Commits · 5beacce2eae2cb36e44498e2ea80b2d3ee34fe61 · OpenDAS / vllm_cscc

22 Oct, 2025 2 commits
- [BugFix] bugfix for Flash Attention MLA with full cuda graph IMA following pr-25490 (#27128) · 5beacce2
  Daisy-Ma-coder authored Oct 22, 2025
```
Signed-off-by: qqma <qqma@amazon.com>
Co-authored-by: qqma <qqma@amazon.com>
```
  5beacce2
- [Bugfix] Disable FlexAttention direct block mask building for encoder-only models (#27344) · 084a9dae
  Isotr0py authored Oct 23, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  084a9dae
21 Oct, 2025 1 commit

[Bugfix] fixes the decoding metadata of dense mla's fp8 kvcache. (#27144) · 250fb1b8

Tao He authored Oct 22, 2025


Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>

250fb1b8

18 Oct, 2025 2 commits
- [Chore] Clean up pytorch helper functions in `vllm.utils` (#26908) · 6ac5e06f
  Isotr0py authored Oct 19, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: isotr0py <2037008807@qq.com>
```
  6ac5e06f
- [fix][cpu] fix prefill attention in CPU attention backend (#27035) · ab4be40f
  Fadi Arafeh authored Oct 18, 2025
```
Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>
```
  ab4be40f
16 Oct, 2025 4 commits
- [small][batch invariance] Rename the env and internal flags to simplify usage (#26855) · b2f78cba
  Bram Wasti authored Oct 16, 2025
```
Signed-off-by: Bram Wasti <bwasti@meta.com>
```
  b2f78cba
- [Feature] Add process_weights_after_loading to AttentionImpl (#26870) · 5afd3276
  rongfu.leng authored Oct 16, 2025
```
Signed-off-by: rongfu.leng <rongfu.leng@daocloud.io>
```
  5afd3276
- Deepseek-v3 Batch Invariant on 8xH100 (#26609) · 7d8975de
  Bram Wasti authored Oct 15, 2025
```
Signed-off-by: Bram Wasti <bwasti@meta.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
```
  7d8975de
- [PERF] Qwen3-next MTP speedup (change bool mask indexing to index_select /... · 785d8b64
  Vadim Gimpelson authored Oct 16, 2025
```
[PERF] Qwen3-next MTP speedup (change bool mask indexing to index_select / index_copy to reduce d2h) (#26437)
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
```
  785d8b64
15 Oct, 2025 1 commit

Move query quantization to attention layer for Flashinfer & Triton. (#26534) · 0a9ef0cf

Adrian Abeyta authored Oct 15, 2025


Signed-off-by: adabeyta <aabeyta@redhat.com>
Signed-off-by: Adrian Abeyta <aabeyta@redhat.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>

0a9ef0cf

14 Oct, 2025 3 commits

remove attn output view kernel (#26680) · a86b4c58

Boyuan Feng authored Oct 14, 2025

Signed-off-by: Boyuan Feng <boyuan@meta.com>
Signed-off-by: Boyuan Feng <fby.1994@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

a86b4c58

[Attention][Spec Decode] FlashMLA spec decode support (#26541) · 82af928c
Matthew Bonanni authored Oct 14, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
82af928c

[DCP] Support Decode Context Parallel (DCP) for GQA with FlashAttention (#24864) · ea97940d

Jaya Yuan authored Oct 14, 2025


Signed-off-by: yuanyongjie.yyj <yuanyongjie.yyj@antgroup.com>
Signed-off-by: FENP <32334296+FENP@users.noreply.github.com>
Signed-off-by: Jaya Yuan <yuanyongjie.yyj@antgroup.com>

ea97940d

13 Oct, 2025 2 commits
- [Log] Optimize Startup Log (#26601) · e251e457
  Wentao Ye authored Oct 13, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  e251e457
- [unrevert] Add batch invariant kernel override for FlashInfer backend [2/n] (#26373) · 32637990
  Bram Wasti authored Oct 13, 2025
```
Signed-off-by: Bram Wasti <bwasti@meta.com>
Signed-off-by: Bram Wasti <bwasti@fb.com>
```
  32637990
12 Oct, 2025 1 commit
- Update `Optional[x]` -> `x | None` and `Union[x, y]` to `x | y` (#26633) · 8fcaaf6a
  Harry Mellor authored Oct 12, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  8fcaaf6a
10 Oct, 2025 1 commit
- [deepseek] kernel block size for UniformTypeKVCacheSpecs (#26559) · 6f0f570c
  Chen Zhang authored Oct 10, 2025
```
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
```
  6f0f570c
09 Oct, 2025 5 commits
- [Bugfix] Fix CUDA graph selection bug in FlashInfer at high concurrency (#26499) · 6e783bc5
  Benjamin Chislett authored Oct 09, 2025
```
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
```
  6e783bc5
- [Attention][DCP] Support DCP with query length > 1 (MTP) with FA3 (#25049) · 3b736e1c
  Ming Yang authored Oct 09, 2025
```
Signed-off-by: Ming Yang <minos.future@gmail.com>
```
