Commits · 097978a15dc3757e6fdbbae6ad752b97691581bd · OpenDAS / vllm_cscc

20 Dec, 2025 1 commit

[MoE Refactor][5/N] Isolate zero expert to LongCatFlash (#28891) · 54c89243

baonudesifeizhai authored Dec 20, 2025

Signed-off-by: baonudesifeizhai <85092850+baonudesifeizhai@users.noreply.github.com>
Signed-off-by: Dongjie Zou <85092850+baonudesifeizhai@users.noreply.github.com>
Signed-off-by: baonudesifeizhai <baonudesifeizhai@gmail.com>
Signed-off-by: Robert Shaw <robertgshaw2@gmail.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <robertgshaw2@gmail.com>

54c89243

16 Dec, 2025 3 commits
- [Kernel][Quantization][MoE] add marlin kernel support for turing (sm75) (#29901) · ce96857f
  Jinzhen Lin authored Dec 17, 2025
```
Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
```
  ce96857f
- [Perf] Do FP4 quant before All gather on flashinfer trtllmgen MOE (#30014) · 254a7f8f
  jiahanc authored Dec 16, 2025
```
Signed-off-by: jiahanc <173873397+jiahanc@users.noreply.github.com>
```
  254a7f8f
- [TRTLLM] Remove the MoE GEMM weight name change (#30713) · ce12b407
  Ming Yang authored Dec 16, 2025
```
Signed-off-by: Ming Yang <minos.future@gmail.com>
```
  ce12b407
14 Dec, 2025 1 commit

Nvidia ModelOpt workaround for issue 28072 (#30164) · 0bb0bae4

Shengliang Xu authored Dec 14, 2025


Signed-off-by: Shengliang Xu <shengliangx@nvidia.com>
Co-authored-by: Pavani Majety <pmajety@nvidia.com>

0bb0bae4

12 Dec, 2025 1 commit

[MoE-FP8-modelopt] Add FlashInfer alignment padding for intermediate dimensions (#29748) · 13618626

danielafrimi authored Dec 12, 2025


Signed-off-by: Daniel Afrimi <dafrimi@pool0-00589.cm.cluster>
Signed-off-by: dafrimi <dafrimi@nvidia.com>
Co-authored-by: Daniel Afrimi <dafrimi@pool0-00589.cm.cluster>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>

13618626

11 Dec, 2025 1 commit

[EPLB] Support EPLB w/ NVFP4 (#29804) · a00d8897

Andrew Briand authored Dec 11, 2025


Signed-off-by: Andrew Briand <abriand@nvidia.com>
Co-authored-by: Andrew Briand <abriand@nvidia.com>

a00d8897

09 Dec, 2025 1 commit

[MoE][Refactor] Remove most arguments to FusedMoEMethodBase.apply (#29066) · 00e5cbb9

bnellnm authored Dec 09, 2025

Signed-off-by: Bill Nell <bnell@redhat.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>

00e5cbb9

30 Nov, 2025 1 commit
- [Feat] Support non-gated activations in NVFP4 modelopt path (#29004) · 39d28108
  Omer Ullman Argov authored Nov 30, 2025
  
  39d28108
29 Nov, 2025 1 commit

[Kernel][Quantization] add w4a8 support for marlin kernel (#24722) · 1656ad37

Jinzhen Lin authored Nov 29, 2025


Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Michael Goin <mgoin@redhat.com>

1656ad37

27 Nov, 2025 1 commit
- [Attention] Update attention imports (#29540) · fc1d8be3
  Matthew Bonanni authored Nov 27, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  fc1d8be3
25 Nov, 2025 1 commit
- [Kernel] Add NVFP4 MoE CUTLASS support for SM120 (#29242) · e5020986
  Michael Goin authored Nov 25, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
```
  e5020986
24 Nov, 2025 1 commit
- [MoE][Refactor] Make select_experts a non-static method (#29067) · 8f066146
  bnellnm authored Nov 24, 2025
```
Signed-off-by: Bill Nell <bnell@redhat.com>
```
  8f066146
21 Nov, 2025 1 commit

Add TRTLLM MoE NVFP4 kernel to CompressedTensorsW4A4MoeMethod (#28892) · b4c8fbaa

Mingyuan Ma authored Nov 21, 2025


Signed-off-by: mingyuanm <mingyuanm@nvidia.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>

b4c8fbaa

20 Nov, 2025 1 commit
- Consolidate Nvidia ModelOpt quant config handling for all quantization methods (#28076) · a8c53682
  Shengliang Xu authored Nov 19, 2025
```
Signed-off-by: Shengliang Xu <shengliangx@nvidia.com>
```
  a8c53682
19 Nov, 2025 3 commits
- [Refactor] Optimize `select_experts` (#28069) · 5031cd5d
  Wentao Ye authored Nov 19, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  5031cd5d
- [MoE] Nvfp4 Masked Gemm: Add flashinfer grouped_gemm_nt_masked (#25990) · 613abb50
  Shu Wang authored Nov 19, 2025
```
Signed-off-by: Shu Wang. <shuw@nvidia.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
```
  613abb50
- [Feat][Perf] Enable deepep-low-latency with round-robin expert placement. (#28449) · da2f6800
  Chen Bruce authored Nov 19, 2025
```
Signed-off-by: bruceszchen <bruceszchen@tencent.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  da2f6800
17 Nov, 2025 1 commit

[Performance][Fix] update nvfp4 code to support renorm routing (#28569) · 561253b3

jiahanc authored Nov 16, 2025


Signed-off-by: jiahanc <173873397+jiahanc@users.noreply.github.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>

