Commits vllm-project/vllm

[Core][KV events] Report prefix-cache-reused blocks in full report mode (#45261)

GongLei-HW committed 37 days ago

Verified e5588e49

DCP supports hybrid attention (#40996)

Yancey0623 committed 38 days ago

Verified 95ed0fea

[kv_offload] Emit tier-owned BlockStored events from FS/OBJ secondary tiers (#47923)

Change72 committed 38 days ago

Verified 2d814a00

[CI/Build][AMD] Fix ROCm OOM in eagle_correctness_heavy by reserving CUDA graph memory (#47366)

peizhang56 committed 38 days ago

Verified 88e5e2c5

[bugfix] bge-m3-sparse-plugin mismatch requests (#48112)

staugust committed 38 days ago

Verified feb384ad

[ROCm][CI] Move remaining engine/samplers AMD steps to mi325_1 (#48169)

peizhang56 committed 38 days ago

Verified a0f6d767

update marlin M size for EP (#48144)

gnovack committed 38 days ago

Verified f1a5addd

Correct model layer aliasing for Bert style models (#43896)

ap9272 committed 38 days ago

Verified cac3e70c

[CI] Fix cargo-deny config flag ordering (#48170)

LucasWilkinson committed 38 days ago

Verified e12b91b0

[ROCm] Revert Part of `[ROCm] Fix pooling startup workspace lock` #47912 (#48154)

micah-wil committed 38 days ago

Verified 766469a4

[CI] Increase extract hidden states TP2 timeout (#48161)

LucasWilkinson committed 38 days ago

Verified ea0fa34f

[ROCm] Synchronize sparse MLA metadata before graph replay (#47404)

zihaomu committed 38 days ago

Verified bbb0f945

[KV Connector][Mooncake] Apply SWA lookup mask before hashing/key build (#47317)

zhewenl committed 38 days ago

Verified 2ded1b24

[ROCM][DSV32][Perf][MTP] Enable UNIFORM_BATCH CG mode in rocm_aiter_mla_sparse (#45149)

tvirolai-amd committed 38 days ago

Verified b0dec2a1

[Bugfix][MRV2] Reset num_accepted_tokens on add_request in all modes (#48132)

njhill committed 38 days ago

Verified ff8d3488

[KVConnector] MultiConnector: give every sub-connector the request's real blocks in `update_state_after_alloc` (#46865)

deng451e committed 38 days ago

Verified 2285cfca

[Bugfix] Preserve tensor causal metadata for grouped attention (#48135)

LucasWilkinson committed 38 days ago

Verified e08a9151

[ROCm][CI] Set all timeout_in_minutes to 180 (#48146)

charlifu committed 38 days ago

Verified 67e7ea89

[Bugfix] Guard CUDA-only rms_norm_per_block_quant in FUSED_OPS for non-CUDA builds (#47296)

tsvikas committed 38 days ago

Verified 429f4057

Pin PyNvVideoCodec to tested 2.0.4 wheel (#48056)

brandonpelfrey committed 38 days ago

Verified 753c5039

[CI] Annotate built Docker image tags on the Buildkite build page (#48101)

khluu committed 38 days ago

Verified 299d2b56

[Bugfix][Model Runner V2] Order uniform decodes first so spec decodes aren't misclassified as prefills (#47381)

WoosukKwon committed 38 days ago

Verified 85b3a726

Migrate Olmo and Olmo2 to the Transformers modeling backend (#48100)

hmellor committed 38 days ago

Verified b83be00c

Remove PersimmonForCausalLM and FuyuForCausalLM model architectures (#48096)

xianbaoqian committed 38 days ago

Verified 412414d8

[P/D][Bugfix] Fix PD async KV load lookahead handling for MTP spec decode (#46694)

chaunceyjiang committed 38 days ago

Verified ae6170f8

Sanitize server file paths from validation error responses (#46415)

muhammadfawaz1 committed 38 days ago

Verified e8752162

[Bugfix] Fix race condition in KVBlockZeroer (#48085)

benchislett committed 38 days ago

Verified 1cd75b3d

Add Intel XPU Docker release pipeline (#47880)

wendyliu235 committed 38 days ago

Verified 0206f108

Remove TeleChatForCausalLM (#47989)

xianbaoqian committed 38 days ago

Verified ab7961a1

[Bugfix] Fix Qwen3-ASR transcription streaming postprocessing (#42478)

BWAAEEEK committed 38 days ago

Verified a07765c6