Commits · e5e9067e61600eedd4e75bd1c512ec52872916aa · OpenDAS / vllm_cscc

08 Nov, 2025 5 commits

[Performance][gpt-oss] Revert gpt-oss max cudagraph size to 1024 (#28345) · 404d7a9d
Mohammad Miadh Angkad authored Nov 09, 2025
```
Signed-off-by: Mohammad Miadh Angkad <MAngkad.BSDSBA2027@aim.edu>
```
404d7a9d
[Bugfix] Update device name for H200 detection (#28349) · 26990d25
Robert Shaw authored Nov 08, 2025
```
Signed-off-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
```
26990d25

[Model] Consolidate Deepseek-MoE implementation with DeepSeek-v2 (#28101) · 934a9c3b

Isotr0py authored Nov 08, 2025


Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>

934a9c3b

[Perf][DeepSeek] Add sigmoid+bias fusion to fused_grouped_topk from TRTLLM (#28124) · 08525276

Michael Goin authored Nov 08, 2025

Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>

08525276

[XPU] Enable Expert parallel for MoE models (#28263) · 1aaecda0
Kunshang Ji authored Nov 08, 2025
```
Signed-off-by: Yan Ma <yan.ma@intel.com>
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
```
1aaecda0

07 Nov, 2025 7 commits
- [Bugfix] Use latency MOE backend as default for Flashinfer and other misc fixes (#27439) · 72b1c2ae
  Pavani Majety authored Nov 07, 2025
```
Signed-off-by: Pavani Majety <pmajety@nvidia.com>
```
  72b1c2ae
- [Core][MM] Add mechanism to configure multimodal fields which should stay on CPU (#28168) · e0919f33
  Lukas Geiger authored Nov 07, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  e0919f33
- [fix] Revert "fixing mm placeholder replacement issue with gemma3" (#28285) · 8e19d470
  Kevin H. Luu authored Nov 07, 2025
```
Signed-off-by: Kevin H. Luu <khluu000@gmail.com>
```
  8e19d470
- [Misc][Model][Refactor] Pass the prefix into Linear layers (#28259) · 1958bda9
  Mengqing Cao authored Nov 07, 2025
```
Signed-off-by: MengqingCao <cmq0113@163.com>
```
  1958bda9
- [amd][gptoss] Perf gain because of block alignment (#28024) · 11fd69dd
  smit kadvani authored Nov 06, 2025
```
Signed-off-by: Smit Kadvani <smit.kadvani@gmail.com>
Co-authored-by: Smit Shaileshbhai Kadvani <kadvani@meta.com>
```
  11fd69dd
- Fix issues from #28242 (#28257) · c0a4b95d
  Harry Mellor authored Nov 06, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  c0a4b95d
- [Multimodal][torch.compile] Add compilation config field for turning off ViT/MM compile (#28242) · 4bf56c79
  Lucas Kabela authored Nov 06, 2025
```
Signed-off-by: Lucas Kabela <lucaskabela@meta.com>
```
  4bf56c79
06 Nov, 2025 9 commits

[ROCm] triton fp8 kernel (#27058) · 449de900

Aleksandr Malyshev authored Nov 06, 2025

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

449de900

Add llama 4 scaling support (#28145) · 7a8375f8
Julien Denize authored Nov 06, 2025
```
Signed-off-by: Julien Denize <julien.denize@mistral.ai>
```
7a8375f8
[Kernel][Model] Tune fused_moe Triton configs for MiniMax-M2 on H100 (#28200) · 0370679c
Eric Yue authored Nov 06, 2025
```
Signed-off-by: minatoaquaMK2 <jiacheng.yue@foxmail.com>
```
0370679c
[CPU]Improve cpu fused moe perf (#27244) · c757a15f
xiangze-arm authored Nov 06, 2025
```
Signed-off-by: Zhang Xiangze <Xiangze.Zhang@arm.com>
```
c757a15f

Fix hard-coded parameter name in gemma3n.py (#27946) · 201dc98a

Seungduk Kim authored Nov 06, 2025

Signed-off-by: Seungduk Kim <seungduk.kim@yanolja.com>
Signed-off-by: Biswa Panda <biswa.panda@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Biswa Panda <biswa.panda@gmail.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>

201dc98a

[flashinfer] fix FI all2all with FI cutlass moe (#28166) · e31946f8
Xiaozhu Meng authored Nov 05, 2025
```
Signed-off-by: Xiaozhu <mxz297@gmail.com>
```
e31946f8
[Chore] Clean up deepseek v2/v3 config copy (#28055) · 43ecd0a9
Isotr0py authored Nov 06, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
43ecd0a9

