All pull requests

Pull requests

• [Converter:Bugfix] Report external weight write failures

  #4943·hulkbig opened Oct 8, 2026Contributor·6 of 12 tasks·

  ·

• [LLM:Bugfix] Honor Eagle draft RoPE configuration

  #4942·ValiantDiligent opened Oct 8, 2026Collaborator·12 tasks·

  ·
  1

• [Doc:Bugfix] Fix stale fork wording and branch targets in CONTRIBUTING

  #4941·ZhuoenX opened Oct 8, 2026·5 of 12 tasks·

  ·
  1

• [Vulkan:Bugfix] Use buffer path for multi-image softmax

  #4940·Qxinyu opened Oct 8, 2026Collaborator·6 tasks done·

  ·
  1

  1

• [Converter:Bugfix] Honor legacy ONNX Softmax defaults and inherited opsets

  awaiting contributormodule:convertertype:bug
  #4936·hulkbig opened Oct 5, 2026Contributor·

  ·
  2

  

• [Converter:Bugfix] Honor ONNX OneHot default axis

  awaiting contributormodule:convertertype:bug
  #4934·hulkbig opened Oct 4, 2026Contributor·

  ·
  1

  2

  

• [Converter:Bugfix] Preserve legacy ONNX Pad fill values

  module:convertertype:bug
  #4932·hulkbig opened Oct 3, 2026Contributor·5 of 12 tasks·

  ·
  

• [Converter:Bugfix] Saturate ONNX Shape int64 bounds before narrowing

  module:convertertype:bug
  #4930·hulkbig opened Oct 3, 2026Contributor·5 of 12 tasks·

  ·
  

• [Converter:Bugfix] Preserve ONNX Slice steps when axes is omitted

  module:convertertype:bug
  #4929·hulkbig opened Oct 3, 2026Contributor·5 of 12 tasks·

  ·
  

• [Converter:Bugfix] Honor constant axes in ONNX composite reductions

  module:convertertype:bug
  #4927·hulkbig opened Oct 2, 2026Contributor·6 of 12 tasks·

  ·
  

• [CPU:Bugfix] Validate complete mmap weight caches before reuse

  type:bug
  #4926·kerry0520 opened Oct 2, 2026·5 of 12 tasks·

  ·
  2

  

• [Converter:Bugfix] Parse all ONNX ScatterElements attributes

  module:convertertype:bug
  #4925·hulkbig opened Oct 2, 2026Contributor·6 of 12 tasks·

  ·
  

• [CPU:Perf] Optimize RVV memory access for packing and depthwise convolution

  awaiting contributortype:performance
  #4913·typer-J opened Sep 24, 2026Contributor·

  ·
  1

  

• [CPU] Optimize large 4-byte raster bitcopy broadcasts with RVV

  awaiting contributortype:performance
  #4907·typer-J opened Sep 23, 2026Contributor·

  ·
  1

  

• [CPU:Bugfix] Fix DFlash cross-attention KV truncation

  type:bug
  #4898·yydhYYDH opened Sep 20, 2026Collaborator·5 of 12 tasks·

  ·
  

• [CPU:Perf] Vectorize the conv int8 bias-float init with a corrected reciprocal

  awaiting contributormodule:quantizationtype:performance
  #4893·typer-J opened Sep 19, 2026Contributor·

  ·
  1

  

• [CPU:Perf] Vectorize the conv int8 bias-float init with a reciprocal

  awaiting contributormodule:quantizationtype:performance
  #4892·typer-J opened Sep 19, 2026Contributor·

  ·
  1

  

• [CPU:Bugfix] Preserve RVV INT8 sum accumulator tails

  awaiting contributormodule:quantizationtype:bug
  #4869·luke-opus opened Sep 13, 2026Collaborator·5 of 12 tasks·

  ·
  2

  

• [CPU:Perf] Add an RVV kernel for conv int8 weight kernel sum

  awaiting contributormodule:quantizationtype:performance
  #4865·typer-J opened Sep 12, 2026Contributor·

  ·
  4

  

• [Pymnn:Test] Remove the dead cv.calcHist copy of test_histogram

  awaiting contributormodule:cvmodule:pymnn
  #4863·Anai-Guo opened Sep 11, 2026·

  ·
  2

  

• MNN:Speed: Add virtual ref opt

  #4794·jxt1234 opened Aug 21, 2026Collaborator·12 tasks·

  ·
  

• [QNN:Bugfix] Preserve calibrated MUL activation ranges for QNN export

  awaiting contributor
  #4765·yydhYYDH opened Aug 13, 2026Collaborator·5 of 12 tasks·

  ·
  9

  

• [QNN:Feature] Auto ADSP_LIBRARY_PATH setup and device creation fallback

  awaiting contributor
  #4584·lanling-47 opened Jun 30, 2026Contributor·

  ·
  1

  

• [QNN:Bugfix] Fix StridedSlice NC4HW4 layout and improve QNNPerf robustness

  awaiting contributor
  #4583·lanling-47 opened Jun 30, 2026Contributor·

  ·
  1

  

• [ARM:Bugfix] Enable MNNRankOneUpdate fallback on Apple Silicon

  awaiting contributor
  #4579·lanling-47 opened Jun 29, 2026Contributor·6 tasks done·

  ·
  1