  3b736e1c
- Bump Flashinfer to v0.4.0 (#26326) · 5e49c3e7
  elvischenv authored Oct 09, 2025
```
Signed-off-by: elvischenv <219235043+elvischenv@users.noreply.github.com>
```
  5e49c3e7
- [Hybrid]: Decouple Kernel Block Size from KV Page Size (#24486) · d24cf322
  Zhiyuan Li authored Oct 09, 2025
```
Signed-off-by: lizhiyuan <uniartisan2017@gmail.com>
Signed-off-by: Zhiyuan Li <uniartisan2017@gmail.com>
```
  d24cf322
- Separate MLAAttention class from Attention (#25103) · e614ab78
  Naveenraj Kamalakannan authored Oct 08, 2025
```
Signed-off-by: Naveenraj Kamalakannan <therealnaveenkamal@gmail.com>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
```
  e614ab78
08 Oct, 2025 4 commits

[Attention] Register FLASHMLA_SPARSE (#26441) · 2a03f93d
Matthew Bonanni authored Oct 08, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
2a03f93d

[Bugfix][Flashinfer] fix VLLM_USE_TRTLLM_ATTENTION issue for models with diff... · b82f4307

elvischenv authored Oct 09, 2025


[Bugfix][Flashinfer] fix VLLM_USE_TRTLLM_ATTENTION issue for models with diff hyperparameters (#25924)
Signed-off-by: elvischenv <219235043+elvischenv@users.noreply.github.com>

b82f4307

[TPU] Rename tpu_commons to tpu_inference (#26279) · 335b28f7

Utkarsh Sharma authored Oct 08, 2025


Signed-off-by: Utkarsh Sharma <utksharma@google.com>
Co-authored-by: Utkarsh Sharma <utksharma@google.com>
Co-authored-by: Chengji Yao <chengjiyao@google.com>

335b28f7

[Misc] Clean up cruft from previous FlashMLA sparse implementation (#26125) · f80e7866
Lucas Wilkinson authored Oct 07, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
f80e7866

07 Oct, 2025 2 commits
- [Spec Decode] Enable efficient speculative decoding with FlashInfer-MLA (#25984) · 3d1f6761
  Benjamin Chislett authored Oct 07, 2025
```
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
```
  3d1f6761
- [Bugfix] Add missing sink tensor into flash attn cascade attn implementation (#26325) · eb577e46
  Pei-Lun Liao authored Oct 07, 2025
  
  eb577e46
06 Oct, 2025 6 commits
- [Perf] Add decode full-graph support to FlashInfer-MLA backend (#26313) · f77df946
  Benjamin Chislett authored Oct 06, 2025
```
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
```
  f77df946
- [ROCm] Split AITER unified attention into its own backend (#25507) · f231e5bc
  Gregory Shtrasberg authored Oct 06, 2025
```
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
```
  f231e5bc
- [Attention] Remove unused reorder_batch method (#24463) · 4727a8af
  Matthew Bonanni authored Oct 06, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  4727a8af
- [Misc] Clean up unnecessary E501 ignore (#26274) · 43c146ca
  Roger Wang authored Oct 06, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
```
  43c146ca
- Fix per file ruff ignores related to line length (#26262) · 6c046382
  Harry Mellor authored Oct 06, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  6c046382
- [V1] [Hybrid] Some additional clean-up in Mamba2 prefix caching (#26222) · 778f5541
  Thomas Parnell authored Oct 06, 2025
```
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
```
  778f5541
05 Oct, 2025 3 commits
- Fix per file ruff ignores related to typing (#26254) · 1c0c6820
  Harry Mellor authored Oct 05, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  1c0c6820
- Remove all references to `yapf` as it's no longer used (#26251) · 4e256cad
  Harry Mellor authored Oct 05, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  4e256cad
- Convert formatting to use `ruff` instead of `yapf` + `isort` (#26247) · d6953beb
  Harry Mellor authored Oct 05, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  d6953beb
04 Oct, 2025 3 commits
- [CPU] Refine batch reorder of CPU attention backend (#26096) · 5c057e06
  Li, Jiang authored Oct 04, 2025
```
Signed-off-by: jiang1.li <jiang1.li@intel.com>
```
  5c057e06
- Revert "Add batch invariant kernel override for FlashInfer backend [2/n]" (#26220) · 1838cd48
  Cyrus Leung authored Oct 04, 2025
  
  1838cd48
- [V1] [Hybrid] Mamba2 Automatic Prefix Caching (#25752) · ea507c3a
  Stan Wozniak authored Oct 04, 2025
```
Signed-off-by: Stanislaw Wozniak <stw@zurich.ibm.com>
Signed-off-by: Thomas Ortner <boh@zurich.ibm.com>
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
Co-authored-by: Thomas Ortner <boh@zurich.ibm.com>
Co-authored-by: Thomas Parnell <tpa@zurich.ibm.com>
```
  ea507c3a