Commits vllm-project/vllm

perf: add push-based allreduce for small tensor reductions

alexm-redhat committed 1 day ago

297ba8e4

[LoRA][Bugfix] Dedup LoRA wrapping for modules referenced from multiple attribute paths (MoE gate) (#42757)

jeejeelee committed 25 days ago

Verified 32b71779

fix: add API key authorization to /v2 endpoints (#42594)

dusthunter committed 25 days ago

Verified 39c67d71

[Misc] Add common random prefix option to structured-output serving benchmark (#41632)

viktorpusTT committed 25 days ago

Verified 87a2adcb

[Bugfix] Respect explicit --kv-cache-dtype over checkpoint kv_cache_scheme (#42782)

mgoin committed 25 days ago

Verified 852f5674

[Kernel][UX] Add `--linear-backend` arg for linear kernel selection (#39538)

mgoin committed 25 days ago

Verified b2a27b82

[Bugfix] Unwrap VLM wrappers for EPLB on Model Runner V2 (#42706)

JasonKeyiL committed 25 days ago

Verified d0921baf

[Bugfix] Fix layerwise reload alias-buffer corruption (#42481)

rasdani committed 25 days ago

Verified 1ccdf875

[ROCm][Bugfix] Fix fused_mla_dual_rms_norm for AITER API rename _fused_qk_rmsnorm (#42606)

rbrugaro-amd committed 25 days ago

Verified bd9dbe60

[Build] Switch CUDA 12.9 wheel builds to PyTorch manylinux_2_28 base (#41668)

mgoin committed 25 days ago

Verified de2d76f3

Add HumanEval and GSM8K benchmarks to datasets (#42648)

southfreebird committed 25 days ago

Verified 9a7a273d

[FlashAttn] Fix supports_kv_cache_dtype() accepting unhandled fp8 kv-cache dtype variants (#42685)

liulanze committed 25 days ago

Verified b2c58ee9

[ROCm] Restore fast top_k_per_row kernels for sparse MLA when topk_tokens=2048 (#42072)

frida-andersson committed 25 days ago

Verified 4d67d3bd

[Bugfix] Fix SM121 (DGX Spark) exclusion from Marlin/CUTLASS FP8 paths (#35568)

blake-snc committed 25 days ago

Verified 06d020bb

[LMCacheMPConnector] Prioritize importing the lmcache_mp_connector from lmcache (#42596)

chunxiaozheng committed 25 days ago

Verified f45c2108

[ROCm] Widen AITER fused AR RMSNorm 1-stage gate (#42409)

akii96 committed 25 days ago

Verified be7a03ea

[Model Runner v2] Support reload weights (sleep mode) (#42673)

yewentao256 committed 25 days ago

Verified 6147c702

[Model Runner V2] FP32 gumbel sampling. (#41775)

PatchouliTIS committed 25 days ago

Verified 01625966

[ROCm][MLA] FP8 ASM prefill for AITER dense MLA backend on gfx950 (#42509)

maeehart committed 25 days ago

Verified 46a95815

[Perf] Set IR Op Priority Once at Worker Init (#42631)

BadrBasowid committed 25 days ago

Verified fb5bd03f

[Bugfix] Fix DeepGEMM context lens contiguity in MLA indexer (#42135)

mmangkad committed 25 days ago

Verified ee58665a

[Perf] Optimize MLA attention `_v_up_proj` bmm by removing additional copy (#42561)

yewentao256 committed 25 days ago

Verified 491e8d85

[Model Runner V2] Fix kv_connector `pre_forward` order (#42676)

yewentao256 committed 25 days ago

Verified af9616d8

[ROCm] Widen OAI Triton MoE capability range to include gfx12 (RDNA4) (#37826)

laudney committed 26 days ago

Verified d792d993

[Feat][RL] IPC weight sync optimizations: multigpu support and chunked packed tensors (#37476)

hao-aaron committed 26 days ago

Verified e0a45f14

[Bugfix] DFlash FP8 KV-Cache (#42692)

benchislett committed 26 days ago

Verified 0fe75502

[Bugfix] Ensure embeding model compilation on CPU (#42709)

bigPYJ1151 committed 26 days ago

Verified 95cfe102

gemma3 multi-gpu bug-fix (#42630)

pmaybank committed 26 days ago

Verified 1dc3fe08

fix: propagate revision/code_revision pins to all artifact boundaries (#42616)

jperezdealgaba committed 26 days ago

Verified d26a28ab

[ROCm][CI] Stage B gating (#42025)

AndreasKaratzas committed 26 days ago

Verified d735968f