All pull requests

Pull requests

• perf(cohere2): fuse the parallel-block residual adds with the next norm

  area:modelsplatform:macospriority:mediumstatus:reviewtype:performance
  #1948·inureyes opened Sep 21, 2026·

  ·

• perf(metal): raise the command-buffer input budget for decode only

  area:coreplatform:macospriority:mediumstatus:reviewtype:performance
  #1947·inureyes opened Sep 21, 2026·

  ·
  2

• fix : handle missing sequence entries

  #1936·saitejagirada opened Sep 20, 2026·

  ·

• feat(webui): redesign the model library as a dense list with row actions

  area:architecturepriority:highstatus:donetype:enhancement
  #1930·inureyes opened Sep 19, 2026·

  ·
  1

  2

• feat(webui): make Activity read as a status page

  area:architecturepriority:mediumstatus:donetype:enhancement
  #1929·inureyes opened Sep 19, 2026·

  ·
  1

  2

• fix(bench): stop hardcoding the run date and replace GNU-only grep -oP in run_bench.sh

  #1899·yujung7768903 opened Sep 15, 2026·

  ·
  1

• ci(deps): bump actions/setup-node from 5 to 7

  type:choretype:dependency
  #1879·dependabot[bot] opened Sep 14, 2026·

  ·

• test(rocm): promote the FFT probe to a reusable example

  area:coreplatform:linuxpriority:mediumstatus:reviewtype:test
  #1878·inureyes opened Sep 13, 2026·

  ·

• test: add reason strings to bare #[ignore] attributes

  #1855·saitejagirada opened Sep 12, 2026·

  ·
  1

• feat(vlm): add vision feature caching for LFM2-VL

  area:modelsmodeltype:vlmpriority:mediumstatus:reviewtype:enhancement
  #1135·MrFelt0385 opened Aug 13, 2026·

  ·
  2

• feat(xla): add Gemma3n audio execution

  area:inferencearea:modelspriority:lowstatus:blockedtype:enhancement
  #922·inureyes opened Jul 24, 2026·

  ·
  1

  4

• feat(xla): add Qwen3-VL DeepStack vision path

  area:inferencearea:modelspriority:lowstatus:reviewtype:enhancement
  #921·inureyes opened Jul 24, 2026·

  ·
  1

  4

• feat: add Qwen2.5-VL OpenXLA vision path

  area:inferencearea:modelspriority:lowstatus:reviewtype:enhancement
  #920·inureyes opened Jul 24, 2026·

  ·
  1

  4

• feat: add Youtu-VL XLA vision and MLA inference (#872)

  area:inferencearea:modelspriority:lowstatus:reviewtype:enhancement
  #919·inureyes opened Jul 24, 2026·

  ·
  1

  2

• feat(xla): support Molmo sparse additive visual embeddings

  area:inferencearea:modelspriority:lowstatus:reviewtype:enhancement
  #918·inureyes opened Jul 24, 2026·

  ·
  1

  8

• feat: add Gemma3 VLM OpenXLA path (#869)

  area:inferencearea:modelspriority:mediumstatus:reviewtype:enhancement
  #917·inureyes opened Jul 24, 2026·

  ·
  1

  3