Commits · 7afe0faab1eb2ab84cda5cab29b24046e516f7b8 · OpenDAS / vllm_cscc · GitLab

13 Mar, 2026 6 commits
- [Frontend][Core] Re-add shutdown timeout - allowing in-flight requests to finish (#36666) · 7afe0faa
  Mark McLoughlin authored Mar 13, 2026
```
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Nick Hill <nickhill123@gmail.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Co-authored-by: Nick Hill <nickhill123@gmail.com>
```
  7afe0faa
- [2/N] Elastic EP Milestone 2: Integrating NIXL-EP (#35627) · d5af196c
  Itay Alroy authored Mar 13, 2026
```
Signed-off-by: Itay Alroy <ialroy@nvidia.com>
Co-authored-by: Yongji Wu <wuyongji317@gmail.com>
Co-authored-by: Ron Tourgeman <rtourgeman@nvidia.com>
```
  d5af196c
- [kv_offload+HMA][1/N]: Support multiple KV groups in OffloadingSpec (#36610) · cfaf4668
  Or Ozeri authored Mar 13, 2026
```
Signed-off-by: Or Ozeri <oro@il.ibm.com>
```
  cfaf4668
- Enable RoPE+KV cache fusion for ROCm AITER FA (non-shuffle layout) (#35786) · a4ad9db5
  Rohan Potdar authored Mar 13, 2026
```
Signed-off-by: Rohan138 <rohanpotdar138@gmail.com>
```
  a4ad9db5
- fix(kv-cache): increase hybrid attention grouping threshold from 1.25 to 1.5 (#36684) · 891c60dc
  jaime campos salas authored Mar 12, 2026
```
Signed-off-by: Jaime Campos Salas <jaime.campos.salas@gmail.com>
```
  891c60dc
- [Model Runner V2] Some code simplification (#36929) · cd32d6f5
  Nick Hill authored Mar 12, 2026
```
Signed-off-by: Nick Hill <nickhill123@gmail.com>
```
  cd32d6f5
12 Mar, 2026 6 commits
- [Attention] Use FA4 for MLA prefill (#34732) · f444c05c
  Matthew Bonanni authored Mar 12, 2026
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  f444c05c
- [Kernel] Add FP8 KV cache support to Triton MLA decode attention (#34597) · a1257fd1
  grimulkan authored Mar 12, 2026
```
Signed-off-by: grimulkan <grimulkan@gmail.com>
```
  a1257fd1
- [Hardware] Replace torch.cuda.device_count/current_device/set_device API (#36145) · 53ec16a7
  Kunshang Ji authored Mar 12, 2026
```
Signed-off-by: Kunshang Ji <jikunshang95@gmail.com>
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
```
  53ec16a7
- [BugFix] Fix multiple/duplicate stdout prefixes (#36822) · 36735fd7
  Nick Hill authored Mar 11, 2026
```
Signed-off-by: Nick Hill <nickhill123@gmail.com>
```
  36735fd7
- [Model Runner V2] Do not initialize sampler for non-last PP ranks (#36824) · 2f8b4ce0
  Woosuk Kwon authored Mar 11, 2026
```
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
```
  2f8b4ce0
- [Perf] Optimize compute maxsim using batched version, 3.2% E2E throughput improvement (#36710) · c34ba6b9
  Wentao Ye authored Mar 11, 2026
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  c34ba6b9
11 Mar, 2026 21 commits
- [BUG] Fix async rlhf tests (#35811) · d6b61e51
  Aaron Hao authored Mar 11, 2026
```
Signed-off-by: ahao-anyscale <ahao@anyscale.com>
```
  d6b61e51
- [Model Runner V2] Add WhisperModelState [6/N] (#35790) · 55eed6b7
  Woosuk Kwon authored Mar 11, 2026
```
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
```
  55eed6b7
- [Model Runner V2] Add probabilistic rejection sampling for spec decoding (#35461) · c77181e5
  Giancarlo Delfin authored Mar 11, 2026
```
Signed-off-by: Giancarlo Delfin <gdelfin@inferact.ai>
```
  c77181e5
- [Refactor] Remove dead code in KV connector (#36424) · 35bdca54
  Wentao Ye authored Mar 11, 2026
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  35bdca54
- [KVConnector] Support worker -> scheduler metadata (#31964) · a1a3523a
  Or Ozeri authored Mar 11, 2026
```
Signed-off-by: Or Ozeri <oro@il.ibm.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
```
  a1a3523a
- Fix routed experts capture for hybrid models (Mamba + Attention) (#35744) · bea02cdf
  Hongxin Xu authored Mar 11, 2026
```
Signed-off-by: arlenxu <arlenxu@tencent.com>
Signed-off-by: xhx1022 <1737006628@qq.com>
Co-authored-by: arlenxu <arlenxu@tencent.com>
```
  bea02cdf
- Kimi k2.5 MLA based eagle3 (#36361) · 55738947
  Jhao-Ting Chen authored Mar 11, 2026
```
Signed-off-by: Izzy Putterman <iputterman@nvidia.com>
Signed-off-by: Jhao-Ting Chen <jhaotingc@nvidia.com>
Co-authored-by: Izzy Putterman <iputterman@nvidia.com>
```
  55738947
- [Model Runner V2] Remove unused warmup_for_prefill method (#36762) · 8ccbcda5
  Woosuk Kwon authored Mar 11, 2026
```
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
```
  8ccbcda5
