Commits · cbb799e314178abf2fa99e7306087198c5f43580 · OpenDAS / vllm_cscc · GitLab

12 Nov, 2025 1 commit
- [Model][Qwen3VL] Simplify `get_mrope_input_positions` using numpy (#28302) · cbb799e3
  Lukas Geiger authored Nov 12, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  cbb799e3
11 Nov, 2025 4 commits
- [LoRA][1/N]Remove LoRA extra vocab (#28382) · 9d1c4747
  Jee Jee Li authored Nov 12, 2025
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  9d1c4747
- [Model] Pass `mm_features` directly into `get_mrope_input_positions` (#28399) · afffd3cc
  Cyrus Leung authored Nov 11, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  afffd3cc
- [Attention] Refactor CUDA attention backend selection logic (#24794) · b30dfa03
  Matthew Bonanni authored Nov 11, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
```
  b30dfa03
- [Model][Qwen3VL] Slighly speedup `fast_pos_embed_interpolate` (#28434) · 9973e6e0
  Lukas Geiger authored Nov 11, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  9973e6e0
10 Nov, 2025 1 commit
- [V0 Deprecation] Remove unused `context_len` and `seq_len` from M-RoPE (#28395) · d0e186c1
  Cyrus Leung authored Nov 11, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  d0e186c1
07 Nov, 2025 1 commit
- [Core][MM] Add mechanism to configure multimodal fields which should stay on CPU (#28168) · e0919f33
  Lukas Geiger authored Nov 07, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  e0919f33
29 Oct, 2025 1 commit

[Model] Fix Qwen3VL and Qwen3Omni after torch.compile changes (#27705) · 0d8161b0

Lukas Geiger authored Oct 29, 2025


Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Roger Wang <hey@rogerw.io>

0d8161b0

27 Oct, 2025 1 commit
- [Model] Use merge_by_field_config for MM models (Qwen series) (#27546) · cbd5e07a
  Cyrus Leung authored Oct 27, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  cbd5e07a
25 Oct, 2025 1 commit

[CI/Build] Refactor processing tests (#27470) · 66a168a1

Cyrus Leung authored Oct 26, 2025


Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>

66a168a1

24 Oct, 2025 1 commit
- [MM][Bugfix] Replace `PatchEmbed`'s conv3d to linear layer (#27418) · 42efe609
  Isotr0py authored Oct 24, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Roger Wang <hey@rogerw.io>
```
  42efe609
22 Oct, 2025 1 commit
- [Bugfix] Make `get_mrope_input_positions` instance methods (#27342) · 14e2f123
  Cyrus Leung authored Oct 22, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  14e2f123
21 Oct, 2025 1 commit
- [MM][Core] Decouple ViT backend from LM backend (#27061) · c3a2c6ac
  Roger Wang authored Oct 21, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
```
  c3a2c6ac
19 Oct, 2025 1 commit
- [Misc] Move utils to avoid conflicts with stdlib, and move tests (#27169) · d31f7844
  Cyrus Leung authored Oct 19, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  d31f7844
17 Oct, 2025 1 commit
- [bugfix] Qwen3-VL fix video incorrect timestamp calculations while do_sample_frames=True (#27104) · 4c91a28e
  燃 authored Oct 18, 2025
```
Co-authored-by: 松灵 <wpf272043@alibaba-inc.com>
```
  4c91a28e
16 Oct, 2025 2 commits
- [Model] Fix Qwen3VL mm mapping (#27027) · 9f4e3090
  Jee Jee Li authored Oct 17, 2025
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  9f4e3090
- [Chore] Separate out `vllm.utils.collections` (#26990) · d2740faf
  Cyrus Leung authored Oct 16, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  d2740faf
14 Oct, 2025 1 commit
- [Bugfix] Standardize merging multimodal embeddings (#26771) · d2f816d6
  Cyrus Leung authored Oct 14, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  d2f816d6
12 Oct, 2025 2 commits
- [Models][Qwen3VL] Speedup `fast_pos_embed_interpolate` (#26647) · a6049be7
  Lukas Geiger authored Oct 12, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  a6049be7
- 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][Qwen3VL] fix deepstack in qwen3vl (#26626) · 9d6cff3e

JJJYmmm authored Oct 11, 2025