561253b3

16 Nov, 2025 1 commit
- Feature: Support Relu2 in FusedMoE fp8 cutlass path (#27261) · 03ee4811
  amirkl94 authored Nov 16, 2025
  
  03ee4811
07 Nov, 2025 1 commit
- [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
05 Nov, 2025 1 commit

Bugfix: Cutlass FP8 FusedMoE bad scaling factors (#27255) · 6b7a8118

amirkl94 authored Nov 05, 2025


Signed-off-by: Amir Klein <203507526+amirkl94@users.noreply.github.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>

6b7a8118

04 Nov, 2025 1 commit
- [Kernels] Isolate modular kernel code from FusedMoEMethodBase subclasses. (#27123) · 938772af
  bnellnm authored Nov 04, 2025
  
  938772af
31 Oct, 2025 1 commit
- Flashinfer_CUTLASS_MOE fuses quantization for TP (#27223) · fc16f1c4
  Shu Wang authored Oct 31, 2025
```
Signed-off-by: Shu Wang. <shuw@nvidia.com>
```
  fc16f1c4
23 Oct, 2025 1 commit
- [Model] Add MoE support for NemotronH (#25863) · 61089465
  tomeras91 authored Oct 23, 2025
```
Signed-off-by: Tomer Asida <57313761+tomeras91@users.noreply.github.com>
```
  61089465
21 Oct, 2025 1 commit

[ModelOpt] Load w13/w2_input_scale for all experts, nvfp4 (#26135) · f95da13c

Shu Wang authored Oct 21, 2025


Signed-off-by: Shu Wang <shuw@nvidia.com>
Signed-off-by: Shu Wang. <shuw@nvidia.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>

f95da13c

16 Oct, 2025 1 commit
- [NVIDIA] [Perf] Update to leverage flashinfer trtllm FP4 MOE throughput kernel (#26714) · 41d30719
  jiahanc authored Oct 16, 2025
```
Signed-off-by: jiahanc <173873397+jiahanc@users.noreply.github.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
```
  41d30719
15 Oct, 2025 2 commits
- [NVIDIA] Add support for cudnn fp4 gemm via flashinfer (#26107) · de92d916
  Kaixi Hou authored Oct 15, 2025
```
Signed-off-by: kaixih <kaixih@nvidia.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
```
  de92d916
- [ModelOpt] Remove NVFP4 MoE K%16==0 constraint (#26891) · d7963752
  XiaobingZhang authored Oct 16, 2025
```
Signed-off-by: XiaobingSuper <xiaobingzhangupc@gmail.com>
```
  d7963752
14 Oct, 2025 1 commit

[torch.compile] Unwrap fused_marlin_moe custom op (#26739) · 8ae16928

Varun Sundar Rabindranath authored Oct 13, 2025


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

8ae16928

12 Oct, 2025 1 commit
- Update `Optional[x]` -> `x | None` and `Union[x, y]` to `x | y` (#26633) · 8fcaaf6a
  Harry Mellor authored Oct 12, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  8fcaaf6a
06 Oct, 2025 1 commit
- [Bugfix] Allow skipping MoE in NVFP4 (fix for MTP) (#25987) · 2161efe9
  Benjamin Chislett authored Oct 06, 2025
```
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
```
  2161efe9
05 Oct, 2025 1 commit
- Convert formatting to use `ruff` instead of `yapf` + `isort` (#26247) · d6953beb
  Harry Mellor authored Oct 05, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  d6953beb
03 Oct, 2025 1 commit

[Quantization/NVFP4] Speed up TRTLLM NVFP4 MOE weight loading and fix K/V... · a2691733

Pavani Majety authored Oct 03, 2025


[Quantization/NVFP4] Speed up TRTLLM NVFP4 MOE weight loading and fix K/V scale loading for MLA Attn (#25968)
Signed-off-by: Pavani Majety <pmajety@nvidia.com>

a2691733

29 Sep, 2025 1 commit
- [Bugfix] Use correct key "ignore" for config.json non-quantized layers (#25706) · d5ab2851
  Lee Nau authored Sep 29, 2025
```
Signed-off-by: Lee Nau <lnau@nvidia.com>
```
  d5ab2851
25 Sep, 2025 2 commits
- Fix routing_bias dtype (#25711) · 081b5594
  Shu Wang authored Sep 25, 2025
```
Signed-off-by: Shu Wang. <shuw@nvidia.com>
```
  081b5594
- [Model] Add LongCat-Flash (#23991) · 845adb3e
  XuruiYang authored Sep 25, 2025
```
Signed-off-by: yangxurui <yangxurui@meituan.com>
Co-authored-by: yangxurui <yangxurui@meituan.com>
```
  845adb3e
19 Sep, 2025 1 commit

Enable modelopt gemma3 nvfp4/fp8, make workflow more robust (#22771) · 431535b5

Zhiyu authored Sep 19, 2025


Signed-off-by: Zhiyu Cheng <zhiyuc@nvidia.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>

431535b5

17 Sep, 2025 1 commit
- [Kernel] Delegate construction of FusedMoEQuantConfig to FusedMoEMethodBase subclasses (#22537) · 5963b98b
  bnellnm authored Sep 17, 2025
```
Signed-off-by: Bill Nell <bnell@redhat.com>
```
  5963b98b
16 Sep, 2025 1 commit

[Perf] Reuse workspace for FP8+FP4 Marlin MoE (#20500) · dbebb7f8

Michael Goin authored Sep 16, 2025


Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>

dbebb7f8