Commits vllm-project/vllm

parser

ywang96 committed 57 days ago

1cbbf14c

[Bugfix] Redo Qwen3.5/Qwen3-Next GDN projector fusion (#34697)

Isotr0py committed 57 days ago

795bd0b2

[Bugfix] Fix prefix creation for Qwen3.5 (#34723)

mgoin committed 57 days ago

0f828e0f

Revert "[Models] Fuse Qwen3.5 GDN's qkvz_proj and ba_proj" (#34683)

ZJY0516 committed 57 days ago

4c5cc4aa

[Models] Fuse Qwen3.5 GDN's qkvz_proj and ba_proj (#34492)

Isotr0py committed 57 days ago

4f5a4e81

[KV Connector] Add temporary, off-by-default `VLLM_DISABLE_REQUEST_ID_RANDOMIZATION` workaround (#34415)

eicherseiji committed 61 days ago

Verified 79c7e092

[Bugfix] Handle num_expert_group=None in flashinfer block-scale FP8 MoE (#34494)

haosdent committed 61 days ago

Verified 79f3fab0

[BUGFIX] Fix accuracy regression for NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4 with TP>1 (#34476)

vadiklyutiy committed 61 days ago

Verified 604b9eae

[bugfix] Fix critical bug when reporting for all paths where handler.create_error_response is used (#34516)

kizill committed 61 days ago

Verified 50dbd6c9

[CI][Entrypoints] Validate detokenize token IDs to prevent int64 overflow causing 500 (#34468)

AndreasKaratzas committed 61 days ago

Verified 98bcc6ca

[Kernels] Fix Helion GPU utils to use platform-agnostic device name API (#34537)

AndreasKaratzas committed 61 days ago

Verified f13e86d8

[Model Runner V2] Minor cleanup for Sampler (#34563)

WoosukKwon committed 61 days ago

Verified 9ca768c7

[Hybrid] Enable mamba prefix cache "align" mode with async scheduling (#33997)

tdoublep committed 61 days ago

Verified d5fe3f70

[Renderer] Move InputPreprocessor into Renderer (1/2) (#34510)

DarkLight1337 committed 61 days ago

Verified 73391a1b

[ROCm][CI] Guard sparse MLA backend imports for ROCm compatibility in tests (#34538)

AndreasKaratzas committed 61 days ago

Verified b3c14229

[Bugfix] Fix Qwen3.5 config loading (#34554)

ywang96 committed 62 days ago

Verified 2f186635

[Misc] Update tests and examples for Prithvi/Terratorch models (#34416)

christian-pinto committed 62 days ago

Verified 342a7cda

[new model] add COLQwen3 code & Inference (#34398)

craftsangjae committed 62 days ago

Verified d1ea65d0

[CI] Heavy refactoring of Voxtral multimodal audio model tests (#34294)

AndreasKaratzas committed 62 days ago

Verified de42abb3

Add explicit validation error for tool calls. (#34438)

juliendenize committed 62 days ago

Verified 60ca7981

fix: use `__annotations__` instead of `get_type_hints()` for dynamic `kwargs` detection (#34527)

perone committed 62 days ago

Verified 0ef5b914

[bug] Make sure get_modality_with_max_tokens is deterministic (#34533)

842974287 committed 62 days ago

Verified ed242652

[Feature][Perf] Support Selective CPU Weight Offloading (#34535)

wzhao18 committed 62 days ago

Verified b37b6797

[Bugfix] Fix ROCm UVA CPU weight offloading broken by #32993 (#34543)

AndreasKaratzas committed 62 days ago

Verified a0638d05

[Hybrid] Enable spec decoding in mamba cache align mode (#33705)

peakcrosser7 committed 62 days ago

Verified c027541e

[Bugfix]: Fix structured output in multi-turn gpt-oss (#34454)

bbrowning committed 62 days ago

Verified fd267bc7

Revert "[Bugfix] Fix fused MoE IMA (sans chunking) by using int64 for strides" (#34530)

mgoin committed 62 days ago

Verified bfaa5593

[Misc] vLLM's --enforce-eager should turn off compile and cudagraphs only (#34523)

zou3519 committed 62 days ago

Verified 87789c83

[Bugfix] Replace c10::optional with std::optional in topk kernel (#34467)

FloatingVertex committed 62 days ago

Verified bcd65c1f

[Feature] Support CPU Offloading without Pytorch Pinned Memory that leads to doubled allocation (#32993)

wzhao18 committed 62 days ago

Verified 59d53066