpt-oss] Enhance error msg on attention sink init (#22335) · 31f5dc5b

Yongye Zhu authored Aug 06, 2025


Signed-off-by: simon-mo <xmo@berkeley.edu>
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: simon-mo <xmo@berkeley.edu>

31f5dc5b

[gpt-oss] flashinfer attention sink init (#22330) · 90ec0069

Yongye Zhu authored Aug 05, 2025


Signed-off-by: simon-mo <xmo@berkeley.edu>
Co-authored-by: LiuXiaoxuanPKU <lilyliupku@gmail.com>
Co-authored-by: simon-mo <xmo@berkeley.edu>
Co-authored-by: Chen Zhang <zhangch99@outlook.com>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: Hongxia Yang <62075498+hongxiayang@users.noreply.github.com>
Co-authored-by: Minseok Lee <47620120+minseokl@users.noreply.github.com>

90ec0069

Add attention sink in attention backends (#22320) · 6e209243

Woosuk Kwon authored Aug 05, 2025


Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: LiuXiaoxuanPKU <lilyliupku@gmail.com>
Co-authored-by: simon-mo <xmo@berkeley.edu>
Co-authored-by: Chen Zhang <zhangch99@outlook.com>
Co-authored-by: Hongxia Yang <62075498+hongxiayang@users.noreply.github.com>
Co-authored-by: Minseok Lee <47620120+minseokl@users.noreply.github.com>
Co-authored-by: Yongye Zhu <zyy1102000@gmail.com>

6e209243

05 Aug, 2025 3 commits
- [V1] port xformers backend to v1 (#21342) · 469b3ffa
  Giancarlo Delfin authored Aug 05, 2025
```
Signed-off-by: Giancarlo Delfin <gdelfin@meta.com>
```
  469b3ffa
- [NVIDIA] Support Flashinfer TRT-LLM Prefill Attention Kernel (#22095) · 83156c7b
  elvischenv authored Aug 05, 2025
```
Signed-off-by: elvischenv <219235043+elvischenv@users.noreply.github.com>
```
  83156c7b
- [Bugfix] Misaligned params in TreeAttentionImpl (#22226) · cdfd6871
  Cyrus Leung authored Aug 05, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  cdfd6871
04 Aug, 2025 2 commits
- Add tree attention backend for v1 (part 1) (#20401) · aa7012eb
  Giancarlo Delfin authored Aug 03, 2025
```
Signed-off-by: Giancarlo Delfin <gdelfin@meta.com>
```
  aa7012eb
- [fix] fix correct assertion syntax error in attention utils. (#22154) · e27d25a0
  ZiTian.Zhao authored Aug 04, 2025
```
Signed-off-by: zitian.zhao <zitian.zhao@tencentmusic.com>
```
  e27d25a0
02 Aug, 2025 2 commits
- [V1] [Hybrid] Validate compatibility of attention backend batch reordering at init time (#21557) · 4abfd879
  Thomas Parnell authored Aug 02, 2025
```
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
```
  4abfd879
- [Attention][DBO] Add support for "splitting" the CommonAttentionMetadata (#21153) · 0edaf752
  Sage Moore authored Aug 01, 2025
```
Signed-off-by: Sage Moore <sage@neuralmagic.com>
```
  0edaf752