Commits · 874d2aaa5060e46a33e9d09b30925e909da758bf · OpenDAS / vllm_cscc

06 Jan, 2026 1 commit

[feat] 支持 mRoPE 的 fused RMSNorm+RoPE 路径，并修正 torch.compile 动态维度标注 · 874d2aaa

zhuwenwen authored Jan 06, 2026

实现了用于优化张量计算的 rms_mrope_fuse 和 rms_mrope_fuse_fake 方法
更新了 forward：在满足条件时走新的 M-RoPE 融合路径
增强了 Qwen3MoeModel 对动态参数维度的支持，以适配该功能

874d2aaa

17 Dec, 2025 1 commit

Synchronize the modifications from the 12th to the 17th: · b66c8e4b

zhuwenwen authored Dec 17, 2025

修复CompressedTensorsLinearMethod中的w4a16的冲突问题
feat(moe): add Marlin W16A16 fused MoE behind VLLM_USE_MARLIN_W16A16_MOE
replace the fp8_mqa_logits and fp8_paged_mqa_logits interfaces in deepgemm with mqa_logits and paged_mqa_logits from lightop

b66c8e4b

11 Dec, 2025 1 commit
- Standardise `get_rope` to use `rope_parameters["partial_rotary_factor"]`, not `rotary_dim` (#30389) · cf3eacfe
  Harry Mellor authored Dec 11, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  cf3eacfe
10 Dec, 2025 1 commit
- [bugfix][quantization] fix quark qwen3 kv_cache quantization (#30308) · 06462392
  haoyangli-amd authored Dec 10, 2025
```
Signed-off-by: Haoyang Li <lihaoyang0109@gmail.com>
```
  06462392
26 Nov, 2025 1 commit
- [Attention] Remove imports from `vllm/attention/__init__.py` (#29342) · 430dd4d9
  Matthew Bonanni authored Nov 26, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  430dd4d9
19 Nov, 2025 1 commit
- Update `rope_scaling` to `rope_parameters` in preparation for Transformers v5 (#28542) · a8b70304
  Harry Mellor authored Nov 19, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  a8b70304
13 Nov, 2025 2 commits
- add VLLM_USE_OPT_ZEROS to replace triton_ (torch.zeros) · e7f2785f
  zhuwenwen authored Nov 13, 2025
```
set default_max_num_batched_tokens = 10240
update qwen3_moe of layernorm
off lightop of moe_fused_gate
```
  e7f2785f
- Rename clashing method names for vLLM model protocol (#27583) · 97d1c993
  Harry Mellor authored Nov 13, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  97d1c993
10 Nov, 2025 1 commit
- [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
05 Nov, 2025 1 commit

[BugFix] Support EP/DP + EPLB with MTP (#25311) · e50c4546

Ilya Markov authored Nov 05, 2025


Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>

e50c4546

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
11 Oct, 2025 2 commits
- [Bugfix] Fix qwen-moe packed_modules_mapping (#26634) · f0a30a06
  Jee Jee Li authored Oct 11, 2025
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  f0a30a06
- Add EAGLE-3 Speculative Decoding Support for Qwen3 MoE (#26485) · d2a71530
  Rahul Tuli authored Oct 11, 2025
```
Signed-off-by: Rahul Tuli <rtuli@redhat.com>
```
  d2a71530
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
28 Sep, 2025 3 commits
- [Bugfix] Fix Qwen3-VL regression from #24982 (#25814) · 19e7ab73
  Roger Wang authored Sep 27, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
Signed-off-by: simon-mo <simon.mo@hey.com>
```
  19e7ab73
- [Bugfix][WideEP] Apply TP Attn + EP MoE fix to other models (#24982) · 26a7a33b
  Tyler Michael Smith authored Sep 27, 2025
```
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Signed-off-by: simon-mo <simon.mo@hey.com>
```
  26a7a33b
- [Bugfix] Fix Qwen3-VL regression from #24982 (#25814) · 61447540
  Roger Wang authored Sep 27, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
```
  61447540
27 Sep, 2025 1 commit
- [Bugfix][WideEP] Apply TP Attn + EP MoE fix to other models (#24982) · a5354b3e
  Tyler Michael Smith authored Sep 27, 2025
```
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
```
  a5354b3e
26 Sep, 2025 1 commit
- [Quantization] Add field to skip unquantized modules for GPTQ config (#25455) · d4d98998
  Isotr0py authored Sep 26, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  d4d98998
21 Sep, 2025 1 commit
- [V0 Deprecation] Remove V0 sampling metadata (#25345) · 1c3ffdbe
  Woosuk Kwon authored Sep 21, 2025
```
Signed-off-by: Woosuk Kwon <woosuk@thinkingmachines.ai>
```
  1c3ffdbe
17 Sep, 2025 3 commits

