All pull requests

Pull requests

• Accept collective config in XCCL c10d op kernels

  disable_e2edisable_win
  #5587·d4l3k opened Oct 1, 2026Contributor·

  ·

• Validate grad_output in grid_sampler_{2d,3d}_backward

  disable_distributeddisable_e2e
  #5584·pbielak opened Oct 1, 2026Contributor·

  ·
  2

• [4/N] Synchronize test files with upstream generalization changes- #5423

  disable_autodisable_distributeddisable_e2e
  #5582·BBBela opened Oct 1, 2026Contributor·

  ·
  

• Align upstream test class TestWrapperSubclassAliasing rename change with torch-xpu-ops codebase.

  disable_autodisable_distributeddisable_e2e
  #5581·BBBela opened Oct 1, 2026Contributor·

  ·
  

• Fix type narrowing in BatchNorms Float2 ctor

  disable_distributeddisable_e2e
  #5580·dominikjastrzebskix opened Oct 1, 2026Contributor·

  ·

• [UT] Align test_output_aliases_input_multi_output_view with upstream

  disable_builddisable_distributeddisable_e2edisable_win
  #5579·jkosnox opened Oct 1, 2026Contributor·

  ·
  1

  1

• Add Float8_e5m2 to abs_xpu

  disable_distributeddisable_e2e
  #5572·tszulist-hbn opened Sep 30, 2026Contributor·

  ·
  1

• Apply sycl::reduce_over_group for RMSNorm

  disable_distributeddisable_e2e
  #5561·CuiYifeng opened Sep 29, 2026Contributor·

  ·
  6

• Fixed #5440

  disable_builddisable_distributeddisable_e2edisable_win
  #5560·mwiktor-intel opened Sep 29, 2026Contributor·

  ·
  1

• Fix group norm forward on unaligned inputs

  disable_distributeddisable_e2e
  #5558·pbielak opened Sep 29, 2026Contributor·

  ·
  1

• Inprove RNG performance and probabilistic guarantees

  #5557·frost-intel opened Sep 29, 2026Contributor·

  ·

• Enable XPU dummy NestedTensor MHA tests

  disable_acceleratedisable_autodisable_e2edisable_windisable_windows_ut
  #5556·tszulist-hbn opened Sep 29, 2026Contributor·

  ·
  1

  4

• Add proper dtype checks for histc_out.

  disable_distributeddisable_e2e
  #5554·AKloniecki opened Sep 29, 2026Contributor·

  ·
  2

  5

• Support channels-last inputs in group norm kernels

  disable_autodisable_distributeddisable_e2e
  #5551·pbielak opened Sep 28, 2026Contributor·

  ·
  1

  1

• CI: skip known P0 UT failures

  #5550·libohao1201 opened Sep 28, 2026Contributor·

  ·
  1

• Enable int64_t for upsample bilinear 2d channels last

  disable_autodisable_distributeddisable_e2e
  #5549·pbielak opened Sep 28, 2026Contributor·

  ·
  1

  4

• [bot] Gate fix on a named list, and let a comment amend it

  #5548·ZhaoqiongZ opened Sep 28, 2026Contributor·

  ·

• LayerNorm Tree-Reduce Optimization

  disable_distributeddisable_e2e
  #5546·CuiYifeng opened Sep 28, 2026Contributor·

  ·
  16

• batch_norm: SLM param cache for channels-last fwd 1D kernel, fix small-C perf

  disable_distributeddisable_e2e
  #5545·weishi-deng opened Sep 28, 2026Contributor·

  ·
  7

• Update to adapt to compiler 26.2

  disable_distributeddisable_e2e
  #5537·NeoZhangJianyu opened Sep 28, 2026Contributor·

  ·
  8

• Migrate Upstream PR https://github.com/intel/torch-xpu-ops/issues/5205 done UT into XPU 8 files

  disable_builddisable_distributeddisable_e2edisable_win
  #5534·shangerxin opened Sep 26, 2026Contributor·

  ·
  8

• Migrate Upstream PR https://github.com/intel/torch-xpu-ops/issues/5205 done UT into XPU 11 files, test/autograd/test_*.py

  disable_builddisable_distributeddisable_e2edisable_win
  #5533·shangerxin opened Sep 25, 2026Contributor·

  ·
  5

• Migrate Upstream PR https://github.com/intel/torch-xpu-ops/issues/5205 done UT into XPU 11 files, ao/sparsity/test_*.py

  disable_builddisable_distributeddisable_e2edisable_win
  #5532·shangerxin opened Sep 25, 2026Contributor·

  ·
  7

• Remove triangular_solve.X fallback as no longer necessary after proper registration upstream.

  disable_distributeddisable_e2e
  #5530·AKloniecki opened Sep 25, 2026Contributor·

  ·
  1

• Fix CompilerBisector test_pre_grad and test_joint_graph by using CustomGraphPass

  disable_builddisable_distributeddisable_e2edisable_win
  #5521·reginaldalfret opened Sep 24, 2026Contributor·

  ·
  1