MLX

Workflow runs for MLX

• Fix mixed-dtype cat and stack in to_edge and the edge check (#24450)

  currently running:
  MLX#5481:Commit3668b0cpushedbyJacobSzwejbka
  release/1.6

  Nov 2, 2026In progress
  More actions for workflow run 5481

• Keep Half and BFloat16 sums in float in portable matmul, var, cumsum …

  completed successfully:
  MLX#5480:Commitb4d713bpushedbyJacobSzwejbka
  release/1.6

  Nov 2, 20261m 5s
  More actions for workflow run 5480

• NXP backend: Adjust getitem conversion when it produces a partition…

  completed successfully:
  MLX#5479:Commit4058814pushedbyJacobSzwejbka
  release/1.6

  Nov 2, 202649s
  More actions for workflow run 5479

• Python runtime: enforce CUDA stream selection rules on the stream its…

  completed successfully:
  MLX#5478:Commitb3a7307pushedbyJacobSzwejbka
  release/1.6

  Nov 2, 202646s
  More actions for workflow run 5478

• Limit the default thread pool to the allowed CPUs on Linux (not Andro…

  queued:
  MLX#5477:Commit10514b8pushedbyJacobSzwejbka
  release/1.6

  Nov 2, 2026Queued
  More actions for workflow run 5477

• Move MLX, Build Presets and macOS wheel CI to macOS 15 runners (#24441)

  completed successfully:
  MLX#5476:Commitb9089ffpushedbyJacobSzwejbka
  release/1.6

  Nov 2, 202634s
  More actions for workflow run 5476

• CUDA backend: lower fused attention variants (#24426)

  queued:
  MLX#5475:Commite40e64cpushedbyshoumikhin
  main

  Nov 2, 2026Queued
  More actions for workflow run 5475

• Precompute quantized constant ops after Q/DQ folding (#24258)

  completed successfully:
  MLX#5474:Commitb0d1268pushedbymeta-codesync[bot]Bot
  main

  Nov 2, 202642s
  More actions for workflow run 5474

• MLX

  queued:
  MLX#5473:Manually runbyJacobSzwejbka
  release/1.6

  Nov 2, 2026Queued
  More actions for workflow run 5473

• Fix crashes, wrong outputs and lowering failures in several backends and in exir

  currently running:
  MLX#5472:Pull request#24298reopenedbyJacobSzwejbka
  cherry-pick-24222-by-pytorch_bot_bot_

  Nov 2, 2026In progress
  More actions for workflow run 5472

• Fix CUDA mask indexing, CUDA graph outputs and the runtime output-buffer check

  currently running:
  MLX#5471:Pull request#24320reopenedbyJacobSzwejbka
  cherry-pick-24299-by-pytorch_bot_bot_

  Nov 2, 2026In progress
  More actions for workflow run 5471

• CUDA backend: per-program weights file names and antialiased resize

  currently running:
  MLX#5470:Pull request#24324reopenedbyJacobSzwejbka
  cherry-pick-24314-by-pytorch_bot_bot_

  Nov 2, 2026In progress
  More actions for workflow run 5470

• Fix 0-d tensors, dynamic output sizes and integer reductions, and default to float32 on CPU, in the OpenVINO backend

  currently running:
  MLX#5469:Pull request#24377reopenedbyJacobSzwejbka
  cherry-pick-24294-by-pytorch_bot_bot_

  Nov 2, 2026In progress
  More actions for workflow run 5469

• Move MLX, Build Presets and macOS wheel CI to macOS 15 runners

  currently running:
  MLX#5468:Pull request#24441reopenedbyJacobSzwejbka
  cherry-pick-24402-by-pytorch_bot_bot_

  Nov 2, 2026In progress
  More actions for workflow run 5468

• OpenVINO: fix bool min/max and masked read lowering, and follow-ups to #24294

  currently running:
  MLX#5467:Pull request#24446reopenedbyJacobSzwejbka
  cherry-pick-24396-by-pytorch_bot_bot_

  Nov 2, 2026In progress
  More actions for workflow run 5467

• Add a portable kernel for antialiased bicubic resize

  currently running:
  MLX#5466:Pull request#24447reopenedbyJacobSzwejbka
  cherry-pick-24400-by-pytorch_bot_bot_

  Nov 2, 2026In progress
  More actions for workflow run 5466

• CUDA backend: delete each output handle once when a program returns one tensor twice

  completed successfully:
  MLX#5465:Pull request#24448reopenedbyJacobSzwejbka
  cherry-pick-24420-by-pytorch_bot_bot_

  Nov 2, 202657m 4s
  More actions for workflow run 5465

• Extend vision exporter to support all models (#24351)

  completed successfully:
  MLX#5464:Commit822cb60pushedbymeta-codesync[bot]Bot
  main

  Nov 2, 20261m 17s
  More actions for workflow run 5464

• Fix crashes, wrong outputs and lowering failures in several backends and in exir

  completed successfully:
  MLX#5463:Pull request#24298synchronizebyshoumikhin
  cherry-pick-24222-by-pytorch_bot_bot_

  Nov 2, 202659m 45s
  More actions for workflow run 5463

• Limit the default thread pool to the allowed CPUs on Linux (not Andro…

  completed successfully:
  MLX#5462:Commite68a824pushedbyshoumikhin
  main

  Nov 2, 202652s
  More actions for workflow run 5462

• Fix mixed-dtype cat and stack in to_edge and the edge check (#24381)

  completed successfully:
  MLX#5461:Commitecc4197pushedbyshoumikhin
  main

  Nov 2, 202626m 3s
  More actions for workflow run 5461

• OpenVINO: fix bool min/max and masked read lowering, and follow-ups t…

  completed successfully:
  MLX#5460:Commitefbd460pushedbyshoumikhin
  main

  Nov 2, 202655s
  More actions for workflow run 5460

• Keep ops above 6 dims out of XNNPACK and fix quantized ops it leaves …

  completed successfully:
  MLX#5459:Commit1549787pushedbyshoumikhin
  main

  Nov 2, 202644s
  More actions for workflow run 5459

• Move MLX, Build Presets and macOS wheel CI to macOS 15 runners

  completed successfully:
  MLX#5458:Pull request#24441openedbypytorchbot
  cherry-pick-24402-by-pytorch_bot_bot_

  Nov 2, 202628m 27s
  More actions for workflow run 5458

• Add a portable kernel for antialiased bicubic resize (#24400)

  completed successfully:
  MLX#5457:Commita5476a5pushedbyshoumikhin
  main

  Nov 2, 202657s
  More actions for workflow run 5457