[Kernel] Delegate construction of FusedMoEQuantConfig to FusedMoEMethodBase subclasses (#22537) · 5963b98b
bnellnm authored Sep 17, 2025
```
Signed-off-by: Bill Nell <bnell@redhat.com>
```
5963b98b
[Model] Pass param prefix to LLMHead (#24862) · 4a9375fe
whx authored Sep 17, 2025
```
Signed-off-by: whx-sjtu <2952154980@qq.com>
```
4a9375fe

[Model] Support Qwen3-VL Model Series (#24727) · 0f7acdd7

Roger Wang authored Sep 16, 2025


Signed-off-by: Roger Wang <hey@rogerw.io>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Huang Jie <92386084+JJJYmmm@users.noreply.github.com>
Co-authored-by: 松灵 <26085463+wulipc@users.noreply.github.com>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>

0f7acdd7

13 Sep, 2025 1 commit
- [Compilation Bug] Fix Inductor Graph Output with Shape Issue (#24772) · da3fa78d
  Wentao Ye authored Sep 12, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  da3fa78d
12 Sep, 2025 1 commit
- [Compilation Bug] Fix Inductor Graph Output with Shape Issue (#24772) · 3beadc2f
  Wentao Ye authored Sep 12, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  3beadc2f
01 Sep, 2025 1 commit

[BUGFIX] GPTQ quantization compatibility for Qwen3 MOE models (AutoGPTQ and... · 183a7096

JartX authored Sep 01, 2025


[BUGFIX] GPTQ quantization compatibility for Qwen3 MOE models (AutoGPTQ and AutoRound-GPTQ) (#23994)
Signed-off-by: JartX <sagformas@epdcenter.es>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>

183a7096

29 Aug, 2025 1 commit
- [Models] Improve iteration over layers (#19497) · de533ab2
  Lukas Geiger authored Aug 29, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  de533ab2
25 Aug, 2025 1 commit
- [Bugfix] Fix Qwen3 MoE GPTQ inference (#23490) · a9082a4d
  Isotr0py authored Aug 25, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  a9082a4d
20 Aug, 2025 1 commit

[Feature] use --eplb_config to set eplb param (#20562) · 4fbda0b2

rongfu.leng authored Aug 21, 2025


Signed-off-by: rongfu.leng <rongfu.leng@daocloud.io>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: rongfu.leng <lenronfu@gmail.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

4fbda0b2

19 Aug, 2025 1 commit
- [Model] Removes redundant all-reduce operation in Qwen3MoeSparseMoeBlock (#23169) · 4f510bc2
  yiz-liu authored Aug 20, 2025
```
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
```
  4f510bc2
13 Aug, 2025 1 commit
- Fix GGUF loader for Qwen3 MoE. (#22785) · b159c0a6
  Gh0u1L5 authored Aug 13, 2025
```
Signed-off-by: Gh0u1L5 <Gh0u1L5@outlook.com>
```
  b159c0a6
12 Aug, 2025 1 commit

Enable 4bit bnb prequant MOE (#21548) · 9b94d6ec

Andy Chen authored Aug 11, 2025


Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>

9b94d6ec

11 Aug, 2025 1 commit
- [BUGFIX] KeyError 'layers.14.mlp.gate.g_idx' for Qwen3-MoE with GPTQ on ROCm (#22017) · 1e55dfa7
  JartX authored Aug 11, 2025
  
  1e55dfa7
09 Aug, 2025 1 commit
- [Misc] Use config definitions from Transformers library (#21913) · 65552b47
  Cyrus Leung authored Aug 09, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  65552b47
08 Aug, 2025 1 commit
- Add ModelOpt Qwen3 nvfp4 support (#20101) · d57dc236
  Zhiyu authored Aug 07, 2025
```
Signed-off-by: Zhiyu Cheng <zhiyuc@nvidia.com>
```
  d57dc236
07 Aug, 2025 1 commit
- [Qwen3] Enable dual-chunk-attention support for Qwen3 models. (#21924) · 7377131a
  Tao He authored Aug 07, 2025
```
Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com>
```
  7377131a
30 Jul, 2025 1 commit
- [Feature][EPLB] Add eplb support for Qwen3 (#20815) · d979dd6b
  aladerran authored Jul 30, 2025
```
Signed-off-by: aladerran <aladerran@gmail.com>
```
  d979dd6b
17 Jul, 2025 1 commit
- [Misc] Qwen MoE model supports LoRA (#20932) · a3a6c695
  Jee Jee Li authored Jul 18, 2025
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  a3a6c695
14 Jul, 2025 1 commit
- [Quantization] add BNB for MixtralForCausalLM (#20893) · a99b9f7d
  Jee Jee Li authored Jul 14, 2025
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  a99b9f7d
12 Jul, 2025 1 commit
- remove qk contiguous · e34d3444
  zhuwenwen authored Jul 12, 2025
  
  e34d3444