Commits · f9a4087182ffcd9404779fcda876f820b3b26d5f · OpenDAS / vllm_cscc

11 Nov, 2025 19 commits
- 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
- Fix Fused MoE LoRA Triton kernel bug (#28450) · 7dbe6d81
  Chaojun Zhang authored Nov 11, 2025
```
Signed-off-by: chaojun-zhang <chaojun.zhang@intel.com>
```
  7dbe6d81
- [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
- [EPLB] Refactor balance_packing to use numpy and optimize GPU-CPU transfers in EPLB (#28369) · 798c7beb
  Sage Moore authored Nov 11, 2025
```
Signed-off-by: Sage Moore <sage@neuralmagic.com>
```
  798c7beb
- [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] Avoid calling KV connector layer APIs when metadata is unset (#28253) · cc079763
  David Ben-David authored Nov 11, 2025
```
Signed-off-by: David Ben-David <davidb@pliops.com>
Co-authored-by: David Ben-David <davidb@pliops.com>
Co-authored-by: Mark McLoughlin <markmc@redhat.com>
```
  cc079763
- [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
- [Frontend] Add sagemaker_standards dynamic lora adapter and stateful session... · bca74e32
  Zuyi Zhao authored Nov 10, 2025
```
[Frontend] Add sagemaker_standards dynamic lora adapter and stateful session management decorators to vLLM OpenAI API server (#27892)
Signed-off-by: Zuyi Zhao <zhaozuy@amazon.com>
Signed-off-by: Shen Teng <sheteng@amazon.com>
Co-authored-by: Shen Teng <sheteng@amazon.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
```
  bca74e32
- [Misc] FlattenLogprobs -> FlatLogprobs (#28335) · 8d706cca
  Zhuohan Li authored Nov 10, 2025
  
  8d706cca
- Only register rocm_aiter_ops if aiter is found (#28428) · f2d9ad06
  Michael Goin authored Nov 10, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  f2d9ad06
- [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
- [CI/Test Fix] Fix CP tests on Blackwell (#28404) · 39029d51
  Lucas Wilkinson authored Nov 10, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
```
  39029d51
- [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 21 commits

[Bugfix] Ensure calculated KV scales are applied in attention. (#27232) · a5a790ee
Adrian Abeyta authored Nov 10, 2025
```
Signed-off-by: adabeyta <aabeyta@redhat.com>
```
a5a790ee
[Perf] Move gc.freeze logic from EngineCoreProc to EngineCore for better coverage (#27896) · b30372cb
Jialin Ouyang authored Nov 10, 2025
```
Signed-off-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>
```
b30372cb

[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
[CI] Fix Plugin Tests Tests (#28413) · 30700b1c
Robert Shaw authored Nov 10, 2025
```
Signed-off-by: Robert Shaw <robertgshaw2@gmail.com>
```
30700b1c
[Frontend][2/n] remove empty content from _parse_tool_calls_from_content (#28331) · 4b94ed8f
Andrew Xia authored Nov 10, 2025
```
Signed-off-by: Andrew Xia <axia@fb.com>
Co-authored-by: Andrew Xia <axia@fb.com>
```
4b94ed8f
[BugFix] Fix DeepGEMM over-allocating workspace (#28254) · 6dec9f61
Lucas Wilkinson authored Nov 10, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
6dec9f61
[Misc] Add more scoping for improved trace (#28329) · bf6a3d0f
Wei Wei authored Nov 10, 2025
```
Signed-off-by: Wei Wei <wwei6@meta.com>
```
bf6a3d0f

[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

[Bugfix] Fix llguidance backend, rollback when EOS was encountered (#25905) · 6d54336a

Rémi Delacourt authored Nov 10, 2025


Signed-off-by: Rémi Delacourt <remi@mistral.ai>
Signed-off-by: remi <remi@mistral.ai>
Co-authored-by: Russell Bryant <rbryant@redhat.com>

6d54336a

[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

[Kernel] Optimization of the mm_k operator. (#28280) · 40e2eeeb
caozuoba authored Nov 11, 2025
```
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
```
40e2eeeb

[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
[Metrics] Refactor LoRA state tracking (#26801) · 6f7de33b
Mark McLoughlin authored Nov 10, 2025
```
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
```
6f7de33b
[V0 deprecation] Remove no longer used `get_metadata_cls` (#28370) · e8697faf
Lucas Wilkinson authored Nov 10, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
e8697faf

[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