Commits · eb577e465589bf4394bc4621228483d0ef34704c · OpenDAS / vllm_cscc

07 Oct, 2025 18 commits
- [Bugfix] Add missing sink tensor into flash attn cascade attn implementation (#26325) · eb577e46
  Pei-Lun Liao authored Oct 07, 2025
  
  eb577e46
- [Bug] Fix Shape Validation for Fallback while Enabling E8M0 for DeepGEMM (#26322) · 8f36850f
  Wentao Ye authored Oct 07, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  8f36850f
- [deepseek] add EP8 FusedMOE config for H200 and B200 (#26331) · 29fd2662
  Chen Zhang authored Oct 08, 2025
```
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
```
  29fd2662
- [Model] Lfm2Moe (#26344) · 320feae6
  Paul Pak authored Oct 08, 2025
```
Signed-off-by: Paul Pak <paulpak58@gmail.com>
```
  320feae6
- [V0 Deprecation] Remove `VLLM_USE_V1` from tests (#26341) · 1e4ecca1
  Cyrus Leung authored Oct 07, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  1e4ecca1
- [Misc] Move `LRUCache` into its own file (#26342) · c0a7b89d
  Cyrus Leung authored Oct 07, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  c0a7b89d
- [Model] Add support for ModernBertForTokenClassification (#26340) · 6f59beaf
  antrec authored Oct 07, 2025
```
Signed-off-by: Antoine Recanati Le Goat <antoine.recanati@sancare.fr>
Signed-off-by: antrec <antoine.recanati@gmail.com>
Co-authored-by: Antoine Recanati Le Goat <antoine.recanati@sancare.fr>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
```
  6f59beaf
- [Feature][OCP MX] Support mxfp6 and mixed mxfp6-mxfp4 (#21166) · 41f1cf38
  fxmarty-amd authored Oct 07, 2025
  
  41f1cf38
- [Model] Use `merge_by_field_config` for MM models (Ovis family) (#26308) · 08d26a1b
  Isotr0py authored Oct 07, 2025
```
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
```
  08d26a1b
- Add topk logits torch op for DS3.2. (#25945) · e1098ced
  Daniel Cámpora authored Oct 07, 2025
```
Signed-off-by: Daniel Campora <961215+dcampora@users.noreply.github.com>
Signed-off-by: Daniel Cámpora <961215+dcampora@users.noreply.github.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
```
  e1098ced
- Optimize KV cache distribution for asymmetric pipeline parallelism (#25164) · d100d78e
  Grant Holmes (Ren) authored Oct 07, 2025
```
Signed-off-by: gholmes829 <g.holmes429@gmail.com>
```
  d100d78e
- [Model] Define merge_by_field_config MM interface (R-T) (#26260) · de342585
  Ayush Satyam authored Oct 07, 2025
```
Signed-off-by: Ayush Satyam <ayushsatyam146@gmail.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  de342585
- [responsesAPI][bugfix] serialize harmony messages (#26185) · 185d8ed4
  Andrew Xia authored Oct 07, 2025
```
Signed-off-by: Andrew Xia <axia@meta.com>
Co-authored-by: Ye (Charlotte) Qi <yeq@meta.com>
```
  185d8ed4
- [Deprecation] Deprecate `LLM.set_tokenizer` (#26333) · d9836d45
  Cyrus Leung authored Oct 07, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  d9836d45
- [Model] Define merge_by_field_config MM interface (U-Z) (#26261) · 5f7e8a91
  Ayush Satyam authored Oct 07, 2025
```
Signed-off-by: Ayush Satyam <ayushsatyam146@gmail.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  5f7e8a91
- [BUG] Fix file parsing for load_format runai_streamer_sharded (#26324) · 4dbdf4a2
  ahao-anyscale authored Oct 06, 2025
```
Signed-off-by: ahao-anyscale <ahao@anyscale.com>
```
  4dbdf4a2
- [UX] Support nested dicts in hf_overrides (#25727) · c6873c4e
  Michael Goin authored Oct 06, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  c6873c4e
- [Core] Simplify the Dp padding/should ubatch coordination logic (#25768) · 2111b464
  Sage Moore authored Oct 06, 2025
```
Signed-off-by: Sage Moore <sage@neuralmagic.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
```
  2111b464
06 Oct, 2025 21 commits

[Perf] Add decode full-graph support to FlashInfer-MLA backend (#26313) · f77df946
Benjamin Chislett authored Oct 06, 2025
```
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
```
f77df946
[ROCm] Split AITER unified attention into its own backend (#25507) · f231e5bc
Gregory Shtrasberg authored Oct 06, 2025
```
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
```
f231e5bc
[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

[BugFix] Pad input buffers in _dummy_run (#26209) · f23b4c04

Varun Sundar Rabindranath authored Oct 06, 2025


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

f23b4c04

[Benchmark] Enable MM Embedding benchmarks (#26310) · 44b9af5b
Cyrus Leung authored Oct 07, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
44b9af5b

[Bugfix] Fix gemma3 with transformers backend (#23178) · 7cd95dc8

Raushan Turganbay authored Oct 06, 2025

Signed-off-by: raushan <raushan@huggingface.co>
Signed-off-by: Raushan Turganbay <raushan@huggingface.co>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

7cd95dc8

Add bias handling to CPUFusedMOE kernel (#26289) · c02058c2

Crefeda Rodrigues authored Oct 06, 2025


Signed-off-by: Crefeda Rodrigues <crefeda.rodrigues@arm.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: Crefeda Rodrigues <65665931+cfRod@users.noreply.github.com>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Sharif Inamdar <Sharif.Inamdar@arm.com>
Co-authored-by: Isotr0py <2037008807@qq.com>

c02058c2

[Bugfix][Spec Decode] Fix wrong valid_mask for padded speculation when chunked... · b2ea5ba6

7mile authored Oct 07, 2025


[Bugfix][Spec Decode] Fix wrong valid_mask for padded speculation when chunked prefill occurs (#26231)
Signed-off-by: seven-mile <i@7li.moe>
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
Co-authored-by: Benjamin Chislett <bchislett@nvidia.com>