All workflows

Workflow runs

• Fix premature rounding in addmm

  requires action with the application:
  Build and Test#5106:Pull request#4769openedbytetsuo
  tetsuo:fix-metal-addmm-precision

  Oct 25, 2026Action required
  More actions for workflow run 5106

• Update bypass list

  completed successfully:
  Update bypass list#424:Scheduled
  main

  Oct 25, 202618s
  More actions for workflow run 424

• Release build

  failed:
  Release build#322:Scheduled
  main

  Oct 25, 20269m 37s
  More actions for workflow run 322

• Fix fast.cross_entropy VJP on non-contiguous Metal inputs

  requires action with the application:
  Build and Test#5105:Pull request#4702synchronizebyStockerMC
  StockerMC:fix-cross-entropy-vjp-strides

  Oct 25, 2026Action required
  More actions for workflow run 5105

• [jaccl] Detect dead RDMA links instead of crashing

  requires action with the application:
  Build and Test#5104:Pull request#4767synchronizebyipazarbasi
  ipazarbasi:fix-4192

  Oct 25, 2026-1s
  More actions for workflow run 5104

• Check scatter and gather indices are in bounds

  requires action with the application:
  Build and Test#5103:Pull request#4768openedbyturingH
  turingH:fix-scatter-index-oob

  Oct 25, 2026Action required
  More actions for workflow run 5103

• [jaccl] Detect dead RDMA links instead of crashing

  requires action with the application:
  Build and Test#5102:Pull request#4767openedbyipazarbasi
  ipazarbasi:fix-4192

  Oct 25, 2026Action required
  More actions for workflow run 5102

• Route of per-channel gate in gated_delta_update fixed

  requires action with the application:
  Build and Test#5101:Pull request#4766openedbyExorust
  Exorust:fix/gated-delta-vector-gate-routing

  Oct 25, 2026Action required
  More actions for workflow run 5101

• Fix gguf tensor to f16 size overflow

  requires action with the application:
  Build and Test#5100:Pull request#4764openedbyturingH
  turingH:fix-gguf-tensor-to-f16-overflow

  Oct 25, 2026Action required
  More actions for workflow run 5100

• Fix GPU concatenate for outputs larger than 2^31 elements

  requires action with the application:
  Build and Test#5099:Pull request#4762openedbycitruslab
  Sogni-AI:fix-concatenate-large-offsets

  Oct 24, 2026-1s
  More actions for workflow run 5099

• [CUDA] Expose compilation targets via device_info

  requires action with the application:
  Build and Test#5098:Pull request#4761openedbydhiltgen
  dhiltgen:codex/cuda-architecture-info

  Oct 24, 2026Action required
  More actions for workflow run 5098

• Fixes MLX/ErrorHandler.swift:345: Fatal error due to missing stream argument

  requires action with the application:
  Build and Test#5097:Pull request#4746synchronizebySatchitananda
  Satchitananda:fix-missing-streams

  Oct 24, 2026Action required
  More actions for workflow run 5097

• Fixes MLX/ErrorHandler.swift:345: Fatal error due to missing stream argument

  requires action with the application:
  Build and Test#5096:Pull request#4746synchronizebySatchitananda
  Satchitananda:fix-missing-streams

  Oct 24, 2026Action required
  More actions for workflow run 5096

• Fixes MLX/ErrorHandler.swift:345: Fatal error due to missing stream argument

  requires action with the application:
  Build and Test#5095:Pull request#4746synchronizebySatchitananda
  Satchitananda:fix-missing-streams

  Oct 24, 2026Action required
  More actions for workflow run 5095

• Keep g and beta in float32 in gated_delta_update

  requires action with the application:
  Build and Test#5094:Pull request#4760openedbyjvmenen
  jvmenen:fix/gdn-float32-gates

  Oct 24, 2026Action required
  More actions for workflow run 5094

• Update bypass list

  completed successfully:
  Update bypass list#423:Scheduled
  main

  Oct 24, 202616s
  More actions for workflow run 423

• Release build

  completed successfully:
  Release build#321:Scheduled
  main

  Oct 24, 202618m 45s
  More actions for workflow run 321

• Fix silent Metal grid truncation in softmax, logsumexp, rms_norm and layer_norm past 2^27 rows

  requires action with the application:
  Build and Test#5093:Pull request#4717synchronizebyArupKantiDas
  ArupKantiDas:fix-row-kernel-grid-overflow

  Oct 24, 2026-1s
  More actions for workflow run 5093

• Fix GPU concatenate for outputs larger than 2^31 elements

  requires action with the application:
  Build and Test#5092:Pull request#4758openedbytivris
  tivris:fix-concatenate-large-offsets

  Oct 24, 2026-1s
  More actions for workflow run 5092

• Add optional source IP binding for ring connections

  requires action with the application:
  Build and Test#5091:Pull request#4757openedbyronaldmannak
  PicoMLX:ip-binding

  Oct 24, 2026Action required
  More actions for workflow run 5091

• Fix FP8 NaN preservation in to_fp8 and E8M0 scale dequantization

  requires action with the application:
  Build and Test#5090:Pull request#4756openedbyFaizPalwala
  FaizPalwala:fix-fp8-nan-handling

  Oct 23, 2026Action required
  More actions for workflow run 5090

• Fix FFT read/write past buffer end in partial threadgroups (#4740)

  completed successfully:
  Update bypass list#422:Commit56c2bf9pushedbyzcbenz
  main

  Oct 23, 202618s
  More actions for workflow run 422

• Fix FFT read/write past buffer end in partial threadgroups (#4740)

  failed:
  Build and Test#5089:Commit56c2bf9pushedbyzcbenz
  main

  Oct 23, 202642m 10s
  More actions for workflow run 5089

• [Metal] Add SDPA fast paths for speculative decoding

  requires action with the application:
  Build and Test#5088:Pull request#4723synchronizebydhiltgen
  dhiltgen:sdpa-nax-split

  Oct 23, 2026Action required
  More actions for workflow run 5088

• Adding sliding window attention (#4734)

  completed successfully:
  Update bypass list#421:Commit360078apushedbyjagrit06
  main

  Oct 23, 202616s
  More actions for workflow run 421