Commits · 8060bb0333855739d5bdee0e67f7eb227fa04ebb · OpenDAS / vllm_cscc

07 Apr, 2026 3 commits
- [vLLM IR] rework gemma_rms_norm (#39014) · 8060bb03
  Jiangyun Zhu authored Apr 07, 2026
```
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
Signed-off-by: Jiangyun Zhu <riverclouds.zhu@qq.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
```
  8060bb03
- [Kernels][MoE] Fix legacy_routing to use bitmatrix-based routing path (#38504) · 2df2c85b
  Andreas Karatzas authored Apr 06, 2026
```
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
```
  2df2c85b
- [MoE Refactor] Split up compressed_tensors_moe.py (#38960) · b2b2c523
  bnellnm authored Apr 06, 2026
```
Signed-off-by: Bill Nell <bnell@redhat.com>
```
  b2b2c523
06 Apr, 2026 5 commits

[NVFP4] Support NVFP4 dense models from `modelopt` and `compressed-tensors` on... · 00d7b497

fxmarty-amd authored Apr 07, 2026


[NVFP4] Support NVFP4 dense models from `modelopt` and `compressed-tensors` on AMD Instinct MI300, MI355X and Hopper through emulation (#35733)
Signed-off-by: Felix Marty <Felix.Marty@amd.com>
Signed-off-by: fxmarty-amd <felmarty@amd.com>
Co-authored-by: Kyle Sayers <kylesayrs@gmail.com>

00d7b497

[Quantization] Add FlashInfer CuteDSL batched experts backend for NVFP4 MoE (#38251) · e8ebbdde

Yongye Zhu authored Apr 06, 2026


Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Roger Wang <hey@rogerw.io>

e8ebbdde

[EASY] Drop duplicate KV-cache initialization (#38799) · 94fbb098
namgyu-youn authored Apr 07, 2026
```
Signed-off-by: namgyu-youn <namgyu.dev@gmail.com>
```
94fbb098

[MoE Refactor][Test] FusedMoE layer test (#24675) · f0148240

bnellnm authored Apr 06, 2026


Signed-off-by: Bill Nell <bnell@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

f0148240

[MoE Refactor] Split of DefaultMoERunner class (#35326) · 93bada49

bnellnm authored Apr 06, 2026


Signed-off-by: Bill Nell <bnell@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

93bada49

05 Apr, 2026 3 commits

[Perf] Change Trtllm fp8 MoE to use Shuffled Weights and BlockMajorK Layout (#38993) · 1af6f78a

Wei Zhao authored Apr 05, 2026


Signed-off-by: wzhao18 <wzhao18.sz@gmail.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

1af6f78a

[Bugfix][MoE] Fix 6-8% decode regression: prefer multi-stream shared expert overlap (#38990) · 228023b3

Martin Vit authored Apr 05, 2026


Signed-off-by: Martin Vit <martin@voipmonitor.org>
Signed-off-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

228023b3

[Quantization][Deprecation] Remove Petit NVFP4 (#32694) · 968ed02a
Robert Shaw authored Apr 04, 2026
```
Signed-off-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>
```
968ed02a

04 Apr, 2026 4 commits

Revert "[vLLM IR] gemma_rms_norm" (#38998) · 7d266abb
Robert Shaw authored Apr 04, 2026

7d266abb
[vLLM IR] gemma_rms_norm (#38780) · 156405d2
Xiaoshuang Wang authored Apr 05, 2026
```
Signed-off-by: Icey <1790571317@qq.com>
```
156405d2

[Perf][GDN] Align TMA usage with upstream FLA (#38981) · 99e5539a

Artem Perevedentsev authored Apr 04, 2026


Signed-off-by: Artem Perevedentsev <aperevedents@nvidia.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

