Commits · 7f1bcd18ffd9313a72e1d72c55a3d44f4fbb6157 · OpenDAS / vllm_cscc

16 Jan, 2026 1 commit

[Bugfix] [DeepSeek-V3.2] fix sparse_attn_indexer padding (#32175) · 5de6dd06

Kebe authored Jan 16, 2026


Signed-off-by: Kebe <mail@kebe7jun.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>

5de6dd06

09 Jan, 2026 1 commit
- [1/N][Attention] Restructure attention: move files (#31916) · 2612ba92
  Matthew Bonanni authored Jan 09, 2026
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  2612ba92
07 Jan, 2026 1 commit
- [BugFix] LoRA: Support loading base_layer of experts (#31104) · 48291484
  ℍ𝕠𝕝𝕝𝕠𝕨 𝕄𝕒𝕟 authored Jan 07, 2026
```
Signed-off-by: Hollow Man <hollowman@opensuse.org>
```
  48291484
24 Dec, 2025 2 commits
- [Bugfix][ROCm] Fix load issue on deepseek quark quantization when shared expert enabled (#31261) · b41aeb34
  Pleaplusone authored Dec 24, 2025
```
Signed-off-by: ganyi <ygan@amd.com>
```
  b41aeb34
- [Bug] Fix `Number of dimensions of tensors must match.` for Deepseek V3.2 (#31160) · 76e6a951
  Wentao Ye authored Dec 23, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  76e6a951
19 Dec, 2025 1 commit
- [Bug] Fix `error 'Dynamo failed to run FX node with fake tensors` for Deepseek V3.2 (#31046) · 4cf94298
  Wentao Ye authored Dec 19, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  4cf94298
17 Dec, 2025 1 commit
- [Bugfix] deepseek-V3.2 self.weights_proj has no bias (#30841) · 84896fda
  baoqian426 authored Dec 17, 2025
```
Signed-off-by: baoqian <1354987947@qq.com>
Signed-off-by: baoqian426 <1354987947@qq.com>
```
  84896fda
12 Dec, 2025 1 commit
- [Attention] Use sparse prefill kernel for fp8 kv-cache in DeepSeek-v3.2 (#27532) · 3e41992f
  Lucas Wilkinson authored Dec 12, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
  3e41992f
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
08 Dec, 2025 1 commit

[DeepSeek v3.2] Make top-k work for any logit values. (#27568) · 184076c3

Daniel Cámpora authored Dec 08, 2025


Signed-off-by: Daniel Campora <961215+dcampora@users.noreply.github.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>

184076c3

02 Dec, 2025 2 commits

[BUGFIX] llama_4_scaling wrongly passed to DeepseekAttention (#29908) · 5e5646e2
Julien Denize authored Dec 02, 2025
```
Signed-off-by: juliendenize <julien.denize@mistral.ai>
```
5e5646e2

Add Mistral Large 3 and Ministral 3 (#29757) · d8c6210e

Julien Denize authored Dec 02, 2025


Signed-off-by: Julien Denize <julien.denize@mistral.ai>
Signed-off-by: Julien Denize <40604584+juliendenize@users.noreply.github.com>
Signed-off-by: Mickael Seznec <mickael@mistral.ai>
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Mickael Seznec <mickael@mistral.ai>

d8c6210e

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
24 Nov, 2025 1 commit
- [Bugfix][Rocm] Fix shared expert weight loading failure in DeepSeek-MTP (#27563) · 8005e606
  杰兮 authored Nov 24, 2025
```
Signed-off-by: zhyajie <yajizhan@amd.com>
Co-authored-by: zhyajie <yajizhan@amd.com>
```
  8005e606
20 Nov, 2025 2 commits
- [ROCm] Add AMD GPU support on Deepseek v3.2 and SparseMLA (#26670) · 06c20c99
  Pleaplusone authored Nov 20, 2025
```
Signed-off-by: ganyi <ygan@amd.com>
```
  06c20c99
- [ROCm][BugFix] Fix shared expert loading error when disable... · 7218f839
  Pleaplusone authored Nov 20, 2025
```
[ROCm][BugFix] Fix shared expert loading error when disable `VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS` (#28633)
Signed-off-by: ganyi <ygan@amd.com>
```
  7218f839
19 Nov, 2025 3 commits
- [CI] Fix precommit `rope_theta` issue (#29040) · 0075bfff
  Wentao Ye authored Nov 19, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  0075bfff
- [DeepSeek] Fix DeepSeek V3.2 Rope Embedding (#28968) · 88f5b19f
  Yongye Zhu authored Nov 19, 2025
```
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
```
  88f5b19f
- 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
15 Nov, 2025 1 commit
- Add support for Eagle with separate lm-head and embed_tokens layers (#28549) · e439c784
  Eldar Kurtić authored Nov 15, 2025
```
Signed-off-by: Eldar Kurtic <8884008+eldarkurtic@users.noreply.github.com>
```
  e439c784
13 Nov, 2025 1 commit
- 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