Signed-off-by: liuye.hj <liuye.hj@alibaba-inc.com>
Signed-off-by: JJJYmmm <92386084+JJJYmmm@users.noreply.github.com>
Co-authored-by: liuye.hj <liuye.hj@alibaba-inc.com>

9d6cff3e

[Refactor]: Use M-RoPE interface directly while defining model class instead... · 727144be

dsinghvi authored Oct 11, 2025


[Refactor]: Use M-RoPE interface directly while defining model class instead of maintaining model specific M-RoPE implementation in mrope.py (#24172)
Signed-off-by: Divyansh Singhvi <divyanshsinghvi@gmail.com>
Signed-off-by: dsinghvi <divyanshsinghvi@gmail.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: wwl2755 <wangwenlong2755@gmail.com>

727144be

10 Oct, 2025 1 commit

[Model][Qwen3VL] Compute `cu_seqlens` on CPU to remove (#26496) · b2155ed3

Lukas Geiger authored Oct 10, 2025


Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

b2155ed3

09 Oct, 2025 2 commits
- [Models][Qwen] Replace `pad` with `cat` for better performance (#26486) · 2c1c7dfb
  Lukas Geiger authored Oct 09, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  2c1c7dfb
- [Models][Qwen3VL] Optimise `_validate_and_reshape_mm_tensor` (#26426) · 0426e3c5
  Lukas Geiger authored Oct 09, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  0426e3c5
08 Oct, 2025 1 commit
- [Models] Improve iteration over layers (#26425) · 93f2c0aa
  Lukas Geiger authored Oct 08, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  93f2c0aa
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
04 Oct, 2025 1 commit
- [Bugfix] Fix qwen3 vl dummy data generation with overrides (#26193) · 67bc0c00
  Roger Wang authored Oct 03, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
```
  67bc0c00
03 Oct, 2025 3 commits
- [Multi Modal] Configurable MM Profiling (#25631) · 79aa2446
  Wenlong Wang authored Oct 03, 2025
```
Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  79aa2446
- [ROCm] [VL] [Bugfix] Fix vit flash attn dispatcher logic for ROCm (#26104) · 9c5ee91b
  TJian authored Oct 02, 2025
```
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
```
  9c5ee91b
- [Attention] Move Backend enum into registry (#25893) · 2aaa4238
  Matthew Bonanni authored Oct 02, 2025
```
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
```
  2aaa4238
01 Oct, 2025 1 commit
- [MM] Add text-only mode for Qwen3-VL (#26000) · 66bca9b8
  Roger Wang authored Sep 30, 2025
  
  66bca9b8
29 Sep, 2025 2 commits
- [V0 Deprecation][Models] Remove all V0 condition for mm embeddings merge (#25331) · bd51f78e
  Isotr0py authored Sep 29, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: isotr0py <2037008807@qq.com>
```
  bd51f78e
- [Bugfix] Fallback ViT attn backend to SDPA for blackwell (#25851) · 65ecb4f1
  Roger Wang authored Sep 28, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
```
  65ecb4f1
28 Sep, 2025 1 commit

[VLM] Update Qwen3-VL max_num_video_tokens calculation for configurable video profiling (#25557) · 0efd540d

Isotr0py authored Sep 28, 2025


Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Roger Wang <hey@rogerw.io>

0efd540d

27 Sep, 2025 2 commits

[Bugfix] Merge MM embeddings by index instead of token IDs (#16229) · 27d7638b

Cyrus Leung authored Sep 27, 2025


Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Signed-off-by: NickLucche <nlucches@redhat.com>
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: NickLucche <nlucches@redhat.com>
Co-authored-by: Roger Wang <hey@rogerw.io>

27d7638b

[Bugfix] Allow Only SDPA Backend for ViT on B200 for Qwen3-VL (#25788) · c242c980
Wentao Ye authored Sep 26, 2025

c242c980

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
25 Sep, 2025 2 commits
- [Bugfix] Fix Qwen3-VL max_num_video_tokens calculation for video profiling (#25648) · 17b4c668
  Isotr0py authored Sep 25, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  17b4c668
- [Misc] Fix Qwen3-VL `video_grid_thw` typing (#25646) · 7be9ffcd
  Roger Wang authored Sep 25, 2025
```
Signed-off-by: Roger Wang <hey@rogerw.io>
```
  7be9ffcd