99e5539a

[IR][RmsNorm] pass None if not has_weight (#38961) · a88ce94b

Linkun authored Apr 04, 2026

Signed-off-by: Linkun Chen <github@lkchen.net>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>

a88ce94b

03 Apr, 2026 10 commits

[Bugfix] Re-enable Renormalize routing for TRT-LLM MoE experts (#38859) · a5a623d9
yzong-rh authored Apr 03, 2026
```
Signed-off-by: Yifan Zong <yzong@redhat.com>
```
a5a623d9
[Frontend] new online quantization frontend (#38138) · 7b1a7423
Vasiliy Kuznetsov authored Apr 03, 2026
```
Signed-off-by: Vasiliy Kuznetsov <vasiliy@meta.com>
```
7b1a7423

[Bugfix] Fix AWQ models batch invariance issues (#38670) · 46f02e00

Yusuf Mohammad authored Apr 03, 2026


Signed-off-by: yusuf <yusuf@deeplearningmachine.mynet>
Signed-off-by: <>
Co-authored-by: yusuf <yusuf@deeplearningmachine.mynet>

46f02e00

[XPU] bump up xpu-kernel v0.1.5, transpose moe weights (#38342) · 6b487224

Qiming Zhang authored Apr 03, 2026


Signed-off-by: mayuyuace <qiming1.zhang@intel.com>
Signed-off-by: Qiming Zhang <qiming1.zhang@intel.com>
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>

6b487224

[GDN] Eliminate GPU->CPU sync in prepare_chunk_indices during prefill (#38361) · cb10b7e8

Artem Perevedentsev authored Apr 03, 2026


Signed-off-by: Artem Perevedentsev <aperevedents@nvidia.com>
Signed-off-by: Vadim Gimpelson <156319763+vadiklyutiy@users.noreply.github.com>

cb10b7e8

[Intel][Triton] Support `round_int8` for Intel backend (#38825) · bf8b022e

Mieszko Dziadowiec authored Apr 03, 2026


Signed-off-by: Mieszko Dziadowiec <mdziadowiec@habana.ai>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Stefano Castagnetta <scastagnetta@nvidia.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Stefano Castagnetta <scastagnetta@nvidia.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>

bf8b022e

[ROCm][Quantization][1/N] Refactor quark_moe w_mxfp4 w/ oracle (#38774) · 103f0de5

Bowen Bao authored Apr 02, 2026


Signed-off-by: Bowen Bao <bowenbao@amd.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

103f0de5

[mla] Support fused FP8/NVFP4 output quantization in MLA attention (#35792) (#36205) · 1f5ec288

Carl Y authored Apr 02, 2026

Signed-off-by: Carl You <4531192+carlyou@users.noreply.github.com>
Signed-off-by: Carl Y <4531192+carlyou@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

1f5ec288

[XPU] Initial support for GDN attention on Qwen3-next/Qwen3.5 (#33657) · ee3cf457

Yan Ma authored Apr 03, 2026


Signed-off-by: Yan Ma <yan.ma@intel.com>
Signed-off-by: Chendi Xue <chendi.xue@intel.com>
Co-authored-by: Chendi Xue <chendi.xue@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>

ee3cf457

[New Model]: add support for telechat3 (#38510) · 71a9125c

1096125073 authored Apr 03, 2026


Signed-off-by: xiayongqiang <xiayq1@chinatelecom.cn>
Co-authored-by: xiayongqiang <xiayq1@chinatelecom.cn>

71a9125c

02 Apr, 2026 8 commits

[Kernel] Mamba support different layout for Conv state (#37416) · 66e86f1d
Nicolò Lucchesi authored Apr 03, 2026

66e86f1d

feat(models): implement Google Gemma 4 architecture support (MoE, Multimodal,... · 08ed2b96

Luciano Martins authored Apr 02, 2026

feat(models): implement Google Gemma 4 architecture support (MoE, Multimodal, Reasoning, Tool-Use) (#38826)
Signed-off-by: Luciano Martins <lucianommartins@users.noreply.github.com>
Signed-off-by: Luciano Martins <lucianomartins@google.com>
Co-authored-by: Luciano Martins <lucianommartins@users.noreply.github.com>
Co-authored-by: Isotr0py <2037008807@qq.com>

08ed2b96

[Bugfix] Enable batch-invariant Triton matmul on all Ampere GPUs (SM 8x) (#38427) · 16a65e41

