Commits vllm-project/vllm

[CI] Mirror external test assets in vLLM S3 (#52064)

khluu committed 3 days ago

Verified 8eb35c52

[Bugfix][ROCm][CI] Restore the DeepSeek-V4 input GEMM override point (#51821)

stefankoncarevic committed 3 days ago

Verified b369f10d

[Bugfix] Report FULL_ATTENTION for uniform-base UniformTypeKVCacheSpecs groups instead of UNKNOWN (#51218)

yifjiang committed 3 days ago

Verified 5936bac7

[CI] Force source builds for hybrid dependencies (#52043)

AndreasKaratzas committed 3 days ago

Verified caf9e8f7

[ROCm] [bugfix] Chunked prefill paged decode masked load perf (#50017)

afriedri committed 3 days ago

Verified 98f86b9c

[CI Bug] Fix ci moe test (#52009)

yewentao256 committed 3 days ago

Verified 23f360ed

[Build] Update DeepGEMM pin to deepseek-ai nv_dev tip (#52035)

zyongye committed 3 days ago

Verified 025d56a1

[Bugfix] Pin DeepEP by its full commit hash (#52028)

tlrmchlsmth committed 3 days ago

Verified 34735ace

[Bugfix][Kernel] Fix persistent top-k histogram reuse after short rows (#49139)

fxfxfxfxfxfxfxfx committed 3 days ago

Verified e62abc37

[Spec Decode] DSpark confidence-scheduled verification (#47808)

LucasWilkinson committed 3 days ago

Verified 7f7a32cf

[K3 Perf] Flash kda out kernel for prefill, 1.1~1.4x kernel performance improvement (#51311)

yewentao256 committed 3 days ago

Verified fe889ac9

[Bugfix] Initialize DeepGemmQuantScaleFMT oracle lazily; bound QuantFP8 UE8M0 packed path to group_size 128 (#51359)

BabyDrangoner committed 4 days ago

Verified 19d61b12

[Bugfix][ROCm][MoE] Update AITER MXFP4 W4A16 tests to the renamed expert_mask (#51980)

stefankoncarevic committed 4 days ago

Verified d20a031d

[CI Bug] Fix ci qwen3.5 (#52007)

yewentao256 committed 4 days ago

Verified 0fb168e6

[ROCm][K3] Dequantize the fp8 decode query for MLA backends without quant-query support - TRITON_MLA (#51860)

hongxiayang committed 4 days ago

Verified b745d08d

[ROCm] update triton in base docker for gluon compatibility (#51464)

hongxiayang committed 4 days ago

Verified 324f452f

[Bugfix] Bound Anthropic stop sequences (#51997)

taneem-ibrahim committed 4 days ago

Verified 6ea6c426

[Docs] Warn that --api-key does not gate all endpoints (#51999)

russellb committed 4 days ago

Verified 8151f2ad

[Model] Add native Dots3 NOTE multimodal support (#51255)

KurodaKanbei committed 4 days ago

Verified 9035151d

Avoid long-blocking H2D copies in ViT (#51841)

maxyanghu committed 4 days ago

Verified b1b75204

[XPU][CI/Release][2/N] add triton shim in xpu requirements (#51935)

jikunshang committed 4 days ago

Verified 7aa248fc

[Bugfix] Disable fine-grained prefix-cache hits for incompatible hybrid KV layouts (#51843)

mgoin committed 4 days ago

Verified 8e958902

[Bugfix][Multimodal] Invalidate retained PyNvVideoCodec decoder after failure (#51139)

dmai-afk committed 4 days ago

Verified 4c51ceb1

[Bugfix][Frontend] Return 400 for invalid PyNvVideoCodec video input (#51120)

dmai-afk committed 4 days ago

Verified 86e2ab50

[Bugfix] Fix MiniMax-M3 compressed-tensors FP8 MoE SwiGLU params (#46845)

tanpinsiang committed 4 days ago

Verified e60f3c4c

[CI] Add registry layer cache to x86 CPU image build (#51911)

bigPYJ1151 committed 4 days ago

Verified 6accb779

[XPU] Route block-quantized FP8 weights to the W8A8 kernel (#50787)

chaojun-zhang committed 4 days ago

Verified 8c011da6

[Bugfix][XPU] Run GDN attention as eager break under breakable cudagraph (#51928)

ccrhx4 committed 4 days ago

Verified aeece10c

[Model] Support R3 capture with DeepGEMM MegaMoE (#51831)

aoshen02 committed 4 days ago

Verified 4eef91c0

[Attention] Move context_lens_tensor compute into GDN prefill path (#51913)

xyang16 committed 4 days ago

Verified 10b7766a