Commits · eaa82a709a963ab744647a701fe267223ed7b02b · OpenDAS / vllm_cscc

16 Dec, 2025 9 commits
- [Core][MM] Optimize encoder cache manager by operating with embeddings only (#30475) · f5f51e59
  Roger Wang authored Dec 16, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Sun Kim <sunytokki@gmail.com>
```
  f5f51e59
- [Perf] Do FP4 quant before All gather on flashinfer trtllmgen MOE (#30014) · 254a7f8f
  jiahanc authored Dec 16, 2025
```
Signed-off-by: jiahanc <173873397+jiahanc@users.noreply.github.com>
```
  254a7f8f
- [CI] Generalize gsm8k test args and add Qwen3-Next MTP B200 test (#30723) · 10ee1c64
  Michael Goin authored Dec 16, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  10ee1c64
- Remove `head_mask` from Ultravox and Swin (#30764) · 0b0acc75
  Harry Mellor authored Dec 16, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  0b0acc75
- [TRTLLM] Remove the MoE GEMM weight name change (#30713) · ce12b407
  Ming Yang authored Dec 16, 2025
```
Signed-off-by: Ming Yang <minos.future@gmail.com>
```
  ce12b407
- [Feat] Enable eplb with default all2all backend (#30559) · 59bd5f6a
  Wentao Ye authored Dec 16, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  59bd5f6a
- Don'e assume `position_embedding_type` will be present for BERT and RoBERTa models (#30770) · 6f15ac5d
  Harry Mellor authored Dec 16, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  6f15ac5d
- [Bugfix] Fix broken ViT attention selection for Blackwell device (#30731) · e94384bb
  Isotr0py authored Dec 16, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  e94384bb
- [CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873) · 3bd9c491
  Shanshan Shen authored Dec 16, 2025
```
Signed-off-by: shen-shanshan <467638484@qq.com>
Co-authored-by: gcanlin <canlinguosdu@gmail.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
```
  3bd9c491
15 Dec, 2025 9 commits

Update batch invariant to use attention config (#30704) · 60dbf7d8

Matthew Bonanni authored Dec 15, 2025


Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

60dbf7d8

[MoE][Refactor 1/N] Separate Online Quantization (#30627) · d0502b49
Robert Shaw authored Dec 15, 2025
```
Signed-off-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>
```
d0502b49
[Bugfix] Fix multimodal configuration for Qwen3VL MOE model (#30670) · 3f175f18
Max Hu authored Dec 15, 2025
```
Signed-off-by: Max Hu <hyoung2991@gmail.com>
```
3f175f18
[BugFix] Add embed_input_ids method to make QWenLMHeadModel a vllm model (#30674) · e4806d97
duke authored Dec 15, 2025
```
Signed-off-by: root <iwzbi@zju.edu.cn>
Co-authored-by: root <iwzbi@zju.edu.cn>
```
e4806d97
[Model] Automatic conversion of TokenClassification model (#30666) · 4429d934
wang.yuqi authored Dec 15, 2025
```
Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
```
4429d934

[New Model] BAGEL support (AR only) (#28439) · 1adeb3b8

汪志鹏 authored Dec 15, 2025


Signed-off-by: princepride <wangzhipeng628@gmail.com>
Signed-off-by: 汪志鹏 <wangzhipeng628@gmail.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>

1adeb3b8

[Feat] Refactor for `parallel_config` in `FusedMoEModularKernel` (#30282) · 3778673e

Wentao Ye authored Dec 14, 2025


Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

3778673e

[CustomOp][MM] Extract MMEncoderAttention as CustomOp and replace the backend... · 87b4d155

Shanshan Shen authored Dec 15, 2025


[CustomOp][MM] Extract MMEncoderAttention as CustomOp and replace the backend of QwenVisionAttention with it. (#30125)
Signed-off-by: shen-shanshan <467638484@qq.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: tjtanaa <tunjian.tan@embeddedllm.com>

87b4d155

[CustomOp] Support object-level enable for CustomOp (#30547) · 738648fb
Shanshan Shen authored Dec 15, 2025
```
Signed-off-by: shen-shanshan <467638484@qq.com>
```
738648fb

14 Dec, 2025 7 commits

[Feature]Add EVS (Efficient Video Sampling) Support for Qwen3-VL (#29752) · ae88aada
ZiTian Zhao authored Dec 14, 2025
```
Signed-off-by: zitian.zhao <zitian.zhao@tencentmusic.com>
Co-authored-by: deitxfge <huhaibo1990@126.com>
```
ae88aada
[Bugfix] Revert Qwen2-VL part of change in #28271 (#30542) · 48b8456f
zifeitong authored Dec 14, 2025
```
Signed-off-by: Zifei Tong <zifeitong@gmail.com>
```
48b8456f
[Bugfix] fix _get_quant_method of FusedMoE for deepseekV3.2 on non-NV… (#30057) · 6ecc1e41
tjp_zju authored Dec 14, 2025
```
Signed-off-by: tjp_zju <tanjianpingzju1990@gmail.com>
```
6ecc1e41

Nvidia ModelOpt workaround for issue 28072 (#30164) · 0bb0bae4