- [ROCm][Perf] Allow MTP lens > 1 in Sparse MLA (#36681) · a9e532af
  tvirolai-amd authored Mar 11, 2026
```
Signed-off-by: Teemu Virolainen <teemu.virolainen@amd.com>
```
  a9e532af
- Add XPU MLA Sparse backend for DeepSeek v3.2 (#33230) · e584dce5
  Wuxun Zhang authored Mar 11, 2026
```
Signed-off-by: Zhang, Wuxun <wuxun.zhang@intel.com>
```
  e584dce5
- [XPU]Bug fix for some unexpected error when use AgRs backend on XPU device. (#36593) · c910eeb1
  YiSheng5 authored Mar 11, 2026
```
Signed-off-by: yisheng <yi.sheng@intel.com>
```
  c910eeb1
- [NIXL][1/N] Refactor `kernel_block_size` detection (#35752) · 098d8447
  Nicolò Lucchesi authored Mar 11, 2026
```
Signed-off-by: NickLucche <nlucches@redhat.com>
```
  098d8447
- [Bugfix] Add Multiple of 16 block_size to triton fallback on rocm Attention to... · a40ee486
  JartX authored Mar 11, 2026
```
[Bugfix] Add Multiple of 16 block_size to triton fallback on rocm Attention to support qwen3_5 (#35923)
Signed-off-by: JartX <sagformas@epdcenter.es>
Co-authored-by: akaratza <akaratza@amd.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
```
  a40ee486
- AITER MLA backend: Avoid CPU sync in _build_decode (#35765) · eac2dc2b
  pschlan-amd authored Mar 11, 2026
```
Signed-off-by: Patrick Schlangen <pschlan@amd.com>
```
  eac2dc2b
- feat(spec_decode): fuse EAGLE step slot mapping and metadata updates (#33503) · 4aaaf8c8
  Sladyn authored Mar 10, 2026
```
Signed-off-by: sladynnunes <snunes@usc.edu>
```
  4aaaf8c8
- [Doc] Fix duplicate words in comments (#36713) · 4bf53362
  Hongbin Guo authored Mar 11, 2026
```
Signed-off-by: Hongbin10 <jdmjdm1998@163.com>
```
  4bf53362
- [Misc][Attention] Clean up unused method in `CPU_ATTN` (#36673) · 5f77ef15
  Matthew Bonanni authored Mar 11, 2026
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  5f77ef15
- [Perf] Optimize scheduler overhead for PD disaggregation, around 5% E2E perf improvement (#35781) · a8ff2cca
  Wentao Ye authored Mar 11, 2026
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: Or Ozeri <oro@il.ibm.com>
```
  a8ff2cca
- [DSV3.2][MTP] Optimize Indexer MTP handling (#36723) · 9040cd40
  Benjamin Chislett authored Mar 11, 2026
```
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
```
  9040cd40
- [Bugfix] Surface exceptions from non-blocking execute_model in UniProcExecutor... · fa0d353a
  fangyuchu authored Mar 11, 2026
```
[Bugfix] Surface exceptions from non-blocking execute_model in UniProcExecutor to avoid DP deadlocks (#35194)
Signed-off-by: fangyuchu <fangyuchu@qq.com>
```
  fa0d353a
- [Bugfix] Fix DeepSeek V3.2 OOM during CG memory profiling (#36691) · 8ab3d742
  Matthew Bonanni authored Mar 10, 2026
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  8ab3d742
10 Mar, 2026 7 commits
- [Minor] Enhance error message for TRTLLM decode uniformity check (#36609) · 195d1ca3
  Woosuk Kwon authored Mar 10, 2026
```
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
```
  195d1ca3
- [Core] Simplify core kv-cache blocks initialization logic (#36521) · 65b2f405
  Nick Hill authored Mar 10, 2026
```
Signed-off-by: Nick Hill <nickhill123@gmail.com>
```
  65b2f405
- [Model Runner V2] Use unpadded num_tokens for PW CUDA graph attn metadata (#36626) · f088a831
  Woosuk Kwon authored Mar 10, 2026
```
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
```
  f088a831
- [CI] Bump `mypy` version to 1.19.1 (#36104) · f83b933b
  Harry Mellor authored Mar 10, 2026
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  f83b933b
- [ROCm][Perf] Enable `sparse_mla`'s cudagraph on ROCm platform (#35719) · 82f3f30e
  Pleaplusone authored Mar 11, 2026
```
Signed-off-by: ganyi <ygan@amd.com>
```
  82f3f30e
- [Bugfix][Sparse MLA] report indexer CG support properly (#36519) · 9095cbbf
  Matthew Bonanni authored Mar 10, 2026
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  9095cbbf
- feat(kv-offload): Strategy A — StoreReusedOffloadingManager gates CPU stores... · 106ff69c
  Srinivasoo7 authored Mar 10, 2026
```
feat(kv-offload): Strategy A — StoreReusedOffloadingManager gates CPU stores on reuse frequency (#35342)
Signed-off-by: srinivas_oo7 <Sriusa4414@gmail.com>
Signed-off-by: Sriusa4414@gmail.com
Signed-off-by: Srinivasoo7 <158864704+Srinivasoo7@users.noreply.github.com>
Co-authored-by: srinivas_oo7 <sklinkedin0120@gmail.com>
Co-authored-by: Srinivasoo7 <158864704+Srinivasoo7@users.noreply.github.com>
Co-authored-by: Or Ozeri <oro@il.ibm.com>
```
  106ff69c