All pull requests

Pull requests

• [do not review, for test now - dflash opt]move FC transform to main, avoiding excessive memory alloc/free

  category: speculative decoding
  #4616·songbell opened Oct 10, 2026Contributor·4 tasks·

  ·
  3

• Preserve KV cache parameter runtime info

  category: speculative decodingno-match-files
  #4615·mangguo321 opened Oct 10, 2026Contributor·4 tasks done·

  ·
  1

• Text2SpeechPipeline: Allow optional SpeechGenerationConfig to be passed into generate

  category: C APIcategory: CPP APIcategory: Python APIcategory: speech generation
  #4613·RyanMetcalfeInt8 opened Oct 9, 2026Contributor·2 of 4 tasks·

  ·
  11

• Bump the pip-dependencies group across 1 directory with 4 updates

  category: tools_dependenciescategory: WWBdependenciespython
  #4612·dependabot[bot] opened Oct 9, 2026Bot·

  ·

• [CB] Fix partial preemption for hybrid models with prefix caching

  category: continuous batchingcategory: GGUF
  #4611·mzegla opened Oct 8, 2026Collaborator·4 tasks done·

  ·
  9

• [genai-cli] Introduce genai-cli tool for HF models export

  category: cmake / buildcategory: GGUFcategory: GH Pages Docscategory: GHAcategory: Python APIcategory: visual languagecategory: WWB
  #4609·rkazants opened Oct 8, 2026Collaborator·4 tasks·

  ·
  13

• [llm_bench] Add configurable output image path to llm_bench

  category: GGUFcategory: llm_bench
  #4608·andrew-k-park opened Oct 8, 2026Contributor·3 tasks done·

  ·
  24

• Fix VLM MTP Reporting in LLM Bench

  category: llm_bench
  #4607·apaniukov opened Oct 7, 2026Contributor·4 tasks·

  ·
  4

• Bump the pip group across 2 directories with 2 updates

  category: GGUFcategory: tests dependenciescategory: tools_dependenciescategory: WWBdependenciespython
  #4601·dependabot[bot] opened Oct 7, 2026Bot·

  ·

• [VLM] Add Qwen3-Omni sample tests

  category: GGUFcategory: GHA
  #4599·sgonorov opened Oct 7, 2026Contributor·3 of 4 tasks·

  ·
  13

• Intel Instrumentation and Tracing Technology for GenAI

  category: automatic speech recognitioncategory: cmake / buildcategory: continuous batchingcategory: LLMcategory: tokenizerscategory: visual languageno-match-files
  #4597·freddychiu opened Oct 7, 2026·1 of 4 tasks·

  ·
  13

• [VLM][CB] Roll back pipeline state when generate() throws

  category: continuous batchingcategory: GGUFcategory: LLMcategory: visual language
  #4596·sgonorov opened Oct 6, 2026Contributor·4 tasks done·

  ·
  17

  

• [JS] Implement EmbeddingPipeline

  category: GH Pages Docscategory: GHAcategory: JS APIno-match-files
  #4595·Retribution98 opened Oct 6, 2026Contributor·4 tasks done·

  ·
  12

• Bump the npm-dependencies group across 1 directory with 2 updates

  dependenciesjavascriptno-match-files
  #4593·dependabot[bot] opened Oct 6, 2026Bot·

  ·
  1

  

• Support FunAudioLLM/SenseVoiceSmall model

  category: automatic speech recognitioncategory: cmake / buildcategory: CPP APIcategory: GGUFcategory: GH Pages Docscategory: GHAcategory: JS APIcategory: Python APIcategory: samplingno-match-files
  #4592·byekrang opened Oct 5, 2026Contributor·3 of 4 tasks·

  ·
  38

  

• Add SenseVoice support in WWB.

  category: WWB
  #4590·popovaan opened Oct 5, 2026Contributor·4 tasks·

  ·
  7

  

• Add minicpm v 4 5 video

  category: GGUFcategory: visual languagecategory: WWB
  #4589·pavel-esir opened Oct 5, 2026Contributor·3 of 4 tasks·

  ·
  25

  

• Add tool_call_benchmark to validate tool calling of exported models

  category: GHAno-match-files
  #4584·dhandhalyabhavik opened Oct 3, 2026·

  ·
  138

• Fix stateful (SDPA) batch KV-cache offsets after a request finishes early

  category: GGUFcategory: LLMcategory: visual languageno-match-files
  #4582·10vulture1005 opened Oct 2, 2026·4 tasks done·

  ·
  1

  

• Gemma4 MTP Continuous Batching

  category: cmake / buildcategory: continuous batchingcategory: CPP APIcategory: GGUFcategory: GH Pages Docscategory: GHAcategory: LLMcategory: prompt lookupcategory: Python APIcategory: samplingcategory: speculative decodingcategory: visual languageno-match-files
  #4579·apaniukov opened Oct 2, 2026Contributor·4 tasks·

  ·
  6

• [CB] Query KV cache capacity in tokens and reject requests exceeding it before prefill (#4545)

  category: continuous batchingcategory: CPP APIcategory: LLMcategory: Python APIcategory: visual languageno-match-files
  #4578·uvv-01 opened Oct 2, 2026·4 tasks done·

  ·
  1

  1

  

• [llm_bench][wwb] Remove tiny Qwen3-Omni xfails after checkpoint fix

  category: GGUFcategory: GHAcategory: llm_benchcategory: WWB
  #4575·sgonorov opened Oct 2, 2026Contributor·4 tasks done·

  ·
  17

  

• [GHA] Testing HuggingFace caching proxy

  category: GHAdo_not_mergedo_not_reviewdo-not-merge
  #4574·ababushk opened Oct 2, 2026Contributor·

  ·
  17

• [EXPERIMENT][WIP] Add k2-fsa/OmniVoice WWB speech-generation wrapper

  category: WWBdo-not-mergeexperimental
  #4568·mlukasze opened Sep 30, 2026Contributor·

  ·
  5

• [TP_GPU] Initial approach to interaction between GenAI and Tensor Parallel plugin

  category: continuous batchingdo-not-mergeno-match-files
  #4567·Lyamin-Roman opened Sep 30, 2026·

  ·
  5