Commits · cc1dc7ed6d2d901df46847724ad90147872a9d79 · OpenDAS / vllm_cscc

23 Sep, 2025 4 commits

[Core/DBO][2/N] Dual-Batch Overlap add DeepEP High Throughput support and Prefill support (#24845) · cc1dc7ed

Lucas Wilkinson authored Sep 23, 2025


Signed-off-by: Sage Moore <sage@neuralmagic.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>

cc1dc7ed

[V1] Remove V0 code paths for Hybrid models (#25400) · a903669e
Thomas Parnell authored Sep 23, 2025
```
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
```
a903669e
[Bug] Fix Long Context OOM Issue (#25290) · 4741239d
Wentao Ye authored Sep 22, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
4741239d
[Performance] Remove input pads in cutlass_mla and optimize v_proj output handling (#25184) · 0b7bed9c
Alexander Matveev authored Sep 22, 2025
```
Signed-off-by: Alexander Matveev <amatveev@redhat.com>
```
0b7bed9c

22 Sep, 2025 3 commits
- [CLI env var] Add VLLM_FLASH_ATTN_MAX_NUM_SPLITS_FOR_CUDA_GRAPH in env variables (#25274) · cfbee3d0
  Daisy-Ma-coder authored Sep 22, 2025
```
Signed-off-by: qqma <qqma@amazon.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: qqma <qqma@amazon.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
```
  cfbee3d0
- [V1][Attention] Split triton_attn in triton-only and rocm specific backends (#24648) · 175811e3
  Burkhard Ringlein authored Sep 22, 2025
```
Signed-off-by: Burkhard Ringlein <ngl@zurich.ibm.com>
```
  175811e3
- [V0 Deprecation] Remove `MultiModalPlaceholderMap` (#25366) · f92d9526
  Cyrus Leung authored Sep 22, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  f92d9526
21 Sep, 2025 1 commit
- [V1] Add sliding window support to Flex Attention backend (#24089) · cf56cf78
  Isotr0py authored Sep 21, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  cf56cf78
19 Sep, 2025 2 commits
- allow disable flashinfer prefill (#25276) · ee7a66dd
  Lucia Fang authored Sep 19, 2025
```
Signed-off-by: Lu Fang <fanglu@fb.com>
```
  ee7a66dd
- [BugFix] Make FlashInferMetadataBuilder non-blocking (#25040) · b1a63d1b
  nvjullin authored Sep 20, 2025
```
Signed-off-by: Julien Lin <jullin@nvidia.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
```
  b1a63d1b
18 Sep, 2025 4 commits
- [Bug] Fix `returned_lse` not Defined issue (#25106) · 75fb112d
  Wentao Ye authored Sep 18, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
```
  75fb112d
- [ROCm][AITER][Bugfix] Switch AITER to use PIECEWISE_AND_FULL compilation (#25104) · bbdc0f23
  Rohan Potdar authored Sep 18, 2025
```
Signed-off-by: Rohan138 <rohanpotdar138@gmail.com>
```
  bbdc0f23
- [New Model] Support BertForTokenClassification / Named Entity Recognition (NER) task (#24872) · 5f696c33
  wang.yuqi authored Sep 18, 2025
```
Signed-off-by: wang.yuqi <noooop@126.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  5f696c33
- [PERF] Add `conv1d` metadata to GDN attn (#25105) · 072d7e53
  Vadim Gimpelson authored Sep 18, 2025
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
```
  072d7e53
17 Sep, 2025 4 commits
- [Bugfix] Refactor Flashinfer TRTLLM attention kernel selection logic (#24600) · e67a79db
  elvischenv authored Sep 18, 2025
```
Signed-off-by: elvischenv <219235043+elvischenv@users.noreply.github.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
```
  e67a79db
- Aiter mha fp8 fix (#24991) · 1a456c7c
  Douglas Lehr authored Sep 17, 2025
```
Signed-off-by: Doug Lehr <douglehr@amd.com>
Co-authored-by: Doug Lehr <douglehr@amd.com>
```
  1a456c7c
- Remove old cutlass mla (#23961) · 8f3616f4
  Matthew Bonanni authored Sep 17, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  8f3616f4
- [Bugfix][Qwen3-Next] fixes the varlen issue in qwen3-next's MTP implementation. (#24957) · dd6a910a
  Tao He authored Sep 17, 2025
```
Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com>
```
  dd6a910a
16 Sep, 2025 3 commits

fp8 kv cache support fix for torch.compile (#22758) · 3053a22b

Aleksandr Malyshev authored Sep 16, 2025

Signed-off-by: Aleksandr Malyshev <maleksan@amd.com>
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Co-authored-by: Aleksandr Malyshev <maleksan@amd.com>
Co-authored-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Co-authored-by: Gregory Shtrasberg <156009573+gshtras@users.noreply.github.com>

3053a22b

[Core/DBO][1/N] Add Dual-Batch Overlap mechanism to VLLM (#23693) · 56793995

Sage Moore authored Sep 16, 2025


Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Sage Moore <sage@neuralmagic.com>
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Co-authored-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

56793995

[Bug] Fix `is_flashmla_supported` Check Error (#24774) · b42566f4
Wentao Ye authored Sep 15, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
b42566f4

15 Sep, 2025 2 commits
- [Performance] Remove redundant clone() calls in cutlass_mla (#24891) · aae725af
  Alexander Matveev authored Sep 15, 2025
  
  aae725af
- [XPU] Set consistent default KV cache layout (#24745) · 2e41f5ab
  Nicolò Lucchesi authored Sep 15, 2025
```
Signed-off-by: NickLucche <nlucches@redhat.com>
```
  2e41f5ab
12 Sep, 2025 3 commits
- [Qwen3Next] Fixes the cuda graph capture conditions under large batch sizes (#24660) (#24667) · 8226dd56
  Tao He authored Sep 13, 2025
```
Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com>
```
  8226dd56
- [Attention][FlashInfer] Enable FP8 FlashInfer (TRTLLM) MLA decode (#24705) · 7ba32aa6
  Matthew Bonanni authored Sep 12, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
```
  7ba32aa6
- [Kernel] [CPU] refactor `cpu_attn.py:_run_sdpa_forward` for better memory access (#24701) · 7a1c4025
  Ignacio Sica authored Sep 12, 2025
```
Signed-off-by: ignaciosica <mignacio.sica@gmail.com>
```
  7a1c4025
11 Sep, 2025 4 commits
- [flashinfer] [kernel] support for fp8 kv cache for trtllm prefill attention (#24197) · e42af78b
  Xiaozhu Meng authored Sep 11, 2025
```
Signed-off-by: Xiaozhu <mxz297@gmail.com>
```
  e42af78b
- fix some typos (#24616) · e26fef83
  co63oc authored Sep 12, 2025
```
Signed-off-by: co63oc <co63oc@users.noreply.github.com>
```
  e26fef83
- Add the support for the qwen3 next model (a hybrid attention model). (#24526) · e93f4cc9
  Tao He authored Sep 11, 2025
```
Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
```
  e93f4cc9
- [Doc]: fixing doc typos (#24635) · e2b1f863
  Didier Durand authored Sep 11, 2025
```
Signed-off-by: Didier Durand <durand.didier@gmail.com>
```
  e2b1f863
10 Sep, 2025 8 commits

