vllm-project/vllm

Pull Requests Commits

yewentao256 committed 3 hours ago

c1879c1f

yewentao256 committed 3 hours ago

7e73d268

optimize sampled_token_ids using numpy and remove tolist

yewentao256 committed 4 hours ago

c7201f5b

add mixed precision support for modelopt (#35047)

sychen52 committed 4 hours ago

Verified d0105b84

Nemotron: use per-layer config in NemotronHMLPDecoderLayer for heterogeneous models (#35396)

danielafrimi committed 4 hours ago

Verified 832a780f

[Performance] Extract KV cache update op from flashinfer forward (#35422)

ElizaWszola committed 5 hours ago

Verified 98217b09

fix(reasoning): Qwen3ReasoningParser returns truncated output as reasoning (#35230)

stakeswky committed 5 hours ago

Verified 967572dd

[Model Runner V2] Prepare attn metadata in ModelState [2/N] (#35383)

WoosukKwon committed 6 hours ago

Verified 3d66502e

[Model Runner V2] Add model states [1/N] (#35350)

WoosukKwon committed 7 hours ago

Verified c66aa48e

[Model Runner V2] Fix error-handling (#35063)

njhill committed 7 hours ago

Verified b6d5a172

[Bugfix] Remove erroneous lower bound on LoRA vocab size constraint (#35354)

LucasWilkinson committed 7 hours ago

Verified 5e58bdc7

[BugFix] Align fused MoE-LoRA kernel config with actual weight shapes (#34396)

RunkaiTao committed 8 hours ago

Verified a1f53add

[Refactor] Remove dead code for attention benchmark script (#35418)

yewentao256 committed 8 hours ago

Verified 05970c77

[Core] Support `min_tokens` with speculative decoding (#32642)

qianlihuang committed 8 hours ago

Verified d9406076

[Perf] Optimize maxsim scores computation for pooling models, 13.9% E2E throughput improvement (#35330)

yewentao256 committed 9 hours ago

Verified 99c7892c

Add GlmOcrConfig for GLM-OCR model type recognition (#34982)

hujia177 committed 9 hours ago

Verified ec8f943d

[BugFix][kv_offload]: Fix kernel block size detection (#35125)

orozery committed 10 hours ago

Verified f2ad952f

[ROCm] Update the torch version in rocm_build.txt to use the official 2.10 release (#34387)

SageMoore committed 10 hours ago

Verified 9e2cabdf

[ROCm] Add dynamic mxfp4 quantization for DeepSeek V2 projection layers (#34157)

dllehr-amd committed 10 hours ago

Verified ec8ab9d2

[Refactor] Remove dead or duplicate func utils or variables (#35318)

yewentao256 committed 10 hours ago

Verified 05972ea7

[Model] Add nvidia/llama-nemotron-embed-vl-1b-v2 multimodal embedding model (#35297)

jzakrzew committed 12 hours ago

Verified 111d8690

[Bug] Fix missing <think> tag after tool call in MiniMax 2.1 (#35352)

stingoChen committed 12 hours ago

Verified 7fea7250

[Misc] Standardize handling of `mm_processor_kwargs.size` (#35284)

DarkLight1337 committed 13 hours ago

Verified 845ee348

[Bugfix] Fix uint32 overflow in Mamba selective scan state pointer arithmetic (#35275)

Josephasafg committed 14 hours ago

Verified ec13e549

[Bugfix] fix device_name for routing replay (#34336)

Li-Yongwen committed 14 hours ago

Verified c6ca5159

[Bugfix] Fix Qwen2.5-Omni and Qwen3-Omni mixed-modality embed regression (#35368)

linyueqian committed 14 hours ago

Verified c0615a29

Remove `bc-lint` (#35274)

hmellor committed 15 hours ago

Verified 01914445

[XPU] use fixed UMD version in dockerfile.xpu (#35392)

jikunshang committed 15 hours ago

Verified 5281713e

[Bugfix] [Qwen3.5]Fix Qwen3.5 FP8 quantization: tuple shard_id weight loading (#35289)

Alibaba-HZY committed 16 hours ago

Verified 32693db8

[Hardware][Powerpc]Enable prefix caching and chunked prefill for ppc64le (#35081)

Akashcodes732 committed 16 hours ago

Verified e03ddcfb

Older