Commits vllm-project/vllm

[Rust Frontend] Run utility calls to completion when the caller is cancelled

BugenZhao committed 4 days ago

Verified 9c78e6bb

[Rust Frontend] Move the utility call fan-out onto ClientInner

BugenZhao committed 4 days ago

Verified cba227e9

[Bugfix][KV Offload][P2P] Stop DEALER monitor before closing its reader (#63207)

vMaroon committed 4 days ago

Verified f4fd1ab6

[CI] Move Attention, Quantization and MoE kernel tests and their AMD mirrors from fixed to automatic shards (#62077)

Thangnguyenvn98 committed 4 days ago

Verified 5ea83019

[Bugfix][KV Offload] Count bounded-replay SWA allocations in the lazy walk target (#64229)

Juntian777 committed 4 days ago

Verified 77f58e53

[Bugfix][Multimodal] Fix DeepSeek-OCR zero-sum patch guard in _encode… (#64091)

nrenvo committed 4 days ago

Verified 76c4ad2f

[Bugfix][Benchmark] Restore --input-len/--output-len/--prefix-len in vllm bench throughput (#64227)

subhashpolisetti committed 4 days ago

Verified deb1a97e

[Bugfix][DSA] Reserve the decode indexer logits in the memory profile (#62860)

gcanlin committed 4 days ago

Verified 317081fb

[KV Connector][NIXL] Select push mode for NixlConnector via transfer_mode (#64230)

Juntian777 committed 4 days ago

Verified 066d9799

[ROCm][MiniMax-M3.1] Fuse MXFP8 activation quant into the Gemma input_layernorm for qkv_proj (#63637)

RuibinCheung committed 4 days ago

Verified 1cb31be7

[Bugfix][Spec Decode] Align adaptive draft prefill graph padding (#62796)

amenomori committed 5 days ago

Verified 5118d990

[Frontend] Support /v1/completions in run-batch (#64097)

paolo-perrone committed 5 days ago

Verified 9fc019a7

[Perf][DSv4.1] Tile the NVFP4 compressed-KV dequant-gather (#63018)

nbommi19 committed 5 days ago

Verified 6abdff59

[XPU] bump xpu-kernels to v0.1.16.2 (#61956)

yma11 committed 5 days ago

Verified fd3469f4

[XPU][CI] Fix test_tp1_fp8_fusions rms_quant_fusion on XPU (#63975)

Yejing-Lai committed 5 days ago

Verified 41dd82e6

[CI][XPU] Deselect Mamba deep-sleep case in Intel CI (#64057)

chaojun-zhang committed 5 days ago

Verified 08a236fa

[Bugfix][Pooling] Reject prompt-wide truncation for Jina ranking (#55200)

bvolpato committed 5 days ago

Verified e82e6d42

[Security] Bound FunAudioChat runtime audio to the profiled duration (#57411)

jperezdealgaba committed 5 days ago

Verified 26302eb9

[Security] Fix ReDoS in validation error path sanitizer (#52120)

jperezdealgaba committed 5 days ago

Verified 665bfc8a

[Bugfix] Omit null `annotations` and `usage.prompt_tokens_details` in chat completions (#64177)

vadiklyutiy committed 5 days ago

Verified f60edccd

[Bugfix] Merge supported KV cache layouts across pipeline stages instead of asserting they match (#61294)

vadiklyutiy committed 5 days ago

Verified edad8b03

[Perf][SM120][MLA] Use compact FP8 KV rows for NoPE models (#58305)

lucifer1004 committed 5 days ago

Verified dbcf5688

[Feature][Core] Add opt-in multi-worker CUDA checkpoint sleep (#64186)

AlanFokCo committed 5 days ago

Verified 88d3995e

[CI][Bugfix] Pin the ColUltraVBERT test checkpoint to the revision vLLM supports (#64189)

AndreasKaratzas committed 5 days ago

Verified 7e3e3f6a

[Doc] Add autotrust/JEV-27B-VL multimodal recipe (vLLM >= 0.32.0) (#64180)

yiminghub2024 committed 5 days ago

Verified 9bc1c8b4

[Bugfix][Frontend] fix Anthropic inline-system probe crashing startup (#64020)

Hritik003 committed 5 days ago

Verified b766f159

[Rust Frontend] Export vllm:num_requests_kv_fetch_by_stage like Python (#63962)

BugenZhao committed 5 days ago

Verified 68eca9af

[CI][XPU] Give test_mrv2_weight_offloading more GPU memory headroom (#63895)

chaojun-zhang committed 5 days ago

Verified 84e45536

[ROCm][Bugfix][Kimi-K3] Build the folded q_b_proj / W_UK weight on the root model (#64158)

seungrokj committed 5 days ago

Verified b71c1d5f

[RL] Support per-update source ranks for NCCL weight transfer (#62557)

giannis2two committed 5 days ago

Verified 2b678ac7