Search results

Unnecessary FP64 cast for <code>padding_idx</code> in <code>EmbeddingDenseBackward</code>

Status: Open.

#9392 In pytorch/xla;

· unterumarmung opened on Jun 18, 2025

Changes needed for <code>device_assignment</code> in <code>PjRtComputationClient::Compile</code> to support submeshing/localized spmd

Status: Open.

#9391 In pytorch/xla;

· kvshbg-aws opened on Jun 18, 2025

Create and Expose the <code>torch_xla::OpSharding</code> wrapper class instead of <code>xla::OpSharding</code> class

distributedSPMD and other distributed things.SPMD and other distributed things.

enhancementNew feature or requestNew feature or request

Status: Open.

#9390 In pytorch/xla;

· kvshbg-aws opened on Jun 18, 2025

Normalize <code>tile_assignment</code> after constructing the <code>xla::OpSharding</code> object

distributedSPMD and other distributed things.SPMD and other distributed things.

enhancementNew feature or requestNew feature or request

Status: Open.

#9389 In pytorch/xla;

· kvshbg-aws opened on Jun 18, 2025

Update pins for week of June 16, 2025

Status: Open.

#9387 In pytorch/xla;

· yaoshiang opened on Jun 18, 2025

[v2.8] Neuron inference trace analyzer/bucketing unit tests hanging at GetParameterIdTensorMapping/TransferFromDevice

Status: Open.

#9378 In pytorch/xla;

· jeffhataws opened on Jun 17, 2025

Error handling on <code>supported</code> XLA operations.

tech debtTechnical Debt Is EvilTechnical Debt Is Evil

Status: Open.

#9377 In pytorch/xla;

· ysiraichi opened on Jun 17, 2025

Failing <code>torch_xla._XLAC._xla_custom_call()</code> with <code>RuntimeError: Bad StatusOr access: UNIMPLEMENTED: No registered implementation for custom call to my_lib.my_op.default for platform CUDA</code>

bugSomething isn't workingSomething isn't working

stablehloStableHLO related workStableHLO related work

Status: Open.

#9371 In pytorch/xla;

· hsjts0u opened on Jun 16, 2025

consider per-compilation-pass verification to reduce XLA crashes

usabilityBugs/features related to improving the usability of PyTorch/XLABugs/features related to improving the usability of PyTorch/XLA

Status: Open.

#9370 In pytorch/xla;

· zhanyong-wan opened on Jun 16, 2025

PyTorch/XLA custom Triton kernel export to StableHLO

enhancementNew feature or requestNew feature or request

Triton

xla:gpu

Status: Open.

#9366 In pytorch/xla;

· annabellej opened on Jun 16, 2025

Torchax should be a separate wheel or a submodule

buildBuild process related matters (e.g. build system).Build process related matters (e.g. build system).

Status: Open.

Feature

#9363 In pytorch/xla;

· pganssle-google opened on Jun 16, 2025

Op(s) not lowered: aten::glu_backward

enhancementNew feature or requestNew feature or request

loweringATen Operation loweringATen Operation lowering

Status: Open.

#9359 In pytorch/xla;

· joris-vaneyghen opened on Jun 15, 2025