[Kernel] Flashinfer MLA (trtllm-gen) decode kernel integration (#21078) · dcb28a33

Hanjie Qiu authored Sep 10, 2025


Signed-off-by: hjjq <hanjieq@nvidia.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>

dcb28a33

[Perf] Warmup FlashInfer attention during startup (#23439) · fba78565

Michael Goin authored Sep 10, 2025


Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Matthew Bonanni <mbonanni001@gmail.com>

fba78565

[gpt-oss] raise error for flashinfer backend without trtllm (#24482) · b5e383cd
Chen Zhang authored Sep 10, 2025
```
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
```
b5e383cd

[torch.compile][ROCm][V1] Enable attention output FP8 fusion for V1 attention backends (#19767) · 9a161307

Gregory Shtrasberg authored Sep 10, 2025


Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>

9a161307

[v1] Add Whisper model support (encoder-decoder) (#21088) · 37e8182b
Russell Bryant authored Sep 10, 2025
```
Signed-off-by: Russell Bryant <rbryant@redhat.com>
Co-authored-by: NickLucche <nlucches@redhat.com>
```
37e8182b
[Bugfix] Enable FP8 KV cache for FlashInfer and Triton backend on non-sm100 GPUs (#24577) · a0933c3b
Thien Tran authored Sep 11, 2025
```
Signed-off-by: Thien Tran <gau.nernst@yahoo.com.sg>
```
a0933c3b

[Attention] add DCP support for FLASH_ATTN_MLA backend (#24453) · 0ae43dbf

Lucas Wilkinson authored Sep 10, 2025


Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>

0ae43dbf

[Perf] Convert np array to torch tensor to index into block table for attn chunking (#24474) · dc625ea6
Yong Hoon Shin authored Sep 09, 2025
```
Signed-off-by: Yong Hoon Shin <yhshin@meta.com>
```
dc625ea6

09 Sep, 2025 2 commits

[Feature] Disallow FlashMLA on Blackwell (#24521) · 15de5ff9

Wentao Ye authored Sep 09, 2025

Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

15de5ff9

[Flashinfer] Support Flashinfer TRTLLM FP8-qkv BF16/FP16-out Attention Kernel (#23647) · bba1042c
elvischenv authored Sep 09, 2025
```
Signed-off-by: elvischenv <219235043+elvischenv@users.noreply.github.com>
```
bba1042c