Commits vllm-project/vllm

Fix CUDA image cache invalidation and reuse published ancestor caches

khluu committed 4 days ago

6e510468

[Bugfix][SleepMode] Exact-size, reusable pinned CPU backup for level-1 sleep (#64630)

aoshen02 committed 4 days ago

Verified deacc367

[KV Offload] Emit self-describing events for recurrent state groups (#62613)

luckyq committed 4 days ago

Verified fac2ce2e

[Bugfix] Preserve multimodal processor-cache hits during request eviction (#64495)

SherifWaly committed 4 days ago

Verified 8472c01d

[Perf][Frontend] Build run-batch TypeAdapters once instead of per request (#63005)

inc-jeong committed 4 days ago

Verified 96c80f8a

[Rust Frontend] Support --root-path (#64222)

BugenZhao committed 4 days ago

Verified a23fd990

[CI] Continue automatic PyTorch Fullgraph sharding (#64635)

Thangnguyenvn98 committed 4 days ago

Verified b8141bd3

[Test][UMBP][6/10] Add embedded runtime and MORI integration coverage (#61794)

charxwu committed 4 days ago

Verified 617135ee

[Rust Frontend] Report the generate response envelope like Python (#64513)

BugenZhao committed 4 days ago

Verified e152c705

[Rust Frontend] Support --max-log-len (#64223)

BugenZhao committed 4 days ago

Verified 515b8753

[XPU] Use GitHub release page index URL for vllm-xpu-kernels (#64633)

jikunshang committed 4 days ago

Verified 00ba900a

[Rust Frontend] Report prompt logprobs on streamed completions echo like Python (#64220)

BugenZhao committed 4 days ago

Verified d8a79457

[Rust Frontend] Support return_weight_versions (#64218)

BugenZhao committed 4 days ago

Verified 52737e03

[Rust Frontend] Export UMBP connector transfer metrics like Python (#64512)

BugenZhao committed 4 days ago

Verified 2b313cd9

[Rust Frontend] Fix bad_words handling in /v1/completions (#48951)

reidliu41 committed 4 days ago

Verified 8bad43e2

[CI] Enroll the remaining LM Eval jobs in automatic sharding (30 jobs) (#64590)

Thangnguyenvn98 committed 4 days ago

Verified 0e34db70

[CI] Enroll the remaining Misc and Samplers jobs in automatic sharding (17 jobs) (#64596)

Thangnguyenvn98 committed 4 days ago

Verified 15a96773

[CI] Enroll the remaining Compile, PyTorch, CUDA and ABI jobs in automatic sharding (29 jobs) (#64595)

Thangnguyenvn98 committed 4 days ago

Verified 4a8dc13a

[CI] Enroll the remaining Distributed, EP and fault-tolerance jobs in automatic sharding (25 jobs) (#64593)

Thangnguyenvn98 committed 4 days ago

Verified c5e37f19

[CI] Enroll the remaining Spec Decode jobs in automatic sharding (21 jobs) (#64592)

Thangnguyenvn98 committed 4 days ago

Verified fc84d05a

[CI] Enroll the remaining Kernels jobs in automatic sharding (30 jobs) (#64591)

Thangnguyenvn98 committed 4 days ago

Verified a9a82229

[CI] Enroll 6 steps and 10 AMD mirrors in automatic sharding (#64569)

Thangnguyenvn98 committed 4 days ago

Verified 336d883f

[Kernel] Add size-gated PDL to silu_and_mul_nvfp4_quant (#64494)

gau-nernst committed 4 days ago

Verified 8b10bda5

[Compile] Make the compile cache device-neutral with compile_on_one_rank (#63247)

kouroshHakha committed 4 days ago

Verified a20280f3

[Rust Frontend] Support hybrid and external DP LB in `vllm-rs serve` (#64451)

BugenZhao committed 4 days ago

Verified 85591000

[Rust Frontend] Let the handshake owner serve a range of DP ranks (#64450)

BugenZhao committed 4 days ago

Verified c20fc8a8

[Bugfix][Rust Frontend] Send the DP rendezvous in coordinated-DP INIT (#64449)

BugenZhao committed 4 days ago

Verified aa85ff7f

[Bugfix] Retry incomplete HTTP response bodies (#62674)

bzsuni committed 4 days ago

Verified 8b9a4341

[Perf][DSv4.1] Bump FlashMLA to pass int64 positions to MegaAttn (#61971)

Juntian777 committed 4 days ago

Verified 53624fc9

[Perf][MoE] Use 32-row tiles in prepare_megamoe_inputs from 1024 tokens (#64467)

Juntian777 committed 4 days ago

Verified 2bfb8f16