Shengliang Xu authored Dec 14, 2025


Signed-off-by: Shengliang Xu <shengliangx@nvidia.com>
Co-authored-by: Pavani Majety <pmajety@nvidia.com>

0bb0bae4

[torch.compile] Add encoder tag for compilation (#30489) · 3224ea99
Ilya Markov authored Dec 14, 2025
```
Signed-off-by: ilmarkov <markovilya197@gmail.com>
```
3224ea99

Add AudioFlamingo3 model support (#30539) · 3a20450d

Lasha Koroshinadze authored Dec 14, 2025


Signed-off-by: Lasha <26011196+lashahub@users.noreply.github.com>
Signed-off-by: Lasha Koroshinadze <26011196+lashahub@users.noreply.github.com>
Co-authored-by: Isotr0py <2037008807@qq.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>

3a20450d

[Doc]: fixing typos in various files (#30540) · 1a55cfaf

Didier Durand authored Dec 14, 2025

Signed-off-by: Didier Durand <durand.didier@gmail.com>
Signed-off-by: Didier Durand <2927957+didier-durand@users.noreply.github.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>

1a55cfaf

13 Dec, 2025 5 commits

[Logs] Optimize startup logs 4 (#29903) · 6e78ed6b

Wentao Ye authored Dec 13, 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>

6e78ed6b

[Bugfix] Qwen3-next with --hf-overrides \{\"num_hidden_layers\":8\} (#30433) · ace34e37
Chen Zhang authored Dec 13, 2025
```
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
```
ace34e37
[Chore] Adjust tokenizer import to avoid circular imports (#30601) · 64251f48
Cyrus Leung authored Dec 13, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
64251f48
[Misc][Quantization] Clarify the intent of GGUF `FusedMoE` weight materialization (#30310) · fdc135d7
Tsukasa OI authored Dec 13, 2025
```
Signed-off-by: Tsukasa OI <floss_llm@irq.a4lg.com>
```
fdc135d7

[Feature] Add SM103 (Blackwell Ultra) Support to vLLM (#30484) · 4fa7ce46

Roberto L. Castro authored Dec 13, 2025

Signed-off-by: LopezCastroRoberto <robertol.c510@gmail.com>
Signed-off-by: Roberto L. Castro <38211239+LopezCastroRoberto@users.noreply.github.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>

4fa7ce46

12 Dec, 2025 10 commits

[CI/Build][Kernel][BugFix][AMD] Fix per_token_group_quant_fp8 to use correct... · 08f8a562

rasmith authored Dec 12, 2025


[CI/Build][Kernel][BugFix][AMD] Fix per_token_group_quant_fp8 to use correct fp8 min/max values and update atol/rtol in test_quantfp8_group_functionality  (#30292)
Signed-off-by: Randall Smith <ransmith@amd.com>
Co-authored-by: Randall Smith <ransmith@amd.com>

08f8a562

[MoE-FP8-modelopt] Add FlashInfer alignment padding for intermediate dimensions (#29748) · 13618626

danielafrimi authored Dec 12, 2025


Signed-off-by: Daniel Afrimi <dafrimi@pool0-00589.cm.cluster>
Signed-off-by: dafrimi <dafrimi@nvidia.com>
Co-authored-by: Daniel Afrimi <dafrimi@pool0-00589.cm.cluster>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>

13618626

[Perf] Set split_k to 1 for triton_kernels (#30528) · 1f19d8f8
Xin Yang authored Dec 12, 2025
```
Signed-off-by: Xin Yang <xyangx@amazon.com>
```
1f19d8f8
[ROCm] Enable Triton ScaledMM fallback + kernel selection fix (#26668) · cd7740ac
shivampr authored Dec 12, 2025
```
Signed-off-by: Shivam <shivampr.dev@gmail.com>
Signed-off-by: Shivam <shivamprasad91@gmail.com>
```
cd7740ac

fix(gguf): Disable bfloat16 for GGUF on blackwell device (#30408) · dc13c99e

Christina Norman authored Dec 12, 2025


Signed-off-by: Christina <truffle@gmail.com>
Signed-off-by: Isotr0py <2037008807@qq.com>
Signed-off-by: Christina Norman <christina@example.com>
Co-authored-by: Isotr0py <isotr0py@users.noreply.github.com>
Co-authored-by: Isotr0py <2037008807@qq.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

dc13c99e

[Attention] Use sparse prefill kernel for fp8 kv-cache in DeepSeek-v3.2 (#27532) · 3e41992f
Lucas Wilkinson authored Dec 12, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
3e41992f
[Bugfix] Schedule failure due to wrong get_image_size_with_most_features (#29692) · f90319d5
Jaehwang Jung authored Dec 12, 2025

f90319d5

[Bugfix][Model] Fix Afmoe rope_parameters issue (#30505) · 9f2fc16a

Michael Goin authored Dec 11, 2025


Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

9f2fc16a

gptq marlin quantization support for fused moe with lora (#30254) · 6a6fc41c
Bhanu Prakash Voutharoja authored Dec 12, 2025
```
Signed-off-by: Bhanu068 <voutharoja.bhanu06@gmail.com>
```
6a6fc41c
[fix] fix SM check for Flashinfer TRTLLM MOE (#30314) · 0ab23c2b
jiahanc authored Dec 11, 2025
```
Signed-off-by: jiahanc <173873397+jiahanc@users.noreply.github.com>
```
0ab23c2b