Commits · 700a5ad6c616358f42db7d9b55e8bc9caa140ca5 · OpenDAS / vllm_cscc · GitLab

18 Dec, 2025 14 commits
- [MM Encoder]: Migrate legacy ViT `MultiHeadAttention` to new... · 700a5ad6
  Isotr0py authored Dec 19, 2025
```
[MM Encoder]: Migrate legacy ViT `MultiHeadAttention` to new `MMEncoderAttention` interface (#30684)
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  700a5ad6
- fix fp8 online quantization streaming with tp > 1 (#30900) · f4ee2c3d
  Vasiliy Kuznetsov authored Dec 18, 2025
```
Signed-off-by: vasiliy <vasiliy@fb.com>
```
  f4ee2c3d
- [LoRA] Set default MXFP4 LoRA backend to Marlin (#30598) · 9a5e9652
  Xin Yang authored Dec 18, 2025
```
Signed-off-by: Xin Yang <xyangx@amazon.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
```
  9a5e9652
- adds jais 2 support (#30188) · 28d15ab5
  sarathc-cerebras authored Dec 18, 2025
```
Signed-off-by: sarathc-cerebras <sarath.chandran@cerebras.net>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
```
  28d15ab5
- [Bug] Fix batch invariant in torch 2.10 (#30907) · 66287582
  Wentao Ye authored Dec 18, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
```
  66287582
- Filter safetensors files to download if .safetensors.index.json exists (#30537) · 100f93d2
  Michael Goin authored Dec 18, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  100f93d2
- [moe] Use enable_chunking func (to support disabling chunking) (#29935) · 8372be28
  Ming Yang authored Dec 18, 2025
```
Signed-off-by: Ming Yang <minos.future@gmail.com>
```
  8372be28
- [CPU] Refactor CPU fused MOE (#30531) · e3ab93c8
  Li, Jiang authored Dec 18, 2025
```
Signed-off-by: jiang1.li <jiang1.li@intel.com>
```
  e3ab93c8
- [PERF] Qwen3-next. Add fp8 cutlass MoE tuned configs. `chmod -x *MI308X.json` (#29553) · 717ac33d
  Vadim Gimpelson authored Dec 18, 2025
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
```
  717ac33d
- [refactor] Add prefix support to embed_tokens in DeepSeek MTP (#30788) · b166ef20
  zzhxxx authored Dec 18, 2025
```
Signed-off-by: zzhx1 <zzh_201018@outlook.com>
```
  b166ef20
- [UX] Reduce DeepGEMM warmup log output to single progress bar (#30903) · 4a8412f7
  Matthew Bonanni authored Dec 17, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  4a8412f7
- [Quantization] Support Quark int4-fp8 w4a8 for MoE (#30071) · 0c738b58
  Bowen Bao authored Dec 17, 2025
```
Signed-off-by: Bowen Bao <bowenbao@amd.com>
```
  0c738b58
- [Chore] Remove v0 dead code for Qwen2.5-omni (#30883) · 6fe58876
  Isotr0py authored Dec 18, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  6fe58876
- [v1] Add PrefixLM support to TritonAttention backend (#30386) · 74a1ac38
  Isotr0py authored Dec 18, 2025
  
  74a1ac38
17 Dec, 2025 11 commits
- [BugFix] Workspace allocation during profile run : DeepEPHighThroughput + DeepGEMM (#30899) · e3fc374a
  Varun Sundar Rabindranath authored Dec 17, 2025
  
  e3fc374a
- 2.9.1 PyTorch release update (#28495) · e06d0bf0
  Andrey Talman authored Dec 17, 2025
  
  e06d0bf0
- [Model] Gemma3: Support untied word embeddings (#30827) · 196cdc32
  KimHyemin authored Dec 18, 2025
```
Signed-off-by: www-spam <panmahm@naver.com>
```
  196cdc32
- [Bugfix] deepseek-V3.2 self.weights_proj has no bias (#30841) · 84896fda
  baoqian426 authored Dec 17, 2025
```
Signed-off-by: baoqian <1354987947@qq.com>
Signed-off-by: baoqian426 <1354987947@qq.com>
```
  84896fda
- [Bug] Fix AttributeError: 'ColumnParallelLinear' object has no attribute... · f284d7bd
  Wentao Ye authored Dec 17, 2025
```
[Bug] Fix AttributeError: 'ColumnParallelLinear' object has no attribute `weight_scale_inv` (#30823)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  f284d7bd
- CustomOp: grouped topk (#29575) · 3b1d440e
  Xinyu Chen authored Dec 17, 2025
```
Signed-off-by: Xinyu Chen <xinyu1.chen@intel.com>
```
  3b1d440e
- [Mamba] Removed disable cascade attn in MambaModelConfig (#30712) · a9e15c21
  Asaf Joseph Gardin authored Dec 17, 2025
```
Signed-off-by: asafg <39553475+Josephasafg@users.noreply.github.com>
```
  a9e15c21
- [XPU] fix broken fp8 online quantization for XPU platform (#30831) · 4f735bab
  Yan Ma authored Dec 17, 2025
```
Signed-off-by: Yan Ma <yan.ma@intel.com>
```
  4f735bab
- [Bugfix][CPU] Fix CPU backend ROPE dispatch for VL models (#30829) · 0cd53536
  Li, Jiang authored Dec 17, 2025
```
Signed-off-by: jiang1.li <jiang1.li@intel.com>
Signed-off-by: Li, Jiang <bigpyj64@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
```
  0cd53536
- Update note comment for flashinfer attention warmup (#30711) · d4d27517
  Michael Goin authored Dec 17, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  d4d27517
- Fix nemotron_nas intermediate_size computation (#30795) · f5db6385
  Grzegorz K. Karch authored Dec 17, 2025
```
Signed-off-by: Grzegorz Karch <gkarch@nvidia.com>
```
  f5db6385
16 Dec, 2025 10 commits
- [Kernel][Quantization][MoE] add marlin kernel support for turing (sm75) (#29901) · ce96857f
  Jinzhen Lin authored Dec 17, 2025
```
Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
```
  ce96857f
- [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 5 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