Commits · 3982bc2cd0bd9d633060b22e9ff683d8316a0f82 · OpenDAS / vllm_cscc · GitLab

21 Mar, 2026 9 commits
- [ROCm] Enable DeepEP ROCm as all2allbackend for AMD GPUs. (#34692) · 3982bc2c
  Chaitanya Sri Krishna Lolla authored Mar 21, 2026
```
Signed-off-by: Tej Kiran <vpolamre@amd.com>
Co-authored-by: Tej Kiran <vpolamre@amd.com>
```
  3982bc2c
- [ROCm][CI] Update GSM8K eval config to use fp8-and-mixed models list (MI355) (#37721) · 02eec7ec
  Andreas Karatzas authored Mar 21, 2026
```
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
```
  02eec7ec
- [Responses API] Add kv_transfer_params for PD disaggregation (#37424) · 17ee641c
  Bongwoo Bak authored Mar 21, 2026
```
Signed-off-by: bongwoobak <bongwoobak@gmail.com>
Co-authored-by: Chauncey <chaunceyjiang@gmail.com>
```
  17ee641c
- [ROCm][CI] Mark gemma3 as large GPU test to avoid OOM on MI250 (#37610) · 0d50fa1d
  Andreas Karatzas authored Mar 20, 2026
```
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
```
  0d50fa1d
- Revert "[compile] Initialize passes at VllmBackend init" (#37733) · 1fa1e53a
  Simon Mo authored Mar 20, 2026
  
  1fa1e53a
- [ROCm][CI] Guard CudaPlatform/RocmPlatform imports to fix test collection on... · 3ffa5200
  Andreas Karatzas authored Mar 20, 2026
```
[ROCm][CI] Guard CudaPlatform/RocmPlatform imports to fix test collection on cross-platform builds (#37617)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
```
  3ffa5200
- [MoE Refactor] Mxfp4 oracle rebased (#37128) · 87bd9189
  Yongye Zhu authored Mar 20, 2026
```
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
```
  87bd9189
- [Frontend] Remove librosa from audio dependency (#37058) · c7f98b4d
  Isotr0py authored Mar 21, 2026
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  c7f98b4d
- Add get_device_uuid for rocm (#37694) · 1c472f8f
  tmm77 authored Mar 20, 2026
```
Signed-off-by: Tiffany Mintz <Tiffany.Mintz@amd.com>
```
  1c472f8f
20 Mar, 2026 31 commits
- elastic_ep: Fix issues with repeated scale up/down cycles (#37131) · c57d38d6
  Itay Alroy authored Mar 21, 2026
```
Signed-off-by: Itay Alroy <ialroy@nvidia.com>
Co-authored-by: Ron Tourgeman <rtourgeman@nvidia.com>
```
  c57d38d6
- [BugFix] Allow qk_nope_head_dim=192 in FlashInfer MLA backend checks (#37475) · e5ed6c6c
  Kaihang Jiang authored Mar 20, 2026
```
Signed-off-by: Kaihang Jiang <kaihangj@nvidia.com>
```
  e5ed6c6c
- [Refactor] Remove unused dead code (#36171) · b3d0b379
  Wentao Ye authored Mar 20, 2026
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  b3d0b379
- [Model Runner V2] Support Streaming Inputs (#37028) · 85f671b8
  Santino Ramos authored Mar 20, 2026
```
Signed-off-by: Santino Ramos <elsantinoramos@gmail.com>
```
  85f671b8
- [ROCm][CI] Setting some mi325_4 tests back to optional (in parity with upstream) (#37711) · 8bc6b5cd
  Andreas Karatzas authored Mar 20, 2026
```
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
```
  8bc6b5cd
- [Bugfix] Disable monolithic TRTLLM MoE for Renormalize routing (#37591) (#37605) · 4f16ebbb
  Vadim Gimpelson authored Mar 20, 2026
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
```
  4f16ebbb
- [compile] Initialize passes at VllmBackend init (#35216) · 12fd17eb
  Angela Yi authored Mar 20, 2026
```
Signed-off-by: angelayi <yiangela7@gmail.com>
```
  12fd17eb
- [Model] Update Kimi-K25 and Isaac processors to fit HF-style (#37693) · 37aadf62
  Cyrus Leung authored Mar 21, 2026
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  37aadf62
- [Bugfix] Disable --calculate-kv-scales for hybrid GDN/Mamba+Attention… (#37565) · d7d2b5e4
  Le Yang authored Mar 21, 2026
```
Signed-off-by: Young-Leo <562593859@qq.com>
```
  d7d2b5e4
- refactor: abstract deepgemm support into platform (#37519) · 6ec5e9fd
  SherryC41 authored Mar 21, 2026
```
Co-authored-by: sherryC41 <sherry.c.c41@gmail.com>
```
  6ec5e9fd
- [Attention] Support distinguishing between short extends and decodes (#37303) · e1d85e5c
  Lucas Wilkinson authored Mar 20, 2026
