Commits vllm-project/vllm

Merge branch 'main' into woosuk/rm-compile-sp

WoosukKwon committed 8 hours ago

Verified 62841aca

Keep all_gather_tensors in the isend_tensor_dict test mock

WoosukKwon committed 8 hours ago

c1b8eb0b

[Platform] Allow BackendEnum dynamic register (#41091)

wangxiyuan committed 8 hours ago

Verified a4b14e57

[Compile] Remove torch.compile-based sequence parallelism and AsyncTP

WoosukKwon committed 8 hours ago

f0d114e0

[Bugfix][ROCm] Keep MLA derived weights in place on weight reload (#60389)

mawong-amd committed 8 hours ago

Verified 9dba339d

[Core] Bound execute_dummy_batch RPC waits by the execute-model timeout (#60398)

NickLucche committed 9 hours ago

Verified 8733a21d

[Bugfix][ToolParser] Stream every tool call when one delta completes several (#60354)

Kylinny committed 9 hours ago

Verified fdfc171a

[Bugfix][Frontend] Do not modify caller params in LLM.score() (#59959)

apex-mochen committed 9 hours ago

Verified 2dd88742

[TEST][XPU][CI] disable tests of unsupported quantization (#60383)

microslaw committed 9 hours ago

Verified 9e3e37cb

[Bugfix][Frontend] Apply output_config and thinking in Anthropic count_tokens (#59180)

ejrtks1020 committed 10 hours ago

Verified 1995c0fc

[CI] Add headroom and selection coverage for H200 fast lanes (#56775)

khluu committed 10 hours ago

Verified 1283362e

[Perf] JIT monitor: hook TileLang JITImpl.compile instead of __call__ (#58685)

NickLucche committed 10 hours ago

Verified a58bdd0d

[Bugfix] [Spec Decode] DeepSeek V4/V4.1 NVFP4: keep ignored MTP/DSpark experts on MXFP4 (#59655)

v-shobhit committed 10 hours ago

Verified 8faac3dd

[Bugfix][Frontend] Fix Responses namespace aliases in named tool calling (#59723)

YueshenZ committed 10 hours ago

Verified ed825aaf

[Bugfix] Validate tool-parser/tokenizer compatibility at startup (#59749)

www6v committed 10 hours ago

Verified 38dc8ee5

[Bugfix][Parser] Emit buffered post-reasoning text when no tool parser is configured (#58911)

dhairyashilRG committed 11 hours ago

Verified d6fe5dca

[CI/Build][NVIDIA] Build NIXL with NIXL EP from source in Rubin image (#60090)

arpera committed 11 hours ago

Verified 4ac5a36a

[Bugfix][DSv4.1] Skip SWA bounded replay for KV loads that carry the window (#59197)

ivanium committed 12 hours ago

Verified ef63c23d

[Docs] Add disagg PD flow-example guidance to /pr-checklist skill (#60111)

NickLucche committed 12 hours ago

Verified d169c5a9

[quantization] Add per-token NVFP4 quantization MoE support for ReLU2 (#56740)

xuantengh committed 12 hours ago

Verified f38c6778

[Model] Streamline EmbeddingGemma 2 config resolution and harmonize Triton prefill attention (#60289)

lucianommartins committed 12 hours ago

Verified b803edbc

[CI] Fix failing test_deepseek_v4_rocm_dspark* tests (#60462)

simondanielsson committed 13 hours ago

Verified 947dd62c

[CI] Stamp provenance labels on the torch-nightly image so Initialized Snapshot E2E can run (#59487)

Thangnguyenvn98 committed 13 hours ago

Verified b47ec440

[Bugfix] Limit GlmMoeDsaForCausalLM fp8 KV cache default to SM100 (#60281)

mgoin committed 13 hours ago

Verified a982c81e

[Bugfix][KV Offload] Restore KVCR construction through the secondary-tier factory (#58088)

ai-yang committed 14 hours ago

Verified 08567505

[UX] Skip checkpoint shards an MTP head does not load (#60152)

robertgshaw2-redhat committed 15 hours ago

Verified e572e02d

[KV Offload] Update KVCR adapter to pool-and-index API (#59899)

mkhazraee committed 15 hours ago

Verified 17c70c89

[Bugfix] Respect Inductor deterministic mode in combo-kernel defaults (#59074)

cheese-cakee committed 15 hours ago

Verified b9767299

[Bugfix][Quantization] Accept folded NVFP4 scales for Humming A16 (#60337)

jinzhen-lin committed 16 hours ago

Verified 32fbfa15

[Spec Decode] Rename AR speculators to StandaloneAR / TargetDependentAR (#60335)

TheEpicDolphin committed 16 hours ago

Verified 6ee9be6b