[RFC][ROCm][AITER] Keep all AITER kernels in `_aiter_ops` class like... · f080a835

vllmellm authored Nov 10, 2025


[RFC][ROCm][AITER] Keep all AITER kernels in `_aiter_ops` class like `_custom_ops` and `_ipex_ops` (#24490)
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>

f080a835

08 Nov, 2025 1 commit

[Model] Consolidate Deepseek-MoE implementation with DeepSeek-v2 (#28101) · 934a9c3b

Isotr0py authored Nov 08, 2025


Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>

934a9c3b

06 Nov, 2025 1 commit
- [Chore] Clean up deepseek v2/v3 config copy (#28055) · 43ecd0a9
  Isotr0py authored Nov 06, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  43ecd0a9
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

21 Oct, 2025 4 commits

[Deepseek v3.2] Remove extra logics in indexer (#26465) · 09a7e6f6

Lain authored Oct 21, 2025


Signed-off-by: Siyuan Fu <siyuanf@nvidia.com>
Signed-off-by: Daniel Campora <961215+dcampora@users.noreply.github.com>
Signed-off-by: Lain <siyuanf@nvidia.com>
Co-authored-by: Daniel Campora <961215+dcampora@users.noreply.github.com>

09a7e6f6

[Performance] Dual stream execution of "shared_experts" and "selected_experts"... · 344a0017

Alexander Matveev authored Oct 21, 2025


[Performance] Dual stream execution of "shared_experts" and "selected_experts" inside FusedMoE (#26440)
Signed-off-by: Alexander Matveev <amatveev@redhat.com>

344a0017

[Deepseek v3.2] Optimize top_k_per_row (#26763) · 80e94529
Daniel Cámpora authored Oct 21, 2025
```
Signed-off-by: Daniel Campora <961215+dcampora@users.noreply.github.com>
```
80e94529

[Feature][Kernel]FusedMoE LoRA (#21229) · 5f6cbf60

Chen Wu authored Oct 21, 2025


Signed-off-by: wuchen <cntryroa@gmail.com>
Signed-off-by: banjuede <lmklhc@163.com>
Signed-off-by: Chen Wu <cntryroa@gmail.com>
Signed-off-by: Danielle Robinson <dmmaddix@amazon.com>
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Signed-off-by: bk-201 <joy25810@foxmail.com>
Co-authored-by: wuchen <wuchen@zetyun.com>
Co-authored-by: Nathan Van Gheem <vangheem@gmail.com>
Co-authored-by: banjuede <lmklhc@163.com>
Co-authored-by: Danielle Robinson <dmmaddix@amazon.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: bk-201 <joy25810@foxmail.com>

5f6cbf60

18 Oct, 2025 2 commits
- [Chore] Clean up pytorch helper functions in `vllm.utils` (#26908) · 6ac5e06f
  Isotr0py authored Oct 19, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: isotr0py <2037008807@qq.com>
```
  6ac5e06f
- [Misc] Refactor `get_kv_cache_spec` into `AttentionLayerBase` (#26587) · b26b70be
  Nicolò Lucchesi authored Oct 18, 2025
```
Signed-off-by: NickLucche <nlucches@redhat.com>
```
  b26b70be
16 Oct, 2025 1 commit

[ROCm][FEAT] Fuse DeepSeek shared experts into AITER fused_moe ops (#24097) · 13170343

kliuae authored Oct 16, 2025


Signed-off-by: chenjun <junchen2@amd.com>
Signed-off-by: kliuae <kuanfu.liu@embeddedllm.com>
Co-authored-by: valarLip <103567126+valarLip@users.noreply.github.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>