vllm-project/vllm

Pull Requests Commits

Bump nvshmem to 3.3.24 and fix CUDA 13 installation (#30149)

dmitry-tokarev-nv committed 52 minutes ago

Verified 799804d1

online fp8 quant with streaming weight post-processing (#29196)

vkuzo committed 1 hour ago

Verified 0d402d26

[TPU] Bump tpu-inference to 0.12.0 (#30221)

jcyang43 committed 1 hour ago

Verified d1b5e7af

Add latent MoE support (#30203)

shaharmor98 committed 3 hours ago

Verified fcd5306f

[MP executor] fix get device count for multi node of mp executor feature (#30042)

weiguihua2 committed 3 hours ago

Verified 398a596e

[Misc] Split the LoRA code (#30253)

jeejeelee committed 4 hours ago

Verified 67312cad

Add evaluate_guards option to DynamicShapesConfig (#27432)

laithsakka committed 5 hours ago

Verified 87aee9ed

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

dcampora committed 6 hours ago

Verified 184076c3

[ROCm] Guard group quant RMS norm fusion patterns (#30239)

yeqcharlotte committed 6 hours ago

Verified eb1051fb

[LoRA] Reduce the loading time of MoE LoRA (#30243)

jeejeelee committed 7 hours ago

Verified 80433e22

Add tip for `mypy` and `markdownlint` to the pre-commit comment (#30259)

hmellor committed 8 hours ago

Verified 5c2433a6

[docs] governance documents (#24801)

simon-mo committed 9 hours ago

Verified 77072e93

[Frontend] Binary embedding response does not return metadata by setting encoding_format to bytes_only. (#30249)

noooop committed 9 hours ago

Verified 2e660c24

[CI] Prevents triggering of an inactive issue/PR check for forked repository. (#29654)

wzshiming committed 10 hours ago

Verified 408cf42f

[Model][7/N] Improve all pooling task | Deprecation as_reward_model. Extract hidden states prefer using new multi-vector retrieval API (#26686)

noooop committed 13 hours ago

Verified 9e77ffca

[Perf] Remove sync point in vit torch sdpa attn backend (#30232)

DamonJiang777 committed 14 hours ago

Verified bcb6f594

[Misc] Rename TensorRT Model Optimizer to Model Optimizer (#30091)

Edwardf0t1 committed 14 hours ago

Verified cd00c443

[Bugfix] fix fuse_allreduce_rms when tp =1 (#30178)

ZJY0516 committed 14 hours ago

Verified d1432712

[ROCm] [Fused Moe EP] Use binary expert mask for aiter fused moe kernel (#29773)

ZhiweiYan-96 committed 15 hours ago

Verified c6df05eb

[BugFix] Unblock use of LoRA with data parallel mode (#30220)

njhill committed 16 hours ago

Verified d726a7b0

Address comment to mergify.yml in #30117 (#30219)

ZhijianJiang committed 17 hours ago

Verified 344b50d5

[responsesAPI][7] Browser, Container MCP tools for non harmony models (#29989)

qandrew committed 19 hours ago

Verified 735284ed

[Frontend] Add MCP type support infrastructure to Responses API (#30054)

daniel-salib committed 19 hours ago

Verified 444f0e3f

[Performance] Fused blockwise quant RMS norm (#27883)

ElizaWszola committed 1 day ago

Verified af0444bf

[BugFix][DeepSeek-V3.2] Fix backend selection logic for Blackwell (#30195)

LucasWilkinson committed 1 day ago

Verified 0044c403

[v1] Add PrefixLM support to FlexAttention backend (#27938)

Isotr0py committed 1 day ago

Verified b952f4d3

[Perf] Deepgemm fused layout kernel for activations, 4.3% throughput improvement, 10.7% TTFT improvement. (#29546)

yewentao256 committed 1 day ago

Verified 541a2ef8

[CI/Build]Temporary workaround for test_default_mm_loras timeout (#30202)

jeejeelee committed 1 day ago

Verified b0f4866a

[Kernel][MoE] optimize `moe_align_block_size` (#29642)

jinzhen-lin committed 1 day ago

Verified 879ddb09

[Misc][Core] Remove unused `req_index` increment in scheduler (#30176)

ivanium committed 1 day ago

Verified 1b0482b9

Older