Commits · 05576df85c5274ee3045d90b0779d4adeecc09b9 · OpenDAS / vllm_cscc

11 Nov, 2025 14 commits
- [ROCm][Quantization] extend AMD Quark to support mixed-precision quantized model (#24239) · 05576df8
  xuebwang-amd authored Nov 12, 2025
```
Signed-off-by: xuebwang-amd <xuebwang@amd.com>
Co-authored-by: fxmarty-amd <felmarty@amd.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
```
  05576df8
- [Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165) · 68c09efc
  zhrrr authored Nov 12, 2025
```
Signed-off-by: zhuhaoran <zhuhaoran.zhr@alibaba-inc.com>
```
  68c09efc
- Remove weight_scale.T special case for SM90 Block FP8 CUTLASS kernel (#28431) · f9a40871
  Michael Goin authored Nov 11, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  f9a40871
- [BugFix] Fix RuntimeError in PixtralHFAttention on CPU/XPU (#28444) · b8860680
  Fanli Lin authored Nov 11, 2025
```
Signed-off-by: Lin, Fanli <fanli.lin@intel.com>
```
  b8860680
- [Kernels] Split up fused_moe/layer.py, isolate more modular kernel code (#28064) · a1448b4b
  bnellnm authored Nov 11, 2025
  
  a1448b4b
- [Model] Pass `mm_features` directly into `get_mrope_input_positions` (#28399) · afffd3cc
  Cyrus Leung authored Nov 11, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  afffd3cc
- [Attention] Refactor CUDA attention backend selection logic (#24794) · b30dfa03
  Matthew Bonanni authored Nov 11, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
```
  b30dfa03
- [Model][Qwen3VL] Slighly speedup `fast_pos_embed_interpolate` (#28434) · 9973e6e0
  Lukas Geiger authored Nov 11, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  9973e6e0
- [BugFix] 'DeepseekV2Config' object has no attribute 'use_mla'` (#28387) · c7991269
  Fanli Lin authored Nov 11, 2025
```
Signed-off-by: Lin, Fanli <fanli.lin@intel.com>
```
  c7991269
- [Bugfix] fix qwen3-next crash (#28202) · f0359fff
  Jiangyun Zhu authored Nov 11, 2025
```
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
```
  f0359fff
- [Bugfix] Fix max image size for PaddleOCR-VL (#28442) · 4fd4b743
  Roger Wang authored Nov 11, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
```
  4fd4b743
- [Bugfix] Fix Stream Sync for Shared Expert Overlap (#28430) · e605e8e3
  Robert Shaw authored Nov 11, 2025
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
Signed-off-by: Robert Shaw <robertgshaw2@gmail.com>
Co-authored-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
```
  e605e8e3
- [Feature] Add env var `VLLM_MOE_USE_DEEP_GEMM` (#28422) · de540c03
  Wentao Ye authored Nov 10, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  de540c03
- [Feature] Refactor batch invariant fp8 DeepGEMM (#27606) · 35d801f1
  Wentao Ye authored Nov 10, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  35d801f1
10 Nov, 2025 12 commits

[PERF] Allreduce fusion. Support torch native matching. Tuning of the thresholds (#24248) · d17ecc6b

Ilya Markov authored Nov 11, 2025

Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Co-authored-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

d17ecc6b

[ROCm] Add missing gemm_a8w8_blockscale import (#28378) · 02114356
Yong Hoon Shin authored Nov 10, 2025
```
Signed-off-by: Yong Hoon Shin <yhshin@meta.com>
```
02114356
[BugFix] Fix DeepGEMM over-allocating workspace (#28254) · 6dec9f61
Lucas Wilkinson authored Nov 10, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
6dec9f61

[Bugfix][EPLB] Disabled shared expert overlap when EPLB is enabled (#28377) · 40d33264

Sage Moore authored Nov 10, 2025

Signed-off-by: Sage Moore <sage@neuralmagic.com>
Signed-off-by: Sage Moore <sagemoore@utexas.edu>
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>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>

40d33264

[Performance] Support FP8 flashinfer TRTLLM MOE on Qwen3 and Qwen-3next (#27492) · 34553b9d
jiahanc authored Nov 10, 2025
```
Signed-off-by: jiahanc <173873397+jiahanc@users.noreply.github.com>
```
34553b9d

[Bugfix] Fix persistent_masked_m_silu_mul_quant tests (#28366) · b039bfda

Varun Sundar Rabindranath authored Nov 10, 2025


Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>

b039bfda

[V0 Deprecation] Remove unused `context_len` and `seq_len` from M-RoPE (#28395) · d0e186c1
Cyrus Leung authored Nov 11, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
d0e186c1

[RFC][ROCm][AITER] Keep all AITER kernels in `_aiter_ops` class like... · f080a835

vllmellm authored Nov 10, 2025


[RFC][ROCm][AITER] Keep all AITER kernels in `_aiter_ops` class like `_custom_ops` and `_ipex_ops` (#24490)
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>

f080a835

[Rocm][fused_moe][fp4] view weight to torch.float4_e2m1fn_x2 when running... · b06b9470

zejunchen-zejun authored Nov 10, 2025


[Rocm][fused_moe][fp4] view weight to torch.float4_e2m1fn_x2 when running aiter fused moe for fp4 model (#27474)
Signed-off-by: zejunchen-zejun <zejun.chen@amd.com>

b06b9470

[Fix] optimize visual token mask with caching and multi-token support (#28374) · 912744d0

Ferrebo authored Nov 10, 2025

Signed-off-by: Ferrebo <itachi971009@gmail.com>
Signed-off-by: kebo01 <kebo01@baidu.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

912744d0

[bugfix] fix siglip batch text output error (#28365) · 15be507c
Yu Jiaqi authored Nov 10, 2025
```
Signed-off-by: piood <2477084691@qq.com>
```
15be507c

[Hardware][AMD][Model] Add Triton MoE tuning support and optimized configs for... · 03fa4d3f

Xiake Sun authored Nov 10, 2025


[Hardware][AMD][Model] Add Triton MoE tuning support and optimized configs for Qwen3 omni for MI308X (#28373)
Signed-off-by: Xiake Sun <xiake.sun@amd.com>
Signed-off-by: Xiake Sun <xisun@amd.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

03fa4d3f

09 Nov, 2025 3 commits
- [Kernel] Fix fused_gdn_gating (#28343) · c4768dcf
  Jiangyun Zhu authored Nov 10, 2025
```
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
```
  c4768dcf
- [Misc] Add some comments in qwen3-next (#28267) · 7ae5a5fb
  Jiangyun Zhu authored Nov 09, 2025
```
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
```
  7ae5a5fb
- [ROCm] Add env to enable/disable aiter triton gemm (#28321) · de2b7830
  Yong Hoon Shin authored Nov 08, 2025
```
Signed-off-by: Yong Hoon Shin <yhshin@meta.com>
```
  de2b7830
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 6 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