vllm-project/vllm

Pull Requests Commits

[Docs] Bring README updates into docs README (#39397)

hmellor committed 38 days ago

Verified d87fb264

[Frontend][4/n] Improve pooling entrypoints | pooling. (#39153)

noooop committed 38 days ago

Verified 66c079ae

[CI] fix possible user permission issues in nightly index generation (#39390)

Harry-Chen committed 38 days ago

Verified b6c9be50

Fix NUMA binding on non-CDMM Grace-Blackwell systems (#39361)

soodoshll committed 38 days ago

Verified ed733802

[ROCm] Remove unnecessary fp8 roundtrip in gather cache NHD dequant (#39122)

Bortlesboat committed 38 days ago

Verified 8a34c508

[Perf] Optimize redundant sync for pooling model, 3.7% Throughput Improvement (#39113)

yewentao256 committed 38 days ago

Verified ed2f282b

[Docker] Add fastsafetensors to NVIDIA Dockerfile (#38950)

zhewenl committed 38 days ago

Verified 9e785557

[XPU] Skip VLLM_BATCH_INVARIANT for XPU in EAGLE DP test (#39164)

1643661061leo committed 38 days ago

Verified e80e6339

[Multimodal] Fix nested_tensors_equal: add length check for lists and tuple support (#38388)

khairulkabir1661 committed 38 days ago

Verified 490f17d0

[Refactor] Improve indexer decode path metadata preparation (#38865)

zyongye committed 38 days ago

Verified 2e984060

[PD][HeteroArch]Fix accuracy issue with CPU_ATTN as Decoder and Flash_ATTN as prefiller (#38935)

xuechendi committed 38 days ago

Verified ef5a2268

[CI] Fix mypy for `vllm/v1/ops` (#39219)

yewentao256 committed 38 days ago

Verified aec18492

Fix Responses JSON schema alias serialization (#38519)

noobHappylife committed 38 days ago

Verified 2a49284c

[Model] Update ColModernVBERT to support latest HF checkpoint (#39307)

ieBoytsov committed 38 days ago

Verified d37b3787

[Bug] Fix routing bias dtype for trtllm per-block fp8 moe (#38989)

wzhao18 committed 38 days ago

Verified 92fbec39

[Bugfix] Fix cpu-offload-gb assertion with non-default block sizes (#36461)

AjAnubolu committed 38 days ago

Verified 2f41d6c0

[Gemma4] Support quantized MoE (#39045)

dsikka committed 38 days ago

Verified 3aecdf08

[UX] Integrate DeepGEMM into vLLM wheel via CMake (#37980)

mgoin committed 38 days ago

Verified eb4205fe

[XPU][UT] update UTs in CI (#39296)

zhenwei-intel committed 38 days ago

Verified 83aea214

[W8A8 Block Linear Refactor][2/N] Remove W8A8Fp8BlockLinearOp and adopt Fp8 block linear kernel selections. (#33892)

maralbahari committed 38 days ago

Verified 2e9034c9

[Bugfix] FlashInfer MXINT4 MoE crashes, missing do_finalize (#39315)

benchislett committed 38 days ago

Verified 8332078c

[torch.compile] Allow usage of Opaque Objects in PyTorch 2.11 (#39286)

zou3519 committed 38 days ago

Verified ba4a78eb

[Bugfix]Fix EP precision for Qwen3.5, Qwen3-Next (#39181)

USTCKAY committed 38 days ago

Verified f3c7941e

[CI Bug] Fix pre-commit issue in main (#39347)

yewentao256 committed 38 days ago

Verified 3352bf8b

[BugFix] --max-model-len=-1 causes over-limit requests to hang and starve the entire service (#39102)

triangleXIV committed 38 days ago

Verified 7c94ae16

`tests/v1/e2e/spec_decode`: assert async scheduling is used (#39206)

puririshi98 committed 38 days ago

Verified ad05edfb

[Feature] Batch invariant nvfp4 linear support (#39322)

yewentao256 committed 38 days ago

Verified 20181372

[MoE] Move DEEP_GEMM into experts/ subdirectory (#39005)

Jackmin801 committed 38 days ago

Verified a776a48b

[Tool] `adjust_request` to reasoning parser, and Gemma4 fixes (#39027)

bbrowning committed 38 days ago

Verified 8477fe42

[Attention] relax the head dim 512 and paged kv for sm90+FA4 (#38835)

IwakuraRein committed 38 days ago

Verified e24e0a43

Newer Older