All pull requests

Pull requests

• [Frontend][Metrics] Add vllm:request_reasoning_effort counter

  documentationfrontendrust
  #62315·robertgshaw2-redhat opened Oct 18, 2026Collaborator·6 tasks done·

  ·
  1

• [Performance] Memoize waiting prefix cache hits across retries

  kv-cache-manager
  #62314·RichterWang opened Oct 18, 2026·

  ·
  1

• [Bugfix][KV Connector][ROCm] MoRIIO READ: skip zeroing READ destinations for mixed-precision caches

  bugkv-connectorreadyrocm
  #62313·benenzhu opened Oct 18, 2026Contributor·5 of 6 tasks·

  ·

• [Rust Frontend] Add --reasoning-effort-map to rewrite request reasoning efforts

  readyrust
  #62312·robertgshaw2-redhat opened Oct 18, 2026Collaborator·6 tasks done·

  ·
  2

• [Bugfix][KV Connector][ROCm] MoRIIO: get the attention backend from the model layers

  bugkv-connectorreadyrocm
  #62311·benenzhu opened Oct 18, 2026Contributor·6 tasks done·

  ·
  1

• [XPU] support adaptive verification for deepseek-v4

  deepseekdflashDSv4intel-gpumrv2
  #62309·yma11 opened Oct 18, 2026Contributor·6 tasks·

  ·

• [Bugfix][Tool Parser] Fix invalid streamed JSON for MiniCPM5 string args with leading or trailing spaces

  bugtool-calling
  #62306·Pranavv-dev opened Oct 18, 2026·6 tasks done·

  ·
  1

• [Doc] Document /inference/v1/generate request/response fields and DP rank header

  documentationready
  #62305·aoshen02 opened Oct 18, 2026Contributor·

  ·
  4

• [Frontend] Serve Gemma 4 on /v1/systemone and /v1/decisions

  documentationfrontendreadyverified
  #62303·cipheraxat opened Oct 18, 2026Contributor·6 tasks done·

  ·
  7

• [Bugfix][Spec Decode] Preserve DP replica index for standalone draft models

  bugmrv2speculative-decoding
  #62301·MuonChaser opened Oct 18, 2026·6 tasks done·

  ·
  3

• [CI] Avoid nested CUPTI control in bench forward profiles

  performance
  #62300·khluu opened Oct 18, 2026Member·

  ·

• [Bugfix][Parser] Honor a prompt-closed reasoning block for PLaMo3

  bugtool-calling
  #62299·kushaldabbe opened Oct 18, 2026Contributor·6 tasks done·

  ·

• [CI][ROCm] Run ROCm-capable tests that are currently skipped on AMD

  ci/buildcpudeepseekdflashinklingkv-connectorminimaxmistralmulti-modalitynvidiaqwenrocmspeculative-decodingtorch.compile
  #62298·AndreasKaratzas opened Oct 18, 2026Member·2 of 6 tasks·

  ·

• [Backport][0.32] Preserve Clef Flash context and cold-cache behavior

  documentationfrontendmulti-modalitynew-modelpooling
  #62297·webcodes-cz opened Oct 18, 2026·6 tasks done·

  ·
  2

• [Bugfix][Multimodal] Normalize native-rate PyAV audio

  bugmulti-modality
  #62296·LingYi-Liang opened Oct 18, 2026·6 tasks done·

  ·
  1

• [Bugfix][Structured Output][v0.32] Honor resolved backends and mask widths

  bugstructured-output
  #62295·webcodes-cz opened Oct 18, 2026·5 of 6 tasks·

  ·
  1

• [Bugfix][Anthropic] Map vLLM error types to Anthropic error.type vocabulary

  bugfrontend
  #62294·zyulai opened Oct 18, 2026·6 tasks done·

  ·
  1

• [Bugfix][Frontend] Render disabled-thinking Messages requests with the server's reasoning effort

  bugfrontendready
  #62293·albertoperdomo2 opened Oct 18, 2026Contributor·6 tasks done·

  ·
  3

• [Bugfix][Benchmark] Fix standalone dataset statistics plotting

  bugperformance
  #62292·LingYi-Liang opened Oct 18, 2026·6 tasks done·

  ·
  1

• [Bugfix][Multimodal] Treat a torchcodec that cannot load its native library as unavailable

  bugmulti-modality
  #62291·kimseunghyun-kr opened Oct 18, 2026Contributor·6 tasks done·

  ·

• [Bugfix][Rust Frontend] Drop an incomplete DeepSeek DSML tool call at max_tokens instead of leaking it into content

  bugdeepseekDSv4.1rust
  #62290·JohnZhao01 opened Oct 18, 2026·6 tasks done·

  ·
  1

• [Frontend][Tool] Add nemotron_v3_compact parser that keeps edge newlines in tool arguments

  documentationtool-calling
  #62289·Mramaran opened Oct 18, 2026·5 of 6 tasks·

  ·
  1

  2

• [Perf][Attention] Fuse DCP correction output layout for reduce-scatter

  #62288·wjinxu opened Oct 18, 2026Contributor·6 tasks done·

  ·
  1

• [Bugfix][KV Connector] Handle load failures after preemption

  bugkv-connectorscheduler
  #62287·Lewis-panda opened Oct 18, 2026·6 tasks done·

  ·
  1

  3

• [ROCm][CI] Contain the MI355 DP + EP exit-time abort

  ci/buildrocm
  #62286·AndreasKaratzas opened Oct 18, 2026Member·

  ·
  2