[Feature] Enable TP + EP `shared_experts` overlap with router, 3.7% E2E... · d71af5f5

Wentao Ye authored Nov 05, 2025


[Feature] Enable TP + EP `shared_experts` overlap with router, 3.7% E2E performance improvement (#28164)
Signed-off-by: yewentao256 <zhyanwentao@126.com>

d71af5f5

[PERF] Decouple projections from GDN custom op. Attempt 2 (#28083) · b6a248bd
Vadim Gimpelson authored Nov 06, 2025
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
```
b6a248bd

05 Nov, 2025 13 commits
- [Bugfix] Fix Qwen3-Reranker-8B load (#28117) · 802748bd
  wang.yuqi authored Nov 06, 2025
```
Signed-off-by: wang.yuqi <noooop@126.com>
```
  802748bd
- [Feature] Extend batch invariant torch.compile to B200 (#27856) · faedbb4d
  Paul Zhang authored Nov 05, 2025
```
Signed-off-by: PaulZhang12 <paulzhan@fb.com>
```
  faedbb4d
- [FlashInfer] Avoid FlashInfer block_size 16 + head_size 256 on blackwell (#27994) · c765f0b4
  Chen Zhang authored Nov 05, 2025
```
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
```
  c765f0b4
- [Kernel] Fuse computation of g and beta for Gated Delta Net (#28095) · c18f88c6
  Jiangyun Zhu authored Nov 06, 2025
```
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
```
  c18f88c6
- [Bugfix] Validate custom logits processor xargs for online serving (#27560) · 3f5a4b64
  Isotr0py authored Nov 06, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  3f5a4b64
- [BugFix] Support EP/DP + EPLB with MTP (#25311) · e50c4546
  Ilya Markov authored Nov 05, 2025
```
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
```
  e50c4546
- [XPU] Enable custom routing functions in IPEX for Llama4 (#28004) · 6e97eccf
  Frost Mitchell authored Nov 05, 2025
```
Signed-off-by: frost-intel <frost.mitchell@intel.com>
```
  6e97eccf
- Bugfix: Cutlass FP8 FusedMoE bad scaling factors (#27255) · 6b7a8118
  amirkl94 authored Nov 05, 2025
```
Signed-off-by: Amir Klein <203507526+amirkl94@users.noreply.github.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
```
  6b7a8118
- [Model, Core] Support Granite Speech & LoRA for STT (#24455) · b7cbc254
  Alex Brooks authored Nov 05, 2025
  
  b7cbc254
- [Bugfix] Fix encoder-only model support for transformers backend (#28021) · 0ff05e37
  Isotr0py authored Nov 05, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  0ff05e37
- [V0 deprecation] Remove VLLM_USE_V1 usage in most modules (#27955) · 428bc7bf
  wangxiyuan authored Nov 05, 2025
```
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
```
  428bc7bf
- [XPU] Add gpt-oss model support for Intel GPU (#27786) · 18b39828
  Kunshang Ji authored Nov 05, 2025
```
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
```
  18b39828
- [Qwen3-Next] MOE configs for A100-SXM4-80GB TP4 TP8 (#27740) · 4ea62b77
  tou authored Nov 05, 2025
  
  4ea62b77
04 Nov, 2025 6 commits
- Revert "[PERF] Decouple projections from GDN custom op" (#28080) · d4e547bb
  Vadim Gimpelson authored Nov 05, 2025
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
```
  d4e547bb
- [ROCm] gemm_a16w16 upstreaming (#26969) · 2d977a7a
  Aleksandr Malyshev authored Nov 04, 2025
```
Signed-off-by: Aleksandr Malyshev <maleksan@amd.com>
Co-authored-by: Aleksandr Malyshev <maleksan@amd.com>
```
  2d977a7a
- [model] Add support for openPangu_Ultra_MoE (#27521) · 05cae69f
  yt0428 authored Nov 05, 2025
```
Signed-off-by: yuantao <2422264527@qq.com>
Signed-off-by: yt0428 <51468697+yt0428@users.noreply.github.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
```
  05cae69f
- [PERF] Decouple projections from GDN custom op (#27512) · 5fd8f02e
  Vadim Gimpelson authored Nov 04, 2025
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
```
  5fd8f02e
- [Kernels] Isolate modular kernel code from FusedMoEMethodBase subclasses. (#27123) · 938772af
  bnellnm authored Nov 04, 2025
  
  938772af
- [Model] add optimal triton fused moe configs for NemotronH MoE (#27967) · e4ee6586
  tomeras91 authored Nov 04, 2025
```
Signed-off-by: Tomer Asida <57313761+tomeras91@users.noreply.github.com>
```
  e4ee6586