vllm-project/vllm

Pull Requests Commits

[Kernel] Move Marlin lock storage to WorkspaceManager

mgoin committed 90 days ago

49cdd64d

[ROCm][Quark][7/N] Use MXFP4 linear kernel abstraction for `emulation` backend (#48949)

fxmarty-amd committed 91 days ago

Verified e67a2e0a

[Compressed-Tensors] Support Kimi-K3 quantized models (#50500)

kylesayrs committed 91 days ago

Verified d87d2ca7

[ROCm][CI] Restore Mistral tool-parser compatibility after unification (#50515)

AndreasKaratzas committed 91 days ago

Verified 8d8a4e0f

Upgrade tpu-inference to v0.26.0 (#50522)

meiyeh123 committed 91 days ago

Verified 7c08664f

Update torch version to 2.13.0+cpu (#50412)

ylangtsou committed 91 days ago

Verified 10ad649d

[Doc] Add BgeM3EmbeddingModel to embedding supported models (#50571)

LG-0927 committed 91 days ago

Verified c036cb25

[Bugfix] Universally align block table width to 128 tokens (#50302)

MatthewBonanni committed 91 days ago

Verified a0cd2b69

[Bugfix] Don't transpose fused MoE quantization scales in `RoutedExperts.load_weights` (#50137)

hmellor committed 91 days ago

Verified 94e9ef07

[DSv4 Perf] Optimize workspace reuse for eager break, 3.9% E2E TTFT improvement. (#49236)

yewentao256 committed 91 days ago

Verified df71917c

[ROCm][CI] Fall back to lossless Kimi K3 MXFP4 emulation on gfx942 (#50516)

AndreasKaratzas committed 91 days ago

Verified 5233368d

[2/N][Attention] Enable masked MHA for sparse MLA prefills (#48770)

MatthewBonanni committed 91 days ago

Verified 82ae4164

[Kernel][CI] `--jit-monitor-mode error` e2e tests for kernel warmup infra (#50109)

NickLucche committed 91 days ago

Verified 864a87fe

[CI] Remove default_torch_num_threads workaround from llava-onevision-transformers test (#50560)

oguzhankir committed 91 days ago

Verified 7fdc1ab2

[Bugfix][Frontend] Raise VLLMValidationError for user-facing errors in chat_utils.py (#50491)

latent-9 committed 91 days ago

Verified 0b5b49dd

[Renderer] Warm up the renderer properly. (#50408)

noooop committed 91 days ago

Verified f5ffc59b

[KV Connector] Add per-layer canonical KV page mappings for parallelism-agnostic offload (#48408)

Etelis committed 91 days ago

Verified 2c4d3488

[Attention]: Use KVCacheSpec for AttentionMetadataBuilder type hints (#50148)

hickeyma committed 91 days ago

Verified b2fb83e7

K3 DSpark AR fusion (#50242)

jeejeelee committed 91 days ago

Verified 92643d68

[Misc] Clarify mono audio requirement (#50141)

NickLucche committed 91 days ago

Verified 17beffd5

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941)

bnellnm committed 91 days ago

Verified 6e311c6e

[ROCm][CI] Update Transformers AR+RMS fusion expectation (#50517)

AndreasKaratzas committed 91 days ago

Verified c911120a

[chore] clean-up weight prepack for INT8 MoE (#50116)

fadara01 committed 91 days ago

Verified 0e9b5007

[CPU][s390x] Optimize inference perf and add oneDNN INT8 GEMM for s390x (#50219)

R3hankhan123 committed 91 days ago

Verified 88bc8fb5

[XPU] Unify XPU RMSNorm kernels with vllm_c and drop redundant XPU-specific implementation (#46981)

chaojun-zhang committed 91 days ago

Verified 482cfc20

[UT] add skipif for rocm aiter sampler UT (#50530)

mayuyuace committed 91 days ago

Verified 5d7647a1

[CI] Add M3 MSA tests to CI (#49143)

gau-nernst committed 91 days ago

Verified 34bb795f

[Bugfix] Re-land MiniMax M3 default video processor (#50305)

taneem-ibrahim committed 91 days ago

Verified f727951d

[CPU][BugFix] Remove redundant kv cache write (#50437)

fadara01 committed 91 days ago

Verified 10e6b400

Fix duplicate HunyuanVL image boundary tokens (#49691)

Mi-Jiazhi committed 91 days ago

Verified 3ee2bd13

Older