All pull requests

Pull requests

• [2026.5] Patch Qwen3.5 chat templates for minja

  category: GGUFcategory: tokenizerscategory: visual languageno-match-files
  #4883·mvafin opened Nov 10, 2026Contributor·4 tasks done·

  ·
  3

• [Not Ready for Review][NPU][ASR] Support SenseVoiceSmall on NPU with static shapes

  category: automatic speech recognitioncategory: GGUF
  #4882·GuoliangShiIntel opened Nov 10, 2026Contributor·4 tasks·

  ·
  25

  

• [2026.5] Make sampled speculative decoding exact for FastDraft, EAGLE3 and MTP

  category: continuous batchingcategory: GGUFcategory: GH Pages Docscategory: samplingcategory: speculative decodingno-match-files
  #4881·imargulis opened Nov 10, 2026·4 tasks done·

  ·
  10

• Streaming from language model to speech model in Omni- #4333

  category: continuous batchingcategory: CPP APIcategory: GGUFcategory: GH Pages Docscategory: JS APIcategory: omnicategory: Python APIcategory: samplingcategory: text streamercategory: visual languagecategory: VLM samples
  #4880·Wovchena opened Nov 10, 2026Collaborator·4 tasks done·

  ·
  21

  

• Accept NumPy scalars in GenerationConfig and pipeline kwargs

  category: GGUFcategory: Python APIcategory: sampling
  #4879·sohailshk opened Nov 10, 2026·1 of 4 tasks·

  ·
  

• Accept list and multi-word string stop_strings and stop_token_ids kwargs

  category: GGUFcategory: sampling
  #4878·sohailshk opened Nov 10, 2026·1 of 4 tasks·

  ·
  1

  

• Robotics integ

  category: automatic speech recognition samplescategory: automatic speech recognitioncategory: GGUFcategory: llm_benchcategory: visual languageno-match-files
  #4877·cwzrad opened Nov 10, 2026·4 tasks·

  ·
  5

• Introduce of Memory Sampler for MacOS (physical footprint)

  category: llm_bench
  #4873·krzyczar opened Nov 9, 2026Contributor·4 tasks·

  ·
  3

• [genai-cli] Introduce genai-cli export tool and Molmo2 support

  category: buildcategory: GGUFcategory: GH Pages Docscategory: GHAcategory: Python APIcategory: samples dependenciescategory: speech generationcategory: tests dependenciescategory: visual languagecategory: VLM samplescategory: WWBno-match-files
  #4870·rkazants opened Nov 9, 2026Collaborator·4 tasks done·

  ·
  10

• Support DFlash and DSpark drafts through continuous batching on PagedAttention (#4790)

  category: continuous batchingcategory: CPP APIcategory: GGUFcategory: GH Pages Docscategory: GHAcategory: LLMcategory: samplingcategory: speculative decodingcategory: visual languageno-match-files
  #4869·ofirzaf opened Nov 9, 2026Contributor·4 tasks done·

  ·
  6

• [WIP] Add Continuous Batching support for Qwen3 TTS

  category: continuous batchingcategory: CPP APIcategory: Python APIcategory: samplingcategory: speech generationno-match-files
  #4867·RyanMetcalfeInt8 opened Nov 8, 2026Contributor·1 of 4 tasks·

  ·
  27

• [C API] Fix eos and stop token id setters in generation config

  category: buildcategory: C APIno-match-files
  #4866·yasharth-0910 opened Nov 7, 2026·4 tasks done·

  ·
  1

  8

  

• Validate num_inference_steps and num_images_per_prompt are greater than 0 in ImageGenerationConfig

  category: GGUFcategory: image generation
  #4864·Aditya365x opened Nov 7, 2026Contributor·4 tasks done·

  ·
  1

  1

  

• [ASR][Whisper] Fix performance metrics to include all generated tokens

  category: automatic speech recognitioncategory: GGUF
  #4863·byekrang opened Nov 7, 2026Contributor·3 of 4 tasks·

  ·
  2

• Add MTP support for Qwen3.8-Flash-Next (aka qwen4_exp) in PA/CB

  category: continuous batchingcategory: GGUFcategory: LLMcategory: speculative decodingcategory: visual languageno-match-files
  #4862·sbalandi opened Nov 6, 2026Contributor·4 tasks·

  ·
  11

• [wwb] Fix divergency crash on single-token answers

  category: WWB
  #4860·soujasK opened Nov 6, 2026·3 of 4 tasks·

  ·
  1

• [llm_bench] Add possibility to use dataset for text/visual gen

  category: llm_benchcategory: tools_dependencies
  #4859·sbalandi opened Nov 6, 2026Contributor·4 tasks·

  ·
  5

• [TextRerankPipeline] Slice last token and fold 1-row (yes - no) score head before Qwen3 lm_head

  category: RAG
  #4857·wilson-seok opened Nov 6, 2026·4 tasks done·

  ·
  16

  

• [VLM] Support Qwen3.8-Flash-Next for PA / CB

  category: continuous batchingcategory: GGUFcategory: LLMcategory: visual languageno-match-files
  #4853·yatarkan opened Nov 5, 2026Contributor·2 of 4 tasks·

  ·
  8

• [ASR] Rename Whisper pipeline tests to ASR pipeline tests

  category: automatic speech recognitioncategory: GGUFcategory: GHA
  #4852·byekrang opened Nov 5, 2026Contributor·4 tasks done·

  ·
  3

• Support DFlash-style drafts with a pruned draft vocabulary

  category: continuous batchingcategory: CPP APIcategory: GGUFcategory: GH Pages Docscategory: GHAcategory: LLMcategory: Python APIcategory: samplingcategory: speculative decodingcategory: visual languageno-match-files
  #4848·imargulis opened Nov 5, 2026·3 of 4 tasks·

  ·

  · 2026.5
  20

  

• Add DSpark and DFlash v2 drafts and sampled drafting to DFlash speculative decoding

  category: continuous batchingcategory: CPP APIcategory: GGUFcategory: GH Pages Docscategory: GHAcategory: LLMcategory: Python APIcategory: samplingcategory: speculative decodingcategory: visual languageno-match-files
  #4847·imargulis opened Nov 5, 2026·3 of 4 tasks·

  ·

  · 2026.5
  33

  

• Add ClassificationPipeline

  category: buildcategory: CPP APIcategory: GH Pages Docscategory: GHAcategory: Python APIcategory: RAGno-match-files
  #4846·as-suvorov opened Nov 5, 2026Collaborator·4 tasks·

  ·
  6

• [GGUF] Support embedded MTP speculative decoding

  category: continuous batchingcategory: CPP APIcategory: GGUFcategory: GH Pages Docscategory: LLMcategory: speculative decodingcategory: visual languageno-match-files
  #4842·mvafin opened Nov 5, 2026Contributor·4 tasks done·

  ·
  9

• Bump pytest in /tests/python_tests

  category: GGUFcategory: tests dependenciesdependenciespython
  #4835·dependabot[bot] opened Nov 4, 2026Bot·

  ·
  4