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
  e1d85e5c
- fix CUDAGraph memory being counted twice (#37426) · 79eb9369
  Peter Pan authored Mar 21, 2026
```
Signed-off-by: Peter Pan <Peter.Pan@daocloud.io>
Signed-off-by: Peter Pan <peter.pan@daocloud.io>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>
```
  79eb9369
- [MRV2] Avoid recompilation of _gather_block_tables_kernel (#37645) · e80cfe57
  Woosuk Kwon authored Mar 20, 2026
```
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
```
  e80cfe57
- [Perf] Eliminate redundant SparseMatrix creation in gpt_oss_triton_kernels (#37683) · d0532bf3
  Xin Yang authored Mar 20, 2026
```
Signed-off-by: Xin Yang <xyangx@amazon.com>
```
  d0532bf3
- [ROCm][CI] Fix accuracy for llama-nemotron-vl pooling tests (#37613) · fb4e8bf4
  Andreas Karatzas authored Mar 20, 2026
```
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
```
  fb4e8bf4
- Fix various config related issues for Transformers v5 (#37681) · 6ade4bc5
  Harry Mellor authored Mar 20, 2026
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  6ade4bc5
- [compile] Add compiled artifact counter for VLLM_USE_MEGA_AOT_ARTIFACT=1. (#37589) · 2e089b96
  Zhengxu Chen authored Mar 20, 2026
```
Signed-off-by: zhxchen17 <zhxchen17@fb.com>
```
  2e089b96
- [Metrics] Some small refactoring for better maintainability (#33898) · 880be2b1
  Martin Hickey authored Mar 20, 2026
```
Signed-off-by: Martin Hickey <martin.hickey@ie.ibm.com>
```
  880be2b1
- [compile] Fix aot test failures with torch 2.12. (#37604) · c0f5fae6
  Zhengxu Chen authored Mar 20, 2026
```
Signed-off-by: zhxchen17 <zhxchen17@fb.com>
```
  c0f5fae6
- [Pixtral] Enable Pixtral language model support Eagle3 (#37182) · aa84e43c
  Rémi Delacourt authored Mar 20, 2026
```
Signed-off-by: remi <remi@mistral.ai>
```
  aa84e43c
- [Bugfix] Fix ConchLinearKernel channelwise quantization (group_size=-1) (#37329) · 5e806bcf
  Matthias Gehre authored Mar 20, 2026
```
Signed-off-by: Matthias Gehre <matthias.gehre@amd.com>
```
  5e806bcf
- [Bugfix] Reject channelwise quantization (group_size <= 0) in ExllamaLinearKernel (#37331) · 56a62c31
  Matthias Gehre authored Mar 20, 2026
```
Signed-off-by: Matthias Gehre <matthias.gehre@amd.com>
```
  56a62c31
- [ROCm] Enable wvSplitK skinny GEMM kernel for RDNA4/gfx1x decode (#34709) · 1779c098
  L.B.R. authored Mar 20, 2026
```
Signed-off-by: L.B.R. <lbr@mmonad.com>
Co-authored-by: L.B.R. <lbr@mmonad.com>
```
  1779c098
- [ROCm][Quantization] make quark ocp mx dtype parser robust for weight-only quantization (#36232) · 44eea10f
  xuebwang-amd authored Mar 20, 2026
```
Signed-off-by: xuebwang-amd <xuebwang@amd.com>
```
  44eea10f
- [Model] Add LFM2-ColBERT-350M support (#37528) · 8b6c6b95
  Ilya Boytsov authored Mar 20, 2026
```
Signed-off-by: Ilya Boytsov <ilyaboytsov1805@gmail.com>
```
  8b6c6b95
- Fix attribute error in `isaac_patch_hf_runner` (#37685) · 9f6d9dd3
  Harry Mellor authored Mar 20, 2026
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  9f6d9dd3
- [UX] Enable torch_profiler_with_stack (#37571) · dd20ee4e
  Jee Jee Li authored Mar 20, 2026
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  dd20ee4e
- [Misc] Use logger.info_once for auto tool choice log message (#37661) · 0523449c
  Chauncey authored Mar 20, 2026
```
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
```
  0523449c
- [Refactor] Relocate tests from tests/v1/entrypoints/ to tests/entrypoints/ (#37500) · b4c1aef2
  Flora Feng authored Mar 20, 2026
```
Signed-off-by: sfeng33 <4florafeng@gmail.com>
```
  b4c1aef2
- [Refactor] Move serve entrypoint tests under tests/entrypoints/serve/ (#37595) · 6050b93b
  Flora Feng authored Mar 20, 2026
```
Signed-off-by: sfeng33 <4florafeng@gmail.com>
```
  6050b93b
- [ROCm][CI] Fix granite_speech test for gfx90a by selecting compatible attention backend (#37611) · 5a4a1795
  Andreas Karatzas authored Mar 20, 2026
```
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
```
  5a4a1795