Yusuf Mohammad authored Apr 02, 2026


Signed-off-by: yusuf <yusufmohammad@live.com>
Signed-off-by: yusuf <yusuf@deeplearningmachine.mynet>
Signed-off-by: Yusuf Mohammad <79484377+YM2132@users.noreply.github.com>
Signed-off-by: <>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: yusuf <yusuf@deeplearningmachine.mynet>

16a65e41

[Feature] KV cache per-token-head INT8/FP8 quantization (#38378) · 2ce3d0ce

JartX authored Apr 02, 2026


Signed-off-by: JartX <sagformas@epdcenter.es>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: yangyang4991 <yangyang4991@gmail.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: Isotr0py <2037008807@qq.com>

2ce3d0ce

[Perf] fuse kernels in gdn (#37813) · 4eefbf96
Jiangyun Zhu authored Apr 02, 2026
```
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
```
4eefbf96

[ROCm] Enable VLLM triton FP8 moe for gfx1201, tuned for Qwen3-30B-A3B-FP8... · 551b3fb3

vllmellm authored Apr 02, 2026


[ROCm] Enable VLLM triton FP8 moe for gfx1201, tuned for Qwen3-30B-A3B-FP8 tp=2 and Qwen/Qwen3.5-35B-A3B-FP8 tp=2 (#38086)
Signed-off-by: big-yellow-duck <jeffaw99@hotmail.com>
Co-authored-by: vllmellm <vllm.ellm@embeddedllm.com>

551b3fb3

[CPU] Support gelu act in cpu_fused_moe (#38770) · c6f722b9
Li, Jiang authored Apr 02, 2026
```
Signed-off-by: jiang1.li <jiang1.li@intel.com>
```
c6f722b9
Revert "[Kernel] Add gpt-oss Router GEMM kernel (#37205)" (#38778) · 9bd72311
Xin Yang authored Apr 01, 2026
```
Signed-off-by: Xin Yang <xyangx@amazon.com>
```
9bd72311

01 Apr, 2026 6 commits
- [Kernel] Add MXFP8 to Marlin GEMM/MoE and refactor Mxfp8LinearOp (#34664) · db5d0719
  Michael Goin authored Apr 01, 2026
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  db5d0719
- [MoE Refactor] Make SharedExperts class for use with DefaultMoERunner (#35153) · 7cf56a59
  bnellnm authored Apr 01, 2026
```
Signed-off-by: Bill Nell <bnell@redhat.com>
```
  7cf56a59
- [Bugfix][MLA] Add logits size budget to sparse indexer prefill chunking (#36178) · eb474549
  Lucas Wilkinson authored Apr 01, 2026
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
  eb474549
- [1/N][Cleanup] Standardize on use of `is_quantized_kv_cache` (#38659) · 116f4be4
  Matthew Bonanni authored Apr 01, 2026
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  116f4be4
- [Perf] Optimize mean pooling using chunks and index_add, 5.9% E2E throughput improvement (#38559) · 7b01d97a
  Wentao Ye authored Mar 31, 2026
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  7b01d97a
- [vLLM IR] 1/N Implement IR skeleton and rms_norm op (#33825) · 40bb1750
  Luka Govedič authored Mar 31, 2026
```
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Signed-off-by: Xinyu Chen <xinyu1.chen@intel.com>
Signed-off-by: chzhang <chaojun.zhang@intel.com>
Signed-off-by: Luka Govedic <luka.govedic@gmail.com>
Co-authored-by: Xinyu Chen <xinyu1.chen@intel.com>
Co-authored-by: Chaojun Zhang <chaojun.zhang@intel.com>
Co-authored-by: Luka Govedič <ProExpertProg@h100-01.nemg-001.lab.rdu2.dc.redhat.com>
```
  40bb1750
31 Mar, 2026 1 commit
- [Perf] Fix DBO overlap: capture DeepEP event before yield (#38451) · 517b769b
  czhu-cohere authored Mar 31, 2026
```
Signed-off-by: root <conway.zhu@cohere.com>
```
  517b769b